The Impact of C-Reactive Protein-To-Albumin Ratio on Mortality in Patients with Acute Kidney Injury Requiring Continuous Renal Replacement Therapy: A Multicenter Retrospective Study

Abstract Introduction: C-reactive protein-to-albumin ratio (CAR) is a prognostic marker in various diseases that represents patients’ inflammation and nutritional status. Here, we aimed to investigate the prognostic value of CAR in critically ill patients with severe acute kidney injury requiring continuous renal replacement therapy (CRRT). Methods: We retrospectively collected data from eight tertiary hospitals in Korea from 2006–2021. The patients were divided into quartiles according to CAR levels at the time of CRRT initiation. Cox regression analyses were performed to investigate the effect of CAR on in-hospital mortality. The mortality prediction performance of CAR was evaluated using the area under the curve (AUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Results: In total, 3,995 patients who underwent CRRT were included, and the in-hospital mortality rate was 67.3% during the follow-up period. The 7-day, 30-day, and in-hospital mortality rates increased toward higher CAR quartiles (all p < 0.001). After adjusting for confounding variables, the higher quartile groups had an increased risk of in-hospital mortality (quartile 3: adjusted hazard ratio [aHR], 1.26, 95% confidence interval [CI], 1.10–1.43, p < 0.001; quartile 4: aHR, 1.22, 95% CI, 1.07–1.40, p = 0.003). CAR combined with Acute Physiology and Chronic Health Evaluation II or Sequential Organ Failure Assessment scores significantly increased the predictive power compared to each severity score alone for AUC, NRI, and IDI (all p < 0.05). Conclusions: A high CAR is associated with increased in-hospital mortality in critically ill patients requiring CRRT. The combined use of CAR and severity scores provides better predictive performance for mortality than the severity score alone. Plain Language Summary The mortality rate of critically ill patients with severe acute kidney injury (AKI) requiring continuous renal replacement therapy (CRRT) remains high. In addition, predicting the prognosis of these patients, which is crucial for determining the appropriate treatment level and timing, is difficult. Herein, we evaluated the C-reactive protein-to-albumin ratio (CAR) as a prognostic factor and compared its predictive performance with those of traditional severity scores. Our study demonstrated that high CAR was associated with increased in-hospital mortality. In particular, the addition of CAR to the APACHE II and SOFA scores was superior to traditional severity scores alone in predicting mortality. Consequently, CAR can be used to enhance the accuracy of predicting mortality in patients with severe AKI who require CRRT.


Introduction
The incidence of acute kidney injury (AKI) has increased over the past few decades, raising concerns worldwide [1,2].AKI occurs in approximately 60% of critically ill patients admitted to the intensive care unit (ICU) and significantly increases the risk of morbidity and mortality [3].Continuous renal replacement therapy (CRRT) is the preferred modality for renal replacement therapy in critically ill patients with severe AKI because it provides hemodynamic stability through a slow shift of excessive water, osmolarity, and electrolytes [4,5].Despite advances in the CRRT technique and insights into critical care, the mortality rate in patients with severe AKI requiring CRRT remains high [6,7].Therefore, there is a need to develop early predictors of mortality in critically ill patients requiring CRRT.
Several biomarkers, such as neutrophil gelatinaseassociated lipocalin, kidney injury molecule-1, and growth differentiation factor-15, and severity scoring systems, such as the Acute Physiology and Chronic Health Evaluation II (APACHE II) and Sequential Organ Failure Assessment (SOFA) scoring systems, can predict mortality in critically ill patients with AKI [8][9][10][11].However, these prognostic predictors have limitations in critically ill AKI patients requiring CRRT because the etiology of AKI varies and the clinical course is difficult to predict [12][13][14][15].
C-reactive protein (CRP) is an acute-phase reactant and representative inflammatory marker.CRP is synthesized by the liver in response to factors released by macrophages and is used for the diagnosis and prognostic prediction of sepsis [16,17].Albumin is an indicator of a patient's nutritional status and decreases in the catabolic status, such as acute infection, liver cirrhosis, and malignancy [18,19].Recently, better mortality predictive power of the CRP-to-albumin ratio (CAR) in critically ill patients has been reported in several studies [20][21][22] because using CRP or albumin alone can be less specific [23].In the present study, we aimed to evaluate the predictive value of CAR for inhospital mortality in critically ill patients with severe AKI who underwent CRRT.

Study Participants and Data Collection
This was a multicenter cohort study of patients with severe AKI requiring CRRT.Patients over 18 years of age who underwent CRRT for over 24 h were included; patients who were already on maintenance dialysis before CRRT were excluded.A total of 4,995 adult patients with severe AKI who received CRRT between 2006 and 2021 were enrolled from eight teaching hospitals in South Korea (the Asan Medical Center, Daejeon Eulji Medical Center, Dongguk University Ilsan Hospital, Inha University Hospital, Keimyung University Dongsan Medical Center, Kyungpook National University Chilgok Hospital, Seoul National University Hospital, and the Catholic University of Korea Eunpyeong St. Mary's Hospital).All the institutions participating in this study were university-affiliated tertiary hospitals so there was no significant disparity in healthcare resources including medical specialists, equipment, and medications.Among them, patients with end-stage kidney disease (n = 651) and those without baseline information (n = 349) were excluded.A total of 3,995 patients were included in the final analysis (online suppl.Fig. S1; for all online suppl.material, see https://doi.org/10.1159/000534970).
Patient information was retrospectively collected, and the data included demographics and comorbidities, clinical parameters, and laboratory results, including complete blood counts, blood urea nitrogen, creatinine, electrolytes, CRP, albumin, calcium, and lactate at the initiation of CRRT.Serum CRP and albumin were quantified using latex-enhanced immunoturbidimetric assay and bromocresol green assay, respectively.Both were analyzed on the Atellica ® CH 930 Analyzer (Siemense Healthcare Diagnostics, Germany).The Charlson Comorbidity Index (CCI) and disease severity scores, such as APACHE II and SOFA scores, were also collected.The CCI consists of 19 diseases with different weights and shows the severity of underlying comorbidities [24].The diagnosis of AKI was based on the KDIGO guideline [25].All patients were diagnosed with AKI stage 3 needing CRRT by a nephrologist.The causes of AKI were reviewed by nephrologists and classified as septic or nonseptic AKI.Septic AKI was defined as an infection that met the criteria for systemic inflammatory response syndrome [26].Information on in-hospital mortality, including at 7 and 30 days, was also collected.

Study Outcomes
The primary outcome was the in-hospital mortality rate.The secondary outcomes were the 7-and 30-day mortalities.Inhospital mortality was defined as death during hospitalization.Deaths occurring after patients were transferred to another hospital were not counted as in-hospital mortality.The 7-and 30-day mortalities were defined as death occurred within 7 and 30 days of hospitalization, respectively.In addition, the predictive power of CAR for mortality was compared with that of the traditional prognostic factors, and the integrative effects of CAR with other prognostic factors were evaluated.

Statistical Analyses
Collected data are expressed as the mean ± standard deviation or median (interquartile range) for continuous variables, while categorical variables are expressed as numbers (percentages).The normality of the data was determined using the Kolmogorov-Smirnov test.Participants were divided into quartiles based on their CAR at the time of CRRT initiation.Comparison of parameters according to quartiles was performed using one-way analysis of variance or the Kruskal-Wallis test for continuous variables and Pearson's χ 2 test or Fisher's exact test for categorical variables.To analyze CAR as a quantitative variable, the relationship between CAR and in-hospital mortality was evaluated using a Cox proportional hazard model with restricted cubic spline functions to capture the potential nonlinear effects.Survival analyses using Kaplan-Meier curves and log-rank tests were performed to investigate the impact of CAR on mortality.Cox proportional hazard regression models were used to estimate the adjusted hazard ratio (aHR) of in-hospital mortality according to CAR.In the multivariate Cox regression model, clinically important variables affecting mortality, such as age, sex, hypertension, the CCI, APACHE II score, and mechanical ventilator use, were adjusted.The proportional hazards assumption was assessed using statistical test and graphical diagnostics based on the scaled Schoenfeld residuals.An interaction term between the variable which does not satisfy the proportional hazards assumption (APACHE II score) and follow-up time was included in the model to account for the nonproportional hazards.Differences between the area under the curve (AUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were analyzed to compare the predicted probabilities among the prognostic factors.The AUC values were compared using De-Long's test [27].We also analyzed the hazard ratio for in-hospital mortality by subgroup to estimate the differences in the effects of CAR.Linear regression analyses were used to evaluate the factors associated with CAR, and significant variables in the univariate analysis were selected for multivariate analysis.Statistical significance was set at p < 0.05.Statistical analyses were performed using SAS for Windows (version 9.4; SAS Institute Inc., Cary, NC, USA) and R software (R Foundation for Statistical Computing, Vienna, Austria; www.r-project.org).

Baseline Characteristics of the Study Participants
The mean age of 3,995 patients was 66.4 ± 15.0 years, and 60.8% were male.The median CAR was 5.05 (IQR, 1.16-7.72),and the study cohort was divided into quartiles according to the CAR (quartile 1: 0.39 [IQR, 0.13-0.74];quartile 2: 2.17 [IQR, 1.63-2.76];quartile 3: 5.32 [IQR, 4.42-6.40];quartile 4: 11.06 [IQR, 9.28-13.39]).Table 1 presents the baseline characteristics of the study patients according to the CAR quartiles.Patients in the higher CAR quartiles were older (p = 0.007) and had a higher incidence of septic AKI (p < 0.001).The male ratio and body mass index did not differ among CAR groups.The prevalence of comorbidities was different for each disease, but there was no difference in the CCI and comorbid liver disease among the quartile groups.The severity scoring systems for critically ill patients, such as SOFA and APACHE II scores, increased with increasing CAR quartiles (p < 0.001).

Association between CAR and Mortality
In total, 2,687 (67.3%) critically ill patients with severe AKI requiring CRRT died during hospitalization, and the in-hospital mortality rate increased with higher CAR quartiles (quartile 1, 58.5%; quartile 2, 66.2%; quartile 3, 69.8%; quartile 4, 74.6%; p < 0.001) (Table 2).Among patients, the 7-and 30-day mortality rates were 40.7 and 59.1%, respectively; both the 7-and 30-day mortality rates also showed significant differences according to CAR quartiles (both p < 0.001).We performed Cox regression analysis using restricted cubic splines to evaluate the relationship between CAR and in-hospital mortality and found that a higher CAR was associated with increased in-hospital mortality in patients with    severe AKI requiring CRRT (Fig. 1). Figure 2 includes the Kaplan-Meier curves showing the survival rate according to the length of hospitalization between the CAR groups.The higher quartile groups showed higher mortality from the early period after CRRT (within 7 days), and the mortality was consistently higher during hospitalization (all log-rank p < 0.001).Table 3 shows that the higher CAR quartile groups had significantly increased in-hospital mortality compared with the lowest CAR quartile after adjustment for confounding factors (model 4; quartile 3, aHR, 1.26, 95% confidence interval [CI], 1.10-1.43,p < 0.001; quartile 4: aHR, 1.22, 95% CI, 1.07-1.43,p = 0.003).In particular, the highest quartile group of CAR (quartile 4) showed consistently higher 7-day, 30-day, and in-hospital mortality rates compared to those of the first quartile group (model 4; 7-day mortality: aHR, 1.19, 95% CI, 1.01-1.37,p = 0.042; 30-day mortality: aHR, 1.25, 95% CI, 1.10-1.43,p < 0.001) (online suppl.Table S1).
In the subgroup analyses defined by patient age, sex, body mass index, cause of AKI, hypertension, diabetes, and APACHE II score, consistent associations were observed between CAR and in-hospital mortality, with higher CAR quartile groups having higher in-hospital mortality (Fig. 3).The interaction analyses revealed that the impact of CAR on in-hospital mortality did not differ within subgroups.

Prognostic Value of CAR Compared with Other Prognostic Factors
We evaluated the efficacy of CAR in predicting mortality in critically ill AKI patients requiring CRRT by comparing it with traditional prognostic factors.Table 4 presents the AUC, NRI, and IDI results for in-hospital mortality.CAR had a higher AUC than CRP and it was lower than the APACHE II and SOFA scores.However, the addition of CAR to the conventional severity scores (APACHE II and SOFA) significantly improved the predictability of in-hospital mortality in all AUC, NRI, and IDI analyses (all p < 0.05).

Associated Factors of CAR
In the multivariate linear regression analysis, CAR was positively correlated with septic AKI, chronic obstructive pulmonary disease, leukemia, and metastatic cancer, while congestive heart failure and chronic liver disease were negatively correlated with CAR (all p < 0.05) (Table 5).Among these variables, septic AKI had the highest standardized coefficient value for CAR (β = 0.188).

Discussion
This is the first study to investigate the association between CAR and mortality, and its predictive power for mortality in critically ill patients who underwent CRRT.The present study demonstrated that a high CAR value at CRRT initiation was associated with increased in-hospital mortality in patients with severe AKI requiring CRRT.CAR has improved the predictability of traditional prognostic indicators in critically ill patients.Precise prediction of mortality using CAR, which is an easily accessible marker, in critically ill patients receiving CRRT will help improve the prognosis by applying timely and proper treatment.
CRP is the most widely used inflammatory biomarker and has been evaluated as a diagnostic and prognostic marker for infection, particularly in critically ill patients [28][29][30].Serum albumin is one of the most important proteins in humans because it maintains the plasma colloid osmotic pressure, is involved in the transport of substances, and facilitates communication among the tissue, intracellular, and extracellular fluids [31].The serum albumin level reflects the nutritional and catabolic status of the patient and is the most representative indicator of protein-energy wasting syndrome [32].For these reasons, CRP and albumin have been identified as prognostic factors in patients with various diseases [33][34][35][36][37].However, the use of CRP or albumin alone is too nonspecific to reflect the various conditions of critically ill patients, so they are insufficient for outcome prediction [38,39].Therefore, we tried to predict the patient's status more delicately by using CRP and albumin together.
CAR is a simple index that combines CRP and albumin, and it is associated with mortality in critically ill patients [20][21][22].Ranzani et al. [20] reviewed 334 patients admitted to the ICU with severe sepsis or septic shock.They revealed a long-term association between CAR and outcome using the 90-day mortality.CAR at ICU discharge was also an independent risk factor for long-term mortality.Moreover, Park et al. [40] evaluated the predictability of CAR for 28-day mortality in critically ill Korean patients.In an analysis of 875 patients, CAR was an independent predictor of 28-day mortality.However, only 30% of patients in the cohort had AKI.Wang et al. [22] evaluated the association between CAR and mortality in ICU-admitted AKI patients.They collected survival data over a 2-year period and found that an increase in CAR was associated with an increased risk of all-cause death.However, the number of patients was small (n = 580) and most patients (88.3%) had mildto-moderate AKI without dialysis.In addition, severity scores, such as APACHE II and SOFA scores, were much lower than those in our study cohorts, and in-hospital mortality was also lower than in this study (Wang et al. [22] vs. present study; 25% vs. 67%).Therefore, our study differs from previous studies in that we identified the mortality predictive power of CAR in critically ill patients requiring CRRT using a large number of patients with severe disease.
CAR has been reported to increase in various underlying diseases such as diabetic nephropathy and chronic inflammation [41,42].According to our results, CAR in critically ill patients was more closely related to severity scores such as SOFA and APACHE II scores than to the underlying comorbidities.Since albumin synthesis occurs in hepatocytes, CAR values may be affected by underlying liver disease.However, the proportion of chronic liver disease did not differ among the CAR quartile groups, and CAR was negatively correlated with underlying chronic liver disease.These results suggest that CAR is more affected by the patients' status at the time of CRRT than the baseline liver disease.In particular, septic AKI showed the greatest correlation with CAR, indicating that CAR is closely associated with infection.
Estimation of mortality in critically ill patients is crucial for prompt determination of the appropriate treatment degree to improve patient outcomes.Since the 1990s, SOFA [43] and APACHE II scores [44] have been widely used as mortality prediction tools for ICU-  admitted patients.However, these traditional ICU severity scores are insufficient for predicting mortality in patients with severe AKI who undergo CRRT.This is because the etiology of AKI is diverse, and the clinical course of these critically ill patients is usually unexpected owing to the high morbidity and mortality [11,15].Therefore, a lot of effort has gone into improving the predictability of mortality in patients with severe AKI requiring CRRT [45][46][47][48].The key features of a good prognostic marker are the ease of measurement and interpretation, reproducibility, low cost, and high predictability [49].CAR is an easily measurable, interpretable, reproducible, and inexpensive prognostic marker that improves mortality predictability by integrating it with the traditional severity scores.In particular, CRP has high daily variability in critically ill patients, resulting in large differences depending on the measurement time point.Our results show that CAR at the time of CRRT initiation is an important prognostic indicator that can represent the condition of severe AKI patients.In addition, CAR showed good prognostic performance even in the subgroup with lower APACHE II scores.Therefore, CAR could be used in conjunction with other prognostic indicators.
This study has several limitations.First, there may have been selection and information bias owing to the retrospective study design.However, our study addressed this issue by enrolling a large number of patients from multiple centers nationwide.Second, we could not assess the effects of CAR change on mortality because we only used a single-measure CAR at the initiation of CRRT.We used CAR at the initiation of CRRT, which may reflect the patients' critical illness to predict mortality in the CRRT population.Studying the time course of CAR may be helpful to evaluate patients' clinical course and treatment response.Finally, this study focused on the analysis of phenomena and failed to provide a pathophysiological causal relationship between CAR and mortality.Future well-designed studies are required to overcome these limitations.Despite these limitations, this study is meaningful because it is the first to confirm the predictive power of CAR for mortality in a large cohort of AKI patients requiring CRRT.
In conclusion, among the critically ill patients requiring CRRT, those with a high CAR had an increased in-hospital mortality rate.CAR enhanced the predictive performance when combined with the conventional ICU severity scoring system, such as APACHE II and SOFA scores.Overall, CAR can be a complementary predictor of mortality in critically ill patients with severe AKI.

Fig. 1 .
Fig. 1.Relationship between CAR and in-hospital mortality hazard ratio according to the restricted cubic spline regression model.The reference value is the upper limit of quartile 1 (1.57).The red line indicates the estimated hazard ratio; the dashed green line indicates the reference line of null hypothesis that the hazard ratio is 1; the dashed black lines indicate the upper and lower 95% confidence limits.CAR, C-reactive protein-to-albumin ratio.

Fig. 3 .
Fig.3.Adjusted hazard ratios for in-hospital mortality in the subgroups.Forest plots show that the higher CAR quartile groups have the greatest risk of in-hospital mortality across subgroups.CAR, C-reactive protein-to-albumin ratio; HR, hazard ratio; BMI, body mass index; AKI, acute kidney injury; HTN, hypertension; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment.

Table 1 .
Baseline characteristics AKI, acute kidney injury; SOFA, Sequential Organ Failure Assessment; APACHE II, Acute Physiology and Chronic Health Evaluation II; CRRT, continuous renal replacement therapy; WBC, white blood cell; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; CRP, C-reactive protein.

Table 2 .
In-hospital information for the CAR quartile groups CAR, C-reactive protein-to-albumin ratio.

Table 3 .
Cox regression analyses for in-hospital mortality in CAR quartile groups Model 2: adjusted for age and sex.Model 3: adjusted for age, sex, hypertension, and CCI.Model 4: adjusted for age, sex, hypertension, CCI, APACHE II score, mechanical ventilator use, and cause of AKI.CAR, C-reactive protein-to-albumin ratio; HR, hazard ratio; aHR, adjusted hazard ratio; CI, confidence interval; CCI, Charlson Comorbidity Index; APACHE II, Acute Physiology and Chronic Health Evaluation II; AKI, acute kidney injury.

Table 4 .
Comparison of the AUC and predictive power of prognosis for in-hospital mortality AUC, area under the curve; CI, confidence interval; NRI, net reclassification improvement; IDI, integrated discrimination improvement; CAR, C-reactive protein-to-albumin ratio; CRP, C-reactive protein; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment.

Table 5 .
Factors associated with CAR in the linear regression analysis