“Patients fall out because the drug isn’t working well or they’re having a side effect, so over time, you tend to enrich for patients who are doing very well with the as-observed analysis,” the dermatologist explained.
Historically, many industry-sponsored clinical trials reported efficacy outcomes using the as-observed analysis; however, the FDA is increasingly unwilling to accept that approach as the sole analytic method.
At the other extreme is the nonresponder imputation method.
“This is the most stringent statistical package that exists. In fact, when a patient isn’t observed at one of the observation points – for example, at week 8 say the patient has a flat tire and can’t make it to the clinic – they’re counted as a treatment failure. So it’s a very tough statistical package,” according to Dr. Leonardi.
Bruce Jancin/MDedge News
Dr. Craig L. Leonardi
Seventy-seven percent of the initial 1,346 randomized patients in UNCOVER-3 completed 156 weeks of follow-up. To illustrate the importance of paying attention to the details of statistical methodology utilized in reporting efficacy outcomes, he noted that the PASI 75 rate at 156 weeks in study completers on the approved dosing regimen was 97% by the as-observed method, dropping to a still robust 81% by nonresponder imputation. The PASI 90 and -100 rates and static Physician’s Global Assessment (sPGA) results followed suit (see graphic).