Low-Fat Abiraterone Food Effect Is of Little Consequence
Mené sur 72 patients atteints d'un cancer de la prostate de stade métastatique et résistant à la castration, cet essai international de phase II compare l'efficacité, du point de vue des modifications du taux de PSA, de faibles doses d'acétate d'abiratérone dispensé avec un repas pauvre en graisses, et des doses standard d'acétate d'abiratérone
Szmulewitz et al1 report a phase II noninferiority clinical trial that compared the activity of low-dose abiraterone administered with food (LOW) to the currently US Food and Drug Administration–approved standard-dose abiraterone administered in the fasted state (STD) in patients with castrate-resistant prostate cancer (CRPC). Given that abiraterone commonly is used in the first-line setting of CRPC and that its monthly cost is approximately $10,000, a 75% dose reduction would entail substantial cost savings if the doses achieve equivalent clinical outcomes. The primary end point in this trial was prostate-specific antigen (PSA) change between study entry and 12 weeks of therapy by comparing the log-transformed mean change in PSA between the STD and LOW groups. The LOW group had a mean log change of −1.59 versus −1.19 in the STD group, which is sufficient to establish the LOW group as noninferior to STD within a 90% confidence limit. Although the prespecified noninferiority margin of 15% was met, the clinical significance of the chosen end point is unclear. As the authors acknowledge, log PSA change over 12 weeks in not a clinically validated surrogate end point. Reliance on local PSA testing in this trial is a substantial concern, particularly given the noninferiority trial design the investigators chose. PSA results can vary by up to 20%2 for the same sample, and the Prostate Cancer Working Group has recommended central testing to minimize interassay variability.3 Of note, the 20% variability in locally obtained PSA tests is greater than the predetermined noninferiority margin of 15%. In addition, when clinical trial end points are variable, the bias is toward the null (ie, not detecting a difference). In a noninferiority trial design, failure to reject the null establishes noninferiority of the treatments.4 In other words, assay variability may lead to erroneously determining two treatments to be noninferior. Given the large expected interassay variation in PSA results, a conclusion of noninferiority in this trial may be spurious.(...)