The rate of achieving a PR or SD≥ 24 weeks, but not CR as no CRs were recorded, was higher in the palbociclib and fulvestrant arm compared with the placebo and fulvestrant arm: 66.6% (95% CI: 61.3-71.5) versus 39.7% (95% CI: 32.3-47.3) based on investigator assessment. The odds ratio was 3.05 (95% CI: 2.07-4.61) also in favour of palbociclib plus fulvestrant with a 1-sided p-value of <0.0001.
Comment: ‘CBR’ is defined as being a CR, PR or SD for >24 weeks. As no CRs were achieved in the palbociclib and fulvestrant, the PI should clearly state this in conjunction with any statement made about the CBR rate and its statistical significance. This clarification in the PI thus avoids overstating the apparent benefits of treatment with palbociclib and fulvestrant from this trial (especially as there were 3 CRs in the fulvestrant arm with only half the number of patients enrolled).
Duration of response (investigator-assessed)
None of the results from the 3 time points provided was statistically significant, with overlapping 95% confidence intervals, but as the data have matured, the difference between the 2 arms has continued to decrease:
23 October 2015 data cut-off date
median DOR was 10.4 months (95% CI: 8.3-NE) in the palbociclib plus fulvestrant arm and 9.0 months (95% CI: 5.6-NE) in the placebo plus fulvestrant arm.
16 March 2015 data cut-off date
median of 9.3 months (95% CI: 4.0, not estimable) in the palbociclib plus fulvestrant arm versus 7.6 months (95% CI: 5.5, 9.3) in the placebo plus fulvestrant arm.
5 December 2014 study cut-off:
median of 9.3 months (95% CI: 4.0, not estimable) in the palbociclib plus fulvestrant arm versus 5.7 months (95% CI: 3.7-5.7) in the placebo plus fulvestrant arm
there is a progressive narrowing of the difference between the two arms over time indicating the value of mature clinical data.
the investigator-assessed duration of response was not statistically significant (overlapping 95% confidence intervals) in either Study 1008 or Study 1023. This would suggest that adding in palbociclib to existing endocrine therapies potentially improves response rates, but not the durability of that response.
All data for these variables are from the main CSR with a data cut-off date of 5 December 2014. Completion rates were calculated on the percentage of patients who completed at least 1 question from baseline to cycle 14.
Furthermore, the following was specified in the SAP as to the handling of missing values in PROs:
‘For QLQ-C30 and QLQ-BR23, if at least half of the constituent items for the multi-item functional or symptom scale have been answered, then the score for that scale may be pro-rated based on the non-missing items.’
Comment: Completing just one question (or even 15-20/30) is a very low requirement to be eligible to be included in the completion rate, and where it is not complete extrapolations from other parts of the questionnaire have been deemed significant for those who have completed 15 or more responses. It is not possible to determine how many study participants have completed sufficient questions to allow a meaningful evaluation of the resulting data, nor for how many sponsor has ‘pro-rated’ the data. The value of such ‘pro-rated’ data is uncertain as the very nature of these values is that they are subjective. Further clarification is only being required for participation rates and potential impact of pro-rating if the sponsor wishes to pursue retention of the information in the PI, as this evaluator’s analysis of the data does not support the PI claims currently made – see below, Clinical Question.
Higher scores indicate a positive effect on quality global and functioning in the scales for this assessment, while higher symptom scores indicate worsening symptoms. The SAP specifies that a 10-point or higher change in scores from baseline is considered clinically significant.
Completion rates were calculated on the percentage of patients who completed at least 1 question from baseline to cycle 14.
96.9% to 100% in the palbociclib plus fulvestrant arm and 95.8% to 100% in the placebo plus fulvestrant arm
The estimated difference in overall change from baseline score for global QOL was 3.1 (95% CI: 0.3, 6.0). The sponsor used the 95% CI to determine there was a statistically significant difference in favour of the investigational treatment.
The CSR states: ‘The difference between the two treatment arms in change from baseline scores for emotional functioning was found to be statistically significant (2.7 [95% CI: 1.1, 4.3] versus -1.9 [95% CI: -4.2, 0.5]; p=0.0016) favoring palbociclib plus fulvestrant. The estimated difference in overall change from baseline score for emotional functioning was 4.6 (95% CI: 1.7, 7.4).
None of the values for change from baseline for the 9 symptoms assessed (fatigue, nausea and vomiting, pain, dyspnea, insomnia, appetite loss, constipation diarrhea, and financial difficulties) reached the required change from baseline to be considered of clinical signficance.
Comment: The 95% confidence intervals for all the functional scales overlap 1, and the change in global health status is only 3.1 points (Figure 11 of the CSR). No parameters reach the SAP’s stipulated change of ≥ 10 point change from baseline required to be considered of clinical significance. Accordingly, all the claims in the last paragraph of the Clinical Trials Section of the PI should be removed.
If the sponsor wishes to retain this assessment, a justification against the criteria in the SAP (>10 point shift from baseline) as well as the following information needs to be included in the s31 response: the sponsor is requested to provide the following information: the number of patients who completed all questions of the EORTC-QLQ-C30/the number of patients completing<29 questions. See Clinical Questions
Time to deterioration in pain
A time to event analysis was prespecified for pain. An analysis of TTD in pain defined as time between baseline and first occurrence of increase of ≥ 10 points in pain was carried out based on survival analysis methods using a Cox Proportional hazards model and log rank test.
Median TTD in pain was 8 months (95% CI 5.6, not estimable) in the palbociclib plus fulvestrant arm compared with 2.8 months (95% CI, 2.3, 5.4) in the placebo plus fulvestrant arm.
HR of 0.642 [95% CI 0.487, 0.846]; p <0.001) indicating palbociclib plus fulvestrant significantly delayed TTD in pain symptom compared with placebo plus fulvestrant.
Comment: This prespecified analysis represents a clinically meaningful delay in worsening of symptoms with palbociclib and fulvestrant treatment. Inclusion of a brief statement in the PI would be considered acceptable.
Functional Scale – QLQ-BR23
The EORTC QLQ-BR23 functional scales consist of the 4 scales body image, sexual functioning, sexual enjoyment, and future perspective. The sponsor reports no statistically significant changes occurred as a result of treatment and has no claims in the PI for this assessment.
Symptoms scales - QLQ-BR23
The EORTC QLQ-BR23 symptom scales consist of the 4 symptoms systemic therapy side effects, breast symptoms, arm symptoms and upset by hair loss. No symptom scales reached the prespecified change from baseline.
Comment: It is noted that alopecia was a cause of some distress and this is listed as an AE in the PI for palbociclib and fulvestrant so clinicians will be aware. This should be included in the CMI as endocrine therapies often result in some thinning of hair, but significant hair loss is uncommon.
The Side Effects section of the CMI is poorly written: the wording for possible side effects is very long and does not provide a simple list and clear instructions. Nor does it contain any reference to neutropenia and a risk of infection.
EQ-5D Index and Visual Analog scores
No minimally important difference was pre-specified in the SAP to guide as to a significant change in this instrument.
Comment: In the absence of a prespecified figure for minimal important difference, a treatment effect cannot be established.
EQ-5D Health State profile
The first part of the EQ-5D consists of 5 descriptors of current health state, mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The percentage of patients with extreme problem was low in both treatment arms at baseline and did not change notably from baseline.
EQ-5D Index Scores
The overall EQ-5D index score on treatment was found to be statistically significantly greater in the palbociclib plus fulvestrant arm compared with the placebo plus fulvestrant arm (0.74 [95% CI: 0.72, 0.76] versus 0.70 [95% CI: 0.67, 0.73]; p=0.0308). Similarly, the between-treatment comparison based on change from baseline in overall EQ- 5D index score favored the palbociclib plus fulvestrant arm over the placebo plus fulvestrant arm (0.006 [95% CI: -0.01, 0.03] versus -0.031 [95% CI: -0.06, 0.00]; p=0.0308).
Comment: The 95% confidence intervals overlap for each of sets of results, rendering them not statistically significantly different.
EQ-5D Visual Analog scores
The visual analog scale within the EQ-5D assesses general health status. No significant difference was reported between the treatment arms, although the sponsor reported both groups experienced a statistically significant decrease from baseline.
ER, PR, HER2 from local laboratory results
these were available in 97.7% of the palbociclib and fulvestrant arm and 96% of the placebo and fulvestrant arm;
3 patients had ER-negative, and 2 of these had PR-positive disease; the other was unknown;
70% of the palbociclib and fulvestrant arm and 67.2% of the placebo and fulvestrant arm had PR-positive disease;
HER2 was ultimately confirmed to be negative for all patients except one
ER and PR were assessed by IHC while HER2 expression was first assessed with IHC assay. A FISH test was applied if the IHC results showed 2+ staining (equivocal).
Tissue slides/blocks from 490 of 521 enrolled patients were sent to the central laboratory for analysis. The samples from 87 patients either had no breast tumour cells in the tissue or were not evaluable based on hematoxylin and eosin stain assessment.
224 (64.6%) patients in the palbociclib plus fulvestrant arm and 115 (66.1%) patients in the placebo plus fulvestrant arm were categorized as having ER-positive disease.
Overall, 41 patients were shown to have ER-negative disease, 26 (7.5%) patients in the palbociclib plus fulvestrant arm and 15 (8.6%) patients in the placebo plus fulvestrant arm;
In total, 28 patients were shown to have centrally confirmed ER- and PR-negative disease: 20 patients in the palbociclib plus fulvestrant arm and 8 patients in the placebo plus fulvestrant arm.
By IHC, a total of 166 (47.8%) patients in the palbociclib plus fulvestrant arm and 90 (51.7%) patients in the placebo plus fulvestrant arm were confirmed to have HER2-negative disease;
A total of 72 (20.7%) patients in the palbociclib plus fulvestrant arm had equivocal IHC results (IHC 2+ staining):
10/72 patients were confirmed to have HER2-positive disease by FISH test, respectively; FISH failed in 4 patients and rest were found to have HER2-negativedisease
A total of 34 (19.5%) patients in the placebo plus fulvestrant arm also had equivocal HER2 results:
5 of these 34 patients were confirmed as having and HER2-positive disease by FISH, the FISH test failed in 1 patient and rest were HER2 non-amplified.
Table 17: Study A5481023 Local laboratory review of ER, PR and HER2 status
Progression-free survival and central laboratory results
The investigator-assessed median PFS in the ER-positive subset of the palbociclib plus fulvestrant arm was 9.2 months, which is substantially longer than the median PFS (3.7 months) of the ER-positive subset of the placebo plus fulvestrant arm. Due to only 8 PFS events in the ER negative group, the median PFS was not reached at the time of analysis. The median PFS for the ER negative group was 3.8 months in the placebo plus fulvestrant arm.
For patients in the palbociclib plus fulvestrant arm assessed for PR expression at the central laboratory, the investigator-assessed median PFS was 9.2 months and 5.9 months in the PR-positive subset and PR-negative subset, respectively.
In the palbociclib plus fulvestrant arm, the median PFS of the HER2-negative subset was longer than the median PFS of the HER2-positive subset (9.2 months versus 7.5 months). But there were only 5 PFS events in the 19 centrally confirmed HER2-positive patients.
The median PFS of the HER2-negative group was 3.7 months in the placebo plus fulvestrant arm. The median PFS of the HER2-positive subset was not reached due to the low number of only 2 PFS events.
The inclusion criteria required patients to have advanced or metastatic breast cancer that was ER-positive and/or PR-positive, and HER2-ve. 403/521 patients’ tumours were evaluable for central testing and of these 28 were ER-/PR- and 15 were confirmed to be HER2 amplified with a further 5 uncertain due to failure of FISH testing. These revised outcomes were relatively balanced between the arms but the sponsor’s review of the outcomes (which appear to be based on immature data) do not suggest a balanced outcome.
Thus there have been very significant protocol violations for at least 43 patients out of the 403 (10.7%) who had evaluable samples, subsequently being found to be ineligible to participate. A further 5 remain uncertain due to difficulties with the FISH test for HER2. Missing data prevent an analysis and confirmation of the status of the remaining 118 patients, and these too are a protocol deviation as provision of adequate samples was an inclusion criterion.
No information was found by the evaluator as to how many of the biopsy samples used in the central testing were from a biopsy sample taken following their most recent episode of progression to determine ER/PR/HER2 status and the sponsor is requested to provide this information as it has been shown that a discordant rate between primary and secondary breast cancers has been reported to be as high as 25-30%: Amir et al (2012) reported rates of 16% such discordance for ER and PR status in a prospective study.
ER, PR and HER2 centrally determination has been shown to be discordant between primary and secondary breast cancer in a high proportion of patients, which may be in part due to prior treatments as well as clonal heterogeneity.
A review of the protocol deviations in the CSR does not include these events as a major protocol deviation (none could be found under ‘Inclusion/Exclusion’), yet it would appear more significant than many of the events listed there in terms of potential to affect efficacy. Triple negative and HER2-positivebreast cancers are aggressive subtypes and the former does not respond to endocrine therapy at all, and the latter may if ER-positive but that information is not provided; in any case, failure of prior last endocrine therapy may select for and/or indicate those where HER2 is the predominant driver of cell division and disease progression.
This information is a clinically relevant ‘real world’ issue, where patients have not been treated with the standard of care (for example, anti-HER2 therapy) due to tumour testing issues.
Failure to provide an adequate sample for central testing is also a protocol violation, and does not appear to be recorded as such.
Given the importance of these, both in terms of potential effect and the numbers involved (10.7% of the evaluable population), the evaluator believes these should be regarded as, and included in the protocol violations. The results are too important to be considered ‘exploratory’, especially as central testing was prespecified.
Had central testing been required prior to randomisation then these issues would not have arisen.
A sensitivity analysis, removing all those who were ineligible for enrolment, to determine whether there was any effect on the ITT PFS analysis has not been performed and is requested. Furthermore, although limited by small numbers, an analysis of the PFS in this subgroup should be performed, as well as a further sensitivity analysis excluding those whose data were missing as well as those who were ineligible should also be performed (Clinical Question). These should be presented as a forest plot against the observed PFS outcome for the ITT population.
No information has been provided on the patients whose samples were PR-positive but not ER-positive. The sponsor is requested to provide the numbers in each arm and the outcomes for this small subset as the indication is currently seeking registration in a population described as ‘hormone receptor positive’ which could mean ER-negative/PR-positive.
Table 18: Study A5481023 Central Laboratory review of ER, PR and HER2
Study A5481023 was a randomised, placebo-controlled, double blind study examining the effects of adding palbociclib to fulvestrant compared with fulvestrant alone in women with locally recurrent or metastatic ER-positive and/or PR-positive, HER2- breast cancer who were postmenopausal and whose disease had most recently progressed on an aromatase inhibitor, and also included premenopausal women whose disease had progressed following tamoxifen (with or without ovarian suppression), with ovarian suppression with goserelin required in conjunction with the fulvestrant.
The primary endpoint of investigator-assessed PFS was robustly demonstrated for the ITT population, and the benefit appeared consistent regardless of menopausal status prior to the commencement. While no complete responses were achieved with the combination of palbociclib and fulvestrant, 4 were seen in the fulvestrant alone arm; there was an improvement in the PR and SD rates, with these two figures contributing to the high figure for clinical benefit rate. The OR (made up of PR and CR) was statistically significantly improved in the palbociclib and fulvestrant arm for the first time at the most recent data cut-off (that is, no overlapping 95% confidence intervals).
No improvement in duration of response was observed between the treatment arms (a finding also reported in the first line studies). Overall survival data are too immature for a meaningful analysis, and the study is ongoing and remains blinded to report this at a later time point. The sponsor should provide this information when available.
A blinded independent central review was conducted in 40.5% of the ITT population, selected at random and did not demonstrate within that subpopulation that there was any apparent investigator bias with respect to the PFS outcomes. Extrapolation of this assumption to the entire ITT population is not considered valid, nor any analyses that rely upon partially replaced investigator data for the ITT population with data with BICR assessments; while the chance of bias is considered low for the remaining population, any remaining uncertainty is a consequence of undertaking only a limited independent review process.
The sponsor has been requested to provide additional information and sensitivity analyses after a central review demonstrated discordance between the central and local laboratories in determining ER, PR and/or HER2 status, with 10.7% of those with evaluable samples not meeting the study inclusion criteria as per central review. The sponsor has also been requested to provide a breakdown of the numbers and outcomes for those with ER-/PR-positive disease as this population is currently encompassed by the proposed indication, but no efficacy outcomes are provided specifically for this group. Until this information is provided, any potential recommendation may require that the indication is restricted to those with ER-positive disease only.
Similarly, it is not clear what percentage of patients had local or locoregional recurrence (without metastatic disease), and whether they derived the same benefit as those with metastatic disease. The sponsor has been requested to provide the breakdown of numbers in each arm and efficacy outcomes as this population is identified specifically in the indication. Any recommendation regarding approval in this group cannot be made without this information (note is made that the population should be identified as ‘locally advanced’ not ‘advanced’ in the indication which can encompass distant disease as well, and therefore is used less commonly in Australia).
The evaluation of the patient-reported outcomes was hampered by methodological issues:
the presentation of participation and completion rates (if 1 or more questions were answered, then this was deemed to have been completed);
pro-rating to fill in data in missing responses for those who had answered >50% of the questions in a questionnaire;
either the absence of nominating a minimal important difference or disregard for that value as nominated in the SAP.
The only patient-reported outcome endpoint the evaluator considered established as having both statistical and clinical significance was the time to deterioration in pain scores, which was prolonged in those receiving palbociclib and fulvestrant compared with those receiving fulvestrant alone.
The PI is currently not satisfactory, and requires amendments and updating and re-evaluation once those have been done. This includes updating in line with the latest data cut-off date presented (23 October 2015 cut-off date for the text, figures and tables), taking note of the comments made in regard to the text in the PI comments by this evaluator.