Deteniendo ensayos clinicos para beneficio. Demasiado bueno para ser verdad?

Stopping Trials Early for Benefit — Too Good to Be True?

Be skeptical of results from trials stopped early for benefit.

The practice of stopping randomized clinical trials (RCTs) early for benefit overestimates treatment effects, accrues inadequate data on adverse events, and can lead to misguided treatment recommendations. The number of trials that were stopped early for benefit has increased markedly since 1990 (JAMA 2005; 294:2203). Motives for stopping a trial early include a perceived ethical obligation to offer the treatment to participants and lower trial costs. However, might investigators have less noble motives?

In this systematic review, Italian investigators analyzed 25 RCTs in which new anticancer treatments were tested and that were stopped early for benefit. In 22 studies, the interim endpoint that drove trial truncation was the same endpoint that was proposed for the final analysis. However, the sample sizes used to generate interim efficacy results varied greatly; five trials were stopped when only 43% or less of the planned sample size was reached, and in another five trials, researchers did not provide this information. Of 14 trials published in the last 3 years, 11 were used to support applications for marketing approval by the FDA or a European counterpart. Notably, six trials did not involve data and safety monitoring committees.

Comment: These results affirm the growing phenomenon of trials stopped early for benefit. Furthermore, the authors assert that most of these trial results were used for drug approval, which suggests a commercial motive for stopping early. These findings, coupled with concerns about scientific validity, indicate that clinicians should be skeptical of results of trials stopped early for benefit.

— Paul S. Mueller, MD, MPH, FACP

Published in Journal Watch General Medicine May 1, 2008

Citation(s):

Trotta F et al. Stopping a trial early in oncology: For patients or for industry? Ann Oncol 2008 Apr 9; [e-pub ahead of print]. (http://dx.doi.org/doi:10.1093/annonc/mdn042)