Article type
Year
Abstract
Background: Randomised trials that deviate from intention-to-treat approach are often encountered in meta-analyses. This type of trial also is more likely to report exclusions than trials that declare standard intention-to-treat.
Objectives: To assess differences in the types and the characteristics of exclusions between these two categories of trials.
Methods: We used a meta-epidemiological sample of 50 meta-analyses and 310 randomised trials (322 comparisons). Trials were classified based on reporting of the intention-to-treat: ITT trials (trials reporting a standard intention-to-treat), mITT trials (trials reporting a deviation from intention-to-treat), no-ITT trials (trials reporting no intention-to-treat). In addition, based on the exclusions, trials were classified as: a) no apparent exclusions; b) exclusions without further description; c) exclusions related to deviation from protocol; d) exclusions related to missing data.
Results: The distribution of the exclusions among the three groups of trials is shown in the Table. The proportion of trials with exclusions, without providing reasons, was higher in the mITT category than in the ITT trials (27% vs 20%; adjusted odds ratio (OR) 3.2, 95% confidence interval (CI) 1.4 to 6.9). This also held for the proportion of trials that declared exclusions that could be ascribed as deviation from protocol (mITT 30% vs ITT 17% respectively; adjusted OR 3.6, 95% CI 1.6 to 8.1) and the proportion of trials with exclusions that were attributed to missing data (mITT 23% vs ITT 9%; adjusted OR 5.8, 95% CI 2.2 to 15.1).
Conclusion: Post-randomisation exclusions are specific to trials that deviate from intention-to-treat and comprise deviation from protocol and missing data, which are two well-known problems often associated with randomised trials. However, a substantial amount of mITT trials (24%) did not report exclusions and another 28% did not provide information about the type of exclusions. ITT trials reported more exclusions related to protocol deviation than exclusions related to missing data.
Objectives: To assess differences in the types and the characteristics of exclusions between these two categories of trials.
Methods: We used a meta-epidemiological sample of 50 meta-analyses and 310 randomised trials (322 comparisons). Trials were classified based on reporting of the intention-to-treat: ITT trials (trials reporting a standard intention-to-treat), mITT trials (trials reporting a deviation from intention-to-treat), no-ITT trials (trials reporting no intention-to-treat). In addition, based on the exclusions, trials were classified as: a) no apparent exclusions; b) exclusions without further description; c) exclusions related to deviation from protocol; d) exclusions related to missing data.
Results: The distribution of the exclusions among the three groups of trials is shown in the Table. The proportion of trials with exclusions, without providing reasons, was higher in the mITT category than in the ITT trials (27% vs 20%; adjusted odds ratio (OR) 3.2, 95% confidence interval (CI) 1.4 to 6.9). This also held for the proportion of trials that declared exclusions that could be ascribed as deviation from protocol (mITT 30% vs ITT 17% respectively; adjusted OR 3.6, 95% CI 1.6 to 8.1) and the proportion of trials with exclusions that were attributed to missing data (mITT 23% vs ITT 9%; adjusted OR 5.8, 95% CI 2.2 to 15.1).
Conclusion: Post-randomisation exclusions are specific to trials that deviate from intention-to-treat and comprise deviation from protocol and missing data, which are two well-known problems often associated with randomised trials. However, a substantial amount of mITT trials (24%) did not report exclusions and another 28% did not provide information about the type of exclusions. ITT trials reported more exclusions related to protocol deviation than exclusions related to missing data.