Regorafenib for previously treated unresectable or metastatic gastrointestinal stromal tumours

The patient and clinical experts explained that gastrointestinal stromal tumours (GIST) are a rare disease that often takes years to diagnose because people have non-specific symptoms. The clinical expert advised that first-line treatment for metastatic GIST is imatinib, which is generally well tolerated. Once disease has progressed, patients are switched to sunitinib, which on average gives 6 to 12 months benefit before the disease progresses again but has more side effects than imatinib and requires careful monitoring. The patient and clinical experts explained that there are no alternative third-line treatment options after disease progression or if patients are intolerant to imatinib and sunitinib other than best supportive care. Regorafenib is currently available on the Cancer Drugs Fund for metastatic GIST after disease progression or intolerance to imatinib and disease progression after sunitinib, only if the patient has an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 1. The clinical expert explained that the only alternative to best supportive care and regorafenib (through the Cancer Drugs Fund) was to participate in a clinical trial but noted that these were rare. The committee concluded that best supportive care is the most appropriate comparator for this appraisal.

The patient experts advised that regorafenib is well tolerated, with side effects (such as sore feet and hands, constipation and diarrhoea) that can be managed. The clinical experts described how dose adjustments and scheduling were important in treating GIST and that these are managed on an individual basis to achieve a balance between the minimal effective dose and side effects that can be tolerated. The patient experts explained that regorafenib allows people to remain actively engaged in work and family life, while reducing anxiety and increasing quality of life. Regorafenib also provides an important further treatment option to people who cannot tolerate sunitinib. One patient expert described how extending life was a priority to enable patients to spend more time with their families and it was important to be able to lead a near-normal life while having treatment. The committee recognised the importance of a potentially life-extending treatment with manageable side effects. The committee concluded that a treatment option for people with metastatic or unresectable GIST after imatinib and sunitinib would be greatly valued, particularly if it extended life and had manageable side effects.

The evidence for regorafenib submitted by the company came from GRID, a multicentre double-blind randomised controlled trial with 199 patients who had previously had treatment with imatinib and sunitinib. It compared regorafenib plus best supportive care (133 patients) with placebo plus best supportive care (66 patients). Best supportive care included various treatments such as analgesics, radiation therapy for pain control and palliative surgery. After disease progression, patients receiving placebo were given the option to cross over to regorafenib, and patients already receiving regorafenib were offered open-label regorafenib if it was considered clinically beneficial. In the trial, treatment could be delayed or reduced according to a pre-specified schedule. The clinical experts confirmed that dose reductions were common in clinical practice, because of the side effects associated with regorafenib, and managed on an individual basis. The clinical experts explained that treatment would only be stopped in clinical practice if there was clear disease progression and worsening clinical symptoms. Clinical guidelines recommend continued treatment with tyrosine kinase inhibitors as long as there is continued benefit. The clinical experts also advised that there was uncertainty around whether patients in the GRID trial could have other off-label treatments after disease progression that are not standard practice in the NHS. The committee concluded that using regorafenib after disease progression was in line with the marketing authorisation and current clinical practice.

Patients in the GRID trial had an ECOG performance status of 0 to 1 and the summary of product characteristics for regorafenib states that there are limited data in patients with a performance status of 2 or above. Regorafenib is currently provided by the Cancer Drugs Fund to people who have a performance status of 0 to 1, in line with trial evidence. The clinical experts explained that very few people for whom regorafenib would be considered present with a performance status of 2 in clinical practice; this is because treatment is actively managed in specialist centres to prevent performance status from deteriorating, particularly given the tolerability issues associated with sunitinib. The committee concluded that in line with the clinical evidence and the use of regorafenib in clinical practice, it would take performance status into account when making its final recommendations.

In the GRID trial there was a high level of crossover (88%) from the placebo arm to open-label regorafenib after disease progression. The company and the ERG agreed that a statistical correction for crossover was needed to produce unbiased estimates of overall survival in the placebo arm. The company carried out 2 crossover corrections using the iterative parameter estimation (IPE) and rank preserving structural failure time (RPSFT) methods. Both methods aimed to reconstruct individual patient data for overall survival in the placebo arm as if there had been no crossover. The ERG advised that both the IPE and RPSFT methods were appropriate for correcting crossover. The committee concluded that given the high level of crossover from placebo to regorafenib, an intention-to-treat analysis was not appropriate and a statistical correction for overall survival was needed.

Data from 2015 from the GRID trial showed that regorafenib improved progression-free survival (hazard ratio [HR] 0.27, 95% confidence interval [CI] 0.19 to 0.39). Data from 2017 showed that after crossover correction, median overall survival in the placebo arm was around 8 months compared with 17.4 months for regorafenib. The 2017 data showed a larger overall survival benefit after correcting for crossover (unadjusted HR 0.90; 95% CI 0.68 to 1.19, RPSFT HR 0.48; 95% CI 0.07 to 3.33, IPE HR 0.45; 95% CI 0.06 to 3.69). The committee understood that although not statistically significant, the point estimates did show improved overall survival with regorafenib. The committee concluded that there was evidence that regorafenib is clinically effective in improving progression-free survival compared with best supportive care but that there was uncertainty about the magnitude of overall survival benefit.

The company submitted a partitioned survival model with 3 health states (progression free, progressed and death) that used survival data from the GRID trial. The company's base case modelled treatment duration by using the treatment discontinuation curve from GRID. Therefore the company's base case included the additional cost of regorafenib in the treatment arm after disease progression in line with the GRID trial (open-label regorafenib was offered to patients randomised to regorafenib after progression if it was considered clinically beneficial). Use of regorafenib after disease progression is in line with clinical practice (see section 3.3). The company's base case also incorporated dose intensity by including the mean observed dose of regorafenib from GRID (including doses of 0 mg). The ERG agreed that the company's approach to modelling treatment duration and dose intensity was appropriate. The committee concluded that the company's model, using trial data to model treatment duration and dose intensity and including the additional cost of regorafenib after disease progression, was appropriate for decision-making.

The company's base case included overall survival estimates based on the most recent data from GRID (2017 data). The ERG initially had several concerns with using the 2017 data related to:

• methodological assumptions in the treatment switching adjustment

• the impact of recensoring on adjusted overall survival and the cost effectiveness of regorafenib

• reasons for a large decrease in overall survival in the placebo arm after adjusting for treatment switching when using the 2017 data compared with the 2015 data.
  
  Using different data cut-offs for estimating overall survival had a large impact on the cost-effectiveness results. Regorafenib appeared to be more cost effective when using the 2017 data for overall survival than when the 2015 data were used. The company responded to the ERG's concerns with additional explanations, and the ERG accepted the company's rationale and justifications of the methods used. The ERG confirmed that there was no conclusive evidence that the company had performed the treatment switching adjustment incorrectly, and accepted it was appropriate to use the most recent (2017) data. The committee concluded that while there was some remaining uncertainty, the analysis using 2017 data for estimating overall survival was acceptable.

The company's base case included an estimated treatment effect for overall survival with an IPE crossover correction. The ERG advised that the IPE method is an extension to the RPSFT method using parametric methods and the assumption of a common treatment effect (that is, people receive the same treatment benefit regardless of when they receive treatment). This assumption may not hold if people in the trial were able to receive a variety of off-label treatments after disease progression (see section 3.3). The ERG noted that both the IPE and RPSFT methods were appropriate and gave similar estimates of overall survival for the placebo arm (median overall survival IPE 8.0 months and RPSFT 8.4 months). In its revised explorations, the ERG considered both IPE- and RPSFT-corrected overall survival estimates as acceptable. The committee recognised that using the RPSFT rather than the IPE method did not have a large impact on the cost-effectiveness estimate of regorafenib. The committee concluded that cost-effectiveness results using either IPE or RPSFT adjustments should be considered in its decision-making.

The company's base case included an IPE adjustment with recensoring (in line with guidance from the Decision Support Unit). Recensoring is used to reduce the risk of bias associated with a treatment-switching adjustment, and recensoring data at an earlier time point aims to avoid informative censoring (because treatment switching is not random). The ERG advised that recensoring may lead to biased estimates of the average treatment effect when the proportional treatment effect assumptions do not hold, and there is some academic debate on whether to use recensoring because the estimated treatment effect is generally larger when it is used. The company provided further analyses that explored the impact of recensoring. The ERG advised that it was appropriate to consider analyses with and without recensoring. The committee noted that recensoring had a large impact on the cost-effectiveness estimate and that regorafenib appeared to be more cost effective if recensoring was used. The committee concluded that the arguments for and against recensoring were evenly balanced and that results both with and without recensoring should be considered when assessing the cost effectiveness of regorafenib.

The company's base case included a log-logistic extrapolation of overall survival because it provided the best fit to the trial data from GRID. The company highlighted that log-logistic, Weibull and Gompertz models were considered clinically plausible by the clinical experts. The ERG advised that clinical plausibility of extrapolations was critical and the evidence for longer-tailed and shorter-tailed distributions appeared to be evenly balanced. The clinical experts explained that very few people with unresectable or metastatic GIST whose disease has progressed on, or who are intolerant to, treatment with imatinib and sunitinib will survive for 10 years. The choice of extrapolation method for estimating overall survival had a large impact on the cost-effectiveness results for regorafenib. A log-logistic extrapolation resulted in a larger proportion of people surviving after 10 years compared with the Weibull extrapolation and the clinical experts and committee considered this to be optimistic. The committee agreed that a Weibull extrapolation would be more appropriate because it had a shorter tail length and more plausibly reflected the survival of people at this stage of the disease. It noted that in the company's base case, regorafenib became less cost effective when the Weibull and Gompertz models were used. The committee concluded that an overall survival extrapolation that used a shorter-tailed Weibull model was most appropriate because it best reflected the survival outcomes of the population.

The ERG explained that causes of mortality in the GRID trial would largely be related to GIST but it was likely that additional general mortality would have occurred after the trial ended. The ERG's base case incorporated this additional background mortality but the company's did not. The committee understood that the impact of background mortality was likely to be lower when a shorter-tailed Weibull extrapolation was used. The committee concluded that including an adjustment for general mortality was appropriate.

Utility estimates for health-related quality of life were taken from the GRID trial, which reported the EQ-5D and European Organisation for Research and Treatment of Cancer Core Quality of Life Questionnaire (EORTC QLQ-C30). Both the company and ERG base cases used EQ-5D data from paired samples that were not split by treatment arm. The company stated that a repeated measures analysis would have been biased because more measurements were taken when patients did not have disease progression and there were no clinically meaningful differences in EQ-5D between the treatment arms. The ERG's base case also applied age-related utility decrements because utility often declines with age; this was not included in the company's base case. The committee understood that the addition of age-related utility decrements did not have a large impact on the cost-effectiveness results. The committee concluded that including age-related utility decrements was appropriate.

The committee considered the advice about life-extending treatments for people with a short life expectancy in NICE's Cancer Drugs Fund technology appraisal process and methods. In the GRID trial the adjusted median overall survival in the placebo plus best supportive care arm was less than 12 months (RPSFT-corrected median overall survival 8.4 months and IPE-corrected median overall survival 8.0 months). The company's economic model predicted mean undiscounted life years of 0.97 for the IPE-adjusted placebo arm and 1.01 for the RPSFT-adjusted placebo arm when recensoring was applied. The committee concluded that regorafenib was indicated for people with a short life expectancy. The median overall survival improvement in the GRID trial was at least 9 months for patients treated with regorafenib compared with those having best supportive care, depending on which crossover correction method was used. The crossover-corrected mean overall survival benefit for regorafenib was around 1.2 years longer compared with best supportive care in the ERG's revised base case, but less than 1.2 years when recensoring was used (estimates taken from the company's economic model). The committee concluded that regorafenib offered an extension to life of at least 3 months compared with best supportive care.

The committee considered the cost effectiveness of regorafenib as estimated using its preferred analyses, which included:

• 2017 data for overall survival (see section 3.8)

• crossover adjustment using either RPSFT or IPE (see section 3.9)

• analyses with and without recensoring (see section 3.10)

• overall survival extrapolation using a Weibull distribution (see section 3.11)

• background mortality (see section 3.12)

• age-related utility decrements (see section 3.13).
  
  The committee discussed that, using its preferred analyses, the incremental cost-effectiveness ratio (ICER) was likely to lie between £40,000 and £48,000 per quality-adjusted life year (QALY) gained. It concluded that the most plausible ICER was the midpoint of that range: around £44,000 per QALY gained. The committee concluded that, given that regorafenib met the criteria for being a life-extending, end-of-life treatment, it could be recommended as a cost-effective use of NHS resources.

The committee considered whether its recommendations were associated with any potential issues related to equality. The committee concluded that healthcare professionals should take into account any physical, sensory or learning disabilities, or communication difficulties that could affect ECOG performance status and make any adjustments they consider appropriate.

The Pharmaceutical Price Regulation Scheme (2014) payment mechanism was not relevant in considering the cost effectiveness of technology.