, Introduction
Cancer has emerged as a critical medical and social issue in the 21st century. The aging global population has led to a continuous increase in cancer incidence, with projections estimating approximately 35 million new cancer cases annually by 2050 [1]. Previous studies have demonstrated that patients with end-stage kidney disease (ESKD) undergoing hemodialysis (HD) are at a significantly higher risk of developing cancer compared with the general population [2–6]. Furthermore, cancer-specific mortality rates in patients with ESKD are 2–3 times higher than those in the general population [7,8].
While the incidence of ESKD has remained stable, its prevalence has increased owing to population aging and improved survival rates among patients with ESKD [9,10]. HD remains the most commonly used kidney replacement therapy in this population [9,10]. Since 2019, >50% of patients in South Korea undergoing dialysis are aged ≥65 years, nearly half of whom have been on dialysis for more than a decade [11]. These trends underscore the critical need for effective cancer screening of patients undergoing dialysis.
Serum albumin, an endogenous antioxidant, is associated with cancer development and also reflects nutritional status; thus, it may predict cancer survival in the general population [12–15]. Additionally, hypoalbuminemia is a marker of poor prognosis in patients with ESKD and is associated with higher mortality and cardiovascular disease [16–19]. However, its association with cancer incidence has not been extensively investigated in this population. The association of serum albumin levels with cancer risk in patients with ESKD may suggest their usefulness as a clinically relevant biomarker for optimizing cancer screening strategies in this population. Therefore, in the present study, we investigated the association of serum albumin levels with cancer risk and subsequent mortality in patients undergoing maintenance HD using data from the Health Insurance Review and Assessment Service (HIRA), a comprehensive national registry in South Korea.
Methods
Data source and study population
Data from adult patients (age ≥18 years) undergoing maintenance HD for ≥3 months and ≥2 times per week were retrospectively analyzed. The participants in the fourth (July and December 2013), fifth (July and December 2015), sixth (March and August 2018), and seventh (October 2020 and March 2021) HD quality assessment programs conducted by the HIRA in South Korea were included [20]. The fourth, fifth, sixth, and seventh HD quality assessment programs included 21,839, 35,496, 31,238, and 38,729 patients, respectively. The following patients were excluded to ensure data integrity: repeated participants (13,789 in the fifth, 18,518 in the sixth, and 21,421 in the seventh HD quality assessment programs), those with insufficient dataset or invalid data (n = 271), those who underwent HD with a catheter (n = 1,763), as such patients often have limited life expectancy or unstable clinical conditions, those without data on serum albumin (n = 6), those with upper and lower 1% extreme values for serum albumin levels (n = 1,347), and those with International Classification of Diseases, 10th Revision (ICD-10) codes for cancer for 1 year period, including the 6 months before and after the HD quality assessment (n = 5,459). Hypoalbuminemia can serve as an early marker of undiagnosed cancer or indicate the progression of an existing cancer. To reduce the potential for reverse causality, patients diagnosed with cancer within 6 months after cohort enrollment were excluded from the analysis. Finally, the analysis included 64,728 patients.
Ethics statement
This study was approved by the Institutional Review Board (IRB) of Yeungnam University Medical Center (No. YUMC 2023-12-012) and was conducted in accordance with the principles of the Declaration of Helsinki. The IRB waived the requirement for informed consent because patient records and information were anonymized and de-identified before analysis.
Exposure
The primary exposure of interest was serum albumin level (g/dL), which was measured as part of routine laboratory testing. For the analysis, serum albumin levels were treated as categorical variables. Serum albumin levels were measured monthly during the 6 months of the HD quality assessment program. We calculated mean serum albumin levels and sorted the patients into quintiles based on these levels: 1Q, 3.12–3.72 g/dL; 2Q, 3.73–3.92 g/dL; 3Q, 3.93–4.06 g/dL; 4Q, 4.07–4.21 g/dL; and 5Q, 4.25–4.73 g/dL.
Study variables
Data on age, sex, HD vintage (months), diabetes mellitus as an underlying cause of ESKD, and vascular access type (autologous arteriovenous fistula or graft) were obtained. Hemoglobin levels (g/dL), body mass index (kg/m2), Kt/Vurea, serum phosphorus levels (mg/dL), serum creatinine levels (mg/dL), and ultrafiltration volume (L/session) were recorded. These data were collected monthly, and all laboratory values were averaged from the monthly collected values. Kt/Vurea was calculated using the Daugirdas equation [21].
The codes for the medications are listed in Supplementary Table 1 (available online). The medications included renin-angiotensin system blockers (RASBs), aspirin, clopidogrel, and statins. Use of these medications was defined as one or more prescriptions during the 6 months of the HD quality assessment program. Comorbidity was assessed using the Charlson comorbidity index (CCI), which was evaluated the year before the HD quality assessment program. CCI scores were computed for all patients using ICD-10 codes as previously described [22,23]. Myocardial infarction (MI) or congestive heart failure (CHF) was also defined using the ICD-10 codes.
Outcomes
The patients were followed up until June 2024. Data on patient deaths were obtained from the HIRA, and patients who switched to peritoneal dialysis or received kidney transplantation without experiencing an event were censored at the time of transfer. The incidence of any cancer was defined as the presence of ICD-10 codes for 12 cancers between the endpoint of each HD quality assessment program and the endpoint of follow-up. The 12 cancers were selected from the most common in South Korea based on the 2020 database [24] and included thyroid (C73), lung (C33 and C34), colorectal (C18–21), stomach (C16), breast (C50), prostate (C61), liver (C22), pancreas (C25), gallbladder or biliary duct (C23 and C24), kidney (C64 and C65), uterus or cervix (C53–55), and bladder (C67) cancers. The time of diagnosis of any cancer was defined as the time when an ICD-10 code for one of the 12 cancers was first recorded on a prescription. Any cancer-free interval was calculated using the endpoint of each HD quality assessment program and the time of diagnosis of any cancer, death, or censoring time. Survival was evaluated in patients diagnosed with cancer during follow-up. The survival duration was calculated as the time between the diagnosis of any cancer and the time of death or censoring.
Statistical analysis
We analyzed the data using SAS Enterprise Guide (version 7.1; SAS Institute) and R (version 3.5.1; R Foundation for Statistical Computing). Categorical variables are presented as frequencies and percentages, whereas continuous variables are presented as means and standard deviations. We applied Pearson chi-square or Fisher exact tests to evaluate the statistical significance of the differences between categorical variables. Differences in continuous variables among the five groups were assessed using one-way analysis of variance, followed by Tukey post-hoc test.
The baseline characteristics differed significantly among the five groups. We used propensity score weighting to balance these characteristics and to ensure that the results of our analyses were not biased. We created a balanced cohort for the five groups using generalized boosted models for the following variables: age; sex; body mass index; underlying cause of ESKD; types of vascular access; CCI score; HD vintage; ultrafiltration volume; Kt/Vurea; hemoglobin, creatinine, phosphorus, and calcium levels; administration of RASBs, aspirin, clopidogrel, or statins; and presence of MI or CHF. We used the propensity scores to calculate the inverse probability treatment weights. Finally, we defined the balanced cohort as a sample with weights assigned to each case. Continuous variables are presented as means and standard errors. The p-values were calculated using a general linear model with a complex survey design that included sample weights.
Survival curves were estimated using the Kaplan-Meier curves and compared using p-values determined via the log-rank test with weights. We calculated hazard ratios (HRs) and confidence intervals (CIs) using Cox regression analysis. The multivariate Cox regression analyses were adjusted for age; sex; body mass index; type of vascular access; CCI score; HD vintage; underlying cause of ESKD; ultrafiltration volumes; Kt/Vurea; hemoglobin, creatinine, calcium, and phosphorus levels; presence of MI or CHF; and use of RASBs, aspirin, clopidogrel, or statins and were performed using the enter mode. To minimize the potential for reverse causality, we performed sensitivity analyses in which cancer cases diagnosed within the first 2 years of follow-up were excluded. We used a restricted cubic spline curve to evaluate nonlinear relationships between serum albumin levels and the incidence of any cancer or patient death, which were unadjusted or adjusted for covariates. Statistical significance was set at p < 0.05.
Results
Clinical characteristics
All variables differed significantly among the five groups in the original cohort (Supplementary Table 2, available online); therefore, we performed analyses after propensity score weighting. We assessed the balance among the five groups by calculating the maximum pairwise absolute standardized mean differences (ASMDs) of the covariates before and after balancing (Supplementary Fig. 1, available online). After applying the weights, the maximum ASMDs and differences in baseline characteristics decreased for most covariates. The 1Q, 2Q, 3Q, 4Q, and 5Q groups using the weighted cohort included 60,987, 62,948, 63,386, 62,907, and 61,326 patients, respectively. The baseline characteristics after weighting are shown in Table 1. Although the differences in characteristics were attenuated, the 1Q and Q2 groups had older patients compared with the other groups, and the proportion of males or arteriovenous fistulas increased as the quintile increased.
Risks of cancer and all-cause mortality after cancer diagnosis according to serum albumin level
The follow-up durations of the 1Q, 2Q, 3Q, 4Q, and 5Q groups were 62.3 ± 0.4, 64.1 ± 0.3, 64.9 ± 0.3, 64.7 ± 0.3, and 65.3 ± 0.4 months, respectively. The distribution of cancer types among patients with newly diagnosed cancers is shown in Supplementary Table 3 (available online). The proportions of colorectal, liver, and kidney cancers were the highest among the 12 cancers. The cancer-free rates were 88.0%, 88.1%, 89.5%, 89.6%, and 89.8% (Fig. 1A), while the survival rates of patients with cancers were 39.1%, 42.8%, 44.3%, 44.1%, and 41.3% in the 1Q, 2Q, 3Q, 4Q, and 5Q groups (Fig. 1B), respectively. The 1Q group had the highest incidence of cancer and the poorest survival in patients with cancer among the five groups.
The results of the univariable and multivariable Cox re gression analyses revealed higher risks of cancer in the 1Q and 2Q groups than in the 3Q group (adjusted HR [aHR], 1.12; 95% CI, 1.08–1.16 for 1Q vs. 3Q; aHR, 1.08; 95% CI, 1.04–1.12 for 2Q vs. 3Q), while the 4Q group had a lower risk of cancer than the 3Q group (aHR, 0.94; 95% CI, 0.91–0.98) (Table 2). Regarding all-cause mortality among patients with cancer, the 1Q group had a higher risk of all-cause mortality than the 3Q group (aHR, 1.16; 95% CI, 1.11–1.22). The restricted spline curves also showed an inverse association between serum albumin level and the risks of cancer or subsequent mortality (Fig. 2).
To minimize the potential for reverse causality due to delayed cancer diagnosis, we conducted an additional analysis after excluding patients who were diagnosed with cancer within 2 years (Supplementary Table 4, Supplementary Fig. 2; available online). Even after excluding cancer cases diagnosed within the first 2 years of follow-up, the 1Q and 2Q groups continued to exhibit an increased risk of incident cancer compared with the 3Q group. A similar pattern was observed for all-cause mortality, with higher risks in the 1Q and 2Q groups, consistent with the main analyses.
Subgroup analyses based on age or sex
We performed subgroup analyses based on age (≥60 years) and sex (Table 3). Multivariable analyses showed that low serum albumin levels were associated with a higher risk of cancer and subsequent all-cause mortality in both the age and sex subgroups compared with the reference group. These associations were more pronounced in patients aged <60 years than in those aged ≥60 years, and in females than in males. In males, higher serum albumin levels were associated with a lower risk of all-cause mortality after a diagnosis of cancer. In contrast, higher albumin levels in females were associated with an increased risk of all-cause mortality, showing a U-shaped association.
Subgroup analyses based on cancer type
Supplementary Table 5 (available online) shows the trends of cancer and mortality compared with the 3Q group. Overall, lower serum albumin levels were associated with a higher risk of cancer and increased all-cause mortality among patients with cancer. The lowest albumin quintile (1Q) was consistently associated with increased risks of various types of cancers, including thyroid, colorectal, stomach, breast, gallbladder, bile duct, and uterine or cervix cancers, compared with the middle quintile (3Q). Among all cancer types, the association between lower albumin levels and cancer risk was strongest for gallbladder or bile duct cancer (aHR, 1.54; 95% CI, 1.27–1.87 for 1Q vs. 3Q) and for uterine or cervical cancer (aHR, 1.64; 95% CI, 1.34–2.00 for 1Q vs. 3Q). In contrast, higher albumin levels (4Q–5Q) were generally associated with lower risks of cancer, although this trend was not uniform across all cancer types. For instance, the highest quintile (5Q) demonstrated a significantly increased risk of prostate, gallbladder, bile duct, uterine, cervical, and bladder cancers.
Lower albumin levels were associated with higher risks of all-cause mortality after cancer diagnosis across most types of cancers. However, in some cancers, such as thyroid, breast, and liver cancers, the highest quintile (5Q) was associated with an increased risk of mortality, suggesting that baseline serum albumin levels may not be uniformly associated with better survival across all types of cancers.
Discussion
In this large cohort study, we investigated the association between serum albumin levels and cancer risk as well as overall mortality after cancer diagnosis in patients undergoing maintenance HD. Patients in the lowest serum albumin quintile (1Q group) showed the highest risk of cancer and subsequent all-cause mortality. These associations remained significant even after propensity score weighting and adjustment for a range of clinical and dialysis-related variables. Furthermore, the restricted spline curves demonstrated a continuous inverse association between serum albumin levels and the risks of cancer and subsequent mortality. Although this pattern varied depending on the type of cancer, the overall trend remained consistent across the subgroup analyses stratified by age and sex.
Hypoalbuminemia may result from a variety of conditions, including liver cirrhosis, inadequate protein intake, nephrotic syndrome, and protein-losing enteropathy. However, some patients develop hypoalbuminemia without an obvious underlying condition; in such cases, hypoalbuminemia is caused by increased capillary permeability and enlarged interstitial volume, as well as decreased synthesis and increased catabolism of albumin, primarily driven by inflammation [25]. A widely accepted hypothesis in carcinogenesis suggests that free radical generation driven by chronic inflammation contributes to cancer development and progression [26,27]. Recent large cohort studies in the United Kingdom demonstrated inflammation was associated with increased incidence of cancer [28,29]. In patients on HD, serum albumin levels are closely associated with inflammation [30,31]. However, as data reflecting inflammatory status, such as C-reactive protein levels, were not collected in our study, we were unable to establish an association between hypoalbuminemia and inflammation. In addition to inflammation, malnutrition, including micronutrient deficiencies, may also contribute to cancer development [32,33], and the antioxidant properties of serum albumin may further influence carcinogenesis [34,35]. Therefore, further research is needed to elucidate the mechanisms underlying the association between hypoalbuminemia and increased cancer risk in patients on HD.
Previous prospective cohort studies conducted in the general adult population in Japan, China, and Europe have established the relationship between hypoalbuminemia and cancer [12–14]. Although hypoalbuminemia has been linked to a higher mortality rate in patients with ESKD [16,17], its association with cancer risk in patients on maintenance HD has not been previously demonstrated. Our study findings align with those reported previously in the general population, extending the inverse association between serum albumin levels and cancer risk in patients undergoing maintenance HD. While previous studies have shown that patients on dialysis are at increased risk of certain cancers (e.g., kidney, bladder, liver, thyroid, and oral cavity cancers) compared with the general population, current evidence remains insufficient to support recommendations for additional cancer screening, specifically for patients on dialysis [2,4–6]. Although we demonstrated an association between serum albumin levels and cancer incidence, this relationship was only modest and may be influenced by numerous confounding factors. Therefore, it is difficult to recommend any modification of cancer screening strategies based solely on our findings. Such recommendations would require a thorough analysis of diagnostic performance metrics, including sensitivity and specificity. However, our study is significant because it paves the way for further research regarding the recommendations for enhanced cancer screening among patients with low serum albumin levels undergoing HD.
Notably, we observed a stronger association between serum albumin level and cancer incidence in patients aged <60 years. Previous studies have shown that serum albumin levels are generally lower in patients on HD and tend to decrease more rapidly with age [36,37]; however, large international cohort studies have paradoxically demonstrated a higher relative risk of cancer in younger individuals undergoing HD compared with the general population, and that the risk decreases with age [5,6]. Considering these findings, hypoalbuminemia in younger patients on maintenance HD is likely to reflect a more severe underlying pathological condition such as systemic inflammatory conditions, rather than being confounded by other age-related factors like comorbidities or metabolic disorders. Our results suggest that age should be considered when interpreting the association between serum albumin levels and cancer risk in patients on HD, and support further investigation into age-specific risk stratification strategies.
The predictive value of serum albumin level for clinical outcomes varies according to sex [38–40]. Female patients have lower serum albumin levels than male patients [37,39], while albumin and inflammatory markers are more strongly associated with clinical outcomes in men than in women [38]. In a European cohort study, female patients with inflammation tended to have better clinical outcomes than male patients [40], suggesting the protective effects of female hormones against inflammation. Interestingly, in our study, the association between hypoalbuminemia and cancer was more pronounced in females than in males. A large cohort study in Taiwan also reported a trend toward a higher cancer incidence in female patients on dialysis than in male patients [6]. The sex-related differences in carcinogenesis associated with serum albumin levels seem to oppose those observed in inflammation, suggesting that other complex sex-specific mechanisms may influence the link between serum albumin levels and cancer development. Furthermore, the association between serum albumin levels and cancer incidence varies according to cancer type. The complex relationships among sex, serum albumin level, and cancer risk require further study.
This study has certain limitations. First, despite adjustments using propensity score weighting and multivariable analysis to attenuate differences in baseline characteristics across the five groups, the inherent limitations of observational cohort studies make it impossible to exclude unmeasured residual confounders. Second, we did not evaluate the potential causes of hypoalbuminemia, thus limiting its ability to assess the mechanisms underlying its association with an increased risk of cancer. Notably, we did not include inflammatory markers such as C-reactive protein, which limited our ability to determine whether hypoalbuminemia indirectly reflects an elevated inflammatory state. Third, the follow-up period of approximately 5 years may be insufficient to fully assess the incidence of some types of cancer. However, we observed an association between serum albumin levels and the incidence of most cancers within this study period. Fourth, several important confounding variables related to cancer development—including smoking status, alcohol consumption, socioeconomic status, dietary patterns, and physical activity—were not available in our dataset. Therefore, the observed associations should be interpreted with caution due to the potential for residual confounding. Fifth, in our study, cancer diagnosis was defined as the first occurrence of an ICD-10 code in the claims data, which may not precisely reflect the actual clinical onset of cancer. This approach, although carrying a risk of diagnostic delay or reverse causality, is widely accepted and commonly used in large epidemiological studies utilizing national cancer registries or administrative databases [12,13,28,29]. While imaging-confirmed or histologically verified diagnoses were unavailable in our dataset, our definition aligns with prior studies and reflects a pragmatic balance between data accessibility and diagnostic accuracy. Despite these limitations, this study is significant because it utilized a large national dataset, and the results of the subgroup analyses suggest directions for future research.
In conclusion, the results of this study demonstrated the inverse association of serum albumin levels with cancer risk and overall mortality after cancer diagnosis in patients undergoing maintenance HD. These findings support the value of regular monitoring of serum albumin as part of a comprehensive patient assessment in patients undergoing HD. However, further prospective studies are required to clarify the potential role of serum albumin in cancer screening.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement table 1
Print
