Randomisation by interactive randomisation technology (IRT) following receipt and approval by the sponsor of the study documentation including patients meeting the above criteria and written informed consent, will occur using the stratification factors described below.
The IRT assigned a unique patient identification number and also assigned study medication. If a patient does not receive the correct study treatment for their allocated treatment arm, the reason was to be clearly documented in CRF.
Blinding codes should only be broken in emergency situations for reason of patient safety. Blinding codes may also be broken after a patient discontinues treatment due to disease progression, as determined by the treating investigator using RECIST v.1.1 criteria, but only if deemed essential to allow the investigator to select the patient's next treatment regimen and after discussion and agreement with the sponsor. Code should not be broken in the absence of emergency situations or progressive disease as per RECIST v.1.1 (for example, in case of clinical deterioration, increase in tumour markers or any other evidence suggestive of disease progression but in the absence of RECIST-defined disease progression). When the blinding code is broken, the date and reason for unblinding must be fully documented in source documents and entered on the case report form. However, every effort should be made by the site staff to ensure that the treatment arm in which the unblinded patient is assigned is not communicated to any sponsor personnel or designee involved in the conduct of the trial.
Intent-to-Treat Population (ITT)
The ITT population was to include all patients who are randomised, with study drug assignment designated according to initial randomisation. The ITT population will be the primary population for evaluating all efficacy endpoints and patient characteristics.
As-Treated Population (AT)
The AT population was to include all patients who receive at least 1 dose of study treatment (ie,
PD-0332991/placebo or letrozole/placebo), with treatment assignments designated according to actual study treatment received. The AT population will be the primary population for evaluating treatment administration/compliance and safety. Efficacy endpoints may be assessed in this population as well.
All efficacy analyses will be based on intent-to-treat (ITT) population. Some efficacy analyses will also be performed on the AT population.
At least 650 patients will be randomised 2:1 between the experimental arm (Arm A: at least 433 patients treated with PD-0332991 plus letrozole) and the control arm (Arm B: at least 217 patients treated with placebo plus letrozole).
Patients will be stratified by:
site of disease (visceral versus non-visceral)
disease-free interval since completion of prior (neo)adjuvant therapy:
the nature of prior (neo)adjuvant anticancer treatment received
The sample size for this study is determined based on the assumptions that the median PFS for patients with advanced/metastatic breast cancer receiving placebo plus letrozole in the first line treatment setting is 9 months and a risk reduction by 31% (hazard ratio of 0.69) or an improvement by 44% to a median PFS of 13 months in the palbociclib plus letrozole treatment arm is clinically meaningful. A total of 347 events are required in the 2 arms of the study based on a 2:1 randomisation to have 90% power to detect a difference assuming a true hazard ratio of 0.69 in favor of the palbociclib plus letrozole arm using a one-sided log-rank test at a significance level of 0.025. Assuming a 15% drop-out rate on either treatment arm, a non uniform accrual accomplished over a 15-month period and follow-up that will continue for about 10 months after the last patient is enrolled, a total sample size of approximately 650 patients (approximately 433 patients in the palbociclib plus letrozole arm and approximately 217 patients in the placebo plus letrozole arm) is required.
The sample size described above will also allow the assessment of differences in the secondary endpoint of overall survival (OS) with a high level of significance. The OS outcome of a Phase III clinical trial in a similar patient population demonstrated a median OS of 34 months for the arm receiving letrozole. Using this value as an assumption with a hypothesized 26% risk reduction (a hazard ratio of 0.74) or 35% improvement in median OS (from 34 months to 46 months) in patients randomised to receive PD-0332991 plus letrozole and a follow-up period of approximately 68 months, evaluation of 390 events using a one-sided log-rank test is required for a significance level of 0.025 and power of 80% to detect a difference. OS will be hierarchically tested for significance at its interim analysis, provided the primary endpoint, PFS, is statistically significant at the interim PFS analysis, or at the final PFS analysis.
Comment: The ability to detect an effect on OS will be affected by the following:
numerous lines of therapy, including some with proven effect on OS are available for use upon progression;
although switching to palbociclib was not permitted within the study, it is approved for second line usage in combination with fulvestrant in the US;
The Statistical Analysis Plan version 2 dated December 2014 was provided but states on page 6 ‘The SAP is being amended to reflect the changes in Protocol Amendments 2 (January 3, 2014), 3 (March 21, 2014), 4 (September 18, 2014), and 5 (December 2 2014).’
The Protocol Amendments documents provided in this submission included Protocol Amendments 6 and 7, and any resulting changes to the SAP have not yet been made. All information about the study conduct and statistical analysis is taken from Protocol amendment 7.
Both the date of the SAP document and this first statement imply that required amendments to the SAP are not yet complete for Amendments 2-5, and Amendments 6 and 7 are not discussed at all in this document. Amendments 6 and 7 pertain to collection of data beyond progression and are thus unlikely to affect the sample size for the primary endpoint analysis. However, Amendments 2- 5 fundamentally affected the study design and conduct, including amongst other changes, an alteration to the number of patients to be recruited and changes to the interim analysis efficacy boundary. The impact of such changes not yet finalized require a more up to date, finalised version of the SAP to be provided with a clear indication as to what has been changed compared with the currently available SAP (as was done for Study A5481003 updated SAP).
The SAP states the study is designed to have one interim analysis and the final analysis based on the primary endpoint of PFS. A formal efficacy stopping boundary (Haybittle-Peto) for rejecting the null hypothesis will be used for the interim analysis. The purposes for the interim analysis are to allow early stopping of the study for futility or efficacy, to assess safety of the combination regimen, and to potentially adjust the sample size. The interim analysis will be performed after approximately 226 patients have documented progressive disease or die (approximately 65% of the total events expected). If the value of the test statistic exceeds the Haybittle-Peto efficacy boundary (z >4.2059, p <0.000013), the trial may be stopped for efficacy. Under exponential distribution assumption, this boundary equates to a hazard ratio of ~0.55 or smaller in favor of the palbociclib plus letrozole arm versus the letrozole alone arm. Alternatively, as appropriate, the sample size of the study may be adjusted using the method outlined by Cui et al. If the results of the interim analysis indicate serious safety concerns, the sponsor will communicate with the Health Authorities regarding stopping the clinical trial.
An interim analysis for efficacy is also planned for the secondary endpoint of OS. The analysis will be performed at the time of the interim or final PFS analyses if the primary endpoint PFS analysis is positive. The nominal significance levels for the interim and final analyses for OS will be determined by using the Lan-DeMets procedure with an O’Brien-Fleming stopping rule. The overall significance level for the efficacy analysis of OS will be preserved at 0.025 (one-sided test). OS will be hierarchically tested for significance at the time of PFS analyses, provided the primary endpoint, PFS, is statistically positive at the interim or final PFS analyses. If OS does not yield a significant result at these analyses, OS will be tested at the final OS analysis. If PFS is not significant at the interim and/or final PFS analyses, OS will not be statistically evaluated.
Between 28 February 2013 and 29 July 2014, 666 women were randomised at 186 sites in 17 countries (Australia, Belgium, Canada, France, Germany, Hungary, Ireland, Italy, Japan, Republic of Korea, Poland, Russia, Spain, Taiwan, Ukraine, United Kingdom, and United States of America).
444 patients were randomised to the palbociclib plus letrozole arm, and 222 patients were randomised to the placebo plus letrozole arm.
The study is ongoing and the data cut-off date is February 26, 2016 and report date is 21 April 2016.
On 12 September 2015, the External Data Monitoring Committee recommended that the study be continued as planned after reviewing results from the prespecified interim analyses of efficacy and safety. The number of PFS events for the interim analysis was 236, which represented about 68% of the expected events for the study. It is stated the sponsor accepted the E-DMC recommendation but remained blinded to the results of the interim analysis.
Major protocol violations/deviations
No data or summary information was presented specifically addressing these issues.
The clinical evaluator has identified that for at least 6% of patients (established from the difference in numbers between subgroup who relapsed >12 months after completion of adjuvant endocrine therapy as determined at randomisation and by CRF information), there was discordance between the baseline information provided at randomisation and that subsequently recorded on the CRF. As these were stratification factors, this implies that patients were not stratified correctly which then makes establishment of balance in each arm difficult and potentially compromises the accuracy of any subgroup analyses where discordance is noted. It also introduces uncertainty regarding which is the more accurate dataset.
It is noted that the presentation on the updated PFS outcomes of this study at the annual meeting of the 2016 American Society of Clinical Oncology, that Dr Richard Finn presented analyses using the CRF-derived population.
No information is presented about investigator audit sites.
Table 19: Study A5481008 Summary of demographic and baseline characteristic by treatment (ITT population)
In the above table, the distribution across the arms is well-balanced for the factors presented, except for the following prognostic factors:
ECOG 0 (45.9% control arm versus 57.9% for the palbociclib and letrozole arm): +12% favouring the treatment arm;
>3 disease sites (46.9% control arm versus 42.5% for the palbociclib and letrozole arm): -4.4% favouring the treatment arm;
≥ 65 years of age: 36.5% control arm versus 40.8% for the palbociclib and letrozole arm); +4.3% in favour of the treatment arm (identified in Study 1003 as predicting a poorer prognosis).
A review of the tables identified some inconsistencies across the tables for example, with respect to the rates presenting with Stage IV disease at diagnosis compared with de novo metastatic disease. Further discrepancies are noted between the stratification factors based on randomisation and as reported in the CRF:
Reported de novo metastatic disease rates
From Table 220.127.116.11.1 31.5% for the total population (‘Stage IV at initial diagnosis’) – the apparent discrepancy from other rates is not corrected by adding in locoregional advanced disease
From Table 18.104.22.168 33% for total population from randomisation data
37.3% according to the CRF (a discrepancy of 19 patients in the treatment arm and 7 in the control arm cf randomisation data)
The sponsor is requested to state which reported rate was used for primary efficacy analysis of the data, and how such discrepancies are handled in the statistical analysis. (Clinical Questions)
The treatment arms were balanced with respect to prior surgeries, radiation, systemic therapies, and prior aromatase usage in adjuvant setting (48.1% of total population).