BATON-CRC: A Phase II Randomized Trial Comparing Tivozanib Plus mFOLFOX6 with Bevacizumab Plus mFOLFOX6 in Stage IV Metastatic Colorectal Cancer

Colorectal cancer is a prevalent and deadly cancer with an estimated 136,830 new cases and 50,310 deaths in the United States in 2014 (1). Although 5-year survival rates remain low (9.8% to 17.7% across major racial/ethnic population and age groups), survival time for metastatic CRC (mCRC) has been improving (2). Additional therapeutic options including combination therapy and therapies with agents targeting the EGFR and VEGFR signaling pathways have created a significant change in colorectal cancer therapy over the last decade, particularly for mCRC (3, 4). In particular, the addition of bevacizumab has become a standard in combination therapy for mCRC (5).

Tivozanib is an oral VEGFR tyrosine kinase inhibitor (TKI) with selective inhibition of VEGFR-1, -2, and -3 (6). It is more selective and potent than other VEGFR TKIs and was designed to optimize blockade of VEGFR while minimizing off-target toxicities (6–8). Tivozanib has a long half-life of approximately 4.5 days, which allows once-daily dosing to maintain serum concentrations (7, 9).

In a phase III trial in patients with renal cell carcinoma, tivozanib met its primary endpoint of progression-free survival (PFS), but failed to meet the overall survival (OS) secondary endpoint (10). One possibility is that crossover to sorafenib may have confounded the survival results (11). Tivozanib has also shown single-agent activity in patients with other solid tumors, including colorectal cancer (7, 12–14) and in combination with temsirolimus in patients with renal cell carcinoma (15), with paclitaxel in patients with metastatic breast cancer (16), and with everolimus in patients with metastatic colon cancer (17). Preclinical data demonstrated additive antitumor activity of tivozanib when combined with fluorouracil. A pharmacokinetic drug–drug interaction study showed that dosing of tivozanib with a CYP3A4 pathway inhibitor had no influence on tivozanib concentrations; however, a strong CYP3A4 inducer reduced tivozanib plasma concentration (9).

Tivozanib plus mFOLFOX6 (modified treatment regimen of: FOLinic acid, Fluorouracil, and OXaliplatin) was evaluated in a phase Ib study in patients with advanced gastrointestinal cancers, including mCRC (18). Results from this study showed that tivozanib can be safely combined at the recommended dose with mFOLFOX6 in patients with advanced gastrointestinal tumors and also demonstrated encouraging antineoplastic activity with 30.8% of all patients achieving a partial response (18). A randomized, open-label, phase II trial of tivozanib/mFOLFOX6 (Arm A) versus bevacizumab/mFOLFOX6 (Arm B) in patients with previously untreated mCRC [BATON (Biomarker Assessment of Tivozanib in ONcology) -CRC; NCT01478594] was initiated to evaluate tivozanib activity versus bevacizumab and to evaluate a predefined set of biomarkers for potential prediction of clinical response following tivozanib therapy.

Male or female patients ≥18 years with histologically or cytologically confirmed mCRC, ≥1 measurable lesions by RECIST (Version1.1), no prior systemic chemotherapy, no fluorouracil-containing adjuvant therapy in the previous 6 months, and an Eastern Cooperative Oncology Group performance status (ECOG PS) ≤1 were included in the study. Patients were excluded if they had prior VEGF therapy (including bevacizumab) or a history of significant thromboembolic or vascular disorders within 6 months of study entry. Patients with any of the following were also excluded: hematologic abnormalities; significant cardiovascular disease; history of serious gastrointestinal toxicity, diarrhea, or stomatitis within 6 weeks of study entry; a gastrointestinal condition with increased risk of perforation; history of abdominal fistula, gastrointestinal perforation, or intra-abdominal abscess within 4 weeks prior to study entry; and peripheral neuropathy ≥grade 2.

All patients provided informed consent prior to performing any study-related procedures. Approval was obtained by the research ethics committee at each participating institution for the protocol, informed consent, patient information, and amendments. The study was conducted in accordance with the International Conference on Harmonization Good Clinical Practice Guideline with the ethical principles of the current version of Declaration of Helsinki.

This was a randomized, open-label, phase II trial designed to evaluate superiority of tivozanib/mFOLFOX6 over bevacizumab/mFOLFOX6 in patients with mCRC (ClinicalTrials.gov registration: NCT01478594). Patients were randomized 2:1 to tivozanib/mFOLFOX6 (Arm A) or bevacizumab/mFOLFOX6 (Arm B) and stratified by lactate dehydrogenase (LDH) status (<1.5 ULN or >1.5 ULN), origin of cancer (rectal or colon), and number of metastatic sites (1 or ≥2). Patients received mFOLFOX6 every 2 weeks on days 1 and 15 of each 28-day cycle with either tivozanib orally 1.5 mg once daily for 21 days followed by 7 days off treatment (Arm A) or bevacizumab intravenously 5 mg/kg every 2 weeks on days 1 and 15 (Arm B).

PFS by investigator-assessed radiologic disease progression was the primary endpoint. Assessed secondary endpoints included safety, OS, overall response rate (ORR), duration of response (DOR), and time to treatment failure (TTF). PFS subgroup analyses and biomarker analyses of LDH; VEGF A, C, D, and their ratios; CD68; myeloid-derived gene signature; and serum soluble cytokines were evaluated for any potential to predict response to tivozanib therapy. For this study, “predictive” was referred to as a marker status above a clinically derived cut-off point that implies future therapeutic efficacy, and “prognostic” refers to serving to predict the likely outcome of a disease or ailment; of or relating to the status of a medical condition. The trial initiated on December 20, 2011 and interim efficacy analyses were based on a September 13, 2013 data analysis cutoff, which led to the termination of the study due to futility. Secondary endpoints of PFS by independent radiologic review and health-related quality of life were not assessed due to early termination of the study. Disposition, safety, and biomarker analyses include data through trial follow-up to February 28, 2014.

Safety and tolerability was assessed by monitoring vital sign measurements, physical examinations, 12-lead electrocardiograms (ECG), clinical laboratory tests (including hematology, urinalysis, serum chemistry, thyroid, coagulation, and pregnancy tests), cardiac biomarkers (troponin, creatine phosphokinase), ECOG PS, and adverse events (AE) which were coded using NCI-CTCAE (v4.03). Dose reduction for toxicity was allowed for tivozanib and mFOLFOX6, but not for bevacizumab. Dose interruptions for all agents were allowed; however, interruption for longer than 2 weeks could result in treatment discontinuation as determined by the study investigator. Dose reductions of tivozanib and mFOLFOX6 could have occurred independently. Tivozanib was reduced to 1.0 mg/day for patients with grade ≥3 AEs, and could not be re-escalated.

Disease progression was assessed on the basis of radiologic imaging (e.g., diagnostic computed tomography scans, magnetic resonance imaging scans) of the chest, abdomen, and pelvis every 8 weeks from the start of study medication for the first 18 months on study and then every 12 weeks until study end, disease progression, or death. PFS was defined as the time from the date of randomization until objective tumor progression (defined by RECIST v1.1) or death. Secondary efficacy variables were defined as follows: OS from the date of randomization until death from any cause, ORR was the proportion of patients with confirmed complete response (CR) or partial response (PR), DOR from the date of first documented CR or PR to the date of progression, and TTF from randomization to treatment discontinuation for any reason.

A secondary objective to identify biomarkers that potentially could predict clinical response following tivozanib/mFOLFOX6 therapy or preferential response versus bevacizumab/mFOLFOX6, included analysis of VEGF A, C, D, and their ratios; CD68; myeloid-derived gene signature (used for tumor values of VEGF A, C, D, and PIGF); serum soluble cytokines; and neuropilin-1 (NRP-1). Serum protein measurement for the biomarker analyses were measured using the Myriad Rules-Based Medicine system (Myriad RBM). These multiplexed assays are based on the capture-sandwich format using capture antibodies attached to fluorescently encoded microspheres. After capture of antigen from a biological sample, such as serum, the antigen is then detected using specific detection antibodies coupled to a fluorescent probe. Biomarker measurement was categorized as below or above the median. All biomarkers were analyzed via RBM and were analyzed only once. No longitudinal testing was performed in the study. Please see the Appendix to this article for a table containing the 7 biomarkers tested and the ranges for each biomarker, as well as a detailed description on the methods of sample collection and testing.

Approximately 252 subjects were to be randomized in this study. Randomized subjects were assigned to receive tivozanib in combination with mFOLFOX6 or bevacizumab in combination with mFOLFOX6 using a 2:1 randomization scheme. A total of 166 PFS events were planned to provide 80% power at a one-sided type I error of 0.2 to detect a statistically significant treatment effect of tivozanib/mFOLFOX6 over bevacizumab/mFOLFOX6 using a log-rank test assuming a median PFS for bevacizumab/mFOLFOX6 of 9.4 months (19) and a 3-month improvement for tivozanib/mFOLFOX6. An interim futility analysis for superiority was planned for when approximately 83 PFS events (50% of the total) were observed with a futility boundary of HR > 1.0581. The actual number of PFS events at the time of the interim analysis (95, 57.2% of the required 166 events) changed the futility stopping boundary to an HR > 1.004 according to the prespecified Lans-DeMets beta spending function for O'Brien-Fleming boundary. The full analysis set was defined as all randomized patients and was used for primary analyses of efficacy as well as demographic and baseline characteristics. The safety analysis set consisted of all patients who received ≥1 dose of study drug. The distribution of the primary endpoint, PFS, was estimated for each treatment arm using the Kaplan–Meier method and compared between arms using a stratified log-rank test. A Cox proportional hazard model along with the 95% confidence interval (CI) was used to assess PFS, OS, TTF, and DOR as well as the association between potential serum biomarkers and PFS with treatment arms as the covariate. ORR and corresponding CIs were calculated for each treatment arm and compared between 2 arms using a stratified Cochran–Mantel–Haenszel test. All data processing, summarization, and analyses were performed using SAS Version 9.2 or above in a UNIX environment.

Exploratory Biomarker analysis was prespecified before the interim analysis was performed. The objective of the biomarker analysis was to identify potential predictive biomarkers (whether tivozanib treatment effect over bevacizumab is better in one biomarker subgroup than the other) as well as potential prognostic biomarkers (within each treatment arm, one biomarker subgroup had better efficacy outcome than the other). All biomarkers were measured at baseline and were dichotomized by their respective sample median values (≥median, <median). In each biomarker subgroup, the HR, along with a 95% CI, were calculated and compared using an unadjusted Cox proportional hazard model utilizing the treatment arms as covariates. No adjustment was made for testing multiple hypotheses, as these are exploratory comparisons.

The study was conducted at 73 sites in 12 countries including Australia (8), Austria (3), Belgium (4), Canada (5), Czech Republic (2), Finland (2), Hungary (4), Italy (4), the Netherlands (1), Spain (6), United Kingdom (7), and the United States (27). From December 2, 2011 through April 28, 2013, 265 subjects were randomized, 177 to tivozanib/mFOLFOX6 (Arm A) and 88 to bevacizumab/mFOLFOX6 (Arm B) (Supplementary Fig. S1). In Arm A, all 177 patients were included in both the full and safety analysis set and 158 patients (89%) discontinued study treatment [61 patients (38.6%) primarily because of an AE and 55 (34.8%) because of progressive disease]. In Arm B, 88 patients were included in the full analysis set, but one patient did not receive study drug and was not included in the safety analysis set; 81 patients (92%) discontinued study treatment [28 patients (34.6%) primarily due to an AE and 34 (42.0%) due to progressive disease]. Median dose intensity for the combination dose was similar between the 2 arms at 83% in Arm A and 86% in Arm B. Likewise, median duration of treatment was similar (168.0 days Arm A and 162.0 days Arm B). In Arm A, 41 patients (23.2%) required dose reduction in tivozanib while dose reduction in bevacizumab was not allowed in Arm B. Patient baseline demographic and disease characteristics are presented in Table 1 and are generally balanced between the two treatment groups except that Arm A had a higher proportion of patients with an ECOG PS of 1 (46.3 vs. 34.1%).

The overall safety profile was comparable between treatment arms, although some differences were noted. In both treatment arms, the most common all grade treatment-emergent adverse event (TEAE) was diarrhea (58.2% Arm A and 57.5% Arm B) and the most common grade 3/4 TEAE was neutropenia (39.5% Arm A and 24.1% Arm B; Table 2). A slightly higher proportion of patients in Arm A experienced TEAEs of neutropenia (39.4% vs. 24.1%), hypertension (16.4% vs. 10.3%), vomiting (5.6% vs. 1.1%), and thrombocytopenia (5.6% vs. 2.3%). The most common treatment-related AE in both arms was hypertension (39.5% Arm A and 25.3% Arm B). An AE led to permanent discontinuation in 41.2% of patients in Arm A and 34.5% of patients in Arm B (although not necessarily the primary reason for discontinuation). Pulmonary embolism and deep vein thrombosis were the most common AEs leading to discontinuation for Arm A and Arm B, respectively.

Serious AEs (SAE) were reported for 46.3% of patients in Arm A compared with 48.3% in Arm B. The two most common SAEs in Arm A were diarrhea (4.0% Arm A, 5.7% Arm B) and pulmonary embolism (4.0% Arm A, 0% Arm B), and in Arm B were pyrexia (2.3% Arm A, 8.0% Arm B) and diarrhea (as reported above). Serious treatment-related AEs were reported in 21.5% of patients for tivozanib (most common being pulmonary embolism at 2.8%) and in 17.2% patients for bevacizumab (most commonly abdominal pain at 3.4%).

A total of 9 patients died while on treatment or within 30 days of last dose. There were 7 deaths (4.0%) in Arm A with 3 patients having at least one fatal AE considered to be probably related to tivozanib (pulmonary hemorrhage, gastrointestinal hemorrhage, and duodenal neoplasm) or possibly related (asthenia). There were 2 deaths (2.3%) in Arm B patients, both of which were due to AEs probably related to bevacizumab (hepatic hemorrhage and large intestine perforation).

A prespecified interim analysis included 95 PFS events and met the prespecified futility criteria. Median PFS (Arm A vs. Arm B) was 9.4 vs. 10.7 months (HR = 1.091; CI, 0.693–1.718; P = 0.706; Table 3; Fig. 1). Complete responses were achieved for 4 patients (2.3%) in Arm A and 1 patient (1.1%) in Arm B and ORR (Arm A vs. Arm B) was 45.2% versus 43.2% (P = 0.718; Table 3). In addition, Arm A versus Arm B median DOR was 7.4 versus 9.3 months (HR = 1.389; CI, 0.604–3.194; P = 0.437), median TTF was 5.5 versus 5.4 months (HR = 1.006; CI, 0.764–1.358; P = 0.967), and median OS was not reached for either arm. Because the data from the prespecified interim analysis demonstrated futility for superiority, the study was discontinued. Post hoc analysis of final data of 91 and 50 events for Arm A and Arm B, respectively, showed a median PFS of 9.8 versus 9.5 months (HR = 0.908; CI, 0.624–1.301; P = 0.598) and an ORR of 46.9% in Arm A versus 43.2% in Arm B. Subgroup analysis did not show significant differences in PFS (Fig. 2A). A trend of benefit was observed in the predefined LDH high [>1.5 upper limit of normal (ULN)] subgroup.

A secondary objective of the study was to evaluate the value of potential predefined biomarkers for preferential response of tivozanib/mFOLFOX6 versus bevacizumab/mFOLFOX6. None of the biomarker analyses showed significant differences at the time of the interim analysis using a data cutoff of September 13, 2013. Subsequent post hoc exploratory biomarker analyses were conducted on 162 (61%) patient samples based on final efficacy data collected through February 28, 2014. With 65% (62) total PFS events included, the PFS HR along with the 95% CI is presented for each biomarker subgroup (Fig. 2B). Baseline characteristics for this biomarker population are presented in Supplementary Table S1. Most of the characteristics analyzed in the intent-to-treat (ITT) population were analyzed in this population and were very similar to the ITT population.

NRP-1, the biomarker found to be most associated with increased PFS, was analyzed further to determine whether it had potential prognostic value for tivozanib and bevacizumab and potential predictive value for tivozanib over bevacizumab in the low NRP-1–selected population. In both treatment arms, there was an increase in PFS in patients with NRP-1 below the median (NRP-1 low; Fig. 3A and B), consistent with other studies (20, 21) and suggesting a potential prognostic biomarker. More importantly, there is a statistically significant difference in PFS in patients with low NRP-1 in favor of Arm A with no difference in high NRP-1 (above median NRP-1), suggesting that NRP-1 could be a potential predictive marker as well as a prognostic marker (Fig. 3C and D); for NRP-1–low patients, PFS was 17.9 months in the tivozanib-treated patients and 11.2 months in the bevacizumab-treated patients (HR = 0.38; unstratified P value 0.0075); for NRP-1–high patients, PFS was 7.3 months in the tivozanib-treated patients, and 7.5 months in the bevacizumab-treated patients (HR = 1.00).

In this randomized, open-label, phase II trial of tivozanib/mFOLFOX6 versus bevacizumab/mFOLFOX6 in patients with previously untreated mCRC, a prespecified interim futility analysis demonstrated that tivozanib was not superior to bevacizumab for PFS in the ITT population, resulting in discontinuation of the study. At this interim analysis, the combination of tivozanib/mFOLFOX6 resulted in PFS, ORR, TTF, and DOR comparable with that of bevacizumab/mFOLFOX6. Also at this interim analysis, subgroup and biomarker analyses revealed no significant association with PFS, although LDH-1 and NRP-1 showed a trend of PFS benefit. At the post hoc final analysis, NRP-1 low (below the median) did demonstrate a statistical difference in PFS favoring the tivozanib/mFOLFOX6 arm. The safety profile of the combination of tivozanib/mFOLFOX6 was consistent with the safety profile reported to date in other tivozanib studies of patients with advanced disease (7, 10, 15, 16, 18) and comparable with that of bevacizumab/mFOLFOX6. Some small differences between the tivozanib and bevacizumab arms were observed in the following TEAEs: neutropenia, hypertension, vomiting, and thrombocytopenia.

Bevacizumab plus chemotherapy has been shown to improve OS compared with chemotherapy alone in patients with mCRC (3, 4). Specifically, bevacizumab added survival benefits when added to treatment with various FOLFOX therapies (19, 22), although the benefits may be specific to PFS according to a recent meta-analysis (23). Bevacizumab/mFOLFOX6 is considered to be a standard treatment for mCRC, and the combination has been used as a comparison arm in several first-line mCRC clinical trials, including those involving panitumumab (24), cediranib (25), axitinib (26), cetuximab (19, 27), and XELOX (capecitabine and oxaliplatin)/SOX (oxaliplatin and S-1; refs. 28, 29). In the current study at the interim futility analysis, the efficacy of tivozanib/mFOLFOX6 was comparable with that of bevacizumab/mFOLFOX6 with a PFS of 9.4 versus 10.7 months (HR = 1.091; CI, 0.693–1.718; P = 0.706) and ORR 45.2% versus 43.2%; the post hoc final analysis of PFS was 9.8 versus 9.5 months (HR = 0.908; CI, 0.624–1.301; P = 0.598). Furthermore, the bevacizumab/mFOLFOX6 results were comparable with those previously observed in recent comparison studies in similar patient populations treated with first-line therapy, which reported a bevacizumab/FOLFOX6 PFS range of 10.1 to 15.9 months and an ORR range of 47.3% to 63% (21, 24–26, 28, 29).

In the current study, none of the subgroup analyses revealed significant differences in PFS. However, a trend of PFS benefit was observed in the predefined LDH high (≥1.5 ULN) subgroup (HR = 0.67; CI, 0.37–1.19). LDH levels were investigated in this study based on previous studies demonstrating increased efficacy associated with high LDH in patients treated with vatalanib combined with FOLFOX4 (30, 31). High LDH levels have been associated with survival in patients with mCRC treated with bevacizumab. Yin and colleagues (2014) reported an improvement in PFS in patients treated with first-line bevacizumab associated with high serum LDH levels (32). Similarly, high LDH level was designated a factor for improved prognosis in patients treated with first-line bevacizumab (33) and in patients treated with FOLFIRI (irinotecan, fluorouracil, and folinic acid) or XELOX plus anti-VEGF therapy (34). In addition, a high level of LDH was reported to be one of only 5 parameters associated with survival in ≥3 studies included in a systematic review of 29 bevacizumab studies (35). However, the association is not definitive. High LDH levels were associated with worse prognosis in the HORIZON I study (36) and in a large single-center analysis (37).

PFS was not significantly changed in this study in the ITT population. It is possible that the high percentage of AEs leading to permanent discontinuation seen in the current study (41.2% Arm A and 34.5% Arm B), which led to a median duration of treatment of 168 and 162 days, may have confounded the estimate of median PFS in both study arms. Previous studies with bevacizumab suggest high discontinuation observed for reasons other than disease progression could have potentially affected survival endpoints, but not the treatment response (19, 38).

In light of the move toward personalized treatments in mCRC, potential biomarkers are an important aspect of treatment investigation. RAS is the only established predictive biomarker in the treatment of mCRC (39). To date, although LDH and NRP-1 levels have been shown to be prognostic for antiangiogenesis therapy (32–35, 40), no predictive factor for anti-VEGF therapy has been identified (39). In the current study, the biomarkers that were investigated were largely based on tivozanib mechanism (VEGF-C, VEGF-A, sVEGFR2, and sVEGFR3) and antiangiogenic properties (IL8). These biomarkers were investigated to find a potential predictor of clinical response following tivozanib therapy based on the premise that reduced ligand and their respective soluble receptor levels have been associated with longer PFS and ORR in previous sunitinib studies (41, 42). RNA expression levels from patient archival formaldehyde fixed-paraffin embedded tumor samples and circulating serum protein levels were evaluated for VEGF-A, -C, -D, and placental growth factor ligand levels as well as VEGF-C/VEGF-A ratios. In addition, circulating NRP-1 protein, which was reported to be highly correlated with PFS following treatment of renal cancer with tivozanib (40), was also evaluated in serum. NRP-1 is a VEGFR-2 coreceptor and is involved in the regulation of VEGFR-2–mediated angiogenesis (43, 44). Although there was no statistical difference in the ITT analysis, the results presented here following the preplanned, post hoc biomarker efficacy analysis suggest that serum NRP-1 could be a potential predictive biomarker, where a subset of patients with low NRP-1 who would benefit from treatment with tivozanib rather than bevacizumab in combination with chemotherapy. A PFS HR of 0.38 for tivozanib compared with bevacizumab, both in combination with FOLFOX chemotherapy, was observed in patients with serum NRP-1 levels below the median for the patients included in the trial analysis while no treatment benefit was observed in the other group. With the small sample sizes associated with each of these NRP-1 subgroups and without a prespecified approach to adjust for multiple comparisons, it is important to notice that a definitive conclusion on whether NRP-1 is a real predictive biomarker cannot be reached. In addition, due to small sample sizes, impact of baseline imbalance on the results is hard to evaluate. It should be noted that as the comparator in this study is not a typical placebo and rather, is an active control, interpretation of the predictive or prognostic values of this biomarker can be different from those trials with a typical placebo arm. Finally, given the number of tested variables, it is possible that the difference observed could be due to chance. Validation of this analysis will be required. Additional efforts are underway to screen antibodies suitable for use as a companion diagnostic to further validate serum NRP-1 as a predictive biomarker to identify those patients most likely to respond to tivozanib.

This study has demonstrated that the tivozanib/FOLFOX6 combination treatment is tolerable in patients with mCRC. The AEs noted with the combination were comparable with those observed in previous studies with tivozanib in combination with other therapies (15–17) and with tivozanib alone (7, 10). The safety of the tivozanib/FOLFOX6 combination was also comparable with the bevacizumab/FOLFOX6 combination. The most common treatment-related AE in both study arms was hypertension, which is characteristic of VEGF-targeted therapy (45).

In conclusion, data from the prespecified interim analysis demonstrated futility for superiority and resulted in discontinuation of this study; however, the efficacy of tivozanib/mFOLFOX6 was comparable with but not superior to that of bevacizumab/mFOLFOX6 in patients with previously untreated mCRC. Moreover, the safety profile of tivozanib/mFOLFOX6 was consistent with that seen with other trials with tivozanib in similar patient populations. On the basis of the results, further study of tivozanib is not warranted in an unselected population. In exploratory analyses, low NRP-1 emerged as a potential predictive biomarker for tivozanib activity. Although these results require further validation, they may warrant further study to determine whether tivozanib may be a potential oncology treatment in selected patient populations.

A. Krivoshik holds stock in Abbott and AbbVie. A.B. Benson III reports receiving compensation for research from Advanced Accelerator Applications, Amgen, Astellas, Aveo, Bayer HealthCare, Genentech, Gilead, and Novartis; is an advisory board member for Bristol-Myers Squibb, Celgene, Eli Lilly and Co., ImClone Systems Inc., Merck and Co., Pharmacyclics, Precision Therapeutics, Sanofi-Aventis, Taiho Pharmaceuticals, and Vicus; and has served as a paid consultant to Genomic Health and the National Cancer Institute, and in the role of DMC to Alchemia and Infinity. J. Bridgewater is an advisory board member for AstraZeneca, Merck Serono, Sanofi, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.B. Benson, I. Kiss, C. Sasse, J. Chen, H.A. Ball, A. Keating, A. Krivoshik

Development of methodology: A.B. Benson, J. Chen, C. Van Sant, H.A. Ball, A. Krivoshik

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.B. Benson, J. Bridgewater, F.A.L.M. Eskens, S. Vossen, C. Van Sant, H.A. Ball

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.B. Benson, F.A.L.M. Eskens, C. Sasse, J. Chen, C. Van Sant, H.A. Ball, A. Keating, A. Krivoshik

Writing, review, and/or revision of the manuscript: A.B. Benson, I. Kiss, J. Bridgewater, F.A.L.M. Eskens, C. Sasse, S. Vossen, J. Chen, H.A. Ball, A. Keating, A. Krivoshik

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Chen, H.A. Ball

Study supervision: C. Sasse, S. Vossen, H.A. Ball, A. Krivoshik

Editorial assistance was provided by Melissa Kirk, PhD, of Scientific Connexions, a member of Ashfield Healthcare Communications. The authors would like to thank Alan Rong, PhD, of Astellas Pharma for his outstanding biostatistical support in the resubmission process of this manuscript.

This study was supported by Astellas Pharma Inc. and AVEO Oncology.

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.