Higher body mass index is not a protective risk factor for 28-days mortality in critically ill patients with acute kidney injury undergoing continuous renal replacement therapy

Abstract Background: Acute kidney injury (AKI) requiring continuous renal replacement therapy (CRRT) is a fatal and common clinical disorder in critically ill patients. Recent studies have shown that the relationship between BMI and the outcome of patients with AKI undergoing CRRT is conflicting. Methods: A retrospective cohort study based on data reuse. Univariate analysis, multi-factor regression analysis and subgroup analyses were used to explore the association of the BMI with the 28-days mortality risk in patients with AKI undergoing CRRT. Results: From January 2009 to September 2016, a total of 1120 cases met the inclusion criteria and were enrolled in this study. The univariate analysis showed that BMI was associated with 28-days mortality of patients with AKI undergoing CRRT, its HR value was 0.98 (0.96, 0.99). The multi-factor regression analysis showed that BMI was not associated with 28-days mortality of patients with AKI undergoing CRRT in the four models, the adjusted HR value of four models were 1.00 (0.96, 1.04), 1.01 (0.97, 1.04), 1.00 (0.96, 1.04) and 1.00 (0.96, 1.04), respectively. The subgroups analyses showed that the BMI was a risk factor of the 28-days mortality in patients with AKI undergoing CRRT when GFR ≥30 mL/min, its HR value was 1.04 (1.01, 1.09). Conclusion: Higher BMI was not a protective risk of 28-day mortality in patients with AKI undergoing CRRT. Especially, when GFR ≥30 mL/min, higher BMI increased the risk of the 28-day mortality rate in patients with AKI undergoing CRRT.


Introduction
Body mass index (BMI) is a simple and useful index for assessing the nutritional status [1]. It has been reported that BMI may be associated with the prognosis of critically ill patients [2]. However, the relationship of BMI with the outcome of critically ill patients is conflicting. In critically ill patients, some studies had shown that obese patients were likely to have higher mortality than others, while other studies had shown that the BMI was not associated with the mortality rate [3,4]. Recent studies had shown that obese patients had lower mortality rates compared to underweight patients [5][6][7].
Acute kidney injury (AKI) requiring continuous renal replacement therapy (CRRT) is a serious disorder and common in critically ill patients [8]. Additionally, patients with AKI had been found to have worse outcomes which increases the medical costs [9]. In a retrospective single-center study, it was found that BMI was significantly associated with the development of AKI, overweight patients showed higher incidence of AKI and hospital mortality compared to underweight or normal patients [10]. Recently, two studies had shown that high BMI conferred survival benefits to AKI patients undergoing CRRT compared to underweight or normal patients [6,7]. Therefore, researchers have pondered on whether a higher BMI would be a protective risk for the prognosis of patients with AKI undergoing CRRT, and this inspired our interest to explore the relationship of the BMI with mortality in patients with AKI undergoing CRRT.

Objection
To explore the relationship of BMI with the 28-day mortality of patients with AKI undergoing CRRT.

Data source
Data was provided by Seung Hyeok Han, which was stored in the dryad database (https://datadryad.org// resource/doi:10.5061/dryad.6v0j9) [11]. The database is a public data repository which contains data uploaded by authors to make their research data available for future research.

Inclusion criteria
Patients with stage 2 AKI according to the Acute Kidney Injury Network (AKIN) criteria; treated with CRRT.

Exclusion criteria
Age less than 18 years; preexisting CKD or dialysis or CRRT before the study; pregnancy or lactation; postrenal obstruction; prior kidney transplantation; the data that the value of BMI was missing.

CRRT protocol
Nephrologists decided whether or not to initiate CRRT, upon the development of AKI in ICU patients. The indications of CRRT included uncontrolled volume overload, intractable hyperkalemia or metabolic acidosis. The applied model of CRRT was CVVH (continuous venovenous Hemofiltration) through the internal jugular, subclavian, or femoral vein. CRRT was started at a blood flow rate of 100 mL/min, and up to 150 mL/min. The total effluent volume as a sum of dialysis and replacement dose was targeted to deliver !35 mL/kg/h in all patients.

Statistical analysis
(1) Statistical description: Mean ± standard deviation (x ± s) was used for continuous variables of baseline data in the groups, and counts data were shown by numerical values and percentages. (2) Univariate analysis was carried out to detect the possible risk that may be associated with 28-day mortality. (3) In multifactor analysis, we adjusted the possible variables that may affect the prognosis of patients to determine the relationship between BMI and 28-day mortality. (4) Sensitivity analysis was carried out for diabetes, hypertension, congestive heart failure, sepsis, GFR, mechanical ventilation, CRRT dose, CRP and SOFA score to further verify the relationship between BMI and 28-day mortality. All statistical analyses were performed by EmpowerStats (version numbers: 2018-05-05, Copyright 2009 X&Y Solutions, Inc) and R software. p < 0.05 was considered as a statistical difference.

Baseline characteristics
The clinical characteristics and laboratory findings of all patients were shown in Table 1 CRP, GFR, CRRT dose and AKIN score between the four groups were significant, p < 0.05. The difference in other clinical characteristics and laboratory findings between the groups were not significant, p > 0.05 (see Table 1).

The results of univariate analysis
The univariate analysis showed that BMI, congestive heart failure, diabetes mellitus, hypertension, MAP, phosphate (0 h), phosphate (24 h), mechanical ventilation, Hb, Cr, Alb, 2 h urine output at CRRT initiation, APACHE II score, SOFA score and CRRT causes were associated with the 28-day mortality of patients with AKI undergoing CRRT (see Table 2).

The results of multi-factor regression analysis
In the multi-factor regression analysis, we found that when BMI was used as a continuous variable, it was not associated with the 28-day mortality of patients with AKI undergoing CRRT. When BMI was employed as a continuous variable, the adjusted HR value in the four models were separately 1.00 (0.96, 1.04), 1.01 (0.97, 1.04), 1.00 (0.96, 1.04) and 1.00 (0.96, 1.04). When BMI was applied as a classification variable, it was also not associated with the 28-day mortality of patients with AKI undergoing CRRT in the four models (see Table 3).
The results of subgroup analysis of multi-factor regression analysis The sensitivity analysis was carried out for diabetes, hypertension, congestive heart failure, sepsis, mechanical ventilation, CRRT dose, CRP and SOFA score, which revealed that a higher BMI did not reduce the risk of the 28-day mortality rate in critically ill patients. The subgroup analyses showed that the BMI was associated with the risk of 28-days mortality in patients with AKI  Table 4).

Discussion
The multiple factor regression analysis showed that BMI did not decrease the risk of 28-day mortality in critically ill patients with AKI undergoing CRRT. Further sensitivity analyses showed that when GFR !30 mL/min, a higher BMI increased the risk of the 28-day mortality in patients with AKI undergoing CRRT. A prospective study including 82 severely obese patients (mean BMI, 42 ± 6 kg/m 2 ) and 124 nonobese patients (mean BMI, 24 ± 4 kg/m 2 ) with mechanical ventilation in ICU showed that obesity was not associated either with increased ICU mortality or with hospital mortality [4]. In this study, the researchers controlled the potential confounding factors using multiple factor logistic regression analysis and matched the following variables: age, gender, and the simplified acute physiology (SAPS) II score between obese patients and nonobese patients. In another study which included 4698 patients mechanically ventilated, it was found that being overweight was not associated with high mortality in ICU patients after adjusting for the following variables: age, sex, SAPS II, body mass index category and type of ventilation [3]. In our study, we also found that obesity was not associated with the 28-day mortality of patients with AKI undergoing CRRT.
A cohort study that was conducted for over 2 years in six medical-surgical ICUs that enrolled 1698 patients showed that BMI below 18.5 kg/m 2 was independently  ) and (25-29.9 kg/m 2 ) were not associated with high mortality [5]. The conclusion of this study was inconsistent with that of our study. There are several possible reasons: (1) the study populations were different, our study only included ICU patients with AKI undergoing CRRT, while this study included all the ICU patient; (2) the outcome indicators were not similar, the outcome in our study was 28-day mortality, while the outcomes in this study were ICU and hospital mortality; (3) we adjusted for the possible confounding factors to determine the independent effect of BMI on the 28-day mortality, but the possible confounding factors were not adjusted for in this study. Another observational study including 212 patients with AKI undergone CRRT found that a higher BMI was beneficial to AKI patients unlike a low BMI value [6]. In study by Seung Hyeok Han published in 2018, they also found that a survival benefit of high BMI in AKI patients (SOFA score ! 12) undergoing CRRT [7]. The above two studies were performed by Seung Hyeok Han. However, our conclusions were inconsistent with theirs. This may be due to the following reasons: (1) At the beginning of statistical analysis, all cases with missing BMI data and outliers were excluded from the study; (2) The first study by Seung Hyeok Han enrolled 212 patients with AKI undergone CRRT, which was significantly smaller compared to our study, thus the conclusion of our study was more reliable. (3) The second study by Seung Hyeok Han adjusted the following variables: age, sex, CCI score, septic AKI, MAP, eGFR, SOFA score, WBC, Alb and CRRT dose, but did not adjust for the following variables: myocardial infarction, diabetes, congestive heart failure and hypertension, CRP, AKI cause, mechanical ventilation and phosphate which had been reported to occur in critically ill patients [13][14][15][16][17]. (4) We tested the collinearity of the variables included in the statistical analysis, and found that VIF of all variables was less than 2, hence there was no statistical collinearity in the included variables. Nevertheless, further analysis of the included variables showed that SOFA scores comprised of MAP, CR and AKIN stage, and that age and  complications exhibited collinearity with CCI scores. Therefore, to avoid the instability of the model caused by collinearity among variables, we did not include MAP, Cr, AKIN stage and CCI in the statistical analysis. The collinearity between variables was not taken into account in the study by Seung Hyeok Han, which was likely to affect the reliability of their results. Therefore, our conclusion was more reliable compared to the above studies.
There several studies that had shown that the following factors: age, platelet count, APACHE II score, serum creatinine level, a urine output of <0.05 mL/kg/h the first day, eGFR <45 mL/min, et al were associated with the prognosis of patients with AKI undergoing CRRT [18,19]. Moreover, there was study that reported that BMI was a risk for AKI, but was not associated with prognosis of the patients with sepsis treated with CRRT [20].
In our subgroup analysis, the BMI was found to be a risk factor for the 28-days mortality in patients with AKI undergoing CRRT when GFR !30 mL/min. A meta-analysis including 39 general population cohorts (n ¼ 5 459 014) revealed that higher BMI was an independent risk factor for GFR decline and death in patients who have normal or reduced levels of estimated GFR [21]. This may explain why the BMI was a risk factor for the 28-days mortality in patients with AKI undergoing CRRT when GFR !30 mL/min. Even so, we found that the BMI was not associated with 28-days mortality of patients with AKI undergoing CRRT when GFR <30 mL/min. A possible rationale was that when GFR <30 mL/min, the renal function was impaired, which affected the prognosis.

Strength of the study
This study present new findings that BMI was not associated with the 28-day mortality in patients with AKI undergoing CRRT. Here, we controlled more possible confounding factors than previous studies which avoided obvious statistical errors, making our conclusions more reliable.

Limitations of the study
Only patients with AKI and undergoing CRRT were enrolled, which limited the application of our conclusions. There was no data on other risk factors of mortality such as other interventions, cardiac support during the ICU/hospital stay etc. which might influence our conclusions.

Conclusion
Higher BMI was not a protective risk of 28-day mortality in patients with AKI undergoing CRRT. Especially, when GFR !30 mL/min, higher BMI increased the risk of the 28-day mortality rate in patients with AKI undergoing CRRT.