This retrospective and the observational study included 131 male patients, who were diagnosed with AKA at the emergency department of Wonju Severance Christian hospital from September 2009 to May 2014. All patients with the word “alcoholic ketoacidosis”, or “ketoacidosis” in the discharge codes registered in the computerized hospital records were selected. We then confirmed the diagnosed of AKA by several findings, including the history of chronic alcohol abuse and a recent episode of binge drinking with metabolic acidosis on arterial blood gas analysis and increased serum anion gap in serum chemistry1). A total of 13 patients were excluded from analyses due to other diagnoses or the inability to obtain patient history (AKA combined with organophosphate intoxication (n=3), diabetic ketoacidosis (n=1), obstructive nephropathy (n=1), and unable to obtain history due to lack of guardian or unconscious patient (n=8)). Unfortunately, there were only 2 female patients during evaluation periods. One hundred eighteen male patients were finally included in the evaluation. Underlying disease, demographic characteristics, and laboratory results were investigated on the day of hospital arrival. We categorized survival and non-survival groups according to the patient's survival. We defined renal survival patients as those who did not receive renal replacement therapy during hospitalization. Patient informed consent was not required for this retrospective, observational study, and the patient records were anonymized prior to the analysis. This study was approved by the institutional review board of Wonju College of Medicine, Yonsei University (CR318085).

Data were collected by retrospectively reviewing medical records. Demographic data and clinical variables including, age, sex, hospitalization duration, initial clinical parameters (blood pressure, heart rate, respiratory rate, and 24 hour urine output), presence of gastrointestinal (GI) symptoms (including nausea, vomiting, melena, hematochezia, diarrhea, and abdominal pain), central nervous system(CNS) symptoms (including mental change, dizziness, seizure, and dysarthria), dyspnea, general weakness, past disease history, alcohol intake history (duration, amount, last drinking time, and amount), use of inotropes (norepinephrine, vasopressin, and epinephrine) during hospitalization, need for mechanical ventilation or renal replacement therapy during hospitalization, total sodium bicarbonate (NaHCO₃) infused (mEq) during the first 24 hours after admission, complications during hospitalization, and mortality were recorded. Initial Acute physiology and chronic health evaluation (APACHE) II scores were calculated to measure the severity of patients' conditions11).

Blood samples for the analyses of DNI and other laboratory parameters were obtained from arterial or venous puncture within the first 2 hours of hospital admission. The blood samples were immediately transferred to the laboratory department, and the DNI assays were performed within 2 hours of blood sampling. DNI was measured through an automated specific hematology cell analyzer (ADVIA 2120, Hematology System, Siemens Healthcare Diagnostics, Forchheim, Germany). The DNI was calculated in leukocyte differentials using the following formula: DNI=(the neutrophil subfraction and the eosinophil subfraction measured in the myeloperoxidase (MPO) channel by cytochemical MPO reaction)−(the PMN subfraction measured in the nuclear lobularity channel by the reflected light beam)81213).

Data are expressed according to the properties of the variable. Continuous variables are presented as mean and standard deviation (SD). Categorical variables are presented as frequency and percentage. In order to compare groups, we performed the two-sample t-test, ANOVA, and Chi-square test (Fisher's exact test) as appropriate. Prognostic variables for mortality and renal survival were analyzed by using the univariate Cox proportional hazard model. Considering the Cox proportional hazards model in univariate analysis and collinearity, multivariate Cox regression analysis was performed. The univariate and multivariate Cox regression analysis results are presented as hazard ratios (HR) and 95% confidence intervals (CI). Considering univariate Cox regression results, collinearity, and clinical importance age, history of diabetes, and history of liver cirrhosis were used in multivariate Cox regression of in hospital day renal replacement therapy (Renal Survival) and hospital day mortality (Patient Survival). The discrimination of DNI, pH, and GFR for all-cause mortality and renal survival were evaluated using the area under the receiver operating characteristics (ROC) curve. Decreased pH and GFR values are associated with poor prognosis in AKA patients. Therefore, we used reverse pH(-pH), GFR(-GFR) in ROC curves. The optimal cut off point for ROC curves was selected for maximizing the sensitivity and specificity of the selected values. Kaplan-Meier survival and renal survival curves were then drawn and Log-rank values were calculated to assess their statistical significance. A p-value less than 0.05 was considered statistically significant. All statistical analyses were conducted using the IBM Statistical Package for the Social Science (SPSS) version 23.0 (IBM Corporation, Armonk, NY, USA).

The average duration of hospitalization for the total 118 male patients was 10.9 days (range 0–71 days) (Table 1). All-cause mortality was 32.2% (38 patients). The most common cause of death was alcoholic ketoacidosis itself (52.6%) (Table 2). On average, mortality occurred 6.7 days after hospitalization. During the hospital day, 64 (54.2%) patients had infectious complications. The most common pathogen of infection was Methicillin-Resistant Staphylococcus aureus (23.4%) (Table 3). The most common symptoms among patients were GI symptoms (48.3%), such as nausea, vomiting, diarrhea, melena, hematochezia, hematemesis, and abdominal pain. History of diabetes, hypertension, liver cirrhosis and the amount of alcohol ingested were not different between survival and non-survival patients. Clinical characteristics that differed between the survival group and the non-survival group are shown in Table 1. On laboratory results, CRP, creatine kinase (CK), and bicarbonate (HCO₃) were not statistically different between survival and non-survival patients. Laboratory results that differed between survival and non-survival patients are shown in Table 1. The DNI level in the survival group (4.8±6.4) was statistically lower than in the non-survival group (11.4±12.5) (p<0.001).

In Cox regression analysis, initial laboratory values (DNI, MAP, pH, hemoglobin, platelet counts, albumin, PT (INR), lactate, and Cr), use of ventilator during hospitalization, use of inotropic during hospitalization, cardiac arrest, and APACHE II score on first hospital day were significantly associated with renal survival. In multivari ate adjustment analysis, initial DNI level and initial pH remained as independent predictors of in-hospital renal replacement therapy (Renal Survival) (Table 4). ROC curves were drawn for initial DNI (AUC 0.671, 95% CI 0.553–0.789), initial pH(AUC 0.738, 95% CI 0.637–840) and CRP (AUC 0.557, 95% CI 0.431–0.683) to determine cut-off values predicting renal survival (Fig. 1). Patients with an initial pH value above 7.044 had higher renal survival days (47.0 days) than other patients (33.9 days) (p=0.006). Mean renal survival days were significantly lower in the high initial DNI group (above 4.75%, 27.8 days) compared with the low initial DNI group (below 4.75%, 53.8 days) (p<0.001) (Fig. 2).

In Cox regression analysis, initial DNI level, initial MAP, initial pH, initial lactate, received renal replacement therapy (RRT) or mechanical ventilation during hospitalization, complications of rhabdomyolysis, arrest during hospitalization, and APACHE II score on first hospital day were significantly associated with mortality. In multivariate Cox regression analysis, DNI and initial pH remained as independent predictors of hospital day mortality (Table 5). ROC curves were drawn for initial DNI (AUC 0.705, 95% CI 0.6–0.811) and, initial pH(AUC 0.761, 95% CI 0.665–0.857) to determine cut-off values predicting mortality (Fig. 3). Mean survival days were significantly lower in the high initial DNI group (above 4.5%, 30.5 days) compared with the low initial DNI group (below 4.5%, 56.3 days) (p<0.001).