Trifluridine–tipiracil for treating metastatic gastric cancer or gastro-oesophageal junction adenocarcinoma after 2 or more treatments

3.1 The initial symptoms of gastric cancer are vague and similar to other stomach conditions, but for advanced disease they may include lack of appetite, weight loss, fluid in the abdomen and blood in the stool. The clinical experts estimated that life expectancy after 2 previous treatments is between 2 and 4 months in current practice. They explained that there is no standard treatment for previously treated metastatic gastric cancer but in clinical practice in the NHS in England, treatment is usually in line with the European Society for Medical Oncology (ESMO) guideline for gastric cancer. The clinical experts advised that paclitaxel is generally used after 1 treatment, and irinotecan may be used after 2 treatments but for most people it is not appropriate because of the risk of side effects. They estimated that third-line chemotherapy is used in about 10% of people, with most people having best supportive care alone. The committee was aware that the ESMO guideline had recently been updated to recommend trifluridine–tipiracil as a third-line treatment option for people with an Eastern Cooperative Oncology Group (ECOG) performance score of 0 or 1. The committee noted that there was no patient expert submission for this appraisal, but the clinical experts explained that maintaining health-related quality of life is very important to people with metastatic gastric cancer. They advised that an oral treatment such as trifluridine–tipiracil would be preferred because it does not need many hospital visits, allowing people to spend more time at home. The committee concluded that there is an unmet need for third-line treatment options for gastric cancer.

3.2 The company submitted cost-effectiveness analyses comparing trifluridine–tipiracil and best supportive care with placebo and best supportive care. It advised that there is a lack of evidence to support the use of third-line chemotherapy and that its expert advice suggested this is usually restricted to clinical trials. The committee recalled that third-line chemotherapy is appropriate but is used in only a small proportion of people in current practice, with most people having best supportive care alone (see section 3.1). The clinical experts explained that although there is no clear definition of best supportive care, it usually includes treatments to control symptoms such as pain. The committee concluded that the most appropriate comparator is best supportive care.

3.3 The clinical evidence for trifluridine–tipiracil came from TAGS, a phase 3 randomised controlled trial. It compared trifluridine–tipiracil and best supportive care with placebo and best supportive care in 507 adults with metastatic gastric cancer (including 29% with gastro-oesophageal junction cancer) who had had at least 2 treatments for advanced disease, and who had an ECOG performance score of 0 or 1. The committee was aware of several issues that may impact the generalisability of the full intention-to-treat analysis from TAGS to the NHS in England:

3.4 The EAG advised that the committee's preferred third-line, European subgroup (see section 3.3) had imbalances in patient characteristics between the 2 arms of the TAGS trial. Some of these could favour survival after treatment with trifluridine–tipiracil and some could favour survival after treatment with placebo (the exact numbers are confidential and cannot be reported here). The company accepted that any subgroup analysis may be at risk of imbalances in characteristics, but it felt there were no imbalances in verified prognostic factors in this analysis. After the second meeting, the committee requested additional analyses to adjust the third-line data for imbalances in:

3.5 The company included a partitioned survival cost-effectiveness model in its evidence submission. The model comprised 3 health states representing progression-free disease, progressed disease and death. Health-state occupancy over time was informed by survival functions from TAGS data. The EAG advised that the model was generally clear and appropriate. The committee concluded that the company's model was suitable for decision making.

3.6 In TAGS, the most common side effects included nausea, anaemia, decreased appetite, vomiting, diarrhoea, fatigue, neutropenia, asthenia and thrombocytopenia. Anaemia was considerably more common in the trifluridine–tipiracil group than the placebo group (45% compared with 19%). Neutropenia was also more common (53% compared with 4%). The company included adverse events such as neutropenia in the model to capture their effect on health-related quality of life. The committee noted that in the summary of product characteristics for trifluridine–tipiracil, neutropenia was one of the most common side effects that led to treatment being stopped, delayed or interrupted. It concluded that neutropenia may affect health-related quality of life.

3.7 The company extrapolated overall survival in both treatment arms using an accelerated failure time model, which included a dependent variable to capture the effect of treatment. This approach assumes that the relative treatment effect is constant over time. In its base-case analysis the company used a log-normal function that was applied for the entire duration of the model. The EAG explained that the Kaplan−Meier estimates from the intention-to-treat population, new analyses at consultation and analyses in the committee's preferred population (the third-line, European subgroup) all showed that trifluridine–tipiracil survival either crossed or almost converged with best supportive care survival. This indicates that the treatment effect was not constant over time. The committee heard that because of this, the EAG preferred separate functions that were fitted independently to each treatment arm. This had little difference in statistical fit compared with the dependent models. The company maintained its preference for the dependent model in its base-case analysis but accepted that other approaches may also be valid. The committee concluded that the model should use survival functions fitted independently to each trial arm to extrapolate overall survival.

3.8 The company used a log-normal function to extrapolate overall survival for the entire duration of the model in its base-case analysis. The clinical experts predicted that 20% to 25% of people survive to 6 months in current practice, which reduces to 10% to 15% at 1 year. The committee noted exploratory analyses that modelled overall survival using the relatively mature Kaplan−Meier estimates for the first 12 or 18 months of the model, then applied a parametric function to extrapolate beyond each timepoint. The EAG advised that using the Kaplan−Meier estimates was problematic because the timepoint when the observed data was replaced by the parametric function was arbitrary. Also, the available parametric functions had been estimated using the full duration of trial data rather than the end portion. The committee noted that the EAG's preferred method for extrapolating the overall survival was a parametric model used for the entire time horizon. It concluded that a full log-normal function was most plausible, and should be considered for decision making.

3.9 The company's revised base-case analysis, using the committee's preferred approach (see section 3.4) used a generalised gamma function to model treatment duration, fitted to the adjusted TAGS third-line European subgroup. The committee noted that the Kaplan−Meier estimates showed that no patients were having trifluridine–tipiracil at 1 year. At the third meeting, the company confirmed that there were no other data on treatment duration for this population with metastatic gastric cancer. It also clarified that in the full TAGS population, the maximum time on trifluridine–tipiracil was around 1.2 years, indicating that very few people would be expected to remain on treatment for a long time. The EAG explained that it preferred to use an extrapolated function, rather than the Kaplan−Meier estimates, to better reflect uncertainty in the data. This is because in clinical practice there may be a small number of people who do stay on treatment for a long time, but this was not reflected in the Kaplan−Meier estimates. The EAG advised that the generalised gamma function was reasonable, but other parametric functions could not be ruled out based on their statistical fit to the data, including some that predicted more people having long-term treatment. The committee agreed that people are unlikely to remain on treatment with trifluridine–tipiracil for very long, and so functions with long tails were not appropriate. It agreed that the generalised gamma function was acceptable for decision making, but noted that using the Kaplan−Meier estimates may also be plausible.

3.10 The company's base-case utility values were 0.764 for the progression-free health state and 0.652 for progressed disease. These values came from TAGS data on EORTC QLQ‑C30. This is a disease-specific measure, mapped onto the generic EQ‑5D‑3L scale using an algorithm from a small Greek study that included people with non-metastatic gastric cancer. The committee was aware that at the clarification stage, the company did not provide cost-effectiveness results using alternative mapping studies from Versteegh et al. (2012) or Longworth et al. (2014), as requested by the EAG. The company clarified that this was because neither study was in gastric cancer and Versteegh et al. (2012) did not use the UK value set. The committee noted that the company's preferred utility values were higher than those used in NICE's technology appraisal guidance on ramucirumab, particularly for progressed disease (0.652 compared with 0.587). The utility values in that appraisal were based on EQ‑5D data from a trial (RAINBOW) and included people with metastatic disease after 1 previous treatment. The company did not consider those utility values appropriate because they did not account for correlation between utility scores for the same patient over time. The committee noted that the preferred utility values in the ramucirumab appraisal included data from multiple timepoints for the progression-free health state but not for progressed disease. The clinical experts advised that, in their opinion, the most appropriate data source would be the population from the TAGS trial who had at least 2 previous treatments, no previous treatment with ramucirumab and had good performance status. The committee concluded that the company's mapped utility values from TAGS were acceptable for decision making.

3.11 At consultation, the company highlighted a Turkish study of 72 patients with gastric cancer and 72 caregivers. This reported improvement in the carers' quality of life after a nursing care intervention. The company noted that the benefit for carers and families from delaying disease progression with trifluridine–tipiracil was not captured in its model. However, the committee concluded that there was no evidence that the quality-of-life gain would be significant and so carer quality-of-life improvement should not be considered in the model.

3.12 For this update, the company provided evidence that metastatic gastric cancer after 2 or more treatments is a severe condition. The severity modifier allows the committee to give more weight to health benefits in the most severe conditions. The company provided absolute and proportional quality-adjusted life year (QALY) shortfall estimates in line with NICE's health technology evaluations manual. Absolute QALY shortfall is the future health that is lost by people living with a condition, including quality and length of life, compared with the expected future health of people without the condition over their remaining lifetime. Proportional QALY shortfall represents the proportion of future health that is lost by people living with the condition, including quality and length of life. The company provided evidence for the committee's preferred population, the third-line European subgroup from the TAGS trial (see section 3.3). This population had a mean age of 62 and was 33.3% female. People with these characteristics without metastatic gastric cancer after 2 or more treatments would be expected to gain 11.69 QALYs. The company's model estimated that people with metastatic gastric cancer after 2 or more treatments who have best supportive care would be expected to gain 0.37 QALYs. The company used these estimates to calculate an absolute QALY shortfall of 11.3 and a proportional QALY shortfall of 0.97. The committee considered the advice about severity as a decision modifier, which allows it to apply a greater weight to the QALYs for technologies indicated for conditions with a high degree of severity. It noted that if either the absolute or the proportional QALY shortfall calculated falls on the cut-off between severity levels, the higher severity level will apply. The company stated that although the absolute QALY shortfall was less than 12, which would imply no QALY weight, the proportional QALY shortfall of 0.97 implies the highest QALY weight of 1.7. The EAG confirmed that the company's analyses had been implemented correctly and supported the use of the 1.7 QALY weight. The committee noted that uncertainty around inputs had not been fully explored. But it recognised that uncertainty had been explored in the original appraisal and that the new analyses were based on its preferred assumptions. So, the committee concluded that the highest severity weight of 1.7 applied to the QALYs was appropriate.

3.13 The company's updated incremental cost-effectiveness ratio (ICER) for trifluridine–tipiracil compared with best supportive care for the third-line European subgroup was £29,347 per QALY gained, including the confidential commercial discount for trifluridine–tipiracil and a 1.7 QALY severity weight (see section 3.12). The committee noted that this analysis was based on its preferred assumptions, including the third-line European subgroup and the generalised gamma function for extrapolating trifluridine–tipiracil treatment duration. The committee concluded that the most plausible cost-effectiveness result was less than £30,000 per QALY gained.

3.14 The committee considered that the acceptable decision-making threshold was £20,000 to £30,000 per QALY gained. The committee concluded that the most plausible ICER based on its preferred assumptions, was less than £30,000 per QALY gained. Therefore, trifluridine–tipiracil is recommended for routine use in the NHS.

3.15 The committee recalled the poor prognosis for people with metastatic gastric cancer and that there is an unmet need for treatment options after 2 or more treatments (see section 3.1). The company considered trifluridine–tipiracil to be innovative because it provides an alternative oral treatment option that increases overall survival. The committee recalled that trifluridine–tipiracil was clinically effective compared with best supportive care (see section 3.3), but it had not seen evidence of additional benefits that were not captured in the model. It concluded that all relevant benefits had been captured in the cost-effectiveness estimates.

3.16 The committee understood that no equalities issues were raised during scoping and technical engagement. It also noted that no potential equality issues were identified in the company submission. The committee concluded that no equalities issues were identified relevant to the recommendation.