"3 Randomized Discontinuation Designs in Early Clinical Drug Screening – how to Adapt if the Drug Works too well? The Cabozantinib Experience" Weitzman, et al.
Published Nov 2012 Eoropean Journal of Cancer
Abstract: Cabozantinib (cabo) is a potent, targeted therapy that inhibits MET & VEGFR2. MET & VEGF signaling are implicated in tumor angiogenesis, invasion & metastasis. In the initial Phase I clinical study, treatment with cabozantinib resulted in a high rate of stable disease (SD) across multiple tumor types. In order to broadly evaluate efficacy in a single Phase 2 study, an adaptive randomized discontinuation trial (RDT) design was implemented (Ratain et al. 2006). Nine tumor types were selected for evaluation based on the role of MET & VEGFR in tumor biology. The trial design incorporated stopping rules based on the week-12 disease control (partial response or SD per RECIST) rates for the first 20 patients (pts) in each tumor cohort, & up to 2 cohorts could be fully expanded based upon review of the clinical activity observed. Treatment beyond wk 12 was based on wk 12 response: pts with confirmed response continued open-label cabo, pts with stable disease were randomized to cabo vs. placebo, & pts with progressive disease (PD) were discontinued. The primary endpoints were objective response rate (ORR) per RECIST in the 12 wk Lead-in Stage, & PFS in the Randomized Stage. Response per RECIST criteria was assessed every 6 wks. As originally outlined in the protocol, accrual to any cohort could be halted for either high rates of ORR or PD. Overall, 526 pts were enrolled from September 2009 to July 2011. Pts with progressive measurable disease (RECIST) received 100mg cabo PO qd over a 12 wk Lead-in Stage. Three cohorts-small cell lung, pancreatic, & gastric cancer were not expanded due to low rates of disease control at wk 12. The remaining 6 cohorts (prostate, ovarian, hepatocellular, breast, non-small cell lung, & melanoma) showed encouraging levels of clinical activity that warranted further expansion. The prostate, ovarian, & hepatocellular cancer cohorts showed the highest rates of disease control at wk 12. (cont.)
The unique & unexpected observations of bone scan improvement, pain improvement, & a narcotic-sparing effect in the prostate cancer cohort, as well as high rates of soft tissue regression in the prostate & ovarian cancer cohorts, led to the randomization in these two cohorts being halted. In addition, the protocol was amended to add non-randomized expansion cohorts for both tumor types, including incorporation of a novel primary endpoint of bone scan response in the prostate cohort. This adaptive RDT design provided distinct advantages as a signal-search study, including determination of ORR in 9 well-defined populations. However, certain disadvantages relative to more traditional Phase 2 designs became evident early in the trial, including randomization to placebo of pts who had stable disease by RECIST but were clearly experiencing other measures of clinical benefit. Thus, the success of this trial design depends on the timely retrieval & analysis of data, allowing for the potential suspension of the randomized discontinuation component in select tumor types where a stronger than anticipated efficacy signal is observed.