Skip Navigation U.S. Department of Health and Human Services
Agency for Healthcare Research Quality
Archive print banner
Report on the Relative Efficacy of Oral Cancer Therapy for Medicare Beneficiaries Versus Currently Covered Therapy

This information is for reference purposes only. It was current when produced and may now be outdated. Archive material is no longer maintained, and some links may not work. Persons with disabilities having difficulty accessing this information should contact us at: Let us know the nature of the problem, the Web address of what you want, and your contact information.

Please go to for current information.

Part 2. Imatinib for Gastrointestinal Stromal Tumors (GISTs) (continued)


Search Strategy

The search strategy was constructed by combining three concepts:

  1. The intervention imatinib.
  2. The disease gastrointestinal stromal tumor.
  3. Prospective clinical trials.

To identify the intervention concept, since these new drugs lack a specific term in the MeSH® lexicon, we used text word searching for the following text strings: imatinib or gleevec or glivec or STI571. The disease concept was implemented using the text word searching for gist or adjacent text strings for gastro$ within two words of stromal adjacent to (tumo$ or cancer$). This is designed to detect various spellings such as gastrointestinal stromal tumor or gastrointestinal stromal cancer or gastro-intestinal stromal cancer, etc. A published strategy, validated for finding randomized controlled trials (RCTs), was used to identify prospective clinical trials. This strategy is designed to find all prospective clinical trials (maximize sensitivity), rather than to eliminate non-randomized trials (maximize specificity), and so is appropriate for this study's goal of finding phase II and III prospective clinical trials. Finally, the three concepts were combined (Boolean "or"). The strategy was executed in MEDLINE® (1966 through September 2004, updated February 2005) and limited to articles published in the English language. The exact text of the OVID MEDLINE® versions of the search strategy is provided in Appendix A.

Supplemental searches were conducted in International Pharmaceutical Abstracts, The Cochrane Library (Central Register of Controlled Trials (CENTRAL) and Health Technology Assessment22 database), American Society of Hematology 2004 annual meeting abstracts database, and in the American Society of Clinical Oncology 2004 annual meeting abstracts database. References lists of identified studies and relevant systematic reviews and meta-analyses were hand-checked. Additional articles not indexed in the major bibliographies by September 2004 were identified through ongoing searches and discussions with field experts and monitoring new sources.

Selection Criteria

Each citation identified from the search strategies was evaluated according to the following selection criteria. Evaluations were performed by the authors.

Inclusion criteria were as follows:

Patients: Patients with unresectable or metastatic GIST
Interventions: Imatinib (Gleevac™ or Glivec™ or [STI571])
Comparators: Any

Study designs:

  • For efficacy questions: Prospective clinical trials; may be phase II uncontrolled, or phase III randomized controlled trials.
  • For studies of adverse effects: May be retrospective or prospective case series, cohort studies, or clinical trials, provided the number of patients treated (at risk for adverse effects) as well as the number with adverse effects can be ascertained.
  • For studies of predictors of response: May be retrospective or prospective case series, cohort studies, case-control studies, or clinical trials, provided the response can be ascertained for patients with and without the predictor.


  • For efficacy questions: Survival, disease-free survival, tumor response, and quality of life (QOL). Tumor response was defined according to Table 2.
  • For studies of adverse effects: Adverse effects, tolerability, and compliance with treatment.
  • For studies of predictors of response: Predictive value of patient or tumor characteristics that are associated with clinically important differences in treatment response that are:
    1. Related to the mechanism of action of the drug (i.e., molecular target).
    2. Candidates for diagnostic testing (even if not commercially or clinically available currently (e.g., Polymerase Chain Reaction)).

Data Abstraction

The following data were abstracted from included studies: study design, population characteristics (including sex, age, and diagnosis), eligibility and exclusion criteria, interventions (dose and duration), outcomes assessed and results for each outcome.

We developed data collection forms in Excel® (Microsoft®; Redmond, WA) and summarized the data in evidence tables formatted like those in a 2003 report from NICE.16

Quality Assessment

We assessed the quality of included studies by evaluating elements of internal validity (e.g., randomization and allocation concealment; similarity of compared groups at baseline; specification of eligibility criteria; blinding of assessors, care providers, and patients) and external validity (e.g., description of the patient population, similarity to the target population of the report, use of highly selective criteria).

We used as a framework the quality assessment criteria from NICE.16 These are displayed in Appendix B. They provide specific criteria for the range of study designs used in this report including experimental studies, cohort studies, case-control studies, and case series.

Point scores were allocated by assigning one point for each quality category. There were a total of 6 possible categories. Quality ratings of "yes" to a quality criteria were assigned 1 point; no and unknown were both assigned 0 points. The last category, adequate description of subseries, was not applicable to all studies. Hence, the total possible quality points were 5 or 6 depending upon the applicability of the subseries category. High quality studies were those with ≥ 3/5 or 4/6 points.

Abstract quality was not scored.

Data Synthesis

In addition to the data abstraction and quality analysis, a narrative description of study findings was prepared. Further quantitative analyses were considered, but the available data were not adequate to support these.

Return to Contents


The search strategy yielded 88 articles. The selection process is described below:

Identified by search strategy (N = 88)

     |------ Excluded based on review of abstract (N = 54)

Included based on review of abstract (N = 34)

     |------ Excluded based on full-text review (N = 11)
     |          4 not phase II-III for efficacy
     |          3 no primary or original data (review article)
     |          1 wrong disease
     |          3 wrong outcome

Included in full-text review and evidence tables (N = 23)

The 23 included reports comprised 16 full reports and 7 abstract-only publications (Table 3). Study designs included two phase III controlled clinical trials (ongoing, three abstract reports), five phase II uncontrolled clinical trials (four trials in eight full reports plus one trial in one abstract), and eight studies of other designs. Four of these studies with other designs evaluated the role of surgery plus imatinib and four were investigating the role of imaging techniques in predicting GIST outcomes. There were several sub-studies of the phase II trials assessing predictors of GIST outcomes—one assessing radiological predictors, four of tumor characteristics, and two of other clinical predictors. All of the adverse events data were derived from the four phase II clinical trials that were published in full reports.

Quality of the studies varied by outcome category. All of the main imatinib efficacy studies published in full were of high quality. Only one of the four imatinib plus surgery studies was of high quality. Two of the four molecular predictor studies were of high quality, with one low quality study and one abstract. Two of five radiological predictor studies were of high quality, with one low quality study and two abstracts. With the exception of the one abstract, all of the studies reported other types of predictors were of high quality. In summary, study quality was generally high, with surgical and radiological studies being the most suspect.

Return to Contents


Imatinib Alone for Advanced GIST

There are four completed high quality studies of imatinib alone for the management of unresectable or metastatic CD117-positive GIST (Table 4) involving a total of 1156 patients with GIST. The patient populations were nearly identical except for the studies by van Oosterom, et al. and Verweij, et al. (2001), which included some patients with non-GIST soft tissue sarcomas (STS). When described, 83-98 percent of patients had prior surgery, 33-71 percent of patients had prior chemotherapy, and 8-24 percent of patients had prior radiotherapy. The average age was 55 with a range of 18-83; 62 percent were male.

The completed trials comprise several different designs. The first is a mixed phase design with a dose escalation phase from 400 mg to 1000 mg and phase III follow through at 800 mg (total N = 40).6,45 The second is a randomized phase II with patients randomized to 400 or 600 mg daily (total N = 147).43,46,47 The third is a standard phase II design evaluating efficacy of 800 mg daily (N = 50).48 The fourth is randomized phase III with patients randomized to 400 or 800 mg daily (total N = 946).49 Followup lasted 9-25 months. There is one incomplete phase II trial reported in abstract form only.51 This trial appears to be a standard phase II with 135 patients (mean age 65, 70 percent male) receiving an unspecified dose of imatinib.

Tumor response was the most consistent outcome reported. In the completed phase II and phase III studies, complete response (CR) rates ranged from 0-6 percent, partial response (PR) rates ranged from 45-67 percent, and stable disease (SD) rates ranged from 19-47 percent. Overall response (CR + PR) was 49-71 percent.

Overall survival (OS), disease-free survival (DFS), event-free survival (EFS), and progression-free survival (PFS) were variably reported. Verweij, et al. (2004) had the longest median followup (760 days), and reported 1 year OS as 85 percent for either the 400 mg/d or 800 mg/d doses and 2 year OS as 69 percent for the 400 mg/d dose and 74 percent for the 800 mg/d dose (p not significant).49 Two-year PFS on that study was 44 percent for the 400 mg/d dose and 50 percent for the 800 mg/d dose (p = 0.026). The study initially described by Demetri, et al. in 2002 had a subsequent report by Heinrich, et al. in 2003 that reported median followup of 594 days.46, 47 The OS reported for all patients was 85 percent at 76 weeks, with a median EFS of 17 months.47

Two phase III efficacy studies of imatinib are currently ongoing. Both of these are published in abstract form only. Both compare different dosing regimens of imatinib. Blay and colleagues are testing continuous vs. intermittent imatinib dosing schedules52 and Rankin, et al. are testing 400 mg/d vs. 800 mg/d.53 Results are preliminary. The study of 400 mg/d vs. 800 mg/d involved 746 participants and reports 2-year OS at 78 percent (CI 73-82%) for the 400 mg/d group and 73 percent (CI 68-77%) for the 800 mg/d group. The 2-year PFS is 50 percent (CI 45-55%) for the 400 mg/d group and 53 percent (CI 47-58%) for the 800 mg/d group.

The only significant difference in outcomes identified between different dosing schedules was that seen in the Verweij, et al. (2004) study where patients on 800 mg/d have significantly higher PFS rates.49 Two studies have evaluated the potential to increase the dose from 400 mg/d to 800 mg/d in the setting of tumor progression at the lower dose. In a subsequent abstract report of the Verweij, et al. (2004) study, 220 of 473 patients on the lower dose of 400 mg/d progressed; 65 percent (143) crossed over to the 800 mg/d dose with 26 percent progression-free at one year.50 Toxicity required dose reductions in 31 percent of those who crossed over to the higher dose. In the ongoing Rankin, et al. phase III trial, 164 patients on the lower dose of 400 mg/d progressed; 54 percent (88) crossed over to the 800 mg/d dose with median PFS of 4 months and median OS of 19 months after crossover. The overall interpretation of these studies is that crossover is feasible with some tumor responses.

Imatinib Plus Surgery

Four studies were identified that reviewed the role of imatinib in peri-surgical settings (Table 5) including neoadjuvant imatinib prior to planned tumor resection, adjuvant imatinib for tumors at high risk of recurrence after complete resection, and palliative imatinib after incomplete surgery. Only one of these was prospective (Bumming55), while the others were retrospective reviews of prospectively collected registry data. The summary finding from these small lower-quality studies is that neoadjuvant and adjuvant imatinib is feasible. The efficacy is still to be determined.

Quality of Life (QOL) on Imatinib With GIST

None of the studies reviewed reported formal QOL analyses. Only Demetri, et al. reported any outcomes that could be categorized as QOL.46 This study looked at the effect of imatinib on Eastern Cooperative Oncology Group (ECOG) performance status (Table 6), comparing scores between baseline and 4 months. Some authors consider performance status to be a crude measurement of QOL.67 At baseline, 42 percent of participants were ECOG 0 and 19 percent were ECOG 2-3. This substantially improved after imatinib, such that at 4 months 64 percent of participants were ECOG 0 and 5 percent were ECOG 2-3.

Table 6. The Eastern Cooperative Oncology Group (ECOG) Performance Status Scale

0 = fully active
1 = restricted in strenuous activity only
2 = unable to work; up and about more than 50% of waking hours
3 = confined to bed or chair more than 50% of waking hours
4 = completely disabled; totally confined to bed or chair

Adverse Effects/Harms

Adverse effects that are possibly or likely associated with imatinib for GIST are described in Tables 7a and 7b. The most common adverse effects are edema, nausea, and diarrhea. The edema is predominantly superficial, with some authors breaking this out to highlight the periorbital edema commonly described with imatinib. Few patients (< 36 percent) experienced any grade 3 or 4 toxicity at lower doses, and when they did this was predominantly hematological or hemorrhage (< 8 percent). These studies were often conducted in patients with bulky, advanced GIST, making it difficult to ascertain which adverse events were truly due to imatinib and which might have occurred due to the disease process itself. Toxicity was somewhat more common at the higher doses, with 800 mg/d being the most toxic dose fully described. These toxicities compare favorably to those of traditional cytotoxic agents.28,31,32

Predictors of Response

All reports on predictors of response or survival from the clinical studies of imatinib for GIST are shown in Tables 8-10. Response predictors were divided into three groups:

  1. Tumor characteristics.
  2. Radiological studies predicting GIST response to imatinib.
  3. Other clinical prognostic factors. Three factors had data from two studies evaluating each factor (c-kit and PDGFRA mutational status, positive emission tomography (PET) as a complementary tool to computed tomography (CT), and performance status). All other factors evaluated were presented in one study only.

The most relevant predictors of response relate to the mechanism of action of imatinib (Table 8). Patients with GIST (presumably expressing CD117) were more likely to respond to imatinib than patients with other soft tissue sarcomas (presumably CD117 negative).48 Patients with identified c-kit mutations were less likely to progress on imatinib and had longer overall survival.59 Both exon 11 and exon 9 mutations positively influenced response to imatinib.47,59 Some other genetic predictors such as SPRY4 and MAFbx may be clinically relevant, but current studies are too small to make valid conclusions at this time.60

Early response on PET scan at day 8 predicted clinical response to imatinib (Table 9).61 Patients with positive evidence of response on day 8 PET had a 92 percent PFS rate, while those with no evidence of response had a 12 percent PFS rate. Studies of the differential performance of PET vs. CT were mixed, one suggesting PET was a more sensitive indicator of the presence of tumor and the other suggesting no difference between the two modalities. In the high quality study comparing multiple types of PET and CT, combined PET-CT was the most sensitive.62,63 This study was specifically evaluating the differential ability of the modalities to predict response to imatinib.62 Promising new radiological techniques are on the horizon, some of which may be more sensitive than CT (e.g., angio-echography with BR-1 contrast) and some of which may be as good as CT but less invasive (e.g., ultrasound with perfusion software).64,65

Other clinical factors associated with poor response to imatinib included poor performance status, lower doses of imatinib, renal dysfunction, and prior chemotherapy (Table 10).47,49

Return to Contents
Proceed to Next Section

The information on this page is archived and provided for reference purposes only.


AHRQ Advancing Excellence in Health Care