No firm evidence that lack of blinding affects estimates of mortality in randomized clinical trials of intensive care interventions: a systematic review and meta-analysis

Abstract

Objective: To evaluate the effect of blinding on mortality effect estimates in randomized clinical trials (RCTs) in adult intensive care unit (ICU) patients.

Study design and setting: A systematic review and meta-analysis of RCTs reporting mortality effect estimates of ICU interventions in adult ICU patients. We assessed differences in summarized risk ratios with 95% confidence intervals between blinded and unblinded RCTs. P < 0.10 was considered statistically significant (test of interaction).

Results: We included 22 ICU interventions assessed in 269 RCTs enrolling a total of 42,007 adult ICU patients in the primary analyses. We observed statistically significant differences between blinded and unblinded RCTs for all-cause mortality at longest follow-up in one of 22 interventions (5%), for in-hospital mortality in one of 12 interventions (8%) but not for in-ICU mortality. Combining all interventions, unblinded trials reported larger summary effect estimates on all-cause mortality at longest follow-up compared to blinded trials (test of interaction, P = 0.09). However, the difference was not statistically significant following adjustment for other risk of bias domains.

Conclusions: We observed no firm evidence that lack of blinding affects estimates of mortality in RCTs of ICU interventions.

Keywords: Bias; Blinding; Effect estimates; Intensive care; Mortality.