Nivolumab for treating relapsed or refractory classical Hodgkin lymphoma

Table 1 Clinical data from CheckMate 205 and CA209‑039

The committee was concerned that the single-arm design of the trials, the small number of patients included and short follow-up meant that the results were potentially biased. The committee accepted that the results from the latest data cut-off for both trials (April 2016 for CheckMate 205 and August 2015 for CA209‑039) showed that nivolumab was clinically effective based on response rates but agreed that there was a large degree of uncertainty in the clinical evidence.

Indirect treatment comparison of nivolumab with standard of care

4.6 The committee was aware that there were no data providing direct comparative evidence for the clinical effectiveness of nivolumab compared with current practice (standard of care), because nivolumab for Hodgkin lymphoma had only been studied in single-arm trials. It noted that the company had done an indirect treatment comparison of nivolumab compared with standard of care by comparing the pooled outcomes from the nivolumab trials with standard of care. The outcomes for standard of care came from Cheah et al. (2016), a retrospective real-world study done in the US. The study aimed to determine progression-free survival and overall survival in patients with Hodgkin lymphoma following disease relapse after brentuximab vedotin therapy; a secondary outcome was the efficacy of subsequent treatments.

4.7 The committee considered whether the population and composition of treatments in the Cheah study reflected clinical practice in the UK. The committee noted that the study population partially matched the population of interest because around 70% of patients had previous autologous stem cell transplant and brentuximab vedotin. The committee noted a lack of detail on the precise combinations of chemotherapies given as standard of care in the study, and the inclusion of platinum-based therapies and 'other alkylators'. It considered that the study may not reflect UK practice, particularly regarding subsequent rates of allogeneic stem cell transplant. The committee noted that in response to consultation, the company had explored UK standard-of-care data from the Haematological Malignancy Research Network and surveyed clinicians actively treating relapsed or refractory classical Hodgkin lymphoma in the UK. The committee considered that both the network data and the clinician survey somewhat supported the Cheah study as reflecting UK practice, but it recognised that the data were limited. The committee concluded that the Cheah study was the best available evidence for standard of care and considered it appropriate for its decision-making, but overall the clinical effectiveness of nivolumab compared with standard of care was highly uncertain because the comparator data may not fully represent UK clinical practice.

4.8 The committee noted that the company's indirect treatment comparison excluded results from patients who had investigational agents in the Cheah study. It heard from the company that including investigational agents could have confounded the results, because they increased survival benefit above that expected from treatments used in current practice, and that the investigational agents were likely to include PD‑1 inhibitors, one of which is nivolumab itself. It also heard from the company that use of investigational agents tended to be restricted to large centres and that they could not therefore be considered current NHS practice. The committee acknowledged that the investigational agents used in the study could have included treatments not available in the UK, but noted that there was little detail about which specific therapies were defined as 'investigational agents'. It heard from the evidence review group (ERG), who had contacted the authors of the study, that only 'a couple' of patients in the study had a PD‑1 inhibitor, and so it considered that the overall population should be used for comparator data. The committee considered that selectively excluding potentially the fittest patients from the Cheah dataset could bias the results of the indirect treatment comparison more than including some treatments that may not be used in UK current practice. The committee concluded therefore that the overall population of the Cheah study was the most appropriate dataset for standard of care to use in the indirect comparison.

4.9 The committee was aware of the results of the company's comparison of the pooled nivolumab data with the standard-of-care data from the overall population in the Cheah study for progression-free and overall survival (results are academic in confidence because the company plan to publish them before the end of 2017, and so they cannot be reported here). The committee noted that in the company's original submission, the results of the indirect treatment comparison were obtained from an unadjusted or 'naive' comparison, but that it had also presented the results of a matched-adjusted indirect comparison as a scenario analysis. The committee was aware that a matched-adjusted indirect comparison took account of different distributions of prognostic factors and effect modifiers arising from any differences in baseline characteristics of patients in the trials. It concluded that a matched-adjusted indirect comparison would produce more robust results, but considered that the unadjusted comparison used in the company's base-case analysis was similar and so was acceptable for its decision-making. Furthermore, the committee acknowledged that the company had used the results of the matched-adjusted comparison in its revised cost-effectiveness analysis submitted in response to consultation.

4.10 In conclusion, the committee acknowledged that nivolumab was clinically effective, but noted that the available evidence was highly uncertain because the data were immature and from single-arm studies. The committee noted that the company's indirect comparison with standard of care had compared pooled outcomes from the nivolumab trials with outcomes from the Cheah et al. study. The committee considered that Cheah et al. may not fully reflect UK current practice, but it acknowledged that the published evidence for comparator treatments used in the UK was limited. When considering the new evidence received from the company in response to consultation, it acknowledged that Cheah et al. was currently the best available evidence for standard of care and considered it appropriate for decision-making. The committee concluded that there was a large degree of uncertainty in the clinical evidence.

Modelling survival data

4.12 The committee noted that to model progression-free survival and overall survival, the company used the outcome data from the matched-adjusted indirect treatment comparison of nivolumab compared with the treatments in the Cheah study (see section 4.6). It was concerned that a large proportion of the survival benefit of nivolumab compared with standard of care was based on extrapolation rather than on trial data, because the trial data were very immature. It was aware that the company had extrapolated beyond the trial follow-up for nivolumab by fitting a lognormal curve to progression-free survival data (investigator-assessed) and a Weibull curve to overall survival, and that for standard of care, exponential curves had been fitted to the progression-free and overall survival data from the Cheah study. The committee heard from the ERG that the extrapolation curves used were plausible, but it also considered the plausibility of the Gompertz curve fit to the nivolumab overall survival curve, which represented a more pessimistic assumption about long-term survival. The committee concluded that the Gompertz curve may not be clinically probable, but it was not at all clear that the outcomes would be as favourable as the company's estimates. It concluded that all the parametric curves fitted to the data had a reasonable fit, but that they needed to be considered alongside survival modelling that included the long-term survival benefit of subsequent allogeneic stem cell transplant.

Subsequent allogeneic stem cell transplant

4.13 The committee considered those patients who had a partial or complete response to nivolumab and went on to have a potentially curative allogeneic stem cell transplant, and how these patients may have affected overall survival in the model. The committee recalled that allogeneic stem cell transplant was potentially curative (see section 4.2), and that because nivolumab could be used as salvage therapy to enable allogeneic stem cell transplant, the modelling should include the projected long-term survival benefit of transplant. It was aware that the survival modelling used in the company's original base-case analysis included both patients who had allogeneic stem cell transplant and those who had not, in both treatment arms, but that the company had only modelled the effect of subsequent allogeneic stem cell transplant on long-term survival in a scenario analysis. The committee noted that in this analysis, the company had used non-UK data from the Cheah study to project long-term survival for patients who had subsequent allogeneic stem cell transplant. The committee understood that the survival data for subsequent allogeneic stem cell transplant had been extrapolated independently from the overall survival extrapolation used in the base case. It acknowledged that there would be some double counting because the overall survival extrapolation used in the base case included some patients who had allogeneic stem cell transplant, but agreed that it was an acceptable approach. The committee noted that in its revised cost-effectiveness analyses received in response to consultation, the company had censored overall survival data for patients having nivolumab who went on to have subsequent allogeneic stem cell transplant. It considered that censoring data in only 1 arm of a model introduced a more substantial bias than a small amount of double counting, and was not therefore methodologically appropriate.

4.14 The committee noted that the company had used data from a UK study to model long-term survival after allogeneic stem cell transplant in its revised cost-effectiveness analyses. The committee considered the UK study of 13 patients with classical Hodgkin lymphoma having allogeneic stem cell transplant after 3 previous therapies (Laffety et al. 2017), and noted the small number of patients included in the study. It agreed that any modelling of long-term overall survival beyond the median trial follow-up of approximately 28 months would therefore be subject to significant uncertainty. It noted that the company had fitted a Gompertz parametric curve to the data from the Laffety study to extrapolate long-term overall survival but that this projected an infinite median survival for patients having allogeneic stem cell transplant (all-cause mortality excluded), which the committee considered implausible. It therefore considered that the lognormal and Weibull parametric curves used by the ERG in its new exploratory analysis were more clinically plausible because these curves did not predict infinite median survival.

4.15 The committee considered the proportion of patients who were likely to have an allogeneic stem cell transplant in the UK, if their disease had partially or completely responded to treatment after autologous stem cell transplant failed. The committee was aware that in its original scenario analyses, the company had obtained response-specific proportions of patients having subsequent allogeneic stem cell transplant (22.2% of those with complete response, 14.1% with partial response and 5.56% with stable disease) from a study in France (Perrot et al. 2016), and applied them to the response rates seen in the nivolumab and Cheah studies to generate transition probabilities for each treatment arm for use in the model. The committee understood that the ERG had assumed the proportion of patients having subsequent allogeneic stem cell transplant would be equivalent to the proportion who had subsequent allogeneic stem cell transplant in the nivolumab and Cheah studies, which was overall slightly higher than the proportions in the Perrot study. However, it heard from the clinical experts that UK rates of allogeneic stem cell transplant were much higher than those in the US. The committee considered the results of the company's clinician survey, which it received in response to consultation. It noted that the response-specific proportions of patients expected to have subsequent allogeneic stem cell transplant, as reported by clinicians working in the UK, were significantly higher than those used in the original analyses by both the company and ERG. However, because these rates were expected rather than actual, and the results of the survey included a wide range of expected transplant rates (which were very different from the actual rates reported in the nivolumab trials and Cheah study), the committee agreed that rates of allogeneic stem cell transplant in the UK remained uncertain. The committee was also aware from supportive evidence provided by the company in response to consultation that caution may be warranted about using allogeneic stem cell transplant following treatment with nivolumab or another PD‑1 inhibitor, because of the potential for increased risk of complications from transplant linked to PD‑1 inhibitors' immunomodulatory mechanism of action. The committee also heard that recent NHS referrals for allogeneic stem cell transplant were lower than those reported in the survey. The committee concluded that UK rates of allogeneic stem cell transplant may lie somewhere between the high rates reported in the results of the survey, and the considerably lower rates of actual transplants reported in the nivolumab trials and Cheah study.

Treatment costs

4.16 The committee recognised that some patients in the nivolumab trials and the Cheah study had subsequent allogeneic stem cell transplant, and that because the survival benefit from allogeneic stem cell transplant was captured in the survival data for both arms of the model, the costs should also be included. It recognised that the company had used 2 different sources to calculate costs of allogeneic stem cell transplant: a weighted average of NHS reference costs and the costs included in the Radford et al. (2016) paper. The committee agreed that the Radford costs were more appropriate for its decision-making because they were consistent with the costs used for allogeneic stem cell transplant in guidelines currently in development. The committee also considered the costs of comparator treatments, and agreed with the ERG that the costs of mini-BEAM (carmustine, etoposide, cytarabine and melphalan) and DexaBEAM (dexamethasone, carmustine, etoposide, cytarabine and melphalan) should be excluded because they are not used in UK clinical practice, and their benefits would not significantly affect the progression-free or overall survival projections.

Utility values

4.17 The committee was aware that CheckMate 205 (cohort B) collected health-related quality-of-life data for patients having nivolumab using the EQ-5D, which were then converted to utility data. It was also aware that utility data for patients having standard of care were taken from published literature (Swinburn et al. 2015). The committee recognised that response-specific utility values from CheckMate 205 and Swinburn et al. diverged, and that the ERG had instead used the response-specific utility values from CheckMate 205 to estimate utility values for standard of care. The committee agreed that this was a more consistent approach but heard from the clinical experts that pre-progression quality of life was likely to be better with nivolumab than with existing treatments because of nivolumab's potential to improve quality of life (see section 4.3). The committee recognised that the pre-progression utility values used by the ERG in its base case maintained a difference between the treatment arms and concluded that they were therefore more appropriate for its decision-making.

4.18 The committee considered the post-progression utility values and noted the large difference in values between the nivolumab and standard of care treatment arms. It heard from the clinical experts that this large difference was not clinically plausible. The committee preferred the ERG's assumption that post-progression utility values were the same across all treatments.

4.19 The committee noted the ERG's identification of an error in the company's revised analyses received in response to consultation, whereby utility values for patients who discontinued treatment and transitioned to best supportive care reflected the company's original base case and not the committee's preferred assumptions about pre- and post-progression utilities (see sections 4.17 and 4.18). The committee was aware that the ERG had corrected this in its new exploratory analysis and agreed that this was appropriate.

Results of cost-effectiveness analyses

4.20 The committee noted that the company had presented deterministic and probabilistic incremental cost-effectiveness ratios (ICERs) in its revised base-case analysis provided in response to consultation. For nivolumab compared with standard of care, including the confidential discount agreed for nivolumab, the company's deterministic base-case ICER was £15,181 per quality-adjusted life year (QALY) gained and the probabilistic ICER was £17,826 per QALY gained.

4.21 The committee was aware that the company's revised base case included the committee's preferred assumptions about using the overall population from the Cheah study for the standard-of-care data (see section 4.8), the method of indirect treatment comparison (see section 4.9), costs (see section 4.16) and utilities (see sections 4.17 to 4.18), and incorporated subsequent allogeneic stem cell transplant overall survival data (see section 4.13). However, the committee considered that this revised base-case analysis was flawed because of errors in the utility values for best supportive care (see section 4.19) and inappropriate censoring of nivolumab overall survival data (see section 4.13). The committee noted that when the ERG corrected these errors, the company's revised base-case ICER increased to £26,664 per QALY gained. The committee also noted that the company's revised base case had extrapolated long-term survival after allogeneic stem cell transplant using a Gompertz curve, which it had considered implausible (see section 4.14). It was aware that the ERG's new exploratory analyses using the lognormal and Weibull curves increased the ICER to £30,366 and £31,031 per QALY gained respectively. However, the committee was also aware that these analyses included the high rates of allogeneic stem cell transplant from the clinician survey; because the committee considered that actual rates may be lower (see section 4.15), it concluded that the most plausible ICER was likely to be around £30,000 per QALY gained. But this ICER was associated with a large degree of uncertainty because of the immaturity of the nivolumab trial data, the lack of comparator data fully relevant to UK practice, and uncertain outcomes and rates of subsequent allogeneic stem cell transplant in the UK.