NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Hislop J, Quayyum Z, Elders A, et al. Clinical Effectiveness and Cost-Effectiveness of Imatinib Dose Escalation for the Treatment of Unresectable and/or Metastatic Gastrointestinal Stromal Tumours that have Progressed on Treatment at a Dose of 400 Mg/Day: A Systematic Review and Economic Evaluation. Southampton (UK): NIHR Evaluation, Trials and Studies Coordinating Centre (UK); 2011 Jun. (Health Technology Assessment, No. 15.25.)

Cover of Clinical Effectiveness and Cost-Effectiveness of Imatinib Dose Escalation for the Treatment of Unresectable and/or Metastatic Gastrointestinal Stromal Tumours that have Progressed on Treatment at a Dose of 400 Mg/Day: A Systematic Review and Economic Evaluation

Clinical Effectiveness and Cost-Effectiveness of Imatinib Dose Escalation for the Treatment of Unresectable and/or Metastatic Gastrointestinal Stromal Tumours that have Progressed on Treatment at a Dose of 400 Mg/Day: A Systematic Review and Economic Evaluation.

Show details


Statement of principal findings

Review of clinical effectiveness

This review is a part update of a previous review on imatinib for the treatment of patients with unresectable and/or metastatic GISTs.55 We focused on patients with KIT (CD117)-positive, unresectable and/or metastatic GISTs whose disease had progressed on treatment with imatinib at a dose of 400 mg/day. Five studies involving 669 patients from within the relevant treatment arms met the inclusion criteria. Of these studies, four involving 318 patients reported imatinib outcomes and one involving 351 patients, who had received a prior imatinib dose of ≤ 400 mg/day, reported sunitinib outcomes. No studies reporting BSC were identified that met our inclusion criteria.

Although the study designs for most of the included trials were RCTs (plus one retrospective cohort study) none of these trials had, as their primary objective, the assessment of the effects of dose escalation following progression on 400 mg/day imatinib. Only a proportion of the overall patient populations received an escalated dose, and these patients were not randomised at the point of dose escalation to receive either an escalated dose of imatinib or remain on 400 mg/day. Therefore, the nature of the evidence base for patients who progress on 400 mg/day imatinib and receive escalated doses of 600 or 800 mg/day is observational and open to bias.

The sample sizes of the studies from which the 669 patients were drawn from ranged from 2479 to 111786 participants. Each study had more male than female participants. The vast majority of participants in each study had an ECOG performance status of ≤ 2, meaning that they were ambulatory and confined to bed for less than 50% of their waking hours.116 Of the studies that reported the proportion of the study population receiving prior surgery,39,44,79 most patients had undergone prior surgery for treatment of their disease. Information on the characteristics of all the 669 patients relevant to this review was not provided separately.

From the data on imatinib it can be seen that approximately one-third of patients progressing on 400 mg/day of imatinib will respond to escalated doses. With 600 mg/day, between 25.6% (11/43)39 and 41.7% (5/12)79 of patients with unresectable and/or metastatic GIST, who had previously progressed on a dose of 400 mg/day of imatinib, either developed a PR or maintained SD. With 800 mg/day, the proportions achieving PR or SD ranged between 29.3%44 and 33.3%.79 These data were used to inform transition probabilities of non-response to imatinib at escalated doses of 600 and 800 mg/day, respectively. However, response data were not available for patients receiving sunitinib following treatment with imatinib at a dose of ≤ 400 mg/day. As an alternative to excluding sunitinib entirely, which could arguably have been appropriate given the lack of data, the economic model used data that did not meet the inclusion criteria for the review of clinical effectiveness because it failed to report response data separately for those progressing on a 400 mg/day dose. A further assumption made in the economic model was that response was unaffected by prior treatment received. This assumption was made because of a lack of data on how response might change over time and be affected by prior treatments other than imatinib at 400 mg/day.

Median OS data were not reported for those receiving an escalated imatinib dose of 600 mg/day upon progression at a 400 mg/day dose. Therefore, the economic model calculated the probability of death from the available trial data on median OS according to best response, and the proportion of patients receiving escalated doses who will have had a response to imatinib at the initial 400 mg/day dose prior to eventual progression and dose escalation.

For those receiving an escalated imatinib dose of 800 mg/day upon progression, median OS was reported to be 19 months (95% CI 13 to 23 months) in the S0033 trial.41 Median OS was not reported for the EORTC-ISG-AGITG study44 for the population of interest, (CiC information has been removed). For those receiving sunitinib after a prior imatinib dose of ≤ 400 mg/day, median OS was reported as 22.5 months (95% CI 18.3 to 26.5 months).86

Figure 3 provided a visual comparison of the median OS times for imatinib at an escalated dose of 800 mg/day and sunitinib, showing overlapping CIs until 33 months from commencement of treatment, at which point the estimated proportion of sunitinib patients surviving appeared to be less than the proportion surviving on the 800 mg/day imatinib dose. (CiC information has been removed.) It is difficult to draw any conclusions with regard to possible differences in OS between imatinib at an escalated dose of 800 mg/day and sunitinib at 50 mg/day (with a 4-weeks-on/2-weeks-off cycle), owing to the lack of data, but as the 95% CIs for median OS overlap, there does not appear to be any significant difference in median OS with dose escalation, compared with sunitinib.

The median time to progression and PFS was reported for imatinib 600 mg/day as 1.7 months (range 0.7–24.9 months),79 and for imatinib 800 mg/day it ranged between 2.9 months (reported without CIs as ‘81 days’)44 and 5 months (95% CI 2 to 10 months).41 A visual representation of these data for imatinib 800 mg/day in Figure 4 gives 95% CIs that do not overlap, for all time points between 12 and 21 months, indicating that PFS was significantly shorter in the EORTC-ISG-AGITG study reported by Zalcberg et al.44 than in the S0033 trial reported by Blanke et al.41

In addition, those studies looking at an 800 mg/day dose of imatinib reported that between 16.1% (19/118) and 18.8% (25/133) of patients were progression free at the time of the analysis. This represented a proportion of between 52.8% (19/36) and 64.1% (25/39) of all of those achieving PR and SD on the 800 mg/day dose. This suggests that a small proportion (i.e. < 20%) of those receiving an escalated dose of 800 mg/day of imatinib on progression may maintain their response/SD for a median time period of at least 25 months (i.e. the shorter of the median follow-up times reported by these trials), and those who achieve a response or maintain SD on the escalated dose may have a greater than 50% likelihood of maintaining this in the longer term.

For those receiving an escalated dose of 800 mg/day, the study by Zalcberg et al.44 reported a median duration of ‘stabilisation’ among those showing response or SD with treatment of 153 days (range 37–574 days). For sunitinib, the treatment duration for all patients receiving sunitinib (i.e. regardless of the dose of prior imatinib therapy) was 126 days (range 1–618 days).86

Data on adverse events were not available from any of the studies where the population of interest received imatinib at 600 mg/day or sunitinib following progression at 400 mg/day. For the trials reporting outcomes following dose escalation from 400 to 800 mg/day after progression at the lower dose, it was reported that the vast majority (88.4%) of study discontinuations were due to disease progression and not study drug toxicity.44 (CiC information has been removed.)

Nevertheless, it was also reported that between 15.6%77 and 31%78 of patients receiving an escalated imatinib dose of 800 mg/day required a dose reduction. It was also reported that 23.3% (18/77) of patients required at least one dose delay.77 However, it was not possible to take possible dose reductions into account with regard to any of the outcomes. This was because information on the dose provided following reduction, the median duration of any dose delay or dose reduction, and any other factors, besides toxicity, contributing to any of the dose delays or reductions were not reported.

These data on discontinuations and dose modifications indicate that, although disease progression is far more likely than adverse events to contribute to the decision to stop escalated imatinib treatment at the 800 mg/day dose, approximately one-third of patients will require dose modifications (i.e. dose reduction or interruption) during treatment at this escalated dose.

With regard to specific adverse events, data were reported by Zalcberg et al.44 showing that a higher proportion of patients with skin rash, nausea, leucopenia, neutropenia and thrombocytopenia reported a reduction in the severity of these events following dose escalation compared with the proportion of patients reporting an increase in these events. This reduction was significant in the case of neutropenia (p = 0.002). However, the proportion of patients with oedema, fatigue, dyspnoea and anaemia who reported an increase in severity of these events following dose escalation was greater than the proportion of patients who reported a reduction in these events. This increase in severity was significant in the case of fatigue (p < 0.001) and anaemia (p = 0.015).44 (CiC information has been removed.) It is difficult to draw any conclusions about specific adverse events from these data, aside from noting that fatigue and anaemia may significantly increase upon dose escalation from 400 mg/day imatinib to 800 mg/day.

The only data available for any of the prespecified subgroups of interest were reported by Debiec-Rychter et al.14 for the EORTC-ISG-AGITG trial, which looked at imatinib dose escalation from 400 to 800 mg/day following progression at the lower dose. They noted that patients with wild type, and those with exon 9 mutations, were significantly more likely to have a response to dose escalation than those with exon 11 mutations, but no numerical data were reported for the population of interest. (CiC information has been removed.) Furthermore, it has been argued that subgroups with certain exon mutations might have improved response and/or survival outcomes if they initially receive an escalated imatinib dose, rather than receiving dose escalation only if there is progression at the 400 mg/day dose.114

It was outwith the remit of this review to consider outcomes for patients receiving escalated dosing other than following progression on the initial 400 mg/day dose. The lack of data available meant it was not possible to assess for specific mutational population subgroups the effects of escalation to an imatinib dose of 800 mg/day following progression at the initial 400 mg/day dose.

Review of cost-effectiveness

The economic component of this study included both a review of the existing economic evaluations and an economic modelling exercise. The evidence from the review of economic evaluations was sparse and there was no published economic evaluation conducted for a UK context that compared all of the interventions for the patient group of interest.

The modelling exercise compared alternative treatment pathways for patients with unresectable GIST who failed to respond to imatinib 400 mg/day. Over almost all the sensitivity analyses, Path-1, BSC, is the least costly and least effective intervention. Similarly, Path-4, imatinib 600 mg/day, typically has an incremental cost per QALY that is less than £30,000 compared with Path-1, BSC. Path-2 (imatinib 600 mg/day to 800 mg/day to sunitinib) is the only other pathway that is not dominated or extendedly dominated over most of the analyses conducted. However, in this case the incremental cost per QALY (compared with the next most costly option: Path-4, imatinib 600 mg/day) tends to be > £40,000.

When society's willingness to pay for a QALY is < ~£25,000, Path-1, BSC, is the most cost-effective intervention. When society's willingness to pay for a QALY is between approximately £25,000 and £45,000, Path-4, imatinib 600 mg/day, is most likely to be considered cost-effective. Beyond a threshold of approximately £45,000, Path-2, imatinib 600 mg/day to imatinib 800 mg/day to sunitinib, is most likely to be cost-effective.

As discussed below, these data should be treated cautiously, as the data used are observational and non-comparative. Furthermore, the data on sunitinib and imatinib 600 mg/day are particularly sparse and potentially unreliable. For example, data on treatment with sunitinib show a lower life expectancy than those on treatment with BSC (although sunitinib has greater QALYs). This means that when the cost of sunitinib is reduced it becomes more cost-effective than BSC, as the potentially unreliable source data for life expectancy on sunitinib mean that patients on sunitinib will not survive long enough to incur higher costs of treatment. Although sufficient evidence on the effectiveness of sunitinib compared with BSC following treatment on imatinib at a 400 mg/day dose was not available, evidence on the effectiveness of sunitinib compared with BSC regardless of prior imatinib dose suggests that life expectancy with sunitinib is superior.

In addition, the data available for imatinib at a dose of 600 mg/day suggested superior effectiveness compared with the 800 mg/day dose. This is because the evidence on imatinib at the 600 mg/day dose was based on a smaller sample size (43 patients), making the model results for this pathway potentially counterintuitive if we expect higher drug doses to have greater effectiveness than lower doses.

Strengths and limitations of the assessment

In terms of strengths, the review of the evidence base was detailed and thorough. It was unclear from the information provided in a substantial number of abstracts whether the studies met the inclusion criteria and full-text papers for all of these reports were obtained and assessed. Non-English language studies were not excluded. Authors were contacted in an attempt to obtain additional information concerning their studies. For the review of economic evaluations, a rigorous systematic approach was adopted. The economic model considered a large number of plausible alternative treatments and also incorporated both probabilistic and deterministic estimates of cost effectiveness. The former was limited to clinical effectiveness parameters but this limitation was chosen specifically to draw attention to the uncertainties surrounding these data.

In terms of limitations, there was a dearth of evidence available on the specific population of interest, despite the overall large evidence base on the treatment of GISTs with imatinib or sunitinib. The quality of reporting of dose information in reports of imatinib or sunitinib for GISTs was poor and the data on the population of interest for the studies that were included were non-randomised, non-comparative and therefore observational. Therefore, lack of quality data, as well as lack of data itself, severely limited both assessments of clinical effectiveness and cost-effectiveness.

There was also a lack of evidence on QoL outcomes, which may be of fundamental importance to patients given the potentially palliative nature of treatment following progression, and there was also a lack of evidence on BSC. This is important as since the development of imatinib and sunitinib, it no longer represents the only treatment option for those with unresectable/metastatic disease. There was little evidence on response to escalated doses of imatinib based on mutational status, specifically for those who had already failed on an initial imatinib dose of 400 mg/day. It was also not possible to account for the effects of required dose interruptions and reductions, or the effects of sunitinib on those intolerant to imatinib, owing to the lack of available data. This lack of data also prevented comparative analysis of adverse events between the intervention and comparator treatments.

For sunitinib, it was also necessary to assume that the vast majority of those receiving sunitinib after imatinib treatment at ≤ 400 mg/day had actually received imatinib at 400 mg/day, which may not be a valid assumption. However, it was not possible to confirm the validity of the assumption despite contacting the study authors (P Reichardt, HELIOS Klnikum Bad Saarow, Germany, 2010, personal correspondence). In addition, much of the evidence base for sunitinib generally relates to its use following the failure of escalated doses of imatinib rather than failure on 400 mg/day, suggesting that the role of sunitinib is seen more as a third-line treatment rather than a potential comparator to 600 mg/day or 800 mg/day imatinib treatment. This was highlighted by the manufacturer of imatinib in their submission for this technology appraisal, and is noted in Chapter 3 of this report.

For the economic model, sufficient sound comparative data for the different plausible treatments were not available, despite conducting an extensive review of relevant studies. This necessitated a number of simplifying assumptions being made with respect to the model and also the use of data that were potentially unreliable. The model assumes that patients entering a pathway will remain in that treatment for one cycle only if they do not respond and survive in the treatment arm. In these cases they are considered to move to the escalated doses, move to another alternative (if allowed by a treatment pathway) or continue with BSC for the remainder of the model time horizon or until they die. The care pathways considered in the economic model are not an exhaustive list of all possible treatment options available but represent plausible treatment scenarios. Some are likely to be more representative of clinical practice than others. Whilst additional clinical advice during the development of the care pathways might have increased the extent to which the chosen scenarios reflect true clinical practice, it may also have increased the level of complexity required within the model. Given the lack of robust data it was felt that a more sophisticated model would be difficult to populate.

Within the model, several simplifying assumptions had to be made for individual parameters. For example, it was necessary to consider the costs and utilities associated with BSC as consistent across all care pathways despite the fact that in clinical practice the costs of BSC may increase as an individual's health deteriorates. Unfortunately, there were no data available to model how costs of BSC might increase and QoL might fall over time.

A further simplifying assumption was not to model the complications and side effects of therapy. This latter assumption was made owing to the very limited evidence available. This is coupled with the assumption made that the utility associated with stable response or progression did not vary between treatments. One impact of this assumption is that no utility decrement has been assumed for the arguably worse side effect profile of sunitinib. This means that pathways involving sunitinib may overestimate QALYs.

Perhaps a more important limitation is caused by the limited evidence base available. With respect to the clinical effectiveness data used to derive transition probabilities these data, as already noted, were based upon non-randomised, non-comparative data. Such data are potentially biased as well as being imprecise. In particular, it is worth noting that point estimates of death and response used within the model may be misleading, for example the point estimates used suggest that sunitinib has a higher mortality rate than BSC.


For the assessment of clinical effectiveness:

  • The diagnosis of GIST as stated in the final scope document was based on a positive KIT (CD117) test. However, this is not a perfect test and in a small (< 5%) number of cases a patient can have a GIST despite having a negative KIT (CD117) test.4,7,25 More recent tests (e.g. PDGRFA and DOG1) may clarify diagnosis. However, the WHO classification of GI tumours recommends that a diagnosis of GIST should only apply to those patients testing positive for the KIT (CD117) protein.
  • It was not possible to conduct any subgroup analysis for patients with particular mutations, or consider the methods used to identify response (e.g. FDG-PET or CT scanning), or possible factors related to the provision of dose-escalated imatinib in an adjuvant or neoadjuvant setting.
  • It was not possible within the time frame of this review for sufficient information to be provided that would have enabled meta-analysis of outcomes for the 800 mg/day dose of imatinib. This evidence may have enabled more robust estimates of survival following dose escalation to 800 mg/day. However, the data would still be prone to bias (being taken from data from a non-randomised patient population) and uncertainty surrounding other parameters (e.g. BSC, sunitinib and imatinib at 600 mg/day) would still be likely to make the model difficult to interpret.
  • Following progression, the proportion of patients subsequently progressing on escalated doses, who are kept on the study drug on the basis that progression of disease might be slower than if the patient were to be taken off the drug, is not known. It is also not clear whether there is a standard dose used for this purpose. Within the economic model it has been assumed that this would be the case (400 mg/day).
  • This review only considered drug treatments that were licensed for patients with GISTs and did not consider other drugs that may be being used in the treatment of GISTs, or licensed drugs that are being used ‘off licence’ to treat GIST (e.g. imatinib at doses exceeding 800 mg/day, or sunitinib provided in a continuous daily dosing regime).
  • Surgical interventions were also not considered even though surgery is an important treatment option for GIST patients, and even though those with unresectable disease may be eligible for surgery if their tumours become resectable following treatment with an escalated dose of imatinib. The role of emergency surgery as part of BSC was also not considered.

The economic model has also not considered three main areas of uncertainty due to lack of data:

  • alternative assumptions about how probabilities of death and response change over time
  • reductions in utility associated with side effects of treatment
  • impact on cost-effectiveness for people with different gene mutations.

The impact of making alternative assumptions about how probabilities for death and response change is unknown but it is anticipated that the assumption of constant probabilities over time will exaggerate estimated life expectancy (and hence QALYs and cost) for all pathways. The net impact on relative cost-effectiveness is unclear as it depends upon the magnitude of any changes in both costs and QALYs that might occur.

A further uncertainty is the probability of death for BSC. No studies for this comparator met the inclusion criteria for the review. The only sources available for this parameter were from studies published in the pre-imatinib era where the population could not have been exposed to a prior 400 mg/day dose of imatinib, and the proportion of the study populations with KIT-positive GIST was not known. With regard to the impact of this uncertainty on the economic model, it is reasonable to assume that if, for example, there was an increase in mortality for BSC, the costs and QALYs associated with each of the pathways would fall because BSC is included within each pathway.

The net impact of adjusting utility scores for side effects is also uncertain but it might be expected that it will reduce the QALYs associated with each medication and, although there are limited data available from the systematic review of effectiveness, this reduction may be greater for pathways involving sunitinib because its side effect profile is believed to be worse than that of imatinib.

A further factor not considered by the economic model was the cost-effectiveness of treating patients with specific gene mutations. Again this was not addressed owing to lack of data.

No studies looking at plasma monitoring met our inclusion criteria, but its potential, along with that of mutation testing, as an early predictor of the need for escalated imatinib dosing may have implications for both the costs and effects of escalated doses, because it may allow the identification of those people who are expected to respond better to escalated doses quickly and hence they may be given escalated doses immediately rather than waiting for progression to occur at the 400 mg/day dose. If either of these practices become widely adopted within the NHS then the evidence on the effect of imatinib dose escalation following progression at the standard 400 mg/day dose will become less relevant to clinical practice. Should mutation testing and plasma monitoring allow the tailoring of dose escalation then we might expect the benefits to those who receive therapy to be increased, particularly at earlier stages of treatment (although this also means that there may be fewer remaining treatment options following failure at the escalated dose). Costs would also increase owing to both the cost of mutation testing or plasma monitoring, and also the costs of escalated doses that are incurred earlier. The net impact of this on cost-effectiveness is unclear.

Finally, the economic evaluation has assumed that patients who move on to BSC still receive medication to prevent tumour flare. This has the impact of increasing the cost of BSC. It is further assumed that there is no impact on effectiveness (the implicit assumption is that discontinuing the medication would reduce life expectancy). Within the analysis it has been assumed that all patients on BSC or moving on to BSC after failing to respond to a medication would receive imatinib 400 mg/day. This assumption appears reasonable for Path-1, BSC only, but may not be appropriate for the other pathways where patients would move on to BSC after failing to respond to an escalated dose of imatinib, or on sunitinib. Should these patients continue with the last active medication that they received then costs, and incremental costs per QALY, would increase.

Image ch4f3
Image ch4f4
© 2011, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK99523
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.5M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...