Metastatic Gastrointestinal Stromal Tumors in the Era of Imatinib: Improved Survival and Elimination of Socioeconomic Survival Disparities

Gastrointestinal stromal tumors are a recently recognized disease entity and represent the most common mesenchymal neoplasms of the gastrointestinal tract (1, 2). The annual incidence of gastrointestinal stromal tumor is estimated at 11 to 14.5 cases per million per year (3, 4). Gastrointestinal stromal tumor lesions vary widely with respect to malignant potential, ranging from small, slow-growing, incidentally detected lesions to large, rapidly progressive tumors (2). Approximately 10% to 25% of patients present with metastatic disease (4, 5). Of the remaining patients with localized disease, ∼15% to 43% recur on long-term follow-up after complete surgical resection (6-10). Before the introduction of targeted biological therapies, treatment options in patients with advanced gastrointestinal stromal tumor were limited to conventional cytotoxic chemotherapy and/or radiotherapy, with historically poor response rates and outcomes (7, 11-13).

In early 2002, imatinib mesylate (Gleevec, Novartis Pharmaceuticals), a small-molecule tyrosine kinase inhibitor, was approved for the treatment of unresectable and metastatic gastrointestinal stromal tumor based on a phase II randomized trial that showed a high rate of radiological response and improved survival at 1 year compared with historical controls (14). Subsequent phase II and phase III clinical trials showed an improvement in progression-free survival and overall survival at 2 years (15, 16), and recently completed long-term studies showed a survival advantage at 5 years in select patients (17, 18).

Although the benefit of imatinib therapy is clear, it is uncertain to what degree imatinib treatment has been available to the general population and to what extent a broader improvement in outcome has occurred. Because biological therapies can be expensive, the additional financial burden incurred by such treatment can potentially limit widespread access, especially in low-income and minority patients (19). Given these issue, we have done a U.S. population-based analysis to evaluate overall survival from metastatic gastrointestinal stromal tumor in the era of imatinib therapy. We hypothesized that the introduction of imatinib therapy has led to a population-based improvement in overall survival independent of other clinical factors. An additional objective was to determine whether improvements in survival have been consistent across all racial and income groups.

The Surveillance, Epidemiology, and End Results (SEER) registry (20) was queried for all patients diagnosed with histologically confirmed gastrointestinal stromal sarcomas (International Classification of Disease for Oncology, Third Edition, code 8936; ref. 21) with malignant behavior between 1995 and 2004. All other gastrointestinal mesenchymal neoplasms were excluded. Of the 2,467 patients diagnosed with gastrointestinal stromal tumor, 552 presented with metastatic disease. This subset of metastatic patients was then subjected to statistical analysis.

The primary outcome measure of interest was overall survival. Prognostic factors that were evaluated included year of diagnosis, income, and race. To account for the introduction of imatinib therapy, year of diagnosis was categorized into two groups: 1995 to 2000, corresponding to the period preceding imatinib availability, and 2001 to 2004, corresponding to the era of imatinib use. Although imatinib was not widely available until early 2002, the year 2001 represents a point in which a paradigm shift occurred in the treatment of gastrointestinal stromal tumor. A subset of patients diagnosed with metastatic disease during 2001 received off-label treatment with imatinib. In addition, given historical survival rates, most patients diagnosed with metastatic gastrointestinal stromal tumor in 2001 likely survived long enough to receive imatinib therapy when the drug was approved. For this reason, patients diagnosed in 2001 were included in the imatinib group. Characteristics of the two subgroups were compared using independent sample Student's t tests for continuous variables and the χ² test for categorical variables. Overall survival estimates were determined using the Kaplan-Meier method. Differences in survival across individual subgroups were compared using the log-rank test. Results were reported as median survival estimates, 36-month overall survival estimates, and 47-month estimates, the longest comparable follow-up available for the two periods. Multivariate Cox regression analysis was used to determine the significance of time adjusted for other clinical factors. Results were reported as hazard ratios with 95% confidence intervals (95% CI).

To evaluate socioeconomic factors, the survival analysis was stratified by race and income. Of note, each individual case was assigned an income value equivalent to the median household income of his or her county of residence based on year 2000 U.S. Census Bureau data using the SEER median household income county attribute variable for the year 2000 (20). Therefore, this figure represents an approximation of income at the census-tract level, rather than a true measure of individual household income. Income groups were then categorized into low, middle, and high, with cutoffs defined at the 33rd and 66th percentiles using all 2,467 patients diagnosed with gastrointestinal stromal tumor between 1995 and 2004. Race was categorized initially into Black, White, Hispanic, and other groups. By convention, the Hispanic category included only non-Black Hispanics. Those individuals coded as “Black” by the race-ethnicity variable who were of Hispanic origin were included in the Black race category. However, within our sample population, all patients of Hispanic origin were of “White” ethnicity. The “other” racial designation consisted of 293 Asian or Pacific Islanders, 2 American Indian/Alaskan natives, and 1 individual identified as other with unspecified race. The other category was therefore relabeled "Asian/Pacific Islander," although the data from the additional 3 patients were retained.

Multivariate analysis was done to determine the prognostic significance of income and race before and after the introduction of imatinib therapy. Because of power limitations, race and income were analyzed separately in the multivariate model, and interaction was not examined. All categorical variables were coded as indicator variables. Analyses were adjusted for age, gender, resection, and tumor size. A P value of less than 0.05 was considered statistically significant. The statistical analysis was done using SPSS Version 12.0 (SPSS Inc.) and STATA Version 8.0 (Stata Corporation).

Mean age of diagnosis was 62.5 ± 15.2 years for the entire cohort and 62.5 ± 15.1 years for the metastatic subset. There was a slight male predominance (52% and 57%, respectively), and most patients (63%) were White. Median tumor size was 7.9 cm (mean, 9.5 ± 7.3 cm) for the entire cohort and 10.0 cm (mean, 12.1 ± 9.8 cm) for the metastatic subgroup. Table 1 presents a comparison of characteristics by time for patients with metastatic disease. There was no difference in age, tumor size, or the distribution of gender and race between the two periods. There was, however, a significant difference in the income distribution by time, with a greater proportion of low-income patients in the earlier period (54% versus 40%; P = 0.004). There was also a trend toward a decrease in resection rate over time in patients with metastatic disease (62% versus 53%, 1995-2000 versus 2001-2004, respectively; P = 0.061).

There was a statistically significant increase in survival from the pre-imatinib period (1995-2000) to the era of imatinib therapy (2001-2004; median survival, 12 versus 33 months, respectively; log-rank P < 0.01). Accordingly, 36-month actuarial survival increased from 24% to 48%, and 47-month survival increased from 21% to 41%, respectively (P < 0.001). On multivariate analysis (Table 2), time was an independent prognostic factor for survival adjusted for age, gender, tumor size, and surgical resection, with a 60% reduction in the risk for death in the imatinib era (hazard ratio, 0.40; 95% CI, 0.29-0.55; P < 0.001). Interestingly, the size of the primary tumor was not significantly associated with survival in patients with metastatic disease (hazard ratio, 1.01; 95% CI, 0.99-1.03; P = 0.34) nor did the addition of the variable change the relationship between time and survival (data not shown).

Figures 1 and 2 present survival by time stratified by income and race, respectively. A significant increase in survival was noted in the era of imatinib therapy in all income groups, with median survival increasing from 11 to 38 months in the low-income group (P < 0.001; Fig. 1A), 16 to 35 months in the middle-income group (P = 0.034; Fig. 1B), and 18 to 32 months in the high-income group (P = 0.046; Fig. 1C). Similarly, a significant increase in survival was noted in all racial groups in the era of imatinib. Median survival in the White, Black, Hispanic, and Asian or Pacific Islander populations increased from 16 to 30, 7 to 39, 12 to 23, and 7 to 36 months, respectively (all P < 0.05; Fig. 2A-D).

To determine whether outcome disparities have changed before and after the introduction of imatinib, Kaplan-Meier curves comparing survival by race and income, respectively, were constructed for each period (Fig. 3A-D). Before the introduction of imatinib therapy (1995-2000), there were significant survival disparities based on socioeconomic status (Fig. 3A and C). During this period, low-income patients had a significant survival disadvantage compared with the high-income group (median survival, 11 versus 18 months, respectively; P = 0.042). No statistically significant survival differences were present when comparing the low- and middle-income groups (median survival, 11 versus 16 months; P = 0.71) or the middle- and high-income groups (16 versus 18 months; P = 0.21; Fig. 3A). Conversely, in the era of imatinib therapy (2001-2004), there were no significant survival difference by income and the previously worst outcome group, that is, low-income patients, had a small nonsignificant survival advantage (MS, 38, 35, and 32 months for low-, middle-, and high-income groups, respectively; P for all pairwise comparisons >0.05; Fig. 3B). Similarly, before the introduction of imatinib, Black patients had a significant survival disadvantage compared with the White population (median survival, 7 versus 16 months; P = 0.033; Fig. 3D). Both the Hispanic and Asian or Pacific Islander groups (median survival, 12 and 7 months, respectively) also had decreased survival compared with the White population, although these comparisons did not reach statistical significance (P > 0.05; Fig. 3D). In the era of imatinib therapy, again, there was no significant difference in survival on pairwise comparison. In addition, the worst outcome group from the pre-imatinib era, the Black population (median survival, 39 months), had a small, but nonsignificant, survival advantage compared with the other groups (median survival, 30, 23, and 36 months in White, Hispanic, and Asian or Pacific Islander groups, respectively; all Ps for pairwise comparison > 0.05).

Table 3 presents the multivariate analyses for income and race, respectively, before and after the introduction of imatinib therapy. In the 1995 to 2000 period, both income (overall P = 0.029) and race (overall P = 0.021) were significant prognostic factors for survival, with the low-income and the Black and Hispanic groups defining the groups with the worst survival, respectively. In the era of imatinib, neither income (P = 0.63) nor race (P = 0.40) were significant prognostic factors for survival.

The approval of imatinib mesylate represented a major breakthrough in the treatment of unresectable and metastatic gastrointestinal stromal tumor (14). Imatinib is a small-molecule tyrosine kinase inhibitor that prevents KIT, PDGFRA, ABL, and a number of other transmembrane tyrosine kinase receptors from binding with their respective ligands (22). Molecular and cytogenetic studies have shown a prominent role for KIT in the development and progression of gastrointestinal stromal tumor (23-25). Specifically, activation of KIT has been shown to increase cellular growth and proliferation (26-28). More than 95% of gastrointestinal stromal tumors express KIT (29, 30), and ∼75% to 80% have KIT mutations that lead to constitutive activity of the receptor (15, 23–25). Binding of KIT by imatinib results in a decrease in cell proliferation and cell survival, especially in tumors with KIT mutations (15, 31, 32). In addition, ∼5% of gastrointestinal stromal tumors have activating mutations of PDGFRA (33) and often respond to imatinib therapy (15, 33).

Before the introduction of imatinib, therapeutic options for the treatment of gastrointestinal stromal tumor were extremely limited (1, 12, 34). After curative surgical resection, recurrence was common, and effective adjuvant treatments were unavailable (7, 34). In the setting of unresectable and metastatic disease, conventional doxorubicin-based chemotherapy and radiation were the only options; and treatment in this manner resulted in extremely low response rates and frequent disease progression (7, 11-13, 34). After showing the in vitro imatinib efficacy in gastrointestinal stromal tumor, an open-label, phase II, randomized, multicenter clinical trial was done that showed a significant radiological response rate of >50%, with >25% stable disease in patients with unresectable and/or metastatic tumors (14). Based on this study, the Food and Drug Administration approved imatinib for use in advanced gastrointestinal stromal tumor in February 2002. Subsequent phase II and III studies showed significant radiological response rates in the setting of advanced disease, with improved 1, 2, and 5-year progression-free and overall survival compared with historical controls (15-18).

It is now evident that imatinib therapy improves outcome in patients with advanced gastrointestinal stromal tumor. However, because these studies were done in select patients at specialized centers, it is still uncertain how the introduction of imatinib has impacted survival on a broader level within the U.S. population. Previous population-based data have suggested an improvement in survival after 2000 in patients with gastrointestinal mesenchymal neoplasms (35). However, this study was done using a heterogeneous sample population with various stages of disease and a variety of histologies other than gastrointestinal stromal tumor, including other gastrointestinal smooth muscle tumors and nerve sheath tumors. Given this heterogeneity, these results are difficult to interpret.

Our results show convincingly that overall survival from metastatic gastrointestinal stromal tumor has increased in the U.S. population in the era of imatinib, and that this improvement in survival has been independent of patient age, tumor size, and surgical resection. Although we did not directly measure imatinib use, given that no other effective therapy exists, and given that the analysis was adjusted for other important clinical prognostic factors, the survival advantage we have shown was likely the direct result of imatinib treatment. Interestingly, despite this increase in survival, imatinib was not a panacea, and most patients continued to die of their disease, with a near 4-year overall survival of <50%. This observation is in part related to the inherent biological heterogeneity of gastrointestinal stromal tumor lesions with resultant variability in response to imatinib. For example, tumors that harbor mutations of KIT in exon 11 have a better response to therapy (15, 32), whereas other mutations or wild-type KIT/PDGFRA are predictive of poorer long-term outcome (15, 18, 33).

Few patients are cured with imatinib therapy, and cessation of treatment, even in responders, ultimately results in recurrence (7, 34). Because imatinib treatment is largely suppressive in nature, surgery in patients with metastatic disease is often entertained when complete resection is possible, especially when tumor response to imatinib has diminished. However, definitive evidence to support such an aggressive approach is lacking (36-38). Our data show that, in the setting of metastatic disease, surgical resection remains an independent predictor of improved survival, and this survival advantage persists into the era of imatinib therapy.

Because of the tumor-suppressive, rather than tumoricidal, nature of imatinib, the current recommendation is to continue drug administration indefinitely in responders (39). Patients must undergo extended courses of treatment. The long-term use of imatinib has several implications with respect to cost and, in turn, access to care. Targeted therapies such as imatinib, although highly effective, are costly, especially when treatment is of indefinite duration. The expected result is a decrease in compliance in patients who are unable to pay for therapy (19, 40). The consequence of this additional financial burden is unequal access to novel treatments and a resultant disparity in outcomes from advanced disease. Although such disparities may be based on economic grounds, to the degree that race is associated with income, significant racial disparities can also arise. For this reason, medication assistance programs were developed to provide drugs to individuals with limited economic means (19). Although the cost-effectiveness of such programs has already been shown (19, 41-43), the impact of assistance programs on health outcomes has not been shown.

In light of these issues, we sought to determine whether significant socioeconomic survival disparities were present with metastatic gastrointestinal stromal tumor and if these disparities were affected by the introduction of imatinib. Our results show that the improvement in survival from metastatic gastrointestinal stromal tumor in the era of imatinib occurred in all racial and income groups. Our results also show that, before the introduction of imatinib, significant outcome disparities based on race and income were already present, with notable survival disadvantages in Black, Hispanic, and low-income populations. These disparities may have been related to unequal access to the only potentially effective treatment in the pre-imatinib era, adequate surgical resection. Our data further show that, in the era of imatinib therapy, survival disparities based both on income and race were no longer statistically apparent, with the previous highest-risk groups showing at least equivalent long-term survival compared with the White and high-income populations. Although it is difficult to determine exactly why this shift occurred, changes in access patterns were almost certainly involved. One plausible explanation is the presence of an assistance program initiated by Novartis Pharmaceuticals to provide drug at no cost to patients unable to pay for treatment. Because imatinib was already in use for the treatment of chronic myelogenous leukemia before 2002, this assistance program was already in place and was able to be implemented immediately upon approval of the drug for gastrointestinal stromal tumor. Given the cost-effectiveness of assistance programs, our findings provide a further argument for their continued expansion, especially in light of the ongoing emergence of more advanced and more costly targeted therapies.

The results we have presented should be interpreted in light of several limitations associated with all retrospective database studies and limitations specific to the SEER registry. The SEER limited-use data set does not provide any information regarding chemotherapy. We cannot show directly the relationship between imatinib therapy and improved long-term survival. However, because other clinically relevant predictors of survival for metastatic gastrointestinal stromal tumor were accounted for, imatinib therapy was the mostly likely reason for the observed improvement in survival. An additional limitation is that SEER reports household income at the census tract level and not at the individual level. Although it is generally true that individual incomes are roughly similar in a given census tract, heterogeneity may still exist.

In conclusion, we have shown that long-term overall survival from metastatic gastrointestinal stromal tumor has increased significantly in the U.S. population in the era of imatinib therapy, corroborating the results of large-scale multi-institutional randomized trials. However, survival in this broad cross section of metastatic gastrointestinal stromal tumor still lags behind the results observed in select patients with known KIT expression and KIT/PDGFRA mutations, indicating that treatment has not reached all patients with susceptible tumors and reaffirming that more effective treatments are necessary in high-risk groups. The improvement in survival in the era of imatinib has occurred in all racial and income groups, an indication that imatinib treatment has reached all patients regardless of socioeconomic status. Significant survival disparities from metastatic gastrointestinal stromal tumor were noted before the availability of imatinib, and it has been in the era of imatinib that these disparities have diminished. This shift toward equality is potentially the result of universal access to imatinib facilitated by an assistance program offered by the drug manufacturer. In light of the ongoing emergence of newer and more costly biological therapies, our findings provide further support for the development and expansion of drug-assistance programs to ensure broad availability of effective treatment.

Warren Chow received honoraria from the Interactive Network for Continuing Education Speakers Bureau. The other authors disclosed no potential conflicts of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.