BACKGROUND: Critical care trials report mortality at different, sometimes arbitrarily defined time-points. In addition to that, meta-analysts are often limited to 2x2 tables reporting mortality as dichotomous outcomes. There is an on-going debate among methodologists whether it is appropriate to pool mortality estimates from trials that used mortality outcomes ascertained at different time-points. If the relative effects vary over time, which might especially be the case in critical care, standard pooling would not be appropriate. However, limiting meta-analysis to only specific time-points, or performing separate analyses, would lead to lower sample size and thus lower precision, therefore thwarting the motives and reducing one of the main strengths of meta-analyses. AIM AND METHODS: In a four step approach, we aimed to 1.) describe the current practice of mortality definitions from a representative sample of studies; 2.) simulate the influence of mortality time points on pooled effect estimates; 3.) analyze the influence of mortality time points on pooled effect estimates in actual meta-analyses; and 4.) identify characteristics of meta-analyses which could modify such effects. RESULTS: By analyzing 106 studies on 63,713 patients, a total of 24 different kinds of reported time points were identified. Multilevel mixed effects linear regression and meta-regression showed no influence of time-points on pooled effect estimates. By simulating a total of 25,200 studies with varying characteristics, no effect of different pooling methods was found on point estimates. Using methods limiting the number of eligible studies, however, decreased precision and the likelihood of significant pooled results. By recalculating all 80 Cochrane meta-analyses in the field of critical care using 4 different methods each, these findings were confirmed, and no relevant confounding or interaction by critical care-specific study-characteristics was found. SUMMARY: In contrast to frequently observed practice, our findings advocate the use of all available mortality data. We furthermore provide practicioners of meta-analysis with a graphical tool to predict whether a certain method of pooling would change their findings, based on anticipated magnitude and precision of effect estimates.