4 Evidence and interpretation

The Appraisal Committee (see Appendix A) considered evidence from a number of sources (see Appendix B).

4.1 Clinical effectiveness

4.1.1 No randomised controlled trials (RCTs) were identified that compared imatinib treatment with best supportive care in patients with unresectable and/or metastatic GIST. In the absence of controlled studies, historical controls were identified to provide information on the natural history of advanced GIST. Many patients in these studies had received other treatments (such as chemotherapy or radiotherapy). These studies provide the most appropriate data for comparison with studies of imatinib treatment.

4.1.2 The Assessment Group identified 15 studies of imatinib treatment of c-KIT-positive GIST, which included six ongoing uncontrolled trials. Four of the uncontrolled trials were only available in abstract form at the time of the review, but one was published during the course of the appraisal. The Assessment Group also identified eight single case studies and one case series. In the absence of any controlled trials, the Assessment Group also identified 14 uncontrolled case-series and cohort studies, that is, comparator studies (one of which is currently unpublished).

4.1.3 Key study outcomes included: survival; tumour status (tumour mass measured by computerised tomography [CT] or magnetic resonance imaging [MRI] and classified by response evaluation criteria in solid tumours [RECIST] or SWOG criteria); and Eastern Cooperative Oncology Group (ECOG) performance status (which measures functional status in everyday tasks, and which are also reported in quality-of-life measurement scales). Positron emission tomography (PET) imaging, reported in one study, provided information on the effect of imatinib on tumour metabolism.

Imatinib treatment

4.1.4 Study CSTI571-B2222 is a published ongoing phase II uncontrolled trial of imatinib treatment in 147 patients (91% of whom were c-KIT-positive) with unresectable and metastatic GIST, which formed the basis of the licence application. The manufacturer also provided updated results of this study, which are not yet fully published. Patients were randomly assigned to receive a single dose of 400 mg (n = 73) or 600 mg (n = 74) of imatinib. Patients received imatinib treatment for a median of 21 months (range 7 to 783 days). The study was not powered to distinguish statistically significant differences in the efficacy of imatinib treatment between the two study arms (400 and 600 mg/day). The combined survival rate from the start of treatment was 88% at 1-year follow-up and 78% at 2-year follow-up. Median survival had not been reached after 31 months of follow-up.

4.1.5 Tumour response (based on SWOG criteria, Appendix D) evaluated at 21-month follow-up showed that 66% of patients achieved a partial response, 17% stable disease and 12% progressive disease (with 5% of patients being unevaluable). No patients achieved a complete response. The manufacturer's submission reported resistance to imatinib in 16 patients; 3 patients exhibited primary resistance (no response to imatinib) and a further 13 exhibited secondary resistance to imatinib (loss of response to imatinib).

4.1.6 There were substantial improvements in patients' ECOG performance status. At the 4-month follow-up, the proportion of patients with normal functional status (grade 0) had increased from 42% at baseline to 64%, and the number of patients with impaired functional status of grade 2 (capable of self-care but unable to work) had decreased from 18% at baseline to 5%. These improvements were maintained in the 21% of patients who were followed up to 25 months.

4.1.7 A published ongoing phase I study evaluated imatinib treatment at licensed doses of 400 and 600 mg/day and unlicensed higher doses in 40 patients with advanced GIST (35 of whom were c-KIT-positive). After 9 to 12 months of follow-up, survival from start of treatment was 90%. Tumour response was evaluated using RECIST criteria in the 35 patients who were c-KIT-positive: 51% achieved a partial response; 31% stable disease; and 8.5% progressive disease. Tumour function evaluated in 14 patients using PET imaging after 8 days of treatment showed that eight patients had achieved a complete response, two patients a partial response and one patient no change. The remaining three patients showed disease progression after 28 days.

4.1.8 Of the four trials that were unpublished at the time of the Assessment Group's review, two were non-randomised studies and two were randomised dose-response studies based on large patient samples. An ongoing European study comparing 400 mg/day with (the unlicensed dose of) 800 mg/day in 946 patients reported progression-free survival at 2 years to be approximately 40% (400 mg/day) and 55% (800 mg/day). An interim analysis of this study published in abstract form (median follow-up of 8 months) showed no significant differences between the two treatment arms. A second interim analysis of this trial presented to the Committee (median follow-up of 17 months) showed that progression-free survival was significantly better with initial treatment at 800 mg/day. A study based in the USA, comparing the same daily doses in 746 patients, reported progression-free survival at 6 months to be 80% (400 mg/day) and 82% (800 mg/day).

4.1.9 Study CSTI571-B2222 (registration study) reported that at least one adverse event had been experienced by all 147 patients by 21 months' follow-up. Of these, 37% were 'severe and undesirable' (grade 3) and 15% were 'life threatening and disabling' (grade 4). A total of 15 (10%) patients withdrew from the study because of adverse events; one third of these events were classed as drug-related. The most commonly reported side effects of imatinib include nausea, diarrhoea, periorbital oedema, muscle cramps, fatigue, rash and headache. The most common serious adverse events were unspecified haemorrhage and neutropenia, each event occurring in approximately 5% of patients. Overall, imatinib was well tolerated.

4.1.10 Patient group submissions commented on the dramatic effect of imatinib, with the majority of patients experiencing improvements in disease-related symptoms (such as abdominal distension and pain) and reporting improved appetite and a feeling of well-being. Some patients were able to resume normal daily activities, which included a return to full-time employment.

4.1.11 Expert testimony at the committee meeting provided evidence on the issue of resistance to imatinib. The clinical expert informed the Committee that, although data from the clinical trials are too premature to give definitive answers regarding resistance, they expect that approximately 40% of patients may become resistant to imatinib.

Case-series and cohort studies on the natural history of disease progression

4.1.12 Fourteen primary studies of alternative treatments (surgical resection, chemotherapy, radiotherapy and chemoembolisation) for patients with advanced GIST provided information on the natural history of disease.

4.1.13 Median survival reported in 12 papers (983 patients) ranged from 2 to 39 months, with survival rates of 37–80% at 1-year follow-up, 6–45% at 3-year follow-up and 0–45% at 5-year follow-up. However, 12 out of 14 studies did not have the GIST diagnosis confirmed by c-KIT testing.

4.1.14 Data from an unpublished cohort study were used to represent the natural history of patients with GIST. A total of 143 c-KIT-positive patients (132 patients with recurrent or metastatic GIST) were included in the retrospective analysis. Of these, 91% had previously undergone surgical resection of the tumour, and all patients had received chemotherapy for a median duration of 55 weeks. Patients surviving to the time when imatinib treatment became available were transferred to the treatment (n = 67). Survival outcomes were presented for all patients (irrespective of whether they received imatinib) and for patients who never received imatinib.

4.2 Cost effectiveness

4.2.1 No published cost-effectiveness analyses or quality-of-life studies for patients with advanced GIST were identified in the literature. The manufacturer submitted an economic model, and the Assessment Group re-analysed this model to overcome identified shortcomings. The Assessment Group also developed its own economic model, which was revised after discussion at the committee meeting to answer questions raised about some of the assumptions underpinning all the models.

4.2.2 The manufacturer's model estimated the incremental cost effectiveness of imatinib treatment compared with best supportive care. Patients in the control arm were assumed to start in a state of progressive disease, where they remained until death, on the basis of survival estimates extrapolated from the cohort study (using selected patients who never received imatinib only). All patients in the imatinib treatment arm were assumed to respond to imatinib treatment immediately. The probability of developing progressive disease after initial response was based on the results of the CSTI571-B2222 study. The data from both the trial and the cohort study were extrapolated to 10 years using exponential curves. Estimates of health-related utility were derived by using clinical judgement to map ECOG performance status to a generic measure of health status (the EuroQol EQ-5D). The results of the model showed the incremental cost-effectiveness ratio (ICER) estimates to be £59,000 at 2 years, £24,000 at 5 years and £14,000 at 10 years.

4.2.3 The Assessment Group modified the manufacturer's model to overcome some concerns that the Group considered would overestimate the cost effectiveness of imatinib. The first of two key amendments was to the model structure; it sought to rectify the overestimation of benefit of imatinib by using both the survival and time-to-treatment-failure curves from the registration study to estimate the proportion of patients in the imatinib health state. The second key change was to estimate survival with progressive disease from all patients included in the cohort study, including those who later went on to receive imatinib. The ICER estimates following all the modifications were £41,000 at 2 years and £30,000 at 10 years.

4.2.4 At the instruction of the Appraisal Committee, the Assessment Group, in conjunction with the NICE Decision Support Unit (DSU), was commissioned to develop its own economic model. Additional data from the cohort study (survival estimates, censored at the time imatinib became available) were sourced to improve the estimates of survival with progressive disease. One of the key differences between the DSU model and the other economic models was that the DSU model was structured so that all patients started in the same health state of progressive disease. Also, all the relevant censored data from the cohort study were used to estimate survival following progressive disease − that is, not just patients who died before they could be prescribed imatinib, but also survivors up to the point at which they were transferred to imatinib treatment. Another key difference was that the extrapolations of both the trial data and the censored cohort study data were based on all the data available and did not assume a constant hazard ratio. The estimates of utility were the same as those included in the manufacturer's economic model. The model was also structured to estimate the cost effectiveness of different policies regarding dose escalation.

4.2.5 The results of the DSU model suggest that the incremental cost per additional quality-adjusted life year is approximately £32,000 for patients on 400 mg/day estimated over 10 years. The incremental cost effectiveness of a policy allowing dose escalation to 600 mg/day after failure of 400 mg/day is approximately £39,000 at 10 years compared with a policy of treatment with 400 mg/day and no dose escalation.

4.3 Consideration of the evidence

4.3.1 The Committee reviewed the evidence available on the clinical and cost effectiveness of imatinib for the treatment of GIST, having considered evidence on the nature of the condition and the value placed by users on the benefits of imatinib from people with GIST, those who represent them, and clinical experts. It was also mindful of the need to ensure that its advice took account of the efficient use of NHS resources.

4.3.2 The Committee heard evidence from experts on the current treatment of patients with GIST, and it was aware that imatinib is the only effective treatment for metastatic and/or unresectable GIST. Experts advised that about half of all patients with GIST have unresectable and/or metastatic disease at presentation. Experts also advised that tumour response is most commonly assessed by CT scan (tumour size and density) or MRI scan (tumour size). The Committee also heard that although PET assessment of tumour metabolism at 1 week post-treatment could provide early information on patients' responses to imatinib treatment, the PET technology is not currently routinely available. The Committee therefore considered that assessment of tumour response should be based principally on the SWOG criteria (Appendix D) because this classification of tumour response was reported in the published clinical study and formed the basis of all of the economic models. The Committee understood that estimates of SWOG response criteria are assessed by CT/MRI and that these include an element of assessment of symptoms and the need to note that tumour enlargement may be related to swelling associated with tissue necrosis. It was also persuaded that changes in the density of the tumour deposits as assessed by CT scanning may also indicate response to therapy, and that these changes should be included as part of the overall assessment of the response to imatinib. The Committee also concluded that further research into the use of PET for assessing tumour response would be beneficial.

4.3.3 The experts informed the Committee that patients with multiple lesions may experience tumour growth in some sites while the bulk of the tumour remains under control. The Committee carefully considered the SWOG criteria in relation to these situations. It acknowledged that the criteria would allow for an increase in tumour size of up to 50% or 10 cm (whichever is smaller) in the sum of products of all measurable lesions over the smallest sum observed. The Committee also noted that lesions that appear to increase in size because of the presence of necrotic tissue are not considered to have progressed using the SWOG criteria. For these reasons, the Committee concluded that the use of SWOG criteria to assess response is appropriate.

4.3.4 The experts advised the Committee on the difficulties involved in diagnosing metastatic and/or unresectable GISTs, the assessment of response to treatment and determining the appropriate mix and timing of treatments (surgery or drug therapy) for these patients. The Committee therefore concluded that imatinib therapy should be used only under the supervision of an expert with experience in the treatment of these patients. It is understood that this may also include shared care of the management of these patients between experts in GIST with other cancer centres.

4.3.5 The Committee heard evidence that mutational analysis to determine whether patients have a mutation in KIT may enable clinicians to predict the patients who are most likely to respond to imatinib treatment, because patients with no mutations in KIT have a poorer prognosis. Although these tests are not widely available in the UK, the Committee believed that further research into their use to identify patients for whom imatinib may be particularly appropriate would be important.

4.3.6 The Committee carefully considered the appropriate length of time for which patients should receive imatinib after initiation of treatment before first assessment of response, and at what stage treatment should be considered to have failed. Experts advised that patients are usually evaluated (by an assessment of diagnostic imaging and clinical symptoms) at about 12 weeks from the initiation of treatment, at which stage the disease is classified as complete response, partial response, stable disease or progressive disease. The experts also advised the Committee that a tumour response to imatinib, including stabilisation of disease, is usually seen within 12 weeks of initiation of treatment. The Committee also considered evidence from various sources that the maximum response of GIST to imatinib may not be reached for up to 6 to 12 months from initiation of treatment. The Committee was, however, persuaded that the time taken to achieve an initial response would be significantly less than that taken to achieve a maximum response, as is evidenced from the PET scanning studies, which show that response to imatinib can be achieved within the first 2 weeks of treatment. Additionally, the Committee noted that all of the clinical trials included early review of patients as part of the assessment of response to treatment. The Committee considered, therefore, that patients should initially receive imatinib treatment for up to 12 weeks before tumour response is evaluated, but that only patients responding to treatment (as defined in Section 1.5) by 12 weeks should continue to be treated with imatinib until there is further evidence of disease progression. In making this judgment, the Committee took into account the fact that the definition of tumour response allows for an increase in the size of tumour(s) due to necrosis and swelling, and for stabilisation (that is, no change) of disease within the SWOG criteria.

4.3.7 The Committee considered the evidence on the effectiveness of the higher initial dose of imatinib of 600 mg/day from the CSTI571-B2222 study and the evidence published in abstract form on the daily dose of 800 mg. It concluded from the evidence reviewed that there was no difference in the effectiveness between initial doses of 400 mg/day and 600 mg/day. The Committee considered the early results from the two trials comparing initial doses of 400 mg/day and 800 mg/day. It acknowledged that the early interim results from one of these trials showed a non-significant benefit in progression-free survival from 800 mg/day compared with 400 mg/day, and that the later analysis of this trial showed a significant benefit from initial treatment at 800 mg/day. The Committee also noted that the other trial showed very little difference between the two daily dosages. The Committee concluded that the data from these studies were too premature to draw firm conclusions. The Committee was also aware that the safety and tolerability of a daily dose of imatinib of 800 mg has not yet been assessed by the Medicines and Healthcare products Regulatory Agency. The Committee concluded that it was unable to issue guidance on a dose of 800 mg/day because this dosage has not yet received a UK marketing authorisation, and it was persuaded that the licensed dosage of 400 mg/day was the most appropriate initial dose of imatinib.

4.3.8 The Committee considered evidence from the trials, and new information provided by the manufacturer, regarding dose escalation in patients with progressive disease. The Committee considered that the data on dose escalation were limited because the number of patients involved was small, the length of follow-up for these patients was short, and patients were not allocated to dose escalation by randomisation, possibly leading to bias in the results. The Committee considered all the evidence on dose escalation in relation to the economic models and concluded that dose escalation is not cost effective.

4.3.9 The Committee also considered views of the experts regarding dose escalation in patients with progressive disease following initial treatment. The experts advised that there is a lack of evidence supporting the effectiveness of dose escalation to 600 mg/day. The Committee acknowledged that studies were ongoing to evaluate the effectiveness of an 800 mg dose of imatinib, and it concluded that there is currently no robust evidence to suggest that continued treatment with imatinib (including dose escalation) in patients with progressive disease is cost effective.

4.3.10 The Committee carefully considered the most appropriate estimates of survival for a control group for imatinib treatment in the cost-effectiveness modelling. All the models used data from the same unpublished historical control study to represent the natural history of GIST, but differed in the selection of patients and whether the data were censored for when imatinib treatment became available. The Committee considered this data to be the most appropriate because all patients' tumours were diagnosed as CD117-positive. The Committee concluded that data on the survival of the control group, censored for when imatinib became available, was the least prone to bias and the best estimate of prognosis of untreated GIST.

4.3.11 The Committee considered the different methods of extrapolating the trial and cohort study data to a 10-year time horizon in the cost-effectiveness models. The DSU model, which did not assume constant hazard and used all of the available data for both the treatment and control groups, was considered to be the most appropriate method of extrapolation. The Committee concluded that the cost-effectiveness estimates based on a 10-year time horizon were the most appropriate, because this time horizon was likely to encompass the key costs and benefits. The Committee considered the results from extrapolating the data in relation to new information (at 152 weeks' follow-up) provided by the manufacturer. The Committee acknowledged that the extrapolated time to progression in the DSU model was shorter, and that the extrapolated time to progression in the manufacturer's model was longer, than the time to progression shown by the new data supplied by the manufacturer. However, the Committee also noted that overall survival predicted using the DSU model was longer than suggested by the new data provided by the manufacturer. Thus, the Committee concluded that the ICER produced using the DSU model was likely to be an underestimate when considering this new data.

4.3.12 The Committee considered that the assumption that all patients would respond to imatinib immediately (as presented in the manufacturer's and Assessment Group model) was unlikely. The Committee concluded that the estimates of cost effectiveness based on the DSU model (in which all patients begin in the progressive phase) used the most appropriate available data and assumptions.

4.3.13 The Committee heard evidence from experts regarding likely tumour response rate in the event of dose escalation. The DSU model assumed that patients who are escalated to 600 mg/day after disease progression on 400 mg/day have the same response as when they initially responded to the lower dose. The experts advised the Committee that the time to treatment failure following dose escalation is likely to be substantially shorter than the initial response at 400 mg/day. The Committee therefore considered that this assumption (of an equivalent length of response following dose escalation) used in the DSU economic model was more optimistic than suggested by the experts. The Committee also considered new evidence on response following dose escalation provided by the manufacturer. The Committee considered that the assumption about the proportion of patients responding to dose escalation in the economic model was more optimistic than supported by the evidence provided by the manufacturer. These conclusions supported their view on the inappropriateness of dose escalation in progressive disease, as in Sections 4.3.8 and 4.3.9.

4.3.14 The Committee heard testimony from the representative for patients with GIST and the clinical experts about the dramatic improvement in health-related quality of life associated with successful imatinib treatment. In addition, the Committee considered advice from a clinical expert that, in a minority of patients, imatinib treatment may be given until unresectable tumours shrink to a size at which they can be surgically resected. The expected survival of these patients is better than the expected survival of patients whose tumours remain unresectable. The Committee considered that both of these factors may not have been fully represented in the economic modelling. The Committee concluded that if data were available to model these issues more robustly, the ICERs would be slightly lower. However, it concluded that these factors did not alter its overall view of the cost effectiveness of imatinib therapy for GIST.

4.3.15 Experts commented that current practice sometimes includes the continuation of imatinib treatment in patients with progressive disease (as assessed by CT/MRI), provided that they report a symptomatic benefit without objective evidence of response. However, the Committee considered that, because there is a lack of robust evidence on the effectiveness of extended treatment in these patients, and because a decrease in the overall rate of response to imatinib treatment would increase the ICERs to an unacceptable level, the use of imatinib in this context should be undertaken only as part of clinical studies.

4.3.16 The Committee was aware of the continuing research in this area and the emergence of new data published in abstract form during the consultation process. This new information included evidence relating to updated results from trials, dose escalation, assessment of response and discontinuation of imatinib therapy. The manufacturer also provided updated results for the CSTI571-B2222 study. The Committee considered that this evidence has not yet been fully peer reviewed and that much of the evidence relates to the unlicensed dose of imatinib of 800 mg/day. The Committee considered this evidence and its relation to the results of the economic models when making its recommendations. The Committee concluded that it would be important to re-evaluate the recommendations when new evidence becomes fully available or if substantial changes are made to the marketing authorisation of imatinib.