A Randomized, Double-Blind, Placebo-Controlled Phase II Study of the Efficacy and Safety of Monotherapy Ontuxizumab (MORAb-004) Plus Best Supportive Care in Patients with Chemorefractory Metastatic Colorectal Cancer

Angiogenesis is a critical step in the growth, progression, and metastatic spread of cancer cells. Thus, understanding the key signals involved in initiation of new vessel formation and increased blood supply to the growing tumor has been the focus of much research. Tumor endothelial markers are a group of proteins that have been associated with tumor-specific angiogenesis (2). Specifically, endosialin [also known as tumor endothelial marker-1 (TEM-1) or CD248], is a 175 kilodalton type I transmembrane protein of the C-type lectin–like receptor family. Endosialin is expressed in the stromal compartment, including pericytes and fibroblasts, of most human tumors and also expressed in the tumor cells in certain cancers (3, 4). In preclinical studies, endosialin has been shown to play a key role in tumor growth and neovascular formation in numerous cancer types (2, 5–8). The overexpression of endosialin in colorectal cancer tissue compared with healthy tissue suggests its role in colorectal cancer blood vessel formation (3, 4, 9, 10, 11).

Endosialin expression in pericytes and fibroblasts may regulate pericyte proliferation, tumor angiogenesis, and stromal development (12), thought to be critical for tumor–stroma organization and tumor blood vessel development. When endosialin expression in normal human pericytes was downregulated, pericyte proliferation was disrupted, suggesting that the endosialin-dependent signaling pathway may be a target for modulating tumor angiogenesis and stromal development (13). Immunohistochemistry studies of human tumor biopsy samples have shown endosialin expression in tumor stromal cells, especially in the mural cell compartment of new vessels and cancer-associated fibroblasts (12, 14). Preclinical studies have shown that tumors grown in endosialin knockout mice showed reduced tumor growth, invasion, and metastases, as well as abnormal vessel size and structure (3). These observations led to the selective blockade of endosialin as an anticancer treatment strategy.

Ontuxizumab (MORAb-004) is a humanized immunoglobulin G (IgG)1/κ mAb directed against human endosialin. Ontuxizumab binds to human endosialin on the surface of cells expressing the antigen. In preclinical studies, immunofluorescent staining of ontuxizumab-treated human pericytes showed cellular internalization of the antibody, with a corresponding reduction of surface endosialin. Using a human endosialin knockin mouse tumor model, treatment with ontuxizumab resulted in reduced tumor growth, inhibited metastases, and reduced new vessel formation (15).

In patients with colorectal cancer, levels of expression of endosialin, TEM-7, TEM-7R, and TEM-8 are higher in tumor tissue compared with normal background tissues (10). The levels of expression of these tumor endothelial markers were associated with both nodal involvement and tumor progression, suggesting that endosialin may be an appropriate target for therapy using an antibody that binds to and disrupts endosialin function. A completed phase I study found ontuxizumab had an acceptable safety profile up to a MTD of 12 mg/kg, with evidence of disease stabilization observed in this group of heavily pretreated patients [18 of 32 evaluable patients (56%) across all doses and treatment durations had a best overall response of stable disease; ref. 1]. An 8 mg/kg dose was used in the current study based on the pharmacokinetic results of the phase I study.

Given the involvement of endosialin in tumorigenesis and phase I study results, the present study was designed to compare ontuxizumab + best supportive care (BSC) versus placebo + BSC in patients with chemorefractory metastatic colorectal cancer (mCRC) and represents the first clinical study of an anti-endosialin agent in mCRC. The study had two primary objectives: to evaluate the safety and efficacy of ontuxizumab in mCRC and to determine if various biomarkers associated with endosialin were predictive of a positive outcome with ontuxizumab.

This was a multicenter, randomized, double-blind, placebo-controlled, phase II, adaptive-design study of the efficacy and safety of ontuxizumab monotherapy plus BSC compared with placebo plus BSC for chemorefractory mCRC. The study was planned to be conducted in two stages according to an adaptive design employing an interim analysis of progression-free survival (PFS) in stage 1 to potentially identify an endosialin-related biomarker which could be used to enrich a responsive patient population to enter stage 2. The objective of stage 2 was to further evaluate PFS in biomarker-selected patients defined by stage 1.

The study protocol was approved by the Institutional Review Board at participating sites and was conducted in accordance with the Declaration of Helsinki and other Good Practice Guidelines. All patients provided written informed consent to participate.

Eligible patients were 18 years of age or older and had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, with histologically confirmed colorectal cancer and radiologically documented disease progression in the metastatic setting, and a life expectancy of at least 3 months. Unless contraindicated, patients had received prior treatment with a fluoropyrimidine (fluorouracil or capecitabine), irinotecan, oxaliplatin, and bevacizumab (patients with KRAS wild-type tumors must have received an EGFR-targeting agent in the metastatic setting, unless contraindicated), with an interval of ≥2 weeks between the last treatment and first infusion of study therapy.

Eligible patients had archived tumor tissue from a previous surgery/biopsy taken prior to anticancer therapy that was positive for tissue endosialin. Patients were required to have adequate hematologic, coagulation, renal, and liver parameters [absolute neutrophil count ≥1.5 x 10⁹/L, platelet count ≥100 x 10⁹/L, hemoglobin ≥8 g/dL, creatinine ≤1.5 × upper limit of normal (ULN)], total bilirubin <1.5 × ULN, aspartate aminotransferase (AST), and alanine aminotransferase (ALT) <2.5 × ULN. For patients with hepatic metastases, AST and ALT <5.0 × ULN, and activated partial thromboplastin time and prothrombin time < 1.5 × ULN; and no active malignancy other than mCRC required systemic or local treatment within the past 2 years (other than basal cell or squamous cell skin cancer).

Patients were randomly assigned in a 2:1 ratio to treatment with either weekly ontuxizumab (8 mg/kg intravenously) plus BSC or placebo (normal saline) plus BSC. Each cycle consisted of two weekly infusions. Patients continued on the assigned treatment until disease progression by RECIST (16) v.1.1 or until discontinuation for any other reason. All patients received BSC, defined as measures intended to provide palliation of all symptoms and improve quality of life.

Dose reductions were permitted if patients experienced adverse events related to study drug. When a reduction was required to manage toxicity, no dose re-escalation was permitted. Only two dose reductions were permitted for each patient.

All patients were followed for disease progression defined by RECIST v. 1.1, based on computerized tomography or magnetic resonance imaging performed at 8-week intervals.

Standard safety assessments were employed during the study, including assessment of adverse events by National Cancer Institute Common Terminology Criteria for Adverse Events (NCICTCAE, v.4.03), laboratory parameters, and physical examinations. Treatment-emergent adverse events (TEAEs, from first dose to 45 days after the last dose of study drug), drug hypersensitivity events, electrocardiographic data, and development of antidrug antibody (ADA) were assessed. Adverse events that were grade 4 or 5 were reported as serious adverse events (SAEs). The causality of adverse events was graded as not related or related to the study drug as determined by the investigator.

During the follow-up period, survival status and additional therapy for mCRC were captured once per month until death or study termination.

Endosialin (tissue and serum) and platelet-derived growth factor receptor-beta (PDGFR-β; tissue and serum) were evaluated for their potential to serve as predictive biomarkers to prospectively define a subpopulation that could benefit most from ontuxizumab treatment. Baseline levels of markers (serum and tissue) were analyzed to determine their relationship, if any, with PFS and overall survival (OS) in these patients. Both endosialin and PDGFR-β biomarkers were evaluated because previous experimental data showed a biological link between these pathways (13).

Tumor tissue from the time of initial diagnosis, prior to any anticancer treatment, was evaluated for baseline expression of biomarkers. Formalin-fixed, paraffin-embedded slides were prepared and assessed for tissue endosialin and PDGFR-β expression by immunohistochemistry using the previously described rat anti-human endosialin 9G5 mAb (18) or clone 28E1 (Cell Signaling Technology) for PDGFR-β (1). Expression was reported a +1, +2, or +3 by a board-certified pathologist. Methods were as described previously (15, 17).

Screening serum samples were assessed for levels of endosialin and PDGFR-β. Soluble endosialin and PDGFR-β were evaluated using electrochemiluminescent assays using mAbs developed at Morphotek, Inc. (18).

ADA assays were performed using an immunoassay bridging format based on the formation of ontuxizumab and ADA complexes detected in a semiquantitative manner via electrochemiluminescence (1). A tiered analysis process was used to confirm the presence of ontuxizumab-mediated ADA through screening, confirmatory specificity, and titration of ADA responses (19).

Patients with at least one ADA sample obtained after study drug administration were included in the analysis. Patients were considered positive for ADA if at least one ADA-positive sample occurred at any time after study drug administration and the sample was either negative or missing at baseline or, if the sample was positive at baseline, at least a fourfold increase in ADA titer occurred after study drug administration compared with baseline values.

The study was planned to be conducted in two stages according to an adaptive design employing an interim analysis of PFS in 120 patients in stage 1 to potentially identify an endosialin-related biomarker which could be used to adapt and select patients to enter stage 2. The objective of stage 2 was to further evaluate PFS in biomarker-selected patients defined by the stage 1 interim analysis (20). After 126 patients were enrolled and followed for 8 weeks, an interim analysis was performed and the study was discontinued based on the futility criterion (log-rank 1-sided P ≥ 0.60 or an HR of approximately 1.1 or greater).

The primary analysis population for all efficacy endpoints was the intent-to-treat population comprising all randomized patients, according to the treatment assignment. Safety was analyzed for all patients receiving at least one dose of study medication, according to the actual treatment received. The primary efficacy endpoint was PFS, based on RECIST v.1.1, and was defined as the time (in weeks) from the date of randomization to the date of the first observation of disease progression (RECIST v.1.1) or date of death, regardless of the cause. The PFS comparison of ontuxizumab versus placebo was based on the log-rank test, with HR estimated based on Cox's proportional hazards model. The two-sided 95% confidence intervals (CI) for median PFS were constructed using the methodology of Brookmeyer and Crowley (21). OS was analyzed using similar methods.

Tumor tissue– and serum-based biomarkers were evaluated during the stage 1 interim analysis in order to identify a responsive subpopulation of patients. The evaluation of the predictive value of biomarkers was determined using subgroup identification based on differential effect search (SIDES) methodology (22, 23). Using the optimal cutpoint for each biomarker, PFS was compared between the two study treatment groups by log-rank tests with HR estimated based on Cox's proportional hazards model. All P values reported are two-sided unless otherwise stated.

After 126 patients in stage 1 were enrolled and followed for at least 8 weeks, an interim futility analysis was conducted. The futility criterion set forth in the protocol was met in the overall population (overall log-rank test for superiority of ontuxizumab versus placebo based on PFS; one-sided P ≥ 0.60, HR ≥ 1.1). Therefore, the study was stopped after stage 1, and no stage 2 patients were enrolled. The analyses shown below include the 126 patients enrolled in stage 1.

The study was conducted at 61 sites in the United States, 36 of whom enrolled at least one patient. Of 154 patients screened for enrollment, 126 patients were randomly assigned to receive ontuxizumab (84 patients) or placebo (42 patients). Of those, 122 received treatment and were included in the safety population (Fig. 1). Of note, one patient was randomly assigned to placebo but was treated with ontuxizumab. This patient was displayed under the placebo treatment group for the efficacy analysis and under the ontuxizumab group for the safety analysis. A total of 126 patients were included in the intent-to-treat population and evaluated for efficacy.

The average age of the patient population was 61.5 years (Table 1). In the total patient population, 53.2% were male, and 87.3% were white. In terms of site of the primary tumor, 62.7% had the colon, 34.1% had the rectum, and 3.2% had both. Median number of metastatic sites at screening was 2 (range, 1–4).

The primary efficacy endpoint was PFS in the intent-to-treat population. There was no significant difference between the ontuxizumab and placebo groups with respect to the primary endpoint of PFS (Fig. 2), with a median PFS of 8.1 weeks in each group (HR, 1.13; 95% CI, 0.76–1.67; P = 0.53).

There were no significant between-group differences with respect to secondary endpoints. Median OS was 20.7 weeks in the ontuxizumab group and 26.7 weeks in the placebo group (HR, 1.43; 95% CI, 0.93–2.19; P = 0.10; Fig. 3). Of note, during the follow-up phase of this study (after progression was noted), the median time (95% CI) to initiation of poststudy antitumor treatment following progression was substantially delayed in the ontuxizumab group compared with the placebo group. Median time to initiation of poststudy antitumor agents was 6 months (95% CI, 3.8–11.1) for the ontuxizumab group [n = 30/84 (35.7%) patients who initiated poststudy agents] and 2.7 months (95% CI, 2.3–5.8) for the placebo group [n = 23/42 (54.8%) patients who initiated poststudy agents]. In the follow-up period, regorafenib was the most commonly used antitumor agent for both the ontuxizumab group [18 patients (21.4%)] and the placebo group [10 patients (23.8%)]. Second in frequency was the use of combinations of antitumor therapies [eight ontuxizumab patients (9.5%) and eight placebo patients (19.0%)].

Ontuxizumab-treated patients showed no significant difference between right- and left-sided cancer for either PFS or OS. No correlation with PFS was found for time from last treatment or last treatment received (data not shown).

There were no objective responses in the ontuxizumab group and one partial response in the placebo group. Best overall response of stable disease was 14.3% (12 of 84 patients) and 9.5% (4 of 42 patients) for ontuxizumab and placebo, respectively. Progressive disease occurred in 71.4% of patients in each group (60 of 84 patients and 30 of 42 patients for ontuxizumab and placebo, respectively). The remaining 12 patients (14.3%) in the ontuxizumab group and seven patients (16.7%) in the placebo group were not evaluable for tumor response.

As required for enrollment, all assessable baseline tumor tissues were positive for endosialin expression (76 in the ontuxizumab group and 39 in the placebo group). Only baseline tumor tissue endosialin significantly affected PFS (i.e., was prognostic), based on a univariate regression analysis (P = 0.0222; HR, 0.965; 95% CI, 0.936–0.995); however, it was not predictive of response by ontuxizumab.

Each biomarker was evaluated for predictive ability, and an optimal cutoff point for predictability was determined. PFS was compared between the two study treatment groups in the biomarker-positive subgroup. None of the biomarkers showed a significant difference between control- and ontuxizumab-treated patients for PFS in the biomarker-positive subgroup. No individual biomarker defined a responsive subpopulation that was robust enough for use in enriching a patient population for stage 2 enrollment. Patients with coexpression of serum endosialin (>121 ng/mL) and serum PDGFR-β (≤9.66 ng/mL) provided the optimal predictive value for PFS. This subgroup was comprised of 45 patients. The median PFS for this subgroup was 8.3 weeks (95% CI, 8.1–8.6) among patients treated with ontuxizumab versus 7.7 weeks (95% CI, 6.1–8.3) for placebo (HR, 0.45; 95% CI, 0.23–0.89; P = 0.0149). There were seven patients with ontuxizumab treatment who were progression free at 10 weeks, whereas all patients with placebo treatment progressed by that time (Fig. 4A). Although a significant difference for PFS was observed among this subgroup with ontuxizumab treatment compared with placebo, this difference was not evident for OS (Fig. 4B). The median OS for this subgroup of patients was 16.6 weeks (95% CI, 12.6–23.4) among patients treated with ontuxizumab versus 15.3 weeks (95% CI, 8.3–26.7) for placebo (HR, 0.95; 95% CI, 0.48–1.89; P = 0.8883).

In the safety population, one patient randomized to placebo was treated with ontuxizumab in error, so 82 patients in the ontuxizumab group and 40 patients in the placebo group were analyzed for safety. The median number of treatment cycles (2-week cycles) was 4 in each treatment group (range, 1–18 cycles for ontuxizumab and 1–24 cycles for placebo). Patients received a median of 8.0 infusions in each group. Study drug delays, reductions, or interruptions were similar in both groups and occurred in 26.2% of patients overall.

TEAEs (regardless of relationship to treatment) were reported by most patients: 95.1% in the ontuxizumab group and 100% in the placebo group. The most common TEAEs in the ontuxizumab and placebo groups, respectively, were primarily gastrointestinal in nature and included fatigue (53.7% vs. 47.5%), nausea (39.0% vs. 35.0%), decreased appetite (34.1% vs. 27.5%), and constipation (28.0% vs. 32.5%; Table 2). Grade 3/4 TEAEs occurring in ≥10% of patients in the ontuxizumab or placebo groups, respectively, were back pain (11.0% vs. 0%), fatigue (9.8% vs. 10.0%), and anemia (3.7% vs. 10.0%). Adverse events led to early discontinuation for two patients in the ontuxizumab group (2.4%, both hepatic failure) and one patient in the placebo group (2.5%, fatigue).

Adverse events considered by the investigator to be at least possibly related to study treatment occurred in 65.9% of patients receiving ontuxizumab and 70.0% of patients receiving placebo, with possibly treatment-related grade 3/4 TEAEs occurring in 11.0% and 10.0% of patients, respectively.

A total of 96 of 122 (78.7%) patients died during the study period; of these, 86 died due to progressive disease. The percentage of deaths was similar in both groups (80.5% in the ontuxizumab group and 75.0% in the placebo group), but no deaths were due to a treatment-related adverse event. The frequency of SAEs was similar in both groups (ontuxizumab: 37.8% vs. placebo: 37.5%). The frequencies of the most common SAEs (occurring in 4 or more patients) in the ontuxizumab group were dehydration (6.1%) and back pain (4.9%).

In this study population, 7.3% of dosed patients developed a measurable treatment-induced or boosted ADA response subsequent to ontuxizumab administration, and 6.6% of all study patients were found to have preexisting immunoreactivity to ontuxizumab. Drug hypersensitivity adverse events (DHAE) were defined as clinical hypersensitivity events that occurred within 48 hours of infusion and for which the serum ADA result was positive for at least one sample obtained on or after the onset of the adverse event of interest. One patient receiving ontuxizumab experienced a DHAE (grade 1 dyspnea).

There were no notable differences between treatment groups in clinical laboratory variables, vital signs, or physical exams.

Colorectal cancer remains one of the most common newly diagnosed cancers in the world, as well as in the United States (24). This study represents the first clinical trial assessing treatment with a mAb targeting endosialin in patients with chemorefractory mCRC. The purpose of the study was twofold: to evaluate the safety and efficacy of ontuxizumab monotherapy in mCRC and to determine if various biomarkers associated with endosialin were predictive of a positive outcome with ontuxizumab.

The current study failed to demonstrate a benefit with ontuxizumab monotherapy compared with placebo for clinical response parameters of PFS, OS, or ORR in patients with mCRC. No partial or complete responses were observed in the ontuxizumab treatment group. Although no indication of improved efficacy with ontuxizumab monotherapy was observed in this study, the study has advanced the body of evidence relating to treatment of mCRC with an endosialin-targeted mAb. Results of this study will help inform directions of future studies in this field.

Despite the results of this study, there remains a wealth of evidence implicating the role of endosialin in tumor growth and expansion in patients with colorectal cancer. A number of studies have shown that tumor endosialin levels are associated with tumor progression and poor survival in patients with colorectal cancer (10, 11, 25). Preclinical studies have shown that ontuxizumab, an endosialin-targeted mAb, specifically disrupts tumor growth and tumor metastasis. In a preclinical study by Rybinski and colleagues, endosialin expression was significantly reduced with ontuxizumab treatment resulting in drastically more small and dysfunctional vessels in the treated tumors (15). These results suggest that ontuxizumab reduced endosialin on pericytes, impaired tumor microvasculature maturation, and ultimately suppressed tumor development. Another study by Li and colleagues showed that ontuxizumab-treated mice demonstrated a 70% reduction in tumor burden compared with control animals (26).

Although the results of the current study showed that monotherapy with ontuxizumab does not improve the outcome of patients with mCRC, the approach to treating with an antiangiogenic agent remains valid. Future efforts will focus on understanding mechanisms to optimize the delivery of ontuxizumab effectively. Another antiangiogenic agent, bevacizumab, has been shown to be effective in treating patients with mCRC (27) when combined with chemotherapy, suggesting that the antiangiogenic approach to treating mCRC is valid. Although combination of bevacizumab with chemotherapy was effective, monotherapy with bevacizumab showed little if any efficacy in patients with colorectal cancer (28). Potentially, ontuxizumab will require the combination with chemotherapy to demonstrate improved survival.

Several alternative approaches to augment the efficacy of ontuxizumab can be considered. One approach to improving the efficacy of mAbs has been to use antibody–drug conjugates to selectively deliver cytotoxic agents to tumor sites (29, 30). In preclinical studies, an anti-endosialin antibody was conjugated to cytotoxic agents to treat human tumor xenograft sarcoma models of endosialin-positive disease resulting in a prolonged antitumor response compared with controls (29). In addition, combining inhibition of angiogenesis with inhibition of nonangiogenic vessels (vessel co-option) has been shown to be more effective than antiangiogenesis alone in preclinical models (31).

Although no PFS benefit was seen in the heavily pretreated patients with mCRC in this study, the potential for ontuxizumab to support disease stabilization in less advanced tumors could also warrant further evaluation. It is possible that the antiangiogenic effects of ontuxizumab require treatment earlier in the process of tumor expansion and neovascularization.

The study design and the sample size calculations for the current study were based on another phase III, placebo-controlled study with regorafenib in patients with mCRC (CORRECT study; ref. 32). Compared with this study, the median OS in the placebo treatment group of the current study was longer than expected. In the CORRECT study, the placebo control treatment group had a median OS of 5.0 months (2.8–10.4) compared with 6.2 months (4.7–9.7 months or 26.7 weeks) in the placebo group in the present study. One possible explanation to account for the higher than expected median OS in the control group for the current study was that the CORRECT study was conducted worldwide, whereas this study was limited to the United States. Of note, during the follow-up phase of this study (after progression was noted), the median time (95% CI) to initiation of poststudy antitumor treatment following progression was substantially delayed in the ontuxizumab group. There was no obvious explanation for the delay in poststudy antitumor treatment among the ontuxizumab treatment group, but it may explain why the OS in the ontuxizumab treatment group did not compare favorably with the placebo group.

A secondary objective of our study was to evaluate the potential for specific biomarkers to prospectively identify the most appropriate population to treat with ontuxizumab. The search for predictive biomarkers to use to enrich patient populations has proven to be elusive. Individual biomarkers of tissue and serum endosialin, as well as tissue and serum PDGFR-β, were evaluated for their ability to predict a positive clinical response with ontuxizumab treatment. Expression levels of endosialin were evaluated, given it is the target antigen of ontuxizumab, and PDGFR-β was evaluated because it has been reported to be an important signaling partner of endosialin (13). This study did not identify a serum or tissue biomarker–defined responsive population based on PFS to support adaptation into stage 2. Because of the limited efficacy observed in this study, a biomarker-defined responsive population was difficult to identify.

Tumor tissue biomarkers were not predictive of response. One potential reason was that the tissue for biomarker evaluation was from the time of original diagnosis of colorectal cancer; tissue levels of biomarker expression at time of diagnosis may have changed compared with the tissue microenvironment at the time of treatment for mCRC. Exploratory biomarker analyses found that patients with a combination of serum endosialin (>121 ng/mL) and serum PDGFR-β (≤9.66 ng/mL) provided the optimal predictive value for PFS, but this predictive impact represented only 8% (7 of 84 patients treated with ontuxizumab) and was observed only for PFS and not for OS. The predictive significance of various study biomarkers individually or combined was therefore inconclusive.

The time course of events occurring in the microenvironment of the tumor that lead to angiogenesis and tumor expansion, including the role of endosialin, is still poorly understood. The initiation of angiogenesis in response to tumor growth is a multifaceted process that may arise locally or from bone marrow–derived endothelial cell precursors. In a recent study using a human endosialin knockin mouse tumor model, endosialin was highly expressed in the pericytes of tumor blood vessels following tumor implantation (15). Treatment with ontuxizumab showed internalization and degradation of endosialin in pericytes, using immunofluorescent staining of tumor microvasculature. These changes in endosialin levels at the tissue level and the relationship to changes in endosialin levels in the blood are not yet well understood. Future directions for research include evaluation of endosialin levels localized to specific microenvironment subcompartments to determine if they are associated with improved outcomes. Also, serial evaluations of the time course of endosialin expression in relation to tumor progression are important for furthering the body of evidence supporting the role of endosialin in tumor progression.

The results and conclusions of this single-agent ontuxizumab study will be used to inform future ontuxizumab research programs. mCRC continues to be a disease posing an ongoing serious unmet medical need, and the investigation of potentially effective agents in this setting continues to be a high priority.

L.A. Renfro is a consultant/advisory board member for Bayer and Ignyta. L.A. Diaz has ownership interests (including patents) at PapGene and PGDx, reports receiving speakers bureau honoraria from OncLive and PrimeOncology, and is a consultant/advisory board member for Cell Design Labs, Genocea, and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Grothey, L.A. Renfro, H.I. Hurwitz, J.L. Marshall, H. Safran, G.P. Kim, J. Heyburn, C. Schweizer, D.J. O'Shannessy, L.A. Diaz, Jr

Development of methodology: A. Grothey, L.A. Renfro, G.P. Kim, D.J. O'Shannessy, L.A. Diaz, Jr

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Grothey, J.R. Strosberg, H.I. Hurwitz, J.L. Marshall, H. Safran, M.J. Guarino, G.P. Kim, J.R. Hecht, S.C. Weil, D.J. O'Shannessy, L.A. Diaz, Jr

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Grothey, L.A. Renfro, H.I. Hurwitz, J.L. Marshall, G.P. Kim, J.R. Hecht, S.C. Weil, W. Wang, C. Schweizer, D.J. O'Shannessy, L.A. Diaz, Jr

Writing, review, and/or revision of the manuscript: A. Grothey, J.R. Strosberg, L.A. Renfro, H.I. Hurwitz, J.L. Marshall, H. Safran, M.J. Guarino, G.P. Kim, J.R. Hecht, S.C. Weil, J. Heyburn, C. Schweizer, D.J. O'Shannessy, L.A. Diaz, Jr

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Safran, J.R. Hecht, S.C. Weil

Study supervision: S.C. Weil, J. Heyburn, L.A. Diaz, Jr

Other (patient accrual): A. Grothey

The authors wish to acknowledge the extensive contribution of Daniel Sargent, PhD (deceased), Mayo Clinic, for statistical input and M.L. Skoglund, PhD, Healthcare Consulting, for medical writing. This study was funded by Morphotek, Inc.

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.