case study anemia of chronic disease

Enter search terms to find related medical topics, multimedia and more.

Advanced Search:

• Use “ “ for exact phrases.
• For example: “pediatric abdominal pain”
• Use – to remove results with certain keywords.
• For example: abdominal pain -pediatric
• Use OR to account for alternate keywords.
• For example: teenager OR adolescent
• IN THIS TOPIC

Advertisement

Introduction

Pathophysiology of ai, acknowledgments, anemia of inflammation.

• Split-Screen
• Request Permissions
• Cite Icon Cite
• Search Site
• Open the PDF for in another window

Guenter Weiss , Tomas Ganz , Lawrence T. Goodnough; Anemia of inflammation. Blood 2019; 133 (1): 40–50. doi: https://doi.org/10.1182/blood-2018-06-856500

Download citation file:

• Ris (Zotero)
• Reference Manager

Anemia of inflammation (AI), also known as anemia of chronic disease (ACD), is regarded as the most frequent anemia in hospitalized and chronically ill patients. It is prevalent in patients with diseases that cause prolonged immune activation, including infection, autoimmune diseases, and cancer. More recently, the list has grown to include chronic kidney disease, congestive heart failure, chronic pulmonary diseases, and obesity. Inflammation-inducible cytokines and the master regulator of iron homeostasis, hepcidin, block intestinal iron absorption and cause iron retention in reticuloendothelial cells, resulting in iron-restricted erythropoiesis. In addition, shortened erythrocyte half-life, suppressed erythropoietin response to anemia, and inhibition of erythroid cell differentiation by inflammatory mediators further contribute to AI in a disease-specific pattern. Although the diagnosis of AI is a diagnosis of exclusion and is supported by characteristic alterations in iron homeostasis, hypoferremia, and hyperferritinemia, the diagnosis of AI patients with coexisting iron deficiency is more difficult. In addition to treatment of the disease underlying AI, the combination of iron therapy and erythropoiesis-stimulating agents can improve anemia in many patients. In the future, emerging therapeutics that antagonize hepcidin function and redistribute endogenous iron for erythropoiesis may offer additional options. However, based on experience with anemia treatment in chronic kidney disease, critical illness, and cancer, finding the appropriate indications for the specific treatment of AI will require improved understanding and a balanced consideration of the contribution of anemia to each patient’s morbidity and the impact of anemia treatment on the patient’s prognosis in a variety of disease settings.

Anemia of inflammation (AI), better known as anemia of chronic disease (ACD), is considered the second most prevalent anemia worldwide (after iron deficiency anemia [IDA]) and the most frequent anemic entity observed in hospitalized or chronically ill patients. 1 , 2   Estimates suggest that up to 40% of all anemias worldwide can be considered AI or combined anemias with important AI contributions, which, in total, account for >1 billion affected individuals. 2 , 3   These high numbers reflect that the spectrum of diseases in which inflammation has been recognized as contributing to anemia has expanded over the past years. Originally, AI was linked to chronic infections and autoimmune diseases in which inflammation was easily detectable and sustained 1 , 4 , 5   ( Table 1 ). Some cancers that presented with a strong inflammatory component were recognized to have a similar pathophysiology, although this was often complicated by other cancer-specific and iatrogenic mechanisms. 4   Accumulating data suggest that AI, sometimes with coexisting iron deficiency, is much more prevalent and also affects patients with chronic kidney disease, especially those undergoing dialysis, and patients with congestive heart failure in whom iron deficiency impairs cardiovascular performance. 6 , 7   Other less well-studied examples include chronic obstructive pulmonary disease, pulmonary arterial hypertension, obesity, chronic liver disease, and advanced atherosclerosis with its sequelae of coronary artery disease and stroke 8-10   ( Table 1 ). Despite the high prevalence of AI and its documented associations with the progression of underlying diseases, it remains unclear to what extent AI is merely a marker of disease severity and progression as opposed to a causative factor with a specific impact on the underlying diseases and long-term patient outcomes. Increased understanding of the specific contribution of AI to long-term patient outcomes may only come with human trials of narrowly targeted therapeutic interventions against AI.

Disease groups in which AI is common

AI is caused by 3 major pathophysiological pathways that act through mediators of an activated immune system ( Figure 1 ).

Pathophysiological mechanisms of AI. Systemic inflammation results in immune cell activation and formation of numerous cytokines. Interleukin (IL-6) and IL-1β, as well as lipopolysaccharide (LPS), are potent inducers of the master regulator of iron homeostasis, hepcidin, in the liver, whereas expression of the iron-transport protein transferrin is reduced. Hepcidin causes iron retention in macrophages by degrading the only known cellular iron exporter ferroportin (FP1); by the same mechanism, it blocks dietary iron absorption in the duodenum. Multiple cytokines (eg, interleukin-1β [IL-1β], IL-6, IL-10, and interferon-γ [IFN-γ]) promote iron uptake into macrophages, increase radical-mediated damage to erythrocytes and their ingestion by macrophages, and cause efficient iron storage by stimulating ferritin production and blocking iron export by transcriptional inhibition of FP1 expression. This results in the typical changes of AI (ie, hypoferremia and hyperferritinemia). In addition, IL-1 and TNF inhibit the formation of the red cell hormone erythropoietin (Epo) by kidney epithelial cells. Epo stimulates erythroid progenitor cell proliferation and differentiation, but the expression of its erythroid receptor (EpoR) and EpoR-mediated signaling are inhibited by several cytokines. Moreover, cytokines can directly damage erythroid progenitors or inhibit heme biosynthesis via radical formation or induction of apoptotic processes. Importantly, because of iron restriction in macrophages, the availability of this metal for erythroid progenitors is reduced. Erythroid progenitors acquire iron mainly via transferrin-iron/transferrin receptor (Tf/TfR)-mediated endocytosis. Erythroid iron deficiency limits heme and hemoglobin (Hb) biosynthesis, as well as reduces EpoR expression and signaling via blunted expression of Scribble (Scb). In addition, the reduced Epo/EpoR signaling activity impairs the induction of erythroferrone (Erfe), which normally inhibits hepcidin production.

Although not specifically documented, it can be expected that the contribution of each of these pathways depends on the underlying causes and patterns of inflammation, as well as the genetic makeup and premorbid condition of the patient, including preexisting iron stores, erythropoietic capacity of the marrow and the responsiveness of renal erythropoietin (Epo) production to anemia and hypoxia, and the resistance of erythrocytes to mechanical and antibody-complement induced injury.

Iron restriction

First, systemic immune activation leads to profound changes of iron trafficking, resulting in iron retention in macrophages and in reduced dietary iron absorption. Iron sequestration in macrophages is by far more important, because recycling of iron from senescent erythrocytes by macrophages accounts for >90% of the daily iron requirements for hemoglobin (Hb) synthesis and erythropoiesis. 11   In response to microbial molecules, autoantigens, or tumor antigens, multiple inflammatory cytokines are released by cells of the immune system and alter systemic iron metabolism. Although it is not possible to completely disentangle the iron-regulatory contributions of the multiply cross-regulating cytokine networks, interleukin-6 (IL-6) appears to be the most important, at least in animal models. 12   In laboratory animals and in humans, IL-6 stimulates hepatocytes to produce hepcidin, the master regulator of iron homeostasis, predominantly through STAT3. 11   Other cytokines, including IL-1 and activin B, can also stimulate hepcidin production, but their specific pathological role is less well established. 13   As reviewed elsewhere, the synthesis of systemically acting hepcidin by hepatocytes is also positively regulated by transferrin saturation and iron stores and negatively regulated by erythroid activity in the marrow. 11 , 14  

Hepcidin exerts its iron-regulatory effects by binding to the only known transmembrane iron exporter, ferroportin, causing cellular ferroportin internalization and degradation. 15   Thus, increased hepcidin concentrations inhibit iron absorption in the duodenum where ferroportin is needed to deliver absorbed dietary iron to the circulation, and they also act on macrophages to block the release of iron recycled from senescent erythrocytes into the plasma. 16   Recent evidence indicates that, at higher concentrations, hepcidin may directly block iron export by occluding ferroportin, 17   a mechanism that may be especially important in limiting iron release from cells that lack endocytic machinery (erythrocytes) or in conditions under which endocytosis is slow.

In animal models of AI and in patients suffering from inflammatory diseases, increased hepcidin levels are associated with low ferroportin expression on duodenal enterocytes and macrophages, along with impaired dietary iron absorption and retention of iron in macrophages, 18 , 19   thereby causing decreased iron delivery for erythropoiesis.

In addition, various cytokines directly impact on duodenal or macrophage iron homeostasis. Tumor necrosis factor (TNF) reduces duodenal iron absorption by a poorly characterized, but hepcidin-independent, mechanism. 20   The cytokines IL-1, IL-6, IL-10, or TNF-α promote iron acquisition into macrophages via transferrin receptor–mediated endocytosis, via divalent metal transporter 1, or possibly also via increased iron acquisition by lactoferrin and lipocalin-2. 21   However, the major source for iron for macrophages is senescent erythrocytes. Cytokines, inflammation-derived radicals, and complement factors damage erythrocytes and promote erythrophagocytosis via stimulation of receptors recognizing senescent red blood cells, such as T-cell immunoglobulin domain 4 or possibly CD44. 22 , 23   Recent evidence suggests that, during periods of increased erythrocyte destruction, erythrophagocytosis and iron recycling are primarily carried out by hepatic macrophages differentiating in the liver from circulating monocytes, rather than by resident splenic macrophages. 24   Once iron is acquired by macrophages, it is mainly stored within ferritin, the major iron storage protein whose expression is massively induced by macrophage iron, heme, and cytokines. 5 , 25   Although circulating and, to a lesser extent, macrophage-derived hepcidin are the main regulators of iron export by these cells, 15 , 16 , 18 , 19 , 26 , 27   bacterial lipopolysaccharides and interferon-γ (IFN-γ) block the transcription of ferroportin, thereby reducing cellular iron export. 28 , 29   All of these events lead to iron-restricted erythropoiesis and the characteristic changes in systemic iron homeostasis observed in AI: hypoferremia and hyperferritinemia. These effects are partially counteracted by the stimulation of synthesis of ferroportin in macrophages by retained iron and heme, 30   perhaps explaining why AI rarely reaches the severity seen in pure iron-deficiency anemia.

The specific mechanisms by which iron restriction decreases erythropoiesis have not been fully clarified, but they involve active iron-regulated erythroid-specific mechanisms that decrease the synthesis of heme and Hb, as well as inhibit erythropoiesis, thereby protecting nonerythroid tissues from iron deficiency. 31-33   Heme concentration in erythrocyte precursors functions as a secondary iron-dependent regulator of Hb synthesis and erythropoiesis.

Inflammatory suppression of erythropoietic activity

The second pathogenic factor in AI is iron- and hepcidin-independent impairment of erythropoiesis. It is caused, in part, by reduced production and/or reduced biological activity of the hormone Epo in the inflammatory setting. 34   Observational studies have indicated lower Epo levels than expected for the degree of anemia in most AI subjects. 1   These observations may be due, in part, to the inhibitory effects of cytokines, such as IL-1 and TNF, on hypoxia-mediated stimulation of Epo by interfering with mediated GATA-2 or HNF4 transcription or by causing radical-mediated damage of Epo-producing kidney epithelial cells. 35 , 36   Epo exerts its biological effects after binding to its homodimeric erythroid receptor via JAK/STAT-mediated signaling cascades. 36   Although, the number of Epo receptors (EpoRs) did not appear to be altered in subjects with inflammatory anemia, the efficacy of Epo-mediated signaling is reduced and inversely linked to the circulating levels of IL-1 and IL-6, 37   indicating the inflammation-driven hyporesponsiveness of EpoRs. However, recent evidence suggests that erythroid iron deficiency, because it also occurs in AI, results in downregulation of EpoR, which could be traced back to iron-mediated regulation of the EpoR control element Scribble. 31   Specifically, iron deficiency impairs transferrin receptor-2–mediated iron sensing in erythroid cells, 38   resulting in downregulation of Scribble and reduced EpoR expression.

Acting to promote increased iron supply during intensified erythropoiesis, Epo and hypoxia inhibit hepcidin formation via induction of hypoxia inducible factor 1, erythroferrone, matriptase-2, growth differentiation factor-15, or platelet derived growth factor-BB. 14 , 39-42   Thus, the reduced availability and activity of Epo in AI negatively impact the induction of at least some of these hepcidin blockers, such as erythroferrone, thereby aggravating hepcidin-mediated erythroid iron limitation 43   and impairing, via a vicious cycle, Epo signaling via Scribble ( Figure 1 ).

Erythroid cell proliferation and differentiation are impaired by the blunted Epo effect and by iron limitation via hepcidin and cytokines. In addition, various inflammatory mediators directly target erythroid cells and induce apoptosis via ceramide- or radical-mediated pathways. 1   IFN-γ appears to be central for this process, 44   but this cytokine also downregulates EpoR expression on erythroid progenitors, 45   inhibits their differentiation via stimulation of PU.1 expression, and reduces erythrocyte lifespan, even as it promotes leukocyte production, differentiation, and activation that are important for host defense. 46  

In addition, the severity and appearance of AI can be further modified by different factors ( Table 2 ). Specifically, concomitant bleeding episodes, as well as dietary-, genetic-, hormonal-, age-, treatment-, or disease-specific factors and mechanisms, can impact iron metabolism and erythropoiesis in the setting of AI.

Modifiers of severity of AI

ACE, angiotensin convertin enzyme; CMV, cytomegalovirus; EBV, Epstein-Barr virus; HCV, hepatitis C virus.

Finally, in an acute clinical setting, anemia is detected after hours or a few days in patients with severe infection, which cannot be solely explained by inflammation-driven iron retention or inhibition of erythropoiesis, which may require more time to result in a clinically evident reduction in Hb.

Decreased erythrocyte survival

A shortened erythrocyte lifespan has been extensively documented in the inflammatory setting and has been attributed to enhanced erythrophagocytosis by hepatic and splenic macrophages caused by “bystander” deposition of antibody and complement on erythrocytes, mechanical damage from fibrin deposition in microvasculature, and activation of macrophages for increased erythrophagocytosis. 23 , 46 , 47   Shortened erythrocyte survival is usually a minor factor in chronic AI; however, in acute infections, severe sepsis, or other critical illnesses accompanied by a high level of cytokine activation, anemia is detected after hours or a few days (ie, too rapidly to be accounted for by deficient erythropoiesis). It is reasonable that massive erythrophagocytosis, hemolysis, or pooling of erythrocytes, along with hemodilution, contribute to this entity that awaits systematic scientific analysis. 48   Moreover, remediable iatrogenic factors are common in critical illness and include blood loss from phlebotomy and gastrointestinal blood loss caused by nasogastric tubes, anticoagulation, and the use of medications that promote gastroduodenal erosion or ulceration.

Although not studied in a comparative fashion, symptoms and signs of IDA and AI are similar and include fatigue, weakness, reduced cardiovascular performance and exercise tolerance, and impaired learning and memory capacity. 49 , 50   Of note, it is uncertain to what extent symptoms of anemia are caused by hypoxia and reduced tissue oxygen tension, as opposed to iron deficiency, which impairs mitochondrial function, cellular metabolism, enzyme activities, and neurotransmitter synthesis. 11 , 51 , 52   The important effects of cellular iron deficiency are highlighted by clinical observations of symptomatic improvement in women who are treated for nonanemic iron deficiency, 53   as well as interventional studies in patients with congestive heart failure in whom intravenous iron supplementation provided comparable benefits in subjects with IDA, even without correction of anemia. 7   However, in contrast to IDA, anemia-related symptoms in AI are often attributed to the underlying disease, which may be 1 reason why AI is often not specifically treated.

Anemia, defined by a Hb concentration <120 g/L for women and <130 g/L for men, can be diagnosed as AI or ACD based on the underlying alterations in iron homeostasis along with clinical or biochemical evidence of inflammation; however, it is often necessary to rule out coexisting causes of anemia that may require specific interventions ( Table 2 ). In addition to assessing coexisting iron deficiency, as discussed further below, laboratory evaluations may include renal and liver function tests, thyroid function tests or markers of hemolysis, and determination of folic acid, cobalamin, or vitamin D concentrations. Vitamin D is a negative regulator of hepcidin expression. 54   Elderly subjects with different causes of anemia were often found to suffer from vitamin D deficiency, and anemia can be corrected, in part, by vitamin D supplementation, which decreases hepcidin levels. 54 , 55  

Characteristically, AI presents as a mild to moderate normochromic and normocytic anemia, which clearly separates it from hypochromic and microcytic IDA. 50   AI and IDA have in common reduced circulating iron concentrations and a reduced percentage of iron bound to transferrin, known as transferrin saturation, as well as reduced reticulocyte counts. The most useful differentiating parameter is serum ferritin. Whereas a ferritin level <30 µg/mL is associated with absolute or true iron deficiency, patients with AI present with normal or increased ferritin levels (>100 µg/L), depending on the underlying condition. 1   Serum ferritin is predominantly secreted by macrophages and hepatocytes. 56 , 57   High concentrations of serum ferritin in AI result from increased ferritin secretion by iron-retaining macrophages but also reflect that ferritin is an acute-phase protein that is induced by various inflammatory mediators. 5 , 56   Thus, in the setting of inflammatory diseases, ferritin largely loses its diagnostic value as an indicator of body iron stores. The circulating concentrations of the iron transport protein transferrin are at the upper limit of normal in IDA subjects but are reduced in subjects with AI, because transferrin expression is negatively affected by cytokines. 58   Thus, with these parameters at hand, AI can be diagnosed, 18   as illustrated in Figure 2 .

Evaluation and management of anemia. Evaluation and management of anemia. Once the screening blood count demonstrates anemia, an evaluation is necessary and begins with an assessment of iron status. When ferritin (SF) and/or iron saturation levels (TSAT) indicate absolute iron deficiency, referral to a gastroenterologist or gynecologist to identify a specific source of chronic blood loss may be indicated. When ferritin and/or iron saturation values rule out absolute iron deficiency, and signs of inflammation are evident, AI is likely. Depending on ferritin, transferrin saturation, or values of markers suggesting concomitant true iron deficiency, diagnostic steps to identify the disease underlying AI and/or the reason for iron deficiency should be undertaken. A nephrologist may be consulted in the case of GFR reduction and evidence for chronic kidney disease. When ferritin and/or iron saturation values are indeterminant, further evaluation to rule out absolute iron deficiency vs inflammation/chronic disease is necessary. A successful therapeutic trial of iron would confirm absolute iron deficiency. No response to iron therapy would support the diagnosis of AI, suggesting that ESA therapy may be beneficial. Reprinted from Goodnough and Schrier 8   with permission.

The diagnostic challenge in AI is the identification of patients with concomitant true iron deficiency (AI/IDA patients), because they need specific evaluation for the source of blood loss and iron-targeted management strategies. 1 , 59 , 60   A varying percentage (20%-85%) of patients with AI also suffer from true iron deficiency, which is mainly based on concomitant disease-related or unrelated gastrointestinal or urogenital bleeding episodes, iatrogenic blood draws, or losses in association with therapeutic procedures, such as hemodialysis. 1 , 9 , 10 , 50 , 55 , 59 , 61-65   Children with AI and chronic inflammatory diseases are at particular risk for coexisting true iron deficiency, because their growth and the expansion of the red cell mass require additional iron.

Although one would expect that subjects with AI/IDA may become microcytic and hypochromic, this pattern is much less pronounced in AI/IDA than in IDA, and a significant overlap between AI and AI/IDA makes the differential diagnosis challenging and inaccurate. 18 , 66   Several alternative cellular markers, including the percentage of hypochromic red blood cells, reticulocyte Hb content, red blood cell Hb content, and red cell distribution width, alone or in combination with iron-metabolism parameters, have been studied for their potential to detect iron deficiency in the presence of inflammation. 58 , 59 , 65 , 66   Although some of these tests showed promise, the analyses greatly depend on the availability of appropriate laboratory instruments and on standardized preanalytical procedures. Moreover, most of these tests have never been prospectively studied to evaluate their true diagnostic potential for differentiating AI vs AI/IDA, for the prediction of response to the chosen therapy, or as indicators of potential harm (eg, iron oversupplementation). 65 , 67   Even soluble transferrin receptor (sTfR), a valuable measure of iron requirements for erythropoiesis, is confounded by inflammation, because several cytokines affect sTfR levels independently of iron status. 68   The ferritin index, which is calculated by sTfR/log ferritin values, provided a better discrimination between AI (<1) and AI/IDA (>2) subjects, admittedly with some overlap and a clinically relevant zone of uncertainty. 58 , 64 , 69   Based on our expanding knowledge of the pathophysiology of AI, hepcidin measurement may eventually play a role in the evaluation and management of AI ( Table 3 ). 19 , 61 , 70 , 71   The potential utility of hepcidin measurements in this context was based on observations showing that, although hepcidin levels are increased in AI, they are significantly reduced in the presence of concomitant iron deficiency. The mechanism could be traced back to induction of inhibitory SMAD proteins by iron deficiency, which reduced inflammation-driven hepcidin expression, suggesting that inhibitory erythropoietic signals dominate over inflammation-driven hepcidin induction. 72 , 73   Although serum hepcidin may help to differentiate AI and AI/IDA subjects, 19   a hepcidin assay alone is not definitive ( Table 3 ). Nonetheless, hepcidin determination is a promising diagnostic tool when used in combination with other established tests 16 , 19 , 71 , 74   or emerging novel markers, such as erythroferrone. Of note, hepcidin levels may predict the response to therapy with oral iron, and if treatment with oral iron has not been efficient after 2 weeks, a switch to IV iron is indicated 75   ( Table 3 ). The differentiation between AI and AI/IDA is also an important consideration when managing dialysis patients; however, hepcidin levels are increased in these subjects as a consequence of impaired renal hepcidin excretion and are of limited diagnostic value. 67   Thus, there is still a need for readily available and interpretable biomarkers that clearly differentiate between ACD and ACD/AI patients at beside and help to identify the best therapy and predict the therapeutic response. 75  

Potential role of hepcidin in the diagnosis and management of anemia

CKD, chronic kidney disease; PO, by mouth; Tsat, transferrin saturation.

Any consideration of treatment in AI must take into account its evolutionary context. AI results from a conserved defense strategy of the body directed against invading microbes. Iron is an essential nutrient for humans and other animals, as well as for most microbes, which require iron for their proliferation and pathogenicity. By the combined action of cytokines, hepcidin, and iron-binding peptides, as outlined in Pathophysiology of AI, iron is sequestered in forms that make it less accessible to circulating pathogens, a strategy termed “nutritional immunity.” 76 , 77   In this context, tissue- and cell-specific changes in iron trafficking depend on the type and location of the pathogen (ie, extracellularly or within a specific cellular compartment), 78-80   with details under intensive study. In addition, iron availability exerts subtle effects on immune function by modulating immune cell differentiation and proliferation, and it also affects antimicrobial effector mechanisms of immune cells. 76   Thus, when treating anemias, one has to consider whether such a treatment could also impact the underlying disease; this is of utmost concern when treating patients with infections or cancer.

Optimally, the best treatment for AI is cure of the underlying inflammatory disease, which mostly results in resolution of AI. As much as is feasible, the contribution of concomitant pathologies to AI ( Table 2 ) should also be considered and specifically corrected; however, such fundamental treatments are not always possible or effective. With regard to treatments directed specifically at AI, we lack data from prospective trials about how aggressive such treatments should be or what constitutes the optimal therapeutic end point. Caution is suggested by studies indicating that, in environments with a high endemic burden of infectious diseases, mild anemia and/or iron deficiency may even be beneficial. Infants with mild iron deficiency or anemia were less likely to suffer and die from severe malaria. 81   Indiscriminate dietary iron fortification resulted in increased morbidity and mortality from serious infections, including malaria and enteric infections. 82 , 83   Such studies indicate that care should be taken to identify those patients with AI, with or without iron deficiency, who can benefit from iron supplementation or anemia correction. 84   Although not yet widely used, low pretreatment serum hepcidin concentration appears to be a good predictor of therapeutic response to oral iron supplementation, thus potentially avoiding risks to patients who could not benefit from oral iron supplementation. 85  

Two therapies have been established for the treatment of AI: iron supplementation and treatment with erythropoiesis-stimulating agents (ESAs). Red blood cell transfusion is considered only as an emergency treatment in patients with severe anemia who are clinically unstable and in whom rapid correction of Hb levels is warranted. 48 , 86   Of interest, recent evidence suggested that restrictive use of blood transfusion, specifically in critically ill patients with acute bleeding, is associated with a lower mortality than liberal use achieving higher target Hb levels. 87   Given the increasing evidence that blood transfusions have poor effectiveness and are possibly harmful, the guiding principle for transfusion therapy should be “less is more.” 88   Anemic patients with inflammatory bowel disease 89   and rheumatoid arthritis 77 , 90   have very mild anemias that are corrected by anti-TNF therapy. 89 , 90   Patients with more severe anemia often have other pathophysiologic components that are treatable with iron or ESA therapy, 91   as illustrated in Figure 2 .

Recombinant human ESAs have been used successfully for the treatment of AI for many years, specifically in patients with cancer or renal failure or when iron supplementation alone was ineffective. 1 , 68   However, concerns about the unrestricted use of ESAs in AI arose from studies showing higher mortality in patients with cancer or in dialysis patients not immediately responding to ESA treatment, as well as in nondialysis patients treated with novel erythropoiesis-stimulating drugs. 92-94   The specific mechanisms of increased mortality remain elusive but may include effects of ESAs on coagulation or angiogenesis, direct proliferative effects of ESAs on cancer cells expressing EpoRs, or the immune-modulatory effects of ESAs that may dampen antimicrobial effector function. 92 , 95 , 96   Nevertheless, ESA therapy remains approved for a variety of indications, including patients who are anemic (and in whom iron deficiency has been ruled out) and scheduled for major noncardiac surgery ( Table 4 ). 91  

ESAs: current approval status

Year of approval is shown in parentheses.

CIA, chemotherapy-induced anemia; ESKD, end-stage kidney disease; PAD, preoperative autologous donation.

Iron-replenishing strategies have attracted renewed interest in AI therapy, either alone or in combination with low to moderate doses of ESAs. Although oral or IV iron therapy appears to be justified in patients with combined AI/IDA, new treatment strategies that mobilize sequestered iron from the reticuloendothelial system by targeting hepcidin make more sense for subjects with pure AI. Accurate diagnostic tools will be needed to correctly discriminate AI and AI/IDA subjects for future trials of these approaches.

Oral iron preparations are available as iron salts or iron carbohydrates and are the therapy of choice in AI subjects with true iron deficiency and mild inflammation, in whom they show an efficacy that is comparable to IV preparations. 97 , 98   However, in specific disease conditions, IV iron may be preferable, even with low inflammation (eg, chronic heart failure) or under circumstances when data on the therapeutic efficacy of oral iron therapy are not available, as reviewed recently. 99 , 100   In general, oral iron preparations should be taken once in the morning at a minimum dose of 50 mg of ferrous iron (the total compound dose depends on the specific iron salt or glycan used). More frequent dosing reduces iron bioavailability by increased production of hepcidin, inhibiting iron absorption. 101   Ascorbic acid and overnight fasting can increase iron bioavailability, whereas proton pump inhibitors or several foods, including milk products and tea, decrease iron absorption. 11 , 50 , 63   IV iron therapy is indicated when oral iron treatment is insufficient to correct iron deficiency, such as during ongoing blood loss in dysfunctional uterine bleeding or multiple vascular malformations in the gastrointestinal tract, in the presence of gastrointestinal side effects, or when a rapid replenishment of iron stores is desired, such as prior to a planned surgery. Insufficient absorption of iron in the duodenum is a major problem in AI patients, because of inhibition of iron transfer from enterocytes to the circulation by hepcidin and cytokines. 16 , 19   In addition to older IV iron-carbohydrate formulations, such as ferric gluconate or ferric sucrose, newer glycan-coated nanoparticle drugs, such as iron carboxymaltose, iron isomaltoside, and ferumoxytol, have been introduced into clinical practice, which allow administration of up to 1000 mg of iron per injection. The rare life-threatening anaphylactoid reactions that occurred after iron injections resulted in warning letters from the US Food and Drug Administration and the European Medicines Agency, 102   and strategies have been implemented to identify susceptible patients and to reduce the risk of serious side effects. 103   IV iron has been shown to successfully correct iron deficiency in AI patients, 62   and it is especially effective in patients with inflammatory bowel disease who often suffer from concomitant true iron deficiency as a consequence of chronic intestinal bleeding. 61 , 63   Severe systemic inflammation may even make IV iron therapy less effective, because IV iron-glycan complexes are primarily taken up by macrophages, and iron is subsequently exported from macrophages to transferrin in the circulation via ferroportin. The expression and activity of ferroportin are inhibited by hepcidin, which is present at higher concentration in such patients. 16 , 104   In this instance, ESA therapy may be helpful, because it stimulates erythropoiesis while suppressing hepcidin.

Based on our knowledge of the pathophysiology of AI, with the underlying iron retention in the reticuloendothelial system and the central role of hepcidin in this process ( Figure 1 ), novel therapeutic approaches have emerged that aim to antagonize hepcidin function and to mobilize iron from macrophages to deliver it for erythropoiesis ( Table 3 ). Such strategies have been evaluated in animal models and are beginning to reach human clinical trials. General mechanisms involve inhibition of hepcidin production, neutralization of circulating hepcidin, protection of ferroportin function from hepcidin inhibition, and inhibition of hepcidin-inducing signals, such as IL-6; all were summarized in recent reviews. 9 , 105   Of note, hepcidin-antagonizing strategies and increased egress of iron from the reticuloendothelial system to the circulation may affect the disease underlying ACD, potentially detrimentally (eg, in infections) or beneficially (eg, chronic kidney disease). 79 , 106   Nonetheless, hepcidin-modifying strategies may hold promise in combination with ESA, 107   because the increased endogenous delivery of iron may reduce ESA requirements and limit their adverse effects.

A novel therapeutic principle emerged from the introduction of prolyl hydroxylase inhibitors, which stabilize hypoxia-inducible factors and subsequently ameliorate anemia by promoting endogenous Epo formation and iron delivery from enterocytes and macrophages. 108   These orally available drugs are being studied in phase 3 clinical trials for the treatment of anemia in hemodialysis, 109   but they could also become useful therapeutic options in AI.

Our knowledge on the pathophysiology of AI has expanded dramatically over the past years, and we are gathering information on the therapeutic efficacy of established and novel emerging treatment strategies. However, we are still lacking information on optimal therapeutic start and end points for AI and are in need of identifying biomarkers that help to differentiate patients with AI from patients with AI and true iron deficiency, because different treatment strategies are used for these 2 groups. Moreover, we have to gather information on the effects of any treatment on the course of the specific disease underlying AI. This is essential to choose the best therapeutic options for our patients to achieve an optimal quality of life or cardiovascular performance, together with a neutral, or even beneficial, effect on the primary disease causing AI.

G.W. is grateful for support from the Christian Doppler Society.

Contribution: G.W., T.G., and L.T.G. were responsible for conception and design, as well as the writing and final approval of the manuscript.

Conflict-of-interest disclosure: G.W. has received lecture honoraria from Vifor. L.T.G. is a consultant for Vifor Pharma and InCube Labs. T.G. is a consultant for Keryx Pharma, Vifor Pharma, Akebia Therapeutics, Gilead Sciences, La Jolla Pharma, and Ionis Pharmaceuticals; has received research funding from Keryx Pharma and Akebia Therapeutics; and is a scientific founder and consultant for Silarus Pharma and Intrinsic LifeSciences.

Correspondence: Guenter Weiss, Department of Internal Medicine II, Infectious Diseases, Immunology, Pneumology and Rheumatology, Medical University Innsbruck, Anichstr 35, A-6020 Innsbruck, Austria; e-mail: [email protected] .

This feature is available to Subscribers Only

• Previous Article
• Next Article

Email alerts

Affiliations.

• Current Issue
• First edition
• Collections
• Submit to Blood
• About Blood
• Subscriptions
• Public Access
• Permissions
• Blood Classifieds
• Advertising in Blood
• Terms and Conditions

American Society of Hematology

• 2021 L Street NW, Suite 900
• Washington, DC 20036
• TEL +1 202-776-0544
• FAX +1 202-776-0545

ASH Publications

• Blood Advances
• Hematology, ASH Education Program
• ASH Clinical News
• The Hematologist
• Publications
• Privacy Policy
• Cookie Policy
• Terms of Use

This Feature Is Available To Subscribers Only

Sign In or Create an Account

Learn how UpToDate can help you.

Select the option that best describes you

• Medical Professional
• Resident, Fellow, or Student
• Hospital or Institution
• Group Practice
• Patient or Caregiver
• Find in topic

RELATED TOPICS

INTRODUCTION

The pathogenesis, laboratory findings, and treatment of ACD/AI will be reviewed here. Overviews of the evaluation of anemia are presented separately.

● Children – (See "Approach to the child with anemia" .)

● Adults – (See "Diagnostic approach to anemia in adults" .)

PATHOGENESIS

Reduced iron availability  —  ACD/AI is thought to result from an evolutionary defense strategy of the body to limit the availability of iron for invading microbes [ 1-3 ]. Accordingly, the pathophysiology of ACD/AI involves:

Anaemia of chronic diseases: Pathophysiology, diagnosis and treatment

Affiliations.

• 1 Servicio Cántabro de Salud. Centro de Salud El Alisal, Santander, Cantabria, España.
• 2 Servicio Cántabro de Salud, Santander, Cantabria, España.
• 3 Servicio Cántabro de Salud. Centro de Salud Altamira, Cantabria, España. Electronic address: [email protected].
• PMID: 33358297
• DOI: 10.1016/j.medcli.2020.07.035

Anaemia of chronic disease (ACD) is generated by the activation of the immune system by autoantigens, microbial molecules or tumour antigens resulting in the release of cytokines that cause an elevation of serum hepcidin, hypoferraemia, suppression of erythropoiesis, decrease in erythropoietin (EPO) and shortening of the half-life of red blood cells. Anaemia is usually normocytic and normochromic, which is the most prevalent after iron deficiency anaemia, and it is the most frequent in the elderly and in hospitalized patients. If the anaemia is severe, the patient's quality of life deteriorates, and it can have a negative impact on survival. Treatment is aimed at controlling the underlying disease and correcting anaemia. Sometimes intravenous iron and EPO have been used, but the therapeutic future is directed against hepcidin, which is the final target of anaemia.

Keywords: Antagonistas hepcidina; Ferroportin; Ferroportina; Hepcidin; Hepcidin antagonists; Hepcidina; Interleucina-6; Interleukin-6; Iron metabolism; Metabolismo hierro.

Copyright © 2020 Elsevier España, S.L.U. All rights reserved.

Publication types

• Anemia* / diagnosis
• Anemia* / etiology
• Anemia* / therapy
• Chronic Disease
• Erythropoietin*
• Quality of Life
• Erythropoietin

• Case report
• Open access
• Published: 13 September 2021

Critical iron deficiency anemia with record low hemoglobin: a case report

• Audrey L. Chai   ORCID: orcid.org/0000-0002-5009-0468 1 ,
• Owen Y. Huang 1 ,
• Rastko Rakočević 2 &
• Peter Chung 2  

Journal of Medical Case Reports volume  15 , Article number:  472 ( 2021 ) Cite this article

27k Accesses

6 Citations

2 Altmetric

Metrics details

Anemia is a serious global health problem that affects individuals of all ages but particularly women of reproductive age. Iron deficiency anemia is one of the most common causes of anemia seen in women, with menstruation being one of the leading causes. Excessive, prolonged, and irregular uterine bleeding, also known as menometrorrhagia, can lead to severe anemia. In this case report, we present a case of a premenopausal woman with menometrorrhagia leading to severe iron deficiency anemia with record low hemoglobin.

Case presentation

A 42-year-old Hispanic woman with no known past medical history presented with a chief complaint of increasing fatigue and dizziness for 2 weeks. Initial vitals revealed temperature of 36.1 °C, blood pressure 107/47 mmHg, heart rate 87 beats/minute, respiratory rate 17 breaths/minute, and oxygen saturation 100% on room air. She was fully alert and oriented without any neurological deficits. Physical examination was otherwise notable for findings typical of anemia, including: marked pallor with pale mucous membranes and conjunctiva, a systolic flow murmur, and koilonychia of her fingernails. Her initial laboratory results showed a critically low hemoglobin of 1.4 g/dL and severe iron deficiency. After further diagnostic workup, her profound anemia was likely attributed to a long history of menometrorrhagia, and her remarkably stable presentation was due to impressive, years-long compensation. Over the course of her hospital stay, she received blood transfusions and intravenous iron repletion. Her symptoms of fatigue and dizziness resolved by the end of her hospital course, and she returned to her baseline ambulatory and activity level upon discharge.

Conclusions

Critically low hemoglobin levels are typically associated with significant symptoms, physical examination findings, and hemodynamic instability. To our knowledge, this is the lowest recorded hemoglobin in a hemodynamically stable patient not requiring cardiac or supplemental oxygen support.

Peer Review reports

Anemia and menometrorrhagia are common and co-occurring conditions in women of premenopausal age [ 1 , 2 ]. Analysis of the global anemia burden from 1990 to 2010 revealed that the prevalence of iron deficiency anemia, although declining every year, remained significantly high, affecting almost one in every five women [ 1 ]. Menstruation is considered largely responsible for the depletion of body iron stores in premenopausal women, and it has been estimated that the proportion of menstruating women in the USA who have minimal-to-absent iron reserves ranges from 20% to 65% [ 3 ]. Studies have quantified that a premenopausal woman’s iron storage levels could be approximately two to three times lower than those in a woman 10 years post-menopause [ 4 ]. Excessive and prolonged uterine bleeding that occurs at irregular and frequent intervals (menometrorrhagia) can be seen in almost a quarter of women who are 40–50 years old [ 2 ]. Women with menometrorrhagia usually bleed more than 80 mL, or 3 ounces, during a menstrual cycle and are therefore at greater risk for developing iron deficiency and iron deficiency anemia. Here, we report an unusual case of a 42-year-old woman with a long history of menometrorrhagia who presented with severe anemia and was found to have a record low hemoglobin level.

A 42-year-old Hispanic woman with no known past medical history presented to our emergency department with the chief complaint of increasing fatigue and dizziness for 2 weeks and mechanical fall at home on day of presentation.

On physical examination, she was afebrile (36.1 °C), blood pressure was 107/47 mmHg with a mean arterial pressure of 69 mmHg, heart rate was 87 beats per minute (bpm), respiratory rate was 17 breaths per minute, and oxygen saturation was 100% on room air. Her height was 143 cm and weight was 45 kg (body mass index 22). She was fully alert and oriented to person, place, time, and situation without any neurological deficits and was speaking in clear, full sentences. She had marked pallor with pale mucous membranes and conjunctiva. She had no palpable lymphadenopathy. She was breathing comfortably on room air and displayed no signs of shortness of breath. Her cardiac examination was notable for a grade 2 systolic flow murmur. Her abdominal examination was unremarkable without palpable masses. On musculoskeletal examination, her extremities were thin, and her fingernails demonstrated koilonychia (Fig. 1 ). She had full strength in lower and upper extremities bilaterally, even though she required assistance with ambulation secondary to weakness and used a wheelchair for mobility for 2 weeks prior to admission. She declined a pelvic examination. No bleeding was noted in any part of her physical examination.

Koilonychia, as seen in our patient above, is a nail disease commonly seen in hypochromic anemia, especially iron deficiency anemia, and refers to abnormally thin nails that have lost their convexity, becoming flat and sometimes concave in shape

She was admitted directly to the intensive care unit after her hemoglobin was found to be critically low at 1.4 g/dL on two consecutive measurements with an unclear etiology of blood loss at the time of presentation. Note that no intravenous fluids were administered prior to obtaining the hemoglobin levels. Upon collecting further history from the patient, she revealed that she has had a lifetime history of extremely heavy menstrual periods: Since menarche at the age of 10 years when her periods started, she has been having irregular menstruation, with periods occurring every 2–3 weeks, sometimes more often. She bled heavily for the entire 5–7 day duration of her periods; she quantified soaking at least seven heavy flow pads each day with bright red blood as well as large-sized blood clots. Since the age of 30 years, her periods had also become increasingly heavier, with intermittent bleeding in between cycles, stating that lately she bled for “half of the month.” She denied any other sources of bleeding.

Initial laboratory data are summarized in Table 1 . Her hemoglobin (Hgb) level was critically low at 1.4 g/dL on arrival, with a low mean corpuscular volume (MCV) of < 50.0 fL. Hematocrit was also critically low at 5.8%. Red blood cell distribution width (RDW) was elevated to 34.5%, and absolute reticulocyte count was elevated to 31 × 10 9 /L. Iron panel results were consistent with iron deficiency anemia, showing a low serum iron level of 9 μg/dL, elevated total iron-binding capacity (TIBC) of 441 μg/dL, low Fe Sat of 2%, and low ferritin of 4 ng/mL. Vitamin B12, folate, hemolysis labs [lactate dehydrogenase (LDH), haptoglobin, bilirubin], and disseminated intravascular coagulation (DIC) labs [prothrombin time (PT), partial thromboplastin time (PTT), fibrinogen, d -dimer] were all unremarkable. Platelet count was 232,000/mm 3 . Peripheral smear showed erythrocytes with marked microcytosis, anisocytosis, and hypochromia (Fig. 2 ). Of note, the patient did have a positive indirect antiglobulin test (IAT); however, she denied any history of pregnancy, prior transfusions, intravenous drug use, or intravenous immunoglobulin (IVIG). Her direct antiglobulin test (DAT) was negative.

A peripheral smear from the patient after receiving one packed red blood cell transfusion is shown. Small microcytic red blood cells are seen, many of which are hypochromic and have a large zone of pallor with a thin pink peripheral rim. A few characteristic poikilocytes (small elongated red cells also known as pencil cells) are also seen in addition to normal red blood cells (RBCs) likely from transfusion

A transvaginal ultrasound and endometrial biopsy were offered, but the patient declined. Instead, a computed tomography (CT) abdomen and pelvis with contrast was performed, which showed a 3.5-cm mass protruding into the endometrium, favored to represent an intracavitary submucosal leiomyoma (Fig. 3 ). Aside from her abnormal uterine bleeding (AUB), the patient was without any other significant personal history, family history, or lab abnormalities to explain her severe anemia.

Computed tomography (CT) of the abdomen and pelvis with contrast was obtained revealing an approximately 3.5 × 3.0 cm heterogeneously enhancing mass protruding into the endometrial canal favored to represent an intracavitary submucosal leiomyoma

The patient’s presenting symptoms of fatigue and dizziness are common and nonspecific symptoms with a wide range of etiologies. Based on her physical presentation—overall well-appearing nature with normal vital signs—as well as the duration of her symptoms, we focused our investigation on chronic subacute causes of fatigue and dizziness rather than acute medical causes. We initially considered a range of chronic medical conditions from cardiopulmonary to endocrinologic, metabolic, malignancy, rheumatologic, and neurological conditions, especially given her reported history of fall. However, once the patient’s lab work revealed a significantly abnormal complete blood count and iron panel, the direction of our workup shifted towards evaluating hematologic causes.

With such a critically low Hgb on presentation (1.4 g/dL), we evaluated for potential sources of blood loss and wanted to first rule out emergent, dangerous causes: the patient’s physical examination and reported history did not elicit any concern for traumatic hemorrhage or common gastrointestinal bleeding. She denied recent or current pregnancy. Her CT scan of abdomen and pelvis was unremarkable for any pathology other than a uterine fibroid. The microcytic nature of her anemia pointed away from nutritional deficiencies, and she lacked any other medical comorbidities such as alcohol use disorder, liver disease, or history of substance use. There was also no personal or family history of autoimmune disorders, and the patient denied any history of gastrointestinal or extraintestinal signs and/or symptoms concerning for absorptive disorders such as celiac disease. We also eliminated hemolytic causes of anemia as hemolysis labs were all normal. We considered the possibility of inherited or acquired bleeding disorders, but the patient denied any prior signs or symptoms of bleeding diatheses in her or her family. The patient’s reported history of menometrorrhagia led to the likely cause of her significant microcytic anemia as chronic blood loss from menstruation leading to iron deficiency.

Over the course of her 4-day hospital stay, she was transfused 5 units of packed red blood cells and received 2 g of intravenous iron dextran. Hematology and Gynecology were consulted, and the patient was administered a medroxyprogesterone (150 mg) intramuscular injection on hospital day 2. On hospital day 4, she was discharged home with follow-up plans. Her hemoglobin and hematocrit on discharge were 8.1 g/dL and 24.3%, respectively. Her symptoms of fatigue and dizziness had resolved, and she was back to her normal baseline ambulatory and activity level.

Discussion and conclusions

This patient presented with all the classic signs and symptoms of iron deficiency: anemia, fatigue, pallor, koilonychia, and labs revealing marked iron deficiency, microcytosis, elevated RDW, and low hemoglobin. To the best of our knowledge, this is the lowest recorded hemoglobin in an awake and alert patient breathing ambient air. There have been previous reports describing patients with critically low Hgb levels of < 2 g/dL: A case of a 21-year old woman with a history of long-lasting menorrhagia who presented with a Hgb of 1.7 g/dL was reported in 2013 [ 5 ]. This woman, although younger than our patient, was more hemodynamically unstable with a heart rate (HR) of 125 beats per minute. Her menorrhagia was also shorter lasting and presumably of larger volume, leading to this hemoglobin level. It is likely that her physiological regulatory mechanisms did not have a chance to fully compensate. A 29-year-old woman with celiac disease and bulimia nervosa was found to have a Hgb of 1.7 g/dL: she presented more dramatically with severe fatigue, abdominal pain and inability to stand or ambulate. She had a body mass index (BMI) of 15 along with other vitamin and micronutrient deficiencies, leading to a mixed picture of iron deficiency and non-iron deficiency anemia [ 6 ]. Both of these cases were of reproductive-age females; however, our patient was notably older (age difference of > 20 years) and had a longer period for physiologic adjustment and compensation.

Lower hemoglobin, though in the intraoperative setting, has also been reported in two cases—a patient undergoing cadaveric liver transplantation who suffered massive bleeding with associated hemodilution leading to a Hgb of 0.6 g/dL [ 7 ] and a patient with hemorrhagic shock and extreme hemodilution secondary to multiple stab wounds leading to a Hgb of 0.7 g/dL [ 8 ]. Both patients were hemodynamically unstable requiring inotropic and vasopressor support, had higher preoperative hemoglobin, and were resuscitated with large volumes of colloids and crystalloids leading to significant hemodilution. Both were intubated and received 100% supplemental oxygen, increasing both hemoglobin-bound and dissolved oxygen. Furthermore, it should be emphasized that the deep anesthesia and decreased body temperature in both these patients minimized oxygen consumption and increased the available oxygen in arterial blood [ 9 ].

Our case is remarkably unique with the lowest recorded hemoglobin not requiring cardiac or supplemental oxygen support. The patient was hemodynamically stable with a critically low hemoglobin likely due to chronic, decades-long iron deficiency anemia of blood loss. Confirmatory workup in the outpatient setting is ongoing. The degree of compensation our patient had undergone is impressive as she reported living a very active lifestyle prior to the onset of her symptoms (2 weeks prior to presentation), she routinely biked to work every day, and maintained a high level of daily physical activity without issue.

In addition, while the first priority during our patient’s hospital stay was treating her severe anemia, her education became an equally important component of her treatment plan. Our institution is the county hospital for the most populous county in the USA and serves as a safety-net hospital for many vulnerable populations, most of whom have low health literacy and a lack of awareness of when to seek care. This patient had been experiencing irregular menstrual periods for more than three decades and never sought care for her heavy bleeding. She, in fact, had not seen a primary care doctor for many years nor visited a gynecologist before. We emphasized the importance of close follow-up, self-monitoring of her symptoms, and risks with continued heavy bleeding. It is important to note that, despite the compensatory mechanisms, complications of chronic anemia left untreated are not minor and can negatively impact cardiovascular function, cause worsening of chronic conditions, and eventually lead to the development of multiorgan failure and even death [ 10 , 11 ].

Availability of data and materials

All data generated or analyzed during this study are included in this published article.

Kassebaum NJ. The global burden of anemia. Hematol Oncol Clin. 2016;30(2):247–308.

Article   Google Scholar  

Donnez J. Menometrorrhagia during the premenopause: an overview. Gynecol Endocrinol. 2011;27(sup1):1114–9.

Cook JD, Skikne BS, Lynch SR, Reusser ME. Estimates of iron sufficiency in the US population. Blood. 1986;68(3):726–31.

Article   CAS   Google Scholar  

Palacios S. The management of iron deficiency in menometrorrhagia. Gynecol Endocrinol. 2011;27(sup1):1126–30.

Can Ç, Gulactı U, Kurtoglu E. An extremely low hemoglobin level due to menorrhagia and iron deficiency anemia in a patient with mental retardation. Int Med J. 2013;20(6):735–6.

Google Scholar  

Jost PJ, Stengel SM, Huber W, Sarbia M, Peschel C, Duyster J. Very severe iron-deficiency anemia in a patient with celiac disease and bulimia nervosa: a case report. Int J Hematol. 2005;82(4):310–1.

Kariya T, Ito N, Kitamura T, Yamada Y. Recovery from extreme hemodilution (hemoglobin level of 0.6 g/dL) in cadaveric liver transplantation. A A Case Rep. 2015;4(10):132.

Dai J, Tu W, Yang Z, Lin R. Intraoperative management of extreme hemodilution in a patient with a severed axillary artery. Anesth Analg. 2010;111(5):1204–6.

Koehntop DE, Belani KG. Acute severe hemodilution to a hemoglobin of 1.3 g/dl tolerated in the presence of mild hypothermia. J Am Soc Anesthesiol. 1999;90(6):1798–9.

Georgieva Z, Georgieva M. Compensatory and adaptive changes in microcirculation and left ventricular function of patients with chronic iron-deficiency anaemia. Clin Hemorheol Microcirc. 1997;17(1):21–30.

CAS   PubMed   Google Scholar  

Lanier JB, Park JJ, Callahan RC. Anemia in older adults. Am Fam Phys. 2018;98(7):437–42.

Download references

Acknowledgements

Not applicable.

No funding to be declared.

Author information

Authors and affiliations.

Department of Medicine, University of Southern California, Los Angeles, CA, USA

Audrey L. Chai & Owen Y. Huang

Division of Pulmonary, Critical Care and Sleep Medicine, Department of Medicine, University of Southern California, Los Angeles, CA, USA

Rastko Rakočević & Peter Chung

You can also search for this author in PubMed   Google Scholar

Contributions

AC, OH, RR, and PC managed the presented case. AC performed the literature search. AC, OH, and RR collected all data and images. AC and OH drafted the article. RR and PC provided critical revision of the article. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Audrey L. Chai .

Ethics declarations

Ethics approval and consent to participate, consent for publication.

Written informed consent was obtained from the patient for publication of this case report and any accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Chai, A.L., Huang, O.Y., Rakočević, R. et al. Critical iron deficiency anemia with record low hemoglobin: a case report. J Med Case Reports 15 , 472 (2021). https://doi.org/10.1186/s13256-021-03024-9

Download citation

Received : 25 March 2021

Accepted : 21 July 2021

Published : 13 September 2021

DOI : https://doi.org/10.1186/s13256-021-03024-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Menometrorrhagia
• Iron deficiency
• Critical care
• Transfusion

Journal of Medical Case Reports

ISSN: 1752-1947

• Submission enquiries: Access here and click Contact Us
• General enquiries: [email protected]

An official website of the United States government

Here’s how you know

Official websites use .gov A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

• Entire Site
• Research & Funding
• Health Information
• About NIDDK
• Blood Diseases

Anemia of Inflammation or Chronic Disease

• Español

On this page:

What is anemia of inflammation?

Why is anemia of inflammation also called anemia of chronic disease, are there other types of anemia, how common is anemia of inflammation, who is more likely to have anemia of inflammation, does anemia of inflammation lead to other health problems, what are the symptoms of anemia of inflammation, what causes anemia of inflammation, chronic conditions that cause anemia of inflammation, other causes of inflammation that may lead to anemia, how do health care professionals diagnose anemia of inflammation, how do health care professionals treat anemia of inflammation, can i prevent anemia of inflammation, how does eating, diet, and nutrition affect anemia of inflammation, clinical trials for anemia of inflammation, what are clinical trials for anemia of inflammation, what clinical studies for anemia of inflammation are looking for participants.

Anemia of inflammation, also called anemia of chronic disease or ACD, is a type of anemia that affects people who have conditions that cause inflammation, such as infections, autoimmune diseases , cancer , and chronic kidney disease (CKD) .

Anemia is a condition in which your blood has fewer red blood cells than normal. Your red blood cells may also have less hemoglobin than normal. Hemoglobin is the iron-rich protein that allows red blood cells to carry oxygen from your lungs to the rest of your body. Your body needs oxygen to work properly. With fewer red blood cells or less hemoglobin, your body may not get enough oxygen.

In anemia of inflammation, you may have a normal or sometimes increased amount of iron stored in your body tissues, but a low level of iron in your blood. Inflammation may prevent your body from using stored iron to make enough healthy red blood cells, leading to anemia.

Anemia of inflammation is also called anemia of chronic disease because this type of anemia commonly occurs in people who have chronic conditions that may be associated with inflammation .

There are many types of anemia . Common types include

• iron-deficiency anemia , a condition in which the body’s stored iron is used up, causing the body to make fewer healthy red blood cells. In people with iron-deficiency anemia, iron levels are low in both body tissues and the blood. This is the most common type of anemia.
• pernicious anemia , which is caused by a lack of vitamin B12.
• aplastic anemia , a condition in which the bone marrow doesn’t make enough new red blood cells, white blood cells, and platelets because the bone marrow’s stem cells are damaged.
• hemolytic anemia , a condition in which red blood cells are destroyed earlier than normal.

Anemia of inflammation is the second most common type of anemia, after iron-deficiency anemia. 1

While anemia of inflammation can affect people of any age, older adults are more likely to have this type of anemia because they are more likely to have chronic diseases that cause inflammation. In the United States, about 1 million people older than age 65 have anemia of inflammation. 2

Anemia of inflammation is typically mild or moderate, meaning that hemoglobin levels in your blood are lower than normal but not severely low. If your anemia becomes severe, the lack of oxygen in your blood can cause symptoms, such as feeling tired or short of breath. Severe anemia can become life-threatening.

In people who have CKD, severe anemia can increase the chance of developing heart problems .

Anemia of inflammation typically develops slowly and may cause few or no symptoms. In fact, you may only experience symptoms of the disease that is causing anemia and not notice additional symptoms.

Symptoms of anemia of inflammation are the same as in any type of anemia and include

• a fast heartbeat
• fainting or feeling dizzy or light-headed
• feeling tired or weak
• getting tired easily during or after physical activity
• shortness of breath

Experts think that when you have an infection or disease that causes inflammation, your immune system causes changes in how your body works that may lead to anemia of inflammation.

• Your body may not store and use iron normally.
• Your kidneys may produce less erythropoietin (EPO), a hormone that signals your bone marrow—the spongy tissue inside most of your bones—to make red blood cells.
• Your bone marrow may not respond normally to EPO, making fewer red blood cells than needed.
• Your red blood cells may live for a shorter time than normal, causing them to die faster than they can be replaced.

Many different chronic conditions can cause inflammation that leads to anemia, including

• autoimmune diseases, such as rheumatoid arthritis or lupus
• chronic infections, such as HIV/AIDS and tuberculosis
• inflammatory bowel disease (IBD), such as Crohn’s disease or ulcerative colitis
• other chronic diseases that involve inflammation, such as diabetes and heart failure

In people with certain chronic conditions, anemia may have more than one cause. For example

• Causes of anemia in CKD  may include inflammation, low levels of EPO due to kidney damage, or low levels of the nutrients needed to make red blood cells. Hemodialysis to treat CKD may also lead to iron-deficiency anemia.
• People with IBD may have both iron-deficiency anemia due to blood loss and anemia of inflammation.
• In people who have cancer, anemia may be caused by inflammation, blood loss, and cancers that affect or spread to the bone marrow. Cancer treatments such as chemotherapy and radiation therapy may also cause or worsen anemia.

While anemia of inflammation typically develops slowly, anemia of critical illness is a type of anemia of inflammation that develops quickly in patients who are hospitalized for severe acute infections, trauma, or other conditions that cause inflammation.

In some cases, older adults develop anemia of inflammation that is not related to an underlying infection or chronic disease. Experts think that the aging process may cause inflammation and anemia.

Health care professionals use a medical history and blood tests to diagnose anemia of inflammation.

Medical history

A health care professional will ask about your history of infections or chronic diseases that may lead to anemia of inflammation.

Blood tests

Health care professionals use blood tests to check for signs of anemia of inflammation, other types of anemia, or other health problems. A health care professional will take a blood sample from you and send the sample to a lab to test.

Blood count tests can check many parts and features of your blood, including

• the number of red blood cells
• the average size of red blood cells
• the amount of hemoglobin in your blood and in your red blood cells
• the number of developing red blood cells, called reticulocytes, in your blood

Some of these blood count tests and others may be combined in a test called a complete blood count . A blood smear may be used to examine the size, shape, and number of red blood cells in your blood.

A health care professional may also use blood tests to check the amount of iron in your blood and stored in your body. These tests may measure

• iron in your blood
• transferrin, a protein in your blood that carries iron
• ferritin, the protein that stores iron in your body’s cells

A health care professional may diagnose anemia of inflammation if blood test results suggest that you have anemia, a low level of iron in your blood, and a normal level of iron stored in your body tissues.

If blood test results suggest you have anemia of inflammation but the cause is unknown, a health care professional may perform additional tests to look for the cause.

Health care professionals treat anemia of inflammation by treating the underlying condition and by treating the anemia with medicines and occasionally with blood transfusions.

Treating the underlying condition

Health care professionals typically treat anemia of inflammation by treating the underlying condition that is causing inflammation. If treatments are available that can reduce the inflammation, the treatments may cause the anemia to improve or go away. For example, taking medicines to treat inflammation in rheumatoid arthritis can improve anemia.

A health care professional may prescribe the erythropoiesis-stimulating agents (ESAs) epoetin alpha or darbepoetin alpha to treat anemia related to CKD, chemotherapy treatments for cancer, or certain treatments for HIV. ESAs cause the bone marrow to make more red blood cells. Health care professionals typically give ESAs as shots and may teach you how to give yourself these shots at home. A health care professional may prescribe iron supplements, given as pills or shots, to help ESAs work.

If you’re on hemodialysis, you may be able to receive intravenous (IV) ESAs and iron supplements during hemodialysis. Read more about treatments for anemia in CKD .

Blood transfusions

In some cases, health care professionals may use blood transfusions to treat severe anemia of inflammation. A blood transfusion can quickly increase the amount of hemoglobin in your blood and boost oxygen.

Experts have not yet found a way to prevent anemia of inflammation. For some chronic conditions that cause inflammation, treatments may be available to reduce or prevent the inflammation that can lead to anemia. Talk with your doctor about treatments and follow the treatment plan your doctor recommends.

If you have a chronic condition that is causing anemia of inflammation, follow the advice of your doctor or dietitian about healthy eating and nutrition.

The National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and other components of the National Institutes of Health (NIH) conduct and support research into many diseases and conditions, including blood diseases.

Clinical trials—and other types of clinical studies —are part of medical research and involve people like you. When you volunteer to take part in a clinical study, you help doctors and researchers learn more about disease and improve health care for people in the future.

Researchers are studying many aspects of anemia of inflammation, including new treatments for this condition.

Find out if clinical studies are right for you .

You can find clinical studies on anemia of inflammation at www.ClinicalTrials.gov . In addition to searching for federally funded studies, you can expand or narrow your search to include clinical studies from industry, universities, and individuals; however, the NIH does not review these studies and cannot ensure they are safe. Always talk with your health care provider before you participate in a clinical study.

This content is provided as a service of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), part of the National Institutes of Health. NIDDK translates and disseminates research findings to increase knowledge and understanding about health and disease among patients, health professionals, and the public. Content produced by NIDDK is carefully reviewed by NIDDK scientists and other experts.

The NIDDK would like to thank: Robert T. Means, Jr., M.D., East Tennessee State University Quillen College of Medicine

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Risk factors for incident anemia of chronic diseases: A cohort study

Contributed equally to this work with: Yun-Gyoo Lee, Yoosoo Chang

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing

Affiliation Department of Internal Medicine, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul, Korea

Roles Data curation, Formal analysis, Software, Writing – review & editing

Affiliations Department of Occupational and Environmental Medicine, Kangbuk Samsung Hospital, Sungkyunkwan University, School of Medicine, Seoul, Korea, Center for Cohort Studies, Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University, School of Medicine, Seoul, Korea

Roles Supervision, Writing – review & editing

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Supervision, Validation, Writing – original draft, Writing – review & editing

¶ ‡ These authors also contributed equally to this work.

Roles Conceptualization, Supervision, Writing – review & editing

* E-mail: [email protected] , [email protected]

• Yun-Gyoo Lee, 
• Yoosoo Chang, 
• Jihoon Kang, 
• Dong-Hoe Koo, 
• Seung-Sei Lee, 
• Seungho Ryu, 
• Sukjoong Oh

• Published: May 6, 2019
• https://doi.org/10.1371/journal.pone.0216062
• Reader Comments

Anemia of chronic disease (ACD) refers to hypoproliferative anemia in the context of acute or chronic activation of the immune system. There is a paucity of prospective data addressing the risk factors for ACD development. An association between common chronic diseases and ACD was examined cross-sectionally and longitudinally.

A cohort of 265,459 healthy participants without ACD at baseline were prospectively followed annually or biennially.

During average follow-up period of 62 months, 4,906 participants developed ACD (incidence rate 3.58 per 1000 person-years). Multivariable-adjusted hazard ratio (HR) [95% confidence interval (CI)] for incident ACD comparing estimated glomerular filtration rate 30–60 and < 30 vs. ≥ 60 ml/min/1.73 m 2 were 3.93 [3.18–4.85] and 39.11 [18.50–82.69]; HRs [95% CI] for ACD comparing prediabetes and diabetes vs. normal were 1.19 [1.12–1.27] and 2.46 [2.14–2.84], respectively. HRs [95% CI] for incident ACD comparing body-mass-index (BMI) of < 18.5, 23–24.9 and ≥ 25 vs. 18.5–22.9 kg/m 2 were 0.89 [0.78–1.00], 0.89 [0.80–0.99] and 0.78 [0.66–0.91], respectively. HRs [95% CI] for incident ACD comparing prehypertension and hypertension vs. normal were 0.79 [0.73–0.86] and 1.10 [0.99–1.23], respectively. Metabolic syndrome, hypertension, chronic liver disease, and chronic obstructive pulmonary disease were not associated with incident ACD.

Conclusions

The severity of chronic kidney disease and diabetic status were independently associated with an increased incidence of ACD, whereas prehypertension and an increasing BMI were significantly associated with decreased risk of ACD.

Citation: Lee Y-G, Chang Y, Kang J, Koo D-H, Lee S-S, Ryu S, et al. (2019) Risk factors for incident anemia of chronic diseases: A cohort study. PLoS ONE 14(5): e0216062. https://doi.org/10.1371/journal.pone.0216062

Editor: Gianpaolo Reboldi, Universita degli Studi di Perugia, ITALY

Received: November 27, 2018; Accepted: April 12, 2019; Published: May 6, 2019

Copyright: © 2019 Lee et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the manuscript and its Supporting Information files.

Funding: The authors received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Anemia of chronic disease (ACD) refers to normochromic, normocytic, hypoproliferative anemia in the context of acute or chronic inflammatory states, including infections, cancers, and autoimmune conditions.[ 1 , 2 ] Some epidemiological studies have reported that ACD also occurs in clinical conditions accompanied by mild but persistent inflammation including chronic kidney disease (CKD), diabetes mellitus, and aging.[ 3 – 5 ] The prevalence of anemia from most causes has decreased globally between 1990 and 2010, but ACD is expected to increase as population ages.[ 6 – 8 ]

Although the underlying pathophysiology of ACD is multifactorial, hepcidin may play a central role in ACD.[ 9 ] Chronic inflammation elevates pro-inflammatory cytokines, including interleukin-6, which centrally mediates hepcidin synthesis. Hepcidin inhibits iron absorption in the intestine and release of recycled iron from macrophages, resulting in reduced efficiency of iron recycling from red blood cells. This functional iron deficiency leads to impaired proliferation of erythroid progenitor cells in the marrow, resulting in iron-restrictive anemia.[ 3 ]

ACD is common but often overlooked in actual clinical practice and the risk factors of ACD is not fully understood. CKD leads to dysfunction of renal erythropoietin-producing cells resulting in normocytic normochromic anemia, which was present in nearly half of patients with CKD.[ 10 , 11 ] Type 2 diabetes increases the risk for anemia by two or three times, which affects 10–15% of patients with type 2 diabetes.[ 12 – 14 ] In these studies, anemia in diabetic patients can be considered as ACD, including the exclusion of iron deficiency anemia and other causes of secondary influences on hemoglobin levels.[ 14 ] ACD is also frequently diagnosed in the elderly (>65 years); a few population-based studies have shown that 17% of the elderly are anemic,[ 15 ] and 70% of hospitalized elderly patients with anemia were found to have ACD.[ 5 ] However, most studies focused on specific single disease or elderly population and were cross-sectional studies limited by the temporal ambiguity between risk factors and anemia. Until now, there is a paucity of prospective cohort study to demonstrate the risk factors for the development of ACD in general population. We examined a prospective relationship of common chronic diseases and their severity with the development of ACD in a large cohort of young and middle-aged Korean adults who underwent a regular health screening examination.

Patients and methods

Study population.

The Kangbuk Samsung Health Study (KSHS) is a cohort study of Korean men and women men and women ≥ 18 years of age who underwent a comprehensive regular (annual or biennial) health examination at Kangbuk Samsung Hospital Total Healthcare Centers in Republic of Korea.[ 16 ] The current analyses included all study participants with at least one follow-up visit who underwent a comprehensive health examination between 2005 and 2015 and were followed annually or biennially until December 2016 (n = 304,229). ACD was defined as having anemia without evidence of nutritional anemia or gastrointestinal blood loss.[ 9 ] The selection process of study participants is shown in Fig 1 . We excluded subjects with missing data on hemoglobin, ferritin, or mean corpuscular volume (MCV) at baseline (n = 3,838), subjects with positive for fecal occult blood tests (n = 9,680), subjects with iron deficiency anemia (n = 12,993) or macrocytic anemia (n = 129), subjects with endoscopic findings including gastric ulcers, duodenal ulcer, inflammatory bowel disease, angiodysplasia, or other gastrointestinal malignancies (esophageal cancer, gastric cancer, or colorectal cancer) (n = 5,724), or subjects with a history of physician-diagnosed malignancy (n = 4,335). After excluding 36,699 subjects, 267,530 participants were included in the baseline cross-sectional study. The cohort participants were not registered at once, but were made up of KSHS cohort in the form of new subjects being added each year. ( S1 Table ) For cohort study, we further excluded an additional 2,071 subjects with ACD at baseline and finally analyzed 265,459 subjects.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pone.0216062.g001

Measurements

Anthropometric measurements and procedures for obtaining blood samples were described previously.[ 16 , 17 ] In accordance with WHO criteria, anemia was defined as a hemoglobin level < 13.0 g/dL in men and < 12.0 g/dL in women.[ 18 , 19 ] Iron deficiency was defined as anemia with a transferrin saturation rate < 16%, or a serum ferritin concentration < 30 ng/mL. Transferrin saturation was calculated by dividing serum iron by total iron-binding capacity. Macrocytic anemia was defined as anemia with MCV > 100 fL.[ 20 ] CKD was defined as estimated glomerular filtration rate (eGFR) was <60 ml/min/1.73 m 2 . Body mass index (BMI) was categorized according to Asian-specific criteria;[ 21 ] underweight, BMI < 18.5 kg/m 2 ; normal weight, BMI of 18.5–22.9 kg/m 2 ; overweight, BMI of 23–24.9 kg/m 2 ; and obese, BMI ≥ 25 kg/m 2 . Metabolic syndrome (MetS) was defined using harmonized criteria.[ 22 , 23 ] Blood pressure (BP) was categorized into normal BP (BP of <120/<80mmHg), prehypertension (systolic BP of 120–139 mmHg or diastolic BP of 80–89 mmHg) and hypertension (systolic BP of ≥140mmHg or diastolic BP of ≥90 mmHg). Diabetes mellitus was defined as either fasting serum glucose ≥126 mg/dL, glycated hemoglobin (HbA1c) ≥ 6.5%, or the use of blood glucose-lowering drugs, and prediabetes was defined as either fasting serum glucose 100–125 mg/dL or HbA1c 5.7–6.4%. Subjects were considered to have chronic liver disease if either serum hepatitis B surface antigen or anti-hepatitis C antibody was positive, or they had an ultrasonographic diagnosis of fatty liver, or liver cirrhosis. Chronic obstructive pulmonary disease (COPD) was defined as forced expiratory volume in 1 sec/forced vital capacity ratio <0.7.

Statistical analyses

We first evaluated the cross-sectional relationship between comorbidities and ACD at baseline, and then we analyzed the prospective relationship of comorbidities and incident ACD in a cohort study. To compare the characteristics of the study participants between the groups, we used an independent t-tests for continuous variables or χ 2 tests for categorical variables. To determine the cross-sectional relationship between comorbidities and ACD, we used a logistic regression model to estimate odds ratios (ORs) with 95% confidence interval (CI). Then, the hazard ratio (HR) and 95% CI were calculated for incident ACD according to comorbidities. Each participant was followed from baseline exam until either development of ACD or the last health exam conducted prior to December 31, 2016, whichever came first. The incidence rate was calculated as number of incident cases divided by number of person-years of follow-up. Since ACD was known to have developed between the two visits, but the precise time at which it developed was unknown, a parametric proportional hazard model was used to account for this type of interval censoring ( stpm command in Stata). To determine linear trends of incidence, the number of categories was used as a continuous variable and tested on each model.

The models were initially adjusted for age and sex, and then were adjusted for smoking, alcohol intake, physical activity, education level, center, and year of screening examination. All analyses were performed using STATA version 15.0 (StataCorp, College Station, Texas, USA).

This study was approved by the Institutional Review Board of Kangbuk Samsung Hospital (KBSMC 2015-07-019). The acquisition of informed consent was waived, as we retrospectively accessed only data that were de-identified. All data were fully anonymized before our analyses.

Baseline cross-sectional study

A total of 267,530 participants were included in analyses for evaluation of cross-sectional relationship between chronic diseases and prevalent ACD. The participants who made up of study cohort are presented in S1 Table by the year of registration. Of these, 2,071 (0.77%) had ACD at baseline. The baseline characteristics of the study participants by prevalence of prevalent ACD are presented in S2 Table . Age, HDL-C, and high sensitivity C-reactive protein were positively associated with the prevalent ACD, whereas BMI, uric acid, eGFR, fasting glucose, total cholesterol, LDL-C, triglycerides, alanine aminotransferase, aspartate aminotransferase, gamma glutamyl transaminase, and HOMA-IR were negatively associated with prevalent ACD. The proportions of male, current smoking, alcohol intake, vigorous exercise, higher education level, MeS, hypertension, chronic liver disease, and obesity were also negatively associated with prevalent ACD. The proportion of patients with CKD was positively associated with prevalent ACD.

S3 Table shows the baseline cross-sectional relationships between chronic diseases and prevalent ACD. The baseline severity of eGFR and diabetic status were significantly associated with an increased prevalence of ACD ( P -trend < 0.001). In contrast, the baseline number of MetS traits, BP category, and low BMI category were significantly associated with a decreased prevalence of ACD ( P -trend < 0.001). The chronic liver disease and COPD were not significantly associated with the prevalence of ACD.

In a multivariate model adjusted for age, sex, smoking, alcohol intake, physical activity, education level, center, and year of screening examination, decreased baseline eGFR and severe diabetic status were positively associated with an increased prevalence of ACD; number of MetS traits and higher levels of BP and BMI were inversely associated with prevalent ACD.

Prospective cohort study

After excluding 2,071 subjects with baseline ACD, 265,459 subjects were included in cohort study to investigate the risk factors for incident ACD. The baseline characteristics of the cohort study participants by incident ACD are presented in Table 1 . Table 2 shows the prospective associations between the chronic diseases and their severity with ACD among subjects without ACD at baseline. During 1,368,691.2 person-years of follow-up, 4,906 participants developed ACD (incidence rate 3.58 per 1,000 person-years). The average follow-up period for subjects who did not develop ACD was 5.2 years.

https://doi.org/10.1371/journal.pone.0216062.t001

https://doi.org/10.1371/journal.pone.0216062.t002

Baseline severity of eGFR and diabetic status were associated with an increased risk of incident ACD in a graded and dose-response manner ( P for trend < 0.001). Multivariable-adjusted HR (95% CI) for ACD comparing eGFR 30–60 and < 30 vs. ≥ 60 ml/min/1.73 m 2 were 3.93 (3.18–4.85) and 39.11 (18.50–82.69), respectively ( P for trend < 0.001). And the multivariable adjusted HR (95% CI) for ACD comparing prediabetes and diabetes vs. normal were 1.19 (1.12–1.27) and 2.46 (2.14–2.84), respectively ( P for trend < 0.001). Increasing BMI was inversely associated with incident ACD. HR (95% CI) for ACD comparing BMIs of <18.5, 23–24.9, and >25 vs. 18.5–22.9 kg/m 2 were 0.89 (0.78–1.00), 0.89 (0.80–0.99) and 0.78 (0.66–0.91), respectively. Prehypertension was associated with a decreased risk of ACD with corresponding HR (95% CI) of 0.79 (0.73–0.86). The cumulative incidence of ACD was displayed in Fig 2 by CKD groups ( Fig 2A ), diabetes categories ( Fig 2B ), BMI groups ( Fig 2C ) and BP categories ( Fig 2D ). Metabolic syndrome, hypertension, chronic liver disease, and COPD were not associated with the incidence of ACD in multivariate analyses.

Cumulative incidence of anemia of chronic disease according to chronic kidney disease groups (A), diabetes categories (B), BMI groups (C) and BP categories (D).

https://doi.org/10.1371/journal.pone.0216062.g002

In this large cohort study of young and middle-aged Korean men and women, both the cross-sectional and cohort analyses demonstrated that decreased eGFR and severe diabetic status were associated with increased risk of ACD. Conversely, higher BMI categories were associated with a decreased risk of developing ACD in dose-response manner. These associations persisted even after adjusting for possible confounders. Our results provide information about the relative risk for developing ACD among patients with various common chronic diseases.

In our study of 267,530 participants, the proportion of those with ACD was 0.77% in overall population, 0.9% of participants with diabetes and 6.7% of participants with eGFR < 60 ml/min/1.73 m 2 . However, previous studies of patients with diabetes or CKD reported a higher prevalence of anemia of 10–20%.[ 10 – 13 ] Our lower prevalence of ACD is partly explained by the following reasons: previous studies included nutritional anemia in the anemia outcome and reported a higher prevalence, and chronic diseases in hospital populations from other studies were more severe than in the general population of our study.

Previous studies have shown an association between eGFR, diabetes status and ACD.[ 10 – 13 ] Diabetes and CKD are considered chronic inflammatory states characterized by an increased level of pro-inflammatory cytokines involved in erythroid progenitor cells.[ 24 , 25 ] In our study, subjects with eGFR < 30 ml/min/1.73 m 2 had almost 40 fold increased risk of developing ACD compared with those with a eGFR ≥ 60 ml/min/1.73 m 2 . The risk of ACD in prediabetic and diabetic participants increased by 1.19 and 2.46 times compared to euglycemic participants, respectively. In CKD or diabetes, incident ACD increased as severity of these diseases increased. Contrary to our expectation, our study found that obese subjects and/or those with prehypertension had a lower risk of ACD. Given that obesity is characterized by chronic low-grade inflammation,[ 26 , 27 ] a previous study also hypothesized that hemoglobin concentration might be lower in individuals with overweight or obesity.[ 28 ] However, this cross-sectional study using data from the third National Health and Nutrition Examination Survey (NHANES III) in a US population reported that overweight (BMI 25–29.9 kg/m 2 ) and obese (BMI > 30 kg/m 2 ) subjects were not associated with anemia compared with normal-weight subjects, while increasing BMI was associated with reduced transferrin saturation and higher serum ferritin, suggesting mechanisms of obesity-related inflammation. A recent systematic review summarized that obese subjects tend to have higher hemoglobin levels than non-obese subjects,[ 29 ] which is consistent with our findings.

The mechanisms underlying the inverse association between obesity and the risk of developing ACD are unknown. Therefore, we hypothesized the following reasons. First, subjects who might have developed obesity-induced iron deficiency anemia (IDA) were not included in our study population. Because obesity-associated inflammation affects iron homeostasis and results in an iron deficiency[ 30 ] and our study population excluded subjects with IDA. In turn, obese subjects were less likely to experience anemic outcomes. Second, obese subjects have more comorbidities, which may increase hemoglobin. Obese subjects are more likely to have obstructive sleep apnea and other obesity-related respiratory disorders, which result in chronic tissue hypoxia and lead to polycythemia.[ 31 , 32 ] Third, obese subjects are less likely to be malnourished, because excessive caloric intake can develop into obesity.[ 26 ] Therefore, adequate or overnutrition in obese subjects might be associated with a reduced risk of anemia.

Several limitations of this study should be discussed. First, although we identified ACD cases after excluding nutritional anemia or possible blood loss, we may have included unexplained anemia without a proven etiology or clonal anemia.[ 33 ] However, given the younger age of our study population, the proportion of unexplained or clonal anemia cases would be minimal. Second, we have no data regarding hepcidin, a key ACD regulatory hormone and proinflammatory factors. Further mechanistic studies are required to elucidate the association between chronic diseases and ACD. Third, our findings are limited by selection bias of case definition. For example, the definition of chronic liver disease could be incomplete without consideration of laboratory findings (liver enzymes, albumin, prothrombin time). The etiology of macrocytic anemia includes chronic illness other than nutritional anemia. [ 20 ] However, the number of macrocytic anemia was very low (0.04%) and the impact of selection bias would be minimal. Finally, our study data included young and middle-aged Korean subjects who regularly attended health screening examinations, which could limit the generalizability of our results to other populations with different characteristics of age and race/ethnicity. However, our study provides reliable estimates of ACD risk because of the large sample size, the use of well-defined measurements, and the prospective nature of the cohort study.

This is the one of the largest cohort studies demonstrating a prospective association between chronic diseases and the incidence of ACD. Our cross-sectional and cohort studies identified that prehypertension and increasing BMI are independently associated with a decreased risk of ACD. Although the relationship between CKD or diabetes and anemia is well known, the negative relationships of obesity and prehypertension for incident anemia are novel and interesting findings. Anemia is a non-negligible, but unrecognizable risk. These data will allow clinicians to identify at-risk subjects for intervention. Further studies are warranted to confirm our results.

Supporting information

S1 table. the participants who made up the kshs cohort by the year of registration..

https://doi.org/10.1371/journal.pone.0216062.s001

S2 Table. Baseline characteristics of study participants by prevalence of anemia of chronic disease (ACD).

https://doi.org/10.1371/journal.pone.0216062.s002

S3 Table. Odds ratios (95% CI) by chronic disease and the prevalence of anemia of chronic disease (ACD) in baseline cross-sectional study.

https://doi.org/10.1371/journal.pone.0216062.s003

• View Article
• PubMed/NCBI
• Google Scholar
• 21. Organization WH. The Asia-Pacific perspective: redefining obesity and its treatment. Sydney: Health Communications Australia; 2000.

If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

To log in and use all the features of Khan Academy, please enable JavaScript in your browser.

Health and medicine

Course: health and medicine   >   unit 6.

• What is iron deficiency?
• Microcytic anemia

Iron deficency anemia diagnosis

• Anemia of chronic disease
• Chronic disease vs iron deficiency anemia
• Iron deficiency and anemia of chronic disease treatment

Want to join the conversation?

• Upvote Button navigates to signup page
• Downvote Button navigates to signup page
• Flag Button navigates to signup page

Video transcript

Select a Community

• MB 1 Preclinical Medical Students
• MB 2/3 Clinical Medical Students
• ORTHO Orthopaedic Surgery

Are you sure you want to trigger topic in your Anconeus AI algorithm?

You are done for today with this topic.

Would you like to start learning session with this topic items scheduled for future?

Anemia of Chronic Disease

• A 56-year-old man presents to his annual physical exam with weight loss, fatigue, and weakness. He is a long-term smoker and also recently immigrated from Southeast Asia. He complains of a chronic cough for the past year with occasional bloody mucus. He agrees to undergo imaging for lung lesions and a blood test for tuberculosis. Iron studies reveal decreased serum iron, increased ferritin, and decreased TIBC. Peripheral blood smear shows basophilic stippling.
• normochromic, normocytic anemia
• chronic inflammatory condition
• tuberculosis
• normocytic and normochromic anemia
• iron is stored in macrophages or bound with ferritin
• inflammation causes release of hepcidin by the liver
• ↑ hepcidin inhibits iron absorption from diet and prevents release of iron bound by ferritin from macrophages
• generalized weakness
• shortness of breath
• tachycardia
• ↓ serum iron
• ↓ TIBC, transferrin saturation
• Normal MCV, can progress to ↓ MCV with longstanding disease
• normochromic RBCs
• may be normocytic or microcytic
• can be seen in alcohol abuse, lead poisoning, thalassemias, and hereditary pyrimidine 5'-nucleotidase deficiency
• Anemia of renal disease
• Iron deficiency anemia
• Aplastic anemia
• Treat underlying disease
• make sure iron stores are sufficient
• if insufficient, patients may be resistant to EPO
• Severe anemia
• Varied based on underlying inflammatory condition
• - Anemia of Chronic Disease

Please Login to add comment

Indications and Limitations of Use

Aranesp ® (darbepoetin alfa) is indicated for the treatment of anemia due to chronic kidney disease (CKD), including patients on dialysis and patients not on dialysis.

EPOGEN ® (epoetin alfa) is indicated for the treatment of anemia due to chronic kidney disease (CKD) in patients on dialysis to decrease the need for red blood cell (RBC) transfusion.

Limitations of Use:

• Aranesp ® and EPOGEN ® have not been shown to improve quality of life, fatigue, or patient well-being.
• Aranesp ® and EPOGEN ® are not indicated for use as a substitute for RBC transfusions in patients who require immediate correction of anemia.

Please click to see accompanying Aranesp ® full prescribing information and EPOGEN ® full prescribing information , including Boxed WARNINGS and Medication Guide.

• Resource Library
• Support And Services
• Create Account

Aranesp ® Case Study: Anemia in CKD—New Patient on Hemodialysis*

This hypothetical case study documents the experience of a patient who has anemia due to chronic kidney disease, CKD, and is new to hemodialysis. This program is promotional in nature and subject to Food and Drug Administration rules and Amgen internal policies for providing fair balance on the approved uses of their products, including full disclosure of the associated risks and benefits.

You may also be interested in:

Case Study: Anemia in CKD Management

Addressing Anemia Due to CKD in Patients on Dialysis: About Aranesp ®

Aranesp ® Case Study: Anemia Due to CKD in the Dialysis Patient*

Aranesp ® Case Study: Anemia in CKD—Post-Hospitalized Patient on Dialysis*

Aranesp ® Erythrokinetics

En Español: Living Well—Peritoneal Dialysis Brochure

En Español: Protein Nutrition Card

En Español: Aranesp® Patient Education Brochure

* Case studies and patient profiles are hypothetical

Important Safety Information

WARNING: ESAs INCREASE THE RISK OF DEATH, MYOCARDIAL INFARCTION, STROKE, VENOUS THROMBOEMBOLISM, THROMBOSIS OF VASCULAR ACCESS AND TUMOR PROGRESSION OR RECURRENCE

Chronic Kidney Disease:

• In controlled trials, patients experienced greater risks for death, serious adverse cardiovascular reactions, and stroke when administered erythropoiesis-stimulating agents (ESAs) to target a hemoglobin level of greater than 11 g/dL.
• No trial has identified a hemoglobin target level, ESA dose, or dosing strategy that does not increase these risks.
• Use the lowest Aranesp® or EPOGEN® dose sufficient to reduce the need for red blood cell (RBC) transfusions.
• ESAs shortened overall survival and/or increased the risk of tumor progression or recurrence in clinical studies of patients with breast, non-small cell lung, head and neck, lymphoid, and cervical cancers.
• To decrease these risks, as well as the risk of serious cardiovascular and thromboembolic reactions, use the lowest dose needed to avoid RBC transfusions.
• Use ESAs only for anemia from myelosuppressive chemotherapy.
• ESAs are not indicated for patients receiving myelosuppressive chemotherapy when the anticipated outcome is cure.
• Discontinue following the completion of a chemotherapy course.

Perisurgery (EPOGEN ® ):

• Due to increased risk of Deep Venous Thrombosis (DVT), DVT prophylaxis is recommended.
• Uncontrolled hypertension
• Pure red cell aplasia (PRCA) that begins after treatment with Aranesp®, EPOGEN®, or other erythropoietin protein drugs
• Serious allergic reactions to Aranesp® or EPOGEN®
• EPOGEN® from multidose vials contains benzyl alcohol and is contraindicated in neonates, infants, pregnant women, and lactating women.
• Use caution in patients with coexistent cardiovascular disease and stroke.
• Patients with CKD and an insufficient hemoglobin response to ESA therapy may be at even greater risk for cardiovascular reactions and mortality than other patients. A rate of hemoglobin rise of > 1 g/dL over 2 weeks may contribute to these risks.
• In controlled clinical trials, ESAs increased the risk of death in patients undergoing coronary artery bypass graft surgery (CABG) and the risk of deep venous thrombosis (DVT) in patients undergoing orthopedic procedures.
• Control hypertension prior to initiating and during treatment with Aranesp® or EPOGEN®.
• Aranesp® and EPOGEN® increase the risk of seizures in patients with CKD. Monitor patients closely for new-onset seizures, premonitory symptoms, or change in seizure frequency.
• For lack or loss of hemoglobin response to Aranesp® or EPOGEN®, initiate a search for causative factors. If typical causes of lack or loss of hemoglobin response are excluded, evaluate for PRCA.
• Cases of PRCA and of severe anemia, with or without other cytopenias that arise following the development of neutralizing antibodies to erythropoietin have been reported in patients treated with Aranesp® or EPOGEN®.
• This has been reported predominantly in patients with CKD receiving ESAs by subcutaneous administration.
• PRCA has also been reported in patients receiving ESAs for anemia related to hepatitis C treatment (an indication for which Aranesp® and EPOGEN® are not approved).
• If severe anemia and low reticulocyte count develop during treatment with Aranesp® or EPOGEN®, withhold Aranesp® or EPOGEN® and evaluate patients for neutralizing antibodies to erythropoietin.
• Permanently discontinue Aranesp ® or EPOGEN ® in patients who develop PRCA following treatment with Aranesp ® , EPOGEN ® , or other erythropoietin protein drugs. Do not switch patients to other ESAs.
• Serious allergic reactions, including anaphylactic reactions, angioedema, bronchospasm, skin rash, and urticaria may occur with Aranesp® or EPOGEN®. Immediately and permanently discontinue Aranesp® or EPOGEN® if a serious allergic reaction occurs.
• Blistering and skin exfoliation reactions including Erythema multiforme and Stevens-Johnson Syndrome (SJS)/Toxic Epidermal Necrolysis (TEN), have been reported in patients treated with ESAs (including Aranesp ® and EPOGEN ® ) in the postmarketing setting. Discontinue Aranesp ® or EPOGEN ® therapy immediately if a severe cutaneous reaction, such as SJS/TEN, is suspected.
• Serious and fatal reactions including “gasping syndrome” can occur in neonates and infants treated with benzyl alcohol-preserved drugs, including EPOGEN ® multiple-dose vials. There is a potential for similar risks to fetuses and infants exposed to benzyl alcohol in utero or in breast-fed milk, respectively.
• Adverse reactions (≥ 10%) in Aranesp® clinical studies in patients with CKD were hypertension, dyspnea, peripheral edema, cough, and procedural hypotension.
• Adverse reactions (≥ 5%) in EPOGEN® clinical studies in patients with CKD were hypertension, arthralgia, muscle spasm, pyrexia, dizziness, medical device malfunction, vascular occlusion, and upper respiratory tract infection.

• Privacy Statement
• Your Cookie Preferences
• Terms of Use

YOU ARE NOW LEAVING ANEMIA HUB™

Please know that Amgen, the sponsor of this site, is not responsible for the content on the site you are about to enter. Do you wish to proceed?

Internal You are now leaving AnemiaHub.com . Please know that the sponsors of this site are not responsible for content on the site you are about to enter. Please click the OK button below to continue.

This website is intended for U.S. healthcare professionals only.

What Is Anemia of Chronic Disease?

Anemia of chronic disease (ACD)vis a low red blood cell count (anemia) that occurs due to inflammation from an underlying health condition, such as cancer, kidney disease, or arthritis. Red blood cells are the cells in the blood that contain the protein hemoglobin, which is essential for delivering oxygen to the body. You can develop anemia if your hemoglobin levels are less than 12.0 grams per deciliter (g/dL) for females or less than 13.5 g/dL for males.

ACD is the second most common cause of anemia after iron deficiency . However, it rarely causes severe anemia unless ACD occurs with other types of anemia. The exact symptoms of ACD will depend on the severity of your condition, but most people experience some level of fatigue and shortness of breath. But treating the underlying condition and certain medications can help improve your symptoms.

Symptoms of ACD look similar to other types of anemia . But the symptoms you experience will depend on how severe your condition is and how low your blood cell levels are. In most cases, ACD causes mild to moderate anemia. But if you also have a pre-existing type of anemia (such as iron deficiency anemia) that occurs alongside ACD, your symptoms may become more severe.

With mild anemia, you might not experience any symptoms at all. But if your red blood cell count becomes lower, it becomes more difficult for your body to deliver oxygen to the rest of your body. In turn, this can cause more serious symptoms.

Common symptoms of ACD include:

• Trouble breathing
• Rapid heart rate
• Difficulty performing physical activities

In more severe cases of anemia, chest pain and fainting can occur because your body isn't able to deliver enough oxygen to organs like your heart and brain.

The spongy substances inside your bones (known as bone marrow) are where precursor cells mature and turn into red or white blood cells . As your body develops and produces red blood cells, it requires iron—an essential component of hemoglobin. Hemoglobin is an important protein in red blood cells that carries oxygen through your blood vessels. Some chronic conditions that cause inflammation can affect how your body stores and uses iron.

With inflammation , the body's immune system releases certain proteins, called cytokines. These proteins can interfere with the process of red blood cell production and prevent iron in your bone marrow from entering into red blood cells. Cytokines also lower the bone marrow's response to erythropoietin —a protein that stimulates red blood cell production.

Conditions That Cause ACD

There are many causes of inflammation in the body, but the following conditions can cause anemia of chronic disease and produce inflammation:

• Chronic infections such as HIV , tuberculosis, and hepatitis C
• Inflammatory and autoimmune conditions like arthritis , lupus, inflammatory bowel disease (such as Crohn's disease and ulcerative colitis ), and psoriasis
• Chronic kidney disease
• Cancers, which not only cause inflammation but may also invade the bone marrow and directly interfere with red blood cell production

If you're experiencing symptoms of anemia or have an underlying diagnosis of one of the conditions that can cause ACD, it's a good idea to see your healthcare provider for testing. Your provider will ask about your symptoms, take your medical history, and perform a physical exam.

After they learn more about your condition, they can order blood tests to help identify the type of anemia you're experiencing and confirm a diagnosis. These tests may include:

• Complete blood count ( CBC ): Checks your red blood cell and hemoglobin levels
• Blood smear: Takes a sample of your blood and views it under a microscope to look at the shape, size, and color of your blood cells
• Iron test: Looks at how much iron and ferritin (a protein that stores iron) you have in your body to help differentiate ACD from iron deficiency anemia

Sometimes, more invasive testing is required for diagnosis. A bone marrow biopsy is a procedure that occurs when your provider uses a needle to take a sample of tissue from the inner part of the bone. This provides information on your body's ability to make red blood cells and whether you have certain cancers.

Treatments for Anemia of Chronic Disease

The main goals of treatment of ACD are to treat the underlying cause of the problem, restore red blood cell levels, and improve your symptoms. Since ACD is caused by inflammation, it's essential to find and treat the source of your condition.

Your exact treatment plan will depend on the underlying cause of your ACD. For example, if an infection is the root cause of your ACD symptoms, you may need medications like antibiotics or antivirals. Other medical treatments such as corticosteroid medications and immune-modulating therapies can treat inflammatory conditions. An important part of the diagnostic process is to figure out what's causing your ACD so your provider knows how to treat it properly.

Other common treatments for ACD may include:

• Iron supplements
• Injectable medications called erythropoietin-stimulating agents which can help your kidneys if they are not making enough erythropoietin
• Blood transfusion to improve red blood cell counts if your levels drop to less than 7 grams per deciliter (g/dL)

How to Prevent ACD

It's not always possible to prevent ACD. But if you have an underlying condition, it's important to see your provider regularly for checkups. It's also important to make an appointment with your provider if you experience any symptoms of ACD.

However, you can follow these tips to help lower your risk of ACD and other chronic health conditions:

• Keeping your blood pressure and blood sugar levels in check
• Avoiding IV drug use which can increase your risk of infections
• Getting care for your underlying health conditions and following your treatment plan properly

If your healthcare provider has diagnosed you with any kind of anemia, it's important to pay attention to your symptoms and follow your treatment plan. Eating a diet rich in iron or taking iron supplements can also help prevent complications from anemia.

Complications

If your condition is left untreated or occurs alongside other anemias, ACD can cause complications like angina (chest pain due to low blood flow to the heart). ACD can also worsen underlying health conditions, like heart failure or coronary artery disease .

ACD during pregnancy can also cause certain problems. During pregnancy, the developing fetus requires its own hemoglobin. It's common for pregnant people to have lower blood counts due to the expansion of blood volume that happens during pregnancy , but significant anemia can lead to premature birth and growth problems for the fetus.

A Quick Review

ACD is a condition that causes low levels of red blood cells due to an underlying health condition that causes inflammation. These health conditions may include infections, cancer, arthritis, kidney disease, and lupus, among others. With ACD, it's common to experience symptoms like fatigue and shortness of breath. These symptoms are usually mild unless you have another type of anemia that occurs alongside ACD. But medications and other therapies can help restore blood counts and improve symptoms.

Frequently Asked Questions

When is anemia life-threatening?

Anemia can be life-threatening when blood counts are severely low, or when it occurs rapidly, such as in severe bleeding. Severely low blood counts impair the oxygen delivery to the body's tissues, including the brain and heart, which can cause fainting, abnormal heart rhythms, and death.

What hemoglobin level requires blood transfusion?

Blood transfusions may be necessary if your hemoglobin levels drop below 7 grams per deciliter. However, you may need a blood transfusion at higher blood counts if anemia is causing more severe symptoms or worsening symptoms of underlying conditions like heart failure.

Can anemia cause heart disease?

Anemia and heart conditions are related. Anemia and heart failure commonly occur together, each making the other condition worse. Longstanding and untreated anemia can cause a form of heart failure. Significant anemia can also worsen symptoms and outcomes in coronary artery disease. People with heart conditions should talk with their healthcare provider about treating anemia.

For more Health.com news, make sure to sign up for our newsletter!

Read the original article on Health.com .

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 13 March 2024

An exposome atlas of serum reveals the risk of chronic diseases in the Chinese population

• Lei You 1 , 2 , 3   na1 ,
• Jing Kou 4   na1 ,
• Mengdie Wang 1 , 3 , 5   na1 ,
• Guoqin Ji 1 , 3 , 6   na1 ,
• Xiang Li 4 ,
• Chang Su 7 ,
• Fujian Zheng 1 , 2 , 3 ,
• Mingye Zhang 4 ,
• Yuting Wang 1 , 2 , 3 ,
• Tiantian Chen 1 , 2 , 3 ,
• Ting Li 1 , 2 , 3 ,
• Lina Zhou 1 , 2 , 3 ,
• Xianzhe Shi   ORCID: orcid.org/0000-0001-9306-0130 1 , 2 , 3 ,
• Chunxia Zhao 1 , 2 , 3 ,
• Xinyu Liu   ORCID: orcid.org/0000-0002-0564-7167 1 , 2 , 3 ,
• Surong Mei 4 &
• Guowang Xu   ORCID: orcid.org/0000-0003-4298-3554 1 , 2 , 3  

Nature Communications volume  15 , Article number:  2268 ( 2024 ) Cite this article

15k Accesses

98 Altmetric

Metrics details

• Bioanalytical chemistry
• Environmental chemistry
• Risk factors

Although adverse environmental exposures are considered a major cause of chronic diseases, current studies provide limited information on real-world chemical exposures and related risks. For this study, we collected serum samples from 5696 healthy people and patients, including those with 12 chronic diseases, in China and completed serum biomonitoring including 267 chemicals via gas and liquid chromatography-tandem mass spectrometry. Seventy-four highly frequently detected exposures were used for exposure characterization and risk analysis. The results show that region is the most critical factor influencing human exposure levels, followed by age. Organochlorine pesticides and perfluoroalkyl substances are associated with multiple chronic diseases, and some of them exceed safe ranges. Multi-exposure models reveal significant risk effects of exposure on hyperlipidemia, metabolic syndrome and hyperuricemia. Overall, this study provides a comprehensive human serum exposome atlas and disease risk information, which can guide subsequent in-depth cause-and-effect studies between environmental exposures and human health.

Similar content being viewed by others

Serum biomarkers from cell-based assays for AhRL and MIS strongly predicted the future development of diabetes in a large community-based prospective study in Korea

Hong Kyu Lee, Wook Ha Park, … Youngmi Kim Pak

Gaussian graphical modeling of the serum exposome and metabolome reveals interactions between environmental chemicals and endogenous metabolites

Vincent Bessonneau, Roy R. Gerona, … Ruthann A. Rudel

Bayesian inference of chemical exposures from NHANES urine biomonitoring data

Zachary Stanfield, R. Woodrow Setzer, … John F. Wambaugh

Introduction

The “exposome” 1 , complementing the genome, offers a promising solution for characterizing environmental chemical factors and their nonnegligible influences on chronic diseases 2 . The episome encompasses life-course environmental factors, which are generally difficult to be measured 3 . Thus, a “top-down” strategy, the “blood exposome”, was proposed to directly reflect an individual’s internal chemical environment by allowing measurement of all the chemicals in blood 3 , 4 . Known as “human biomonitoring”, this method is currently being applied to monitor various chemicals in human blood, including organochlorine pesticides (OCPs) 5 , 6 , 7 , organophosphorus pesticides (OPPs) 8 , herbicides 9 , veterinary drugs 10 , perfluoroalkyl substances (PFASs) 11 , polycyclic aromatic hydrocarbons (PAHs) 7 , 12 , polychlorinated biphenyls (PCBs) 5 , 7 , and phthalates 13 . Although these targeted methods accurately monitor contaminants in human blood, each method can measure only a small number of chemicals with similar properties or structures. Indeed, the complexity of co-exposure to multiple categories in the real world highlights the limitations of current methods, especially for exploring the environmental causes of disease.

In addition, the distribution characteristics of environmental chemical levels and risks in different populations have been studied to provide guidance on prevention and control policies as well as protection for susceptible individuals. Previous studies have shown that epidemiological factors, such as age 6 , 7 , 14 , 15 , 16 , sex 6 , 17 , 18 , and education and income levels 6 , 15 , 19 , may influence the level of chemical residues in humans, and different distributions of serum chemical levels are associated with epidemiological factors. Furthermore, the same exposure may present a significantly greater risk effect in specific populations 20 , and sex-specific exposure-risk relationships are of particular interest 21 , 22 , 23 . Nevertheless, existing research has not yet elucidated the relationship between chemical exposure and population diversity, especially in large-scale general population cohorts.

Chronic diseases are indisputably the predominant challenge to health globally 24 . Both genetic and environmental factors contribute to multiple chronic diseases, with the latter being more important 25 . Currently, a high-profile issue is the associations between chemical exposures and disease outcomes. Environmental scientists and epidemiologists have widely studied these associations in multiple chronic diseases, such as hypertension 23 , 26 , diabetes 27 , 28 , hyperlipidemia 29 , 30 , 31 , and hyperuricemia 32 , 33 , among others. However, existing association studies are not sufficiently robust and are sometimes even contradictory, which is probably due to limited sample sizes and large individual differences in exposures. Moreover, the health effects of a single chemical have been the focus of most cohort studies, whereas the “cocktail effects” caused by exposure mixtures have largely been neglected, especially at real-world environmental concentrations.

In this work, we aimed to comprehensively characterize the human serum exposome, provide an exposure characteristic atlas resource, and define the exposure-disease risk relationship. To this end, we first enrolled a large-scale general population of 5696 people and collected serum samples, basic epidemiological information and chronic disease-related clinical parameters. Then, two targeted methods covering 267 environmental chemicals typically reported were established to quantify a portion of the serum exposome: one based on the gas chromatography‒tandem mass spectrometry (GC‒MS/MS) platform for quantification of 97 chemicals and another based on the liquid chromatography (LC)‒MS/MS platform for quantification of 170 chemicals. Next, we assessed levels and the risk of serum exposures in different people stratified by epidemiological information. Finally, single-exposure and multi-exposure models were used to define correlations between chronic diseases and the serum exposome and to reveal key risk factors.

To assess the risk of chronic disease from the serum exposome, (i) we designed a cohort of 5696 healthy and chronic disease patients from 15 provinces in China; the 12 chronic diseases included diabetes, hyperuricemia, obesity, hypercholesterolemia, hypertriglyceridemia, metabolic syndrome, high diastolic blood pressure, high systolic blood pressure, abdominal obesity, hypertension, high low-density lipoprotein cholesterol and hyperlipidemia. Additionally, we collected data on 9 basic epidemiological factors and 9 clinical parameters of chronic diseases and serum samples (Table  1 ). A portion of the human serum exposome, which included 267 chemicals, was comprehensively characterized via GC‒MS/MS and LC‒MS/MS; 74 chemicals were found at high frequencies and further studied as key targets of interest (Fig.  1a , Table  2 ). (ii) The participants were grouped using basic epidemiological information, after which residual levels of chemicals in serum were determined in the stratified population (Fig.  1b ). (iii) Robust associations between exposures and risk of chronic disease were established using single-exposure and multi-exposure models together to specify chemical residues at risk for chronic diseases (Fig.  1c ).

a Study design: Serum samples of 5696 participants were included in the study, 9 basic epidemiological information and 9 chronic disease-related clinical parameters were collected. Two platforms of LC-MS/MS and GC-MS/MS were used to detect 267 exposures, and 74 high-frequency exposures were determined. b Exposures characteristics: Exposures and basic epidemiological factors were associated, according to these factors, participants were stratified to analyze the distribution characteristics of exposures. c Risk discovery. First, single-exposure analysis showed the risk, stratified risk and health risk assessment of each exposure to chronic disease, and then exposure mixtures analysis showed risk effects of exposure mixtures on related chronic diseases based on 3 multi-exposure models. HbA1c glycated hemoglobin, BMI body mass index, LDL-C low density lipoprotein cholesterol, SBP systolic blood pressure, DBP diastolic blood pressure, OCP organochlorine pesticide, OPP organophosphorus pesticide, PAH polycyclic aromatic hydrocarbon, PCB polychlorinated biphenyl, PFAS perfluoroalkyl substance, WQS weighted quantile sum regression, q g-comp quantile g-computation, BKMR Bayesian kernel machine regression.

Determination of serum chemicals and batches

Human biomonitoring is prioritized for chemicals that possibly accumulate in the body and cause health effects based on the literature and public databases. A total of 97 and 170 chemicals, including OCPs, OPPs, herbicides, insecticides, fungicides, veterinary drugs, food additives, PAHs, PCBs, PFASs, and phthalates (Supplementary Fig.  1a ), were selected as priority lists and monitored by GC‒MS/MS and LC‒MS/MS; specific information is given in Supplementary Data  1 . For 74 exposures, 50% higher detection frequency in serum samples was found (Table  2 , Supplementary Fig.  1b ). The detection frequencies and concentration levels of all the exposures are given in Supplementary Data  2 .

In general, routine maintenance of instrumentation is necessary and performed regularly to ensure the quality of the instrument and the repeatability and stability of the data during long-term large-scale sample analysis, and batches were generated accordingly. The composition of each batch is shown in Supplementary Fig.  2 , which shows the running sequences of the calibration curves, quality control (QC) samples, and actual samples. A calibration curve was constructed at the beginning of each batch. The accuracy of each calibration curve was evaluated at low, medium, and high concentrations. Specifically, 2, 5, and 20 ng/mL spiked concentrations were used for accurate evaluation of 29 calibration curves using the GC‒MS/MS platform. Spiked concentrations of 1, 10, and 100 ng/mL were used for evaluating the accuracy of 20 calibration curves using the LC‒MS/MS platform. In total, 49 independent calibration curves were constructed during the whole sample analysis process for batch-specific quantification, and the results revealed good accuracy at low, medium, and high concentrations (Supplementary Fig.  3 , Supplementary Data  3 , 4 ). QC samples were used to evaluate batch effects and the stability and accuracy of the data. Batch effects were observed in the raw signal (Supplementary Fig.  4a, b ) but were greatly reduced after batch-specific calibration curve quantification and correction for multiple internal standards (Supplementary Fig.  4c–f ). Ninety-two percent of the 267 detected exposures and 84% of the 74 high-frequency exposures met the requirement of relative standard deviations (RSDs) less than 30% in the QC samples, indicating good stability of the data (Supplementary Fig.  5a, b ; Supplementary Data  5 , 6 ). In addition, 98% and 79% of the exposures monitored by the GC‒MS/MS and LC‒MS/MS platforms, respectively, met the requirement of accuracy between 80% and 120% in the QC samples, indicating good data accuracy (Supplementary Fig.  5c, d ). Finally, the samples were randomly analyzed using both platforms according to a given disease to achieve sample randomization (Supplementary Data  7 ).

Correlation between basic epidemiological factors and serum exposures

Considering that contact and accumulation of chemicals vary widely across populations, specific relationships between serum exposure and epidemiological factors, such as region, sex, age, other social factors and lifestyle, were examined in this study. First, the influence of each epidemiological factor on serum exposure was determined, samples were stratified according to each epidemiological factor (Supplementary Fig.  6 ), and serum exposures associated with epidemiological factors were identified. Region was the major factor influencing the level of exposure in human serum, which explained 15.6% of the variance in exposure (Fig.  2a ). The chemicals most influenced by regional factors were PFASs, especially perfluorocarboxylic acids, which included perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorotridecanoic acid, and perfluorodecanoic acid (PFDA) (Fig.  2b ). Sampling time was found to be the second most important factor. However, sampling time and region exhibited similar associations with serum exposure (Fig.  2b ), which was probably because the sampling was conducted region by region over a certain period (Supplementary Fig.  6e ). Furthermore, as the sampling time was mainly concentrated across only 3 months, the effect of sampling time on the serum exposome was considered attributable to regional factors without additional discussion. The next important factor causing variation in human serum exposure was age, which accounted for 1.4% of the variation in exposure. Factors other than region and age had relatively limited effects on exposure levels (Fig.  2a, b ). Principal component analysis revealed significant separation trends among samples from people of different regions and ages based on human chemical residues, but no such trends were observed for other factors (Fig.  2c, d ; Supplementary Fig.  7 ). Although human serum exposures were significantly associated with epidemiological factors, they correlated much less than chemicals in the same category. The strongest association was demonstrated among PFASs, followed by OCPs, PCBs and PAHs (Fig.  2e ; Supplementary Fig.  8 ). This suggests that chemicals in the same category are likely from similar sources of contamination.

a Explanation of exposure variations using epidemiological information based on variation partitioning analysis. b Correlations between each exposure and 9 basic epidemiological factors. The correlation coefficients obtained from the partial spearman correlation analysis were used to plot the heatmap. Principal component analysis score plot for 15 provinces ( c ) and different age ranges ( d ). e Correlation network of exposures and epidemiological factors. Red line represents positive correlation, blue line represents negative correlation obtained by Spearman correlation analysis. HCH hexachlorocyclohexane, DDD dichlorodiphenyldichloroethane, DDE dichlorodiphenyldichloroethylene, DDT dichlorodiphenyltrichloroethane, IBA indole-3-butyric acid, PFOA perfluorooctanoic acid, PFNA perfluorononanoic acid, PFDA perfluorodecanoic acid, PFUnDA perfluoroundecanoic acid, PFDoDA perfluorododecanoic acid, PFTrDA perfluorotridecanoic acid, PFOS perfluorooctanesulfonate, PFHxS perfluorohexanesulfonate, 6:2 Cl-PFAES 6:2 chlorinated polyfluoroalkyl ether sulfonate, 6:2 diPAP bis[2-(perfluorohexyl)ethyl] phosphate, PFPeA perfluoro-n-pentanoic acid, PFHpS perfluoroheptanesulfonic acid, MCHP monocyclohexyl phthalate, MEP monoethyl phthalate.

Distribution characteristics of human serum exposures based on region, age and other factors

Considering that region was the most important factor influencing human exposure levels, the population was stratified according to region to explore the distribution characteristics of exposures. Overall, the highest concentrations of total human serum chemicals were found in Shanghai, followed by Zhejiang, Jiangsu and Shandong, which are located in eastern coastal areas and have dense populations and well-developed industries. The lowest concentrations of chemicals were found in Shaanxi and Guizhou, which are inland areas with lower populations and fewer industries (Fig.  3a ). Among the various types of human serum exposure, the concentration levels of drugs were highest, followed by PFASs, which varied greatly among provinces (Fig.  3b ). People in the Yangtze River Delta had higher serum PFASs, especially in Shanghai and Jiangsu (Supplementary Fig.  9a ). Exposure levels of PAHs, PCBs and OCPs also differed greatly among the populations in different regions. For example, serum PAH concentrations in people from Henan, Zhejiang, and Shandong were significantly greater than those in people from other provinces (Supplementary Fig.  9b ). Serum levels of PCBs and OCPs were significantly greater in Chongqing, Shanghai, and Jiangsu (Supplementary Fig.  9c, d ). For OPPs, the highest exposure levels were found in the population of Chongqing (Supplementary Fig.  9e ). Veterinary drugs and food additives had little difference in populations across provinces (Supplementary Fig.  9f ).

a The location of province included in this study and their total concentration of exposures. b Regional distribution of different categories of exposures depicted by stacked bar plot. Exposures that significantly increase ( c , d ) and decrease ( e ) with age. Exposures that significantly increase ( f – h ) with education. For figures ( c – h ), the concentrations scaled by the z-score method were used and error bars represent standard error of mean ( n  = 5696 biologically independent samples). Exposures that significantly change with gender ( i ) and drinking history in male ( j ). The geometric means of the exposures were used for figures ( a , b , i , j ). Gray dash lines of figures ( i , j ) represent fold change less than 0.8 and more than 1.3.

To better understand exposure levels in human serum on a larger regional scale, the exposure information of some high-frequency chemicals was compared with that of seven other countries based on the literature. In comparison with other regions of the world, serum from people who live in China has the highest residue levels of p,p’-dichlorodiphenyldichloroethylene (DDE), beta-hexachlorocyclohexane (HCH), p,p’-dichlorodiphenyltrichloroethane (DDT), PFOA, perfluoroundecanoic acid and perfluoro-n-pentanoic acid (PFHpS). The highest residue levels of hexachlorobenzene (HCB), perfluorooctanesulfonate (PFOS) and PFDA were found in the serum of people in Korea. The highest residue levels of PFNA and perfluorohexanesulfonate (PFHxS) were found in the serum of people in the United States and Canada, respectively (Supplementary Fig.  10 , Supplementary Data  8 ).

Age is an important factor that explains differences in human exposome and may be related to differences in the exposure times and metabolic levels of people of different age ranges. Therefore, exposure levels were analyzed for each age group (Fig.  3c–e ). Accumulation of the majority of chemicals, such as OCPs and PFASs, in human serum increased with age (Fig.  3c, d ). Among them, beta-HCH, p,p’-DDE, pyrene, and indole-3-butyric acid (IBA) increased with age, and the highest concentrations were found in people older than 70 years. The PFAS concentration increased with age until the age of 50 years (Fig.  3d ). In contrast, the concentrations of some chemical agents decreased with age, with the serum of children aged less than 10 years having the highest residual levels of cyclamic acid and acesulfame (Fig.  3e ), which prompted us to investigate the intake of sugar substitutes in these children.

Other factors also have an impact on serum exposure. Stratification of the population according to education and income levels revealed that most chemicals tended to increase as education and income levels increased. Specifically, these included PAHs, OCPs, and PCBs (Fig.  3f–h , Supplementary Fig.  11a ). Only IBA residues showed a decreasing trend (Supplementary Fig.  11b, c ). In addition to education and income, sex is an important factor influencing accumulation and excretion of chemicals in the human body. Significantly higher levels of OCPs were found in females than in males, whereas significantly lower levels of PFASs and phthalates were found in females (Fig.  3i ). The chemicals most influenced by sex were beta-HCH, monocyclohexyl phthalate (MCHP) and PFHxS (Supplementary Fig.  11d–f ). Finally, with regard to smoking and alcohol consumption, only male individuals were considered because women in this study rarely smoked or drank alcohol (Supplementary Fig.  6d, f ). For many chemicals, significantly greater concentrations were detected in drinkers than in nondrinkers (Fig.  3j ); the chemicals most influenced by drinking were cyclamic acid and six PFASs, with higher levels of the latter in drinkers (Supplementary Fig.  11g–m ). Thus, the risk of exposure associated with alcohol consumption should be considered. Unlike drinking, exposure to various chemicals did not significantly differ between smokers and nonsmokers (Supplementary Fig.  11n ).

Risk analysis of chronic diseases associated with a single exposure

Whether exposure is associated with the risk of chronic disease and which chemicals are key risk factors are two questions of wide interest. In this study, 9 clinical parameters of chronic disease were subdivided into 12 related chronic disease outcomes, and the health risk of exposure was analyzed for each outcome (Supplementary Fig.  12a, b ). Among the 12 chronic diseases, three were determined by multiple clinical parameters: hyperlipidemia, metabolic syndrome, and hypertension (Supplementary Fig.  12c–e ). The remaining 9 chronic diseases were determined by single clinical parameters. Well-matched controls were selected for each disease outcome (Supplementary Data  9 , 10 ). The specific analysis scheme is shown in Supplementary Fig.  13a . For 12 disease outcomes, multiple exposure-disease risk associations were observed. Specifically, OCPs and PCBs were associated with an increased risk of hypertension, diabetes, metabolic syndrome and obesity. PFASs were associated with an increased risk of hyperlipidemia, metabolic syndrome, diabetes, and hyperuricemia (Fig.  4 , Supplementary Fig.  13b ). The disease associated with the greatest number of chemicals was hyperlipidemia, as was the related disease (Fig.  4a–d ), followed by metabolic syndrome (Fig.  4e ), demonstrating that these diseases are most affected by environmental chemicals. Most chemicals have positive effects on the risk of developing chronic diseases, except for hypertension and abdominal obesity. In particular, all chemicals had significant effects on the risk of hyperuricemia (Fig.  4h ). Among all chronic disease outcomes, hypertension, and related diseases had the weakest risk relationship with chemicals (Fig.  4j–l ). These results were still similar after additional adjustment for confounders of external environmental factors, including air pollution and meteorological conditions (Supplementary Fig.  14 ). Associations between 193 low-frequency exposures and chronic diseases were also investigated. Most chemicals were associated with an increased risk of chronic diseases, particularly metabolic syndrome, obesity, diabetes, and hyperuricemia (Supplementary Figs.  15 , 16 ). However, compared to the results of high-frequency exposures, the significance of associations between low-frequency chemicals and chronic diseases was weaker, with larger confidence intervals, which highlights the importance of cautious interpretation of these associations. Therefore, in subsequent analyses, we focused solely on association analysis of highly frequently detected chemicals. Similar exposure risk associations were found in analysis of nine continuous clinical parameters, complementing the results of the classified outcomes described above. Thirteen chemicals had risk effects on five or more disease outcomes according to both regression models (Fig.  5a ; Supplementary Fig.  13b ). The predominant adverse effects were mainly OCPs, PFASs, and PCBs, demonstrating the nonspecific risk of these chemicals for multiple chronic diseases (Supplementary Fig.  17 ).

Exposures with a significant risk for hyperlipidemia ( a ), hyper low density lipoprotein cholesterol ( b ), hypercholesterolemia ( c ), hypertriglyceridemia ( d ), metabolic syndrome( e ), obesity ( f ), diabetes ( g ), hyperuricemia ( h ), abdominal obesity ( i ), hypertension ( j ), hyper diastolic blood pressure ( k ) and hyper systolic blood pressure ( l ). The concentrations were log10 transformed so ORs represent odds ratios per one-unit increase in log-transformed exposure levels. Binary logistic regression models adjusted for age, gender, region, sampling time, education and income levels, marital status, smoking and drinking history. The position and color of diamond represent ORs and significant (two-sided, n  = 5696 biologically independent samples), respectively. Error bars represent 95% confidence interval of ORs.

a Relationship between each exposure and 9 clinical disease parameters. Included exposures were significantly associated with at least one disease outcome. Risk of exposures for hyperlipidemia and metabolic syndrome stratified by age ( b ) and gender ( c ). Scaled odds ratios were used in the heat maps of stratified risk, and * represent significant 0.01 <  p  < 0.05, ** represent significant 0.001 <  p  < 0.01, *** represent significant p  < 0.001 (two-sided). Exposures with significant associations for the corresponding diseases were used for this plot based on multiple linear regression and binary logistic regression models. All of regression models adjusted for age, gender, region, sampling time, education and income levels, marital status, smoking and drinking history.

Furthermore, susceptibility to chronic diseases was investigated in populations of different ages and sexes, which can help to identify and protect susceptible patient subgroups. Hyperlipidemia and metabolic syndrome were chosen as target diseases because of the sufficient sample size and the exposure-disease associations described above. Susceptibility to exposure-induced chronic diseases was explored in three age ranges, 30-50, 50–60, and 60–80 years, representing young-aged, middle-aged, and elderly groups, respectively. Compared to those in the middle-aged group, the elderly group presented a greater risk of exposure to hyperlipidemia; the risk factors were mainly PFASs and OCPs, which also had a stronger exposure risk for metabolic syndrome in elderly group (Fig.  5b ). Given that many disease risk factors are sex-specific, we also analyzed the sex association between exposure and disease risk; no significant sex differences were found in terms of exposure risk to hyperlipidemia, but almost all men were at increased risk for metabolic syndrome (Fig.  5c ). Specific differences in odds ratios (ORs) between men and women are shown in Supplementary Fig.  18 .

Finally, health risk assessments of all monitored individuals were performed using reported exposure guidance values. Available reference dose (RfD) values of 11 PFASs were found, and available exposure guidance values for blood (biomonitoring equivalent (BE), Human Biomonitoring II (HBM II) or Biomonitoring Guidance Values (BGV)) of only 8 chemicals were found (see Supplementary Data  11 ). The health risks associated with the population exposure levels in our study were evaluated based on hazard quotients (HQs) (Supplementary Fig.  19 ). Most exposure levels of PFASs were within the safe range compared to those of RfD, except for PFOA, PFOS, PFNA, and PFUnDA, which approached or exceeded risk concentrations in a few individuals (Supplementary Fig.  19a ). Individuals had excessive exposure reference values (BE, HBM II and BGV) for HCB, PFOA, and PFOS, suggesting potential health risks (Supplementary Fig.  19b ). For hyperlipidemia risk, dose‒risk relationship curves revealed 16 related chemical exposure guidance values. The PFOA and PFOS reference concentrations are 3.19 and 5.26 ng/mL, respectively, lower than the European reference concentrations of 10 ng/mL and 20 ng/mL, respectively, for HBM II. In addition, 5 chemical exposure reference values were not reported, including beta-HCH, p,p’-DDE, PFHpS, 6:2 Cl-PFAES, and MEP (Supplementary Fig.  19c ).

Risk effects of exposure mixtures on related chronic diseases

Exposure to chemical mixtures is a real-life scenario of human exposure but has received insufficient attention in recent studies. In our study, associations of exposure mixtures with each disease outcome were explored by three multi-exposure models, including weighted quantile sum regression (WQS) 34 , quantile g calculation (q g-comp) 35 , and Bayesian kernel machine regression (BKMR) 36 , 37 . The results showed that high-frequency exposure mixtures had adverse effects on hyperuricemia, hyperlipidemia and metabolic syndrome (Fig.  6 ; Supplementary Data  12 ). The risk chemicals that contributed more to the model were obtained by weights and posterior inclusion probability (PIP). For hyperuricemia, 14, 10 and 13 preferred risk chemicals were obtained by the above three models (Supplementary Fig.  20a–d ), respectively, with 13 chemicals overlapping in at least two models (Supplementary Fig.  20e ). For hyperlipidemia, 15, 7, and 10 important risk chemicals were obtained by the three models, respectively (Supplementary Fig.  20f–i ), with an overlap of 10 chemicals (Supplementary Fig.  20j ). For metabolic syndrome, 11, 7, and 6 significant risk chemicals were obtained, respectively (Supplementary Fig.  20k–n ), with an overlap of 5 chemicals (Supplementary Fig.  20o ). The overall risks of high-frequency exposure mixtures obtained from the three models are shown in Fig.  6a–c .

For three chronic disease outcomes including hyperuricemia ( a ), hyperlipidemia ( b ), and metabolic syndrome ( c ), all exposed mixtures have a positive risk effect on them based on WQS, q g-comp, and BKMR Models. d – f Odds ratios (ORs) of the joint and each of the priority risk chemicals screened by three multi-exposure models (The chemicals given in figure were defined in at least two models). ORs of the jointed chemicals were obtained by WQS and q g-comp models, and OR of each chemical was obtained by binary logistic regression model. ORs represent odds ratios per one-unit increase in log-transformed exposure mixtures or single exposure levels. All of the three multi-exposure models adjusted for age, gender, region, smoking, and drinking history, and the binary logistic regression model adjusted for age, gender, region, sampling time, education and income levels, marital status, smoking, and drinking history. The position and color of diamond represent ORs and significant (two-sided), respectively. Gray diamonds represent no significance. Error bars represent 95% confidence interval. g Dose-risk relationship of exposures to hyperuricemia, specifically, seven key exposures screened by both single and mixed models. h Dose-risk relationship of exposures to hyperlipidemia, nine key exposures were included. i Dose-risk relationship of exposures to metabolic syndrome, three key exposures were included. The black solid line represents the OR, and gray, blue and dark green shadow represents the 95 % confidence interval of hyperuricemia ( n  = 927 biologically independent samples), hyperlipidemia ( n  = 2842) and metabolic syndrome ( n  = 1284) ORs, respectively.

The chemicals overlapping in the above 3 multi-exposure models were considered as priority risk chemicals and included mainly OCPs (beta-HCH, p,p’-DDT, p,p’-DDE, HCB, pentachlorophenol), PFASs (PFPeA, PFOA, PFDA, PFHxS, PFHpS, PFOS), phthalates (MCHP, monocyclohexyl phthalate (MEP)) and other pesticides (IBA, fipronil sulfone, chlorpyrifos, triclosan, etofenprox) (Fig.  6d–i ; Supplementary Fig.  21 ). Mixtures of these priority risk chemicals were reincorporated into the WQS and q g-comp models to obtain ORs, which were found to be significantly greater for mixture exposure than for single chemicals. This indicated that the mixtures had a nonnegligible risk-enhancing effect. There are several priority risk chemicals with no significant risk according to the single-exposure model (Fig.  6d–f ; Supplementary Fig.  21 ). This suggests that the multi-exposure model is more sensitive in identifying risk chemicals than the other models and can be used as a complement to single-exposure risk analysis.

To better understand the dose‒risk relationship, overlapping priority risk chemicals were analyzed, and they presented monotonically increasing nonlinear associations with the risk of chronic disease (Fig.  6g–i ). Specifically, 7 of the 13 overlapping priority risk chemicals for hyperuricemia had significant risk effects according to the single model: IBA, MCHP, MEP, PFOA, PFNA, PFHxS and PFHpS. All exposures showed a rapid increase in risk with increasing concentration, except for PFHxS, which showed a faster increase in risk only at low concentrations (Fig.  6g ). Nine of the 10 overlapping priority risk chemicals for hyperlipidemia had significant risk effects according to a single model: beta-HCH, p,p’-DDT, p,p’-DDE, HCB, fipronil sulfone, PCB 138, MEP, PFOS, and PFHxS. Among them, fipronil sulfone, PCB 138 and PFHxS showed a faster increase in risk only at low concentrations, and the others increased more rapidly at higher concentrations (Fig.  6h ). Three of the 5 overlapping priority risk chemicals for metabolic syndrome had significant risk effects according to a single model: beta-HCH, p,p’-DDT, and MCHP. All exhibited a faster increase in risk at high concentrations (Fig.  6i ). The risk relationships of chemicals identified only by multi-exposure models were almost nonmonotonic (Supplementary Fig.  21 ), which may be the reason why they cannot be identified by traditional single-exposure models, such as the linear and logistic regression models used in this study.

In this work, we studied the concentration, distribution and disease risk associated with serum exposure in 5696 control individuals and patients via human biomonitoring. Epidemiological factors such as region, age, and sex significantly influence human exposure levels. For exposure risk, hyperlipidemia, metabolic syndrome and hyperuricemia were associated with multiple exposures. In addition, exposure to OCPs, PFASs, PCBs, and phthalates showed nonspecific associations with risk and nonlinear dose-dependent relationships with chronic diseases. To our knowledge, this study is the largest study to date to examine the association between human serum exposure and health outcomes in terms of both chemical coverage and population sample size (Supplementary Data  13 ). This study not only provides the most representative biomonitoring data for the issue of exposure in the Chinese population but also provides a broad understanding of the chemicals present in humans and their risk of chronic diseases.

Region was the most influential factor on exposure levels, especially for OCPs, PFASs, PCBs, and PAHs, which exhibited significant differences in population levels across 15 provinces in China (Fig.  3a, b ; Supplementary Fig.  9 ). Among them, PFASs, PCBs, and PAHs are mostly produced through industrial use, including anti-staining materials, electronic waste, and incomplete fuel combustion; OCPs are typical pesticides. Above all, the residue levels of these exposures in humans are closely related to industrial and agricultural development, and exposure levels are greater in industrially developed coastal areas than in inland areas. This finding is consistent with previous findings 38 . This may be the result of the rapid industrialization and urbanization in coastal areas 39 . Notably, the highest accumulation levels of almost all chemical categories were found in the population in the Yangtze River Delta. The most likely explanation is that persistent organic pollutants (POPs) cannot be effectively removed from wastewater 40 and are transferred to downstream areas through water flows 41 . In addition, the potential health risk through the food chain is of concern, as studies have shown that the concentrations of PCBs in most animal-derived food groups in coastal areas are significantly greater than those in inland areas 39 . Therefore, management of pollution sources should be strengthened; otherwise, serious contamination in downstream areas and further health hazards may occur through food chain accumulation in the human body. The results from the cohort in this study provide the most representative data for chemical biomonitoring in the Chinese population to date. The cohort in this study included people from 15 provinces, including southern and northern regions as well as coastal and inland areas across the country, representing 56% of the Chinese population. The results showed that the level of OCPs in human blood is significantly greater in China than in other countries, which is closely related to the national conditions of agricultural development in China. Most PFASs show the highest serum concentrations in China and Korea, with a few species of PFASs having the highest concentrations in human serum in the United States and Canada. This is closely related to the geographical shift of industrial sources from North America and Europe to emerging Asian economies, especially China, since 2002 42 .

Age was the second most important factor affecting serum chemical levels. The results showed that age and multiple serum chemical concentrations correlated positively, which was consistent with the findings of others, specifically for OCPs 6 , 7 , 15 , 16 , 43 , PCBs 7 , 15 , 16 , and PFASs 14 , 17 , 18 . The age-dependent association of these POPs may be caused by their excretion rates and exposure histories in different age ranges of the population 6 . Higher levels of exposure to multiple chemicals were found in populations with higher education levels and incomes, which is consistent with the findings of other studies in specific categories, including PCBs 15 , OCPs 6 , PFASs 19 , 44 and phthalates 19 . These findings suggest that social factors may have an impact on exposure to POPs 6 . Moreover, sex was found to be an important factor influencing accumulation and excretion of exposures, and residue levels of more fat-soluble OCPs were greater in women than in men, which may be attributed to the generally high body fat percentage in women 6 , 45 . On the other hand, concentrations of nonfat-soluble chemicals such as phthalates and PFASs were significantly greater in men than in women, which is largely attributed to the unique menstrual excretions of females 46 . Finally, higher serum levels of PFASs were detected in alcohol drinkers, which has not been reported in previous studies; future studies should continuously focus on the relationship between drinking and multiple environmental chemicals.

In the exposure and health effects study, associations of hyperlipidemia, metabolic syndrome and hyperuricemia with various exposures were found (Figs.  4 – 6 ). For hyperlipidemia, the present study showed associations with OCPs, PCBs, PFASs, and phthalates. Our findings validate several previous studies; for example, increased concentrations of PFOA, PFNA 31 , PFHxS 47 , OCPs and PCBs are associated with elevated total serum lipids, total cholesterol and triglycerides 29 , 48 . Nevertheless, some new risk effects of exposure were observed in this study. For example, among PFASs, new risk associations were found, including PFHpS, PFOS and 6:2 chlorinated polyfluoroalkyl ether sulfonate for hyperlipidemia. In particular, 6:2 chlorinated polyfluoroalkyl ether sulfonate is a new type of PFAS with few related disease risk association studies. Moreover, few existing studies have focused on the finding that fipronil sulfone significantly increases the risk of hyperlipidemia, possibly because of insufficient biomonitoring coverage, and our study extends previous findings.

Metabolic syndrome is another chronic disease strongly associated with exposure and is a pathological state involving multiple metabolic diseases, including abdominal obesity, dyslipidemia, hypertension, and diabetes 49 . Previous studies have shown an increased risk of metabolic syndrome with various POPs, such as OCPs, PCBs 50 , and PFAs 31 , which was consistent with our findings. In addition, MEP was shown to be associated with an increased risk of hypertriglyceridemia, a disease subtype of metabolic syndrome, in a previous study of exposure to endocrine-disrupting chemicals affecting the risk of metabolic syndrome in adults 51 . An association between MCHP and the risk of metabolic syndrome was also found in our study. Finally, differences in disease susceptibility among different populations were observed in this study but rarely studied in previous researches. There was a stronger effect of exposure on metabolic syndrome in the elderly and in men (Fig.  5b, c ). This may be due to differences in excretion and metabolism among different populations. Previous studies have reported that pesticides are excreted more slowly in elderly individuals and in men 6 , 46 . Moreover, sex-specific associations between exposure and disease-associated lipid changes may explain sex-related differences at the metabolic level 22 .

Hyperuricemia, another chronic disease strongly associated with exposure in this study, is a causative agent for a variety of diseases, including gout, kidney stones, and cardiovascular disease 52 , 53 . The serum uric acid concentration is strongly influenced by environmental factors, and associations between PFASs and hyperuricemia have been reported 32 , 33 , 54 . These findings were again verified in the present study, and a new association of PFHpS with hyperuricemia was found. The effects of other categories of environmental chemicals on uric acid are unknown; significant associations between IBA, MEP, and MCHP and hyperuricemia were found based on both single and mixed exposure models (Fig.  6d, g ), which have not been reported in other studies.

Multi-exposure models were jointly used to collectively identify combinations of exposures that are associated with significant risk effects for multiple chronic disease outcomes. Some key risk chemicals screened by multi-exposure models were not identified in the single-exposure model (Fig.  6d–f ), reflecting the complementary role of the multi-exposure model to the single-exposure model 55 . In addition, three chronic diseases that are most affected by overall exposure and corresponding exposure mixtures, were identified. Three groups of exposure mixtures showed significant risk-enhancing effects on hyperuricemia, hyperlipidemia, and metabolic syndrome, which have not been reported in previous studies. Future studies should pay attention to these chronic diseases and associated exposure mixtures.

Taken together, our results provide comprehensive insights into the residue levels and exposure characteristics of environmental chemicals in human serum, identifying human serum exposures and their specific combinations that are associated with major chronic disease outcomes. These findings provide guidance for further pollution management and protection of susceptible populations. We recognize that our results do not indicate the causal effect of any chemical on adverse outcomes, which requires further investigation. Nevertheless, this study not only demonstrates the potential of the exposome for disease prediction but also provides a useful resource for more in-depth toxicological and longitudinal epidemiological studies.

Study population and epidemiological information

A cross-sectional study comprising 5696 subjects, including 2141 healthy persons and 3555 patients, was conducted in 15 provinces of China. The study was approved by the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention (reference No. 201524); the ethics committee was the Chinese Center for Disease Control and Prevention Institutional Review Board. Moreover, written informed consent was obtained from each participant before the study began. A total of 12 chronic disease outcomes were analyzed, ranging from 243 to 1813; these included diabetes, hyperuricemia, obesity, hypercholesterolemia, hypertriglyceridemia, metabolic syndrome, high diastolic blood pressure, high systolic blood pressure, abdominal obesity, hypertension, high low-density lipoprotein cholesterol and hyperlipidemia (Supplementary Fig.  12b ). The study population in this work is a subset of the cohort recruited from the China Nutrition and Health Survey in 2015. The samples covered different regions, sexes, age groups, and various chronic diseases. The details of the China Nutrition and Health Survey were described in previous studies 56 , 57 . In brief, the cohort was selected using a multistage random cluster design and included 360 communities from 15 provinces with varying income levels. Then, 20 families were randomly selected from each community to participate in the study. Well-trained researchers collected epidemiological data and blood samples, and clinical parameters related to chronic diseases were determined through biochemical assays of the blood samples.

Epidemiological factor information, including sex, age, sampling location, sampling time, education level, income level, marital status, cigarette smoking status, alcohol consumption status and medical history, was collected via questionnaire. Environmental factor information, including air pollution and meteorological conditions, was obtained based on regional and sampling time (month) information from open-source websites. Moreover, 9 chronic disease-related clinical parameters, namely, uric acid levels, glycated hemoglobin levels, low-density lipoprotein cholesterol levels, triglyceride levels, total cholesterol levels, systolic blood pressure, diastolic blood pressure, waist circumference and BMI, were determined. The detailed information is provided in Table  1 . Twelve chronic disease outcomes were subdivided based on these nine related clinical parameters. A total of ten rounds of survey were conducted on the cohort from 1989 to 2015. For the 2015 survey, 14,000 samples were collected, and we randomly selected half of these samples for exposome analysis. After excluding samples with insufficient volume or incomplete information, a total of 5696 samples were ultimately analyzed in this study.

The specific criteria for classification of 12 chronic disease outcomes by 9 clinical parameters

Individuals with uric acid levels equal to or greater than 420 μmol/L for males, and equal to or greater than 360 μmol/L for females are determined to have hyperuricemia.

Individuals with glycated hemoglobin levels equal to or greater than 6.5% are determined to have diabetes.

Individuals with low-density lipoprotein cholesterol levels equal to or greater than 3.4 mmol/L are determined to have hyperLDL-C.

Individuals with triglyceride concentrations equal to or greater than 2.3 mmol/L are determined to have hypertriglyceridemia.

Individuals with total cholesterol levels equal to or greater than 6.2 mmol/L are determined to have hypercholesterolemia.

Individuals with systolic blood pressure equal to or greater than 140 mmHg are determined to have high systolic blood pressure.

Individuals with diastolic blood pressure equal to or greater than 90 mmHg are determined to have high diastolic blood pressure.

Individuals with waist circumference equal to or greater than 90 cm for males, and equal to or greater than 85 cm for females are determined to have abdominal obesity.

Individuals with a BMI (Body Mass Index) greater than 28 kg/m2 are determined to be obese.

Hypertension is determined if either the systolic blood pressure is equal to or greater than 140 mmHg, or the diastolic blood pressure is equal to or greater than 90 mmHg (refer to Supplementary Fig. 12e ).

Hyperlipidemia is determined if any of the three clinical indicators are met: low-density lipoprotein cholesterol equal to or greater than 3.4 mmol/L, triglyceride concentration equal to or greater than 2.3 mmol/L, or total cholesterol equal to or greater than 6.2 mmol/L (refer to Supplementary Fig. 12c ).

Metabolic syndrome is determined if an individual has three or more of the following four chronic diseases: abdominal obesity, hypertriglyceridemia, hypertension, and diabetes (refer to Supplementary Fig. 12d ).

Environmental variable estimates

Environmental factors including air pollution and meteorological conditions were obtained using regional and sampling time (month) information from open-source websites. Specifically, for air pollution, data on three variables including Air Quality Index (AQI), PM2.5, and PM10 were collected from a recent study 58 with open data on website https://quotsoft.net/air/ . Daily data for each region (specific to city/county) were available, and based on these data, monthly values of AQI, PM2.5, and PM10 were calculated for each individual in their respective region (specific to city/county) and sampling month.

For meteorological conditions, data on three temperature variables including daily maximum temperature, daily minimum temperature, and daily mean temperature were collected from a surface climate dataset of China in the website 59 https://www.geodoi.ac.cn/WebCn/doi.aspx?Id=3187 . Daily data for each region (specific to province) were accessible, and based on these data, monthly mean values of Daily_Maximum_Temp, Daily_Mean_Temp, and Minimum_Temp were calculated for each individual in their respective region (specific to province) and sampling month.

Finally, the obtained six environmental variables and nine epidemiological factors were treated as confounding factors and adjusted by binary logistic regression model.

Chemicals and reagents

Ultrapure water was prepared with a Milli-Q water purification system (Millipore, 7 Billerica, MA, USA). HPLC grade acetonitrile and methanol were obtained from Merck (Darmstadt, Germany). Ammonium acetate and fetal bovine serum were purchased from Thermo Fisher Scientific (Rockford, USA). Formic acid was purchased from National Medicines Corporation Ltd. (Beijing, China). Dichloromethane was purchased from J.T. Baker (Rockford, USA). Hexane was purchased from Merck Sigma‒Aldrich (Darmstadt, Germany). The 267 chemical standards for each analyte were acquired from Alta Scientific Co., Ltd. (Tianjin, China) (Supplementary Data  1 ). Twenty-seven internal standards were isotope-labeled chemical standards, which were purchased from several companies; the detailed information is provided in Supplementary Data  14 .

Selection of 267 target chemicals

The principles of prioritized list selection and identification were described in our previous study 60 . Briefly, at least one of the following three conditions should be included. First, chemicals, including pesticides, veterinary drugs and POPs, are often reported in the literatures with high concentration levels and high detection frequency in blood (plasma/serum). Second, according to the Integrated Risk Information System and the International Agency for Research on Cancer, these exposures can have health effects or carcinogenicity. Third, considering their dietary exposure risks, chemicals are often found to exceed residue limits in daily food in routine assays by the authority agency. Based on the above principles, the detection scope was expanded in our study, and two platforms (GC‒MS/MS and LC‒MS/MS) were used to cover prior exposures comprehensively. After grouping by region, chemicals with a detection frequency greater than 50% in any one of 15 provinces was defined as having high-frequency exposure and were the focus of subsequent analysis.

Method development and analysis based on GC‒MS/MS

The method used for the GC‒MS/MS assay reported in a previous study 61 was modified, and the target analytes were expanded from 35 to 97. Before GC‒MS/MS analysis, solid-phase extraction (SPE) was used for pretreatment of serum samples. The specific pretreatment steps were as follows: (a) Adding internal standard: 10 μL isotope standard (Supplementary Data  14 ) mixture was added to 200 μL serum samples and stored at 4 °C overnight for later use. (b) Deproteinization: 200 μL of 15% formic acid aqueous solution was added to the above serum sample, after which the sample was vortexed. (c) Activation of SPE cartridges: before adding the serum sample, 3 mL of dichloromethane, 3 mL of methanol, and 3 mL of ultrapure water were added to Oasis® PRiME HLB cartridges in advance. (d) Adding samples: all the above pre-treated serum samples were transferred to the Oasis® PRiME HLB cartridges. (e) Rinsing and vacuuming: the sample tube was rinsed twice with 1 mL of methanol:water (1:6, v/v) at a flow rate of 0.5–1 mL/min, and the SPE cartridges were rinsed with 1 mL of methanol:water (1:6, v/v) and evacuated for 20 min. (f) Elution: the samples were eluted with 3 mL of dichloromethane and 3 mL of n-hexane, with all the eluent collected. (g) Nitrogen blowing and redissolving: the eluent was nitrogen-blown nearly dry, redissolved in 100 μL of acetone, and transferred into a sample bottle for later analysis.

Parameter settings of target method based on GC-MS/MS

The experiments were performed using an 8890 GC system equipped with a DB-5MS column (30 m × 0.25 mm × 0.25 μm; Agilent, Santa Clara, USA) coupled to a Agilent 7000D triple quadrupole mass spectrometer. For GC system, high purity helium (99.999%) was used as carrier gas with the flow rate of 1.2 mL/min. The injection volume was 1 μL, and was conducted in splitless mode at 270 °C. The analytes were separated by temperature programming: the initial temperature increased from 70 °C to 150 °C at the speed of 25 °C/min, then increased to 200 °C at 3 °C/min and kept for 2 min. Finally, the temperature was increased to 300 °C at 8 °C/min and kept for 6 min. For MS/MS condition: the MS was equipped with electron bombardment ion source with the ionization voltage of 70 eV. The temperature of ion source was 300°C, the temperature of transmission line was 300 °C, and the detector gain was 0.90 kV + 0.30 kV. Quantitative analysis was performed by using multi-reaction monitoring mode with the solvent delay of 6 min. The detailed mass spectrum parameters are presented in Supplementary Data  15 .

Method development and analysis based on LC‒MS/MS

The serum sample pretreatment method was modified from a previously reported method 60 , and the target analytes were expanded from 106 to 170. In brief, a high-throughput sample processing method was used to extract samples from serum. Deproteinization and purification of serum were conducted using a phospholipid removal plate (Phenomenex, Torrance, USA) with 96 wells. A total of 280 μL of acetonitrile containing 19 internal standards (Supplementary Data  14 ) was added to each well and mixed with 70 μL of the serum sample. The 96-well filter plates were covered with aluminum foil and shaken for 10 min at room temperature. Proteins and phospholipids were removed after centrifugation at 1000 × g for 10 min at 4 °C. The supernatant was concentrated with nitrogen flow and reconstituted with 70 μL of methanol/water (1:1, v/v) as the solvent. The final extract was filtered through 0.22 μm centrifugal filters (Biotage, Uppsala, Sweden). Finally, 5 μL of filter liquor was subjected to LC‒MS/MS.

Parameter settings of target method based on LC-MS/MS

Targeted analysis was performed on a Exion LC AD ultrahigh-performance liquid chromatography (UHPLC) (AB SCIEX, Framingham, U.S.A) coupled with triple-quadrupole 6500 plus mass spectrometry (AB SCIEX, Framingham, U.S.A). The separation was conducted on an ACQUITY UPLC BEH C18 Column (Waters, Milford, MA) 2.1 × 50 mm 1.7 μm at 60 °C oven temperature. An ACQUITY UPLC® BEH C18 VanGuardTM Pre-Column (Waters, Milford, MA) 2.1 × 5 mm 1.7 μm was added before analytical column. The flow rate was 0.4 mL/min and the injection volume was 5 μL. Mobile phase A was 5 mM Ammonium acetate in water while mobile phase B was 5 mM Ammonium acetate in methanol. The gradient program for mobile phase B was started at 5% C maintained for 0.5 min, linearly increased to 50% in 4.5 min, then linearly increased to 70% for 4 min, linearly increased to 100% for 3 min, held for 2 min, then dropped sharply to 5% for 0.1 min, and held for 2 min. The total run time was 16.0 min. Valve switching was set so that the eluate of samples before 0.5 min did not enter the mass spectrometry. In the mass spectrometer system, ionization of targeted analytes was performed by electrospray ionization with positive/negative switching mode. The electrospray voltage is set at 5500 V for positive ion scanning mode and -4500 V for negative ion scanning mode. The curtain gas is set at 40.0 psi, and the ion source temperature is set at 350 °C. The ion source gases 1 (GS1) and 2 (GS2) are both set at 50 psi. Declustering potential and collision energy voltages were optimized for each chemical based on corresponding standard. The detailed parameter settings are presented in Supplementary Data  16 .

Quality control and assurance

To ensure the stability of the entire analytical process, we applied the following standard operating procedure: (1) The same internal standards were added to each real sample and QC sample. (2) The spiked serum samples were used as QC samples and were inserted after every 21 and 11 real samples for GC‒MS/MS and LC‒MS/MS, respectively. (3) Routine instrument maintenance, such as blank flushing, linear replacement for the GC‒MS/MS platform, needle replacement for the LC‒MS/MS platform, cleaning of the ion source, and full maintenance when necessary, was performed before each batch of sample analysis. (4) Multiple calibration curves were used throughout the whole sample analysis process, and each independent calibration curve was run for each batch of samples. For the GC platform, each batch consisted of 158 real samples and 21 QC samples; for the LC platform, each batch contained 264 samples and 24 QC samples. Hence, there were 29 and 20 independent calibration curves for the GC‒MS/MS and LC‒MS/MS platforms, respectively, bracketing the entire sample analysis process.

Data quality was evaluated using the following steps. (1) Quantification of the QC samples of the batch using the calibration curve of each batch. (2) Definition of the appropriate internal standard. For the GC platform, each internal standard was used to correct a dozen chemicals within their neighboring retention time regions; for the LC platform, internal standards were selected with the smallest RSD in QCs after internal standard correction 62 . In other words, the defined internal standard was used to correct the corresponding chemical in the calibration curves and QC samples, and the corrected calibration curve was used to quantify the chemicals in the QC samples of the corresponding batch. Finally, the optimal internal standard was selected based on the smallest RSD of the chemicals in the QC samples after correction and quantification. (3) Evaluation of the batch effect with QC samples. For all the QC samples, the mean concentrations of all the chemicals and the principal component score plots were used to evaluate batch effects. The effectiveness of batch effect correction can be determined by comparing the original signals with the batch-specific quantification concentrations and the batch-specific quantification combined with multiple internal standard corrections. A principal component analysis plot was generated using SIMCA-P 14.1 software (Umetrics, Umea, Sweden). (4) Evaluation of the reliability of the quantitative results with respect to the QC samples. Each chemical in the QC was calibrated with the selected internal standard, and the QC and samples of the corresponding batches were quantified by using the calibrated linearity. RSD < 30% for each substance in the QCs was used as the criterion for stable detection. Recovery within 80–120% for each substance in the QCs and calibration curves at low, medium, and high spiked concentrations were used as the criterion for accurate detection.

Methods examination and evaluation

In order to ensure the stability and feasibility of the established analytical method, methodological evaluation was carried out before conducting large-scale sample testing. The methodological examination and evaluation were reported in previous studies for GC-MS/MS 61 and LC-MS/MS 60 platforms, the present study extended the monitoring list and reanalyzed part of the methodological examination as follows. To determine the limit of detection and linearity of calibration, matrix-matched standard solutions at 0, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, and 100 ng/mL were analyzed for GC-MS/MS. As for LC-MS/MS platform, matrix-matched standard solutions at 0, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 25, 50, 100, and 200 ng/mL were analyzed. The minimum concentration satisfying a signal-to-noise ratio greater than 10 on the chromatogram was defined as the limit of quantification (LOQ). The linearity of calibration was evaluated (the number of calibrator levels greater than or equal to 5) by the correlation coefficient (r2) of the calibration curve from LOQ to a suitable concentration. A weighting regression factor of 1/x (x represent concentration) was adopted to minimize calculation error at low concentrations 63 . The precision of the two platforms was evaluated using spiked QC samples with the concentrations of 2 ng/mL and 10 ng/mL, respectively. Due to 49 calibration curves bracketing the entire sample analysis process, accuracy tests were performed for each one. For GC-MS/MS platform, the accuracy of each exposure in each calibration curve was calculated at 2, 5, and 20 ng/mL concentration levels, and for LC-MS/MS platform, the accuracy of each exposure in each calibration curve was calculated at 1, 10, and 100 ng/mL concentration levels. The raw method files of GC-MS/MS and LC-MS/MS were presented in Supplementary Software, which can be directly opened and used by Agilent “MassHunter Workstation” and AB SCIEX “Analyst” software, respectively.

Statistical analysis

For quantitative analysis of chemicals, peak areas of the serum chemicals and internal standards monitored by both the GC‒MS/MS and LC‒MS/MS platforms were obtained from the raw data by “masshunter” (Agilent, Santa Clara, USA) and “SCIEX OS” (AB SCIEX, Framingham, USA) software, respectively. Both software programs generate quantification curves using a weighted linear regression model (1/x) to ensure the accuracy of quantification at low concentrations 63 . The method of internal standard selection and calibration is described in the “QC and assurance” section. The internal standard calibration curve method was used for quantification. Concentrations less than the LOQ were set as LOQ/√2.

Quantification process of the internal standard calibration curve method: The internal standard calibration curve method was applied for quantification, which is commonly used in precise quantitative research of endogenous metabolites and exogenous chemicals in biological matrices. The point of this method is a group of consistent internal standard (isotopic labeled) mixture was added to both the calibration curve and each sample, and quantification was achieved through regression equations between peak area ratio and concentration. The specific process is as follows:

1. Define the appropriate internal standard. For the GC platform, internal standard was selected according to similarities in chemical properties and retention behavior, and used to correct chemicals within its neighboring retention time regions. For the LC platform, the internal standards were selected according to the rule of smallest RSD in QCs after internal standard correction (details see “Quality control and assurance” section). 2. Construct calibration curves after internal standard (IS) correction. Concentrations of analytes in calibration curve (Cc) were set as the independent variable, the ratios between peak areas and corresponding internal standard are set as the dependent variable (AREA Cc / AREA IS). A weighted linear regression model (weight = 1/x) was used to construct calibration curves. 3. Obtain relative responses of the samples after IS calibration. Relative responses (AREA sample / AREA IS) were the ratio between the peak area of the sample (AREA sample) and the peak area of the corresponding internal standard (AREA IS). 4. Calculate the concentration of the analyte in samples. Relative responses of samples (step 3) are substituted into the calibration curve in step 2 to obtain the sample concentration.

In addition, to ensure quantitative accuracy, a calibration curve was inserted at the beginning of each batch as mentioned before, therefore, the quantitative methods described above were further performed in batch-specific way, means that “each calibration curve was used to quantify the actual samples in its own batch.” This practice helps eliminate errors introduced by pauses in routine instrument maintenance. For the few analytes not assigned to the internal standard with smallest RSD in QCs, their quantification was achieved by external standard method.

Effect of epidemiological factors on the exposome: To assess which epidemiological factor influenced the overall exposure variation, variation partitioning analysis was performed using the R package “vegan”, and the exposure levels used were log10 transformed. Principal component analysis was carried out with the R package “FactoMineR”. To investigate the correlation between epidemiological factors and each chemical, PASW Statistics 18 software (SPSS, Chicago, IL) was used to conduct partial Spearman correlation analysis between one epidemiological factor and each chemical; the other eight epidemiological factors were used as confounders. The correlation coefficients were used to construct a correlation heatmap matrix and a correlation network diagram (Cytoscape software 3.7.1).

Stratified analysis of exposure characteristics: To explore differences in serum residue levels in different populations, samples were stratified based on each epidemiological factor. Exposures that differed significantly among groups are shown in Fig.  3 . First, for continuous and ordinal categorical variables, including age, education, and income, significantly correlated chemicals were obtained using partial Spearman correlation and multiple linear regression while controlling for the false discovery rate for multiple corrections ( p & FDR < 0.05). To exclude the interference of age on education level, data for those older 30 years of age were selected for analysis. Significantly different exposures are shown in a line graph using scaled concentration. Second, for binary epidemiological factors such as sex, smoking status, and alcohol consumption, the p values of nonparametric tests were adjusted by controlling for the false discovery rate ( p and FDR < 0.05). Considering the lack of data on female smoking and drinking habits, relationships between smoking and drinking and residue concentration were analyzed only for males. Significantly different exposures are shown in bar plots using the fold change in the geometric means. All statistical tests were two-sided. Third, geometric means were used to reduce the effect of extreme values when studying the regional distribution of exposures, and geometric means were obtained by PASW Statistics 18 software (SPSS, Chicago, IL). A regional heatmap and stacked bar plot were generated to visualize the regional distribution of exposures. A regional heatmap was generated with https://www.bioinformatics.com.cn , an online platform for data analysis and visualization. Other figures were generated with the package “ggplot2” in R software (version 4.2.1, R Foundation for Statistical Computing, Austria).

Risk of disease for each individual chemical: To adjust for confounders, the R package “MatchIt” was used to match a control sample for each disease, and patients with more than 2 diseases were not considered, except for metabolic syndrome; a control group was matched based on all nine epidemiological factors and five major chronic diseases (obesity, hypertension, diabetes, hyperuricemia, and hyperlipidemia). Subsequently, the risk of each chemical for disease was analyzed using binary logistic regression and multiple linear regression and again adjusted for the nine epidemiological factors as confounders. To understand susceptibility to disease risk from exposure in different age and sex populations, samples were grouped by age and sex. Specifically, the population was divided into three age groups: 30–50 years, 50–60 years, and 60–80 years. The participants were also divided into male and female groups. Subsequently, propensity score matching was used to classify the data sets into disease and control groups for each subgroup. Finally, binary logistic regression was used to determine ORs for the exposures in each group. The exposure concentrations used were log10 transformed for all regression analyses. Two-sided t tests and Hosmer–Lemeshow tests were employed for multiple linear regression and binary logistic regression, respectively.

Health risk assessment: First, available exposure guidance values were collected including BE, HBM II, BGV, and RfD 64 . Then health risk assessments were carried out using the HQ 65 : HQ = C serum /C guidance . HQ > 1 suggested exposure levels exceeding published human health benchmarks. Dose-response curve on the one hand can display the linear or nonlinear relationship between exposure dose and risk; on the other hand, a curve can identify the minimum exposure dose associated with increased disease risk (OR > 1). The R package “RSC” was used to determine the dose‒risk relationship of key exposures in this study.

Exposure mixture to disease risk: We used a combination of WQS 34 , q g-comp 35 , and BKMR 36 , 37 to assess the association of exposure mixtures with multiple chronic diseases. Each model was adjusted for the covariates region, age, sex, smoking status, and alcohol use. Unlike the univariate analysis described above, which used 4756 samples from the GC platform and 5513 samples from the LC platform, 4573 common samples were selected for mixed exposures, and propensity score matching was applied to identify case controls. In addition, 35 individuals with a detection frequency > 50% nationwide were selected for analysis to meet the model quartiles. The exposure levels used were log10 transformed and then scaled for multi-exposure analysis.

Brief description of three multi-exposure models

The weighted quantile sum regression (WQS) scores were estimated using the R package “gWQS“ 34 , which groups different chemicals into ordinal variables (quartiles), and calculated a weighted linear index through the WQS regression model, which represents the entire body burden of all chemicals. the WQS regression estimates sum mixture effects in either positive or negative directions, respectively, so the likelihood of an association was assessed in both directions in separate models. 1000 bootstrap runs were performed for each analysis to optimize the association between WQS scores and outcomes in the multivariate linear regression model. Ultimately, the model provides an estimate beta and significance of the total effect and the corresponding weight for each chemical, which shows how much a particular chemical contributes to the WQS index. Odds ratios (ORs) were calculated according to formula: OR = exp (beta). 95% confidence intervals of ORs were calculated according to formula: upper bound of OR = OR + standard error (OR) × 1.96, lower bound of OR = OR – standard error (OR) × 1.96.

Mixture effects was estimated using the R package “qgcomp“ 35 , which is similar to the WQS in that different chemicals are grouped into ordinal variables (quartiles) and a weighted linear index representing the cumulative effect is estimated by the regression model. The model also provides estimates and significance of the total effect and the corresponding weights for each chemical. The difference is that the WQS regression estimates are performed in either positive or negative directions separately whereas quantile g-computation (q g-comp) allows the joint effects of different directions of individual exposure to be assessed simultaneously in a single run. Moreover, the run speed can be greatly improved by G-computation and more robust associations than the WQS model can be obtained in small sample sizes.

BKMR, a non-parametric Bayesian variable selection framework, can investigate flexibly the joint effects of exposure mixtures on human health 36 , 37 . BKMR model can provides overall risk, single variable risk and non-linearity interaction of exposure to responses, and PIP for each exposure. Among them, PIP describe the relative importance of each exposure to the outcome of interest. Here, hierarchical variable selection method was used due to highly correlated exposures, then groupPIP and conditional PIPs (condPIP) were obtained to commonly evaluate the contributor of exposures. BKMR model was conducted with R package ‘bkmr’ using the binomial link function for binary outcomes and 5000 iterations using a Markov chain Monte Carlo algorithm to ensure convergence.

Reporting summary

Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

Data availability

All data supporting the findings of this study are available within the paper, in the supplementary information file, and in the source data file. The air pollution dataset of China can be found in https://quotsoft.net/air/ , and the surface climate dataset of China can be found in https://www.geodoi.ac.cn/WebCn/doi.aspx?Id=3187 . The concentration levels of 74 high-frequency exposures in human serum of Chinese chronic diseases population have been given in Table  2 of the paper, but the generated individual exposure atlas data are considered sensitive biomonitoring data, therefore, can not be publicly available according to the contracts with cooperating institutions (the initiator of the cohort) and the limitations included in the informed consents signed by the study participants. The request of these individual data is suggested by sending an email to the corresponding author Dr. Guowang Xu ([email protected]). Requests should include name, affiliation and contact details of the person requesting the data, which data are requested and the purpose of requesting the data. Requests will be subject to consideration by the management committee of the corresponding institutes and the sample collection institutes, including Dalian Institute of Chemical Physics, Chinese Academy of Sciences, National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention, Huazhong University of Science and Technology. If approved, the corresponding author will send the request data by email. Time frame for a response will be within 3 months. Data requests under agreement will be considered for purposes of reproducing the data and subject to appropriate confidentiality obligations and restrictions. Applicants must promise that these individual data applied for will only be used for scientific research and cannot be publicly released.  Source data are provided with this paper.

Code availability

No custom code was used in this study. The R codes for statistical analysis and figure production have been deposited to the GitHub and Zenodo ( https://doi.org/10.5281/zenodo.10391262 66 ).

Wild, C. P. Complementing the genome with an “exposome”: the outstanding challenge of environmental exposure measurement in molecular epidemiology. Cancer Epidemiol. Biomark. Prev. 14 , 1847–1850 (2005).

Article   CAS   Google Scholar  

Vermeulen, R., Schymanski, E. L., Barabási, A. L. & Miller, G. W. The exposome and health: where chemistry meets biology. Science 367 , 392–396 (2020).

Article   CAS   PubMed   PubMed Central   ADS   Google Scholar  

Rappaport, S. M. & Smith, M. T. Epidemiology. Environment and disease risks. Science 330 , 460–461 (2010).

Rappaport, S. M., Barupal, D. K., Wishart, D., Vineis, P. & Scalbert, A. The blood exposome and its role in discovering causes of disease. Environ. Health Perspect. 122 , 769–774 (2014).

Article   PubMed   PubMed Central   Google Scholar  

Svarcova, A. et al. Integration of five groups of POPs into one multi-analyte method for human blood serum analysis: an innovative approach within biomonitoring studies. Sci. Total Environ. 667 , 701–709 (2019).

Article   CAS   PubMed   ADS   Google Scholar  

Freire, C., Koifman, R. J. & Koifman, S. Serum levels of organochlorine pesticides in blood donors: a biomonitoring survey in the North of Brazil, 2010-2011. Sci. Total Environ. 598 , 722–732 (2017).

Luzardo, O. P. et al. Body burden of organohalogenated pollutants and polycyclic aromatic hydrocarbons in Romanian population: influence of age, gender, body mass index, and habitat. Sci. Total Environ. 656 , 709–716 (2019).

Silver, M. K. et al. Prenatal organophosphate insecticide exposure and infant sensory function. Int. J. Hyg. Environ. Health 221 , 469–478 (2018).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Li, J., Wang, P., Shi, S. & Xue, J. Background biomonitoring of residue levels of 137 pesticides in the blood plasma of the general population in Beijing. Environ. Monit. Assess. 190 , 315 (2018).

Article   PubMed   Google Scholar  

Qie, M., Zhao, Y., Yang, S., Wang, W. & Xu, Z. Rapid simultaneous determination of 160 drugs in urine and blood of livestock and poultry by ultra-high-performance liquid chromatography-tandem mass spectrometry. J. Chromatogr. A 1608 , 460423 (2019).

Article   CAS   PubMed   Google Scholar  

Yao, J. et al. Novel perfluoroalkyl ether carboxylic acids (PFECAs) and sulfonic acids (PFESAs): occurrence and association with serum biochemical parameters in residents living near a fluorochemical plant in China. Environ. Sci. Technol. 54 , 13389–13398 (2020).

Qin, Y. Y. et al. Halogenated POPs and PAHs in blood plasma of Hong Kong residents. Environ. Sci. Technol. 45 , 1630–1637 (2011).

Li, Z. et al. Phthalate metabolites in paired human serum and whole blood. Sci. Total Environ. 824 , 153792 (2022).

Cho, C. R., Lam, N. H., Cho, B. M., Kannan, K. & Cho, H. S. Concentration and correlations of perfluoroalkyl substances in whole blood among subjects from three different geographical areas in Korea. Sci. Total Environ. 512-513 , 397–405 (2015).

Luzardo, O. P. et al. Socioeconomic development as a determinant of the levels of organochlorine pesticides and PCBs in the inhabitants of Western and Central African countries. Sci. Total Environ. 497-498 , 97–105 (2014).

Wang, Y. et al. Concentrations and relationships between classes of persistent halogenated organic compounds in pooled human serum samples and air from Laizhou Bay, China. Sci. Total Environ. 482-483 , 276–282 (2014).

Goralczyk, K. et al. Perfluorinated chemicals in blood serum of inhabitants in central Poland in relation to gender and age. Sci. Total Environ. 532 , 548–555 (2015).

Jin, Q. et al. Biomonitoring of chlorinated polyfluoroalkyl ether sulfonic acid in the general population in central and eastern China: occurrence and associations with age/sex. Environ. Int. 144 , 106043 (2020).

Tyrrell, J., Melzer, D., Henley, W., Galloway, T. S. & Osborne, N. J. Associations between socioeconomic status and environmental toxicant concentrations in adults in the USA: NHANES 2001-2010. Environ. Int. 59 , 328–335 (2013).

Pan, Q. et al. Identification of the susceptible subpopulations for wide pulse pressure under long-term exposure to ambient particulate matters. Sci. Total Environ. 834 , 155311 (2022).

Collaborators, G. L. Age-sex differences in the global burden of lower respiratory infections and risk factors, 1990-2019: results from the Global Burden of Disease Study 2019. Lancet Infect. Dis. 22 , 1626–1647 (2022).

Article   Google Scholar  

Sen, P. et al. Exposure to environmental contaminants is associated with altered hepatic lipid metabolism in non-alcoholic fatty liver disease. J. Hepatol. 76 , 283–293 (2021).

Mi, X. et al. Serum levels of per- and polyfluoroalkyl substances alternatives and blood pressure by sex status: isomers of C8 health project in China. Chemosphere 261 , 127691 (2020).

Bauer, U. E., Briss, P. A., Goodman, R. A. & Bowman, B. A. Prevention of chronic disease in the 21st century: elimination of the leading preventable causes of premature death and disability in the USA. Lancet 384 , 45–52 (2014).

Willett, W. C. Balancing life-style and genomics research for disease prevention. Science 296 , 695–698 (2002).

Lu, X., Xu, X., Lin, Y., Zhang, Y. & Huo, X. Phthalate exposure as a risk factor for hypertension. Environ. Sci. Pollut. Res. Int. 25 , 20550–20561 (2018).

Singh, K. & Chan, H. M. Persistent organic pollutants and diabetes among Inuit in the Canadian Arctic. Environ. Int. 101 , 183–189 (2017).

Predieri, B. et al. Endocrine disrupting chemicals and type 1 diabetes. Int. J. Mol. Sci. 21 , 2937 (2020).

Han, M., Ma, A., Dong, Z., Yin, J. & Shao, B. Organochlorine pesticides and polycyclic aromatic hydrocarbons in serum of Beijing population: Exposure and health risk assessment. Sci. Total Environ. 860 , 160358 (2022).

Lin, P. D. et al. Per- and polyfluoroalkyl substances and blood lipid levels in pre-diabetic adults-longitudinal analysis of the diabetes prevention program outcomes study. Environ. Int. 129 , 343–353 (2019).

Yang, Q. et al. Association of serum levels of perfluoroalkyl substances (PFASs) with the metabolic syndrome (MetS) in Chinese male adults: A cross-sectional study. Sci. Total Environ. 621 , 1542–1549 (2018).

Arrebola, J. P. et al. Associations of multiple exposures to persistent toxic substances with the risk of hyperuricemia and subclinical uric acid levels in BIOAMBIENT.ES study. Environ. Int. 123 , 512–521 (2019).

Shankar, A., Xiao, J. & Ducatman, A. Perfluoroalkyl chemicals and elevated serum uric acid in US adults. Clin. Epidemiol. 3 , 251–258 (2011).

Carrico, C., Gennings, C., Wheeler, D. C. & Factor-Litvak, P. Characterization of weighted quantile sum regression for highly correlated data in a risk analysis setting. J. Agric. Biol. Environ. Stat. 20 , 100–120 (2015).

Article   MathSciNet   PubMed   Google Scholar  

Keil, A. P. et al. A quantile-based g-computation approach to addressing the effects of exposure mixtures. Environ. Health Perspect. 128 , 47004 (2020).

Bobb, J. F. et al. Bayesian kernel machine regression for estimating the health effects of multi-pollutant mixtures. Biostatistics 16 , 493–508 (2015).

Bobb, J. F., Claus Henn, B., Valeri, L. & Coull, B. A. Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression. Environ. Health 17 , 67 (2018).

Kou, J. et al. Accumulative levels, temporal and spatial distribution of common chemical pollutants in the blood of Chinese adults. Environ. Pollut. 311 , 119980 (2022).

Sun, S. et al. Levels and patterns of polychlorinated dibenzo-p-dioxins and dibenzofurans and polychlorinated biphenyls in foodstuffs of animal origin from Chinese markets and implications of dietary exposure. Environ. Pollut. 273 , 116344 (2020).

Chen, S., Zhou, Y., Meng, J. & Wang, T. Seasonal and annual variations in removal efficiency of perfluoroalkyl substances by different wastewater treatment processes. Environ. Pollut. 242 , 2059–2067 (2018).

Liu, Y. et al. Prenatal exposure and transplacental transfer of perfluoroalkyl substance isomers in participants from the upper and lower reaches of the Yangtze River. Environ. Pollut. 270 , 116202 (2021).

Wang, Z., Cousins, I. T., Scheringer, M., Buck, R. C. & Hungerbuhler, K. Global emission inventories for C4-C14 perfluoroalkyl carboxylic acid (PFCA) homologues from 1951 to 2030, Part I: production and emissions from quantifiable sources. Environ. Int. 70 , 62–75 (2014).

Wang, Q. et al. Organochlorine pesticide concentrations in pooled serum of people in different age groups from five Chinese cities. Sci. Total Environ. 586 , 1012–1019 (2017).

Richterova, D. et al. PFAS levels and determinants of variability in exposure in European teenagers - Results from the HBM4EU aligned studies (2014-2021). Int. J. Hyg. Environ. Health 247 , 114057 (2023).

Chang, C. et al. Current pesticide profiles in blood serum of adults in Jiangsu Province of China and a comparison with other countries. Environ. Int. 102 , 213–222 (2017).

Seo, S. H., Son, M. H., Choi, S. D., Lee, D. H. & Chang, Y. S. Influence of exposure to perfluoroalkyl substances (PFASs) on the Korean general population: 10-year trend and health effects. Environ. Int. 113 , 149–161 (2018).

Fisher, M., Arbuckle, T. E., Wade, M. & Haines, D. A. Do perfluoroalkyl substances affect metabolic function and plasma lipids?–Analysis of the 2007-2009, Canadian Health Measures Survey (CHMS) Cycle 1. Environ. Res. 121 , 95–103 (2013).

Perez-Carrascosa, F. M. et al. Historical exposure to persistent organic pollutants and cardiovascular disease: a 15-year longitudinal analysis focused on pharmaceutical consumption in primary care. Environ. Int. 156 , 106734 (2021).

Nilsson, P. M., Tuomilehto, J. & Ryden, L. The metabolic syndrome—what is it and how should it be managed? Eur. J. Prev. Cardiol. 26 , 33–46 (2019).

Lee, Y. M. et al. Prospective associations between persistent organic pollutants and metabolic syndrome: a nested case-control study. Sci. Total Environ. 496 , 219–225 (2014).

Zamora, A. N. et al. Exposure to phenols, phthalates, and parabens and development of metabolic syndrome among Mexican Women in midlife. Front Public Health 9 , 620769 (2021).

Feig, D. I., Kang, D. H. & Johnson, R. J. Uric acid and cardiovascular risk. N. Engl. J. Med. 359 , 1811–1821 (2008).

Edwards, N. L. The role of hyperuricemia in vascular disorders. Curr. Opin. Rheumatol. 21 , 132–137 (2009).

Scinicariello, F. et al. Perfluoroalkyl acids, hyperuricemia and gout in adults: analyses of NHANES 2009-2014. Chemosphere 259 , 127446 (2020).

Li, Y. C. et al. Machine learning-driven identification of early-life air toxic combinations associated with childhood asthma outcomes. J. Clin. Investig. 131 , e152088 (2021).

Zhang, B. et al. The China Health and Nutrition Survey, 1989-2011. Obes. Rev. 15 , 2–7 (2014).

Su, C. et al. Longitudinal association between urbanicity and total dietary fat intake in adults in urbanizing China from 1991 to 2015: findings from the CHNS. Nutrients 12 , 1597 (2020).

van Donkelaar, A. et al. Monthly global estimates of fine particulate matter and their uncertainty. Environ. Sci. Technol. 55 , 15287–15300 (2021).

Article   PubMed   ADS   Google Scholar  

Liao, J., Shi, Y., Chen, J. & Zhao, Y. F. Dataset of Surface Climate Normals from 219 Chinese International Exchange Stations (1991-2020). Digital Journal of Global Change Data Repository https://doi.org/10.3974/geodb.2022.11.06.V1 (2022).

You, L. et al. Metabolome-wide association study of serum exogenous chemical residues in a cohort with 5 major chronic diseases. Environ. Int. 158 , 106919 (2022).

Li, X. et al. Simultaneous determination of 35 organochlorine pesticides and polychlorinated biphenyls in the serum of the general population in wuhan by solid phase extraction-gas chromatography-tandem mass spectrometry. Chin. J. Chromatogr. 40 , 461–468 (2022).

Article   CAS   ADS   Google Scholar  

Sysi-Aho, M., Katajamaa, M., Yetukuri, L. & Oresic, M. Normalization method for metabolomics data using optimal selection of multiple internal standards. BMC Bioinforma. 8 , 93 (2007).

Koponen, J., Rantakokko, P., Airaksinen, R. & Kiviranta, H. Determination of selected perfluorinated alkyl acids and persistent organic pollutants from a small volume human serum sample relevant for epidemiological studies. J. Chromatogr. A 1309 , 48–55 (2013).

Nakayama, S. F. et al. Interpreting biomonitoring data: Introducing the international human biomonitoring (i-HBM) working group’s health-based guidance value (HB2GV) dashboard. Int. J. Hyg. Environ. Health 247 , 114046 (2023).

Aylward, L. L., Kirman, C. R., Schoeny, R., Portier, C. J. & Hays, S. M. Evaluation of Biomonitoring Data from the CDC National Exposure Report in a Risk Assessment Context: Perspectives across Chemicals. Environ. Health Perspect. 121 , 287–294 (2013).

You, L., et al. An exposome atlas of serum reveals risk of chronic diseases in Chinese population. Zenodo. https://doi.org/10.5281/zenodo.10391262 (2023).

Download references

Acknowledgements

This research was supported by the National Key R&D Program of China (2019YFC1605100, G.X.), the National Natural Science Foundation of China (No. 21934006, G.X.), Youth Innovation Promotion Association CAS (2021186, Xinyu Liu), and the innovation program (DICP ZZBS201804, G.X.) of science and research from the DICP, CAS.

Author information

These authors contributed equally: Lei You, Jing Kou, Mengdie Wang, Guoqin Ji.

Authors and Affiliations

CAS Key Laboratory of Separation Science for Analytical Chemistry, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, 457 Zhongshan Road, Dalian, 116023, China

Lei You, Mengdie Wang, Guoqin Ji, Fujian Zheng, Yuting Wang, Tiantian Chen, Ting Li, Lina Zhou, Xianzhe Shi, Chunxia Zhao, Xinyu Liu & Guowang Xu

University of Chinese Academy of Sciences, Beijing, 100049, China

Lei You, Fujian Zheng, Yuting Wang, Tiantian Chen, Ting Li, Lina Zhou, Xianzhe Shi, Chunxia Zhao, Xinyu Liu & Guowang Xu

Liaoning Province Key Laboratory of Metabolomics, Dalian, 116023, China

State Key Laboratory of Environment Health (Incubation), Key Laboratory of Environment and Health, Ministry of Education, Key Laboratory of Environment and Health (Wuhan), Ministry of Environmental Protection, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, # 13 Hangkong Road, Wuhan, Hubei, 430030, China

Jing Kou, Xiang Li, Mingye Zhang & Surong Mei

School of Public Health, China Medical University, No. 77 Puhe Road, Shenbei New District, Shenyang, 110122, China

Mengdie Wang

School of Life Science, China Medical University, No. 77 Puhe Road, Shenbei New District, Shenyang, 110122, China

National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention, Beijing, 100050, China

You can also search for this author in PubMed   Google Scholar

Contributions

G.X., Xinyu Liu. and S.M. developed the concept and designed experiments. C.S. collected and provided the human serum samples. L.Y. M.W. and G.J. performed the experiments of LC-MS/MS part and analyzed the data. F.Z. contributed to data analysis. J.K., Xiang Li., and M.Z. performed the experiments of GC-MS/MS part and analyzed the data. Y.W., T.C., T.L., L.Z., X.S., and C.Z. provided significant intellectual input. L.Y. drafted the manuscript. Xinyu Liu., G.X., and S.M. contributes to article revisions.

Corresponding authors

Correspondence to Xinyu Liu , Surong Mei or Guowang Xu .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature Communications thanks Alex Chao, Yue Qi, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Peer review file, description of additional supplementary files, supplementary data 1–16, supplementary software, reporting summary, source data, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

You, L., Kou, J., Wang, M. et al. An exposome atlas of serum reveals the risk of chronic diseases in the Chinese population. Nat Commun 15 , 2268 (2024). https://doi.org/10.1038/s41467-024-46595-z

Download citation

Received : 01 March 2023

Accepted : 04 March 2024

Published : 13 March 2024

DOI : https://doi.org/10.1038/s41467-024-46595-z

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

• Search Menu
• Volume 2024, Issue 3, March 2024 (In Progress)
• Volume 2024, Issue 2, February 2024
• Case of the Year
• MSF Case Reports
• Audiovestibular medicine
• Cardiology and cardiovascular systems
• Critical care medicine
• Dermatology
• Emergency medicine
• Endocrinology and metabolism
• Gastroenterology and hepatology
• Geriatrics and gerontology
• Haematology
• Infectious diseases and tropical medicine
• Medical ophthalmology
• Medical disorders in pregnancy
• Paediatrics
• Palliative medicine
• Pharmacology and pharmacy
• Radiology, nuclear medicine, and medical imaging
• Respiratory disorders
• Rheumatology
• Sexual and reproductive health
• Sports medicine
• Substance abuse
• Author Guidelines
• Submission Site
• Open Access
• Editorial Board
• Advertising and Corporate Services
• Journals Career Network
• Self-Archiving Policy
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Introduction, case report, acknowledgements, conflict of interest statement, ethical approval.

• < Previous

A case of paroxysmal cold hemoglobinuria complicated by latent syphilis

• Article contents
• Figures & tables
• Supplementary Data

Tsuyoshi Hirata, Naoko Kubota, Kazuaki Fukushima, Erika Takami, Tsuyoshi Kato, Tomomi Okamoto, A case of paroxysmal cold hemoglobinuria complicated by latent syphilis, Oxford Medical Case Reports , Volume 2024, Issue 3, March 2024, omae009, https://doi.org/10.1093/omcr/omae009

• Permissions Icon Permissions

An 80-year-old man presented in December with the main complaint of jaundice. Blood tests revealed hemolytic anemia and renal dysfunction. Positive syphilis serology results led to a diagnosis of untreated latent syphilis. A positive direct Coombs test led to a diagnosis of autoimmune hemolytic anemia (AIHA). Antibiotics were started for the syphilis, with improvement in the anemia and renal dysfunction observed. However, paroxysmal intravascular hemolysis occurred after his discharge. Based on a positive Donath-Landsteiner (D-L) test, paroxysmal cold hemoglobinuria (PCH) diagnosis was made. The hemolytic anemia improved after further treatment for syphilis, and further avoiding exposure to cold.

Paroxysmal cold hemoglobinuria (PCH) is a very rare type of autoimmune hemolytic anemia (AIHA). PCH is classified as a condition that is associated with cold, accounting for only 1% of all AIHAs [ 1 ]. PCH is seen in young children after viral illnesses, in some hematologic malignancies, and in syphilis cases. In the past, there were many syphilitic cases reported; however, at the present time, most reported cases are non-syphilitic. We report a case of acute paroxysmal cold hemoglobinuria in an elderly male with latent syphilis.

An 80-year-old man with a history of hypertension, chronic heart failure, and dyslipidemia presented to our hospital in December with the main complaint of jaundice. Laboratory studies revealed anemia with hemolysis, an elevated LDH level (LDH 1358 U/l), and indirect hyperbilirubinemia (total bilirubin: 9.4 mg/dl, direct bilirubin: 0.7 mg/dl). Syphilis serology was found to have positive qualitative RPR (389.9×) and TPLA (9937.0×). Based on these findings, we diagnosed the patient with untreated latent syphilis. Blood tests showed elevated reticulocytes, decreased serum haptoglobin, and positive IgG and C3b3d in a direct Coombs test. After reviewing his lab results, a diagnosis of AIHA was made. Bone marrow examination showed no findings suggestive of malignancy. After admission, antibiotics (ceftriaxone 2 g/day) were started for the syphilis for 14 days, with a subsequent improvement in the anemia, renal dysfunction, LDH level, and indirect hyperbilirubinemia observed. The baseline (Day 1: at admission) and follow-up (Day 14: at the day before discharge) laboratory studies are shown in Table. On day 15, the patient was discharged to home. However, nausea and vomiting appeared in the morning after discharge, along with a flare-up of his jaundice, thereby resulting in re-hospitalization on day 16 ( Fig. 1 ).

Time course of total bilirubin (T-bil) and lactic acid dehydrogenase (LDH) values.

Blood tests after the re-hospitalization showed recurrent hemolytic anemia ( Table 1 ). Because of these repeated episodes of paroxysmal hemolytic anemia, along with his history of living without a heater in his home during the winter, cold AIHA was suspected. As a result, the patient underwent cold agglutination and D-L tests.

Laboratory Data on Day 1, Day 14, and Day 16. Day 1: Hospitalization, Day 14: The day before discharge, Day 16: Re-hospitalization

The D-L test results were positive and the cold agglutination titer was 64× ( Fig. 2 ). These results confirmed the PCH diagnosis. After an additional antibiotic treatment for the latent syphilis (doxycycline 200 mg/day) along with avoiding any exposure to cold by keeping himself warm, the hemolytic anemia improved. The antibiotics were changed to doxycycline 200 mg/day for oral switch. The patient’s progress after discharge from the hospital is unknown because he transferred to another hospital.

Donath-Landsteiner test. ( A ) 0°C for 30 min, followed by 37°C for 30 min. ( B ) 37°C for 60 min (control).

PCH is one AIHA in which complement activation by D-L antibodies causes intravascular hemolysis. In 1904, Donath and Landsteiner demonstrated the presence of causative autoantibodies and established the original concept for this disease [ 2 ]. Although PCH used to be frequently seen in syphilitic cases, presently it is primarily seen in children and often develops after viral infections and other diseases [ 3 ].

In the current case, hemolytic anemia was suspected based on blood tests and clinical symptoms. As a direct Coombs test was positive, this was suggestive of a diagnosis of AIHA. Urinalysis demonstrated there was a decreased haptoglobin, which is a finding of hemoglobinuria. A CT scan indicated that there was no splenomegaly, which suggested intravascular hemolysis. Previous findings have indicated that warm AIHA is unlikely to be the cause of intravascular hemolysis, as warm AIHA is predominantly associated with extravascular hemolysis [ 4 ]. Although a bone marrow biopsy was negative for malignant disease in this current case, the possibility of remained Paroxysmal nocturnal hematuria (PNH). With regard to the clinical course of the patient, improvement in the hemolytic symptoms that occurred without direct treatment for AIHA after his admission suggested that PNH was not likely [ 5 ]. Cold agglutinin disease (CAD) and PCH are diseases that cause intravascular hemolysis. This patient had repeated episodes of paroxysmal intravascular hemolysis during the winter season when he was admitted, and thus, CAD and PCH also came up in the differential diagnosis. A D-L test was performed in order to further evaluate the patient for PCH, with the positive results ultimately leading to the diagnosis of PCH.

The treatment of PCH is basically supportive therapy, such as avoiding cold exposure, keeping the patient warm, and the use of steroids, which have been reported to be effective and may be considered for use in severe cases [ 6 ]. Although reports of syphilitic PCH have been extremely rare in recent years, it has been reported that antibiotic treatment for syphilis may reduce or eliminate hemolysis. Although the efficacy of steroids in syphilitic PCH is not known, one study reported that steroids were effective during the acute phase of PCH in an elderly patient [ 7 ].

In our current case, the hemolysis improved in conjunction with avoidance of cold exposure by keeping the patient warm along with antibiotic treatment for the syphilis. Treatment with steroids was not considered as the patient improved soon after admission.

However, the disease recurred even after initiation of antibiotic treatment, so avoidance of cold exposure might have been more useful in this case.

Furthermore, it has been suggested that early intervention may prevent severe disease, and thus, it is important to take a history and perform appropriate tests in order to evaluate these types of patients.

When evaluating patients for PCH, which is a rare complication in syphilis-infected patients, consideration of cold exposure and testing for syphilis may be important factors in helping to differentiate from hemolytic anemia.

There is no financial support for this publication.

No conflicts of interest.

None declared.

No ethical approval is required for case report in our center.

Consent from the patient was taken for the writing and publication of this case report.

Dr. Naoko Kubota.

Wilma   B . Immune Hemolysis: diagnosis and treatment recommendations . Semin Hematol   2015 ; 52 : 304 – 12 .

Google Scholar

Donath   J , Landsteiner   K . Uber paroxysmale hämoglobinurie . Münch Med Wochenschr   1904 ; 51 : 1590 .

Erina   N , Hiroshi   Y , Hitoshi   K . Paroxysmal cold hemoglobinuria in Japan: a literature review . Japanese J Transfus Cell Ther   2021 ; 67 : 541 – 8 .

Brodsky   RA . Warm autoimmune Hemolytic Anemia . N Engl J Med   2019 ; 381 : 647 – 54 .

Hill   A , DeZern   AE , Kinoshita   T , Brodsky   RA . Paroxysmal nocturnal haemoglobinuria . Nat Rev Dis Primers   2017 ; 3 : 17028 .

Mantadakis   E , Bezirgiannidou   Z , Martinis   G , Athanassios   C . Recurrence of paroxysmal cold hemoglobinuria in a boy after physical cooling for fever . J Pediatr Hematol Oncol   2011 ; 33 : 40 – 2 .

Patel   P , Guevara   E , Changela   A , Thar   YY . Paroxysmal cold hemoglobinuria in an elderly patient: a rare case with poor prognosis . J Case Rep Images Med   2016 ; 2 : 20 – 3 .

Email alerts

Citing articles via, affiliations.

• Online ISSN 2053-8855
• Copyright © 2024 Oxford University Press
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

The intermittent fasting trend may pose risks to your heart

Intermittent fasting — when people only eat at certain times of day — has exploded in popularity in recent years. But now a surprising new study suggests that there might be reason to be cautious: It found that some intermittent fasters were more likely to die of heart disease.

The findings were presented Monday at an American Heart Association meeting in Chicago and focused on a popular version of intermittent fasting that involves eating all your meals in just eight hours or less — resulting in at least a 16-hour daily fast, commonly known as “time-restricted” eating.

The study analyzed data on the dietary habits of 20,000 adults across the United States who were followed from 2003 to 2018. They found that people who adhered to the eight-hour eating plan had a 91 percent higher risk of dying from heart disease compared to people who followed a more traditional dietary pattern of eating their food across 12 to 16 hours each day.

The scientists found that this increased risk also applied to people who were already living with a chronic disease or cancer. People with existing cardiovascular disease who followed a time-restricted eating pattern had a 66 percent higher risk of dying from heart disease or a stroke. Those who had cancer meanwhile were more likely to die of the disease if they followed a time-restricted diet compared to people with cancer who followed an eating duration of at least 16 hours a day.

The study results suggest that people who practice intermittent fasting for long periods of time, particularly those with existing heart conditions or cancer, should be “extremely cautious,” said Victor Wenze Zhong, the lead author and the chair of the department of epidemiology and biostatistics at the Shanghai Jiao Tong University School of Medicine in China.

“Based on the evidence as of now, focusing on what people eat appears to be more important than focusing on the time when they eat,” he added.

Zhong said that he and his colleagues conducted the new study because they wanted to see how eating in a narrow window each day would impact “hard endpoints” such as heart disease and mortality. He said that they were surprised by their findings.

“We had expected that long-term adoption of eight-hour time restricted eating would be associated with a lower risk of cardiovascular death and even all-cause death,” he said.

Losing lean muscle mass

The data didn’t explain why time-restricted eating increased a person’s health risks. But the researchers did find that people who followed a 16:8 time-restricted eating pattern, where they eat during an eight-hour window and fast for 16, had less lean muscle mass compared to people who ate throughout longer periods of the day. That lines up with a previous clinical trial published in JAMA Internal Medicine , which found that people assigned to follow a time-restricted diet for three months lost more muscle than a control group that was not assigned to do intermittent fasting.

Get Well+Being tips straight to your inbox

Holding onto muscle as you age is important. It protects you against falls and disability and can boost your metabolic health. Studies have found that having low muscle mass is linked to higher mortality rates, including a higher risk of dying from heart disease, said Zhong.

He stressed that the findings were not definitive. The study uncovered a correlation between time-restricted eating and increased mortality, but it could not show cause and effect. It’s possible for example that people who restricted their food intake to an eight-hour daily window had other habits or risk factors that might explain their increased likelihood of dying from heart disease. The scientists also noted that the study relied on self-reported dietary information. It’s also possible that the participants did not always accurately report their eating durations.

A trendy form of dieting and weight control

Intermittent fasting has been widely touted by celebrities and health experts who say it produces weight loss and a variety of health benefits. Another form of intermittent fasting involves alternating fasting days with days of eating normally. Some people follow the 5:2 diet, in which they eat normally for five days a week and then fast for two days.

But time-restricted eating is generally considered the easiest form of intermittent fasting for people to follow because it doesn’t require full-day fasts. It also typically doesn’t involve excessive food restriction. Adherents often eat or drink whatever they want during the eight-hour eating period — the only rule is that they don’t eat at other times of day.

Some of the earliest studies on time-restricted eating found that it helped prevent mice from developing obesity and metabolic syndrome. These were followed by mostly small clinical trials in humans, some of which showed that time-restricted eating helped people lose weight and improve their blood pressure , blood sugar and cholesterol levels. These studies were largely short-term, typically lasting one to three months, and in some cases showed no benefit .

One of the most rigorous studies of time-restricted eating was published in the New England Journal of Medicine in 2022. It found that people with obesity who were assigned to follow a low-calorie diet and instructed to eat only between the hours of 8 a.m. and 4 p.m. daily lost no more weight than people who ate the same number of calories throughout the day with no restrictions on when they could eat. The two diets had similar effects on blood pressure, blood sugar, cholesterol, and other metabolic markers.

The findings suggest that any benefits of time-restricted eating likely result from eating fewer calories.

More questions about intermittent fasting

Christopher Gardner, the director of nutrition studies at the Stanford Prevention Research Center, said he encouraged people to approach the new study with “healthy skepticism.” He said that while the findings were interesting, he wants to see all the data, including potential demographic differences in the study subjects.

“Did they all have the same level of disposable income and the same level of stress,” he said. “Or is it that the people who ate less than eight hours a day worked three jobs, had very high stress, and didn’t have time to eat?”

Gardner said that studying intermittent fasting can be challenging because there are so many variations of it, and determining its impact on longevity requires closely following people for long periods of time.

But he said that so far, the evidence supporting intermittent fasting for weight loss and other outcomes is mixed at best, with some studies showing short-term benefits and others showing no benefit at all. “I don’t think the data are very strong for intermittent fasting,” he added. “One of the challenges in nutrition is that just because something works really well for a few people doesn’t mean it’s going to work for everyone.”

He said that his biggest complaint with intermittent fasting is that it doesn’t address diet quality. “It doesn’t say anything about choosing poorly when you’re eating,” he said. “What if I have an eight-hour eating window but I’m eating Pop Tarts and Cheetos and drinking Coke in that window? I’m not a fan of that long term. I think that’s potentially problematic.”

Do you have a question about healthy eating? Email [email protected] and we may answer your question in a future column.

Read more from Well+Being

Well+Being shares news and advice for living well every day. Sign up for our newsletter to get tips directly in your inbox.

Vitamin B12 for fatigue has no proven benefit unless you have a deficiency that causes anemia.

Flashes, shimmers and blind spots: Here’s what migraine aura looks like.

Sparkling water is a better choice for your teeth than most popular beverages.

You can help your brain form healthy habits and break the bad ones.

Our 7 best tips to build an exercise habit .

• The intermittent fasting trend may pose risks to your heart March 18, 2024 The intermittent fasting trend may pose risks to your heart March 18, 2024
• Should I be worried about lead in my cinnamon? March 15, 2024 Should I be worried about lead in my cinnamon? March 15, 2024
• Are all olive oils equally healthy for you? A look at the research. March 18, 2024 Are all olive oils equally healthy for you? A look at the research. March 18, 2024

• Open access
• Published: 16 March 2024

Development and validation of a machine learning model to predict time to renal replacement therapy in patients with chronic kidney disease

• Jun Okita 1 ,
• Takeshi Nakata 1 ,
• Hiroki Uchida 1 ,
• Akiko Kudo 1 ,
• Akihiro Fukuda 1 ,
• Tamio Ueno 2 ,
• Masato Tanigawa 3 ,
• Noboru Sato 4 &
• Hirotaka Shibata 1  

BMC Nephrology volume  25 , Article number:  101 ( 2024 ) Cite this article

210 Accesses

Metrics details

Predicting time to renal replacement therapy (RRT) is important in patients at high risk for end-stage kidney disease. We developed and validated machine learning models for predicting the time to RRT and compared its accuracy with conventional prediction methods that uses the rate of estimated glomerular filtration rate (eGFR) decline.

Data of adult chronic kidney disease (CKD) patients who underwent hemodialysis at Oita University Hospital from April 2016 to March 2021 were extracted from electronic medical records ( N  = 135). A new machine learning predictor was compared with the established prediction method that uses the eGFR decline rate and the accuracy of the prediction models was determined using the coefficient of determination (R 2 ). The data were preprocessed and split into training and validation datasets. We created multiple machine learning models using the training data and evaluated their accuracy using validation data. Furthermore, we predicted the time to RRT using a conventional prediction method that uses the eGFR decline rate for patients who had measured eGFR three or more times in two years and evaluated its accuracy.

The least absolute shrinkage and selection operator regression model exhibited moderate accuracy with an R 2 of 0.60. By contrast, the conventional prediction method was found to be extremely low with an R 2 of -17.1.

Conclusions

The significance of this study is that it shows that machine learning can predict time to RRT moderately well with continuous values from data at a single time point. This approach outperforms the conventional prediction method that uses eGFR time series data and presents new avenues for CKD treatment.

Peer Review reports

The number of dialysis patients is increasing globally and is expected to be 3.8 million people worldwide by 2021 [ 1 ]. Chronic kidney disease (CKD) is a concept proposed for the early detection of renal dysfunction, and it is estimated that 9.1% of the world’s population is affected by CKD [ 2 ]. The Kidney Disease Improving Global Outcomes (KDIGO) guideline provides a heat map of the risk of progression to end-stage kidney disease (ESKD), and the National Institute for Health and Care Excellence guideline recommends the kidney failure risk equation (KFRE) as a criterion for referral to a nephrologist [ 3 , 4 , 5 ]. However, CKD patients often delay referral to a nephrologist or discontinue seeing a nephrologist because of a lack of subjective symptoms and their reluctance to continue treatment.

In high-risk patients with ESKD, renal function often deteriorates progressively, making it crucial to predict the time to renal replacement therapy (RRT) to achieve a clearer and concrete description of the necessity of treatment. Conventionally, time series graphs of estimated glomerular filtration rate (eGFR) or reciprocal creatinine (Cr) are used to estimate the time to RRT based on the annual decline rate. However, these methods require time series data, which is difficult to generate for first-time patients. In addition, the method of reciprocal Cr has been reported to have low accuracy [ 6 , 7 ].

Recently, there have been several reports on the application of artificial intelligence (AI) technology in medical care, covering a wide range of areas such as genomic medicine, image diagnosis, diagnostic and therapeutic support, and surgical support [ 8 , 9 , 10 , 11 , 12 ]. Machine learning is a technique for constructing a system to process tasks using big data. Supervised learning, which is a subcategory of machine learning, is divided into two types: classification, which predicts discrete values, and regression, which predicts continuous values. The classification accuracy in the field of image diagnosis is high, and the accuracy has already surpassed that of humans [ 13 ]. In risk assessment in the renal field, there were 39 reports on predicting the risk of developing acute kidney injury (AKI) as of March 2021, and Flechet et al. reported that the AKI predictor (available on the web) predicted AKI with a higher accuracy than that of physicians’ predictions [ 14 , 15 ]. As of October 2021, there were 87 reports that predicted the risk of CKD patients developing ESKD within one to five years, with the area under the curve values ranging from 0.90 to 0.96, which indicates an accuracy comparable to that of the existing KFRE [ 16 , 17 , 18 , 19 , 20 ]. As an example of regression, Dai et al. developed a model to predict hospitalization costs for patients with CKD [ 21 ].

There have been several reports of classification models that can predict the risk of ESKD at a specific point in time; for example, “After 2 years, the probability of being ESKD is XX%.” However, no studies have reported a regression model that can predict the time to RRT with continuous values, such as “You will need RRT in XX days.” In this study, we use a regression model to predict the time to RRT with continuous values based on data obtained at a single time point. The proposed model can also be used for first time patients. Furthermore, we examine the accuracy of the conventional prediction method using eGFR time series data and compare it with that of the machine learning model.

Adult patients (aged 20 years or older) with CKD who underwent hemodialysis at Oita University Hospital from April 2016 to March 2021 were selected retrospectively from electronic medical records. These patients were monitored at our hospital for at least three months until the induction of hemodialysis. Patients who had previously undergone RRT other than hemodialysis (peritoneal dialysis or renal transplantation) were excluded.

A total of 35 items were extracted from patient background and laboratory data up to the start of dialysis (Table  1 ). For laboratory data, we used CKD treatment items (anemia, CKD-MBD, blood glucose, and lipids) and indicators of nutritional status (albumin, total lymphocyte count (TLC), and cholinesterase (ChE)). These items have been used in previous studies on predictive models or noted in CKD guidelines as being associated with CKD progression [ 3 , 17 , 22 , 23 , 24 , 25 , 26 ]. For eGFR, we used the glomerular filtration rate estimation formula frequently used in Japan [ 27 ]:

eGFR (ml/min/1.73 m 2 ) = 194 × Cr − 1.094 × age − 0.287 (multiply by 0.739 for women).

Data were processed and analyzed using Microsoft Excel 2021 and Python (version 3.8.16) from Google Colaboratory. The following packages were used in Python: Scikit-Learn (version 1.4.0) and Statsmodels (version 0.14.1). As an exploratory data analysis, the missing values and correlation coefficients were examined, and the data series with missing data were excluded from the analysis. Removal of missing data is necessary for the analysis, but it reduces the overall data volume and may introduce data bias. Categorical variables were converted to numeric values; CKD etiology was a 5-item categorical variable, including diabetic nephropathy (DN), with an additional data column created for each item and converted to 0 or 1 (one-hot encoding). This approach homogenizes the information and tends to increase the number of data items. The other items were continuous variables and were standardized.

Using Scikit-Learn’s GroupShuffleSplit function, the data were randomly split, with 75% for training and 25% for validation, such that data from the same case were not included in either group. Using the training data, supervised learning was performed with the number of days from the date of examination to the start of dialysis as the objective variable and the other items as explanatory variables. The learning algorithms used were linear regression, ridge regression, least absolute shrinkage and selection operator (LASSO) regression, elastic net, random forest, and gradient boosting decision tree (GBDT) based on the cheat sheet in Scikit-Learn [ 28 ]. Using Scikit-Learn’s GroupKfold function, the training data were divided randomly into four groups such that data from the same case were not included in the same group. Then, using Scikit-Learn’s GridSearchCV function the hyperparameters were adjusted by a grid search using cross-validation [ 29 ]. Hyperparameters control model performance and can be adjusted during model development to improve accuracy or address overfitting [ 30 ].

The coefficient of determination (R 2 ) and mean absolute error (MAE) were used in previous reports to evaluate the accuracy of the regression model [ 16 , 31 , 32 ].

where \( {y}_{i}\) , \( {f}_{i}\) , \( \stackrel{-}{y}\) , and n represent the measured value, predicted value, average of measured values, and number of samples, respectively. R 2 assumes a value of 1.0 or less, and the closer it is to 1.0, the higher is the accuracy. R 2 does not mean square, as can be seen from the above definition, and it can be negative if the accuracy is extremely low [ 32 ]. The closer the MAE is to 0, the better is the model. In this study, R 2 and MAE were calculated using the validation data to verify the accuracy of the model. In machine learning models, the smaller the difference in accuracy between the training data and validation data, the higher is the generalization performance. If the accuracy on the validation data is lower than that on the training data, the model is considered to be specialized for the training data and has low generalization performance; this condition is known as overfitting [ 29 ]. A learning curve shows the validation and training score of an estimator for varying numbers of training samples. It is a tool to determine the benefit derived from adding more training data and whether the estimator suffers more from underfitting or overfitting [ 33 , 34 ]. In this study, we evaluated the generalization performance by comparing the R 2 values on the training and validation data and by creating a learning curve.

We selected patients from the participants who were followed for more than two years and had a minimum of three eGFR measurements to examine the accuracy of the conventional prediction method using the eGFR decline rate. We calculated the eGFR decline rate at each time point using the SLOPE function (based on the least squares method) in Excel, referring to previous reports [ 35 , 36 , 37 ]. According to the guidelines of the Japanese Society for Dialysis Therapy, the number of days that the eGFR was estimated to be less than eight from each time point was used as the predicted value [ 38 ].

This study was approved by the Ethics Committee of Oita University, Faculty of Medicine (approval No. 2139: 2021). Additionally, because this is a retrospective study, the committee also approved the waiver of written informed consent and the adoption of the opt-out method. Information was disclosed on the website of the Department of Endocrinology, Metabolism, Rheumatology and Nephrology, Faculty of Medicine, Oita University.

Figure  1 shows the flowchart from patient enrollment to model evaluation. A total of 135 patients met the criteria. Hemodialysis cases were selected for this study, and patients who had previously started other renal replacement therapies were excluded: only one patient was on peritoneal dialysis and no patient underwent post renal transplant. The patient characteristics are listed in Table  2 . The median age at the induction of dialysis was 71 years, median observation period was 496 days, most common CKD etiology was DN, and median laboratory findings at induction were Cr 7.5 mg/dL and eGFR 6.3 ml/min/1.73 m 2 . The details of the exploratory data analysis and machine learning are provided in the supplemental materials. A total of 10,916 data series containing 35 items were obtained from all patients. The number of missing data points for the survey items was examined, and items with numerous missing data points, such as bicarbonate ions and glucose, were difficult to use for the analysis (Supplementary Table S1 ). For the survey items, the Pearson’s correlation coefficient exhibited a correlation between renal function-related items and time to RRT, with a positive correlation for eGFR and a negative correlation for blood urea nitrogen (BUN) and Cr (Supplementary Figures S1 – S2 , Supplementary Table S2 ). Calcium, phosphorus, urinary protein/creatinine ratio, and albumin exhibited extremely weak correlations with time to RRT. These items have been used in prediction equations such as KFRE and were expected to be useful in this study. Among the items, very strong correlations were observed for red blood cells (RBCs), hemoglobin (Hb), and hematocrit (Ht), which are related to anemia, and when the variance inflation factor (VIF) was examined, multicollinearity was suspected for these items (Supplementary Tables S3 – S4 ).

Data series with missing values were excluded in preprocessing. The items were reduced stepwise in the preliminary study because this method reduces the overall number of data series when more items are used. Finally, the highest accuracy was achieved using 3,026 data series containing 20 items (shown in Table  1 ). The training algorithm and cross-validation results for this dataset are summarized in Table  3 and the hyperparameters in Supplementary Table S5. In LASSO regression, the R 2 for cross-validation was 0.59 with moderate accuracy, and the R 2 difference between cross-validation and training was small compared to those of other algorithms. The MAE was 488, implying that the model had a mean error of 488 days. Conversely, ensemble models such as GBDT appeared to be overfitting, with large R 2 differences between cross-validation and training. The learning curves visually confirmed the overfitting, showing that LASSO converged to the same value for the training and validation data (Fig.  2 -c), whereas GBDT converged with a divergence between the two (Fig.  2 -f). Table  4 summarizes the results of validating the accuracy of the created models on the validation data. LASSO regression exhibited an R 2 of 0.60, which is as stable as that of the cross-validation results, while GBDT exhibited an R 2 of 0.51, which is lower than that of the cross-validation results. GBDT was overfitting, which may have reduced its accuracy on unknown data. Figure  3 shows a scatter plot of the relationship between the predictions and the measured values for the validation data. The larger the measured values, the larger was the deviation from the predictions.

LASSO is a linear regression model that uses L1 regularization. A linear regression model is represented by the prediction equation y = a 1   x   1   + a 2   x   2 +… + a n x n + b , where y , x , and a represent the objective variable (predicted value), explanatory variable (data items), and regression coefficient, respectively. LASSO can automatically select explanatory variables by adjusting the regression coefficients with L1 regularization. Regression coefficients for the LASSO regression are listed in Table  5 , with eGFR having the largest value and highest contribution to the prediction, and the coefficients for several items, such as RBC being 0, implying no contribution to the prediction.

After running the prediction, SHAplay Additive exPlanations (SHAP) can be used to visualize the impact of each item on the prediction [ 39 ]. The waterfall plot provides an explanation of the predicted results for each individual case. Figure  4 -a is a waterfall plot of a randomly selected case that specifically shows the output of predicted values by eGFR, BUN, and other inputs. Figure  4 -b is a summary plot (scatter plot) for all cases analyzed in this study, and Fig.  4 -c is a summary plot (bar chart) of the average SHAP absolute value for all cases. The summary plot also shows a high contribution of eGFR. In this way, SHAP is useful for interpreting predictions not only for individual cases but also for entire cases.

The small number of cases in this study did not allow for sufficient subgroup analysis by CKD etiology or stage. However, when reanalyzing only the data series with DN as the CKD etiology in the LASSO regression, the R 2 was 0.60 and MAE was 302 in 1096 data series, indicating a decrease in MAE. Elsewhere, when reanalyzed only in the KDIGO CKD heatmap high-risk data series with LASSO regression, the R 2 was 0.62 and MAE was 396 in the 3025 data series, indicating an increase in R 2 . In addition, we reduced the number of items with suspected multicollinearity and items with small contributions to prediction by referring to the correlation matrix, VIF, regression coefficient, and SHAP values described, and we reanalyzed them with 12 items (age, sex, height, weight, CKD etiology, Hb, albumin, sodium, potassium, chlorine, eGFR, and urinary protein to creatinine ratio.) LASSO regression had an R 2 of 0.59 and MAE of 407, indicating a slight decrease in accuracy.

For the conventional prediction method that uses the rate of eGFR decline, 97 patients met the criteria, with a total of 6,209 eGFR measurements. A scatter plot of the relationship between the predicted and measured values is shown in Fig.  5 . The predicted values tended to be larger than the measured values. The accuracy was R 2  = − 17.1 and MAE = 2466, indicating an extremely low prediction accuracy.

A notable feature of this study is that, unlike existing ESKD risk, we focused on predicting time to RRT using continuous values and created a moderately accurate prediction model. This model can predict the time to RRT (RRT start date) based on data obtained at a single time point, and therefore it can provide concrete information even for first-time patients and clearly indicate the need to start treatment. As conditions change over time after intervention, the repeated prediction model can be used to predict the RRT start date each time. If the predicted RRT start date is extended, the patient will realize the benefits of treatment and will be more motivated to continue treatment. Even in the unfortunate case that the predicted RRT start date is moved up, it may be helpful to identify the reason for this change. Planned dialysis induction has a better prognosis [ 40 ]; if machine learning can predict the RRT start date, it will enable planned therapy selection and access construction. In addition, because the start of dialysis has a significant impact on a patient’s life, predicting the RRT start date is useful for the patient’s own life planning. The prediction of time to RRT based on regression may be a more patient-oriented outcome than the prediction of the risk of ESKD based on classification. For other progressive diseases, e.g., in the case of malignant tumors, prognosis and treatment efficacy are discussed in terms of a five-year survival rate (risk) in the early stages; however, the life expectancy (time) is often considered in advanced stages or when treatment is difficult. Numerous methods have been reported to predict life expectancy in days rather than risk [ 41 , 42 ]. In the case of CKD, life can be maintained with RRT; however, the lifestyle needs to be changed drastically. Therefore, the argument of predicting time to renal death in CKD may be useful, at least in the cases of high risk for ESKD. The number of elderly CKD patients with complications has increased in recent years, causing the concept of conservative kidney management to emerge [ 43 ]. The indication for renal biopsy, immunosuppressive therapy, and RRT should be determined based on the prognosis for time to renal death and complications, and in this regard, prediction of time to RRT is important. Some people become depressed when they are told how long it will take to reach RRT, and therefore care must be taken in actual use. However, it is expected to be a useful tool to realize better treatments of CKD when used effectively.

On the other hand, the accuracy of the conventional prediction method using the eGFR decline rate was extremely low. As illustrated in Fig.  3 -b, the predicted values tended to be larger than the measured values. The difference between the measured and predicted values would be small if the eGFR decline rate is constant during the observation period; however, the predicted values would be larger than the measured values if the eGFR decline rate increased with time. For instance, in diabetic nephropathy, which was the most common etiology in this study, the eGFR decline rate increased after the appearance of a urinary protein, as shown in Supplementary Figure S3 [ 44 , 45 ]. In this case, the eGFR decline rate was small in the initial stage, and the time to RRT was predicted to be long; however, if the eGFR decline rate increased during the course, the time to RRT became shorter than that during the initial prediction. Considering a nonlinear approximate curve instead of a linear regression may be necessary; however, it is difficult to apply a constant rule because the rate of decline varies for each case. The method using the decline rate of the reciprocal Cr cannot be used unless it is limited to patients with advanced CKD whose renal function worsens in a linear manner. Even in that case, this method suffers from an error of approximately one year [ 6 ]. Although time series information on renal function is important, it seems difficult to predict time to RRT based on the eGFR decline rate alone.

In this study, 10,916 data points were extracted from 135 cases and analyzed, assuming that the data were independent; however, data from the same cases are not completely independent and data bias is likely to occur. Data used in machine learning models should be independent and identically distributed for training and validation [ 46 ]. When using multiple time series data from the same case, as in this study, it is necessary to ensure that data from the same case do not leak into both the training and validation groups. In this study, data from the same cases were grouped together to address leakage, but other methods may also be useful, like dividing data by the time axis [ 47 , 48 ]. The learning models used in this study were linear models: linear regression, ridge regression, LASSO regression, and elastic net, and nonlinear models: random forest and GBDT. In general, the use of nonlinear models is expected to increase accuracy when the linear regression model does not adequately represent the characteristics of the data. With reference to the cheat sheet in Scikit-Learn, the aforementioned models were used in this study. However, in this case, the nonlinear model tended to overfit even after hyperparameter adjustment and cross-validation, and its accuracy on validation data was low. By contrast, the LASSO model, a linear model, exhibited low overfitting tendency and stable accuracy on validation data. Overfitting countermeasures include increasing the training data, simplifying the model by adjusting hyperparameters, cross-validation, regularization, and bootstrapping [ 49 , 50 , 51 ]. LASSO is an algorithm that can reduce the number of variables through regularization, which may have led to stable results. The small number of cases and the relatively large number of feature variables may have caused the nonlinear model to overfit [ 34 ]. However, in this study, the LASSO model has a limitation in that regularization prevents the incorporation of useful information such as the CKD etiology. In fact, subgroup analysis in DN has improved accuracy, and if the number of cases is increased and analyzed by etiology in future, the overall accuracy of the model may be improved. Other general machine learning issues include multicollinearity, which is often a problem when using multiple regression models in statistics [ 52 , 53 ]. Multicollinearity is a problem in which the predictors are correlated, and creating a model with multicollinear items makes it impossible to estimate the contribution by the regression coefficient. Multicollinearity is estimated using correlation matrices and VIF, and it is common to remove multicollinear items from the model. However, in machine learning, methods such as regularization and principal component analysis have been used to reduce the effect of multicollinearity, and good results have been obtained [ 54 ]. In this study, multicollinearity was suspected in anemia-related items, and variables were selected by regularization using LASSO regression. The accuracy decreased slightly when we used the conventional method of removing items with multicollinearity. In addition, there is concern that machine learning creates complex predictive models that cannot be interpreted easily by humans, making it a black box. The explainability of predictive models and the interpretability of predictions are especially important when used in medical applications where decisions can be life-threatening [ 55 ]. LASSO regression uses the regression coefficients to explain the contribution of the explanatory variables in the model. In the present model, eGFR contributes the most, indicating that the model reflects the results of eGFR to a large extent, which is a reasonable result. In addition, after running the predictions, SHAP can be used to visualize the actual contribution of the explanatory variables to the predictions [ 39 , 55 , 56 ]. In the case shown in Fig.  4 -a, it is understandable that inputs such as eGFR and BUN had a significant impact on the number of days predicted. When used in actual practice, the machine learning model can be applied to display the predicted number of days to RRT and the contribution of these items. Although the model is not designed for causal inference in this case, it could be clinically useful to examine the items that the AI focused on to make its predictions.

Limitations of this study include the fact that it was a single-center study, the regional nature of the cases, the small number of cases compared to previous studies, and the concern about data bias due to this. In addition, although only hemodialysis cases were included in this study, if peritoneal dialysis and renal transplantation cases can be added to the analysis in the future, it may be possible to apply this study to more general CKD cases. Owing to the small sample size, it was difficult to adequately analyze subgroups by etiology, stage, blood pressure, and glucose control status of CKD. The etiology and high-risk cases of CKD may have different modes of renal function deterioration, and in this study, increased accuracy was obtained in the analysis of DN and high-risk cases. Reanalysis with more cases is desirable in future. In addition, it was difficult to extract drug data at each time point. Validation on an external cohort is also an issue for the future. Another limitation of the machine learning model in this study is that, unlike the eGFR time series method, it does not incorporate time series information. The model calculates the same predicted value if the clinical data at the time of prediction are similar, regardless of whether renal function has been stable for several years or has deteriorated rapidly. Presently, it is safe to use this model in combination with the eGFR time series method, and the incorporation of time series information is an issue for the future. However, there have been no similar reports to date. We believe that this study is useful because it represents a potential new patient-based outcome.

A moderately accurate prediction model was developed by using a machine learning regression model to predict time to RRT with continuous values using data obtained at a single time point. This approach yielded better results than the conventional prediction method that uses eGFR time series data. The ability to specify the time to RRT is useful not only for medical practitioners to make treatment decisions but also for patients to motivate themselves to undergo treatment and for life planning.

Flowchart from patient enrollment to model evaluation

Learning curves for each algorithm.( a ) Linear regression, ( b ) Ridge regression, ( c ) LASSO regression, ( d ) Elastic net, ( e ) Random forest, and ( f ) GBDT. LASSO regression shows that the accuracy of the training data and the validation data converge to a close value as the number of samples increases. On the other hand, in the GBDT, the accuracy of the training data and the validation data remain divergent even as the number of samples increases

Relationship between predicted and measured values for each algorithm. ( a ) Linear regression, ( b ) Ridge regression, ( c ) LASSO regression, ( d ) Elastic net, ( e ) Random forest, and ( f ) GBDT. For all algorithms, the error in the predictions tended to increase as the measured values increased

SHAP value( a ) Waterfall plot of a randomly selected case. The bottom of the waterfall plot begins with the expected value of the model output, each row shows the positive (red) or negative (blue) contribution of each item, and the top shows the final output. This patient already had severe renal dysfunction, and the expected value was 1,234.956; however, the final output was 401.096 because there were changes such as a standardized eGFR of − 695.53 and a standardized BUN of − 139.3. ( b ) Summary plot (scatter plot) of all cases analyzed in this study. As shown at the top of the plot, the larger and redder the eGFR feature, the larger is the positive contribution (SHAP value) to prediction, which indicates a positive correlation. ( c ) Summary plot (bar chart) of the average absolute SHAP values for all cases, which shows that eGFR had the greatest influence

Relationship between predicted and measured values for the conventional eGFR time series method.Predictions tended to be very large, with a maximum predicted value of 73,549, which is considered an outlier, and for readability, the Y-axis is shown with a range up to 50,000

Data availability

The data underlying this article can be shared upon reasonable request to the corresponding author.

Abbreviations

Chronic kidney disease, KDIGO:Kidney Disease Improving Global Outcomes

End-stage kidney disease

Kidney failure risk equation

• Renal replacement therapy

Estimated glomerular filtration rate

• Artificial intelligence

Acute kidney injury

Total lymphocyte count

Choline esterase

Least absolute shrinkage and selection operator

Gradient boosting decision tree

Coefficient of determination

Mean absolute error

Blood urea nitrogen

Variance inflation factor

Red blood cell

SHAplay Additive exPlanations

Mean corpuscular volume

Mean corpuscular hemoglobin concentration

Hemoglobin A1c

Triglyceride

High-density lipoprotein cholesterol

Low-density lipoprotein cholesterol

Unsaturated iron binding capacity

Intact parathyroid hormone

Hydrogen carbonate ion

Urinary occult blood

Urinary protein to creatinine ratio

Estimated urinary salt excretion

Diabetic nephropathy

Nephrosclerosis

Chronic glomerulonephritis

Polycystic kidney disease

Interquartile range

Fresenius Medical Care. Annual report; 2021.

GBD Chronic Kidney Disease Collaboration. Global, regional, and national burden of chronic kidney disease, 1990–2017: a systematic analysis for the global burden of Disease Study 2017. Lancet. 2020;395:709–33.

Article   Google Scholar  

Kidney Disease. Improving global outcomes (KDIGO) CKD Work Group. KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl. 2013;3:1–150.

Google Scholar  

Tangri N, Stevens LA, Griffith J, Tighiouart H, Djurdjev O, Naimark D, et al. A predictive model for progression of chronic kidney disease to kidney failure. JAMA. 2011;305:1553–9.

Article   CAS   PubMed   Google Scholar  

National Institute for Health and Care Excellence. Chronic kidney disease: Assessment and management (NICE Guideline 203).; 2021. http://www.nice.org.uk/guidance/ng203 . (Accessed October 20, 2023).

Szeto CC, Leung CB, Wong TYH, Chow KM, Wang AY, Lui SF, et al. Extrapolation of reciprocal creatinine plot is not reliable in predicting the onset of dialysis in patients with progressive renal insufficiency. J Intern Med. 2003;253:335–42.

Article   PubMed   Google Scholar  

Shah BV, Levey AS. Spontaneous changes in the rate of decline in reciprocal serum creatinine: errors in predicting the progression of renal disease from extrapolation of the slope. J Am Soc Nephrol. 1992;2:1186–91.

Dias R, Torkamani A. Artificial intelligence in clinical and genomic diagnostics. Genome Med. 2019;11:70.

Article   PubMed   PubMed Central   Google Scholar  

Rajpurkar P, Chen E, Banerjee O, Topol EJ. AI in health and medicine. Nat Med. 2022;28:31–8.

Liu PR, Lu L, Zhang JY, Huo TT, Liu SX, Ye ZW. Application of artificial intelligence in medicine: an overview. Curr Med Sci. 2021;41:1105–15.

Yuan Q, Chen K, Yu Y, Le NQK, Chua MCH. Prediction of anticancer peptides based on an ensemble model of deep learning and machine learning using ordinal positional encoding. Brief Bioinform. 2023;24:bbac630.

Kha Q-H, Le V-H, Hung TNK, Nguyen NTK, Le NQK. Development and validation of an explainable machine learning-based prediction model for drug-food interactions from chemical structures. Sens (Basel). 2023;23:3962.

Article   ADS   CAS   Google Scholar  

Ardila D, Kiraly AP, Bharadwaj S, Choi B, Reicher JJ, Peng L, et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nat Med. 2019;25:954–61.

Vagliano I, Chesnaye NC, Leopold JH, Jager KJ, Abu-Hanna A, Schut MC. Machine learning models for predicting acute kidney injury: a systematic review and critical appraisal. Clin Kidney J. 2022;15:2266–80.

Flechet M, Falini S, Bonetti C, Güiza F, Schetz M, Van den Berghe G, et al. Machine learning versus physicians’ prediction of acute kidney injury in critically ill adults: a prospective evaluation of the AKIpredictor. Crit Care. 2019;23:282.

Sanmarchi F, Fanconi C, Golinelli D, Gori D, Hernandez-Boussard T, Capodici A. Predict, diagnose, and treat chronic kidney disease with machine learning: a systematic literature review. J Nephrol. 2023;36:1101–17.

Bai Q, Su C, Tang W, Li Y. Machine learning to predict end stage kidney disease in chronic kidney disease. Sci Rep. 2022;12:8377.

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Zacharias HU, Altenbuchinger M, Schultheiss UT, Raffler J, Kotsis F, Ghasemi S, et al. A predictive model for progression of CKD to kidney failure based on routine laboratory tests. Am J Kidney Dis. 2022;79:217–230e1.

Su CT, Chang YP, Ku YT, Lin CM. Machine learning models for the prediction of renal failure in chronic kidney disease: a retrospective cohort study. Diagnostics (Basel). 2022;12:2454.

Segal Z, Kalifa D, Radinsky K, Ehrenberg B, Elad G, Maor G, et al. Machine learning algorithm for early detection of end-stage renal disease. BMC Nephrol. 2020;21:518.

Dai P, Chang W, Xin Z, Cheng H, Ouyang W, Luo A. Retrospective study on the influencing factors and prediction of hospitalization expenses for chronic renal failure in China based on random forest and lasso regression. Front Public Health. 2021;9:678276.

Bellocchio F, Lonati C, Ion Titapiccolo J, Nadal J, Meiselbach H, Schmid M, et al. Validation of a novel predictive algorithm for kidney failure in patients suffering from chronic kidney disease: the prognostic reasoning system for chronic kidney disease (Progres-CKD). Int J Environ Res Public Health. 2021;18:12649.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Guligowska A, Corsonello A, Pigłowska M, Roller-Wirnsberger R, Wirnsberger G, Ärnlöv J et al. Association between kidney function, nutritional status and anthropometric measures in older people: the screening for CKD among older people across Europe (SCOPE) study. BMC Geriatr. 2020;20;Suppl 1:366.

Gao F, Huang Z, Liang JJ, Kang Y, Ling Y, He Y, et al. Association of malnutrition with all-cause and cardiovascular mortality in patients with mild to severe chronic kidney disease undergoing coronary angiography: a large multicenter longitudinal study. Int Urol Nephrol. 2023;55:3225–36.

Zhang J, Xiao X, Wu Y, Yang J, Zou Y, Zhao Y, et al. Prognostic nutritional index as a predictor of diabetic nephropathy progression. Nutrients. 2022;14:3634.

Ikizler TA, Burrowes JD, Byham-Gray LD, Campbell KL, Carrero JJ, Chan W, et al. KDOQI Clinical practice guideline for nutrition in CKD: 2020 update. Am J Kidney Dis. 2020;76:1–S107.

Seiichi M, Enyu I, Masaru H, Yoshinari Y, Kimio T, Kosaku N, et al. Revised equations for estimated GFR from serum creatinine in Japan. Am J Kidney Dis. 2009;53:982–92.

Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: machine learning in Python. JMLR. 2011;12:2825–30.

MathSciNet   Google Scholar  

Basu S, Faghmous JH, Doupe P. Machine learning methods for precision medicine research designed to reduce health disparities: a structured tutorial. Ethn Dis. 2020;30:217–28.

Pérez-Enciso M, Zingaretti LM. A guide for using deep learning for complex trait genomic prediction. Genes (Basel). 2019;20:553.

Kvålseth TO. Cautionary note about R 2 . Am Stat. 1985;39:279–85.

Chicco D, Warrens MJ, Jurman G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Comput Sci. 2021;7:e623.

Scikit Learn. Validation curves: plotting scores to evaluate models. 2024. https://scikit-learn.org/stable/modules/learning_curve.html , (Accessed January 29, 2024).

Leist AK, Klee M, Kim JH, Rehkopf DH, Bordas SPA, Muniz-Terrera G, et al. Mapping of machine learning approaches for description, prediction, and causal inference in the social and health sciences. Sci Adv. 2022;8:eabk1942.

Nojima J, Meguro S, Ohkawa N, Furukoshi M, Kawai T, Itoh H. One-year eGFR decline rate is a good predictor of prognosis of renal failure in patients with type 2 diabetes. Proc Jpn Acad Ser B Phys Biol Sci. 2017;93:746–54.

Article   ADS   PubMed   PubMed Central   Google Scholar  

Skupien J, Warram JH, Smiles AM, Stanton RC, Krolewski AS. Patterns of estimated glomerular filtration rate decline leading to end-stage renal disease in type 1 diabetes. Diabetes Care. 2016;39:2262–9.

Grams ME, Sang Y, Ballew SH, Matsushita K, Astor BC, Carrero JJ, et al. Evaluating glomerular filtration rate slope as a surrogate end point for ESKD in clinical trials: an individual participant meta-analysis of observational data. J Am Soc Nephrol. 2019;30:1746–55.

Watanabe Y, Yamagata K, Nishi S, Hirakata H, Hanafusa N, Saito C, et al. Japanese society for dialysis therapy clinical guideline for hemodialysis initiation for maintenance hemodialysis. Ther Apher Dial. 2015;19(Suppl 1):93–107.

Lundberg SM, Lee SI. A unified approach to interpreting model predictions. NIPS. 2017.

Chesser AM, Baker LR. Temporary vascular access for first dialysis is common, undesirable and usually avoidable. Clin Nephrol. 1999;51:228–32.

CAS   PubMed   Google Scholar  

Simmons CPL, McMillan DC, McWilliams K, Sande TA, Fearon KC, Tuck S, et al. Prognostic tools in patients with advanced cancer: a systematic review. J Pain Symptom Manage. 2017;53:962–970e10.

Hui D, Paiva CE, Del Fabbro EG, Steer C, Naberhuis J, van de Wetering M, et al. Prognostication in advanced cancer: update and directions for future research. Support Care Cancer. 2019;27:1973–84.

Okada K, Tsuchiya K, Sakai K, Kuragano T, Uchida A, Tsuruya K, et al. Shared decision making for the initiation and continuation of dialysis: a proposal from the Japanese society for Dialysis Therapy. Ren Replace Ther. 2021;7:56.

Ioannou K. Diabetic nephropathy: is it always there? Assumptions, weaknesses and pitfalls in the diagnosis. Horm (Athens). 2017;16:351–61.

Mogensen CE, Christensen CK, Vittinghus E. The stages in diabetic renal disease. With emphasis on the stage of incipient diabetic nephropathy. Diabetes. 1983;32(Suppl 2):64–78.

He Y, Shen Z, Cui P. Towards non-I.I.D. image classification: a dataset and baselines. Pattern Recognit. 2021;110:107383.

Lee H-T, Cheon H-R, Lee S-H, Shim M, Hwang H-J. Risk of data leakage in estimating the diagnostic performance of a deep-learning-based computer-aided system for psychiatric disorders. Sci Rep. 2023;13(1):16633.

Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement. BMJ. 2015;350:g7594.

López OAM, López AM, Crossa J. Multivariate statistical machine learning methods for genomic prediction. Berlin/Heidelberg, Germany: Springer; 2022. Overfitting, model tuning, and evaluation of prediction performance;109–39.

Book   Google Scholar  

Kernbach JM, Staartjes VE. Foundations of machine learning-based clinical prediction modeling: part II-Generalization and overfitting. Acta Neurochir Suppl. 2022;134:15–21.

Gygi JP, Kleinstein SH, Guan L. Predictive overfitting in immunological applications: pitfalls and solutions. Hum Vaccin Immunother. 2023;19:2251830.

Vatcheva KP, Lee MJ, McCormick JB, Rahbar MH. Multicollinearity in regression analyses conducted in epidemiologic studies. Epidemiol (Sunnyvale). 2016;6:227.

Kim JH. Multicollinearity and misleading statistical results. Korean J Anesthesiol. 2019;72:558–69.

Chan JY-L, Leow SMH, Bea KT, Cheng WK, Phoong SW, Hong Z, et al. Mitigating the multicollinearity problem and its machine learning approach: a review. Mathematics. 2022;10:1283.

Petch J, Di S, Nelson W. Opening the black box: the promise and limitations of explainable machine learning in cardiology. Can J Cardiol. 2022;38:204–13.

Sathyan A, Weinberg AI, Cohen K. Interpretable AI for bio-medical applications. Complex Eng Syst. 2022;2:18.

Download references

Acknowledgements

The authors thank Tsuyoshi Shimomura, MD, PhD, Department of Medical Informatics, Faculty of Medicine, Oita University, Oita, Japan, for assistance in data collection and data processing, and Koji Hatano, MD, PhD, Department of Healthcare AI Data Science, Faculty of Medicine, Oita University, Oita, Japan, for their assistance in data analysis.

No funding was received for conducting this study.

Author information

Authors and affiliations.

Department of Endocrinology, Metabolism, Rheumatology and Nephrology, Faculty of Medicine, Oita University, 8795593, 1-1 idaigaoka Hasama-cho, Yufu-shi, Oita-ken, Japan

Jun Okita, Takeshi Nakata, Hiroki Uchida, Akiko Kudo, Akihiro Fukuda & Hirotaka Shibata

Department of Medical Technology and Sciences, School of Health Sciences at Fukuoka, International University of Health and Welfare, Okawa, Japan

Department of Biophysics, Faculty of Medicine, Oita University, Oita, Japan

Masato Tanigawa

Department of Healthcare AI Data Science, Faculty of Medicine, Oita University, Oita, Japan

Noboru Sato

You can also search for this author in PubMed   Google Scholar

Contributions

JO conceptualized and designed the study, acquired data, analyzed and interpreted the data, drafted the manuscript, performed the statistical analysis, and supervised the study. TN conceptualized and designed the study, acquired data, analyzed and interpreted the data, drafted the manuscript, performed the statistical analysis, and supervised the study. TU and MT acquired data and analyzed and interpreted the data. NS acquired data and analyzed and interpreted the data. AF and HS supervised the study. All authors critically reviewed the important intellectual concepts, read, and approved the final manuscript.

Corresponding author

Correspondence to Takeshi Nakata .

Ethics declarations

Ethical approval.

This study was approved by the Ethics Committee of Oita University, Faculty of Medicine (approval No. 2139: 2021). Additionally, because this is a retrospective study, the committee also approved the waiver of written informed consent and the adoption of the opt-out method. Information was disclosed on the website of the Department of Endocrinology, Metabolism, Rheumatology and Nephrology, Fuculty of Medicine, Oita University.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Okita, J., Nakata, T., Uchida, H. et al. Development and validation of a machine learning model to predict time to renal replacement therapy in patients with chronic kidney disease. BMC Nephrol 25 , 101 (2024). https://doi.org/10.1186/s12882-024-03527-9

Download citation

Received : 13 December 2023

Accepted : 28 February 2024

Published : 16 March 2024

DOI : https://doi.org/10.1186/s12882-024-03527-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Chronic kidney disease
• Machine learning
• Predictive model

BMC Nephrology

ISSN: 1471-2369

• Submission enquiries: [email protected]
• General enquiries: [email protected]

You are using an outdated browser. Upgrade your browser today or install Google Chrome Frame to better experience this site.

• Section 4 - Scuba Diving: Decompression Illness & Other Dive-Related Injuries
• Section 4 - Mosquitoes, Ticks & Other Arthropods

High Elevation Travel & Altitude Illness

Cdc yellow book 2024.

Author(s): Peter Hackett, David Shlim

Acclimatization

Altitude illness, medications, preventing severe altitude illness or death.

Typical high-elevation travel destinations include Colorado ski resorts with lodgings at 8,000–10,000 ft (≈2,440–3,050 m); Cusco, Peru (11,000 ft; ≈3,350 m); La Paz, Bolivia (12,000 ft; ≈3,650 m); Lhasa, Tibet Autonomous Region (12,100 ft; ≈3,700 m); Everest base camp, Nepal (17,700 ft; ≈5,400 m); and Mount Kilimanjaro, Tanzania (19,341 ft; ≈5,900 m). High-elevation environments expose travelers to cold, low humidity, increased ultraviolet radiation, and decreased air pressure, all of which can cause health problems. The biggest concern, however, is hypoxia, due to the decreased partial pressure of oxygen (PO2). At 10,000 ft (≈3,050 m), for example, the inspired PO2 is only 69% of that at sea level; acute exposure to this reduced PO2 can lower arterial oxygen saturation to 88%–91%.

The magnitude and consequences of hypoxic stress depend on the elevation, rate of ascent, and duration of exposure; host genetic factors may also contribute. Hypoxemia is greatest during sleep; day trips to high-elevation destinations with an evening return to a lower elevation are much less stressful on the body. Because of the key role of ventilation, travelers must avoid taking respiratory depressants at high elevations.

The human body can adjust to moderate hypoxia at elevations ≤17,000 ft (≈5,200 m) but requires time to do so. Some acclimatization to high elevation continues for weeks to months, but the acute process, which occurs over the first 3–5 days following ascent, is crucial for travelers. The acute phase is associated with a steady increase in ventilation, improved oxygenation, and changes in cerebral blood flow. Increased red cell production does not play a role in acute acclimatization, although a decrease in plasma volume over the first few days does increase hemoglobin concentration.

Altitude illness can develop before the acute acclimatization process is complete, but not afterwards. In addition to preventing altitude illness, acclimatization improves sleep, increases comfort and sense of well-being, and improves submaximal endurance; maximal exercise performance at high elevation will always be reduced compared to that at low elevation.

Travelers can optimize acclimatization by adjusting their itineraries to avoid going “too high too fast” (see  Box 4-08 ). Gradually ascending to elevation or staging the ascent provides crucial time for the body to adjust. For example, acclimatizing for a minimum of 2–3 nights at 8,000–9,000 ft (≈2,450–≈2,750 m) before proceeding to a higher elevation is markedly protective against acute mountain sickness (AMS). The Wilderness Medical Society recommends avoiding ascent to a sleeping elevation of ≥9,000 ft (≈2,750 m) in a single day; ascending at a rate of no greater than 1,650 ft (≈500 m) per night in sleeping elevation once above 9,800 ft (≈3,000 m); and allowing an extra night to acclimatize for every 3,300 ft (≈1,000 m) of sleeping elevation gain. These reasonable recommendations can still be too fast for some travelers and annoyingly slow for others.

Box 4-08 Acclimatization tips: a checklist for travelers

☐ Ascend gradually. ☐ Avoid going directly from low elevation to >9,000 ft (2,750 m) sleeping elevation in 1 day. ☐ Once above 9,000 ft (≈2,750 m), move sleeping elevation by no more than 1,600 ft (≈500 m) per day, and plan an extra day for acclimatization every 3,300 ft (≈1,000 m). ☐ Consider using acetazolamide to speed acclimatization if abrupt ascent is unavoidable. ☐ Avoid alcohol for the first 48 hours at elevation. ☐ If a regular caffeine user, continue using to avoid a withdrawal headache that could be confused with an altitude headache. ☐ Participate in only mild exercise for the first 48 hours at elevation. ☐ A high-elevation exposure (> 9,000 ft [≈2,750 m]) for ≥2 nights, within 30 days before the trip, is useful, but closer to the trip departure is better.

Risk to Travelers

Susceptibility and resistance to altitude illness are, in part, genetically determined traits, but there are no simple screening tests to predict risk. Training or physical fitness do not affect risk. A traveler’s sex plays a minimal role, if any, in determining predisposition. Children are as susceptible as adults; people aged >50 years have slightly less risk. Any unacclimatized traveler proceeding to a sleeping elevation of ≥8,000 ft (≈2,450 m)—and sometimes lower—is at risk for altitude illness. In addition, travelers who have successfully adjusted to one elevation are at risk when moving to higher sleeping elevations, especially if the elevation gain is >2,000–3,000 ft (600–900 m).

How a traveler previously responded to high elevations is the most reliable guide for future trips, but only if the elevation and rate of ascent are similar, and even then, this is not an infallible predictor. In addition to underlying, inherent baseline susceptibilities, a traveler’s risk for developing altitude illness is influenced by 3 main factors: elevation at destination, rate of ascent, and exertion ( Table 4-04 ). Creating an itinerary to avoid any occurrence of altitude illness is difficult because of variations in individual susceptibility, as well as in starting points and terrain. The goal for the traveler might not be to avoid all symptoms of altitude illness but to have no more than mild illness, thereby avoiding itinerary changes or the need for medical assistance or evacuation.

Table 4-04 Risk categories for developing acute mountain sickness (AMS)

RISK CATEGORY

DESCRIPTION

PROPHYLAXIS RECOMMENDATIONS

• People with no prior history of altitude illness ascending to <9,000 ft (2,750 m)
• People taking ≥2 days to arrive at 8,200–9,800 ft (≈2,500–3,000 m), with subsequent increases in sleeping elevation <1,600 ft (≈500 m) per day, and an extra day for acclimatization every 3,300 ft (1,000 m) increase in elevation

Acetazolamide prophylaxis generally not indicated

• People with prior history of AMS and ascending to 8,200–9,200 ft (≈2,500–2,800 m) elevation (or above) in 1 day
• People with no history of AMS ascending to >9,200 ft (2,800 m) elevation in 1 day
• All people ascending >1,600 ft (≈500 m) per day (increase in sleeping elevation) at elevations >9,900 ft (3,000 m), but with an extra day for acclimatization every 3,300 ft (1,000 m)

Acetazolamide prophylaxis would be beneficial and should be considered

• People with a history of AMS ascending to >9,200 ft (≈2,800 m) in 1 day
• All people with a prior history of HAPE or HACE
• All people ascending to >11,400 ft (≈3,500 m) in 1 day
• All people ascending >1,600 ft (≈500 m) per day (increase in sleeping elevation) at elevations >9,800 ft (≈3,000 m), without extra days for acclimatization
• People making very rapid ascents (e.g., <7-day ascent of Mount Kilimanjaro)

Acetazolamide prophylaxis strongly recommended

Abbreviations: HACE, high-altitude cerebral edema; HAPE, high-altitude pulmonary edema

Destinations of Risk

Some common high-elevation destinations require rapid ascent by a non-pressurized airplane to >11,000 ft (≈3,400 m), placing travelers in a high-risk category for AMS. A common travel medicine question is whether to recommend acetazolamide for travelers when gradual or staged acclimatization is not feasible. With rates of altitude illness approaching 30%–40% in these situations, a low threshold for chemoprophylaxis is advised. In some cases (e.g., Cusco and La Paz), travelers can descend to elevations much lower than the airport to sleep for 1–2 nights and then begin their ascent, perhaps obviating the need for medication.

Itineraries along some trekking routes in Nepal, particularly Everest base camps, push the limits of many people’s ability to acclimatize. Even on standard schedules, incidence of altitude illness can approach 30% at the higher elevations. Whenever possible, adding extra days to the trek can make for a more enjoyable and safer climb.

Altitude Illness Syndromes

Altitude illness is divided into 3 syndromes: acute mountain sickness (AMS), high-altitude cerebral edema (HACE), and high-altitude pulmonary edema (HAPE). Some clinicians consider high-altitude headache a separate entity because isolated headache can occur without the combined symptoms that define AMS.

Acute Mountain Sickness

AMS is the most common form of altitude illness, affecting 25% of all visitors sleeping at elevations >8,000 ft (≈2,450 m) in Colorado.

Diagnosis of AMS is based on a history of recent ascent to high elevation and the presence of subjective symptoms. AMS symptoms are like those of an alcohol hangover; headache is the cardinal symptom, usually accompanied by ≥1 of the following: anorexia, dizziness, fatigue, nausea, or, occasionally, vomiting. Uncommonly, AMS presents without headache. Symptom onset is usually 2–12 hours after initial arrival at a high elevation or after ascent to a higher elevation, and often during or after the first night. Preverbal children with AMS can develop loss of appetite, irritability, and pallor. AMS generally resolves within 12–48 hours if travelers do not ascend farther.

The condition is typically self-limited, developing and resolving over 1–3 days. Symptoms starting after 3 days of arrival to high elevation and without further ascent should not be attributed to AMS. AMS has no characteristic physical findings; pulse oximetry is usually within the normal range for the elevation, or slightly lower than normal.

The differential diagnosis of AMS is broad; common considerations include alcohol hangover, carbon monoxide poisoning, dehydration, drug intoxication, exhaustion, hyponatremia, and migraine. Travelers with AMS will improve rapidly with descent ≥1,000 ft (≈300 m), and this can be a useful indication of a diagnosis of AMS.

Although rarely available, supplemental oxygen at 1–2 liters per minute will relieve headaches within about 30 minutes and resolve other AMS symptoms over hours. The popular small, handheld cans of compressed oxygen can provide brief relief, but contain too little oxygen (5 liters at most) for sustained improvement. Travelers with AMS but without HACE or HAPE (both described below) can remain safely at their current elevation and self-treat with non-opiate analgesics (e.g., ibuprofen 600 mg or acetaminophen 500 mg every 8 hours) and antiemetics (e.g., ondansetron 4 mg orally disintegrating tablets).

Acetazolamide speeds acclimatization and resolves AMS, but is more commonly used and better validated for use as prophylaxis. Dexamethasone is more effective than acetazolamide at rapidly relieving the symptoms of moderate to severe AMS. If symptoms worsen while the traveler is at the same elevation, or despite supplemental oxygen or medication, descent is mandatory.

High-Altitude Cerebral Edema

As an encephalopathy, HACE is considered “end stage” AMS. Fortunately, HACE is rare, especially at elevations <14,000 ft (≈4,300 m). HACE is often a secondary consequence of the severe hypoxemia that occurs with HAPE.

Unlike AMS, HACE presents with neurological findings, particularly altered mental status, ataxia, confusion, and drowsiness, similar to alcohol intoxication. Focal neurologic findings and seizures are rare in HACE; their presence should lead to suspicion of an intracranial lesion, a seizure disorder, or hyponatremia. Other considerations for the differential diagnosis include carbon monoxide poisoning, drug intoxication, hypoglycemia, hypothermia, and stroke. Coma can ensue within 24 hours of onset.

In populated areas with access to medical care, HACE can be treated with supplemental oxygen and dexamethasone. In remote areas, initiate descent for anyone suspected of having HACE, in conjunction with dexamethasone and oxygen, if available. If descent is not feasible, supplemental oxygen or a portable hyperbaric device, in addition to dexamethasone, can be lifesaving. Coma is likely to ensue within 12–24 hours of the onset of ataxia in the absence of treatment or descent.

High-Altitude Pulmonary Edema

HAPE can occur by itself or in conjunction with AMS and HACE; incidence is roughly 1 per 10,000 skiers in Colorado, and ≤1 per 100 climbers at >14,000 ft (≈4,300 m).

Early diagnosis is key; HAPE can be more rapidly fatal than HACE. Initial symptoms include chest congestion, cough, exaggerated dyspnea on exertion, and decreased exercise performance. If unrecognized and untreated, HAPE progresses to dyspnea at rest and frank respiratory distress, often with bloody sputum. This typical progression over 1–2 days is easily recognizable as HAPE, but the condition sometimes presents only as central nervous system dysfunction, with confusion and drowsiness.

Rales are detectable in most victims. Pulse oximetry can aid in making the diagnosis; oxygen saturation levels will be at least 10 points lower in HAPE patients than in healthy people at the same elevation. Oxygen saturation values of 50%–70% are common. The differential diagnosis for HAPE includes bronchospasm, myocardial infarction, pneumonia, and pulmonary embolism.

In most circumstances, descent is urgent and mandatory. Administer oxygen, if available, and exert the patient as little as possible. If immediate descent is not an option, use of supplemental oxygen or a portable hyperbaric chamber is critical.

Patients with mild HAPE who have access to oxygen (e.g., at a hospital or high-elevation medical clinic) might not need to descend to a lower elevation and can be treated with oxygen over 2–4 days at the current elevation. In field settings, where resources are limited and there is a lower margin for error, nifedipine can be used as an adjunct to descent, oxygen, or portable hyperbaric oxygen therapy. A phosphodiesterase inhibitor can be used if nifedipine is not available, but concurrent use of multiple pulmonary vasodilators is not recommended. Descent and oxygen are much more effective treatments than medication.

Recommendations for use and dosages of medications to prevent and treat altitude illness are outlined in  Table 4-05 .

Table 4-05 Recommended medication dosing to prevent & treat altitude illness

Abbreviations: AMS, acute mountain sickness; HACE, high-altitude cerebral edema; HAPE, high-altitude pulmonary edema; IM, intramuscular; IV, intravenous; PO, by mouth; SR, sustained release. 1 This dose can also be used as an adjunct to dexamethasone for HACE treatment; dexamethasone remains the primary treatment for HACE. 2 Use only in conjunction with oral medications and not as monotherapy for HAPE prevention.

Acetazolamide

Mechanism of action.

When taken preventively, acetazolamide hastens acclimatization to high-elevation hypoxia, thereby reducing occurrence and severity of AMS. It also enhances recovery if taken after symptoms have developed. The drug works primarily by inducing a bicarbonate diuresis and metabolic acidosis, which stimulates ventilation and increases alveolar and arterial oxygenation. By using acetazolamide, high-elevation ventilatory acclimatization that normally takes 3–5 days takes only 1 day. Acetazolamide also eliminates central sleep apnea, or periodic breathing, which is common at high elevations, even in those without a history of sleep disorder breathing.

An effective dose for prophylaxis that minimizes the common side effects of increased urination and paresthesia of the fingers and toes is 125 mg every 12 hours, beginning the day before ascent and continuing the first 2 days at elevation, and longer if ascent continues. Acetazolamide can also be taken episodically for symptoms of AMS, as needed. To date, the only dose studied for treatment is 250 mg (2 doses taken 8 hours apart), although the lower dosage used for prevention has anecdotally been successful. The pediatric dose is 5 mg/kg/day in divided doses, up to 125 mg, twice a day.

Adverse & Allergic Reactions

 Allergic reactions to acetazolamide are uncommon. Since acetazolamide is a sulfonamide derivative, cross-sensitivity between acetazolamide, sulfonamides, and other sulfonamide derivatives is possible.

Dexamethasone

Dexamethasone is effective for preventing and treating AMS and HACE and might prevent HAPE as well. Unlike acetazolamide, if the drug is discontinued at elevation before acclimatization, mild rebound can occur. Acetazolamide is preferable to prevent AMS while ascending, and dexamethasone generally should be reserved for treatment, usually as an adjunct to descent. The adult dose is 4 mg every 6 hours; rarely is it needed for more than 1–2 days. An increasing trend is to use dexamethasone for “summit day” on high peaks (e.g., Aconcagua and Kilimanjaro) to prevent abrupt altitude illness.

Recent studies have shown that taking ibuprofen 600 mg every 8 hours helps prevent AMS, although not quite as effectively as acetazolamide. Ibuprofen is, however, available over the counter, inexpensive, and well tolerated.

Nifedipine both prevents and ameliorates HAPE. For prevention, nifedipine is generally reserved for people who are particularly susceptible to the condition. The adult dose for prevention or treatment is 30 mg of extended release every 12 hours, or 20 mg every 8 hours.

Phosphodiesterase-5 Inhibitors

Phosphodiesterase-5 inhibitors selectively lower pulmonary artery pressure, with less effect on systemic blood pressure than nifedipine. Tadalafil, 10 mg taken twice a day during ascent, can prevent HAPE. It is also being studied as a possible treatment.

The main point of instructing travelers about altitude illness is not to eliminate the possibility of mild illness but to prevent death or evacuation. Because the onset of symptoms and the clinical course are sufficiently slow and predictable, there is no reason for anyone to die from altitude illness unless they are trapped by weather or geography in situations where descent is impossible. Travelers can adhere to 3 rules to help prevent death or serious consequences from altitude illness:

• Know the early symptoms of altitude illness and be willing to acknowledge when symptoms are present.
• Never ascend to sleep at a higher elevation when experiencing symptoms of altitude illness, no matter how minor the symptoms seem.
• Descend if the symptoms become worse while resting at the same elevation.

For trekking groups and expeditions going into remote high-elevation areas, where descent to a lower elevation could be problematic, a pressurization bag (e.g., the Gamow bag) can be beneficial. A foot pump produces an increased pressure of 2 lb/in2, mimicking a descent of 5,000–6,000 ft (≈1,500–1,800 m) depending on the starting elevation. The total packed weight of bag and pump is about 14 lb (6.5 kg).

Preexisting Medical Conditions

Travelers with preexisting medical conditions must optimize their treatment and have their conditions stable before departure. In addition, these travelers should have plans for dealing with exacerbation of their conditions at high elevations. Travelers with underlying medical conditions (e.g., coronary artery disease, any form of chronic pulmonary disease or preexisting hypoxemia, obstructive sleep apnea [OSA], or sickle cell trait)—even if well controlled—should consult a physician familiar with high-elevation medical issues before undertaking such travel ( Table 4-06 ).

Clinicians advising travelers should know that in most high-elevation resorts and cities, “home” oxygen is readily available. In North America, this requires a prescription that the traveler can carry, or oxygen can be arranged beforehand. Supplemental oxygen, whether continuous, episodic, or nocturnal, depending on the circumstances, is very effective at restoring oxygenation to low elevation values and eliminates the risk for altitude illness and exacerbation of preexisting medical conditions.

Table 4-06 Ascent risk associated with various underlying medical conditions & risk factors

LIKELY NO EXTRA RISK

CAUTION REQUIRED 1

ASCENT CONTRAINDICATED

• Asthma (well-controlled)
• Children and adolescents
• Chronic obstructive pulmonary disease (mild)
• Coronary artery disease (following revascularization)
• Diabetes mellitus
• Hypertension (controlled)
• Neoplastic diseases
• Obesity (Class 1/Class 2) 2
• Obstructive sleep apnea (mild/ moderate)
• Pregnancy (low-risk)
• Psychiatric disorders (stable)
• Seizure disorder (controlled)
• Angina (stable)
• Arrhythmias (poorly controlled)
• Chronic obstructive pulmonary disease (moderate)
• Coronary artery disease (nonrevascularized)
• Cystic fibrosis (FEV1 30%–50% predicted)
• Heart failure (compensated)
• Hypertension (poorly controlled)v Infants <6 weeks old
• Obesity (Class 3) 3
• Obstructive sleep apnea (severe)
• Pulmonary hypertension (mild)
• Radial keratotomy surgery
• Seizure disorder (poorly controlled)
• Sickle cell trait
• Angina (unstable)
• Asthma (unstable, poorly controlled)
• Cerebral space–occupying lesions
• Cerebral vascular aneurysms or arteriovenous malformations (untreated, high-risk)
• Chronic obstructive pulmonary disease (severe/very severe)
• Cystic fibrosis (FEV1 <30% predicted)
• Heart failure (decompensated)
• Myocardial infarction or stroke (<90 days before ascent)
• Pregnancy (high-risk)
• Pulmonary hypertension (pulmonary artery systolic pressure >60 mm Hg)
• Sickle cell anemia

Abbreviations: : FEV1, forced expiratory volume in 1 second

1 Travelers with these conditions most often require consultation with a physician experienced in high-altitude medicine and a comprehensive management plan.

2 Class 1 obesity: Body Mass Index (BMI) of 30 to <35; Class 2 obesity: BMI of 35 to <40

3 Class 3 obesity: BMI of ≥40.

Diabetes Mellitus

Travelers with diabetes can travel safely to high elevations, but they must be accustomed to exercise if participating in strenuous activities at elevation and carefully monitor their blood glucose. Diabetic ketoacidosis can be triggered by altitude illness and can be more difficult to treat in people taking acetazolamide. Not all glucose meters read accurately at high elevations.

Obstructive Sleep Apnea

Travelers with sleep disordered breathing who are planning high-elevation travel should receive acetazolamide. Those with mild to moderate OSA who are not hypoxic at home might do well without a continuous positive airway pressure (CPAP) device, while those with severe OSA should be advised to avoid high-elevation travel unless they receive supplemental oxygen in addition to their CPAP. Oral appliances for OSA can be useful adjuncts when electrical power is unavailable.

There are no studies or case reports describing fetal harm among people who briefly travel to high elevations during their pregnancy. Nevertheless, clinicians might be prudent to recommend that pregnant people do not stay at sleeping elevations >10,000 ft (≈3,050 m). Travel to high elevations during pregnancy warrants confirmation of good maternal health and verification of a low-risk gestation. Advise pregnant travelers of the dangers of having a pregnancy complication in remote, mountainous terrain.

Radial Keratotomy

Most people do not have visual problems at high elevations. At very high elevations, however, some people who have had radial keratotomy procedures might develop acute farsightedness and be unable to care for themselves. LASIK and other newer procedures may produce only minor visual disturbances at high elevations.

The following authors contributed to the previous version of this chapter: Peter H. Hackett, David R. Shlim

Bibliography

Bartsch P, Swenson ER. Acute high-altitude illnesses. N Engl J Med. 2013;369(17):1666–7. 

Hackett PH, Luks AM, Lawley JS, Roach RC. High-altitude medicine and pathophysiology. In: Auerbach PS, editor. Wilderness medicine, 7th edition. Philadelphia: Elsevier; 2017. pp. 8–28. 

Hackett PH, Roach RC. High altitude cerebral edema. High Alt Med Biol. 2004;5(2):136–46. 

Luks AM, Auerbach PS, Freer L, Grissom CK, Keyes LE, McIntosh SE, et al. Wilderness Medical Society clinical practice guidelines for the prevention and treatment of acute altitude illness: 2019 update. Wilderness Environ Med. 2019;30(4S):S3–18. 

Luks AM, Hackett PH. High altitude and preexisting medical conditions. In: Auerbach PS, editor. Wilderness medicine, 7th edition. Philadelphia: Elsevier; 2017. pp. 29–39. 

Luks AM, Hackett PH. Medical conditions and high-altitude travel. N Engl J Med. 2022;386(4):364–73. 

Luks AM, Swenson ER.Medication and dosage considerations in the prophylaxis and treatment of high-altitude illness. Chest. 2008;133(3):744–55. 

Meier D, Collet TH, Locatelli I, Cornuz J, Kayser B, Simel DL, Sartori C. Does this patient have acute mountain sickness? The rational clinical examination systematic review. JAMA. 2017;318(18):1810–19. 

Roach RC, Lawley JS, Hackett PH. High-altitude physiology. In: Auerbach PS, editor. Wilderness medicine, 7th edition. Philadelphia: Elsevier; 2017. pp. 2–8. 

Woolcott OO. The Lake Louise Acute Mountain Sickness score: still a headache. High Alt Med Biol. 2021;22(4):351–2.

File Formats Help:

• Adobe PDF file
• Microsoft PowerPoint file
• Microsoft Word file
• Microsoft Excel file
• Audio/Video file
• Apple Quicktime file
• RealPlayer file
• Zip Archive file

Exit Notification / Disclaimer Policy

• The Centers for Disease Control and Prevention (CDC) cannot attest to the accuracy of a non-federal website.
• Linking to a non-federal website does not constitute an endorsement by CDC or any of its employees of the sponsors or the information and products presented on the website.
• You will be subject to the destination website's privacy policy when you follow the link.
• CDC is not responsible for Section 508 compliance (accessibility) on other federal or private website.

An official website of the United States government

The .gov means it's official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you're on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• Browse Titles

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2024 Jan-.

StatPearls [Internet].

Anemia of chronic renal disease.

Hira Shaikh ; Muhammad F. Hashmi ; Narothama R. Aeddula .

Affiliations

Last Update: February 24, 2023 .

• Continuing Education Activity

Anemia of chronic renal disease, also known as anemia of chronic kidney disease (CKD), is a form of normocytic, normochromic, hypoproliferative anemia. It is frequently associated with poor outcomes in chronic kidney disease and confers an increased mortality risk. Therefore, treatment is directed toward improving renal function, when possible, and increasing red blood cell production. Erythropoiesis-stimulating agents and iron supplementation constitute the treatment of choice for anemia of chronic renal disease. This activity reviews the evaluation and management of the anemia of chronic renal disease. It highlights the role of the interprofessional team in the care of individuals affected by this condition.

• Explain the pathophysiology of anemia of chronic renal disease.
• Describe how to diagnose anemia of chronic renal disease.
• Review the management of the anemia of chronic renal disease.
• Summarize the interprofessional team strategies for improving coordination and communication to enhance the management of patients with anemia of chronic renal disease.
• Introduction

Anemia is generally defined as hemoglobin of less than 13.0 g/dL in men and less than 12.0 g/dL in premenopausal women. [1]  Anemia of chronic kidney disease (CKD) is a form of normocytic normochromic, hypoproliferative anemia. Among other complications of CKD, it is frequently associated with poor outcomes in CKD and increased mortality. [2] [3]

The disorder starts to develop when the glomerular filtration rate drops below 60 mg/ml. The anemia is rare when the GFR exceeds 80 mL/min/1.73 m2. However, as the GFR worsens, the anemia gets more severe.

Anemia is one of the common associations of CKD responsible for poor outcomes. The current management options for anemia in CKD are controversial, with some clinical trials indicating raised morbidity and mortality associated with erythropoiesis-stimulating agents. One hundred seventy years ago, anemia was linked to CKD for the first time by Richard Bright. [4]  With the progression of kidney disease, the prevalence of anemia increases affecting almost all patients with stage 5 CKD. Anemia of chronic renal disease often leads to declined quality of life and increased risk of cardiovascular diseases, cognitive impairment, hospitalizations, and mortality. [5]

Anemia of chronic renal disease is of multifactorial origin, the widely accepted etiology being decreased renal production of erythropoietin (EPO), the hormone responsible for stimulating red blood cell production. Decreased erythropoietin has recently been linked with the downregulation of hypoxia-inducible factor (HIF), a transcription factor that regulates gene expression of erythropoietin. [6] [7]  Other mechanisms include uremia (leading to RBC deformity responsible for hemolysis), folate and vitamin B12 deficiency, iron deficiency, bleeding due to dysfunctional platelets, and rarely blood loss from hemodialysis. [8]  

RBC fragmentation by injured renovascular endothelium in selected conditions such as glomerulopathy and malignant hypertension exacerbates the anemia, which explains why anemia can be particularly severe in renal glomerulopathies, including glomerulonephritis, diabetic nephropathy, for the degree of excretory failure.

• Epidemiology

The condition usually develops following a greater than 50 percent loss of kidney function, typically when the glomerular filtration rate (GFR) decreases to less than 60 mL/min/1.73 m2. [9]  The severity of anemia tends to worsen as chronic kidney disease (CKD) progresses. The deficiency in renal production of erythropoietin and the severity of anemia do not always tend to correlate with the severity of renal dysfunction. At least 90% of patients who end up on dialysis will eventually develop anemia of chronic disease.

Anemia of chronic renal disease is associated with a poor quality of life, worse renal survival, increased morbidity and mortality, and excessive healthcare costs. [10] [11] [12] [13] Several studies report the prevalence of anemia in non-dialysis dependent (NDD) CKD up to 60%.

Anemia becomes more prevalent and severe with a declining estimated glomerular filtration rate (eGFR). The National Health and Nutrition Examination Survey (NHANES) from 2007–2008 and 2009–2010 ( 7 ) observed that anemia was twice as prevalent in CKD patients as in the general population. [14]  Similar data were observed recently in the CKD Prognosis Consortium. [15]

• Pathophysiology

As discussed above, anemia of chronic renal disease is mainly secondary to EPO deficiency; however, numerous studies indicate that the following also contribute to the development of anemia in patients with chronic renal disease:

• Circulating uremia-induced erythropoiesis inhibitors lead to anemia, although this has been challenged in some studies as no specific inhibitors have been identified. [16]
• The shortened lifespan of red blood cells also contributes, as observed in radioisotope labeling studies. [17] [18]
• Although the underlying mechanism is not entirely understood, mechanical and metabolic factors have been proposed. [18]
• Nutritional deficiencies, such as vitamin B12 and folate, due to dialysate losses or anorexia are currently not very common due to the routine supplementation of nutrients in hemodialysis patients.
• Recent studies have uncovered an excessively recognized role of disordered iron homeostasis in anemia of chronic renal disease. Systemic iron content is maintained by regulating gastrointestinal iron absorption and its release from storage sites, such as reticuloendothelial macrophages and the liver. [19]
• CKD patients have exaggerated iron losses, accounting for around 1 to 3 g annually in hemodialysis patients, secondary to chronic bleeding due to uremia-associated platelet dysfunction, blood being trapped in the dialysis apparatus, and frequent phlebotomy. CKD patients are at significant risk of true iron deficiency; therefore, iron supplementation is part of the mainstay of treatment. Hemodialysis patients have impaired dietary iron absorption, which is why intravenous iron is a preferred treatment option. [5]
• In addition to true iron deficiency, CKD patients also have a functional iron deficiency, known as reticuloendothelial cell iron blockade. It is characterized by reduced iron release from body stores unable to meet the requirement for erythropoiesis.
• Hepcidin excess is the main factor behind impaired iron regulation and anemia of chronic renal disease as it affects dietary iron absorption and mobilization of iron from body stores.

In summary, anemia of chronic renal disease is a multifactorial process attributable to relative EPO deficiency, uremia-induced erythropoiesis inhibitors, the shortened lifespan of erythrocytes, and disordered iron homeostasis.

• History and Physical

The clinical presentation of anemia of chronic renal disease is not different from that of anemia due to other causes. Common symptoms include:

• Dyspnea (shortness of breath)
• Fatigue [20]
• Generalized weakness
• Decreased concentration
• Reduced exercise tolerance.

Commonly observable signs include:

• Skin and conjunctival pallor
• Respiratory distress
• Tachycardia
• Chest pain (mostly with severe anemia)
• Heart failure (usually with chronic severe anemia)

Common tests required to diagnose the condition include the following:

• Complete blood count (CBC) with differential
• Peripheral smear
• Iron indices (iron, ferritin, total iron binding capacity, transferrin saturation)
• Iron, vitamin B, and folate levels (included in initial workup to rule out other reversible causes of anemia)
• Thyroid function tests (rule out alternate etiology of hypoproliferative normocytic anemia)

Normocytic normochromic anemia and peripheral reticulocytopenia are observable on CBC with a peripheral smear. 

Unfortunately, due to high serum ferritin levels secondary to chronic inflammation in CKD, serum iron indices are not accurately indicative of the degree of iron deficiency in dialysis patients, thus raising the standard cutoffs of iron responsiveness. [21] [22]  The Dialysis Patients' Response to IV Iron With Elevated Ferritin (DRIVE) study demonstrated that intravenous iron is beneficial in dialysis patients even in the setting of ferritin as high as 1200 ng/mL if the transferrin saturation is less than 30%. [23]

Measuring serum erythropoietin levels are discouraged in CKD. They are not usable as an indicator of a renal source of the anemia because, in kidney disease, there is 'relative erythropoietin deficiency,' that is, an inappropriate rise in erythropoietin levels for the severity of anemia. [24] [25]

Bone marrow may show erythroid hypoplasia, which correlates to the reports of resistance of bone marrow to erythropoietin.

• Treatment / Management

Treatment of anemia of chronic renal disease is directed toward improving renal function (when possible) and increasing red blood cell production. Therefore, erythropoiesis-stimulating agents (ESAs), together with iron supplementation, is the treatment of choice in anemia of CKD.

Treatment of anemia in CKD has come a long way. Before the advanced treatment options available today, the main treatment option used to be blood transfusions, which came with numerous complications, including infections, hemosiderosis, fluid overload, transfusion reactions, etc. It started with the use of androgens in the 1970s to avoid transfusion in patients with CKD. [26] [27]  After that, in the 1980s, the development of recombinant EPO, followed by ESAs, revolutionized the management of anemia in CKD. [28]  Although initially instituted to avoid transfusions, they were soon known to have various positive effects, including improved survival and quality of life, improved cardiac function and mortality associated with it, lower hospitalizations, and lower costs. [29] [30] [31]  

Recombinant human erythropoietin and darbepoetin alfa are the two ESAs generally used in managing anemia in CKD. They are fairly similar in efficacy and side effect profile, except for the longer half-life of darbepoetin alfa, thus allowing for less frequent dosing. [32] [33]

As per KIDGO guidelines, in patients with CKD who are not on dialysis, ESAs are typically considered when hemoglobin level drops below 10 g/dl but are individualized depending on various factors, including symptoms related to anemia, dependence on transfusions, the rate of drop in hemoglobin concentration, and response to iron therapy. In these patients, erythropoietin (50 to 100 units/kg IV or SC) is usually given every 1 to 2 weeks, and darbepoetin alfa dosing is every 2 to 4 weeks.

In patients on dialysis, ESAs are usually avoided unless the hemoglobin level is between 9 and 10 g/dL. In this subset, erythropoietin is given with every dialysis, i.e., three times a week, whereas darbepoetin alfa is dosed once weekly.

Generally, the peak rise in RBCs in response to ESAs occurs at 8 to 12 weeks. However, in around 10% to 20% of cases, anemia can be resistant to ESAs. Common adverse effects of ESAs include seizures, the progression of hypertension, clotting of dialysis access, the progression of malignancy, and higher mortality in cancer patients. [34] [28]

In all patients with CKD, regardless of the need for dialysis, the goal hemoglobin using ESAs is less than 11.5 g/dL. Multiple trials were done to assess the superiority of target hemoglobin to 'high normal' versus lower range. These trials, including CHOIR, NHCT, and TREAT trials, demonstrated higher mortality, thrombosis, and adverse cerebrovascular and cardiovascular events due to higher levels of ESAs when used for target hemoglobin greater than 11 g/dl. [35] [36] [37]  These events are likely related to the effect of ESAs on vascular remodeling and causing vasoconstriction. [38]  CHOIR trial also showed that patients requiring higher levels of ESAs to achieve target hemoglobin had worse outcomes. [35]  The discovery of stated side effects of ESAs, when used to target high normal hemoglobin levels, raised questions about the benefits of ESAs besides avoidance of transfusions, which has led to growing interest in looking for alternative etiologies and, thus, management for anemia of CKD.

Patients with CKD have an increased risk of iron deficiency due to impaired dietary iron absorption, chronic bleeding due to platelet dysfunction from uremia, frequent phlebotomy, and blood being trapped in the dialysis apparatus. This deficiency, in addition to the depletion of the circulating iron pool by stimulation of erythropoiesis by ESAs, makes iron supplementation the core of the treatment of anemia in CKD. Due to decreased oral iron absorption, intravenous iron is preferable in hemodialysis patients. [39] [40]

KIDGO recommends target transferrin saturation between 20 to 30% and ferritin level 100 to 500 ng/mL in patients with CKD who are not on dialysis. In patients with ESRD on dialysis receiving intravenous iron, goal transferrin saturation of 30 to 50% and ferritin higher than 200 ng/mL. [41]  Iron correlates with acute toxicity and infection risk, which should be weighed against the benefits in individual patients.

Unlike the general population, high serum ferritin levels do not predict hemoglobin responsiveness in renal failure patients. Finally, no erythropoietin level can be considered adequate for defining renal anemia. Thus, erythropoietin levels should not be measured regularly in evaluating patients with renal anemia.

• Differential Diagnosis

The following is a comprehensive list of differential diagnoses that need to be considered when diagnosing anemia of chronic renal disease:

• Alcohol misuse disorder
• Aplastic anemia
• Dialysis complications
• Hypothyroidism
• Hyperthyroidism
• Methemoglobinemia
• Sickle cell anemia
• Systemic lupus erythematosus
• Hypoadrenalism
• Panhypopituitarism
• Primary and secondary hyperparathyroidism
• Myelophthisic anemia

Many patients with renal failure will not respond to erythropoietin, which is crucial as it is an important predictor of adverse cardiac events. Two factors that lead to unresponsiveness include iron deficiency and inflammation. High levels of CRP predict resistance to erythropoietin in dialysis patients. Therefore, to enhance responsiveness to erythropoietin, iron supplements are recommended.

Anemia of chronic renal disease is associated with cardiorenal anemia syndrome. Foley et al. observed that for every 1-g decrease in hemoglobin concentration, a 42% increase in left ventricular dilatation is seen in patients with stage 5 CKD. [42]  Cardiovascular disease remains the commonest cause of mortality in such patients, much greater than in the general population. [43]

The Dialysis Outcomes Practice Pattern Study (DOPPS), involving countries, reported that with the decrease in hemoglobin to less than 11 g/dL, there was an increase in hospitalization and mortality in CKD patients. [44]

• Complications

Anemia of renal disease is an independent risk factor for death. It has been shown to promote faster progression of left ventricular hypertrophy, peripheral oxygen demand, and worsening cardiac outcomes. More importantly, anemia of renal failure leads to depression, fatigue, stroke, reduced exercise tolerance, and an increased rate of re-admissions. [45]

Long-term treatment with erythropoietin can cause hypertension, vasoconstriction, and seizures.

• Deterrence and Patient Education

Patients should be given information that when they have chronic kidney disease, their kidneys cannot make enough erythropoietin which causes their red blood cells to drop and anemia. Most such patients develop anemia, which can happen early in the illness and worsen with time. 

Changes to diet can help prevent or manage anemia, and a dietician's help in this regard can play a significant role.

Patients should store ESA or iron as advised by the manufacturer, as some products need to be kept in the fridge. All CKD patients should be encouraged to let their providers know if they notice any bleeding or experience symptoms of anemia.

• Pearls and Other Issues
• Anemia of renal disease is common and is chiefly due to decreased erythropoietin production.
• Investigating other treatable causes of anemia in renal failure patients is necessary.
• Anemia of renal disease is associated with adverse cardiac events, heart failure, MI, and death.
• Erythropoietin levels are not indicative of anemia in renal failure patients. Therefore, one should target a hemoglobin level of no more than 11.5 g/dl.
• Enhancing Healthcare Team Outcomes

The management of the anemia of CKD is complex because it is not a simple matter of giving patients more blood transfusions or erythropoietin. Both these products have serious adverse effects when given chronically. One should never assume that anemia of renal disease is solely due to a lack of erythropoietin; it may be due to poor nutrition or chronic illness- so a thorough workup is essential to determine the cause.

Managing patients on dialysis with anemia requires an integrated approach by an interprofessional team consisting of the nephrologist, PCP including the nurse practitioner, physician assistant, physician, nursing, pharmacy, and occasionally also a hematologist may be necessary to help achieve the best possible outcomes. The dialysis nurse should monitor vital signs and obtain total blood counts to determine the level of anemia. In addition, pharmacists should educate the patient about the importance of iron supplements because, without iron, many patients develop resistance to erythropoietin. 

Further, anemia of renal disease is also associated with adverse cardiac outcomes, so the patient's cardiopulmonary status has to be monitored for life.

Finally, a nutritionist's involvement is essential for avoiding and/or treating the deficiency of vitamins that can exacerbate the anemia of kidney disease.

The outcomes for patients with anemia of renal disease are guarded. Many develop adverse cardiac events that lead to a high mortality rate. Too much iron from blood transfusions also affects outcomes. Finally, chronic use of erythropoietin has been associated with severe hypertension, stroke, and heart failure. An interprofessional team approach will maximize positive outcomes and minimize adverse events. [Level 5]

• Review Questions
• Access free multiple choice questions on this topic.
• Comment on this article.

Disclosure: Hira Shaikh declares no relevant financial relationships with ineligible companies.

Disclosure: Muhammad Hashmi declares no relevant financial relationships with ineligible companies.

Disclosure: Narothama Aeddula declares no relevant financial relationships with ineligible companies.

This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ), which permits others to distribute the work, provided that the article is not altered or used commercially. You are not required to obtain permission to distribute this article, provided that you credit the author and journal.

• Cite this Page Shaikh H, Hashmi MF, Aeddula NR. Anemia of Chronic Renal Disease. [Updated 2023 Feb 24]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2024 Jan-.

In this Page

Bulk download.

• Bulk download StatPearls data from FTP

Related information

• PMC PubMed Central citations
• PubMed Links to PubMed

Similar articles in PubMed

• Review Practical approach to the diagnosis and treatment of anemia associated with CKD in elderly. [J Am Med Dir Assoc. 2006] Review Practical approach to the diagnosis and treatment of anemia associated with CKD in elderly. Agarwal AK. J Am Med Dir Assoc. 2006 Nov; 7(9 Suppl):S7-S12; quiz S17-21.
• Review Epidemiology of chronic kidney disease and anemia. [J Am Med Dir Assoc. 2006] Review Epidemiology of chronic kidney disease and anemia. Robinson BE. J Am Med Dir Assoc. 2006 Nov; 7(9 Suppl):S3-6; quiz S17-21.
• Anemia and chronic kidney disease are potential risk factors for mortality in stroke patients: a historic cohort study. [BMC Nephrol. 2010] Anemia and chronic kidney disease are potential risk factors for mortality in stroke patients: a historic cohort study. Del Fabbro P, Luthi JC, Carrera E, Michel P, Burnier M, Burnand B. BMC Nephrol. 2010 Oct 16; 11:27. Epub 2010 Oct 16.
• Review Use of erythropoiesis-stimulating agents in patients with anemia of chronic kidney disease: overcoming the pharmacological and pharmacoeconomic limitations of existing therapies. [Mayo Clin Proc. 2007] Review Use of erythropoiesis-stimulating agents in patients with anemia of chronic kidney disease: overcoming the pharmacological and pharmacoeconomic limitations of existing therapies. Wish JB, Coyne DW. Mayo Clin Proc. 2007 Nov; 82(11):1371-80.
• Hemodialysis. [StatPearls. 2024] Hemodialysis. Murdeshwar HN, Anjum F. StatPearls. 2024 Jan

Recent Activity

• Anemia of Chronic Renal Disease - StatPearls Anemia of Chronic Renal Disease - StatPearls

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

Connect with NLM

National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894

Web Policies FOIA HHS Vulnerability Disclosure

Help Accessibility Careers