Margaret A.WinkerMD, Senior EditorIndividualAuthorPhil B.FontanarosaMD, Senior EditorIndividualAuthor
To the Editor.—Nobody would disagree
with Dr Pablos-Méndez and colleagues1
that nonadherence in a clinical trial adversely affects statistical efficiency.2,3 We believe, however, that these
authors do not adequately consider the impact of nonadherence on the validity
of a randomized trial, and hence draw inappropriate inferences about generalizability.
Since patients who do not comply obtain no benefit, nonadherence in a clinical
trial will produce bias in the estimation of the true efficacy.3- 5
For this reason clinical trials use a variety of useful strategies, including
a run-in period to identify and enroll subjects who will adhere. In the Physicians'
Health Study,6 adherence during the run-in
was the strongest predictor of adherence during the trial so prerandomization
exclusion of nonadherers greatly enhanced validity. A successful run-in period
in a trial enhances validity, which is a prerequisite for generalizability.
Glynn RJ, Buring JE, Hennekens CH. Concerns About Run-in Periods in Randomized Trials. JAMA. 1998;279(19):1526-1527. doi:10-1001/pubs.JAMA-ISSN-0098-7484-279-19-jbk0520