- Poster presentation
- Open Access
Base excess and lactate as predictors of mortality in medical ICU patients
© BioMed Central Ltd. 2004
- Published: 15 March 2004
- Multivariate Analysis
- Univariate Analysis
- Discriminatory Power
- Base Excess
Base excess (BE) and lactate (LAC) have been used to monitor ICU patients. Each results from different pathophysiological derangements in perfusion, inflammation and renal function. The individual significance of BE or LAC to predict the outcome of the critically ill patients is still uncertain and was the focus of this retrospective study.
We retrieved 333 patients from our prospective collected database from January to December 2000. Age, diagnosis, APACHE II score, BE and LAC at entrance and after 24 hours of admission were recorded. Univariate and multivariate analyses were performed, the former being based on a matrix of collinearity (Pearson coefficient ≥ 0.4 denotes collinearity) and on the results of the univariate analysis. A receiver–operator characteristic (ROC) curve was built to identify the best predictive value for mortality.
The age was 51 ± 18 years, APACHE II was 21 ± 1, BE and LAC at admission were -6.0 ± 7.6 mmol/l and 4.9 ± 9.7 mmol/l, respectively, and the BE and LAC after 24 hours were -5.5 ± 6.2 mmol/l and 5 ± 9.6 mmol/l, respectively. The variations of BE and LAC were calculated as the 24 hours value minus the admission value, and resulted in 0.4 ± 6.6 and 0.03 ± 5.6 mmol/l, respectively. The univariate analysis showed mortality correlation to the APACHE II score (odds ratio [OR] 1.114 [1.081–1.148]), BE at entrance (OR 0.995 [0.991–0.998]) and at 24 hours (OR 0.989 [0.985–0.993]), and LAC at entrance (OR 1.000 [1.000–1.001]) and at 24 hours (OR 1.000 [1.000–1.001]). The multivariate analysis was performed with APACHE II (OR 1.099 [1.064–1.135]), BE after 24 hours (OR 0.994 [0.990–0.999]) and LAC at entrance (OR 1.000 [1.000–1.001]). In the next step APACHE II was excluded from the analysis due to the collinearity with BE after 24 hours and to individualize the value of LAC and BE as a death predictor, and the final result was BE after 24 hours (OR 0.990 [0.986–0.994], P < 0.001) and LAC (OR 1.000 [1.000–1.001], P = 0.057). The ROC curve resulted in a area under the curve of 0.58 to LAC at admission and of 0.65 to BE after 24 hours, with the value of -4 mmol/l showing a sensitivity of 70% and s specificity of 50% to predict death.
Lower values of BE and higher values of LAC are associated with poor prognosis in ICU patients. In a comparative analysis of these two variables, BE measured after 24 hours of admission has the best discriminatory power to predict outcome.