Palbociclib with an aromatase inhibitor for previously untreated, hormone receptor-positive, HER2-negative, locally advanced or metastatic breast cancer

4.3 The committee noted that the clinical-effectiveness evidence for palbociclib plus letrozole compared with letrozole alone came from 2 studies: PALOMA‑1 and PALOMA‑2. PALOMA‑2 was a larger (666 patients) placebo-controlled, double-blind trial, and PALOMA‑1 was a smaller (165 patients) open-label study. The committee discussed the generalisability of the PALOMA trials to UK clinical practice. It noted that PALOMA‑1 contained no UK patients, but 7 of the PALOMA‑2 sites were in the UK. The committee heard from the clinical experts that both trials had a greater proportion of people with metastatic disease when first diagnosed than is seen in UK practice (37% in PALOMA‑2 compared with about 5% to 10% in UK clinical practice). The committee noted that there was no significant difference in treatment response for people with metastatic disease at first diagnosis and were reassured by the clinical experts that a difference would not be expected. The committee agreed that the populations in both PALOMA trials were similar to the population seen in clinical practice in England. But the committee considered that, because PALOMA‑2 was a blinded larger trial, its results are likely to be more reliable for decision-making. The committee raised concerns that the higher incidence of haematological adverse events in the palbociclib arms of the trials would have resulted in some patients and investigators becoming unblinded to patient allocation during PALOMA‑2. The committee heard that to mitigate this, both investigator-assessed and blinded independent central review (BICR) of progression-free survival was carried out. The committee concluded that the BICR results would be more appropriate for decision-making.

4.4 The committee noted that in the overall intention-to-treat population, the BICR median progression-free survival reported in PALOMA‑1 was 25.7 months for palbociclib plus letrozole, and 14.8 months for letrozole alone. This was reported as statistically significant when using a 1‑sided p value (p=0.0286), but not if a 2‑sided p value had been used. The median overall survival from an interim analysis, which was available at the time of the first committee meeting, was 37.5 months for palbociclib plus letrozole compared with 33.3 months for letrozole alone. This was not a statistically significant difference. In response to consultation the company submitted the final analysis for overall survival from PALOMA‑1, which showed a median overall survival of 37.5 months for palbociclib plus letrozole compared with 34.5 months for letrozole alone. The committee noted a slightly smaller difference in median overall survival gain of 3 months for palbociclib plus letrozole compared with letrozole alone than at the interim analysis (4.2 months), but again there was no statistically significant difference between the treatment arms. The committee concluded that in PALOMA‑1 palbociclib improved progression-free survival, but no significant improvement in overall survival had been shown.

4.5 The committee noted that in the BICR intention-to-treat population, the median progression-free survival was 30.5 months for palbociclib plus letrozole compared with 19.3 months for letrozole alone (hazard ratio 0.653; confidence interval 0.505 to 0.844). The committee heard from the company that overall survival results from this trial are not available because the required number of events has not been reached, and the company remains blinded to the results. The committee concluded that in PALOMA‑2 palbociclib statistically significantly improved progression-free survival, but no data on overall survival are available.

4.6 The committee noted that progression-free survival gains were seen in both trials. But an overall survival gain was seen only in PALOMA‑1 and it was not statistically significant. The committee noted the ERG's comment that final overall survival data from PALOMA‑1 should be considered the best available evidence on overall survival. It also noted the small number of patients in PALOMA‑1, and considered that the results are associated with a wide confidence interval. The clinical experts indicated that they would expect an improved progression-free survival with metastatic breast cancer to result in some benefit in overall survival. However, they judged that the situation is complex and difficult to predict because of the number of further lines of treatment that the person would have, and because the precise relationship between progression-free and overall survival is unclear. The committee agreed that data from PALOMA‑2, when available, will reduce the uncertainty around overall survival gain attributable to palbociclib plus letrozole. However the best available evidence at present is from PALOMA‑1, which showed a non-statistically significant survival gain of less than the progression-free survival gain. The committee concluded that palbociclib has a clear and important benefit in improving progression-free survival, and that it is likely that this would result in some improvement in overall survival. However, it reiterated that the size of this benefit remains uncertain.

4.7 The committee noted that the trial evidence suggested a high incidence of haematological adverse events. It was aware that the marketing authorisation states that full blood counts must be done during treatment, so extra visits may be needed for monitoring. However, it heard from the clinical and patient experts that the adverse events are reversible and manageable. The clinical expert highlighted that many incidences of neutropenia observed in the trials were laboratory findings only and did not result in clinical infections. They expect that in clinical practice, many of these people will continue having palbociclib. In the trials people developing neutropenia may have discontinued treatment, because of protocol restrictions. The committee concluded that, although the incidence of neutropenia in particular was high, the adverse events were manageable and in clinical practice treatment discontinuation because of adverse events will be lower than in the trials.

4.9 The committee noted that the company's original model used overall survival from PALOMA‑1, because this is the only source available, but it used progression-free survival data from PALOMA‑2. The ERG stated that it considers the mixing of the 2 data sets to be methodologically flawed, because it assumes that progression-free survival and overall survival were independent of one another. Therefore, the ERG preferred to use the PALOMA‑1 time-to-event data throughout. The committee noted that in the revised analyses submitted during consultation, the company used progression-free survival data from PALOMA‑1 and also the updated final overall survival data from PALOMA‑1 (see section 4.4). The committee noted that in the revised analyses the company had accepted all amendments suggested in the ERG's exploratory base case, except the modelling of overall survival.

4.10 In its revised analyses the company used 2 different approaches for modelling overall survival. The company's 'lower bound' survival was based on a parametric curve (exponential) fitted to the individual patient data from the final analysis of PALOMA‑1. This method closely resembled the ERG's preferred approach. The company also calculated its 'upper bound' survival, by increasing the overall survival gain in PALOMA‑1 to match the modelled progression-free survival gain (that is, 11.2 months). The committee recalled that the relationship between progression-free and overall survival is complex and difficult to predict, but that palbociclib would be expected to improve overall survival. It noted the company's comments that the lack of a statistically significant overall survival gain in PALOMA‑1 could be because overall survival data are confounded by randomness of response to post-progression treatments. The committee concluded that although it is possible that the overall survival gain might be better than that in PALOMA‑1, there is no evidence to support an assumption of overall survival gain equal to the progression-free survival gain without further overall survival data from PALOMA‑2.

4.11 In its original submission the company did not include any treatment-related costs beyond first-line therapy. It included only disease-related costs, estimated using package 2 care from NICE's clinical guideline on advanced breast cancer. After seeking advice from clinical nurse specialists, and making adjustments to reflect current NHS practice and variations in lines of treatment, the company incorporated an average disease-related cost of £573.86 per cycle (28 days) in the post-progression state. The ERG did not agree that treatment-related cost such as drug acquisition costs for second-line therapy and beyond could be ignored. It recommended more precise costing. The company presented revised analyses with higher estimated costs for post-progression states (£2,000; £1,395; and £1,140 per cycle). The ERG estimated an average post-progression cost of £1,200 per cycle for active treatment states, and £975 per cycle for best supportive care. These estimates were based on a retrospective review of medical records for patients with hormone receptor-positive, HER2-negative metastatic breast cancer in the UK (Kurosky et al. 2015), together with clinical advice. They also took into account the company's original estimates of disease-related costs. The CDF clinical lead also submitted average costs for second-, third- and fourth-line treatments, estimated in consultation with experts in the Chemotherapy Clinical Reference Group of NHS England. These estimates were presented as commercial in confidence because they included confidential pricing agreements and are therefore not presented here. The committee noted the different estimates and could not be sure which estimate could be considered the most plausible. However, the committee was reassured by the fact that, despite having used different sources, the estimates from the ERG and the CDF clinical lead are reasonably close. It therefore agreed that the ERG's estimates for post-progression costs are plausible.

4.12 In its original base case, the company used different utility values for people in the progression-free state having palbociclib plus letrozole (0.74) or letrozole alone (0.71). These values were derived from the corresponding treatment arms of PALOMA‑2. Taking into account that the difference in the EQ‑5D values between the 2 arms of PALOMA‑2 was not statistically significant, the ERG estimated an average utility value (0.72) by pooling EQ‑5D values for European patients from the first 21 cycles in PALOMA‑2. The committee noted the company's comments that the ERG's preferred estimate undervalues progression-free survival, because people with progression-free disease can have a near-normal life. The company referred to NICE's technology appraisal guidance on everolimus with exemestane for treating advanced breast cancer after endocrine therapy. This used a utility value of 0.771 for people with hormone receptor-positive, HER2-negative disease that recurred or progressed after treatment with an aromatase inhibitor, who were having second-line treatment with everolimus plus exemestane. It argued that people on first-line treatment (as in this appraisal) should be assumed to have at least the same quality of life as accepted for those having second-line treatment after progression. The company also presented a scenario using a utility value of 0.75 for progression-free survival, a midpoint between 0.72 and 0.77.

4.13 The committee discussed the most appropriate source of utility values for use in economic modelling, particularly those gathered directly in the relevant trials compared with those sourced from elsewhere. It noted that there is a strong preference for people wanting to delay starting chemotherapy (see section 4.2). It also noted that because EQ‑5D measures the health state of people at points in time, it may not fully capture a person's preference to avoid future events. The committee was aware that EQ‑5D data from trials is recommended for use in the NICE reference case. It was, however, aware that there has been inconsistency in the utility values used for similar disease stages across different NICE appraisals for metastatic breast cancer. The committee concluded that it is difficult to precisely predict the quality of life of someone with progression-free disease who is taking endocrine therapy. It agreed to explore a range of utility values for progression-free disease (0.72 to 0.77) for its deliberation on the cost effectiveness of palbociclib.

4.14 The committee discussed the company's revised base case incorporating a confidential patient access scheme. It noted that the company had presented a range of incremental cost-effectiveness ratios (ICERs) using 2 approaches to modelling overall survival, 3 utility values for progression-free state (0.72, 0.75 and 0.77) and 3 estimates for post-progression costs (£1,140; £1,395; and £2,000 per cycle). These ICERs were presented as commercial in confidence to maintain confidentiality around the patient access scheme. The committee agreed that after applying a discount to the list price as agreed in the patient access scheme, and using a more realistic estimation of the subsequent treatment costs, the ICERs would be within the range that can be considered cost effective.

4.15 The committee discussed the innovative nature of palbociclib. It noted that the Medicines and Healthcare products Regulatory Agency recognises palbociclib as a promising innovative medicine. The committee agreed that there is a clinical need for better treatments for this patient group, and that it prolongs progression-free survival in this population. It recognised that this is important to patients and that no weight had been given in the cost-effectiveness analysis to the specific benefit of delaying chemotherapy with its attendant side effects, which patients consider important. The overall survival gain also remains an area of significant uncertainty, and could be greater than that shown in PALOMA‑1.

4.16 The committee noted that there are uncertainties in the calculation for the most plausible ICERs, including:

4.17 The committee was aware of NICE's position statement on the Pharmaceutical Price Regulation Scheme (PPRS) 2014, and in particular the PPRS payment mechanism. It accepted the conclusion 'that the 2014 PPRS payment mechanism should not, as a matter of course, be regarded as a relevant consideration in its assessment of the cost effectiveness of branded medicines'. The committee heard nothing to suggest that there is any basis for taking a different view about the relevance of the PPRS to this appraisal. It therefore concluded that the PPRS payment mechanism was not relevant in considering the cost effectiveness of the technology in this appraisal.