Concordance between independently and investigator-assessed progression-free survival in CLEOPATRA.

Abstract

e11055^ Background: In CLEOPATRA, pertuzumab (P) + trastuzumab (T) + docetaxel (D) was compared to placebo (Pla)+T+D in HER2-positive 1st-line MBC. In MBC trials, PFS rather than overall survival is often defined as the primary endpoint as PFS data are available earlier and later-line therapy does not impact results. The primary endpoint in CLEOPATRA, PFS, was assessed by an Independent Review Facility (IRF), with investigator (inv)-assessed PFS as a secondary endpoint. As a measure of consistency and robustness, concordance between IRF- and inv-assessed PFS was analyzed. Methods: PFS was defined as time from randomization to first documented radiographic evidence of progressive disease (PD) or death from any cause within 18 wk of the last IRF tumor assessment. Tumor assessments took place every 9 wk and continued until IRF-confirmed PD. Treatment decisions were solely based on the inv’s assessment of PD. Indicators of concordance between IRF- and inv-assessed PFS included: PFS event by IRF only; PFS event by inv only; PFS event by IRF and inv; no PFS event by IRF and inv. These indicators were assessed at any time on study, within 30 d or within 9 wk. Several pre-specified sensitivity analyses for PFS were also performed. Results: 808 pt were enrolled and central tumor assessments performed for 781 pt. Compliance with the schedule of tumor assessments was generally high and comparable between arms, with >78% of pt undergoing assessments within 7 d of the scheduled visit over the 1st yr. At any time on study, concordance between IRF- and inv-assessed PFS was 85%. Of the 15% discordant cases, an equal proportion was assessed by the IRF as a PFS event but not by the inv and vice versa. Estimates of the median time to PFS event based on IRF and inv assessments were identical, Pla+T+D: 12.4 mo, P+T+D: 18.5 mo (IRF: HR 0.62, P < .0001; inv: HR 0.65, P < .0001). Conclusions: Concordance between IRF- and inv-assessed PFS was high, supporting consistency and robustness of the PFS analyses. The subjective nature of PD and potential for bias were reduced by the double-blind, placebo-controlled study design; RECIST 1.0 criteria; identical schedule of assessments in both arms; IRF review and continuation of tumor assessments until IRF confirmation of PD.