Imatinib for the adjuvant treatment of gastrointestinal stromal tumours

4 Consideration of the evidence

The Appraisal Committee reviewed the data available on the clinical and cost effectiveness of imatinib as an adjuvant treatment, having considered evidence on the nature of gastrointestinal stromal tumours (GISTs) and the value placed on the benefits of imatinib by people with the condition, those who represent them, and clinical experts. It also took into account the effective use of NHS resources.

4.1 The Committee discussed the clinical treatment pathway for KIT (CD117)‑positive GISTs in England, noting that NICE technology appraisal guidance 196 had not recommended imatinib for the adjuvant treatment of GISTs. It heard from the clinical experts that people at high risk of recurrence had the greatest clinical need for adjuvant treatment after complete resection of KIT (CD117)‑positive GISTs. It heard that these patients were eligible to receive adjuvant imatinib for up to 3 years via the Cancer Drugs Fund and that the remaining people with GISTs would be monitored for signs of recurrence ('watchful waiting', when the person's condition is monitored but they are not given any treatment). The Committee heard from patient experts that people place a high value on the psychological impact of taking a drug after surgery to increase their chances of remaining cancer free, as well as the physical and social advantages that this can bring (such as being able to take part in physical exercise and social activities). The clinical experts explained that later in the treatment pathway, imatinib was standard care at disease recurrence (NICE technology appraisal guidance 86), and that most patients would subsequently receive sunitinib (NICE technology appraisal guidance 179). The Committee heard that regorafenib was now available via the Cancer Drugs Fund at the end of the treatment pathway for certain patients with advanced GISTs who had previously had imatinib and sunitinib.

4.2 The Committee noted the changes in the evidence base since the publication of NICE technology appraisal guidance 196 (that is, the guidance under review).The Committee was aware that the clinical data in the company's submission for the original appraisal had focused on the primary analysis data from ACOSOG Z9001. It recalled that in the original appraisal, the Committee had concluded that these data were too immature to enable conclusions to be drawn about key aspects of imatinib's clinical effectiveness, and observed that the company's current submission included 3 phase III studies: ACOSOG Z9001, SSGXVIII/AIO and EORTC 62024. It further noted that the company's current submission described analyses with median follow‑up of 4‑5 years, and that outcomes included overall survival. The Committee noted the Expert Review Group (ERG)'s opinion that the studies were generally well designed and executed. The Committee concluded that the clinical‑effectiveness evidence was suitable for its decision‑making.

4.3 The Committee discussed how risk of recurrence after complete resection of GIST had been classified in the clinical trials and whether this was generalisable to clinical practice in England. It noted that 2 of the trials had stratified patients according to risk of recurrence using National Institutes of Health or modified National Institutes of Health criteria and that the third trial had enrolled patients at any risk of recurrence. It heard from the clinical experts that the Miettinen 2006 criteria (Miettinen and Lasota 2006^[2]) are the most commonly used tool in clinical practice in England to predict risk of recurrence, and that these had largely superseded the National Institutes of Health criteria. It heard from the clinical experts that patients at high risk of recurrence according to the Miettinen criteria are mostly likely to be considered for adjuvant treatment, based on tumour size, location and mitotic rate. The Committee was aware of concerns raised in response to consultation that, although adopting the Miettinen 2006 criteria to identify patients at high risk is currently appropriate, research is underway that could at some point in the future inform a more accurate way of assessing risk of recurrence. However, the Committee agreed that it had to make recommendations on the basis of currently available information. It noted that technology appraisal review guidance may be reviewed when there is significant new evidence that is likely to change the recommendations, as described in 'Guide to the processes of technology appraisal', and concluded no changes to its recommendations were necessary at present. The Committee was aware that the marketing authorisation for imatinib as an adjuvant treatment for KIT (CD117)‑positive GISTs was for people who were at 'significant' risk of relapse after complete resection, but observed that this group had not been defined by the regulatory agency.

4.4 The Committee considered the company's subgroup analyses according to risk of recurrence, and noted that these were for patients at high risk of recurrence according to the Miettinen criteria (and this subgroup had been included in the company's base case in the economic model). The Committee heard from the clinical experts that this subgroup was appropriate because it consisted of patients who had the greatest clinical need and whose risk had been classified according to criteria that were commonly used in England. The Committee also heard from the company that evidence for patients at only moderate risk of recurrence was limited and that there was no evidence for 3‑year imatinib treatment in this group. The Committee did have some concerns about selection bias because the differences in baseline characteristics between treatment groups were more pronounced in the high‑risk subgroups than in the full patient population. However, it concluded that the evidence relating to the subgroup at high risk of recurrence according to the Miettinen criteria was the most appropriate for its decision‑making.

4.5 The Committee reviewed the clinical‑effectiveness evidence for adjuvant imatinib compared with placebo in people with resected KIT (CD117)‑positive GISTs at high risk of recurrence according to the Miettinen criteria. It observed it was difficult to draw any conclusions about an overall survival benefit using the primary analysis of ACOSOG Z9001 because the data were immature (that is, there had been few deaths). It noted, however, that there was a statistically significant benefit for recurrence‑free survival with 1‑year imatinib in the 5‑year follow‑up analysis (HR 0.608 [95% CI 0.417 to 0.886], p=0.009) and that the treatment effect could have been underestimated because of confounding factors (see section 3.2). The Committee acknowledged that crossing over from placebo to active treatment after the ACOSOG Z9001 primary analysis could confound subsequent analyses of recurrence‑free survival and overall survival, causing imatinib's treatment effect to be underestimated. It was aware that statistical methods could be used to adjust for this crossover effect. It agreed with the company and the ERG that the inverse probability of censoring weights (IPCW) method was an appropriate way to adjust for the crossover effect in the high‑risk group in ACOSOG Z9001. It noted that adjusting the 5‑year analyses using the IPCW method lowered the hazard ratios for recurrence‑free survival and overall survival compared with the unadjusted hazard ratios (that is, the treatment effect of imatinib appeared greater once the crossover effect had been removed; see section 3.13). The Committee noted that after adjusting for crossover, there was still no statistically significant difference in overall survival between the imatinib and placebo high‑risk groups (HR 0.76, 95% CI 0.36 to 1.62). The Committee concluded that the clinical trial evidence showed that 1‑year adjuvant imatinib increased recurrence‑free survival compared with placebo. However, it was unclear if the increase in recurrence‑free survival resulted in longer overall survival, even after the analyses had been adjusted using statistical methods to correct the crossover effect.

4.6 The Committee reviewed the clinical‑effectiveness evidence for 3‑year adjuvant imatinib compared with 1‑year adjuvant imatinib in people with resected KIT (CD117)‑positive GISTs at high risk of recurrence according to the Miettinen criteria. It observed that recurrence‑free survival and overall survival at 5 years was statistically significantly longer in the 3‑year arm compared with the 1‑year arm. However, the Committee noted that there were very small patient numbers towards the end of follow‑up, with fewer than 10 patients per arm from 30 months onwards. It considered that this added uncertainty to the results, and it was unclear if this treatment benefit would persist over time after stopping treatment. The Committee concluded that adjuvant treatment with imatinib for 3 years was more clinically effective than giving it for 1 year during clinical trial follow‑up, but that there was uncertainty whether this benefit would continue over the longer term.

4.7 The Committee discussed the role of genetic mutational analysis in identifying patients who would be more or less likely to benefit from adjuvant imatinib treatment. It heard from the clinical experts that it would be useful to perform mutational analysis before starting treatment because some GISTs were likely to be resistant to treatment with imatinib and sunitinib, such as those with a PDGFRA exon 18 D842V mutation. However, the Committee also heard that this patient population with the PDGFRA exon 18 D842V mutation is relatively small and that data on statistically significant differences in treatment effect for the high‑risk population were not currently available. The Committee therefore concluded that it was currently unable to specify any patient subgroups with differential treatment benefits.

4.8 The Committee considered the adverse events associated with adjuvant imatinib. It heard from the clinical and patient experts that adverse events associated with imatinib treatment were predictable and manageable. The Committee concluded that adjuvant imatinib had an acceptable safety profile.

4.9 The Committee considered the company's approach to modelling the cost effectiveness of adjuvant imatinib. It noted that the Markov model structure was similar to that used in the original appraisal of adjuvant imatinib (NICE technology appraisal guidance 196). It was aware that progression through the model was based on treatment, rather than disease progression, and noted that this was a potential source of bias. However, it heard from the ERG that the modelled health states were, in this instance, a reasonable approximation. The Committee concluded that the structure of the company's economic model was acceptable for assessing the cost effectiveness of adjuvant imatinib.

4.10 The Committee discussed the duration of adjuvant imatinib therapy used by the company in the model. It noted that the summary of product characteristics for imatinib states that, for this indication, 'optimal treatment duration is not yet established', but that imatinib had been given for a maximum of 3 years in the trials supporting the marketing authorisation. Consequently, the Committee agreed that it was appropriate to consider adjuvant treatment with imatinib for up to 3 years. It concluded that the company's approach of using 1‑year or 3‑year adjuvant imatinib in the economic model was acceptable as this reflected the clinical trial evidence available for people at high risk of recurrence after resection of KIT (CG117)‑positive GISTs.

4.11 The Committee discussed how the company had estimated the baseline risk of recurrence (that is, with no adjuvant treatment). It noted that in the base case, the company had used data from the primary analysis of ACOSOG Z9001, rather than the 5‑year follow‑up analysis, to estimate the baseline risk of recurrence in people at high risk according to the Miettinen criteria. The Committee agreed with the ERG that it might have been more appropriate to use the hazard ratio for recurrence‑free survival using 5‑year follow‑up data that had been adjusted for crossover using the IPCW method. However, the Committee was reassured by the company's incremental cost‑effectiveness ratios (ICERs) using the 5‑year follow‑up data for recurrence‑free survival for the placebo arm of ACOSOG Z9001 (unadjusted for crossover), which were similar to the base‑case ICERs (see section 3.34). The Committee accepted the company's estimate of baseline risk of recurrence.

4.12 The Committee discussed how the company had incorporated the relative treatment effect into its economic model and accepted the company's assumption that imatinib's treatment effect was different during treatment compared with after treatment. The Committee initially focused on the on‑treatment hazard ratio for recurrence‑free survival for imatinib (1 and 3 years) compared with no adjuvant treatment. It noted the ERG's concern that the hazard ratio used in the company's model (0.111) was lower than the hazard ratio for the primary analysis of ACOSOG Z9001 (0.265). The company explained that this discrepancy was because the data used in the model had been truncated at 12 months (in line with the duration of imatinib treatment in the trial), and that additional patients, who had been identified as being at high risk of recurrence at the 5‑year follow‑up analysis, had been included. The Committee also heard from the clinical experts that a hazard ratio of 0.111 was plausible. The Committee noted that the company's deterministic sensitivity analyses showed that reducing the modelled clinical effectiveness of imatinib during treatment by using the upper confidence interval of the on‑treatment hazard ratio (0.281) did not increase the ICERs above the range that is typically considered to be cost effective (£20,000–30,000 per QALY gained). The Committee concluded that the on‑treatment hazard ratio of 0.111 in the company's economic model was associated with an acceptable level of uncertainty.

4.13 The Committee then looked at how the company had incorporated relative treatment effect after adjuvant imatinib treatment had finished (the off‑treatment period). The Committee was concerned that there was a continuing differential off‑treatment effect whereas it would intuitively be expected to taper off. It questioned whether the hazard ratios of 0.519 for 1‑year adjuvant imatinib compared with placebo and 0.344 for 3‑year imatinib compared with 1‑year imatinib were too low, and discussed the sensitivity of the ICERs to the off‑treatment hazard ratio. The Committee noted that the company's sensitivity analyses and the ERG's exploratory analyses indicated that the ICERs were sensitive to changes in the off‑treatment hazard ratio, but noted that the ICERs generally remained within the range that is typically considered to be cost effective. It noted that the ERG's exploratory analyses showed that assuming no long‑term treatment benefit (that is, no benefit after the end of trial follow‑up) did not markedly increase the ICERs. The Committee concluded that the off‑treatment hazard ratios used in the company's model were sufficiently robust for generating cost‑effectiveness estimates.

4.14 The Committee discussed how the company had extrapolated the clinical trial data to predict longer‑term outcomes using a parametric survival model. The Committee understood the company's rationale for rejecting the Weibull, exponential and gamma distributions. The Committee noted the ERG's exploratory analysis using the exponential model, and considered that this was likely to underestimate imatinib's long‑term treatment benefit and therefore produce ICERs that were higher than the true value. The Committee was concerned, however, that the company's Gompertz distribution could overestimate imatinib's long‑term benefit. It noted that no justification had been given by the company for rejecting the log‑logistic model, which was less optimistic for imatinib than the Gompertz distribution. The Committee concluded that there was some uncertainty in using the Gompertz model for the long‑term extrapolation of imatinib's treatment benefit, and that this could cause the cost‑effectiveness estimates generated using the company's model to be too optimistic.

4.15 The Committee reviewed the utility values used in the company's model. It acknowledged the ERG's concern about the lack of detail describing these in the company's submission but heard from the clinical experts that most of them seemed to be plausible. However, the Committee heard from the clinical experts that sunitinib had a poorer adverse‑event profile than imatinib and it was not appropriate to assume the same treatment disutility for the 2 treatments. The Committee understood that if the utility values for people taking imatinib had been underestimated, then changing this assumption would favour imatinib. However, the Committee noted that this would have only a limited impact on the ICER and concluded that the utility values in the company's model were generally acceptable.

4.16 The Committee considered the risks for mortality that had been used in the company's model. It heard from the company that mortality rates for the post‑recurrence health states were taken from imatinib and sunitinib clinical trials (see section 3.28). However, because of a lack of data for recurrence after adjuvant treatment, patients in the model were assumed to have the same mortality risk following recurrence, regardless of whether they had received adjuvant imatinib treatment or not. Together with the issues related to the extrapolation of recurrence‑free survival discussed in section 4.14, the Committee was particularly concerned about the large life‑year gains predicted by the model at 3 years, but a clinical expert emphasised that this was consistent with the 5‑year survival in the trial. The Committee noted that the ERG's exploratory analysis, which used lower mortality rates after recurrence to better fit the trial results, caused the ICERs to decrease slightly (see section 3.48). The Committee concluded that the mortality rates used by the company in its model were appropriate for the pre‑recurrence health states, but that it preferred the ERG's values for the post‑recurrence health states.

4.17 The Committee had some concerns about the validity of the total and incremental costs for the cost‑effectiveness estimates generated using the company's model. It recalled that the estimated drug cost was £20,700 and £62,100 for 1 year and 3 years of adjuvant imatinib treatment respectively, producing incremental drug costs of £41,400. However, it noted that the company's base‑case results reported incremental costs for 1‑year adjuvant imatinib compared with no adjuvant treatment of less than £8000. Similarly, the incremental costs for 3‑year adjuvant imatinib compared with 1‑year adjuvant imatinib were around £23,000. The ERG explained that, regardless of any previous adjuvant treatment, health state costs after recurrence were high and that the greater proportion of people who remained recurrence free in the active treatment groups accounted for this apparent anomaly. The Committee concluded that it was satisfied with the way the company had incorporated total costs, including drug acquisition costs, into its economic model.

4.18 The Committee deliberated over the most plausible ICERs presented by the company and the ERG for adjuvant imatinib compared with no adjuvant treatment. The Committee considered that the company's base‑case results (including a minor correction by the ERG) could underestimate the true value of the ICER because they used the Gompertz distribution for extrapolating recurrence‑free survival, which possibly overestimated imatinib's long‑term treatment effect (£3610 per QALY gained for 1‑year adjuvant imatinib compared with no adjuvant treatment; £16,700 per QALY gained for 3‑year adjuvant imatinib compared with 1‑year adjuvant imatinib). However, the Committee considered that the ICERs from the ERG's combined exploratory analyses were too high because they used the exponential distribution to extrapolate long‑term recurrence‑free survival, which gave results that were not clinically plausible because recurrence‑free survival rates were too low for no adjuvant treatment (£12,100 per QALY gained for 1‑year adjuvant imatinib compared with no adjuvant treatment; £30,000 per QALY gained for 3‑year adjuvant imatinib compared with 1‑year adjuvant imatinib). The Committee concluded that the true value of the ICERs was between £3610 and £12,100 per QALY gained for 1‑year adjuvant imatinib compared with no adjuvant treatment, and between £16,700 and £30,000 per QALY gained for 3‑year adjuvant imatinib compared with 1‑year adjuvant imatinib.

4.19 The Committee discussed whether imatinib could be considered an innovative treatment. Although it believed that the introduction of adjuvant treatment for GISTs could potentially be considered a step change in health‑related benefits, it noted that imatinib has been available as a treatment elsewhere in the GIST treatment pathway for many years and consequently the move to adjuvant treatment could not in itself be considered innovative. The Committee concluded that there were no additional health benefits that had not been included in the company's economic model.

4.20 The Committee discussed whether using imatinib as an adjuvant treatment for GISTs represented cost‑effective use of NHS resources. It had noted concerns around the plausibility of the extent of life years gained and the small increase in incremental costs predicted by the company's model for adjuvant imatinib treatment (see section 4.17). However, it was satisfied that, after a step‑by‑step examination, the assumptions and approaches used in the model were defensible. Moreover, it noted that the ICERs were insensitive to changes in many of the parameters in the company's sensitivity analyses and the ERG's exploratory analyses, and noted that they generally remained below £30,000 per QALY gained. The Committee also acknowledged the similarity between the deterministic ICER and the probabilistic ICER. Taking all of these factors into account, the Committee considered that the fully incremental ICER for the comparison of 3‑year adjuvant imatinib with 1‑year adjuvant imatinib, which was between £16,663 and £29,966 per QALY gained, was within the range normally considered to represent cost‑effective use of NHS resources (£20,000–30,000 per QALY gained) and was associated with an acceptable level of uncertainty. Therefore, the Committee recommended adjuvant treatment with imatinib for up to 3 years as an option for KIT (CD117)‑positive GISTs in people considered at high risk of recurrence as defined by the Miettinen 2006 criteria (based on tumour size, location and mitotic rate).