This method was modified from, and used in accordance with, an approach previously used by check this Finney et al to describe hyperglycaemia [19].As the relationship between LacADM, LacMAX, LacTW and mortality was expected not to be linear in nature, categorical variables were created. We divided lactate into four bands: normal (0.00 to 2.00 mmol.L-1); mild hyperlactemia (2.01 to 4.00 mmol.L-1); moderate hyperlactatemia (4.01 to 6.00 mmol.L-1) and severe hyperlactatemia (> 6.01 mmol.L-1) for comparison.The normal range of lactate (0.00 to 2.00 mmol.L-1) was subsequently divided into eight bands. However, due to the small number of patients with values under < 0.75 mmol.L-1 we combined the three lower octiles to achieve adequate size for statistical comparison. We therefore compared: the lower limit of normal (LLN, 0.

00-0.75 mmol.L-1); upper limit of normal (ULN, 1.76 to 2.00 mmol.L-1) and four intermediate categories (0.75 to 1.00 mmol.L-1); (1.01 to 1.25 mmol.L-1); (1.26 to 1.50 mmol.L-1); (1.51 to 1.75 mmol.L-1).To confirm that any association between LacTW levels within the normal range and mortality was not being biased by patients who had individual lactate concentrations above 2 mmol.L-1 while in the ICU, we then examined the association between LacTW and mortality in the cohort of patients whose lactate never exceeded 2 mmol.L-1.The primary outcome for analysis was hospital mortality and the secondary outcome was ICU mortality. We performed crude univariate analysis with lactate as a catagorial variable for comparison between groups according to hospital survival status using chi-square test for proportions, Student t-test for normally distributed outcomes and Wilcoxon rank sum tests otherwise.

In addition, we performed multivariate analysis where we adjusted for all available predictors of hospital mortality included in the models (gender, age, APACHE II, mechanical ventilation, surgical admission and diagnosis type) determined by backward elimination of non-significant variables. Furthermore, to determine if the lactate associations were consistent across patient admission diagnosis subgroups and study hospitals, we examined the interactions between measures of lactate and other variables in the model. We report results from the multivariate models using odds ratios, OR (95% confidence intervals, 95% CI).All analyses were performed using SAS version 9.

2 (SAS Institute Inc, Cary, NC, USA). A two-sided P-value of 0.05 was considered to be statistically significant.ResultsWe studied a heterogeneous cohort of 7,155 critically ill patients with 172,723 blood lactate measurements (Table (Table1).1). The absolute blood lactate concentrations (admission lactate LacADM, maximal lactate LacMAX and time-weighted lactate LacTW), were significantly higher in non-survivors compared to survivors (Table Dacomitinib (Table11).