Content and quality of currently published phase II cancer trials

A number of statistical methods have been proposed for the design and analysis of phase II studies based on dichotomous outcomes. To investigate to what extent such methods are in current use, we conducted a survey of published studies.

We identified studies by conducting a computerized literature search of MEDLINE. We considered trials on systemic antineoplastic treatments described as phase II or pilot. The search was limited to articles written in English and published in 1997.

Three hundred eight trials were identified. The majority, ie, 295 (95.8%), had been conducted as single-arm studies, with objective tumor response as the primary efficacy end point. An identifiable statistical design was reported for only 58 (19.7%) of these trials. The quality in reporting the statistical design and compliance with the design in carrying out the study or results interpretation were frequently poor. The frequency of reporting the statistical design was not shown to increase over the years of study start and was not associated with sample size or study duration. Instead, a significant association was found with trial results (which were less frequently positive among studies with a statistical design) and with the impact factor of the publishing journal.

This survey shows that the quality of the statistical component of published phase II cancer trials is generally poor and raises suspicion that low quality is likely to bias study findings. Journals might improve the methodologic standard of published articles through a more vigilant reviewing policy.