Abstract
Background
The relationship between serum albumin and prognosis in critically ill patients has been studied, however, there is a paucity of exploration into non-linear relationships, particularly in critically ill patients undergoing dialysis. This study intends to investigate the association between serum albumin (ALB) and 28 day mortality in critically ill patients undergoing dialysis.
Methods
We conducted a multi-center retrospective cohort study of patients undergoing dialysis by utilising data from the eICU Collaborative Research Database from 208 distinct ICUs across the United States between 2014 and 2015. The study focused on mortality within 28 days of ICU admission. We employed univariate analysis, multi-factor logistic regression, subgroup analysis, curve-fitting, and threshold effect analysis to examine the correlation between ALB levels and 28 day mortality.
Results
Among the 2,315 patients with a median age of 63 years, 205 (8.86%) died within 28 days of ICU admission. When ALB level was < 2.7 g/dL, the mortality decreased with an adjusted odds ratio (OR) of 0.34 (95% CI 0.22–0.51, P < 0.0001) for every 1 increment in the ALB. However, no significant mortality changes were observed when ALB levels were at or above this threshold.
Conclusion
Our study identifies a nonlinear dose–response relationship between serum ALB levels and 28 day mortality in critically ill patients undergoing dialysis, with a specific turning point observed. This finding underscores a significant negative correlation between ALB levels and mortality risk, with lower ALB levels being associated with higher mortality risk in this particular population.
Supplementary Information
The online version contains supplementary material available at 10.1186/s40001-024-02127-5.
Keywords: Serum albumin (ALB), 28 day mortality, Critically Ill patients, Dialysis
Background
The mortality rate among ICU (Intensive Care Unit) patients stands at approximately 34.6%, underscoring a substantial risk of death that is not always accurately predicted in this population [1]. Patients in the ICU typically require dialysis due to kidney failure and other severe conditions, such as infections or multiple organ dysfunction, which further complicate their treatment and prognosis [2]. Dialysis is a life-saving treatment for managing acute or chronic kidney failure, by removing waste products and excess fluids to restore internal homeostasis. This can be administered either intermittently or continuously, depending on the patient’s condition [3]. Serum albumin (ALB) is a pivotal protein in human plasma, playing a vital role in maintaining plasma colloid osmotic pressure, transporting materials in the bloodstream, and facilitating communication between intracellular, extracellular, and tissue fluids [4]. The ALB can function as an important prognostic indicator in critically ill patients, helping to identify those individuals who are at an elevated risk of mortality and morbidity [5]. Hypoalbuminemia is closely associated with the prognosis of various diseases, including liver diseases, malnutrition, and inflammatory diseases [6–8]. Notably, a significant inverse association has been observed between ALB levels and hospital mortality, with each 1 g/dL reduction in ALB leading to a 0.60-fold increase in the risk of hospital mortality [9]. ALB is frequently employed in conjunction with other biomarkers as a ratio to evaluate mortality in critically ill patients [10–13]. However, it is not currently evaluated as an independent predictor. Moreover, despite the existence of numerous studies examining the correlation between ALB and mortality in ICU patients, there is a paucity of exploration into non-linear relationships, particularly in critically ill patients undergoing dialysis. This paucity of targeted investigation may impede the development of a comprehensive understanding of the correlation between ALB and 28 day mortality in this cohort. We hypothesized that lower serum albumin (ALB) levels in patients undergoing dialysis would be associated with an elevated risk of 28 day mortality following ICU admission. To evaluate this, we conducted a multi-centre retrospective cohort study utilising data from the eICU Collaborative Research Database, which covers 208 ICUs across the United States from 2014 to 2015.
Method
Data source
The data analysed in this study were obtained from the eICU Collaborative Research Database (eICU-CRD), a multicenter ICU database that involves over 200,000 admissions from 208 hospitals across the United States between 2014 and 2015. All data were automatically collected via the Philips Healthcare eICU program and retrieved electronically for research purposes. The eICU-CRD has been widely used in observational studies [14–16]. The application of this database has been authorised by the institutional review boards of the Massachusetts Institute of Technology (Cambridge, MA, USA). One author, Xinglin Chen acquired the access and was responsible for the data extraction (certification number: 40859994). All procedures were conducted in accordance with the pertinent guidelines and regulations of the eICU-CRD. Given that this study was retrospective and involved anonymous data, informed consent was informed consent was not required.
Study population
This study was designed as a multicenter retrospective cohort study. Initially, the study involved 200,859 participants, but 198,544 were excluded based on specific criteria, leaving a final cohort of 2315 patients for analysis. The exclusion criteria were as follows: (1) patients under 18 years of age; (2) ICU admissions lasting less than 24 h; (3) no dialysis performed within the first 24 h of ICU admission; and (4) fewer than one recorded serum albumin (ALB) measurement after ICU admission. A detailed study flow chart is presented in Fig. 1.
Fig. 1.
Flow chart of the study population. This flowchart illustrates the selection process of participants for the multicenter retrospective cohort study, showing the initial number of participants, exclusion criteria, and the final cohort of patients included in the analysis. ICU: intensive care unit
Data variables and covariates
The eICU database contains a wide array of data, including demographic information, physiological parameters from bedside monitors, diagnoses coded using the International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM), and various laboratory results obtained during routine care. For this study, all data were collected within the first 24 h following ICU admission. The primary exposure variable was serum albumin (ALB), collected as the first measurement within 24 h of ICU admission, measured in grams per deciliter (g/dL). Covariates were chosen based on clinical relevance and existing literature, encompassing factors known to influence outcomes in critically ill patients. These covariates included age (recorded at the time of admission), gender (male or female), ethnicity, and body mass index (BMI, kg/m2). Disease severity upon admission was quantified using the Sequential Organ Failure Assessment (SOFA) score, where higher scores indicate more severe illness. Other key covariates included intubation status, use of mechanical ventilation, and the presence of chronic conditions such as chronic obstructive pulmonary disease (COPD), chronic heart failure (CHF), acute myocardial infarction (AMI), and diabetes mellitus (DM), all of which were classified as either present or absent based on the patient’s medical history. The primary outcome of this study was 28 day mortality, defined as death occurring within 28 days of ICU admission.
Statistical analysis
Descriptive statistics were applied to analyze the characteristics and distribution of the study population. Continuous variables are presented as mean with standard deviation (for normally distributed data) or median with interquartile range (IQR) (for non-normally distributed data). Categorical variables are presented as numbers and percentages. To compare differences across the tertiles of albumin levels, a one-way analysis of variance (ANOVA) was used for continuous variables following a Gaussian distribution, the Kruskal–Wallis H test was applied for skewed continuous data, and chi-squared tests were used for categorical data. A multi-factor logistic regression was employed to explore the dose–response relationship between serum albumin (ALB) levels and 28 day mortality. Univariate and multivariate binary logistic regression models were then applied to assess the association between ALB levels and 28 day mortality. To account for potential confounding variables, two adjustment models were utilized, incorporating the following covariates: age (years), gender, ethnicity, BMI, SOFA score, intubation status, use of mechanical ventilation, COPD, CHF, AMI, and DM. To further investigate the association between ALB and 28 day mortality, subgroup analyses were conducted on the following variables: age (years), gender, ethnicity, BMI, SOFA score, intubation, mechanical ventilation use, COPD, CHF, AMI and DM. The results are reported as odds ratios (ORs) with 95% confidence intervals (95% CIs). A two-piece-wise linear regression model was subsequently utilised to evaluate the threshold effect of ALB on 28 day mortality. The turning point for ALB was ascertained through “exploratory” analyses, which entailed moving the trial turning point within a predetermined interval to identify the point that produced the maximum model likelihood. Furthermore, a log-likelihood ratio test was conducted to compare the one-line linear regression model with the two-piecewise linear model. All the statistical analysis was performed using R software version 4.0.0 (http://www.r-project.org) and the Empower Stats (www.empowerstats.com, X&Y solutions, Inc. Boston MA). Two-sided p-values < 0.05 are considered statistically significant.
Results
Demographic and clinical characteristics
A total of 2315 patients were involved in the analysis. The median age of the study population was 63 years (IQR 53–72 years), with 1252 females (54.08%). Age varied significantly across the albumin tertiles (P = 0.012), while there were no significant differences in gender (P = 0.347) and ethnicity (P = 0.189) distributions through the groups. A significant association was found between body mass index (BMI) and albumin levels (P = 0.034). Additionally, the Sequential Organ Failure Assessment (SOFA) scores differed markedly between albumin tertiles, with higher scores observed in patients with lower albumin levels (P < 0.001). The rate of intubation was notably higher in patients within the first albumin tertile (22.69%, P = 0.008). Furthermore, both chronic obstructive pulmonary disease (COPD) and chronic heart failure (CHF) showed significant variation across albumin tertiles, with the lowest CHF incidence observed in the first tertile (P < 0.001). The detailed baseline characteristics and associations between ALB and clinical variables are summarized in Table 1.
Table 1.
Baseline Characteristics of the Study Population by Albumin Tertiles (N = 2315)
| Characteristics | Albumin (g/dL) | |||
|---|---|---|---|---|
| Tertile 1 1.0–2.4 n = 758 |
Tertile 2 2.5–2.9 n = 693 |
Tertile 3 3.0–6.2 n = 864 |
P value | |
| Age (years) | 62.22 ± 13.59 | 63.16 ± 14.01 | 61.03 ± 14.67 | 0.012 |
| Gender, n (%) | 0.347 | |||
| Male | 355 (46.83) | 328 (47.33) | 380 (43.98) | |
| Female | 403 (53.17) | 365 (52.67) | 484 (56.02) | |
| Ethnicity, n (%) | 0.189 | |||
| Caucasian | 409 (53.96) | 417 (60.17) | 457 (52.89) | |
| African American | 204 (26.91) | 165 (23.81) | 247 (28.59) | |
| Hispanic | 58 (7.65) | 41 (5.92) | 71 (8.22) | |
| Asian | 45 (5.94) | 39 (5.63) | 50 (5.79) | |
| Native American | 16 (2.11) | 14 (2.02) | 21 (2.43) | |
| Other/unknown | 26 (3.43) | 17 (2.45) | 18 (2.08) | |
| BMI (kg/m2) | 26.86 (22.78–32.29) | 27.20 (23.50–32.90) | 27.49 (23.47–33.35) | 0.034 |
| SOFA score | 6.58 ± 2.65 | 6.27 ± 2.73 | 5.83 ± 2.38 | < 0.001 |
| Intubated, n (%) | 0.008 | |||
| No | 586 (77.31) | 549 (79.22) | 720 (83.33) | |
| Yes | 172 (22.69) | 144 (20.78) | 144 (16.67) | |
| Mechanical ventilation use, n (%) | 0.100 | |||
| No | 503 (66.36) | 439 (63.35) | 592 (68.52) | |
| Yes | 255 (33.64) | 254 (36.65) | 272 (31.48) | |
| COPD, n (%) | 0.005 | |||
| No | 729 (96.17) | 639 (92.21) | 812 (93.98) | |
| Yes | 29 (3.83) | 54 (7.79) | 52 (6.02) | |
| CHF, n (%) | < 0.001 | |||
| No | 688 (90.77) | 586 (84.56) | 714 (82.64) | |
| Yes | 70 (9.23) | 107 (15.44) | 150 (17.36) | |
| AMI, n (%) | 0.430 | |||
| No | 731 (96.44) | 664 (95.82) | 822 (95.14) | |
| Yes | 27 (3.56) | 29 (4.18) | 42 (4.86) | |
| DM, n (%) | 0.216 | |||
| No | 601 (79.29) | 548 (79.08) | 710 (82.18) | |
| Yes | 157 (20.71) | 145 (20.92) | 154 (17.82) | |
| Time ICU 28 day | 1.45 (1.00–1.94) | 2.67 (1.95–3.78) | 6.04 (3.79–28.00) | < 0.001 |
| 28 day mortality | < 0.001 | |||
| No | 658 (86.81) | 643 (92.78) | 809 (93.63) | |
| Yes | 100 (13.19) | 50 (7.22) | 55 (6.37) | |
Among the 2315 patients, the number of missing values for the covariates was 73 (3.15%) for BMI
Mean (SD): Median (Q1−Q3)/n (%); BMI: body mass index; SOFA: Sequential Organ Failure Assessment; COPD: chronic obstructive pulmonary disease; CHF: chronic heart failure; AMI: Acute myocardial infarction; DM: diabetes mellitus; ICU: Intensive Care Unit
28 day mortality rates by albumin tertiles
The 28 day mortality rate in our cohort was 8.86% (205 out of 2315 patients; 95% CI 7.7–10.01). Mortality rates varied significantly across albumin tertiles, with the lowest tertile (1.0–2.4 g/dL) showing a mortality rate of 13.19% (100 deaths), the middle tertile (2.5–2.9 g/dL) having a mortality rate of 7.22% (50 deaths), and the highest tertile (3.0–6.2 g/dL) having a mortality rate of 6.37% (55 deaths) Table 1.
Univariate analysis for 28 day mortality
There was a significant relationship between age and the risk of mortality, with each additional year increasing the risk by 1% (OR = 1.01, 95% CI 1.00–1.02, P = 0.0362). No significant difference was noted in the risk of mortality by gender (P = 0.9846), nor was there a significant influence of ethnicity distribution on the mortality. However, a trend towards a higher risk of mortality was noted among native American patients (P = 0.0513). A high SOFA score was found to significantly increase the risk of mortality (OR = 1.35, 95% CI 1.28–1.42, P < 0.0001). Furthermore, patients with CHF exhibited a significantly lower mortality rate compared to non-CHF patients (OR = 0.56, 95% CI 0.34–0.92, P = 0.0231) Table 2.
Table 2.
Univariate analysis for 28 day Mortality in critically ill patients
| Exposure | Statistics | OR (95%CI) | P value |
|---|---|---|---|
| Albumin (g/dL) | 2.74 ± 0.66 | 0.59 (0.47, 0.74) | < 0.0001 |
| Albumin Tertile (g/dL) | |||
| Tertile 1 (1.0–2.4) | 758 (32.74) | Reference | |
| Tertile 2 (2.5–2.9) | 693 (29.94) | 0.51 (0.36, 0.73) | 0.0002 |
| Tertile 3 (3.0–6.2) | 864 (37.32) | 0.45 (0.32, 0.63) | < 0.0001 |
| Age (years) | 62.06 ± 14.15 | 1.01 (1.00, 1.02) | 0.0362 |
| Gender, n (%) | |||
| Male | 1063 (45.92) | Reference | |
| Female | 1252 (54.08) | 1.00 (0.75, 1.34) | 0.9846 |
| Ethnicity, n (%) | |||
| Caucasian | 1283 (55.42) | Reference | |
| African American | 616 (26.61) | 0.86 (0.61, 1.22) | 0.4070 |
| Hispanic | 170 (7.34) | 0.95 (0.54, 1.66) | 0.8484 |
| Asian | 134 (5.79) | 0.46 (0.20, 1.06) | 0.0689 |
| Native American | 51 (2.20) | 2.10 (1.00, 4.41) | 0.0513 |
| Other/unknown | 61 (2.63) | 1.07 (0.45, 2.53) | 0.8829 |
| BMI (kg/m2) | 28.84 ± 8.43 | 1.00 (0.99, 1.02) | 0.6114 |
| SOFA score | 6.21 ± 2.60 | 1.35 (1.28, 1.42) | < 0.0001 |
| Intubated, n (%) | |||
| No | 1855 (80.13) | Reference | |
| Yes | 460 (19.87) | 3.28 (2.43, 4.42) | < 0.0001 |
| Mechanical ventilation use, n (%) | |||
| No | 1534 (66.26) | Reference | |
| Yes | 781 (33.74) | 3.39 (2.52, 4.55) | < 0.0001 |
| COPD, n (%) | |||
| No | 2180 (94.17) | Reference | |
| Yes | 135 (5.83) | 0.91 (0.48, 1.71) | 0.7658 |
| CHF, n (%) | |||
| No | 1988 (85.87) | Reference | |
| Yes | 327 (14.13) | 0.56 (0.34, 0.92) | 0.0231 |
| AMI, n (%) | |||
| No | 2217 (95.77) | Reference | |
| Yes | 98 (4.23) | 0.78 (0.36, 1.72) | 0.5430 |
| DM, n (%) | |||
| No | 1859 (80.30) | Reference | |
| Yes | 456 (19.70) | 0.83 (0.57, 1.21) | 0.3230 |
Mean ± SD/n (%)
BMI: body mass index; SOFA: Sequential Organ Failure Assessment; COPD: chronic obstructive pulmonary disease; CHF: chronic heart failure; AMI: Acute myocardial infarction; DM: diabetes mellitus; OR odds ratio; CI confidence interval
Relationship between ALB and 28 day mortality
In the unadjusted model, for each 1 g/dL increase in ALB, the OR for mortality was 0.59 (95% CI 0.47–0.74, P < 0.0001). This finding remained significant even in the adjusted models (OR = 0.66, 95%CI 0.52–0.85, P = 0.0012), indicating that higher ALB is independently associated with a lower risk of mortality. Analysis by ALB tertiles further presented those patients in the middle and highest tertiles of ALB had a significantly lower mortality risk compared to those in the lowest tertile. Specifically, the odds ratios for mortality were 0.48 (95% CI 0.33–0.71, P = 0.0003) for the middle tertile and 0.57 (95% CI 0.39–0.82, P = 0.0027) for the highest tertile Table 3.
Table 3.
Relationship between albumin and 28 day mortality
| Outcome | Crude model | Model I | Model II | |||
|---|---|---|---|---|---|---|
| OR (95%CI) | P-value | OR (95%CI) | P-value | OR (95%CI) | P-value | |
| Albumin (g/dL) | 0.59 (0.47, 0.74) | < 0.0001 | 0.59 (0.47, 0.74) | < 0.0001 | 0.66 (0.52, 0.85) | 0.0012 |
| Albumin Tertile (g/dL) | ||||||
| Low | Reference | Reference | Reference | |||
| Middle | 0.51 (0.36, 0.73) | 0.0002 | 0.50 (0.35, 0.72) | 0.0002 | 0.48 (0.33, 0.71) | 0.0003 |
| High | 0.45 (0.32, 0.63) | < 0.0001 | 0.45 (0.32, 0.63) | < 0.0001 | 0.57 (0.39, 0.82) | 0.0027 |
Model-I adjusted for age, gender and ethnicity. Model-IIadjusted for age, gender, ethnicity, BMI, SOFA, intubated, mechanical ventilation use, COPD, CHF, AMI, DM
OR odds ratio; CI confidence interval
The effect size of albumin on 28 day mortality in prespecified and exploratory subgroups in each subgroup
Stratified analyses were performed based on age, gender, ethnicity, BMI, SOFA score, intubation status, mechanical ventilation use, COPD, CHF, AMI, and DM. The results consistently demonstrated that higher serum albumin (ALB) levels are associated with a reduced 28 day mortality risk across various subgroups. Notably, ALB levels significantly reduced 28 day mortality in older patients (≥ 60 years, OR = 0.57, 95% CI 0.43–0.76, P = 0.0001), female patients (OR = 0.51, 95% CI 0.37–0.69, P < 0.0001), and African American patients (OR = 0.47, 95% CI 0.30–0.75, P = 0.0014). Additionally, patients with lower SOFA scores also benefited significantly from higher ALB levels, with an odds ratio of 0.36 (95% CI 0.18–0.72, P = 0.005). Interaction tests between ALB and other covariates were conducted and no significant interaction effects were identified (Table 4). Furthermore, serum ALB was analysed as a categorical variable using tertiles (2.5–3.0 g/dL), and similarly, no significant interaction effects were found. Details can be found in the Table S1. These findings suggest that the protective effect of ALB on 28 day mortality is robust across different patient characteristics and clinical conditions.
Table 4.
Effect size of albumin on 28 day mortality in prespecified and exploratory subgroups in Each Subgroup
| Characteristic | No. of participants | OR (95%CI) | P -value | P for interaction |
|---|---|---|---|---|
| Age (years) | 0.7293 | |||
| < 60 | 910 | 0.62 (0.42, 0.91) | 0.0143 | |
| ≥ 60 | 1405 | 0.57 (0.43, 0.76) | 0.0001 | |
| Gender, n (%) | 0.1730 | |||
| Male | 1063 | 0.70 (0.50, 0.97) | 0.0300 | |
| Female | 1252 | 0.51 (0.37, 0.69) | < 0.0001 | |
| Ethnicity, n (%) | 0.5429 | |||
| Caucasian | 1283 | 0.64 (0.48, 0.87) | 0.0038 | |
| African American | 616 | 0.47 (0.30, 0.75) | 0.0014 | |
| Other/unknown | 416 | 0.61 (0.36, 1.04) | 0.0720 | |
| BMI (kg/m2) | 0.8986 | |||
| Low | 747 | 0.61 (0.40, 0.92) | 0.0195 | |
| Middle | 747 | 0.55 (0.37, 0.82) | 0.0031 | |
| High | 748 | 0.62 (0.42, 0.92) | 0.0177 | |
| SOFA score | 0.137 | |||
| Low | 1117 | 0.36 (0.18, 0.72) | 0.005 | |
| High | 1198 | 0.65 (0.47, 0.90) | 0.011 | |
| Intubated, n (%) | 0.9083 | |||
| No | 1855 | 0.62 (0.46, 0.82) | 0.0008 | |
| Yes | 460 | 0.60 (0.40, 0.89) | 0.0120 | |
| Mechanical ventilation use, n (%) | 0.0920 | |||
| No | 1534 | 0.46 (0.32, 0.65) | < 0.0001 | |
| Yes | 781 | 0.69 (0.50, 0.95) | 0.0211 | |
| COPD, n (%) | 0.5612 | |||
| No | 2180 | 0.60 (0.48, 0.76) | < 0.0001 | |
| Yes | 135 | 0.44 (0.16, 1.21) | 0.1117 | |
| CHF, n (%) | 0.0917 | |||
| No | 1988 | 0.64 (0.51, 0.81) | 0.0002 | |
| Yes | 327 | 0.31 (0.13, 0.71) | 0.0055 | |
| AMI, n (%) | 0.1860 | |||
| No | 2217 | 0.61 (0.48, 0.76) | < 0.0001 | |
| Yes | 98 | 0.23 (0.06, 0.99) | 0.0485 | |
| DM, n (%) | 0.8539 | |||
| No | 1859 | 0.58 (0.45, 0.74) | < 0.0001 | |
| Yes | 456 | 0.61 (0.36, 1.05) | 0.0770 |
BMI: body mass index; SOFA: Sequential Organ Failure Assessment; COPD: chronic obstructive pulmonary disease; CHF: chronic heart failure; AMI: Acute myocardial infarction; DM: diabetes mellitus; OR odds ratio; CI confidence interval
Identification of nonlinear relationship
A nonlinear dose–response relationship was demonstrated between ALB and mortality (Figs. 2 and Table 5). A positive relationship was observed between the ALB and mortality when the albumin was < 2.7, with an adjusted OR of 0.34 (95% CI 0.22–0.51, P < 0.0001) for each 1 increment in the ALB. When the ALB was ≥ 2.7, the effect was not statistically significant, as indicated by an adjusted OR of 1.43 (95% CI 0.93–2.20, P = 0.1009), suggesting a negligible impact on mortality (Table 5). The application of a generalised additive model indicated a nonlinear association between ALB and 28 day mortality (Table 5). A comparison was conducted between the linear regression model and a two-piece-wise linear regression model, and the P-value of the log-likelihood ratio test was less than 0.001 (Table 5). These findings indicate that the two-piece-wise linear regression model is the optimal choice for modelling the data.
Fig. 2.
Associations between albumin and 28 day mortality in critically ill patients undergoing dialysis. A threshold, nonlinear association between the albumin and 28 day mortality was found in a generalized additive model (GAM). The solid rad line represents the smooth curve fit between variables. Blue bands represent the 95% confidence interval from the fit. Adjusted for age (years), gender, ethnicity, BMI, SOFA, intubation, mechanical ventilation use, COPD, CHF, AMI, and DM
Table 5.
Threshold effect analysis of the albumin and 28 day mortality
| Models | OR (95%CI) | P value |
|---|---|---|
| Model I | ||
| One line effect | 0.66 (0.52, 0.85) | 0.0012 |
| Model II | ||
| Turning point (K) | 2.7 | |
| Albumin < K | 0.34 (0.22, 0.51) | < 0.0001 |
| Albumin ≥ K | 1.43 (0.93, 2.20) | 0.1009 |
| P value for LRT test* | < 0.001 | |
Data were presented as OR (95% CI) P value; Model I, linear analysis; Model II, non-linear analysis. Adjusted for age (years), gender, ethnicity, BMI, SOFA, intubated, mechanical ventilation use, COPD, CHF, AMI, DM. OR, odds ratio
CI: confidence interval; LRT: logarithm likelihood ratio test
*P < 0.05 indicates that model II is significantly different from Model I
Discussion
In this study, we identified a clear inverse correlation between serum albumin (ALB) levels and 28 day mortality in critically ill patients receiving dialysis. Our study is the first to systematically examine this relationship in such a population, demonstrating a nonlinear pattern where mortality risk escalates sharply when ALB falls below 2.7 g/dL but stabilizes beyond this threshold. To the best of our knowledge, this is the first study to meticulously examine the correlation between serum albumin (ALB) levels and 28 day mortality in critically ill patients undergoing dialysis. Our findings reveal a nonlinear pattern: when ALB levels drop below 2.7 g/dL, mortality risk increases significantly. However, once ALB reaches or exceeds 2.7 g/dL, further changes in mortality risk become negligible.
A low ALB is associated with higher mortality and may indicate a poorer prognosis as in previous studies [17–20]. A retrospective cohort study by Praveen Kolumam Parameswaran, involving 381 critically ill patients, identified hypoalbuminaemia as an independent risk element for acute kidney injury (AKI) (OR 1.810, 95% CI 1.102–2.992), and it accelerates the progression of AKI to chronic kidney disease (CKD) [21]. Similarly, a predictive model for AKI determined that ALB is an independent indicator of 28 day mortality in these patients [22]. Supporting this, a Korean hospital study analysing 1,132 AKI patients prior to continuous renal replacement therapy (CRRT) found that higher ALB levels were significantly related to lower 28 day mortality (HR: 0.92, 95% CI 0.89–0.95, P < 0.001) [17]. Furthermore, a retrospective observational study of 740 AKI patients confirmed that higher ALB levels were closely linked to better outcomes [20]. However, this study differs from previous research in that it includes a larger sample size, specifies particular turning points, and focuses on patients undergoing dialysis in the ICU.
A study involving 925 patients with CKD presented that reduced serum albumin (ALB) levels during ICU admission were linked to an increased risk of 1 year mortality. After adjusting for all covariates, the minimum ALB level was associated with a higher risk of 1 year mortality in model 3 (OR 0.579, 95% CI 0.456–0.735, P < 0.001). However, the study did not specify whether CKD patients received dialysis during their ICU stay, whereas our research focuses on short-term (28 day) mortality and includes additional covariates [18]. A retrospective study found on the MIMIC-IV database analysed 5357 ICU patients with sepsis and found a negative correlation between serum ALB and both short-term and long-term mortality [19]. For each 1 g/dL increase in ALB, the risk of 28 day mortality decreased by 39% (OR = 0.61, 95% CI 0.54–0.69). The study also revealed a nonlinear relationship between ALB and clinical outcomes, identifying 2.6 g/dL as a turning point. When ALB were ≤ 2.6 g/dL, each 1 g/dL increase corresponded to a 59% reduction in the risk of 28 day mortality (OR = 0.41, 95% CI 0.32–0.52) [19]. The 28 day mortality observed in our study was lower by 8.86%, and the turning point differed slightly (2.7 g/dL compared to 2.6 g/dL). This may indicate that different populations (patients undergoing dialysis versus sepsis patients) respond differently to changes in ALB.
A study with contrasting results found that patients with low serum albumin (ALB) and high-sensitive C-reactive protein (hs-CRP) had a notably higher mortality risk, with an adjusted hazard ratio (HR) of 1.62 (95% CI 1.06–2.47; p = 0.02). In contrast, groups with either low ALB and normal hs-CRP or normal ALB and high hs-CRP did not show a significant mortality risk after adjustments. This study highlighted low ALB as a significant risk indicator for mortality in stage 5 CKD patients only when systemic inflammation was present [23]. The discrepancies in results may be attributed to differences in the study populations, designs, follow-up durations, and data analysis strategies employed in the various studies. The aforementioned study focused on stage 5 CKD patients, whereas our research concentrated on patients undergoing dialysis in the ICU with a larger sample size. The differing populations may lead to variations in physiological states and risk factors for mortality. Furthermore, the former study analysed ALB in conjunction with hsCRP, which may have increased confounding among variables [24]. In contrast, our study examined ALB independently, aiming for a clearer assessment of its impact on 28 day mortality. Moreover, the covariates included in the two studies differed, with factors such as SOFA score, intubation status, and mechanical ventilation being more critical in ICU patients. The covariates considered in the two studies differed. The previous research did not account for ICU-specific factors such as the SOFA score, intubation status, and mechanical ventilation. Their study also had a longer follow-up period of 60 months, focusing on chronic risks, while our study emphasized short-term changes relevant to ICU patients. Lastly, different statistical methods were employed: the previous study used Cox regression analysis, whereas our research utilized Generalised Additive Models (GAM) and logistic regression, incorporating subgroup and threshold effect analyses. These methodological differences may influence the interpretation of results, particularly in addressing nonlinear relationships and threshold effects [25].
The correlation between low serum ALB and increased mortality risk in critically ill patients undergoing dialysis can be primarily attributed to factors such as malnutrition, inflammatory response, direct loss of ALB, metabolic acidosis, and dialysis-related oxidative stress. Low ALB is a marker of malnutrition-inflammation complex syndrome (MICS), a condition closely linked to cardiovascular mortality in patients undergoing maintenance haemodialysis [23]. Additionally, inflammation can lead to decreased ALB synthesis and increased breakdown [26], with this association being particularly pronounced in patients with end-stage renal disease [23]. Decreased ALB may exacerbate mortality risk by impairing immune function and increasing inflammatory responses and malnutrition [27]. Furthermore, ALB can be directly lost through the dialysate during dialysis, further reducing ALB and consequently elevating mortality risk [28, 29]. Metabolic acidosis may also affect ALB due to a net negative change in nitrogen balance and overall protein equilibrium [30]. Additionally, dialysis-related oxidative stress may compromise the functionality of ALB [31]. Thus, low albumin levels reflect disruptions in both nutritional and inflammatory states, contributing to an increased mortality risk in ICU patients undergoing dialysis through various mechanisms.
Analysis of the nonlinear ALB and 28 day mortality showed that when ALB levels fall below 2.7 g/dL, the increased mortality risk is associated with several factors, including malnutrition and an elevated inflammatory response [27]. Conversely, when ALB levels reach or exceed 2.7 g/dL, changes in mortality risk become less significant, indicating that within this specific range, variations in ALB may impact disease outcomes. Beyond this threshold, the protective effect of ALB might diminish or change [32].
This study addresses a notable gap in the literature by emphasizing the role of serum albumin (ALB) in predicting 28 day mortality risk among ICU patients undergoing dialysis. The results reveal that low ALB levels are strongly related to increased mortality, with patients in the lowest tertile (1.0–2.4 g/dL) experiencing significantly higher mortality rates compared to those in higher tertiles. Serum ALB emerges as an independent predictor of mortality risk, even after accounting for other established factors, such as the SOFA score [27]. Our finding aids clinicians in identifying high-risk patients, particularly among older and female patients, thereby improving risk stratification and enabling timely interventions to enhance patient outcomes.
Furthermore, future research, including prospective studies and randomized controlled trials, is recommended to investigate the mechanisms behind the relationship between ALB and mortality. Such studies would provide more robust evidence for clinical practice. A key strength of this study is its use of multicenter data, which ensures a nationally representative cohort of critically ill dialysis patients and enhances the generalizability of the findings. The study employed odds ratios (OR) and 95% confidence intervals (CI), adjusted for numerous covariates, and included stratified analyses to confirm the robustness and variability of the results. Additionally, the threshold effect of ALB on 28 day mortality was assessed using a two-piece-wise linear regression model, which identified significant turning points in ALB levels.
Despite these strengths, observational studies are inherently limited by potential unmeasured confounders. Notable discrepancies in age and BMI among participants with different ALB levels (P = 0.012, P = 0.034) suggest that unmeasured factors, such as socioeconomic status and insurance coverage, may also influence 28 day mortality. Although the dataset includes information on ethnicity, length of stay and gender, we cannot ascertain the specific impact of unmeasured factors on the ORs. Another limitation is including the missing data for certain variables, to address this and minimise bias, modern statistical methods were employed. Furthermore, the study focused solely on 28 day mortality, lacking an assessment of long-term outcomes. The database did not differentiate between haemodialysis, peritoneal dialysis, or other dialysis modalities, which precluded a detailed analysis of outcomes under different dialysis regimens. As an observational study, the correlation between ALB and 28 day mortality is correlational rather than causal. Furthermore, because the inclusion criteria were limited to ICU patients receiving dialysis, the results may not apply to other types of ICU patients or those undergoing dialysis in non-ICU settings, thereby limiting the generalizability of the findings.
Conclusion
This study identifies a nonlinear dose–response relationship between serum ALB levels and 28 day mortality in critically ill patients undergoing dialysis, with a specific turning point observed. This finding underscores a significant negative correlation between ALB levels and mortality risk, with lower ALB levels being associated with higher mortality risk in this particular population.
Supplementary Information
Acknowledgements
The author is very grateful to the data providers of the study.
Author contributions
XLC performed statistical analysis, DW cleaned the data, LLZ conceived and designed the research, LLZ and TD drafted the manuscript, and GLZ and DW made critical revisions to the manuscript for key intellectual content.
Funding
Not applicable.
Data availability
The data for this study were sourced from the eICU Collaborative Research Database (eICU-CRD). Access to this database was granted through an institutional review board-approved data use agreement. Due to the proprietary nature of the eICU-CRD, data are not publicly available. Researchers interested in accessing the data may apply for access through the eICU-CRD website-https://eicu-crd.mit.edu/ or contact the database administrators directly. For any queries regarding the data or methodology, please refer to the corresponding author.
Declarations
Ethics approval and consent to participate
This study utilized data from the eICU Collaborative Research Database (eICU-CRD), which comprises de-identified patient records from multiple ICUs across the United States with an agreement (our record ID: 40859994) by the PhysioNet review committee. Given that the data are anonymized and the study was retrospective, informed consent from individual patients was not required. All procedures adhered to the relevant guidelines and regulations for research involving secondary data.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Lanlang Zhang and Ting Deng are co-first authors.
References
- 1.Almansour IM, Ahmad MM, Alnaeem MM. Characteristics, mortality rates, and treatments received in last few days of life for patients dying in intensive care units: a multicenter study. Am J Hosp Palliat Med. 2020;37(10):761–6. [DOI] [PubMed] [Google Scholar]
- 2.Saylan S, et al. Factors impacting the likelihood of death in patients in intensive care unit. Int J Anesthesiol Res. 2016;4(4):225–8. [Google Scholar]
- 3.Burton JO, et al. Recent advances in treatment of haemodialysis. J R Soc Med. 2021;114(1):30–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kianfar E. Protein nanoparticles in drug delivery: animal protein, plant proteins and protein cages, albumin nanoparticles. J Nanobiotechnol. 2021;19(1):159. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ghimire B, et al. Serial estimations of serum albumin levels as a prognostic marker in critically ill patients admitted in ICU in tertiary center: an observational study. Med (Baltimore). 2023;102(45):e35979–e35979. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Gremese E, et al. Serum albumin levels: a biomarker to be repurposed in different disease settings in clinical practice. J Clin Med. 2023;12(18):6017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Liu Q, et al. Association between lactate-to-albumin ratio and 28 days all-cause mortality in patients with acute pancreatitis: a retrospective analysis of the MIMIC-IV database. Front Immunol. 2022;13:1076121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Yi X, et al. Association between lactate-to-albumin ratio and 28 days all-cause mortality in patients with sepsis-associated liver injury: a retrospective cohort study. BMC Infect Dis. 2024;24(1):65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Padkins M, et al. Association between albumin level and mortality among cardiac intensive care unit patients. J Intensiv Care Med. 2021;36(12):1475–82. [DOI] [PubMed] [Google Scholar]
- 10.Zhao L, et al. The lactate-to-albumin ratio relationship with all-cause mortality in cerebral infarction patients: analysis from the MIMIC-IV database. Front Neurol. 2024;15:1334097–1334097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wang J, et al. Association between serum creatinine to albumin ratio and short- and long-term all-cause mortality in patients with acute pancreatitis admitted to the intensive care unit: a retrospective analysis based on the MIMIC-IV database. Front Immunol. 2024;15:1373371–1373371. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wang B, et al. The neutrophil percentage-to-albumin ratio is associated with all-cause mortality in critically ill patients with acute kidney injury. Biomed Res Int. 2020;2020(1):5687672. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang J, et al. Relationship between CRP albumin ratio and the mortality in critically Ill patients with AKI: a retrospective observational study. Biomed Res Int. 2021;2021(1):9957563. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Xiong C, et al. Early postoperative acetaminophen administration and severe acute kidney injury after cardiac surgery. Am J Kidney Dis. 2023;81(6):675-683.e1. [DOI] [PubMed] [Google Scholar]
- 15.Boshen Y, et al. Triglyceride-glucose index is associated with the occurrence and prognosis of cardiac arrest: a multicenter retrospective observational study. Cardiovasc Diabetol. 2023;22(1):190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Yuan Z, et al. A nomogram for predicting hospital mortality of critical ill patients with sepsis and cancer: a retrospective cohort study based on MIMIC-IV and eICU-CRD. BMJ Open. 2023;13(9): e072112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lv J, et al. Serum albumin before CRRT was associated with the 28 and 90 day mortality of critically ill patients with acute kidney injury and treated with continuous renal replacement therapy. Front Nutr (Lausanne). 2021;8:717918–717918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Huang F, et al. The association between blood albumin level and cardiovascular complications and mortality risk in ICU patients with CKD. BMC Cardiovasc Disord. 2022;22(1):322. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cao Y, et al. Albumin level is associated with short-term and long-term outcomes in sepsis patients admitted in the ICU: a large public database retrospective research. Clin Epidemiol. 2023;15:263–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yang K, et al. The association between albumin and mortality in patients with acute kidney injury: a retrospective observational study. BMC Nephrol. 2023;24(1):1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Shao M, Wang S, Parameswaran PK. Hypoalbuminemia: a risk factor for acute kidney injury development and progression to chronic kidney disease in critically ill patients. Int Urol Nephrol. 2017;49(2):295–302. [DOI] [PubMed] [Google Scholar]
- 22.Bai ZH, et al. A nomogram to predict the 28 day mortality of critically Ill patients with acute kidney injury and treated with continuous renal replacement therapy. Am J Med Sci. 2021;361(5):607–15. [DOI] [PubMed] [Google Scholar]
- 23.Alves FC, et al. The higher mortality associated with low serum albumin is dependent on systemic inflammation in end-stage kidney disease. PLoS ONE. 2018;13(1): e0190410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zheng LJ, et al. Reduced serum albumin as a risk factor for poor prognosis in critically ill patients receiving renal replacement therapy. BMC Nephrol. 2021;22(1):305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Tu J, et al. Discussion on the interpretation of the results and selection of the effect model in a meta-analysis. Ann Palliat Med. 2022;11(3):1153–4. [DOI] [PubMed] [Google Scholar]
- 26.Kaysen GA, Don BR. Factors that affect albumin concentration in dialysis patients and their relationship to vascular disease. Kidney Int Suppl. 2003;84:S94–7. [DOI] [PubMed] [Google Scholar]
- 27.Liu B, et al. Albumin-globulin ratio is an independent determinant of 28 day mortality in patients with critical illness. Dis Marker. 2021;2021(1):9965124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Yeun JY. Causes of low serum albumin in peritoneal dialysis patients. Int J Artif Organ. 2000;23(3):149–53. [PubMed] [Google Scholar]
- 29.Kalantar-Zadeh K, et al. Slipping through the pores: hypoalbuminemia and albumin loss during hemodialysis. Int J Nephrol Renovasc Dis. 2021;14:11–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Abdul Razak MK, Rasheed JI, Meizel MM. The effect of acidosis on albumin level in patients treated with regular hemodialysis (single center study). Global J Health Sci. 2018;10(11):57. [Google Scholar]
- 31.Lim PS, Cheng YM, Yang SM. Impairments of the biological properties of serum albumin in patients on haemodialysis. Nephrol (Carlton). 2007;12(1):18–24. [DOI] [PubMed] [Google Scholar]
- 32.Wang MY, et al. U-shape relationship between plasma leucine level and mortality in the intensive care unit. Dis Mark. 2022;2022(1):7389258. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data for this study were sourced from the eICU Collaborative Research Database (eICU-CRD). Access to this database was granted through an institutional review board-approved data use agreement. Due to the proprietary nature of the eICU-CRD, data are not publicly available. Researchers interested in accessing the data may apply for access through the eICU-CRD website-https://eicu-crd.mit.edu/ or contact the database administrators directly. For any queries regarding the data or methodology, please refer to the corresponding author.