3 Evidence

The appraisal committee (section 6) considered evidence from a number of sources. See the committee papers for full details of the evidence.

Clinical effectiveness

3.1 The company identified 2 phase III randomised controlled trials (COMBI‑d and COMBI‑v) that assessed the clinical effectiveness of trametinib plus dabrafenib in people with histologically-confirmed stage IIIC (unresectable) or stage IV (metastatic) BRAF V600E/K mutation-positive melanoma. The trials included people who had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, with unresectable or metastatic melanoma that had not been treated before. Fewer than 20 patients had brain metastases.

3.2 COMBI‑d was a double-blind, multicentre randomised controlled trial that compared trametinib (2 mg once daily) plus dabrafenib (150 mg twice daily; n=211) with dabrafenib monotherapy (150 mg twice daily with placebo control; n=212). COMBI‑v was an open label, multicentre randomised controlled trial that compared trametinib (2 mg once daily) plus dabrafenib (150 mg twice daily; n=352) with vemurafenib monotherapy (960 mg twice daily; n=352).

3.3 The company stated that the demographic disease characteristics and prognostic factors in both trials were generally well balanced between the treatment groups at baseline. In COMBI‑v, after the pre-planned cut off, the Independent Data Monitoring Committee recommended stopping the study early due to superior efficacy in the trametinib plus dabrafenib group. As a result the study protocol for COMBI‑v was amended to allow crossover from vemurafenib to the trametinib plus dabrafenib group. Crossover was not permitted in COMBI‑d.

3.4 The company reported statistically significant and clinically meaningful differences in overall survival (OS) between the trametinib and dabrafenib combination group and the monotherapy groups at the final cut-off point in both trials. In COMBI‑d there as a median OS of 25.1 months in the trametinib plus dabrafenib group compared with 18.7 months in the dabrafenib monotherapy group, with a corresponding hazard ratio (HR) of 0.71 (95% confidence interval [CI] 0.55 to 0.92). In COMBI‑v there was a median OS of 25.6 months in the trametinib plus dabrafenib group compared with 18.0 months in the vemurafenib monotherapy group, with a corresponding hazard ratio of 0.66 (95% CI 0.53 to 0.81).

3.5 The progression-free survival (PFS) results based on investigator assessment from COMBI‑d showed, at final cut off, a median PFS of 11.0 months in the trametinib plus dabrafenib group compared to 8.8 months in the dabrafenib monotherapy group (HR=0.67; 95% CI 0.53 to 0.84). Results from COMBI‑v showed a median PFS of 12.6 months in the combination group compared with 7.3 months for vemurafenib alone (HR=0.61; 95% CI 0.51 to 0.73).

3.6 The company presented health-related quality of life results from both trials using 2 measures; the EuroQol-5 dimension questionnaire (EQ‑5D) and the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire-Core 30 (EORTC‑QLQ‑30). The rates of completion of EORTC‑QLQ‑30 among patients in both studies, and of EQ‑5D in COMBI‑v (the company did not present EQ‑5D completion rates for COMBI‑d) were over 90% at baseline, and over 70% at disease progression. For COMBI‑d, EQ‑5D scores at baseline were similar between treatment groups with a statistically significant difference in results favouring the combination group at week 16 only. Results from COMBI‑v showed similar baseline EQ‑5D scores between treatment groups and, for all assessments, the differences between scores were significantly better for the combination group compared with the monotherapy group. EORTC‑QLQ‑30 global health scores in COMBI‑d were significantly better at weeks 8, 16 and 24 for trametinib plus dabrafenib compared with dabrafenib alone but not at other time points. The company reported that the EORTC‑QLQ‑30 global health scores in COMBI‑v were significantly better for patients in the trametinib plus dabrafenib group compared with vemurafenib alone.

3.7 Approximately half of the adverse events experienced by patients having trametinib plus dabrafenib were mild-to-moderate in severity (grade 1 and grade 2). In both studies, the rates of grade 3 and grade 4 adverse events were higher in the monotherapy groups than in the combination groups (50% and 45% in COMBI‑d; 57% and 66% in COMBI‑v). In both studies, pyrexia and hypertension were the most common grade 3–4 adverse events in the trametinib plus dabrafenib group. Fewer skin-related toxicities were reported in the combination group. Adverse events leading to treatment discontinuation, dose reductions or dose interruptions were more common in the trametinib plus dabrafenib group compared with dabrafenib alone in COMBI‑d, but in COMBI‑v the proportions were generally similar between the trametinib plus dabrafenib group and the vemurafenib monotherapy group.

Evidence review group comments

3.8 The evidence review group (ERG) considered COMBI‑d to be well designed and to have a low risk of bias. It reported that COMBI‑v was of good quality but that the trial was open-label and did not include a placebo control or a blinded independent review of the secondary outcomes, which could lead to bias. The ERG commented that, in both trials, patients had not had treatment before and that there is no evidence for the use of trametinib plus dabrafenib as a second line treatment following immunotherapy.

3.9 The ERG agreed with the company that trametinib plus dabrafenib demonstrated greater clinical effectiveness compared with dabrafenib or vemurafenib alone. The ERG commented on the representativeness of the EQ‑5D data collected in the trials. It noted that in the trametinib plus dabrafenib groups of COMBI‑d and COMBI‑v the utility scores at week 40 and week 48 were higher than those at baseline (+0.01 and +0.07 respectively). Similarly, data from both trials showed that the utility scores at the point of disease progression for patients having the combination treatment were higher than those at baseline (+0.04 and +0.06 respectively). The ERG considered that these observations may be because only patients who felt well completed the questionnaires. The ERG also commented that the lack of study blinding in COMBI‑v may have influenced the scores reported by patients.

Cost effectiveness

3.10 The company presented a de novo partitioned survival model to assess the cost-effectiveness of trametinib plus dabrafenib in people with BRAF V600 mutation-positive unresectable or metastatic melanoma. The perspective was that of the NHS and personal social services. The time horizon of the model was life time (30 years), the cycle length was 1 week and half‑cycle correction was not applied. Costs and outcomes were discounted at 3.5% per year. The model included 3 states: progression-free, post-progression and death. The company assumed that people entered the model in the progression-free state and had treatment with trametinib plus dabrafenib or with one of the BRAF-inhibitor monotherapies (dabrafenib or vemurafenib). Transition between states was derived from response to treatment and risk of disease progression or death. Patients that transitioned to a post-progression state were assumed to discontinue therapy and they stayed in that state until transitioning to death. Progression of disease was defined using Response Evaluation Criteria in Solid Tumours (RECIST) criteria, version 1.1.

3.11 The clinical-effectiveness estimates for each of the treatment groups were taken from pooled PFS and OS times at the final cut-off points from COMBI‑v and COMBI‑d. To estimate long-term PFS the company used Kaplan–Meier analysis until a set breakpoint (estimated using the piecewise linear function). Beyond the set breakpoint an assumption of constant hazards was applied to each group separately. This assumed that the observed PFS benefits would continue beyond the trial follow up. To estimate long-term OS the company used Kaplan–Meier data until a set breakpoint (estimated as above), after which an assumption of constant hazards was applied until year 5. After year 5, the company combined exponential extrapolation of constant hazards with long-term survival data from the American Joint Committee on Cancer (AJCC) registry, with case-mix adjustment by melanoma stage in each treatment group. General population mortality, matched by age and gender, was added to the AJCC rates after 20 years to account for increased risk of death in the population.

3.12 The company assigned utility values to each of the health states in the model using EQ‑5D data from COMBI‑v and COMBI‑d. The model assumed that within each health state the quality of life is constant over time, that is, a single utility value applies to that health state. Adverse events were indirectly incorporated in the model because utility data were taken directly from the trials. The model included all data on healthcare resource use associated with treatment and disease progression. The company used the UK MELODY study (a study of resource utilisation in 220 people with melanoma) to inform UK clinical practice in melanoma treatment and for costing additional resource use in pre- and post-progression health states. Costs were inflated to current price levels. The model included post-study anticancer treatment costs, calculated as a weighted sum of the expected total post-study anticancer treatment cost per patient. Results from COMBI‑d showed that a higher proportion of patients had a subsequent anticancer treatment after progression in the monotherapy group, compared with those who had trametinib plus dabrafenib (51% and 33% respectively). The differences in post-study treatments resulted in a difference in expected mean costs in each treatment group. The model incorporated costs of treating adverse events which the company determined were likely to have the greatest impact on NHS resource. This was derived by selecting those grade 3 and above adverse events with an incidence of 5% or over in each treatment group of COMBI‑v or COMBI‑d.

3.13 The company presented the results from the cost-effectiveness analysis based on the list prices of the intervention and comparator technologies. Results showed that trametinib plus dabrafenib was associated with a greater number of quality-adjusted life years (QALYs) compared with the monotherapies (1.345 more QALYs than vemurafenib and 1.298 more QALYs than dabrafenib). Because trametinib, dabrafenib and vemurafenib have confidential patient access schemes, the incremental cost-effectiveness ratios (ICERs) presented are not reflective of the actual cost to the NHS.

3.14 The company did deterministic sensitivity analysis varying the assumptions in the model. Results were sensitive to the time horizon of the model, with a longer time horizon reducing the ICERs for trametinib plus dabrafenib compared with the monotherapies. The company also reported that there was a reduction in the ICERs if an assumption of no continuing benefit for PFS beyond trial follow up was applied (that is, the treatment effect of combination therapy on progression disappears after the end of the trial period and, in the projection phase, if the monthly hazards of progression were the same for combination therapy and monotherapy).

Evidence review group comments

3.15 The ERG commented that by pooling the data from COMBI‑d and COMBI‑v the company had not taken into account the potential for trial design bias. Although there were similarities in the trial designs and patient populations, COMBI‑d was blinded and placebo controlled, while COMBI‑v was not blinded and no placebo control was used. The ERG noted that the 2 trials have different survival profiles and considered that it was not appropriate to pool the data. It considered that the results from the high-quality, blinded placebo-controlled trial were more likely to show the true effectiveness of dabrafenib and vemurafenib, if their clinical equivalence was accepted.

3.16 The ERG did not consider that the company had appropriately extrapolated long-term OS. It considered that it was preferable to use all of the available trial data, rather than extrapolation, over a period for which data exists. It also noted that constant mortality hazards were not applied up to year 5 in the model, but up to month 30 (trametinib plus dabrafenib) or month 31 (monotherapy). The monthly mortality hazards declined in both groups consistently until they converged with the mortality hazards at year 5, underestimating the survival in the monotherapy arm.

3.17 The ERG noted that the mean time for patients to continue having treatment after investigator-assessed progression was over 200 days in the trametinib plus dabrafenib group. The ERG considered that PFS is a poor proxy for time on treatment and treatment costs, and that time to treatment discontinuation would provide a more accurate measure.

3.18 The ERG did not agree that post-study anticancer therapy costs were different between the treatment groups or that a simple average of the post-study anticancer therapy costs should be applied. The ERG noted that the company's post-study anticancer therapy costs did not represent the current clinical pathway because less than 7% of patients in COMBI‑d and COMBI‑v had pembrolizumab.

3.19 The ERG noted that utility values used in the model were derived from the full study population and considered that data from European patients would provide a more representative estimate of health-related quality of life. It also considered the use of utility values from COMBI‑v to be inappropriate considering the open label design of the trial. The ERG commented that it did not consider that there was a statistical reason to assume different utility values for dabrafenib and vemurafenib monotherapies compared with trametinib plus dabrafenib. It considered that it would be more meaningful to apply utility values that are related to being 'on' or 'off' study treatment as determined by the time to treatment discontinuation data. The ERG also considered that the utility values used in the model should reflect declining utility with age.

Evidence review group exploratory analyses

3.20 Using the company's model, the ERG presented deterministic cost-effectiveness results based on the confidential patient access scheme discounted prices of the intervention and comparator technologies. Because the patient access schemes are confidential, ICERs cannot be presented.

3.21 The ERG made several amendments to the company's model:

  • using time to treatment discontinuation data to estimate study treatment costs

  • using equal post-study anticancer treatments for the intervention and comparator therapies

  • applying on and off treatment utility values relating to European patients and adjusted for age

  • applying the ERG's method for estimating OS

  • using COMBI‑d trial data (instead of pooled data) and applying the ERG's preferred method for estimating OS.

    Each amendment increased the incremental costs and the ICERs for trametinib plus dabrafenib compared with dabrafenib or vemurafenib alone. Applying the ERG's preferred method of extrapolating OS reduced the number of QALYs gained for trametinib plus dabrafenib from 1.297 QALYs to 0.878 QALYs compared with dabrafenib alone, and from 1.345 QALYs to 0.925 QALYs compared with vemurafenib alone.