7. In CSR Table 19 using CRF data, the study population is divided by disease free interval (‘since completion of prior (neo) adjuvant therapy’) into: De Novo (n=167 versus 81); ≤12 months (n=99 versus 48); and >12 months (n=178 versus 93). Could the sponsor confirm that by ‘de novo’ in this classification, what is meant is ‘no prior systemic therapy’ presumably equating to no prior neoadjuvant or adjuvant systemic therapy (rather than ‘de novo metastatic disease’)? Reference is also made to CSR Table 15, where n=167 versus 81 had no prior systemic therapies (either chemotherapy or hormonal)38, and to CSR Table 18, where n=139 versus 71 had ‘newly diagnosed’ recurrence type (other categories involving ‘recurrence’ whether locoregional or distant imply earlier diagnosis of breast cancer).
How many patients per arm had no prior systemic therapy for advanced disease despite not having ‘de novo metastatic disease’ on enrolment into the study? Why had these patients not received endocrine or at least systemic therapy at initial diagnosis of ‘advanced’ breast cancer? What was this group’s median length of time since diagnosis? Provide PFS and ORR outcomes per arm for this subgroup. Also provide PFS and ORR outcomes per arm for the subgroup described in CSR Table 18 as ‘newly diagnosed’ (taken by this evaluator to be truly ‘de novo’ advanced disease), based on investigator and BICR assessments (and according to both IMPALA and CRF analyses).
Results for the primary efficacy outcome
The primary efficacy outcome was PFS, as assessed by the investigator, at the final (PFS) analysis in the ITT population. PFS outcomes are summarised in Table 6 on page 54; the median PFS was 24.8 months in the palbociclib + letrozole arm, and 14.5 months in the placebo + letrozole arm, with a PFS hazard ratio (HR) of 0.58 (95% CI 0.46-0.72). The Kaplan Meier (KM) curve is shown below:
Subgroup analyses presented a consistent picture of investigator-assessed PFS benefit, as did 13 sensitivity analyses (including 4 addressing bone-only disease). While the degree of PFS benefit might have varied by subgroup, no particular subgroup analysed by the sponsor showed any dramatic loss of efficacy. For example, for patients in the sponsor-defined ‘de novo metastases’ subgroup, PFS HR was 0.674: ‘less impressive’ than for recurrent subgroups (HR 0.501-0.516) but still delivering PFS benefit. There were two sources of information used to define subgroups: IMPALA and CRF39. The sponsor concluded that CRF was more accurate. Subgroup analyses are based on CRF information; but for subgroup analysis of investigator-assessed PFS, there were no large discrepancies between IMPALA and CRF –based analyses.
The PALOMA-2 CSR explained that primary efficacy assessment of PFS would be based on investigator assessment.
The clinical evaluator of the initial Dossier considered that a deficiency of the top-line summary was the absence of BICR outcomes. The full CSR included BICR outcomes. Overall results based on the BICR were broadly similar to those based on investigator assessment, with a PFS HR of 0.653 (95% CI 0.505-0.844), although median PFSBICR was 30.5 months versus 19.3 months.
Subgroup analysis of BICR-assessed PFS was also presented.
According to IMPALA analysis, those with de novo metastatic disease did not benefit from addition of palbociclib (HR 1.05, 95% CI 0.64-1.74; mPFS not reached (palbociclib-containing arm) versus 33.1 months (placebo-containing arm); according to CRF analysis – considered more accurate – the PFSBICR HR was 0.876.
Comment: In Study 1003, there was no indication from subgroup analysis of investigator-assessed PFS that patients with de novo advanced disease did not benefit in terms of PFS from the addition of palbociclib.
While PFS benefit was maintained in subjects with ‘no visceral metastases’ by hazard ratio (HRBICR IMPALA 0.724, 95% CI 0.479-1.093), median PFS was similar across arms, at ~30.5 months (regardless of source of data, IMPALA or CRF), quite different than with investigator-assessed PFS.
Investigator versus BICR outcomes
A comparison was made between PFS assessed by the investigator versus BICR; there were discordant PFS event assessments in 19.4% (palbociclib arm) versus 27.5% (placebo arm), with most of this imbalance due to a higher number of events declared by the investigator but not by the BICR (64/444 palbociclib; 51/222 placebo).
Comment: Although this suggests the possibility of investigator bias, overall outcomes were similarly favourable for palbociclib + letrozole for PFS when based on BICR assessment.
PFS by biomarker status
By way of background, the Clinical Overview provided in Dossier version  notes:
…preclinical cell line studies have observed that cell lines with high levels of CCND1, low levels of the CDK inhibitor 2A (CDKN2A; also known as ‘p16INK4A’), and the presence of the Rb 1 gene were more sensitive to palbociclib treatment. Sensitive cell lines represented mostly the luminal ER-positive subtype.
Discussion in the EMA Second JRAR (Question 9, about the Rb biomarker) is noted, including the statement:
Further biomarker analyses are planned or ongoing with the specimens collected on Study 1008: Cyclin D1 amplification and p16 deletion by FISH test; CDK4, CDK6, CDK2, Cyclin E1, Cyclin E2, p16, and Rb mRNA expression. A standalone biomarker analysis report will be provided by the end of 2017.
The EMA Second JRAR assessment of this issue also concluded that the SmPC should include information pertaining to uncertainty about significance of Rb-status.
There were no robust differences according to biomarker-defined subgroups, although various biomarkers conferred prognostic advantage / disadvantage. There were very few CCND1-negative patients, but in these patients, the PFS HR was 0.997 (the CSR authors considered that sample size precluded valid comparison for CCND1 negative subjects). There were very few ‘p16 high’ patients, but here the PFS HR was 0.255.
Comment: In Study 1003, enrolment was focused on patients with CCND1 gene amplification and / or loss of CDKN2A (since preclinical data suggested that such tumours were particularly sensitive to palbociclib). However, interim data found no influence of these biomarkers. It is interesting that in the very few CCND1-negative subjects in PALOMA-2, palbociclib did not confer obvious benefit.
Results for other efficacy outcomes
Investigator-assessed confirmed responses were reported in 42.1% (palbociclib) versus 34.7% (placebo), as indicated below:
Figure 15: Confirmed complete response or partial response
Confirmed ORRs were lower using BICR assessments (34.7% palbociclib + letrozole versus 23.9% placebo + letrozole).
The above ORRs include patients with no measurable disease at baseline.
Median duration of response (based on investigator-assessed, confirmed ORs) was 22.5 months (palbociclib + letrozole) versus 16.8 months (placebo + letrozole).
A planned OS interim analysis was performed at the time of the final PFS analysis based on 133 deaths (34% of 390 events for final analysis) from 666 patients. Since the pre-specified level of significance was not met, the OS data will be continuously followed for the final analysis when 390 deaths have been observed. The median follow-up time for the palbociclib plus letrozole arm was 23.0 months (95% CI: 22.6-23.4) and for the placebo plus letrozole arm was 22.3 months (95% CI: 21.9-22.9). No OS conclusions can be made due to the immaturity of the data. The patients will continue to be followed for the final OS analysis
A more detailed discussion of OS outcomes is included in the Delegate’s Overview (see Overall Risk-Benefit analysis in AusPAR), but some basic considerations are set out below.
The CSR states (page 175):
The planned interim OS analysis was performed at the time of the final PFS analyses based on 133 deaths (Table 184.108.40.206) from 666 patients (34% of expected 390 total deaths for the final OS analysis) since the primary analysis for PFS was statistically significant. However, the prespecified level of significance for the interim analysis was not met. The patients continue to be followed for survival and the final OS analysis will be performed when 390 deaths have been reported. Of note, at the time of the data cutoff, the median follow-up time in the palbociclib plus letrozole arm was 23.0 months (95% CI: 22.6, 23.4) and in the placebo plus letrozole arm was 22.3 months (95% CI: 21.9, 22.9) (Table 220.127.116.11).
For deaths within 28 days of the last dose, the following comments help explain those deaths categorised as having an ‘other’ cause (CSR):
Cause of death was reported as ‘unknown’ for 1 patient in the placebo letrozole arm only. ‘Other’ was reported for 7 patients in the palbociclib plus letrozole arm and for 1 patient in the placebo plus letrozole arm. ‘Other’ was further specified as follows for the 7 patients in the palbociclib plus letrozole arm: 1 death during the injection of an unknown medicine to relieve shoulder pain in another study site, 1 death possibly related to acute respiratory viral infection, 1 death due to pneumonia/respiratory failure, 1 death due to cardiogenic shock of unknown origin, 1 death due to cardiopulmonary failure, 1 death due to thrombosis of pulmonary arteries with unknown thrombus, and 1 death due to acute transmural myocardial infarction of left ventricle posterior wall. ‘Oher’ for the 1 patient in the placebo plus letrozole arm was chest infection (see Errata).
The EMA’s Second JRAR Clinical report goes into the issue of OS. The relevant section is attached in Section 17 on page 61. The sponsor explained to the EMA that the applicant is blinded to the interim OS analysis.
In this regard, the TGA has adopted an EU guidance document on Data Monitoring Committees which does note that ‘policies to avoid the dissemination of interim study result prior to unblinding have to be in place’. The guidance further states that ‘In case of a submission the working procedures of a DMC as well as all DMC reports (open and closed sessions) should form part of the submission.’ The EU guidance on statistical principles notes, in Section 4.6, that DMC procedures ‘should also address the control of dissemination of interim trial results within the sponsor organisation’ – which does not seem to preclude limited dissemination of key outcomes.
Questions 8-9 for sponsor
8. Provide the interim OS analysis that was performed at the time of the final PFS analysis, and any update. Provide 12- and 24-month OS rates per arm, estimated median OS per arm, an estimated OS hazard ratio, and estimated KM curves for OS.
9. Please comment in detail on the apparent discordance between mature PFS and immature OS outcomes.
The SAP version 3.0 states:
A supportive analysis will be performed by combining the OS data from this study and from the randomized phase 2 Study A5481003 with similar approaches described above. The Study as a stratification factor (A5481003 vs. A5481008) will also be included in the analysis. Since the median OS time for the studied patient population is relatively long and it is anticipated a small fraction of OS events would be available at the time of OS interim analysis, this analysis will certainly increase the power of detecting the OS difference between two treatment arms, given both randomized studies have similar patient populations and OS follow up processes.
This implies that the supportive analysis will be conducted prior to OS final analysis.
Question 10 for sponsor
10. Please provide this supportive analysis of OS outcomes across studies 1003 and 1008.
Comment: In Study 1003, after 61 deaths across 165 patients and median 28-30 months follow-up, estimated median OS in the palbociclib + letrozole arm was 37.5 months versus 33.3 months in the letrozole-only arm. Median follow-up time in Study 1003 (28-30 months) was not too dissimilar from median follow-up time in Study 1008 (22-23 months); however, the proportion of patients who had died was ~37% for Study 1003 versus ~20% for Study 1008. In both studies, OS outcomes are considered immature.
FACT-B and FACT-G baseline scores were similar across arms. No statistically significant differences were observed between treatment arms in overall FACT-B or FACT-G scores, on treatment.
For the breast cancer subscale (BCS, range 0-36, higher score indicating better quality of life or better well-being), baseline scores were similar across arms; no statistically significant difference was observed overall on treatment, although from cycle 13 to cycle 25 a difference in means of 1.6-2.8 was seen, with higher scores in the placebo arm. However, there was no statistically significant difference in time to deterioration for BCS (i.e. time between baseline and first occurrence of a decrease of 2+ points).
11. In PALOMA-3, there was a delay in time to deterioration for pain symptoms, in the palbociclib-containing arm. What outcomes were observed for pain symptoms in PALOMA-2?
12. The clinical evaluator for the main dossier writes that assessment of benefit in patients with bone-only disease relies on measures of quality of life such as improvement in pain, reduced skeletal event rates, etc. What patient-reported outcomes for pain symptoms were observed in patients with bone-only disease?
Follow-up use of systemic therapies
Follow-up systemic treatments for breast cancer were used in 42.1% (palbociclib + letrozole) versus 60.8% (placebo + letrozole), e.g. fulvestrant (13.5% versus 21.6%), capecitabine (12.6% versus 17.6%), exemestane (10.1% versus 18.5%), everolimus (7.4% versus 13.5%), paclitaxel (8.8% versus 14.4%) and letrozole (7.4% versus 10.4%).
Table 56: Follow up of systemic therapies
Analyses performed across trials: pooled and meta analyses
See question above about pooled analysis of OS across Study 1003 and Study 1008.
In PALOMA-2 there was a large 10.3 month improvement in median progression-free survival (PFS) for patients receiving palbociclib + letrozole, relative to those receiving placebo + letrozole. There was a substantial improvement in objective response (CR + PR; 42.1% versus 34.7%) and in clinical benefit (defined as CR + PR + SD; 89.4% versus 77.9%).
Benefits were seen in both investigator and BICR -based analyses. The most prominent discordance between investigator and BICR –based analyses was for subjects with de novo metastatic disease; based on BICR assessment of tumour response, these patients had little if any PFS improvement with the addition of palbociclib.
Overall survival (OS) outcomes are immature; a relatively small proportion of subjects on study have died. More data are needed to understand the impact of palbociclib on OS outcomes.
There were no large or consistent differences across arms in patient-reported outcomes – but PROs can be difficult to assay with sensitivity.
Overall, addition of palbociclib to letrozole appears to have a major anti-tumour impact in this setting, but there is an important unresolved concern about impact on survival.