Sunitinib for the treatment of gastrointestinal stromal tumours: a critique of the submission from Pfizer.

The submission's evidence for the clinical effectiveness and cost-effectiveness of sunitinib for the treatment of gastrointestinal stromal tumours (GISTs) is based on a randomised controlled trial (RCT) comparing sunitinib with placebo for people with unresectable and/or metastatic GIST after failure of imatinib and with Eastern Cooperative Oncology Group (ECOG) progression status 0-1, and an ongoing, non-comparative cohort study of a similar population but with ECOG progression status 0-4. The searches are appropriate and include all relevant studies and the RCT is of high quality. In the RCT sunitinib arm overall survival was 73 median weeks [95% confidence interval (CI) 61 to 83] versus 75 median weeks (95% CI 68 to 84) for the cohort study. However, time to tumour progression in the cohort study was different from that in the RCT sunitinib arm [41 (95% CI 36 to 47) versus 29 (95% CI 22 to 41) median weeks respectively]. Median progression-free survival with sunitinib was 24.6 weeks (95% CI 12.1 to 28.4) versus 6.4 weeks (95% CI 4.4 to 10.0) on placebo (hazard ratio 0.333, 95% CI 0.238 to 0.467, p < 0.001). The manufacturer used a three-state Markov model to model the cost-effectiveness of sunitinib compared with best supportive care for GIST patients; the modelling approach and sources and justification of estimates are reasonable. The base-case incremental cost-effectiveness ratio (ICER) was 27,365 pounds per quality-adjusted life-year (QALY) with the first cycle of sunitinib treatment not costed; when we included the cost of the first treatment cycle we estimated a base-case ICER of 32,636 pounds per QALY. Pfizer's sensitivity analysis produced a range of ICERs from 15,536 pounds per QALY to 59,002 pounds per QALY. Weaknesses of the manufacturer's submission include that the evidence is based on only one published RCT; that 84% of the RCT control population crossed over to the intervention group, giving rise to the use of unusual rank preserved structural failure time (RPSFT) analysis to correct for possible bias; and that a number of errors and omissions were made in the probabilistic sensitivity analysis, meaning that it is not possible to come to firm conclusions about the cost-effectiveness of sunitinib for GIST in this patient population. In conclusion, during the blinded phase of the RCT, overall survival was significantly longer in the sunitinib arm than in the placebo arm (hazard ratio 0.491, 95% CI 0.290 to 0.831, p <0.007). However, intention-to-treat analysis of the entire study showed no statistically significant difference in overall survival for those who received sunitinib (73 weeks) versus those who received placebo (65 weeks) (hazard ratio 0.876, 95% CI 0.679 to 1.129, p = 0.306).