1). Within the first 48 hours of their ICU stay, we recorded patients’ arterial blood pH, PaCO2, bicarbonate concentration, standard base excess, creatinine, sodium, potassium, chloride, albumin and lactate (Lac). Apparent strong ion differences were calculated as plasma concentrations: [Na+] + [K+] + [Mg2+] + [Ca2+] – [Cl-] – [Lac-]. Effective strong ion differences were calculated as first plasma concentrations: 2.46 �� 10pH 8 �� PaCO2 + [albumin] �� (0.123 �� pH – 0.631) + [phosphate] �� 0.309 �� pH – 0.469] (all concentrations are expressed as mEq/L). The strong ion gap was calculated as the difference between apparent and effective strong ion data [5]. Moreover, we recorded the amount of sodium bicarbonate administered, the amount of required vasopressors and the need for renal replacement therapy, intubation and mechanical ventilation.

We also documented the duration of mechanical ventilation and use of vasopressors while in the ICU, the length of ICU stay, and mortality. The primary end point was the mortality rate at ICU discharge. The secondary end points were the amount of time spent on mechanical ventilation in the ICU, the duration of vasopressor use and the overall length of ICU stay.Table 1Characteristics and main outcomes of the study populationStatisticsData are expressed as means �� SD or medians and 95% CI for continuous variables and raw numbers and percentages for categorical variables. Two main comparisons were performed. First, we compared survivors to nonsurvivors. Second, we compared patients treated with sodium bicarbonate to those who were not.

Continuous data were compared using a Student’s t-test or a Mann-Whitney U test regarding the normality of the population distribution. A ��2 test was used for categorical variables. Comparisons of several means were performed using repeated-measures analysis of variance and the Tukey’s post hoc test. Multivariate analyses were performed using a logistic regression model with forward selection procedures to estimate the odds ratio of death (with the 95% CI) after discretization of the continuous variables according to their median values and also to describe the prescription of sodium bicarbonate. Calibration of the logistic model was assessed using the Hosmer-Lemeshow goodness-of-fit test to evaluate the importance of the discrepancy between observed and expected mortality.

Each variable associated with a P value below 0.20 in the univariate analysis was entered into the model. All values were two-tailed, and P < 0.05 was considered statistically significant. Statistical analysis was performed with MedCalc version 9.4.2.0 statistical software (MedCalc Software bvba, Brefeldin_A Mariakerke, Belgium).ResultsDuring the study period, 2, 550 patients were admitted to the five participating ICUs.