Nivolumab for adjuvant treatment of invasive urothelial cancer at high risk of recurrence

3.1 Muscle-invasive urothelial cancer can have a significant impact on people and their families and carers. Standard care after radical resection is platinum‑based chemotherapy (adjuvant treatment) or best supportive care. Some people have platinum‑based chemotherapy before the surgery (neoadjuvant treatment) and would not be eligible for adjuvant treatment with platinum‑based chemotherapy. Despite resection, the disease can recur. High risk is defined by the MIUC pathologic staging criteria in section 5.1 of nivolumab's summary of product characteristics. The patient experts explained that there is a high unmet need in this area and a new treatment option at this part of the pathway was welcomed. They explained that for some people platinum‑based chemotherapy is not suitable or tolerated, and some people are unwilling to have it. The patient experts explained that extending the duration of disease-free survival is important to patients. This is because it allows them to spend more time with their families and enjoy a good quality of life, and relieves stress on carers. The clinical and patient experts noted that nivolumab was generally well tolerated and that the short infusion time of the treatment compared with chemotherapy was an advantage. The clinical experts explained that immunotherapy at an early stage has the potential to significantly improve outcomes and increase the number of people whose cancer is cured. The committee considered that adjuvant treatment with nivolumab after radical resection may address an unmet need. The committee acknowledged that nivolumab is the first adjuvant immunotherapy available for resected high-risk muscle-invasive urothelial cancer. It concluded that nivolumab is a valued treatment option for people with resected high-risk muscle-invasive urothelial cancer.

3.2 The final NICE scope included adjuvant chemotherapy and best supportive care (active monitoring) as comparators. The CheckMate 274 trial only included a placebo (best supportive care) comparator arm. The company provided cost-effectiveness estimates for nivolumab compared with best supportive care but did not provide cost-effectiveness estimates comparing nivolumab with adjuvant chemotherapy. The company explained that this was because adjuvant chemotherapy use is low in current NHS practice. It explained that this is because of a lack of evidence of clinical benefits and people may refuse chemotherapy because of toxicity concerns. The clinical experts agreed that the use of adjuvant platinum‑based chemotherapy was low for people with bladder cancer. This is because some people have neoadjuvant platinum‑based chemotherapy, and some people are not fit enough to have platinum‑based chemotherapy or they have toxicity concerns. The experts explained that the evidence base for adjuvant platinum‑based chemotherapy was not robust for bladder cancer. The Cancer Drugs Fund clinical lead and the clinical experts explained that for people with urothelial carcinomas of the upper urinary tract, adjuvant platinum‑based chemotherapy was likely to be standard of care. This was because clinical trial evidence from the POUT trial showed an increase in disease‑free survival when platinum‑based adjuvant chemotherapy was given within 90 days of resection (Birtle et al. 2020). In addition, neoadjuvant platinum‑based chemotherapy is usually not suitable for treating urothelial carcinomas of the upper urinary tract. The committee considered that there are several reasons why adjuvant platinum‑based chemotherapy may not be suitable. These reasons included:

3.3 The company's scenario analysis assumed that only people who had best supportive care after resection would have the option of immunotherapy (atezolizumab) if disease recurrence occurred. This treatment option was based on NICE's technology appraisal guidance on atezolizumab for untreated PD-L1-positive advanced urothelial cancer when cisplatin is unsuitable. The Cancer Drugs Fund clinical lead explained that people who have nivolumab after resection may have immunotherapy again after disease recurrence if enough time has passed since nivolumab treatment had stopped (approximately 12 months). The ERG provided a scenario in which atezolizumab was a treatment option for both the nivolumab and best supportive care groups after disease recurrence. The committee concluded that retreatment with immunotherapy would be offered to some people who have disease recurrence after adjuvant treatment with nivolumab.

3.4 CheckMate 274 is an ongoing phase 3 randomised controlled trial comparing nivolumab with placebo in people with resected muscle-invasive urothelial cancer at high risk of disease recurrence. People had nivolumab for up to 1 year. The trial included 353 people in the nivolumab arm, of whom 140 had tumours expressing PD‑L1 at a level of 1% or more. The trial included 356 people in the placebo arm, of whom 142 had tumours expressing PD‑L1 at a level of 1% or more. CheckMate 274 included people who were eligible to have adjuvant cisplatin (platinum‑based chemotherapy). However, they were only allowed to enrol in the trial if they had documented reasons for refusing adjuvant cisplatin. Disease‑free survival was the primary outcome measure. Median disease‑free survival was not reached in the nivolumab arm in the currently available data (95% confidence interval [CI], 22.1 months to not estimable). Median disease‑free survival in the placebo arm was 8.4 months (95% CI, 5.6 months to 20.0 months). At 6 and 12 months, 74.5% and 67.6% of people in the nivolumab arm were disease-free, respectively. This compared with 55.7% and 46.3% being disease-free in the best supportive care arm. The committee noted that in CheckMate 274, evidence for nivolumab was less encouraging for people with upper tract urothelial cancer compared with the intention‑to‑treat population (hazard ratios: renal pelvis tumour origin, 1.25, 95% CI 0.70 to 2.25; ureter tumour origin, 1.54, 95% CI 0.69 to 3.44). The committee concluded that the currently available data showed that nivolumab increased disease‑free survival compared with placebo in people whose tumours express PD‑L1 at 1% or more.

3.5 The committee noted that because overall survival data from CheckMate 274 was event driven, the currently available data did not provide information on survival. The committee was aware that some published evidence suggested that disease‑free survival gains may not necessarily lead to overall survival gains (Sternberg et al. 2015). The clinical and patient experts emphasised the importance of disease-free survival in the adjuvant treatment setting. A clinical expert explained that the Sternberg et al. study may not be a reliable predictor of overall survival gains with nivolumab. This is because the study recruited over a long period of time, and many participants were enrolled a long time after having surgery. The Cancer Drugs Fund clinical lead highlighted that nivolumab has a different mechanism of action to platinum‑based chemotherapy and therefore the extension in disease‑free survival with nivolumab might translate into improved survival. The committee acknowledged these comments but considered that the lack of overall survival data was a key uncertainty in the analysis. But, it noted that it would take several years for the trial to show this data because nivolumab is positioned at a less severe part of the treatment pathway. The committee concluded that it is not certain to what extent a benefit in disease‑free survival translates into a benefit in overall survival.

3.6 CheckMate 274 did not compare nivolumab with platinum‑based chemotherapy (see section 3.4), so the company provided an indirect treatment comparison. One comparison included people in CheckMate 274, in both the nivolumab and placebo groups, who refused cisplatin. Another comparison compared the CheckMate 274 nivolumab group with evidence from published studies of platinum‑based chemotherapy. In both comparisons, the results showed that outcomes associated with nivolumab and platinum‑based chemotherapy were not statistically significantly different. The exact results of the indirect treatment comparisons are considered academic in confidence and cannot be reported here. The company excluded studies that included upper tract urothelial cancer from its indirect comparison. The committee recalled that adjuvant chemotherapy was a relevant treatment option for this group (see section 3.2). The company explained that CheckMate 274 was not powered to detect differences for upper tract disease, and that providing this indirect comparison would reduce the number of patients that would inform the analysis. The committee noted the company's reasons but considered that an indirect treatment comparison for upper tract urothelial cancer would have been informative. The committee concluded that the company's indirect treatment comparison is highly uncertain. It also concluded that a comparison including only upper tract urothelial cancer would have been the most relevant comparison because of the treatment pathway for upper tract disease.

3.7 The company's model was a Markov model with 4 health states: disease‑free survival, long‑term disease-free, recurred disease and death. In the recurred disease state, the company's base case included 1 line of treatment which was assumed to be either cisplatin or carboplatin. The ERG noted that it was assumed this treatment would be continued until death, which it did not consider to be appropriate. The committee noted that the model assumed equal efficacy in both treatment arms in the recurred disease state. This meant that increases to disease-free survival would translate to overall survival improvements. The committee recalled that the extent to which an increase in disease‑free survival translated to an increased overall survival was a key uncertainty (see section 3.5). The company also used a simplified approach to model survival in the recurred disease health state by applying an exponential function (assuming constant hazards) to overall survival using median values from the literature. The committee noted this did not allow survival rates to vary over time. The committee was also aware that there were further treatment lines available which the company's model did not account for. In particular, it did not include NICE's technology appraisal guidance on atezolizumab for treating locally advanced or metastatic urothelial carcinoma after platinum-containing chemotherapy or avelumab for maintenance treatment of locally advanced or metastatic urothelial cancer after platinum-based chemotherapy. The company did include a scenario which included atezolizumab for untreated PD-L1-positive advanced urothelial cancer when cisplatin is unsuitable, but only for people who had not had nivolumab. The committee expressed concern that the company could have modelled post‑recurrence outcomes more robustly, for example, using tunnel states to allow transition rates to vary over time. The committee also highlighted that the model did not include all treatments currently considered standard care in NHS practice, although it understood avelumab had only recently been recommended by NICE, which limited the generalisability of results. However, the committee was reassured by the ERG that the omission of these treatments was unlikely to bias the cost‑effectiveness estimates in favour of nivolumab. This was because the cost‑effectiveness estimates for nivolumab would likely improve if these treatments were included because nivolumab was predicted to cure a higher proportion of people than best supportive care, therefore avoiding less cost‑effective treatments (see section 3.9). The committee concluded that the company's economic model is appropriate for decision making but does not model disease recurrence robustly.

3.8 The economic model provided by the company uses a generalised gamma distribution to model disease‑free survival for both treatment arms. This distribution was deemed appropriate by the company because it had the best fit to the data. It also allows for a better accounting of the protocol-induced features of the hazard profiles, in particular the steep decline seen at 3 months which coincided with tumour assessments. The ERG agreed that this distribution was potentially appropriate. But, it may not have a desirable fit if the patterns of events observed in the trial, which may be influenced by the timing of data collection, are not reflective of what would happen in clinical practice. Additionally, the ERG noted that the generalised gamma distribution produces a higher risk for disease‑free survival events at 5 years than the general population hazard of mortality. This did not match the company's 5‑year cure assumption (see section 3.9). The ERG advised that the Gompertz distribution is also informative but results in a cure point earlier than 5 years. But, it noted that the choice between these 2 potentially appropriate distributions has only a minimal impact on the cost‑effectiveness estimates. The committee agreed that the choice of distribution is unlikely to affect decision making. It concluded that both the generalised gamma and Gompertz distributions are potentially appropriate for estimating disease‑free survival.

3.9 The company's base‑case model assumes a cure point at 5 years. This means that there will not be a disease recurrence after 5 years in a disease‑free survival state. The company highlighted that evidence from CheckMate 274 demonstrates that risk of death approaches that of the general population at 5 years. Clinical and patient experts agree that after 5 years in a disease‑free state the risk of disease recurrence is low; however, there was some evidence to show that disease recurrence can happen after 5 years in a small number of cases. The ERG provided an exploratory analysis using a 10‑year cure point. It cited evidence from Sternberg et al. (2015) which demonstrated an increased mortality risk for people with resected urothelial cancer compared with the general population, even after 5 years in a disease‑free state. The committee noted that there is uncertainty surrounding which of these points in time is the most accurate. But, it agreed that the exploratory analyses provided by the ERG had a minimal impact on the cost‑effectiveness results. The committee concluded that there is evidence that disease recurrence may take place after 5 years and mortality risks remain elevated. But, it remains uncertain at what point it is reasonable to assume that people are cured.

3.10 The model provided by the company assumed that people in the disease‑free survival state after radical surgery have the same health utility as age- and sex‑matched people from the general population. The company highlighted that people in CheckMate 274, who were disease‑free, had higher utility values than those of the age- and sex‑matched general population. Clinical and patient experts agreed that, after a period of adjustment, people with resected urothelial cancer often adapt very well to post‑surgical changes and often achieve a good quality of life with a fully functional lifestyle. They accepted that quality of life for this population is likely to be impacted in the first 1 or 2 years after surgery. The clinical experts noted that while a good quality of life is likely, there is potential for reduced quality of life. This is because of the increased presence of comorbidities in this population, the potential for some persistent effects from radical surgery (such as those impacting on sexual function) and persisting adverse events of treatment. Clinical advice to the ERG also suggested that a reduced quality of life compared with the general population was to be expected. Without evidence to inform a specific disutility value in this population, the ERG applied a disutility of 0.02 to their analyses to test the impact of disutility on the cost effectiveness of nivolumab. This was up until the time at which the cure point is applied. The committee agreed that there was likely to be a health disutility for this population. But, it noted that there was not enough evidence to inform what value this disutility should be, and that the ERG's exploratory analyses had a minimal impact on the cost‑effectiveness results. The committee concluded that disease‑free utility values may be overestimated in the company's analysis.

3.11 The company's model assumed that people in the disease‑free survival state for 5 years have the same life expectancy as the general population. Clinical experts agreed that the risk of disease recurrence is low after 5 years and noted that discharge from follow‑up is typical at this point (see section 3.9). The ERG advised that there is some evidence of an increased risk of death, even after being disease‑free for 5 years. It noted that the generalised gamma distribution used in the company model also indicated a mortality risk greater than the general population. It therefore provided a scenario analysis in which an increased risk of mortality is applied between years 5 and 10 in the model. The committee had concerns with the modelling provided by the company but agreed that there is a lack of definitive evidence in this area. It also agreed that the ERG's analyses did not have a substantial impact on the cost‑effectiveness estimates. The committee concluded that the company's assumption that people in the long-term disease-free health state have the same life expectancy as the general population is optimistic.

3.12 The company provided an indirect comparison of nivolumab with adjuvant chemotherapy (see section 3.6). However, the company did not provide any cost‑effectiveness analysis comparing nivolumab with adjuvant platinum‑based chemotherapy. The company explained that this was because of the limitations of the indirect treatment comparison. The ERG considered that the cost‑effectiveness results provided by the company were only relevant in circumstances in which platinum‑based chemotherapy was not suitable (see section 3.2). However, the committee also recalled that for some people, particularly those with upper tract urothelial cancer, adjuvant platinum‑based chemotherapy would be an appropriate treatment option (see section 3.2). This included carboplatin for people with renal impairment. The committee recalled that in CheckMate 274, the evidence for nivolumab was less encouraging for people with upper tract urothelial cancer compared with the intention‑to‑treat population (see section 3.4). The ERG stated that it was unlikely that nivolumab would be considered cost effective compared with adjuvant platinum‑based chemotherapy, in people for whom it is suitable. This is because the indirect comparison results showed no overall statistically significant difference in treatment effect and the cost of nivolumab is higher. The committee concluded that because it had not been presented with any cost‑effectiveness results comparing nivolumab with platinum‑based chemotherapy, the cost‑effectiveness results only apply when adjuvant platinum‑based chemotherapy is unsuitable.

3.13 The company's base‑case incremental cost‑effectiveness ratio (ICER) for nivolumab was £11,361 per quality‑adjusted life year (QALY) gained compared with best supportive care. The company's base‑case analysis included the following key assumptions:

3.17 The committee considered that nivolumab is a promising new adjuvant treatment for people with muscle‑invasive urothelial cancer at high risk of recurrence after radical resection and whose tumours express PD‑L1 at a level of 1% or more. However, there are uncertainties in the extent that improved disease‑free survival translates into overall survival gains. The committee was not presented with any cost‑effectiveness analyses comparing nivolumab with platinum‑based chemotherapy. It therefore considered that the cost‑effectiveness results were only relevant to situations when platinum‑based chemotherapy was not suitable. These situations included: the lack of evidence of effect in people with bladder cancer in the adjuvant treatment setting; use of neoadjuvant platinum‑based chemotherapy; when platinum‑based chemotherapy is not suitable for the person or will not be tolerated; or if the person refuses platinum‑based chemotherapy after discussing the benefits and risks with their oncologist. The committee considered that platinum‑based chemotherapy was likely to be an appropriate treatment option for people with urothelial carcinomas of the upper urinary tract, given the POUT trial evidence available. Compared with best supportive care, the ICERs for nivolumab were considered a cost‑effective use of NHS resources when platinum‑based chemotherapy is not suitable. Therefore, the committee recommended nivolumab for routine commissioning, but only if platinum‑based chemotherapy is unsuitable.