Hyperchloremia is associated with aute kidney injury in critical ill patients: an analysis of the MIMIC-III database

Objective: Balanced uid with no critical increase of chloride in serum was recommended in clinic. Whether hyperchloremia could make a difference for intensive care unit (ICU) patients with a higher acute kidney injury (AKI) occurrence remains controversial. Methods: The Medical Information Mart for Intensive Care III (MIMIC-III) database was searched to identify patients hyperchloremia or non-hyperchloremia, and relationship between level of chloride and AKI incidence was analyzed using the univariate and multivariate logistic regression. Patients were divided into four disease subgroups based on the diagnosis at admission: cardiac, cerebral, gastrointestinal, respiratory. The association between maximum chloride (chloride_max) and incidence of AKI in each subgroup was evaluated using the Lowess Smoothing technique. Receiver operating characteristic curves were applied to analyze the diagnostic value of hyperchloremia (chloride_max>110mmol/L) in these four subgroup patients. Results: A total of 34,617 patients were included in our study, of which 12667 patients (36.6%) was diagnosed with hyperchloremia. The risk of incidence of AKI was increased in the hyperchloremia group. As the higher level of hyperchlorimia, the bigger adjusted odds ratio (OR) presented in terms of AKI, with the OR increasing from 1.13 (95%CI 1.06-1.21; P<0.001) to 4.09 (95%CI 3.04-5.52; P<0.001). Normal level of chloride (95-110mmol/L) was associated with the lower incidence of AKI rate compared to the hypochloremia (<95mmol/L) or the hyperchloremia (>110mmol/L) in any subgroup of cerebral, cardiac, respiratory and gastrointestinal disease. The diagnostic performance was good for cerebral disease (AUC=0.617), cardiac disease (AUC=0.636), respiratory disease (AUC=0.623) and gastrointestinal disease (AUC=0.633). The optimal cut-off value in terms of chloride_max for diagnosing AKI was 116mmol/L for the subgroup of cerebral, respiratory and gastrointestinal

(AUC=0.617), cardiac disease (AUC=0.636), respiratory disease (AUC=0.623) and gastrointestinal disease (AUC=0.633). The optimal cut-off value in terms of chloride_max for diagnosing AKI was 116mmol/L for the subgroup of cerebral, respiratory and gastrointestinal diseases, and 115 mmol/L for cardiac patients.
Conclusion: Hyperchloremia was associated with increased risk adjusted AKI incidence among critical ill patients. For ICU patients with cerebral, gastrointestinal and respiratory admission diagnose, the predictive threshold was at 116mmoL/L, and cardiac diagnose was at 115 mmol/L.

Background
Acute kidney injury (AKI) is a major challenge in intensive care unit (ICU) settings, which is characterized by elevated serum creatinine and/or decreased urine output due to a sudden loss of renal function [1].
Critically ill patients are particularly at risk, accounting for 57% of AKI cases [2][3][4]. Numerous studies showed that changes in serum chloride concentration, independent of serum sodium and bicarbonate, are associated with increased risk of AKI, morbidity, and mortality [5][6]. Hyperchloremia occurred in 57.4% of the study patients within 48 h to ICU admission [7]. The chloride-restrictive intravenous strategy intervention period was associated with a 50% decrease in the incidence of AKI and a decrease in renal replacement therapy (RRT) use [8][9][10]. Thus balanced uid with no critical increase of chloride in serum was recommended in clinic. However, whether hyperchloremia could make a difference for ICU patients with a higher AKI occurrence remains controversial.
The 0.9% Saline vs Plasma-Lyte 148 (PL-148) for ICU uid Therapy (SPLIT) trial did not demonstrate reduced risk of AKI, rates of RRT, or in-hospital mortality with chloride-restrictive uid compared with normal saline [11]. Besides, recent meta-analysis showed that there was no clear recommendation to support the choice of chloride-restrictive uid versus unbalanced uid [12]. Furthermore, Oh et al. pointed out that increase in chloride levels and perioperative hyperchloremia were not signi cantly related to the development of postoperative AKI [13]. Lower urinary chloride concentration was associated with increased mortality and incidence of AKI in the ICU [14]. This statement discounts the physiological value of balanced chloride uid in clinical. With proper insights, the proper chloride concentration can be a signi cant clinical marker, and the optimal level of chloride for ICU patients to reduce AKI occurrence remains uncertain.
The relationship between hyperchloremia and AKI could make difference in different disease. Hyperchloremia was reported to be a known risk factor for subsequent development of AKI in patients with aneurysmal subarachnoid hemorrhage [15] or septic shock [16], but make no difference in postoperative patients [17]. The effect of chloride level on ICU patients admitted for different systematic diseases was not clear.
In the present study, we aimed to examine the association between hyperchloremia and AKI in ICU patients, further to characterize the optimal scope of chloride concentration with better clinical outcomes, as well as to nd out the higher incidence of AKI in distinguished disease, in order to better understand the signi cance of controlling chloride level in ICU.

Study design
We conducted a retrospective single-center study based on a large US-based database called the Medical Information Mart for Intensive Care III (MIMIC-III) [18], which containing data associated with over 50,000

Study population and strati cation
Patients who were younger than 18 years old were excluded from this analysis. The following information was extracted: age, gender, comorbidity, sequential organ failure assessment score (SOFA), simpli ed acute physiology score (SAPS ),renal replacement therapy (RRT), glomerular ltration rate (eGFR), initial level of creatinine after ICU admission (creatinine initial ), maximum level of creatinine during ICU (creatinine max ), initial level of sodium after ICU admission (sodium initial ), maximum sodium (sodium max ), change of sodium during ICU (ΔSodium), initial level of chloride after ICU admission (chloride_initial), maximum chloride (chloride_max), change of chloride (ΔChloride), hospital mortality.
We rstly used 95-110mmoL/L as the normal range and reference group in the present study. To further examine the effect of hyperchlorimia, chloride_max was further categorized into ve levels for analysis with logistic regression models: level 1 (< 95mmoL/L), level 2 (95-109mmoL/L), level 3 (110-114mmoL/L), level 4 (115-119mmoL/L), level 5 (120-124mmoL/L) and level 6 (≥ 125mmoL/L) to nd out the optimal level of chloride for ICU patients. The data were also analyzed in terms of subgroups based on diagnosis at admission: cerebral, cardiac, respiratory and gastrointestinal disease.

De nitions and outcomes
The primary endpoint was incidence of AKI. An increase in serum creatinine level of more than 1.5 times above baseline was considered to be acute kidney injury according to the Kidney Disease Improving Global Outcome criteria [19]. Secondary endpoints included number of patients underwent renal replacement therapy (RRT), hospital mortality, ICU mortality, hospital length of stay (LOS), ICU LOS, score of simpli ed acute physiology score (SAPS II) and maximum SOFA during ICU stay. For patients with more than one ICU stay, only the rst ICU stay was considered.

Statistical analysis
Values are presented as the means (standard deviations) or medians [interquartile ranges (IQRs)] for continuous variables, and categorical variables are presented as total numbers and percentages. Comparisons between groups were made using the Student's t-test, Wilcoxon rank-sum test as appropriate. Categorical variables were presented as a percentage and were analyzed using the X 2 test. The Lowess Smoothing technique was used to explore the crude relationship between chloride and AKI. A logistic regression model was built for each subgroup, applying the normal range of chloride level (95-110 mmol/L) as the reference group. A stepwise backward elimination method with a signi cance level of 0.1 was used to build the nal model. Potential multicollinearity was tested using a variance in ation factor, with a value of ≥ 5 indicating multicollinearity. Receiver operating characteristic curves were depicted to show the diagnostic performance. All statistical analyses were performed using the software Stata V.15. All tests were two sided, and a signi cance level of P < 0.05 was used.

Results
The MIMIC III database contains records for 61567 admissions, of which 15091 were excluded for duplications. Of the remaining 46476 admissions, 7938 were excluded because of age less than 18 years old, and 3921 were excluded because of data shortage. Finally, 34617 patients were included in this analysis. Among them, 7364 admissions were AKI patients and 27253 admissions were non-AKI patients. The ow diagram of patient selections was presented in Fig. 1.
Demographic characteristics of the AKI and non-AKI were presented in Table 1. The number of patients of each disease subgroup was as follows: cerebral group (8883, 25.7%), cardiac group (28887, 83.4%), respiratory group (16833, 55.0%) and gastrointestinal group (13424, 38.8%). Group of AKI owed more complication of chronic kidney disease (CKD) or sepsis or septic shock but less history of diabetes, hypertension, heart failure, and chronic obstructive pulmonary disease. AKI patients presented higher level of glomerular ltration rate (eGFR), creatinine initial , and creatinine max than non-AKI patients, and more patients in the AKI group underwent RRT. For the level of chloride, higher level of chloride_max and Δchloride but not chloride_initial was shown in the group of AKI patients. Clinical outcome was explored between the hyperchloremia and non-hyperchlomeria patients. Crude outcomes were observed in Table 2 for patients with hyperchloremia and non-hyperchlomeria. Without adjusting for other factors, higher incidence rate of AKI and RRT were observed in the hyperchloremia group. Hyperchloremia patients presented increased mortality during ICU, as well as higher hospital mortality rate compared to the non-hyperchlomeria group. Besides, the hyperchloremia group was associated with longer LOS and ICU length. In addition, patients with hyperchloremia owed higher score of SOFA and SAPS II than the non-hyperchlomeria group. Relationship between level of chloride and AKI incidence was analyzed using the univariate and multivariate logistic regression ( Table 3). As insigni cant chloride_initial was noticed in terms of AKI, chloride_max was categorized into six groups, which were used as design variables in six regression models. The normal level of serum chloride (95-110 mmol/L) was served as the reference group. Results showed that both hypochlorimia ((chloride_max < 95) and hyperchlorimia (chloride_max > 110) were signi cantly associated with increased AKI. As the higher level of hyperchlorimia, the bigger adjusted odds ratio (OR) presented in terms of AKI, with the OR increasing from 1.13 (95%CI 1.06-1.21) to 4.09 (95%CI 3.04-5.52). Besides, the multivariate logistic regression analyses showed a signi cant positive effect of diabetes, chronic kidney disease, heart failure, COPD, and sepsis in terms of AKI. The speci c relationship between chloride_max and incidence rate of AKI for patients in terms of cerebral, cardiac, respiratory and gastrointestinal disease subgroups was analyzed using the Lowess Smoothing technique (Fig. 2). We observed that the normal level of chloride (95-110 mmol/L) was associated with the lower incidence of AKI rate compared to the hypochloremia (< 95 mmol/L) or the hyperchloremia (> 110 mmol/L) in patients with any of the four systematic diseases. Observed lowest AKI rate was shown in the chloride level of 105-110 mmol/L. Thus, it suggests that chloride level of 105-110 mmol/L might be the optimal chlorimia in cerebral, cardiac, respiratory or gastrointestinal patients. Meanwhile, another four logistic regression models were built for analysis of the cerebral, cardiac, respiratory and gastrointestinal disease subgroups. Figure 2 shows the OR and 95%CI for the four subgroups. A similar trend showed that OR was increased either in group of hyperchloremia or the group of hypochlormia, which indicated the normal chlormia contributed to less AKI in ICU.
Receiver operating characteristic curves were applied to analyze the diagnostic value of hyperchloremia (chloride_max > 110 mmol/L) in these four subgroup patients (Fig. 3). The results showed the diagnostic performance was good for cerebral disease (AUC = 0.617), cardiac disease (AUC = 0.636), respiratory disease (AUC = 0.623) and gastrointestinal disease (AUC = 0.633). The optimal cut-off value in terms of chloride_max for diagnosing AKI was 116 for the subgroup of cerebral, respiratory and gastrointestinal diseases. It indicated that hyperchloremia (chloride_max > 116 mmol/L) could be used to predictive diagnostic of AKI in ICU patients with cerebral, respiratory, or gastrointestinal diseases. For the subgroup of cardiac patients, the cut-off value was shown in 115, which suggested that hyperchloremia (chloride_max > 115 mmol/L) could be applied to predictive diagnostic of AKI in ICU.

Discussion
Our results revealed that hyperchloremia was associated with increased AKI patients who were critically ill. A 'U'-shaped relationship between maximal chloride and AKI was found in our study. Hyperchloremia was related to increased risk of AKI for patients with cardiac, cerebral, respiratory, or gastrointerstinal admission disease. To the best of our knowledge, this is the largest study with comprehensive analysis to establish a link between level of chloride imbalance and AKI using MIMIC database.
Chloride plays essential roles in maintaining water balance, muscular activity, acid-base equilibrium, and osmotic balance [20]. Hyperchloraemia induced renal vasoconstriction and reduced glomerular ltration rate in an animal model; Subsequently, associations between hyperchloraemia and mortality have repeatedly been shown in critically ill patients [21][22]. However, the results have been inconsistent and the underlying mechanisms remain unknown. A large observational study conducted by Neyra et al. recently showed that serum chloride concentration at 72 hours from ICU admission-but not at the time of ICU admission-was independently associated with hospital mortality [23]. This is the reason we choose chloride_max as the inference. A potential cause of hyperchloraemia in the ICU may be inappropriate chloride load in uid therapy, given that the superiority of chloride-restrictive uid strategy, compared to chloride-liberal uid strategy, for preventing hyperchloraemia, adverse kidney events.
We found that chloride level of 95-110 was the most optimal serum chloremia for the all-cause critical ill patients, which presented the lowest incidence of AKI. Limited studies regarding the optimal serum chloride level in ICU patients are available. Previous studies showed different opinions about relationship between chloride level and incidence of AKI. Our ndings are consistent with several negative retrospective cohorts assessing the role of hyperchloremia in ICU patients. Mao et al. reported that chloride exposure during the rst 48 hours were independent risk factors for AKI in moderately severe and severe acute pancreatitis patients [24]. Strong associations between high chloride levels and worse outcome, AKI or death, have been reported [9,11,25]; however, in various cohort studies discrepant results exist [26,27]. None of them found a signi cant difference in AKI and mortality. Our result was different from previous studies in perioperative hyperchloremia. Tak et al. showed that increase in chloride levels and perioperative hyperchloremia were not signi cantly related to the development of postoperative AKI [13]. The perioperative study included hundreds of participants and pointed out that perioperative hyperchloremic metabolic acidosis but not chloride level was independently related to an increased incidence of AKI.
Our study showed the positive relationship between hyperchloremia and incidence of AKI. Although some study reported that there was no signi cant relationship between postoperative chloride concentration and AKI, those study lack high evidence RCT or large enough prospective population [28][29]. Among patients with cerebral disease, previous study reported similar results with our study. Tak et al. pointed out that perioperative hyperchloremia was associated with an increased risk of postoperative AKI after craniotomy for primary brain tumor resection [30]. For patients with respiratory and gastrointestinal disease, few study reported ndings in these aspects. The present study innovatively reported relatively higher incidence of AKI in respiratory or gastrointestinal disease admission ICU patients.
The advantage of the present study is the large sample size, which allowed for subgroup analysis and adjustment for confounding factors, but it also has limitations. Firstly, the level of chloride was calculated in the present study rather than being measured directly, which could cause deviation from actual chloride values despite careful consideration of the optimal equation [31]. Secondly, the grouping method was based on diagnosis at admission, and thus overlap within subgroups was unavoidable. Finally, our study provided the association between hyperchloremia and AKI, and pointed out the signi cant link between hyperchloremia and mortality, owing to the nature of retrospective research, a de nitive causal link for further investigation was needed. It also provided compelling evidence to explore that whether correction of the hyperchloremia could reduce AKI or mortality among these patients.

Conclusion
Hyperchloremia was associated with increased AKI patients who were critically ill. A 'U'-shaped relationship between maximal chloride and AKI was found in our study. Hyperchloremia was related to increased risk of AKI for patients with cardiac, cerebral, respiratory, or gastrointerstinal admission diseases.

Declarations
Acknowledge None Authors' contributions B carried out the molecular genetic studies, participated in the sequence alignment and drafted the manuscript. JY carried out the immunoassays. MT participated in the sequence alignment. ES participated in the design of the study and performed the statistical analysis. FG conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the nal manuscript.
FCP and CYB conceived of the study, and participated in its design and coordination and helped to draft the manuscript. FCP and GJ conceptualized the research aims, planned the analyses and guided the literature review. GJ extracted the data from the MIMIC-III database. ZTT, CH and ZQQ participated in processing the data and doing the statistical analysis. GJ and FCP wrote the rst draft of the paper and the other authors provided comments and approved the nal manuscript. The author(s) read and approved the nal manuscript.

Funding
The protocol was nancially supported by the Suzhou Science, Education and Health Project (KJXW2019003). National Natural Science Foundation of China (Grant No: 81802295).

Availability of data and materials
The datasets presented in the current study are available in the MIMIC III database (https://physionet.org/works/MIMICIIIClinicalDatabase/ les/).

Ethics approval and consent to participate
The establishment of this database was approved by the Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA), and consent was obtained for the original data collection. Therefore, the ethical approval statement and the need for informed consent were waived for this manuscript.

Consent for publication
Not applicable