Abiraterone for treating metastatic hormone-relapsed prostate cancer before chemotherapy is indicated

Clinical effectiveness

4.5 The committee considered whether the randomised placebo-controlled trial COU‑AA‑302 was generalisable to clinical practice in England. It noted that the trial recruited 9% of its patients from the UK and that people in both arms had prednisolone/prednisone. In line with advice from the clinical experts, the committee considered that the placebo arm reflected best supportive care in England before treatment with chemotherapy. The committee heard from the clinical experts that the average age of people in COU‑AA‑302 was similar to that of the people who would be offered abiraterone in clinical practice in England. It heard that the reasons for stopping abiraterone treatment in the trial broadly reflect clinical practice in England. The committee noted that patients in the study stopped treatment with abiraterone when their disease progressed radiographically or clinically, at which point they could have other treatments including docetaxel. The clinical experts stated that, in clinical practice in England, people would get abiraterone or best supportive care until clinical progression rather than radiographic progression. The clinical experts advised that people switch to docetaxel within a week of clinical progression if they are fit enough to tolerate chemotherapy. Despite the difference in defining progression-free survival, the committee concluded that COU‑AA‑302 generally reflected clinical practice in the UK and was relevant to address the decision problem.

4.6 The committee discussed the clinical-effectiveness results from COU‑AA‑302, noting that abiraterone delayed the progression of disease (diagnosed radiographically) compared with placebo. The committee was aware that the survival benefit of abiraterone compared with placebo was not statistically significant at the second and third interim analyses, but was statistically significant at the final analysis. The committee was aware that the company unblinded COU‑AA‑302 early, between the second and third interim analyses, based on advice from the Independent Data Monitoring committee (see section 3.3). It also heard that, after unblinding, people in the trial having placebo could cross over to have abiraterone. Following a request from the committee, the company provided data on treatment switching and subsequent treatments in COU‑AA‑302, to clarify the impact of these factors on the final survival estimates.

4.7 The committee discussed the company's additional evidence about treatment switching and subsequent treatments in COU‑AA‑302. It noted that patients had treatments that prolong survival but are not routinely available in the NHS, specifically:

Cost effectiveness

4.8 The committee understood that the company had developed a discrete event simulation model, rather than the more commonly used Markov model, because it allowed more flexibility to reflect a sequence of treatments, and to model response to treatments that depend on previous treatments. The committee agreed that using a discrete event simulation model was not unreasonable, but that the company's model was particularly complex. In particular, for each of the model's 17 equations predicting time to events or disease status, the committee noted that the company made a large number of judgements when determining which covariates to include and which parametric distribution to choose for extrapolation (see section 3.14). The committee noted that, for 2 equations, the company had not followed its own statistical plan when choosing covariates, and the committee agreed that this could introduce bias to the model. The committee concluded that the company's model was complex and lacked transparency, which made it difficult for the evidence review group (ERG) to validate and critique, and for the committee to determine the plausibility of the model outcomes.

4.9 The committee discussed the clinical data used to inform the company's model. It noted that the company preferred to use data from the third interim analysis of COU‑AA‑302 rather than the final analysis because the interim data needed less adjustment for patients who crossed over from the placebo group to the abiraterone group. The committee concluded that it was reasonable to use data from the third interim analysis, but it requested that the company provide further evidence comparing the modelled estimates with the final trial data for time on first treatment and overall survival (see sections 4.12–4.14).

4.10 The committee discussed the company's choice of parametric distribution for each of the 10 equations in the model that needed extrapolating. The committee was aware that the company considered several functions and selected the best fitting distribution using statistical criteria and visual inspection. The committee noted that:

4.11 The committee discussed the sensitivity analyses that used different parametric distributions. Using a Weibull instead of a log-logistic distribution for 1 equation (time from starting to stopping first treatment) increased the company's base-case incremental cost-effectiveness ratio (ICER) from £28,600 to £35,800 per quality-adjusted life year (QALY) gained. Similarly, using Weibull instead of log-logistic distributions for 2 equations (time from starting to stopping first treatment and time from starting docetaxel to death while taking docetaxel) increased the ERG's exploratory base-case ICER from £35,500 to £55,600 per QALY gained. The committee was also aware of the company's analysis, submitted in response to the second appraisal consultation document, using a piecewise curve for predicting time on first treatment (see section 3.22). Using the piecewise curve increased the company's base-case ICER from £28,600 to £32,800 per QALY gained. The committee heard from the company that it preferred to use the log-logistic distribution to predict time on first treatment because, if people's disease responds to abiraterone, they tend to stay on it for a long time. After the third committee meeting, the committee asked the company to submit additional data about treatment durations with abiraterone.

4.12 The committee discussed how long people take abiraterone, noting that the company had provided 3 sets of data: the final analysis of COU‑AA‑302; data from clinical practice in the UK; and data from clinical practice in the US. The UK data are commercial-in-confidence and cannot be presented here.

4.13 The committee discussed how long people remain on best supportive care before starting docetaxel, noting that the company had provided data from the final analysis of COU‑AA‑302. The committee was concerned that neither the log-logistic distribution (used in the base case) nor the Weibull distribution (used in sensitivity analyses) provided a good fit to the final trial data, and both distributions overestimated the time that patients would remain on best supportive care. The committee noted that the piecewise curve (used in sensitivity analyses) was a closer fit to the trial data, but in this analysis the company assumed that all patients stopped having best supportive care at about 1,000 days; the committee was concerned that this assumption may not be clinically plausible. The committee concluded that, for predicting time on best supportive care, it preferred to use the same distribution as was used for abiraterone (that is, log-logistic or piecewise, see section 4.12).

4.14 The committee discussed the predictions of overall survival in the company's model. It inspected the final Kaplan-Meier survival curves from the trial and compared them with the survival curves predicted by the company's base-case model, which extrapolated beyond the trial data. The committee noted that, beyond about 3 years of follow-up, for both treatment groups the trial data showed longer survival times than predicted by the model. The committee acknowledged that this may be because of treatment switching and the use of subsequent treatments in the trial that are not available on the NHS (see section 4.7). It was also aware that, because of these potential confounders, the company had used data from the third interim analysis for modelling rather than the final analysis. The committee was aware of the company's response to the second appraisal consultation document, which stated that the model over-estimated survival times with best supportive care when compared with the placebo group of the trial after adjusting for treatment switching. The committee concluded that the model predictions of overall survival were uncertain and it was unclear whether the model over- or under-estimated survival times.

4.15 The committee discussed the company's method for adjusting modelled survival times to remove the benefit of treatments that were used in the COU‑AA‑302 trial but are not used in the NHS. The committee understood that the survival times of patients in the abiraterone arm of the model were reduced to remove the benefit of cabazitaxel, enzalutamide and re-treatment with abiraterone. The committee accepted that it was appropriate to adjust for treatments that have a survival benefit and which are not available in the NHS. However, it noted that the company's method (see section 3.13) was an approach that neither the ERG nor the committee had seen before. The committee noted that adjusting for subsequent treatments had a modest impact on the ICER (the company's base case with adjustment was £28,600 per QALY gained; a scenario without adjustment was £27,700 per QALY gained). The committee also noted that the company did not adjust for treatment switching and subsequent treatments in the best supportive care arm of the model. Overall, the committee concluded that adjustment for subsequent treatments in the abiraterone arm should be included in the analyses used for decision-making, recognising that such an adjustment was included in the company's base-case and scenario analyses, and the ERG's exploratory base case.

4.16 The committee discussed the trial population used to inform the model. It noted that, for each prediction equation, the company used data from all the patients in the COU‑AA‑302 intention-to-treat (ITT) population who had complete data for the covariates in that equation. Consequently, the sample size varied between equations; the minimum sample was the 902 people for whom complete data were available (the 'full covariate subgroup'). The company's approach is subsequently referred to as 'using the full covariate subgroup'. The committee was aware that the ERG preferred to use the ITT population for modelling, which meant that the ERG's prediction equations did not include covariates. The committee heard that, in the company's opinion, it was appropriate to use the full covariate subgroup because:

4.17 The committee discussed the survival estimates for the post-docetaxel phase of the model. It noted that abiraterone taken after docetaxel had been appraised in abiraterone for castration-resistant metastatic prostate cancer previously treated with a docetaxel-containing regimen (hereafter referred to as TA259), but that the company had not used data from TA259 to check the validity of its model in the current appraisal. It also noted that the modelled post-docetaxel survival times were shorter in the current appraisal (based on data from COU‑AA‑302) than in TA259 (based on data from COU‑AA‑301). The committee was aware that the ERG carried out a scenario analysis in which it fixed post-docetaxel survival to be the same as in COU‑AA‑301, and this increased the ICER. The committee heard from the company that, although the estimates from COU‑AA‑301 came from a larger sample of patients, it did not consider these data to be relevant for the current appraisal because the population in COU‑AA‑301 was different from that in COU‑AA‑302. In particular, the company stated that people in COU‑AA‑301 started docetaxel earlier in their treatment pathway than in COU‑AA‑302 and therefore people in COU‑AA‑301 were also fitter at the point they started post-docetaxel treatments. During the committee meeting, the company stated that the model was designed to follow individual patients through several stages of treatment. In the company's opinion, it was not appropriate to use the COU‑AA‑301 data in the model because doing so would 'break randomisation' and it would not be possible to adjust the data to reflect differences in baseline characteristics between the COU‑AA‑301 trial and the modelled population. On balance, and because of the complex model chosen by the company, the committee agreed with the company and concluded that it was appropriate to use COU‑AA‑302 to estimate post-docetaxel survival times. Nonetheless, it concluded that uncertainty about the modelled survival times persisted because only a small number of patients from COU‑AA‑302 contributed data to this phase of the model.

4.18 The committee discussed the utility values in the company's model. It understood that the company derived utility values, by phase of treatment, from 3 sources:

4.19 The committee discussed the costs used in the model.

4.20 The committee discussed how the company incorporated the abiraterone PAS in its model. It was aware of an existing simple discount PAS for abiraterone, which was agreed as part of TA259 (the appraisal of abiraterone after docetaxel). The committee was also aware that:

4.21 The committee considered whether it should take into account the consequences of the Pharmaceutical Price Regulation Scheme (PPRS) 2014, and in particular the PPRS payment mechanism, when appraising abiraterone. It noted that the company had not made a case for the relevance of the PPRS in this appraisal. The committee noted NICE's position statement in this regard, and 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'. After asking the company, the committee heard nothing to suggest that there is any basis for taking a different view on the PPRS in this appraisal of abiraterone. It therefore concluded that the PPRS payment mechanism was not applicable when considering the cost effectiveness of abiraterone.

4.22 The committee discussed whether abiraterone could be considered a cost-effective use of NHS resources. It had concluded that its preferred approach was:

4.23 The committee considered supplementary advice from NICE that should be taken into account when appraising treatments that may extend the life of patients with short life expectancy and that are licensed for indications that affect small numbers of people with incurable illnesses. For this advice to be applied, all of the following criteria must be met:

4.24 To address whether metastatic hormone-resistant prostate cancer at this stage of therapy is associated with a mean life expectancy of less than 24 months, the committee considered the median overall survival in the control arm of COU‑AA‑302, noting that it was about 30 months. The committee heard from the company in the first meeting that it had estimated the mean survival for the best supportive care arm as 32 months. The committee considered the company's and stakeholder comments received during consultation suggesting that people treated in the NHS would have a lower life expectancy than people in the best supportive care arm of COU‑AA‑302. These comments included:

4.25 The committee considered carefully the company's review of the published literature on alternative estimates of survival for people with prostate cancer. Overall, the committee concluded that current mean life expectancy for people with metastatic hormone-relapsed prostate cancer for whom chemotherapy is not yet indicated was unlikely to be less than 24 months.

4.26 Having determined that abiraterone did not meet the 'end of life' criterion on life expectancy, the committee discussed the criteria of small patient population and whether abiraterone extended life by more than an average of 3 months. It noted that the company, in its response to the appraisal consultation document, estimated that 6,782 people would be eligible for the pre- and post-docetaxel marketing authorisations in England, but that a proportion of the population eligible for abiraterone after docetaxel would not have abiraterone if they had it before docetaxel. The committee concluded that the eligible population for England did not exceed 7,000 and that abiraterone therefore met the end-of-life criterion for a small patient population. The committee noted that the final analysis of COU‑AA‑302 showed a median survival benefit of 4.4 months and the modelled mean survival benefit was longer than this. It concluded that the extension-to-life criterion was met. However, because the 24 month life expectancy criterion had not been met, the committee concluded that the end-of-life criteria did not apply to abiraterone taken before docetaxel in the treatment pathway.

4.27 The committee considered whether abiraterone was innovative and whether it had substantial, demonstrable and distinctive benefits not adequately captured in the modelling of the QALYs. The committee noted that, although abiraterone is not a new technology, it was the first active treatment available for this position in the treatment pathway and, in this regard, was innovative. It then considered whether the model captured the benefits of either having abiraterone at an earlier point in the treatment pathway when people had higher quality of life, or delaying the need for cytotoxic chemotherapy, such as docetaxel. The committee noted that the model predicted that people in the abiraterone arm have more time with better utility before docetaxel than people on best supportive care. However, the committee agreed that the benefit of delaying chemotherapy perceived by patients may not have been fully captured by the utility values included in the modelling and that accounting for this would have reduced the ICER. The committee concluded that abiraterone was innovative and this should be considered in its decision-making.

4.28 The committee noted that NICE's Guide to the methods of technology appraisal states that, if a technology has a most plausible ICER above £30,000 per QALY gained, the committee will need to identify an increasingly stronger case for supporting the technology as an effective use of NHS resources. The committee noted that:

Summary of appraisal committee's key conclusions