Objective: To understand the impact of patient selection criteria used in recent sepsis trials on baseline mortality risk and number of eligible patients.
Design: Observational cohort study, with retrospective analysis of prospectively collected data.
Methods and main results: Using a MEDLINE search, we selected recent randomized controlled trials in patients with severe sepsis and studied the mortality rate in the control groups of these trials. Nine articles fulfilled the search criteria and were used in our analyses. The 28-day mortality rate in the control groups of these trials varied between 28.0% and 89.0%. Differences in this mortality rate might be due to the use of different entry criteria but also to other factors that vary between the trials. To eliminate the influence of these confounding factors when studying the effect of the use of entry criteria on baseline mortality risk and number of eligible patients, we projected the entry criteria of these nine trials on a large independent database of >70,000 Dutch intensive care patients admitted between 1996 and 2003. This yielded nine groups of patients who would have been eligible for the respective trials. The percentage of patients who would have been eligible for these trials varied between 1.5% and 6.0%. Six of these groups had a similar intensive care mortality rate (between 25.0% and 28.9%). The projection of the entry criteria of the three other trials onto the database resulted in groups of patients with considerably higher intensive care mortality. For in-hospital mortality rate in these groups, similar results were found.
Conclusions: The majority of the trials we studied used entry criteria that select patients with a similar mortality risk. This suggests that differences in baseline mortality risk reported in recent sepsis trials are to be attributed to other factors that vary between trials rather than to differences in entry criteria. However, entry criteria do have an important influence on the number of eligible patients for sepsis trials without influencing baseline mortality rate.