Anion gap as a prognostic tool for risk stratification in critically ill patients - a systematic review and meta-analysis

Abstract

Background: Lactate concentration is a robust predictor of mortality but in many low resource settings facilities for its analysis are not available. Anion gap (AG), calculated from clinical chemistry results, is a marker of metabolic acidosis and may be more easily obtained in such settings. In this systematic review and meta-analysis we investigated whether the AG predicts mortality in adult patients admitted to critical care settings.

Methods: We searched Medline, Embase, Web of Science, Scopus, The Cochrane Library and regional electronic databases from inception until May 2016. Studies conducted in any clinical setting that related AG to in-hospital mortality, in-intensive care unit mortality, 31-day mortality or comparable outcome measures were eligible for inclusion. Methodological quality of included studies was assessed using the Quality in Prognostic Studies tool. Descriptive meta-analysis was performed and the I(2) test was used to quantify heterogeneity. Subgroup analysis was undertaken to identify potential sources of heterogeneity between studies.

Results: Nineteen studies reporting findings in 12,497 patients were included. Overall, quality of studies was poor and most studies were rated as being at moderate or high risk of attrition bias and confounding. There was substantial diversity between studies with regards to clinical setting, age and mortality rates of patient cohorts. High statistical heterogeneity was found in the meta-analyses of area under the ROC curve (I(2) = 99 %) and mean difference (I(2) = 97 %) for the observed AG. Three studies reported good discriminatory power of the AG to predict mortality and were responsible for a large proportion of statistical heterogeneity. The remaining 16 studies reported poor to moderate ability of the AG to predict mortality. Subgroup analysis suggested that intravenous fluids affect the ability of the AG to predict mortality.

Conclusion: Based on the limited quality of available evidence, a single AG measurement cannot be recommended for risk stratification in critically ill patients. The probable influence of intravenous fluids on AG levels renders the AG an impractical tool in clinical practice. Future research should focus on increasing the availability of lactate monitoring in low resource settings.

Prospero registration number: CRD42015015249 . Registered on 4th February 2015.

Fig. 1

Flow chart summarising the search and study selection process. DKA = diabetic ketoacidosis; SOFA = sequential organ failure assessment score

Fig. 2

Forest plot of area under the ROC curves (AUCs) for observed AG predicting mortality. Forest plot of a random effects meta-analysis of AUCs for the observed AG predicting mortality; I² = 99 %. In view of the high heterogeneity a pooled effect estimate is not shown

Fig. 3

Forest plot of odds ratios (ORs) for observed AG predicting mortality. Forest plot of a fixed effects meta-analysis of ORs derived by univariate logistic regression for the observed AG predicting mortality; I² = 0 %. In view of the high heterogeneity in meta-analyses of other effect measures a pooled effect estimate is not shown

Fig. 4

Forest plot of mean differences for observed AG predicting mortality. Forest plot of mean differences in observed AG between survivors and non-survivors; I² = 96 %. In view of the high heterogeneity a pooled effect estimate is not shown