Article type
Year
Abstract
Background: Trialists do not always report how they analyzed categories of participants that might have missing outcome data (e.g. those who withdraw consent, non-compliers) resulting in challenges in addressing missing participant data (MPD) in meta-analyses including those trials.
Objectives: To describe how, in a sample of RCTs included in SRs, trialists reported on the analysis of categories of participants that might have MPD.
Methods: We surveyed all trial reports included in 50 Cochrane and 50 non-Cochrane SRs published in 2012 and reporting a statistically significant pooled effect estimate for a dichotomous patient-important efficacy outcome. We followed standard systematic review methodology. We focused on 19 categories of participants that could potentially have MPD (Table 1). We considered participants as potentially having MPD if they fell into any of these categories and trialists explicitly reported that they were not followed-up, or if it was unclear whether they were followed-up. We analyzed: 1) how trialists analyzed those participants and 2) whether they justified the method used.
Results: Out of 653 included RCTs, 400 mentioned one or more of the categories of potential MPD. We analyze here the 1202 instances in the 400 trials in which trialists mentioned the categories in Table 2. With regard to those 1202 instances, the trials did not report any handling method of MPD for 64%; reported using complete case analysis in 34%; and reported using another specific method for 2%. The trials presented a justification for their approach for addressing possible missing data in less than 2% of reports.
Conclusions: Trialists' reporting of how they analyzed categories of participants that might have MPD is suboptimal. The most commonly reported approach was complete case analysis.
Objectives: To describe how, in a sample of RCTs included in SRs, trialists reported on the analysis of categories of participants that might have MPD.
Methods: We surveyed all trial reports included in 50 Cochrane and 50 non-Cochrane SRs published in 2012 and reporting a statistically significant pooled effect estimate for a dichotomous patient-important efficacy outcome. We followed standard systematic review methodology. We focused on 19 categories of participants that could potentially have MPD (Table 1). We considered participants as potentially having MPD if they fell into any of these categories and trialists explicitly reported that they were not followed-up, or if it was unclear whether they were followed-up. We analyzed: 1) how trialists analyzed those participants and 2) whether they justified the method used.
Results: Out of 653 included RCTs, 400 mentioned one or more of the categories of potential MPD. We analyze here the 1202 instances in the 400 trials in which trialists mentioned the categories in Table 2. With regard to those 1202 instances, the trials did not report any handling method of MPD for 64%; reported using complete case analysis in 34%; and reported using another specific method for 2%. The trials presented a justification for their approach for addressing possible missing data in less than 2% of reports.
Conclusions: Trialists' reporting of how they analyzed categories of participants that might have MPD is suboptimal. The most commonly reported approach was complete case analysis.