A secondary analysis of a bi-center, retrospective, and observational study including critically ill patients with acute kidney injury (AKI) and treated with CRRT from January 2009 to September 2016 at Yonsei University Health System Severance Hospital and National Health Insurance Service Medical Center Ilsan hospital, Republic of Korea.

The study aimed to explore the relationship of ALB with the 28- and 90-day mortality of patients with AKI and treated with CRRT and with the removal efficiency of serum phosphorus.

New ethics permission and consent to participate were not applicable since the original author had received ethical approval while performing this research, and our study was a retrospective analysis of data reuse.

The data used in this study were shared by Seung Hyeok Han, which were stored in the Dryad database (https://datadryad.org//resource/doi:10.5061/dryad.6v0j9) (7). The database is a public data repository, which contains data uploaded by the authors to make their research data discoverable, freely reusable, and citable.

(1) Patients were complicated with AKI and treated with CRRT in the intensive care unit (ICU); (2) the stage of AKI was two or more according to the acute kidney injury network (AKIN) criteria, in which serum creatinine and urine outputs were taken into account.

Patients with the following situation were excluded in this study: (1) age <18 years; (2) pregnant or lactating women; (3) with postrenal obstruction; (4) with Stage 5 chronic kidney disease (CKD), kidney transplantation, dialysis, or CRRT; (5) the value of ALB was missing or outliers.

From January 2009 to September 2016, 2,110 patients were presented with AKIN stage 2 or more and treated with CRRT, of which 978 patients were excluded in the ICU at Yonsei University Health System Severance Hospital and National Health Insurance Service Medical Center Ilsan hospital: (1) age <18 years (n = 42); (2) pregnant or lactating women (n = 12); (3) with postrenal obstruction (n = 263); (4) with the history of stage 5 CKD, kidney transplantation, dialysis, or CRRT (n = 585); (5) the value of ALB was missing (n = 9) or outliers (n = 3), the ALB of two patients were 0 g/L, and the another was 5.9 g/L). Finally, a total of 1,132 patients with AKI and treated with CRRT were included in the current study (as shown in Figure 1).

According to ALB before CRRT, patients were divided into three groups: ALB <25 g/L (n = 436), 25 g/L ≤ ALB <30 g/L (n = 401), and 30 g/L ≤ ALB.

(1) 28 and 90 days mortality; (2) the removal efficiency of serum phosphorus, delta phosphate = phosphate (24 h)-phosphate (0 h).

The following variables were included in the current study, such as age, sex, body mass index (BMI), mean arterial pressure (MAP), complications (myocardial infarction, congestive heart failure, cerebrovascular disease, peripheral vascular disease, dementia, diabetes mellitus, hypertension, and chronic obstructive pulmonary disease), Charlson comorbidity index (CCI) score, biochemical laboratory tests, e.g., K⁺, HCO3⁻, phosphate (0 h), phosphate (24 h), delta phosphate, white blood cell (WBC), hemoglobin (HB), blood urea nitrogen (BUN), creatinine (Cr) and C-reactive protein (CRP), and sequential organ failure assessment (SOFA) score. Age, gender, BMI, and CCI scores were collected at the time of admission; and the other variables were collected at 0 h of CRRT.

The nephrologist assessed whether the patients should be treated with CRRT based on the development of AKI in the ICU patients. The CRRT machines were the multiFiltrate (Fresenius Medical Care, Bad Homburg, Germany) or the Prismaflex (Baxter International Inc. Lundia AB, Sweden). The applied dialyzers had a surface area of 1.0–1.4 m² with a sieving coefficient for albumin and ß2-microglobulin of 0.001 and 0.58–0.65, respectively. The parameters of CRRT were the following: (1) model: Continuous venovenous hemofiltration (CVVH) through the internal jugular, subclavian, or femoral vein; (2) blood flow: the start was at 100 ml/min and up to 150 ml/min; (3) The total dialysis and replacement dose were targeted to deliver ≥ 35 ml/kg/h in all the patients.

(1) Statistical description: Mean ± SD (x ± s) was used for the continuous variables of baseline data in the groups, and counts numerical values and percentages were shown in the data. The data were compared using the t-test, if continuous data had a normal distribution and homogeneity of variance. Mann–Whitney's U-test was performed if the continuous data did not meet the normal distribution or homogeneity of the variance. For categorical data, the χ²-test was utilized. (2) Analyze the relationship between ALB and 28- and 90-day mortality of patients, and delta phosphate by univariate and multivariate analysis. A multivariate Cox regression was performed for 28- and 90-day mortality, and multivariate logistic regression was performed for delta phosphate. (3) To further understand the relationship between ALB and 28- and 90-day mortality of patients, subgroup analyses were performed on the age, MAP, congestive heart failure, hypertension, AKIN stage, mechanical ventilation, SOFA score, AKI causes, and CRRT causes. (4) The selection of adjustment variables were by the following: if the confounders influenced the effective estimate of ALB by more than 10% and identified with the literature, we would adjust it. (5) Curve fitting analysis was used to investigate the connection among ALB, 28- and 90-day mortality, and delta phosphate. All statistical analyses were carried out by EmpowerStats 2.0 (Copyright 2009 X&Y Solutions, Inc.) and R software (version 3.4.3). The value of P < 0.05 was statistically significant.

In this research, 1,132 patients with AKI and treated with CRRT were included. The ages of three groups included were as follows: ALB <25 g/L (n = 436), 25 g/L ≤ ALB <30 g/L (n = 401), and 30 g/L ≤ ALB were 63.31 ± 14.21, 64.27 ± 14.09, and 61.82 ± 14.89 years, respectively, with no significant difference (P = 0.091). The male-to-female ratio among the three groups was 280/156 (1.79), 241/160 (1.51), and 178/117 (1.52), with no statistical difference (P = 0.399). The BMI among the three groups was 23.52 ± 4.72, 23.75 ± 4.49, and 24.24 ± 4.48 kg/m², respectively, with statistical significance (P = 0.034). The 28-day mortality among the three groups was 307 (70.41%), 257 (64.09%), and 141 (47.80%), respectively, with statistical significance (P < 0.001). The 90-day mortality among the three groups was 350 (80.28%), 294 (73.32%), and 171 (57.97%), respectively, with statistical significance (P < 0.001). There was a significant difference between myocardial infarction, congestive heart failure, and dementia (P < 0.05). There was a significant difference in HCO³⁻, phosphate (24 h), delta phosphate, HB, CRP, WBC, SOFA score, AKI cause, and CRRT cause across the three groups (P < 0.05). There was no significant difference in the other variables between the groups (P > 0.05) (as shown in Table 1).

When ALB was used as a continuous variable, the higher the serum ALB before CRRT, the lower the mortality of critically ill patients with AKI and treated with CRRT. The adjusted hazard ratio (HR) value for 28-day mortality in the four models were separately 0.92 (0.90, 0.95), 0.91 (0.89, 0.94), 0.92 (0.89, 0.95), and 0.92 (0.89, 0.95). The adjusted HR value for 90-day mortality in the four models were 0.91 (0.89, 0.94), 0.92 (0.89, 0.95), 0.92 (0.89, 0.95), and 0.92 (0.89, 0.96). When ALB was used as a classification variable and ALB <25 g/L as a reference, it was discovered that when 25 g/L ≤ ALB <30 g/L and 30 g/L ≤ ALB, the 28- and 90-day mortality of patients with AKI and treated with CRRT were substantially decreased (as shown in Table 2).

Age, MAP, myocardial infarction, congestive heart failure, diabetes mellitus, hypertension, AKIN stage, mechanical ventilation, SOFA score, AKI causes, and CRRT causes were all subjected to sensitivity analysis. ALB was associated with 28-day mortality in the sensitivity analysis, except in individuals with myocardial infarction, AKIN stage 2, or CRRT cause (hyperkalemia, uremia, and oliguria). Additional sensitivity analysis for 90-day mortality revealed that ALB was linked with 90-day death in all the patients except those with myocardial infarction, congestive heart failure, AKIN stage 2, or CRRT cause (hyperkalemia, uremia, and oliguria; as shown in Table 3).

When ALB was used as a continuous variable, the higher the serum ALB before CRRT treatment, the higher the clearance efficiency of serum phosphorus. The adjusted OR values for delta phosphate in the four models were −0.04 (−0.07, −0.01), −0.05 (−0.08, −0.01), −0.04 (−0.08, −0.01), and −0.04 (−0.08, −0.01). When ALB was used as a classification variable, it was also found that the higher the ALB of patients, the higher the removal efficiency of serum phosphorus (as shown in Table 4).

In this study, we discovered that the higher the ALB, the lower the 28- and 90-day mortality of patients with AKI treated with CRRT, and the higher the delta phosphate in the curve fitting analysis. The following variables were adjusted in curve fitting analysis: age, sex, BMI, CCI, CRP, WBC, HB, phosphate (0 h) (except for delta phosphate), k⁺, HCO³⁻, AKI cause, CRRT cause, AKIN stages, CRRT dose, 2 h urine output before CRRT, and SOFA score (as shown in Figures 2–4).

(1) Through multivariate analysis, sensitivity analysis, and adjusting the potential confounding factors, this study got a more consistent conclusion: The higher the serum ALB, the better the prognosis of patients with AKI and treated with CRRT; (2) This study also made it clear that increasing the serum ALB might improve the clearance efficiency of serum phosphorus, and then improve the prognosis of critically ill patients with AKI and treated with CRRT.

(1) This research belongs to a bi-center, retrospective, and observational cohort study and lacked important data such as length of the sessions, how many sessions throughout the ICU stay, and methods for coagulation. This lead to a certain possible danger of bias in this research, so its conclusion needed to be verified by a prospective study. (2) The AKIN standard was used to diagnose acute renal injury in this research; however, Kidney Disease Improving Global Outcomes (KDIGO) may be a superior AKI diagnostic standard. (3) Due to the lack of data on creatinine and urea nitrogen after CRRT in this study, only the relationship between ALB and serum phosphorus clearance was obtained in this study, while the relationship between ALB and CRRT clearance efficiency was needed to be further studied. (4) Albumin will rely on the condition of extracellular hydration and blood volume, although hydration quantity, nutritional status of the patients, and exogenous protein supplements might influence albumin measurements, and lead to a certain potential risk of bias. (5) In addition, some patients lost to follow-up, which increased the possibility of bias in the results.