Nivolumab with ipilimumab and chemotherapy for untreated metastatic non-small-cell lung cancer

3.1 The clinical experts explained that immunotherapy with platinum-doublet chemotherapy (chemo-immunotherapy) is standard care in the NHS for untreated metastatic NSCLC with no epidermal growth factor receptor (EGFR) or anaplastic lymphoma kinase (ALK) mutations. Most people having chemo-immunotherapy stop treatment before 2 years because their disease progresses or there are tolerability issues, and few survive in the long term. The clinical experts considered that nivolumab combination is likely to have similar efficacy to other first-line chemo-immunotherapy combinations. Limiting the duration of chemotherapy to 2 cycles may reduce renal toxicity and allow platinum-doublet chemotherapy to be offered again as a later-line therapy. The side effects of long-term immunotherapy are usually mild but can sometimes be considerable, needing specialist management. The clinical experts noted that combining 2 immunotherapies is likely to cause more immune-related toxicities than current regimens with only 1 immunotherapy. The committee concluded that nivolumab combination offers another treatment option for untreated metastatic NSCLC with no EGFR or ALK mutations, which may have advantages for some people.

3.2 The clinical experts explained that:

3.3 The committee heard that, in line with NICE's technology appraisal guidance on pembrolizumab with pemetrexed and platinum chemotherapy for non-squamous NSCLC and pembrolizumab with carboplatin and paclitaxel for squamous NSCLC, these combinations are widely used in NHS clinical practice. When the scope for this appraisal was developed, pembrolizumab plus pemetrexed and platinum chemotherapy was recommended for use within the Cancer Drugs Fund. So, it was not included as a comparator in line with NICE's position statement on handling comparators and treatment sequences in the Cancer Drugs Fund. However, it is now recommended for routine commissioning, and is therefore an appropriate comparator for non-squamous NSCLC. NICE's technology appraisal guidance on pembrolizumab with carboplatin and paclitaxel is being reviewed, but this technology is still recommended for use within the Cancer Drugs Fund while the review is ongoing. It is therefore not an appropriate comparator. The committee understood that, in line with the scope, the following comparators were included in the cost-effectiveness analysis:

3.4 The main clinical effectiveness evidence for nivolumab combination came from CheckMate‑9LA. This is an ongoing open-label phase 3 randomised controlled trial, comparing nivolumab combination with standard platinum-doublet chemotherapy. For people with non-squamous NSCLC, platinum-doublet chemotherapy was pemetrexed plus either cisplatin or carboplatin, with optional pemetrexed maintenance therapy. For people with squamous NSCLC, platinum-doublet chemotherapy was paclitaxel plus carboplatin. The committee was aware that CheckMate‑9LA included adults with untreated recurrent or metastatic NSCLC (with no EGFR or ALK mutations) with an Eastern Cooperative Oncology Group performance status of 0 or 1. The trial included people regardless of PD‑L1 status. CheckMate‑9LA did not include these comparators used in NHS clinical practice:

3.5 An interim analysis of CheckMate‑9LA showed a statistically significant difference in overall and progression-free survival in favour of nivolumab combination compared with standard platinum-doublet chemotherapy. At the most recent data cut (March 2020), median overall survival was 15.6 months for nivolumab combination and 10.9 months for standard chemotherapy (hazard ratio [HR] 0.66, 95% confidence interval [CI] 0.55 to 0.80). Median progression-free survival was 6.7 months for nivolumab combination and 5.0 months for standard chemotherapy (HR 0.68, 95% CI 0.57 to 0.82). Some people having nivolumab combination in CheckMate‑9LA had either immunotherapy or a therapy targeted against EGFR, ALK or vascular endothelial growth factor as a subsequent therapy. This did not represent NHS clinical practice, because people would not have immunotherapy again or a targeted therapy on disease progression. The clinical experts stated that this was unlikely to have had a large effect on treatment outcomes. But it meant that the survival of people having nivolumab combination may have been overestimated in CheckMate‑9LA. The committee concluded that nivolumab combination was clinically effective compared with standard chemotherapy.

3.6 The company did indirect treatment comparisons because there were no head-to-head trials comparing nivolumab combination with pembrolizumab monotherapy and atezolizumab combination. The company considered that the proportional hazards assumption (that is, the relative risk of an event is fixed irrespective of time) was not met. It therefore used fractional polynomial models to estimate time-varying hazard ratios. The committee noted that this approach meant that there was a lot of uncertainty with the indirect treatment comparisons. For the indirect comparison with pembrolizumab monotherapy in the subgroup with PD‑L1 TPS at least 50%, the company used the data from the full intention-to-treat (ITT) population from CheckMate‑9LA. For the comparison with atezolizumab combination for the subgroup with non-squamous NSCLC and PD‑L1 TPS below 50%, the company used the data from the relevant subgroup of CheckMate‑9LA. The results of the indirect comparisons are confidential and cannot be reported here. At technical engagement, it was agreed that the CheckMate‑227 data should also be included in the indirect treatment comparisons. CheckMate‑227 is an ongoing open-label phase 3 randomised controlled trial in a similar population to that in CheckMate‑9LA. It includes a nivolumab plus ipilimumab treatment arm and a platinum-doublet chemotherapy treatment arm, both stratified by PD‑L1 TPS. The most recent data cut from CheckMate‑227 (February 2020) had a minimum follow up of 37.7 months compared with 12.7 months for CheckMate‑9LA. The committee noted that some of the company's indirect treatment comparison results had wide confidence intervals and were uncertain, but concluded that they were acceptable for decision making.

3.7 At the first meeting, the committee concluded that pembrolizumab plus pemetrexed and platinum chemotherapy was now widely used in the NHS. It was approved for routine use after the initial scope for this appraisal, and became a relevant comparator for the subgroup with non-squamous NSCLC and PD‑L1 TPS below 50% (see section 3.3). So, the company updated its indirect treatment comparison at consultation. It found that nivolumab combination was more effective than pembrolizumab plus pemetrexed and platinum chemotherapy in this subgroup. The ERG was concerned because:

3.8 The clinical experts explained that immunotherapy is generally well tolerated but is associated with some rare but unpleasant, and potentially serious, adverse events. These were likely to be more common for nivolumab combination (2 different immunotherapies) than current chemo-immunotherapy combinations (only 1 immunotherapy). The company did not agree, so at consultation provided further evidence using data from CheckMate‑227. It explained that:

3.9 The company used a 3‑state partitioned survival model to estimate the cost effectiveness of nivolumab combination compared with platinum-doublet chemotherapy, atezolizumab combination and pembrolizumab monotherapy. People were able to move to different health states; from pre-progression to post-progression and death and from post-progression to death. The ERG agreed with the company's model structure, noting that it was consistent with previous appraisals. For people having nivolumab combination and people having standard platinum-doublet chemotherapy, the company used the results from CheckMate‑9LA to model overall and progression-free survival for the first 13 months. The results from CheckMate‑227 were used after this point because longer-term data was available (see section 3.6). Survival curves were modelled for the ITT population and applied for everyone, regardless of histology and PD‑L1 TPS. Relative risks from the indirect comparisons were then applied to the nivolumab combination data to estimate overall and progression-free survival for atezolizumab combination and for pembrolizumab monotherapy. The committee concluded that the company's model structure was acceptable for decision making. But it noted the uncertainty about whether it was appropriate to use the same survival curves for everyone (see section 3.11).

3.10 The company included a 2‑year treatment stopping rule in its model. The maximum possible duration for nivolumab combination in CheckMate‑9LA was 24 months, and this was also stated in the summary of product characteristics. The committee understood that implementing a 2‑year stopping rule was consistent with other NICE technology appraisal guidance on untreated NSCLC. It also reflected how nivolumab combination would be used in clinical practice. The committee concluded that a 2‑year treatment stopping rule, in line with the clinical- and cost-effectiveness evidence, was acceptable.

3.11 The committee considered whether it was appropriate to use the same overall and progression-free survival curves across all 3 subgroups, based on the ITT data from CheckMate‑9LA and CheckMate‑227 (see section 3.9). The company took this approach because it believed there was a consistent efficacy benefit across subgroups in CheckMate‑9LA, including those based on PD‑L1 and histology. However, the ERG noted that the CheckMate‑9LA and CheckMate‑227 results suggested there were differences in the absolute and relative efficacy of nivolumab combination across some of the subgroups. Clinical advice to the ERG was that people whose NSCLC has a higher PD‑L1 TPS generally have greater benefit with anti‑PD‑L1 immunotherapies. The company considered that using the ITT data was more appropriate than the subgroup data, because the subgroups were not prespecified and included smaller numbers of people. The company also noted that the lack of external clinical data available to validate the subgroup curves placed greater reliance on clinical opinion. Also, it considered that by combining 3 different mechanisms of action, nivolumab combination was not expected to have the same efficacy differences by histology or PD‑L1 TPS as other immunotherapies. The committee considered that because it had agreed to separate the population into 3 subgroups (see section 3.2), it was appropriate to reflect this heterogeneity in the data used to generate the survival curves. It also noted that applying separate survival curves based on the subgroup data had a considerable impact on the cost-effectiveness results. The committee agreed that there was uncertainty about the validity of applying survival curves derived from the ITT data across all the subgroups. It concluded that survival curves should be modelled separately for each subgroup. So, in its updated base case after consultation, the company modelled the survival curves separately for each subgroup, based on the 1‑year data cut for CheckMate‑9LA and 3‑year data cut for CheckMate‑227. The company highlighted that CheckMate‑9LA not was not stratified or powered for analyses of combined histology and PD‑L1 subgroups. The committee concluded that the company's analyses modelling overall survival data for histology and PD‑L1 subgroups separately were appropriate for decision making.

3.12 The ERG noted that median overall survival for people having platinum-doublet chemotherapy was longer in CheckMate‑227 than in CheckMate‑9LA. This could have been because fewer people had subsequent therapy in CheckMate‑9LA than in CheckMate‑227. The ERG considered that the proportion of people having subsequent therapy after platinum-doublet chemotherapy in CheckMate‑9LA was lower than in NHS clinical practice. It was therefore concerned that using the CheckMate‑9LA data to estimate survival for the first 13 months (see section 3.9) may have underestimated survival for people having platinum-doublet chemotherapy. The ERG did a scenario analysis in which survival for people having platinum-doublet chemotherapy was modelled using the CheckMate‑227 data alone. In this, the relative efficacy of nivolumab combination was taken from the indirect treatment comparison. However, the ERG preferred to retain the company's original approach of using the CheckMate‑9LA data followed by the CheckMate‑227 data for its base case. This was because it used the same data sources for people having nivolumab combination and people having platinum-doublet chemotherapy. The committee noted that in NICE's technology appraisal guidance on pembrolizumab with carboplatin and paclitaxel for squamous NSCLC, it was estimated that around 50% of people having first-line chemotherapy have immunotherapy after disease progression. The clinical experts noted that this may be an underestimate based on the proportion of people switching from chemotherapy to pembrolizumab in the KEYNOTE‑024 trial. KEYNOTE‑024 was an open-label phase 3 randomised controlled trial, comparing pembrolizumab with chemotherapy for untreated metastatic NSCLC. At consultation, the company said that it preferred its original approach, in line with the nivolumab combination arm (CheckMate‑9LA data used up to 13 months, then conditional survival from CheckMate‑227 applied). Because CheckMate‑9LA was the registration trial, the company felt it was the most appropriate to estimate survival for people having platinum-doublet chemotherapy. It would preserve the benefits of comparing between arms of the randomised controlled trial. The ERG found that the model predicted 5‑year survival of 10.5% for platinum-doublet chemotherapy using CheckMate‑227 alone, and 9% in the approach using CheckMate‑9LA and CheckMate‑227. This was consistent with previous appraisals. The committee understood that around 28% of people had subsequent immunotherapy after first-line chemotherapy in CheckMate‑9LA, compared with around 41% in CheckMate‑227. It considered that the rate of subsequent immunotherapy in CheckMate‑227 was likely closer to that in NHS clinical practice. The committee concluded that survival for people having platinum-doublet chemotherapy should be modelled using the CheckMate‑227 data alone.

3.13 The company's base case included a lifetime treatment effect with nivolumab combination. The company justified this using pooled data from 4 clinical trials of nivolumab for previously treated NSCLC, which showed that nivolumab had a long-term survival advantage over docetaxel. The ERG noted that a lifetime treatment effect was inconsistent with previous technology appraisals for NSCLC. For those, a treatment effect lasting from 3 to 5 years after starting treatment had been accepted. The ERG also considered the pooled data used by the company to be of limited relevance. This was because the data was from trials of nivolumab as a monotherapy in a population who had previous treatment, and survival outcomes were only reported to 4 years. The ERG preferred a scenario with a treatment effect lasting 5 years after stopping treatment. This was modelled by setting the mortality rate as equal to that of platinum-doublet chemotherapy from this timepoint onwards. The committee understood that the assumption around the duration of treatment effect had a considerable impact on the cost-effectiveness results. It agreed that, although it was biologically plausible for the treatment effect to continue after stopping treatment with nivolumab combination, its duration was uncertain. The clinical experts explained that there was insufficient evidence to suggest the treatment effect of nivolumab combination lasted longer than for other immunotherapies. The committee concluded that a treatment effect lasting 3 to 5 years after starting treatment was appropriate for decision making, for consistency with previous immunotherapy appraisals in NSCLC.

3.14 At consultation, the company's updated base case used a 5‑year treatment effect after stopping treatment. This was more optimistic than what the committee preferred at the first meeting, which was a treatment effect lasting 3 to 5 years after starting treatment (see section 3.13). The company acknowledged the misunderstanding, but explained that a treatment effect of 5 years after stopping would be a conservative assumption. It explained that the new CheckMate‑227 data would support an additional benefit over other chemo-immunotherapy combinations, particularly the number of people whose disease maintained a response at 3 years. However, this new data did not arrive in time to be incorporated into its analyses. The committee recalled the plausible treatment effect duration for single-agent immunotherapy plus platinum-doublet chemotherapy and noted that strong evidence would be needed to accept a longer duration. It considered that there was no evidence that nivolumab combination had a longer treatment effect duration than pembrolizumab plus platinum-doublet chemotherapy. So, a more favourable assumption could not be made for nivolumab combination. The committee concluded that its preferred assumption for nivolumab combination was still a treatment effect lasting 3 to 5 years after starting treatment.

3.15 The ERG explained that the company's time-to-death approach for including utility values was not appropriate. This was because in previous technology appraisals in which this approach had been accepted, health-related quality-of-life data had only been collected for up to 30 days after stopping treatment. This meant that the utility value for the post-progression state may have been overestimated, because the full effects of progression may not yet have been evident. However, in CheckMate‑9LA, health-related quality-of-life data was collected until death, and there were many observations (1,004) contributing to the post-progression health state. In contrast, there were only 114 observations contributing to the utility value for 4 weeks or less to death in the company's approach. The ERG also had concerns with using time to death to determine health-related quality of life. A time-to-death approach meant that people entering the model had a different health-related quality of life depending on the treatment arm they were assigned to. The ERG preferred a progression-based approach, using utility values derived from the EQ‑5D data collected in CheckMate‑9LA. The clinical experts explained that quality of life may not immediately decline after disease progression. The committee was also aware that progression-based utility values may be overestimated because there were fewer observations in people with more severe disease. However, it agreed with the ERG that utility values based on disease progression were more appropriate for decision making, given the large amount of data captured after progression. At consultation, the company maintained that the time-to-death approach was an appropriate way to include utility values. But it accepted the use of health state utility values in line with the committee's preferred assumption.

3.16 To calculate the costs associated with platinum-doublet chemotherapy, the company based the distribution of chemotherapy regimens on the CheckMate‑9LA data. It applied the same assumption for everyone having platinum-doublet chemotherapy, regardless of histology or PD‑L1 TPS. The ERG considered that the distribution of chemotherapy regimens in CheckMate‑9LA may not reflect NHS clinical practice. For example, some people with squamous NSCLC were modelled as having pemetrexed, which they would not have in practice. It noted that there were differences in drug and administration costs between chemotherapy regimens, which meant that the costs calculated from the company's distribution may not have been representative. The ERG preferred to apply a specific distribution of chemotherapy agents for each subgroup (see section 3.2), including different proportions of people having pemetrexed maintenance therapy. It took these distributions from reported UK market share information in:

3.17 In the company's model, 31% of people having nivolumab combination as their first-line treatment had subsequent therapy, based on the CheckMate‑9LA data. The same assumption was used for people having pembrolizumab monotherapy and atezolizumab combination. In the model, 40% of people having platinum-doublet chemotherapy as their first-line treatment had subsequent therapy. The ERG noted that the data on the rates of subsequent therapy from CheckMate‑9LA was immature, and likely an underestimate. The rates from CheckMate‑227 were higher (45% for people having nivolumab combination as first-line treatment, and 61% for people having platinum-doublet chemotherapy). The ERG considered that the CheckMate‑227 rates were more in line with those seen in NHS clinical practice. They were also based on more mature data. Also, using the CheckMate‑227 data was consistent with the approach used for modelling long-term survival. At technical engagement, the company accepted that the rates of subsequent therapy in CheckMate‑9LA may be lower than expected in a clinical trial, but considered that they reflected NHS clinical practice. The committee recalled its earlier conclusion that the proportion of people having subsequent therapy in the clinical trials was likely lower than in NHS clinical practice (see section 3.12). Therefore, it concluded that the higher rates from CheckMate‑227 better reflected clinical practice and should be used in the model.

3.18 Relative dose intensity is the percentage of the prescribed dose of a treatment that people take. The company applied the relative dose intensities to the cost of the drug, after this had been estimated from the number of vials needed based on the treatment dose in the marketing authorisation. The ERG considered it more appropriate to apply the relative dose intensities to the expected treatment dose, and then calculate the number of vials and associated drug costs from these adjusted numbers. Because this may not necessarily reduce the number of vials, the ERG was concerned that the company may have underestimated the drug costs. The committee concluded that the relative dose intensity likely lay between the company's and the ERG's assumptions. However, it noted that this had minimal impact on the cost-effectiveness results.

3.19 The committee considered the advice about life-extending treatment for people with a short life expectancy in NICE's guide to the methods of technology appraisal. It understood that, at the first meeting, the company and the ERG agreed that nivolumab combination did not meet the criteria for end of life treatments for:

3.20 At consultation, the company presented evidence to support that the end of life criteria should apply across all the populations in this appraisal. This included a retrospective study using data from the Flatiron database, for people with stage 3B or stage 4 NSCLC who had chemo-immunotherapy (over 98% had pembrolizumab plus chemotherapy). The reported median overall survival for people with squamous and non-squamous NSCLC was shorter than in the KEYNOTE‑407 and KEYNOTE‑189 trials. However, the ERG considered it uncertain whether the Flatiron evidence reflected the population who would have nivolumab combination for untreated, metastatic NSCLC in the NHS. This was because people in the Flatiron dataset had poorer performance status, were older, and were more likely to have unstable brain metastases. Also, they had treatment in the US, which has different treatment patterns to the UK. The company contrasted the real-world Flatiron evidence with real-world evidence for nivolumab as second-line treatment for NSCLC. It explained that outcomes in clinical practice (from the Systemic Anti-Cancer Therapy Dataset) were similar to those seen in the randomised controlled trials. The committee recalled that in NICE's technology appraisal guidance on pembrolizumab with pemetrexed and platinum chemotherapy for non-squamous NSCLC, the committee did not accept that the end of life criteria were met. The ERG noted that there was little long-term observational evidence on whether having nivolumab first line would extend survival by 3 months in clinical practice. It considered that efficacy might decline in practice (similar to the clinical trial) when people with more severe disease have nivolumab. Also, survival may not be extended by 3 months for people with severe disease. But the ERG noted that this was uncertain and not supported by evidence. The Cancer Drugs Fund clinical lead explained that real-world outcomes for chemo-immunotherapy seem to reflect what is seen in trials. This is because people must be well enough to have chemo-immunotherapy, which has additional selection considerations for clinicians, compared with immunotherapy alone. The committee considered that the evidence presented by the company only showed that the end of life criteria could plausibly be met for the subgroup with squamous NSCLC and PD‑L1 TPS below 50%. It recalled that mean survival was preferable to median survival for decision making around end of life criteria for chemo-immunotherapy. The committee concluded that the criteria were not met for all other populations covered by this appraisal.

3.21 The committee recalled that its preferred assumptions were:

3.24 Nivolumab combination is more clinically effective than standard chemotherapy, and is likely to have similar efficacy to other chemo-immunotherapies. The committee agreed that the most plausible ICERs for nivolumab combination compared with platinum-doublet chemotherapy, atezolizumab combination and pembrolizumab monotherapy were mostly higher than what NICE normally considers a cost-effective use of NHS resources. The results comparing nivolumab combination with pembrolizumab plus pemetrexed and platinum chemotherapy did not seem realistic, and that analysis was not considered robust enough for decision making. Therefore, it concluded that nivolumab combination could not be recommended for routine use as an option for untreated metastatic NSCLC with no EGFR or ALK mutations.

3.25 Having concluded that nivolumab combination could not be recommended for routine use, the committee then considered if it could be recommended within the Cancer Drugs Fund. The committee discussed the arrangements for the Cancer Drugs Fund agreed by NICE and NHS England in 2016, noting NICE's Cancer Drugs Fund technology appraisal process and methods guide (addendum). The company had expressed an interest in the treatment being considered for funding through the Cancer Drugs Fund. At consultation, the company provided the most up-to-date overall survival results for CheckMate‑9LA compared with CheckMate‑227, for subgroups with: