median PFS was 26.1 months (95% CI: 11.2, NR) in the palbociclib plus letrozole arm and 5.7 months (95% CI: 2.6, 10.5) in the letrozole alone arm.
By investigator assessment:
HR 0.508 (95% CI: 0.303, 0.853; stratified 1-sided p=0.0046) in favour of the combination arm;
median PFS was 18.1 months (95% CI: 13.1, 27.5) in the palbociclib plus letrozole arm and 11.1 months (95% CI: 7.1, 16.4) in the letrozole alone arm.
Comments: the median PFS for the letrozole alone arm in Part 1 is nearly half that of the same arm in the second cohort, and the median PFS for the palbociclib arm is much longer in the Part 1 compared with the same arm in Part 2 of the study.
Thus the Part 1 results suggest a treatment effect (with non-overlapping confidence intervals) and influences the statistical significance attributed to the treatment effect for the whole population (see Figure 5).
Any influence of the biomarkers used for selection of the Part 2 cohort on PFS in either arm remains unclear.
It is difficult when there have been so many data-driven protocol amendments to be confident in the value of statistical analyses performed under such conditions and their ability to demonstrate a true treatment effect.
Figure 5: Study A5481003 Kaplan-Meier plot of progression-free survival in the Part 1 and Part 2 cohorts of the Phase II study by Investigator assessment (ITT population)