Trifluridine–tipiracil for previously treated metastatic colorectal cancer

Generalisability of trial results

4.5 The committee was aware that both trials provided evidence for people who had previously had 2 or more therapies, but not for those in whom available therapies were not suitable. The committee recalled that only a few people would not be considered fit enough to have the therapies recommended by NICE first and second line (see section 4.3). Furthermore, it heard from the clinical experts that trifluridine–tipiracil was unlikely to replace those therapies. The committee agreed that the evidence could be generalised to people in whom available therapies are not suitable.

4.6 The committee was aware that in RECOURSE, but not in Yoshino et al. (2012), patients must have had bevacizumab before enrolling in the trial; patients with KRAS wild-type tumours also must have had cetuximab or panitumumab. The committee noted that none of these therapies were recommended by NICE, or funded by the Cancer Drugs Fund, for previously treated metastatic colorectal cancer. Therefore, people in England are unlikely to have had bevacizumab, cetuximab or panitumumab before trifluridine–tipiracil. The company considered that there was no biological reason for the effect of trifluridine–tipiracil to differ in people who do or do not have biological therapies. The committee noted that 22% of patients in Yoshino et al. did not have bevacizumab. In this group, the hazard ratio for overall survival was 0.37 (95% confidence interval [CI] 0.16 to 0.86) compared with 0.63 (95% CI 0.42 to 0.95) in those who did have bevacizumab; there was no statistically significant interaction between the 2 groups. However, the committee considered these results were inconclusive because it was unclear whether previous treatments had differed between the 2 groups in a way that affected survival outcomes. The committee heard from the clinical experts that effectiveness would be expected to diminish with each additional line of treatment. In addition, the clinical experts questioned whether bevacizumab increases survival; therefore, the long-term benefit of previous bevacizumab treatment was unclear. The committee concluded that trifluridine–tipiracil would be similarly effective in people who have or have not had bevacizumab, and that the results of the trials were generalisable to NHS patients in England who have not had bevacizumab.

Meta-analysis of trials

4.7 The committee considered the company's meta-analysis of Yoshino et al. (2012) and RECOURSE, noting that the company 'naively' pooled the efficacy data from the 2 trials. The committee heard from the ERG that the trials had similar designs, and that patients in both trials had similar disease characteristics at baseline. Also, the pooled results mirrored the individual trials. During the committee meeting, the company confirmed that the results were almost the same whether the meta-analysis was or was not stratified (adjusted) by trial. The committee noted the ERG's comments in the meeting that the naive pooling of data was unlikely to be biased. Given the similarity between the 2 trials and the consistency between the results, the committee concluded that the company's original meta-analysis provided valid results.

Results

4.8 The committee discussed the results of the trials, noting that the survival data were 'mature'; that is most patients had died by the end of the follow-up period (Yoshino et al.; 72.8%; RECOURSE: 89.0%). Compared with placebo, trifluridine–tipiracil increased median overall survival by 2.4 months in Yoshino et al. (2012) and by 2.0 months in RECOURSE. The company's meta-analysis of Yoshino et al. and RECOURSE showed that trifluridine–tipiracil led to a statistically significant increase in overall survival (hazard ratio [HR] 0.67, 95% CI 0.58 to 0.78) and progression-free survival (HR 0.46, 95% CI 0.40 to 0.53). The committee heard from the clinical and patient experts that, for an end-of-line treatment, even small extensions to life are important. Also, people at this stage of the disease would be unlikely to tolerate treatments with severe adverse effects. According to the experts, trifluridine–tipiracil offers the advantage of being an oral treatment with minimal adverse effects. The committee concluded that the survival benefit of trifluridine–tipiracil, although relatively small, was clinically meaningful.

Adverse events

4.9 The committee noted that, in clinical trials, trifluridine–tipiracil was associated with a higher incidence of adverse events than placebo (Yoshino et al.; 96.5% compared with 70.2%; RECOURSE: 85.7% compared with 54.7%). It heard that patients considered the safety profile of trifluridine–tipiracil to be acceptable. The clinical experts also stated that trifluridine–tipiracil was well tolerated by people who took part in the named patient programme, and caused less fatigue than regorafenib (whose NICE technology appraisal was terminated), which may also be used at the same stage in the treatment pathway as trifluridine–tipiracil. The committee concluded that trifluridine–tipiracil had an acceptable burden of adverse events.

Survival modelling

4.10 To model overall survival and progression-free survival, the company fitted independent models for each treatment group (trifluridine–tipiracil or placebo); in this method, the treatment effects do not need to meet the proportional hazards assumption. The ERG, however, preferred a single model with a covariate (predictor) for the treatment group because it noted that the log-cumulative hazard plots for overall survival and for progression-free survival indicated that the proportional hazards assumption would hold. The committee was aware that the pooled data for overall survival and progression-free survival were very mature (86% and 89% respectively). This meant that the choice of the model was unlikely to influence the results greatly; when the ERG chose a different model, this had a negligible impact on the incremental cost-effectiveness ratio (ICER). Moreover, the committee did not consider the log-cumulative hazard plots for progression-free survival to strongly indicate that the proportional hazards assumption would hold. Because of this, the committee concluded that modelling each treatment group independently was reasonable and more appropriate than including the treatment group as a predictor in the model.

Body surface area

4.11 The committee was aware that the dose of trifluridine–tipiracil, and hence its cost, is based on body surface area. To model the dose that people would receive, the company estimated the distribution of body surface area from patients in RECOURSE. It grouped patients based on their body surface area, then fitted the log-normal distribution to these data. The ERG considered that using the actual (observed) data on body surface area from RECOURSE was more reasonable. The committee noted that the average body surface area in RECOURSE was 1.78 m², which was lower than the average body surface area among the UK general population. The committee did not consider that a value from the general population was a relevant comparison because people with advanced colorectal cancer would likely be smaller than the general population, as confirmed by the clinical experts. The committee recognised that the modelling of body surface area should minimally influence the cost of treatment because trifluridine–tipiracil is taken orally and patients are grouped into dosing groups. This means that patients with slightly different body surface areas may be in the same dosing group and have the same dose. The committee did not consider there was a need to use parametric methods (notably, a reason to extrapolate) to estimate body surface area, preferring to use the data from the trial for decision-making.

Health-related quality of life

4.12 The committee noted that neither the clinical trials (Yoshino et al. [2012] and RECOURSE) nor the named patient programme provided data on health-related quality of life. To estimate health-related quality of life in the model, the company averaged utility values from the CORRECT trial, which evaluated regorafenib for previously treated metastatic colorectal cancer, and the evidence submission for NICE's technology appraisal guidance on cetuximab for the first-line treatment of metastatic colorectal cancer. The committee understood that the company considered these 2 sources reflected utility values at both ends of the range. The ERG was concerned about using the submission for NICE's technology appraisal guidance on cetuximab to source utility values because the pre-progression utility value from that submission was derived using the Health Utilities Index Mark 3 instrument, which was not in line with the NICE reference case, and it was not ultimately used in the model for that appraisal. Furthermore, the post-progression utility value was derived from people with KRAS wild-type metastatic colorectal cancer that was refractory to chemotherapy, and the ERG could not verify its original source. Because of this, the ERG preferred using CORRECT only to source utility values. The committee heard from the clinical experts that the population in CORRECT reasonably reflected people who would have trifluridine–tipiracil. The committee concluded that, to apply utility values, using the average value from 2 sources is methodologically worse than using data collected from clinical trials, or than using CORRECT alone.

Cost-effectiveness results

4.13 The committee considered the cost-effectiveness results including the patient access scheme discount. It noted that the deterministic ICER from the company's base case revised in response to a request for clarification from the ERG was £42,674 per quality-adjusted life year (QALY) gained for trifluridine–tipiracil compared with best supportive care. The ERG estimated a probabilistic ICER of £52,695 per QALY gained based on clinical data from RECOURSE only, but also presented an ICER based on the pooled dataset, which was £49,392 per QALY gained. The committee was aware that the ERG's ICER of £49,392 per QALY gained used the utility values from CORRECT (see section 4.12), and the data on body surface area seen in RECOURSE (see section 4.11), both of which it preferred. This ICER, however, did not reflect the use of independent models for each treatment group to model survival, but the committee recalled that this modification individually had a negligible impact on the ICER (see section 4.10). Therefore, the committee concluded that the ERG's ICER of £49,392 per QALY gained most closely mirrored its preferred analysis.