Assessing generalizability of trial results in general practice

This editorial refers to ‘XANTUS: a real-world, prospective, observational study of patients treated with rivaroxaban for stroke prevention in atrial fibrillation’, by A.J. Camm et al ., on page doi:10.1093/eurheartj/ehv466. The gold standard for assessing treatment effects is the randomized clinical trial. Providers are fortunate to have a robust set of trials with >80 000 participants showing important benefits and reassuring safety of the non-vitamin K oral anticoagulants (NOACs) relative to warfarin for stroke prevention in atrial fibrillation (AF).1–4 Based on the consistent findings of these trials, providers can be confident that NOACs will be safe and effective if prescribed and taken in a similar way and by patients comparable with the trials. However, unselected patients differ from those in clinical trials, and the way treatments are prescribed and taken in practice may differ from trials. In a general practice setting, a group of older patients started on warfarin had higher rates of bleeding and of warfarin discontinuation than patients in the recent randomized trials, and the time in the therapeutic range on warfarin was lower.5 The Food and Drug Administration (FDA) received reports of serious and fatal bleeding events with dabigatran within several months of the medication becoming available in general practice, and the question was raised as to whether dabigatran might be less safe than was observed in the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial.6 These concerns have been allayed through high-quality observational studies of unselected populations.6–8 It is important, however, to be cautious when estimating treatment effects from observational studies, since there is always residual unmeasured confounding. Even using the same Medicare data …