4. Please confirm that for the ‘investigator’ assessment of PFS, assignment of the overall response category was not made by the investigator, but was ‘independent of the overall response category provided by the investigator’. Who – sponsor, DMC or other agent – assigned the overall response category for the purpose of investigator-assessed PFS and related investigator-assessed outcomes?
5. Please describe the level of concordance between the investigator’s overall response category and the overall response category designated by the sponsor / DMC / other agent based on lesion measurements / assessments provided by the investigator. For example, how many patients per arm had discordance between an objective response and stable disease, or between stable disease and progressive disease?
For final analysis of PFS, there were two assessments: one in the ITT population, and one in the subgroup of patients given palbociclib with food (‘mITT2’), i.e. those patients randomised after 21 January 2014 when guidance about administration of palbociclib with food and prohibition of PPIs had been communicated to study sites – elsewhere in the CSR described as a subpopulation characterised by optimal exposure to palbociclib.
Requirements for a positive study
The CSR states:
This study was considered a positive study if the 1-sided, stratified log-rank test for PFS based on randomization stratification factors is significant at the significance level of 0.000013 at the interim analysis or 0.025 at the final analysis in favor of palbociclib plus letrozole combination.
Also of note:
OS was to be hierarchically tested for significance at the time of PFS analyses, provided the primary endpoint, PFS, was statistically significant at the interim and/or final PFS analyses. If OS did not yield a statistically significant result at the interim analysis, OS will be statistically evaluated at the final OS analysis. However, if PFS was not significant at the interim and/or final PFS analyses, OS would not be statistically evaluated.
Regarding OS, 1-, 2- and 3-year survival probabilities were also to be calculated.
The study was designed to have an interim analysis after ~226 PFS events and a final analysis at 347 PFS events. An interim analysis of OS was also pre-specified, and to be performed at the time of the interim analysis of PFS and of the final PFS analysis. A reasonable time for the OS interim analysis was estimated to be at 131 deaths, at the estimated time for the planned PFS final analysis. If OS was not significant at the time of the final PFS analysis, a final OS analysis was to be performed after 390 deaths.
666 women were randomised: 444 to palbociclib + letrozole; 222 to placebo + letrozole. All randomised patients were treated.
Disposition at end of treatment is shown below:
Table 52: Disposition at the end of treatment
Comment: There was a large imbalance in the percentage of patients who discontinued due to objective progression or relapse: 38.7% (palbociclib-containing arm) versus 56.3% (placebo-containing arm).
Disposition at end of study is shown below:
Table 53: Disposition at the end of study
Comment: This reflects that many patients died after the ‘end of treatment’, i.e. after 28 days after the last dose was received, so deaths were not considered ‘on treatment’.
The numbers of deaths reported here for ‘end of study’ (n=94 deaths in the palbociclib arm, n=38 in the placebo arm) correspond imperfectly with the OS outcome data for PALOMA-2, where n=95 and n=38 respectively died (from start of treatment onwards).
Major protocol violations/deviations
Frequency of major protocol deviations was high. For example, 21.2% of palbociclib + letrozole arm patients were given prohibited concomitant medications during active treatment, versus 12.2% of placebo + letrozole arm patients. 58-59% of patients per arm had some form of deviation from informed consent.