Use of modified intention-to-treat analysis in studies of direct oral anticoagulants and risk of selection bias: a systematic review.

Background: Following their evaluation in randomised controlled trials (RCTs), direct oral anticoagulants (DOACs) have replaced warfarin for stroke prevention in atrial fibrillation (AF), and treatment and prevention of venous thromboembolism (VTE). To avoid selection bias, it is recommended that RCTs use an intention-to-treat (ITT) analysis strategy.

Objective: The objective of this study was to systematically review and compare reported analytical strategies, the proportion of randomised patients included in analyses and the reasons for participant exclusions.

Study Selection: A systematic search of PubMed, EMBASE and the Cochrane library for phase III trials of DOACs was conducted. Titles and abstracts were screened for relevance by two independent reviewers. Patient population, intervention studied, number of patients included in randomisation and analysis, reasons for exclusions from analysis and trial conclusions were extracted from each article.

Findings: Twenty-nine studies were included, five were about stroke prevention in AF, 10 about VTE treatment and 14 about thromboprophylaxis. Trials of AF and VTE treatment had low proportions of postrandomisation exclusions (around 1%). In contrast, surgical and medical thromboprophylaxis trials excluded almost 30% of participants postrandomisation. This was in spite of authors' claims of using an ITT or modified ITT approach. Higher exclusion proportions in these trials were associated with non-clinically defined primary outcomes and incomplete outcome assessments.

Conclusions: Clinicians should be aware that the level of evidence in favour of DOAC use for thromboprophylaxis is weak due to high rates of postrandomisation exclusions and risks of selection bias.