Purpose: To analyze final long-term survival and clinical outcomes from the randomized phase III study of sunitinib in gastrointestinal stromal tumor patients after imatinib failure; to assess correlative angiogenesis biomarkers with patient outcomes.

Experimental Design: Blinded sunitinib or placebo was given daily on a 4-week-on/2-week-off treatment schedule. Placebo-assigned patients could cross over to sunitinib at disease progression/study unblinding. Overall survival (OS) was analyzed using conventional statistical methods and the rank-preserving structural failure time (RPSFT) method to explore cross-over impact. Circulating levels of angiogenesis biomarkers were analyzed.

Results: In total, 243 patients were randomized to receive sunitinib and 118 to placebo, 103 of whom crossed over to open-label sunitinib. Conventional statistical analysis showed that OS converged in the sunitinib and placebo arms (median 72.7 vs. 64.9 weeks; HR, 0.876; P = 0.306) as expected, given the cross-over design. RPSFT analysis estimated median OS for placebo of 39.0 weeks (HR, 0.505, 95% CI, 0.262–1.134; P = 0.306). No new safety concerns emerged with extended sunitinib treatment. No consistent associations were found between the pharmacodynamics of angiogenesis-related plasma proteins during sunitinib treatment and clinical outcome.

Conclusions: The cross-over design provided evidence of sunitinib clinical benefit based on prolonged time to tumor progression during the double-blind phase of this trial. As expected, following cross-over, there was no statistical difference in OS. RPSFT analysis modeled the absence of cross-over, estimating a substantial sunitinib OS benefit relative to placebo. Long-term sunitinib treatment was tolerated without new adverse events. Clin Cancer Res; 18(11); 3170–9. ©2012 AACR.

Translational Relevance

We present complete longitudinal analyses, including final overall survival (OS), from the pivotal phase III study that led to regulatory approval of sunitinib in patients with advanced imatinib-resistant/intolerant gastrointestinal stromal tumor. We show the confounding effect of the cross-over design of the study (patients could switch from placebo to sunitinib) on OS: conventional intent-to-treat analysis showed that OS converged in the treatment arms, showing no statistical difference, despite statistically significant differences in tumor control rates. However, using the exploratory rank-preserving structural failure time method to model the absence of cross-over, we estimated that sunitinib conferred a long-term OS benefit relative to placebo (73 vs. 39 weeks). We also analyzed circulating biomarkers of angiogenesis during sunitinib treatment and found no consistent associations with clinical outcome. Finally, these long-term exposure data showed that no new safety concerns emerged with extended sunitinib treatment.

The key pathogenic event causing the neoplastic phenotype for the majority of gastrointestinal stromal tumors (GIST) has been shown to be mutation of the genes encoding KIT (stem cell factor receptor) or platelet-derived growth factor receptor α (PDGFR-α), resulting in constitutively activated signaling through these receptor tyrosine kinases (RTK; refs. 1–3). The oral multitargeted RTK inhibitor sunitinib malate targets a number of RTKs, including KIT, PDGFRs-α and -β, and VEGF receptors (VEGFR)-1, -2, and -3 (4–9), thereby blocking key kinases involved in maintaining the survival and proliferation of GIST cells and signals of tumor-related angiogenesis.

Previously reported results of the planned interim analysis of a double-blind, placebo-controlled, randomized phase III study of sunitinib in patients with GIST following imatinib failure showed significant clinical benefit from sunitinib treatment with acceptable safety (10). On the basis of these results, sunitinib received multinational regulatory approval for the treatment of advanced imatinib-resistant or imatinib-intolerant GIST (9, 10).

Following unblinding due to the positive impact of sunitinib on time to tumor progression (TTP), all patients randomized to receive placebo were offered the opportunity to cross over to open-label sunitinib treatment (10). This trial was designed knowing that cross-over to an active agent would necessarily confound any estimate of treatment effect on overall survival (OS) using conventional statistical methods. We now report final results from this study, including OS and toxicities observed with long-term sunitinib administration; OS was analyzed using both conventional statistical techniques and an exploratory method to estimate this effect of cross-over on survival. To expand on the previously reported finding that the extent of reduction in plasma soluble (s)KIT during sunitinib treatment correlated strongly with improved clinical outcome in this study (11), we examined possible associations between plasma levels of the angiogenesis-related proteins VEGF-A, sVEGFR-2, and sVEGFR-3 and sunitinib efficacy in a subset of patients.

### Study population

The study population comprised adults with histologically proven GIST for whom prior imatinib treatment had failed due to resistance or intolerance, an Eastern Cooperative Oncology Group performance status (ECOG PS) ≤1, and adequate hepatic, renal, and cardiac function as described previously (10).

### Study design and treatment

Patients were randomized 2:1 to receive double-blind treatment in repeated 6-week cycles of 4 weeks of daily sunitinib (50 mg) or placebo followed by 2 weeks off treatment (Supplementary Fig. S1). At the time of disease progression or termination of the double-blind phase of the study (whichever occurred first), treatment assignments were unblinded, and patients randomized to sunitinib were permitted to continue treatment. Eligible patients randomized to placebo were allowed to cross over to open-label sunitinib treatment. The study conformed to the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. The institutional review boards of participating study centers approved the protocol. All participants provided written informed consent.

### Clinical assessments

The primary efficacy endpoint of this study was TTP. Secondary efficacy endpoints included OS, progression-free survival (PFS), and overall confirmed objective response rate (ORR). Tumor imaging studies, which have been described previously (10), were carried out at baseline, on day 28 of each treatment cycle, and at the end of treatment, or more frequently as required; disease assessments were made using Response Evaluation Criteria in Solid Tumors (12). Primary analysis of response-based efficacy endpoints used assessments from an independent, third-party, core imaging laboratory.

Safety was evaluated by analysis of adverse events (AE), physical examinations, vital signs, ECOG PS, laboratory tests, and cardiac function assessments as previously described (10). AEs were graded using National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0.

### Assessment of plasma protein biomarkers

VEGF-A, sVEGFR-2, and sVEGFR-3 were chosen for evaluation on the basis of their relationship to known sunitinib molecular targets, their established roles in VEGF-related signaling and angiogenesis, and on reproducible plasma pharmacodynamics obtained in sunitinib trials in a number of tumor types (13–15). Blood was collected in heparinized tubes predose at baseline, on days 14 and 28 of cycle 1, and on days 1 and 28 of subsequent cycles. Plasma samples were stored at −70°C; storage duration was within the period covered by stability assessment for each protein analyzed. Plasma concentrations of VEGF-A, sVEGFR-2, and sVEGFR-3 were determined using ELISA kits from R&D Systems. The VEGF-A assay measured the VEGF-A165 and VEGF-A121 isoforms. Plasma sVEGFR-2 and sVEGFR-3 were measured using assays that detect the extracellular domains of the respective full-length receptors without cross-reactivity. The sVEGFR-2 and sVEGFR-3 assays were calibrated against recombinant proteins consisting of extracellular receptor domains. All assays were run under Good Laboratory Practice conditions, and performance specifications of each ELISA were validated for their intended purpose according to established guidelines (16).

In the placebo arm, plasma protein values obtained after cross-over to sunitinib treatment were excluded from analysis, and TTP values from these patients were censored at cross-over. The correlative analyses reported here represent evaluations of specific individual biomarker hypotheses as described above, and corrections for multiple comparisons were not applied. The numbers of samples analyzed (at baseline: sunitinib arm, n = 68–88; placebo arm, n = 33–60) were comparable with those analyzed in earlier sunitinib trials in several tumor types (13–15).

### Statistical analyses

Efficacy analyses were based on the intent-to-treat (ITT) population (all patients randomized), comparing the treatment arms as randomized. Safety analyses were conducted on the per-protocol patient population (who received 1 dose or more of assigned blinded study treatment) and the population that received 1 or more dose of open-label sunitinib treatment (Supplementary Fig. S2). Time-to-event data (including those stratified by exploratory biomarkers) were analyzed using the Kaplan–Meier method, log-rank test, and Cox proportional hazards model.

In addition, OS was analyzed using the rank-preserving structural failure time (RPSFT) method (17). This method assumes that treatment does not change the order of events (i.e., it preserves rank), but alters the time at which events occur, with treatment prolonging survival by a multiplicative factor [exp(ψ) in the model]. Thus, the structural or causal assumption of this failure time (i.e., survival time) model is that time is proportional for an individual patient (in contrast to the assumption underlying the Cox proportional hazards model, in which there is a multiplicative relationship between the hazard rates of any 2 patients, which is constant over time). This analysis (done on the ITT population) accounts for cross-over by estimating the treatment effect in the placebo arm (as randomized) by relating a patient's observed event time to the treatment effect parameter ψ and the event time that would reasonably be estimated to occur in the absence of cross-over. If T is the observed event time for a patient who crossed over from placebo to open-label sunitinib treatment and Tp is the time on placebo, then |$U = T_p + \exp ({\rm \psi })(T - T_p)$|⁠, in which U is the event time that would have been observed if no treatment had been given and exp(ψ) is the multiplicative effect of having started treatment. The patients' times on treatment after cross-over (TTp) are adjusted to reflect what would have happened if they had stayed on placebo. ψ is estimated by computing U for a range of possible values of ψ and finding the value that yields a zero in the log-rank test statistic when the 2 treatment arms are compared using time U (as one would expect no difference between the 2 arms before treatment due to randomization).

The RPSFT method does not require that cross-over and prognosis are independent, and estimation of U would have incurred this type of bias without proper handling (18, 19). The recensoring procedure described by Robins and Tsiatis (14) was therefore employed to overcome this potential bias. The HR comparing treatment with sunitinib and placebo following RPSFT analysis was estimated by carrying out a Cox regression analysis on the observed event times in the sunitinib arm and the estimated U values for cross-over patients or the observed event times for other patients in the placebo arm. Because the P value obtained in this Cox model does not reflect the uncertainty inherent in the treatment effect parameter ψ, the P value obtained in the conventional ITT analysis was used with the HR obtained following RPSFT analysis (18, 19).

From December 2003 through January 2005, 312 patients from Europe, the USA, Australia, and Asia were enrolled and randomized (2:1; sunitinib, n = 207; placebo, n = 105) in double-blind fashion. The double-blind phase was terminated in January 2005 when a planned interim analysis revealed significantly longer TTP for those randomized to sunitinib versus placebo (10). Following the interim analysis, enrollment continued through May 2005, ultimately with 361 patients from 56 sites randomized to either sunitinib (n = 243) or placebo (n = 118); both arms were well balanced for demographic and prior imatinib treatment characteristics (Table 1). Patient disposition and flow throughout of the study are shown in Supplementary Fig. S2. This report presents data analysis from the full ITT population of 361 patients.

Table 1.

Baseline patient characteristics and prior imatinib treatment history

ITT populationBiomarker subset
Sunitinib (n = 243)Placebo (n = 118)Sunitinib (n = 88)Placebo (n = 60)
Median age, y (range) 57 (23–84) 55 (23–81) 57 (28–81) 55 (26–81)
Sex, n (%)
Male 152 (63) 71 (60) 52 (59) 41 (68)
Female 91 (37) 47 (40) 36 (41) 19 (32)
ECOG performance status, n (%)
0 109 (45) 53 (45) 42 (48) 30 (50)
1 131 (54) 63 (53) 45 (51) 30 (50)
2 3 (1) 2 (2) 1 (1) 0 (0)
Median tumor burdena, mm (range) 227 (18–722) 240 (29–749) 218 (45–707) 243 (29–749)
Most common metastatic sites, n (%)
Liver mass/nodule 138 (57) 68 (58) 67 (76) 44 (73)
Peritoneal mass/nodule 100 (41) 45 (38) 44 (50) 29 (48)
Mesenteric adenopathy 26 (11) 15 (13) 11 (13) 11 (18)
Previous imatinib therapy
Median maximum daily dose, mg (range) 800 (300–1,600) 800 (400–1,600) 800 (300–1,600) 800 (400–1,600)
Median cumulative treatment duration, wk (range) 107 (0.3–206) 108 (11–231) 104 (7–195) 105 (12–174)
Treatment outcome, n (%)
Progression ≤6 mo 42 (17) 20 (17) 16 (18) 10 (17)
Progression >6 mo 188 (77) 94 (80) 68 (77) 47 (78)
Intolerance 13 (5) 4 (3) 4 (5) 3 (5)
Best response, n (%)
Complete response 9 (4) 1 (1) 3 (3) 0 (0)
Partial response 62 (26) 43 (36) 20 (23) 22 (37)
Stable disease 97 (40) 40 (34) 44 (51) 20 (33)
Progressive disease 68 (28) 32 (27) 18 (21) 17 (28)
Not applicable/missing 7 (3) 2 (2) 2 (2) 1 (2)
Time since diagnosis
Median, y (range) 3.2 (0.2–26.8) 3.3 (0.2–16.4) 3.2 (0.2–15.7) 3.2 (0.5–9.5)
ITT populationBiomarker subset
Sunitinib (n = 243)Placebo (n = 118)Sunitinib (n = 88)Placebo (n = 60)
Median age, y (range) 57 (23–84) 55 (23–81) 57 (28–81) 55 (26–81)
Sex, n (%)
Male 152 (63) 71 (60) 52 (59) 41 (68)
Female 91 (37) 47 (40) 36 (41) 19 (32)
ECOG performance status, n (%)
0 109 (45) 53 (45) 42 (48) 30 (50)
1 131 (54) 63 (53) 45 (51) 30 (50)
2 3 (1) 2 (2) 1 (1) 0 (0)
Median tumor burdena, mm (range) 227 (18–722) 240 (29–749) 218 (45–707) 243 (29–749)
Most common metastatic sites, n (%)
Liver mass/nodule 138 (57) 68 (58) 67 (76) 44 (73)
Peritoneal mass/nodule 100 (41) 45 (38) 44 (50) 29 (48)
Mesenteric adenopathy 26 (11) 15 (13) 11 (13) 11 (18)
Previous imatinib therapy
Median maximum daily dose, mg (range) 800 (300–1,600) 800 (400–1,600) 800 (300–1,600) 800 (400–1,600)
Median cumulative treatment duration, wk (range) 107 (0.3–206) 108 (11–231) 104 (7–195) 105 (12–174)
Treatment outcome, n (%)
Progression ≤6 mo 42 (17) 20 (17) 16 (18) 10 (17)
Progression >6 mo 188 (77) 94 (80) 68 (77) 47 (78)
Intolerance 13 (5) 4 (3) 4 (5) 3 (5)
Best response, n (%)
Complete response 9 (4) 1 (1) 3 (3) 0 (0)
Partial response 62 (26) 43 (36) 20 (23) 22 (37)
Stable disease 97 (40) 40 (34) 44 (51) 20 (33)
Progressive disease 68 (28) 32 (27) 18 (21) 17 (28)
Not applicable/missing 7 (3) 2 (2) 2 (2) 1 (2)
Time since diagnosis
Median, y (range) 3.2 (0.2–26.8) 3.3 (0.2–16.4) 3.2 (0.2–15.7) 3.2 (0.5–9.5)

aSum of longest diameters of target lesions.

Randomized patients who received drug per protocol before unblinding of study medication included 228 in the sunitinib and 114 in the placebo arms, respectively (Supplementary Fig. S2). Sixty-three percent of placebo patients received open-label sunitinib following disease progression. All patients ultimately discontinued blinded treatment (reasons shown in Supplementary Fig. S2); 63% and 87% of sunitinib and placebo patients, respectively, received open-label sunitinib. By study end, among the 255 patients who received open-label sunitinib, 246 had discontinued treatment, primarily due to disease progression (68%) or AEs (20%; Supplementary Fig. S2), the most common of which were abdominal pain, fatigue, and thrombocytopenia (each n = 4). Nine patients (4%) completed study treatment and enrolled in a continuation protocol.

Treatment was administered across the double-blind and open-label phases of the trial for a median of 22 weeks (range, 0.4–170) among the 241 patients who were randomized to the sunitinib arm and received blinded and/or open-label sunitinib (Table 2). Dosing interruptions and/or dose reductions in this group were required in 42% and 28% of patients, respectively.

Table 2.

Exposure to study drug

Double-blind phaseOpen-label phaseDouble-blind + open-label phases
Sunitiniba (n = 228)Placeboa (n = 114)All patientsb (n = 255)Sunitinib (n = 241)
Median no. of cycles started (range) 2 (1–15) 1 (1–6) 6 (1–37) 6 (1–44)
Median weeks on treatmentc (range) 12 (1–88) 6 (1–36) 34 (1–197) 34 (1–216)
Median weeks on drugd (range) 8 (0.4–57) 4 (1–24) 20 (1–140) 22 (0.4–170)
Dosing interruptions, n (%) 62 (27) 14 (12) 118 (46) 102 (42)
Reason for interruption, n (%)
AE 51 (82e10 (71e98 (83e86 (84e
Other 19 (31e4 (29e54 (46e47 (46e
Dose reductions, n (%) 26 (11) 0 (0) 72 (28) 67 (28)
Double-blind phaseOpen-label phaseDouble-blind + open-label phases
Sunitiniba (n = 228)Placeboa (n = 114)All patientsb (n = 255)Sunitinib (n = 241)
Median no. of cycles started (range) 2 (1–15) 1 (1–6) 6 (1–37) 6 (1–44)
Median weeks on treatmentc (range) 12 (1–88) 6 (1–36) 34 (1–197) 34 (1–216)
Median weeks on drugd (range) 8 (0.4–57) 4 (1–24) 20 (1–140) 22 (0.4–170)
Dosing interruptions, n (%) 62 (27) 14 (12) 118 (46) 102 (42)
Reason for interruption, n (%)
AE 51 (82e10 (71e98 (83e86 (84e
Other 19 (31e4 (29e54 (46e47 (46e
Dose reductions, n (%) 26 (11) 0 (0) 72 (28) 67 (28)

aPer-protocol population.

bTreatment = sunitinib, irrespective of randomization arm.

cFrom date of first dose to the earlier of either the termination date or 2 weeks after the last dose.

dTotal number of weeks on which study drug was actually administered.

ePercentage based on all patients who had dosing interruptions.

### Efficacy

Over the entire trial including open-label treatment, with median follow-up of 41.7 months [95% confidence interval (CI), 40.3–43.8], median OS for the sunitinib versus placebo arms was 72.7 weeks (95% CI, 61.3–83.0) versus 64.9 weeks (95% CI, 45.7–96.0), respectively (Fig. 1A, Table 3) as estimated using the Kaplan–Meier method (with 28% and 24% of patients alive, respectively). It is important to note that the “placebo arm” results in this analysis included survival data for patients who received open-label sunitinib [i.e., 103 of 118 patients randomized (87%); Supplementary Figs. S2 and S3], and this difference was not statistically significant (P = 0.306; HR, 0.876; 95% CI, 0.679–1.129; Table 3). To correct for the expected confounding impact on survival of cross-over of placebo-treated patients to open-label sunitinib, an exploratory analysis of OS was carried out using the RPSFT method, which calculated a median OS for the placebo arm of 39.0 weeks (95% CI, 28.0–54.1; Fig. 1A, Table 3). In this exploratory analysis, sunitinib nearly (1) doubled median OS and (2) halved the hazard of death versus placebo (HR, 0.505; 95% CI, 0.262–1.134; P = 0.306; Table 3).

Figure 1.

OS across the entire study: A, final analysis; B, final and interim analyses. KM, Kaplan–Meier method; RPSFT, rank-preserving structural failure time method.

Figure 1.

OS across the entire study: A, final analysis; B, final and interim analyses. KM, Kaplan–Meier method; RPSFT, rank-preserving structural failure time method.

Close modal
Table 3.

Efficacy results

Median (95% CI)
EndpointSunitinib (n = 243)Placebo (n = 118)HR (95% CI)P
TTP, wk 26.6 (16.0–32.1) 6.4 (4.4–10.0) 0.339 (0.244–0.472) <0.001
PFS, wk 22.9 (10.9–28.0) 6.0 (4.4–9.7) 0.347 (0.253–0.475) <0.001
Best overall response, n (%)
Partial response 16 (7)
Stable disease 128 (53) 50 (42)
Progressive disease 45 (19) 44 (37)
Missing/unevaluable 54 (22) 24 (20)
Overall ORR, % (range) 7 (4–11) 7 (3–10)a 0.004
OS, wk
Conventional methods 72.7 (61.3–83.0) 64.9 (45.7–96.0) 0.876 (0.679–1.129) 0.306
RPSFT method 72.7 (61.3–83.0) 39.0 (28.0–54.1) 0.505 (0.262–1.134)b 0.306c
Median (95% CI)
EndpointSunitinib (n = 243)Placebo (n = 118)HR (95% CI)P
TTP, wk 26.6 (16.0–32.1) 6.4 (4.4–10.0) 0.339 (0.244–0.472) <0.001
PFS, wk 22.9 (10.9–28.0) 6.0 (4.4–9.7) 0.347 (0.253–0.475) <0.001
Best overall response, n (%)
Partial response 16 (7)
Stable disease 128 (53) 50 (42)
Progressive disease 45 (19) 44 (37)
Missing/unevaluable 54 (22) 24 (20)
Overall ORR, % (range) 7 (4–11) 7 (3–10)a 0.004
OS, wk
Conventional methods 72.7 (61.3–83.0) 64.9 (45.7–96.0) 0.876 (0.679–1.129) 0.306
RPSFT method 72.7 (61.3–83.0) 39.0 (28.0–54.1) 0.505 (0.262–1.134)b 0.306c

aDifference between overall ORRs in sunitinib and placebo arms,% (95% CI of difference).

bEmpirical 95% CI was obtained using 100,000 bootstrap samples.

cThe RPSFT method does not alter the P value obtained using the conventional ITT method.

The median TTP among all patients in the final ITT population was 26.6 weeks (95% CI, 16.0–32.1) versus 6.4 weeks (95% CI: 4.4–10.0) in the sunitinib versus placebo arms, respectively (Table 3). Patients in the placebo arm exhibited an almost 3-fold greater risk of disease progression compared with those randomized to sunitinib treatment (HR, 0.339; 95% CI, 0.244–0.472; P ≤ 0.001). Similar results were obtained for PFS (Table 3). At final analysis, among the 99 placebo-treated patients who subsequently received open-label sunitinib treatment (Supplementary Fig. S2), median TTP was 10.4 weeks (95% CI, 4.3–22.0) on open-label sunitinib treatment.

Partial responses (PR) were observed in 16 sunitinib-treated patients (ORR, 7%) versus none in placebo-treated patients (Table 3). The ORR among placebo-treated patients who subsequently received open-label sunitinib treatment was 10%, comprising 9 PRs and 1 complete response.

### Safety

Safety results across the entire study are presented in Table 4. Results shown for the double-blind phase differ from the previously reported interim analysis (10) only by the inclusion of results for patients randomized subsequent to the interim analysis. Sunitinib-treated patients during the double-blind phase had relatively short-term exposure to the drug (median 8 weeks of drug; Table 2). In contrast, the 241 patients who were randomized to and received blinded and/or open-label sunitinib treatment represent the group with the greatest longitudinal exposure to sunitinib (median 22 weeks on drug). In this group, the most common treatment-related nonhematologic AEs were fatigue (47%), diarrhea (43%), and nausea (37%; Table 4). The most frequent treatment-related nonhematologic grade 3/4 AEs among these patients were fatigue (10%), hypertension (8%), and hand–foot syndrome, asthenia, and diarrhea (5% each). The frequencies of nonhematologic AEs in these patients who received extended sunitinib therapy were slightly higher than those reported in the sunitinib arm during the double-blind phase.

Table 4.

Nonhematologic AEs and hematologic laboratory abnormalities

n (%)
Double-blind phaseOpen-label phaseDouble-blind and open-label phases
Sunitiniba (n = 228)Placeboa (n = 114)All patientsb(n = 255)Sunitinib (n = 241)
AE or laboratory abnormality1/23/41/23/41/23/41/23/4
Treatment-related nonhematologic AEsd
Fatigue 67 (29) 18 (8) 23 (20) 2 (2) 102 (40) 26 (10) 88 (37) 25 (10)
Diarrhea 69 (30) 8 (4) 8 (7) 0 (0) 81(32) 12 (5) 91 (38) 13 (5)
Nausea 60 (26) 3 (1) 14 (12) 2 (2) 74 (29) 9 (4) 82 (34) 6 (3)
Skin discoloration 62 (27) 0 (0) 5 (4) 0 (0) 77 (30) 0 (0) 72 (30) 0 (0)
Anorexia 46 (20) 0 (0) 5 (4) 1 (1) 58 (23) 6 (2) 66 (27) 2 (1)
Dysgeusia 50 (22) 0 (0) 3 (3) 0 (0) 54 (21) 1 (0.4) 60 (25) 1 (0.4)
Vomiting 38 (17) 1 (0.4) 7 (6) 1 (1) 44 (17) 5 (2) 51 (21) 3 (1)
Mucosal inflammation 32 (14) 1 (0.4) 0 (0) 0 (0) 42 (17) 6 (2) 44 (18) 4 (2)
Hypertension 18 (8) 9 (4) 4 (4) 1 (1) 39 (15) 21 (8) 29 (12) 18 (8)
Rash 34 (15) 2 (1) 6 (5) 0 (0) 37 (15) 1 (0.4) 43 (18) 3 (1)
Stomatitis 35 (15) 1 (0.4) 1 (1) 0 (0) 36 (14) 4 (2) 42 (17) 3 (1)
Dyspepsia 33 (14) 1 (0.4) 1 (1) 0 (0) 45 (18) 2 (1) 44 (18) 1 (0.4)
Headache 24 (11) 1 (0.4) 7 (6) 0 (0) 43 (17) 3 (1) 41 (17) 3 (1)
Asthenia 25 (11) 6 (3) 2 (2) 2 (2) 31 (12) 12 (5) 30 (13) 13 (5)
Hand–foot syndrome 16 (7) 8 (4) 1 (1) 0 (0) 44 (17) 12 (5) 29 (12) 13 (5)
Hair color changes 19 (8) 0 (0) 2 (2) 0 (0) 50 (20) 0 (0) 37 (15) 0 (0)
Hypothyroidism 6 (3) 1 (0.4) 1 (1) 0 (0) 35 (14) 5 (2) 26 (11) 6 (2)
Cardiac disorders 22 (10) 2 (1) 2 (2) 0 (0) 14 (5) 3 (1) 25 (10) 3 (1)
Decreased EF 15 (7) 2 (1) 2 (2) 0 (0) 3 (1) 0 (0) 16 (7) 2 (1)
LV dysfunction 3 (1) 0 (0) 0 (0) 0 (0) 4 (2) 1 (0.4) 5 (2) 0 (0)
Hematologic laboratory abnormalities
Hemoglobin 124 (54) 9 (4) 60 (53) 2 (2) ND ND 132 (55) 11 (5)
Leukocytes 126 (55) 8 (4) 7 (6) 0 (0) ND ND 136 (56) 8 (3)
Neutrophils 105 (46) 23 (10) 4 (4) 0 (0) ND ND 113 (47) 28 (12)
Lymphocytes 85 (37) 26 (11) 34 (30) 2 (2) ND ND 58 (38) 28 (12)
Platelets 87 (38) 10 (4) 3 (3) 0 (0) ND ND 90 (37) 10 (4)
n (%)
Double-blind phaseOpen-label phaseDouble-blind and open-label phases
Sunitiniba (n = 228)Placeboa (n = 114)All patientsb(n = 255)Sunitinib (n = 241)
AE or laboratory abnormality1/23/41/23/41/23/41/23/4
Treatment-related nonhematologic AEsd
Fatigue 67 (29) 18 (8) 23 (20) 2 (2) 102 (40) 26 (10) 88 (37) 25 (10)
Diarrhea 69 (30) 8 (4) 8 (7) 0 (0) 81(32) 12 (5) 91 (38) 13 (5)
Nausea 60 (26) 3 (1) 14 (12) 2 (2) 74 (29) 9 (4) 82 (34) 6 (3)
Skin discoloration 62 (27) 0 (0) 5 (4) 0 (0) 77 (30) 0 (0) 72 (30) 0 (0)
Anorexia 46 (20) 0 (0) 5 (4) 1 (1) 58 (23) 6 (2) 66 (27) 2 (1)
Dysgeusia 50 (22) 0 (0) 3 (3) 0 (0) 54 (21) 1 (0.4) 60 (25) 1 (0.4)
Vomiting 38 (17) 1 (0.4) 7 (6) 1 (1) 44 (17) 5 (2) 51 (21) 3 (1)
Mucosal inflammation 32 (14) 1 (0.4) 0 (0) 0 (0) 42 (17) 6 (2) 44 (18) 4 (2)
Hypertension 18 (8) 9 (4) 4 (4) 1 (1) 39 (15) 21 (8) 29 (12) 18 (8)
Rash 34 (15) 2 (1) 6 (5) 0 (0) 37 (15) 1 (0.4) 43 (18) 3 (1)
Stomatitis 35 (15) 1 (0.4) 1 (1) 0 (0) 36 (14) 4 (2) 42 (17) 3 (1)
Dyspepsia 33 (14) 1 (0.4) 1 (1) 0 (0) 45 (18) 2 (1) 44 (18) 1 (0.4)
Headache 24 (11) 1 (0.4) 7 (6) 0 (0) 43 (17) 3 (1) 41 (17) 3 (1)
Asthenia 25 (11) 6 (3) 2 (2) 2 (2) 31 (12) 12 (5) 30 (13) 13 (5)
Hand–foot syndrome 16 (7) 8 (4) 1 (1) 0 (0) 44 (17) 12 (5) 29 (12) 13 (5)
Hair color changes 19 (8) 0 (0) 2 (2) 0 (0) 50 (20) 0 (0) 37 (15) 0 (0)
Hypothyroidism 6 (3) 1 (0.4) 1 (1) 0 (0) 35 (14) 5 (2) 26 (11) 6 (2)
Cardiac disorders 22 (10) 2 (1) 2 (2) 0 (0) 14 (5) 3 (1) 25 (10) 3 (1)
Decreased EF 15 (7) 2 (1) 2 (2) 0 (0) 3 (1) 0 (0) 16 (7) 2 (1)
LV dysfunction 3 (1) 0 (0) 0 (0) 0 (0) 4 (2) 1 (0.4) 5 (2) 0 (0)
Hematologic laboratory abnormalities
Hemoglobin 124 (54) 9 (4) 60 (53) 2 (2) ND ND 132 (55) 11 (5)
Leukocytes 126 (55) 8 (4) 7 (6) 0 (0) ND ND 136 (56) 8 (3)
Neutrophils 105 (46) 23 (10) 4 (4) 0 (0) ND ND 113 (47) 28 (12)
Lymphocytes 85 (37) 26 (11) 34 (30) 2 (2) ND ND 58 (38) 28 (12)
Platelets 87 (38) 10 (4) 3 (3) 0 (0) ND ND 90 (37) 10 (4)

Abbreviations: EF, ejection fraction; LV, left ventricular; ND, not determined.

aPer-protocol population.

bTreatment = sunitinib.

cGrade 5 events are described in the text.

dReported in 15% or more of patients based on the sunitinib arm over the entire study or of particular clinical interest.

The frequency of treatment-related hypertension (any grade) on sunitinib increased from 12% during blinded treatment to 20% across the entire study. In contrast, the frequency of treatment-related cardiac AEs overall was 11% during the double-blind phase and 12% over the entire study [comprising decreased ejection fraction (8%) and left ventricular dysfunction (2%) as the most frequent events]. Over the course of the study, one patient experienced congestive heart failure (grade 2) that was considered related to sunitinib treatment; this event resolved during a dosing delay with appropriate supportive therapy. The rate of treatment-related hypothyroidism (all grades) increased with extended sunitinib treatment from 3% during the double-blind phase to 13% for the entire trial. Hypothyroidism did not result in any treatment discontinuations.

Hematologic laboratory abnormalities among patients who received sunitinib across the entire study included reduced levels of hemoglobin (60%), neutrophils (59%), and platelets (41%). Most of these abnormalities were grade 1/2 and were similar in frequency to those seen during blinded sunitinib treatment.

During blinded treatment, 4 treatment-related deaths were reported in the sunitinib arm (2%; cardiac arrest, cerebral ischemia, left ventricular failure, and multiorgan failure) and 2 (2%; cardiac arrest, gastrointestinal hemorrhage) in the placebo arm. In addition, 4 treatment-related deaths were reported during open-label sunitinib treatment or follow-up [hepatic encephalopathy, hepatic failure, melena, and pneumonia (as coded from the Medical Dictionary for Regulatory Activities)].

### Circulating biomarkers of angiogenesis

Levels of plasma VEGF-A, sVEGFR-2, and sVEGFR-3 were measured on days 1 and 28 of cycles 1 to 3 (i.e., after 2 weeks off-treatment and after 4 weeks on-treatment, respectively) and on cycle 1 day 14 (C1;D14) in a subset of patients in both study arms. Overall, this patient subset was similar to the ITT population with regard to baseline demographic and prior imatinib treatment characteristics, with the 2 treatment arms generally well balanced for these parameters (Table 1). Exceptions were higher percentages of patients with metastases in the liver or peritoneum in both treatment arms and a higher proportion of men in the placebo arm of the biomarker subset compared with the ITT population.

Changes from baseline in plasma levels of VEGF-A, sVEGFR-2, and sVEGFR-3 correlated with sunitinib treatment on this intermittent dosing schedule (Supplementary Fig. S4), as has been reported in other sunitinib studies (13–15). In the placebo arm, there were no significant changes in the levels of these proteins at any time point. Between-treatment differences were significant (P < 0.001) at all time points tested after treatment initiation (except for VEGF-A on C2;D1 and C3;D1).

Possible associations between levels of VEGF-A, sVEGFR-2, or sVEGFR-3 and clinical outcomes (TTP or OS) were also evaluated at these time points (Supplementary Tables S1 and S2). Significant associations are shown in Fig. 2. In the sunitinib arm, low baseline levels (below median) of sVEGFR-2 and reduced levels on C1;D14 relative to baseline were associated with significantly prolonged TTP and OS (Fig. 2A–D). Reductions in sVEGFR-3 on C1;D14 compared with baseline were also associated with longer TTP (Fig. 2E). In the placebo arm, reductions in VEGF-A relative to baseline were significantly associated with longer TTP for both C1;D28 (Fig. 2F) and C2;D1 (Supplementary Table S1). No other significant associations between plasma VEGF-A, sVEGFR-2, or sVEGFR-3 levels and TTP or OS through C3;D28 were found.

Figure 2.

Significant associations found between levels of circulating biomarkers of angiogenesis and TTP or OS. A–E, sunitinib arm. F, placebo arm. Median values used for stratification: A and B (baseline sVEGFR-2), 9,159 pg/mL; C and D (sVEGFR-2 ratio to baseline, C1;D14), 0.727; E (sVEGFR-3 ratio to baseline, C1;D14), 0.694; F (VEGF-A ratio to baseline, C1;D28), 0.957.

Figure 2.

Significant associations found between levels of circulating biomarkers of angiogenesis and TTP or OS. A–E, sunitinib arm. F, placebo arm. Median values used for stratification: A and B (baseline sVEGFR-2), 9,159 pg/mL; C and D (sVEGFR-2 ratio to baseline, C1;D14), 0.727; E (sVEGFR-3 ratio to baseline, C1;D14), 0.694; F (VEGF-A ratio to baseline, C1;D28), 0.957.

Close modal

The predictive value of the ratio of each plasma protein biomarker to baseline at each time point analyzed and other baseline patient characteristics (age, sex, ECOG PS, time since diagnosis, maximum daily dose of prior imatinib treatment, and baseline tumor size) was assessed in uni- and multivariate analyses of the 2 treatment arms separately as well as in the entire population. Several parameters were found to have significant associations with TTP or OS in univariate analyses in one arm or the other. However, in multivariate analysis, significant associations were only found between OS in the sunitinib arm and baseline tumor size, baseline sVEGFR-2, and sVEGFR-2 C1;D14 ratio to baseline (Supplementary Table S3).

These longer term analyses from this large randomized, placebo-controlled, multicenter study provide new insights into the significant clinical activity of sunitinib in imatinib-resistant/intolerant GIST patients. The trial was designed with TTP as the primary endpoint, with the expectation that the majority of patients assigned to placebo would subsequently cross over to open-label sunitinib. In an earlier analysis that reflected the double-blind phase of the trial, sunitinib resulted in significantly longer OS versus placebo (HR, 0.49; 95% CI, 0.29–0.83; P = 0.007; ref. 10). In the final analysis following all cross-overs, Kaplan–Meier estimates of OS for the 2 treatment arms converged, with 87% of patients randomized to placebo successfully able to receive sunitinib following progressive disease or unblinding. This cross-over study was not powered to show an OS difference, as cross-over patients were still regarded as part of the ITT placebo arm for the purpose of survival analysis, as the comparison was based on the randomization. Exploratory analysis using the RPSFT method suggested that sunitinib confers a long-term OS benefit relative to placebo. With this method, the estimated median OS for patients in the sunitinib arm was almost twice that of patients in the placebo arm (72.7 vs. 39.0 weeks). Using RPSFT analysis to “correct for cross-over,” this analysis estimated an HR (0.505; 95% CI, 0.262–1.134) that was further from the null than that obtained using conventional statistical methods (0.876; 95% CI, 0.679–1.129), as expected given that cross-over would reduce the treatment effect size between the 2 arms. The data obtained by this method were highly consistent with those obtained for blinded placebo treatment in the interim analysis using conventional methods (Fig. 1B), supporting the validity of this method. RPSFT analysis also avoids potential biases introduced through subgroup analyses, as it is based on the ITT population.

Overall, the patients who crossed over from placebo to open-label sunitinib showed benefits that were only somewhat more limited than those of patients initially randomized to blinded sunitinib treatment, validating the placebo-controlled cross-over design of the study. However, there was some indication that delaying initiation of sunitinib treatment compared unfavorably to earlier dosing with sunitinib, as median TTP after cross-over in the placebo arm was 10.4 weeks.

Sunitinib exhibited a safety profile consistent with prior reports over the extended course of this study, and AEs were generally manageable and reversible using standard medical therapy with or without dosing interruption and/or dose reduction. Without having specifically analyzed the timing or duration of AEs, the results showed that although the frequency of nonhematologic AEs increased slightly with long-term sunitinib treatment, there were no major changes in the overall safety profile. Fatigue remained the most commonly reported AE (Table 4), although a large proportion of fatigue in this population may be attributable to the burden of advanced GIST given the frequency of this AE reported in the placebo arm during the double-blind phase. As with other angiogenesis inhibitors (20), the frequency of hypertension increased with extended sunitinib treatment, whereas the incidence of other cardiac AEs did not increase. Indeed, the most frequent of these (“decreased ejection fraction”) seemed to occur primarily during the double-blind phase of the trial (8% vs. 1% during open-label treatment, with the caveat that although regularly scheduled periodic assessments of cardiac function were only mandated during the double-blind phase of the study, such assessments were conducted if clinically indicated to monitor patients on this trial at all times). As expected with sunitinib (9, 21), the incidence of hypothyroidism increased with time on treatment from 3% during the double-blind phase to 13% for the entire trial (and thyroid dysfunction may have been underreported because thyroid stimulating hormone levels were not required to be monitored prospectively). However, most cases were mild to moderate in severity (11%) and easily managed using thyroid hormone replacement therapy.

A secondary objective of this trial was to examine potential correlations between clinical outcomes and plasma biomarkers of angiogenesis that are related to known molecular targets of sunitinib. The plasma pharmacodynamics of VEGF-A, sVEGFR-2, and sVEGFR-3 during 4-week-on/2-week-off sunitinib dosing were similar to those previously reported for these proteins in renal cell carcinoma, hepatocellular carcinoma, and breast cancer (13–15). In the present study, no consistent associations were observed between changes in plasma levels of these angiogenesis-related proteins and either TTP or OS. These findings contrast markedly with the strong associations obtained at several time points between reductions in plasma levels of sKIT and improved clinical outcome that have been reported previously in this study (11) and in a phase II GIST study in which sunitinib was administered on a continuous daily dosing schedule (22). Taken together, these findings for circulating proteins related to molecular targets of sunitinib support the view that inhibition of activated KIT represents the primary mechanism for the clinical activity of this multitargeted tyrosine kinase inhibitor in GIST (23).

In conclusion, these efficacy and safety results from the complete longitudinal data of this worldwide phase III study further support prior reports that sunitinib provides significant clinical benefit to GIST patients whose disease is resistant to, or who are intolerant of, imatinib therapy.

The authors thank all the participating patients and their families, as well as the global network of investigators (listed in Supplementary Data), research nurses, study coordinators, and operations staff. C.S. Harmon was an employee of Atrium Staffing (New York, NY) and was a paid contractor of Pfizer in the development of this article and the analysis and interpretation of data involving circulating biomarkers of angiogenesis. Medical writing support was provided by Wendy Sacks at ACUMED (New York, NY) and was funded by Pfizer Inc.

This study was sponsored by Pfizer Inc. (New York, NY). Additional funding was provided by the Virginia and Daniel K. Ludwig Trust for Cancer Research, the Rubenstein Foundation, the Katz Foundation, the Quick Family Fund for Cancer Research, the Ronald O. Perelman Fund for Cancer Research at Dana-Farber, the Stutman GIST Cancer Research Fund, Paul's Posse, and Leslie's Links (G.D. Demetri); and grant no. FIS PI 081156 from the Spanish National Health Service (L. Paz-Ares).

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.

1.
Corless
CL
,
Heinrich
MC
.
Molecular pathobiology of gastrointestinal stromal sarcomas
.
Annu Rev Pathol
2008
;
3
:
557
86
.
2.
Heinrich
MC
,
Corless
CL
,
Duensing
A
,
McGreevey
L
,
Chen
CJ
,
Joseph
N
, et al
PDGFRA activating mutations in gastrointestinal stromal tumors
.
Science
2003
;
299
:
708
10
.
3.
Hirota
S
,
Isozaki
K
,
Moriyama
Y
,
Hashimoto
K
,
Nishida
T
,
Ishiguro
S
, et al
Gain-of-function mutations of c-kit in human gastrointestinal stromal tumors
.
Science
1998
;
279
:
577
80
.
4.
Abrams
TJ
,
Lee
LB
,
Murray
LJ
,
Pryer
NK
,
Cherrington
JM
.
SU11248 inhibits KIT and platelet-derived growth factor receptor beta in preclinical models of human small cell lung cancer
.
Mol Cancer Ther
2003
;
2
:
471
8
.
5.
Kim
DW
,
Jo
YS
,
Jung
HS
,
Chung
HK
,
Song
JH
,
Park
KC
, et al
An orally administered multitarget tyrosine kinase inhibitor, SU11248, is a novel potent inhibitor of thyroid oncogenic RET/papillary thyroid cancer kinases
.
J Clin Endocrinol Metab
2006
;
91
:
4070
6
.
6.
Mendel
DB
,
Laird
,
Xin
X
,
Louie
SG
,
Christensen
JG
,
Li
G
, et al
In vivo antitumor activity of SU11248, a novel tyrosine kinase inhibitor targeting vascular endothelial growth factor and platelet-derived growth factor receptors: determination of a pharmacokinetic/pharmacodynamic relationship
.
Clin Cancer Res
2003
;
9
:
327
37
.
7.
Murray
LJ
,
Abrams
TJ
,
Long
KR
,
Ngai
TJ
,
Olson
LM
,
Hong
W
, et al
SU11248 inhibits tumor growth and CSF-1R-dependent osteolysis in an experimental breast cancer bone metastasis model
.
Clin Exp Metastasis
2003
;
20
:
757
66
.
8.
O'Farrell
A-M
,
Abrams
TJ
,
Yuen
HA
,
Ngai
TJ
,
Louie
SG
,
Yee
KW
, et al
SU11248 is a novel FLT3 tyrosine kinase inhibitor with potent activity in vitro and in vivo
.
Blood
2003
;
101
:
3597
605
.
9.
Pfizer: SUTENT (sunitinib malate) prescribing information
[cited 2012 Apr 20]. Available from
: http://www.pfizer.com/files/products/uspi_sutent.pdf
10.
Demetri
GD
,
van Oosterom
AT
,
Garrett
CR
,
Blackstein
ME
,
Shah
MH
,
Verweij
J
, et al
Efficacy and safety of sunitinib in patients with advanced gastrointestinal stromal tumour after failure of imatinib: a randomised controlled trial
.
Lancet
2006
;
368
:
1329
38
.
11.
DePrimo
SE
,
Huang
X
,
Blackstein
ME
,
Garrett
CR
,
Harmon
CS
,
Schöffski
P
, et al
Circulating levels of soluble KIT serve as a biomarker for clinical outcome in gastrointestinal stromal tumor patients receiving sunitinib following imatinib failure
.
Clin Cancer Res
2009
;
15
:
5869
77
.
12.
Therasse
P
,
Arbuck
SG
,
Eisenhauer
EA
,
Wanders
J
,
Kaplan
RS
,
Rubinstein
L
, et al
New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada
.
J Natl Cancer Inst
2000
;
92
:
205
16
.
13.
DePrimo
SE
,
Bello
CL
,
Smeraglia
J
,
Baum
CM
,
Spinella
D
,
Rini
BI
, et al
Circulating protein biomarkers of pharmacodynamic activity of sunitinib in patients with metastatic renal cell carcinoma: modulation of VEGF and VEGF-related proteins
.
J Transl Med
2007
;
5
:
32
.
14.
Harmon
CS
,
DePrimo
SE
,
Raymond
E
,
Cheng
AL
,
Boucher
E
,
Douillard
JY
, et al
Mechanism-related circulating proteins as biomarkers for clinical outcome in patients with unresectable hepatocellular carcinoma receiving sunitinib
.
J Transl Med
2011
;
9
:
120
.
15.
Burstein
HJ
,
Elias
,
Rugo
HS
,
Cobleigh
MA
,
Wolff
AC
,
Eisenberg
PD
, et al
Phase II study of sunitinib malate, an oral multitargeted tyrosine kinase inhibitor, in patients with metastatic breast cancer previously treated with an anthracycline and a taxane
.
J Clin Oncol
2008
;
26
:
1810
16
.
16.
DeSilva
B
,
Smith
W
,
Weiner
R
,
Kelley
M
,
Smolec
J
,
Lee
B
, et al
Recommendations for the bioanalytical method validation of ligand-binding assays to support pharmacokinetic assessments of macromolecules
.
Pharm Res
2003
;
20
:
1885
900
.
17.
Robins
JM
,
Tsiatis
A
.
Correcting for non-compliance in randomized trials using rank-preserving structural failure time models
.
Commun Stat Theory Methods
1991
;
20
:
2609
31
.
18.
White
IR
.
Uses and limitations of randomization-based efficacy estimators
.
Stat Methods Med Res
2005
;
14
:
327
47
.
19.
White
IR
.
Estimating treatment effects in randomised trials with treatment switching
.
Stat Med
2006
;
25
:
1619
22
.
20.
Sica
DA
.
Angiogenesis inhibitors and hypertension: an emerging issue
.
J Clin Oncol
2006
;
24
:
1329
31
.
21.
Desai
J
,
Yassa
L
,
Marqusee
E
,
George
S
,
Frates
MC
,
Chen
MH
, et al
Hypothyroidism after sunitinib treatment for patients with gastrointestinal stromal tumors
.
Ann Intern Med
2006
;
145
:
660
64
.
22.
George
S
,
Blay
JY
,
Casali
PG
,
Le Cesne
A
,
Stephenson
P
,
Deprimo
SE
, et al
Clinical evaluation of continuous daily dosing of sunitinib malate in patients with advanced gastrointestinal stromal tumour after imatinib failure
.
Eur J Cancer
2009
;
45
:
1959
68
.
23.
Heinrich
MC
,
Maki
RG
,
Corless
CL
,
Antonescu
CR
,
Harlow
A
,
Griffith
D
, et al
Primary and secondary kinase genotypes correlate with the biological and clinical activity of sunitinib in imatinib-resistant gastrointestinal stromal tumor
.
J Clin Oncol
2008
;
26
:
5352
9
.