Intraop indicates intraoperative; mFOLFIRINOX, modified FOLFIRINOX, treatment with fluorouracil, irinotecan, and oxaliplatin; postop, postoperative; preop, preoperative.
mFOLFIRINOX indicates modified FOLFIRINOX, treatment with fluorouracil, irinotecan, and oxaliplatin.
eTable 1. Reasons for Ineligibility by Central Radiology Review
eTable 2. Adverse Events Related to Preoperative Chemotherapy
eTable 3. Adverse Events Related to Surgery
eTable 4. Adverse Events Related to Post-Operative Chemotherapy
Data Sharing Statement
Customize your JAMA Network experience by selecting one or more topics from the list below.
Sohal DPS, Duong M, Ahmad SA, et al. Efficacy of Perioperative Chemotherapy for Resectable Pancreatic Adenocarcinoma: A Phase 2 Randomized Clinical Trial. JAMA Oncol. 2021;7(3):421–427. doi:10.1001/jamaoncol.2020.7328
What is the 2-year overall survival (OS) for patients with resectable pancreatic adenocarcinoma treated with perioperative chemotherapy with either mFOLFIRINOX or gemcitabine/nab-paclitaxel?
In this randomized phase 2 clinical trial, perioperative chemotherapy with either regimen led to a 2-year OS of about 47% (median OS of 23 months), which was not statistically significantly improved over historical data from adjuvant trials. We also noted 29% of enrolled patients to have ineligible disease by resectability criteria, upon central radiology review.
This study establishes safety and efficacy parameters regarding perioperative chemotherapy for resectable pancreatic adenocarcinoma and highlights quality control lessons for future trials.
Clinical outcomes after curative treatment of resectable pancreatic ductal adenocarcinoma (PDA) remain suboptimal. To assess the potential of early control of systemic disease with multiagent perioperative chemotherapy, we conducted a prospective trial.
To determine 2-year overall survival (OS) using perioperative chemotherapy for resectable PDA.
Design, Setting, and Participants
This was a randomized phase 2 trial of perioperative chemotherapy with a pick-the-winner design. It was conducted across the National Clinical Trials Network, including academic and community centers all across the US. Eligibility required patients with Zubrod Performance Score of 0 or 1, confirmed tissue diagnosis of PDA, and resectable disease per Intergroup criteria.
Perioperative (12 weeks preoperative, 12 weeks postoperative) chemotherapy with either fluorouracil, irinotecan, and oxaliplatin (mFOLFIRINOX, arm 1) or gemcitabine/nab-paclitaxel (arm 2).
Main Outcomes and Measures
The primary outcome was 2-year overall survival (OS), using a pick-the-winner design; for 100 eligible patients, accrual up to 150 patients was planned to account for cases deemed ineligible at central radiology review.
From 2015 to 2018, 147 patients were enrolled; 43 patients (29%) had ineligible disease, beyond resectability criteria, at central radiology review. There were 102 eligible and evaluable patients, 55 in arm 1 and 47 in arm 2, of whom the median (range) age was 66 (44-76) and 64 (46-76) years, respectively; 36 patients (65%) in arm 1 and 24 (51%) in arm 2 were men. In arm 1, 34 (62%) had Zubrod Performance Score of 0, while in arm 2, 31 (66%) did; and 44 (80%) in arm 1 and 39 (83%) in arm 2 had head tumors. Of 102 patients, 84% and 85% completed preoperative chemotherapy, 73% and 70% underwent resection, and 49% and 40% completed all treatment. Adverse events were expected hematologic toxic effects, fatigue, and gastrointestinal toxicities. Two-year OS was 47% (95% CI, 31%-61%) for arm 1 and 48% (95% CI, 31%-63%) for arm 2; median OS was 23.2 months (95% CI, 17.6-45.9 months) and 23.6 months (95% CI, 17.8-31.7 months). Neither arm’s 2-year OS estimate was significantly higher than the a priori threshold of 40%. Median disease-free survival after resection was 10.9 months in arm 1 and 14.2 months in arm 2.
Conclusions and Relevance
This phase 2 randomized clinical trial did not demonstrate an improved OS with perioperative chemotherapy, compared with historical data from adjuvant trials in resectable pancreatic cancer. Two-year OS was 47% with mFOLFIRINOX and 48% with gemcitabine/nab-paclitaxel for all eligible patients starting treatment for resectable PDA. The trial also demonstrated adequate safety and high resectability rates with perioperative chemotherapy, and challenges in quality control for resectability criteria.
ClinicalTrials.gov Identifier: NCT02562716
Surgical resection represents the only chance of cure for patients with pancreatic ductal adenocarcinoma, a disease with overall dismal outcomes.1 Even with resection, however, outcomes remain suboptimal. Adjuvant chemotherapy with multiagent regimens improves overall survival compared with the previous de facto standard treatment, gemcitabine.2,3 Nonetheless, a surgery-first approach may render a significant proportion of patients unable to tolerate adequate adjuvant therapy, which has significant toxic effects.2 Up-front chemotherapy—neoadjuvant therapy—may allow early control of systemic disease and identify patients who are intolerant of chemotherapy or whose tumors progress despite aggressive chemotherapy.1,4,5 This test of patient physiology and disease biology can allow such patients to be spared a major operation that is unlikely to improve clinical outcomes. Neoadjuvant therapy is delivered with the view that pancreatic cancer is a systemic disease, and may improve the identification of patients for whom a cure may be achieved. However, while there have been many reports of neoadjuvant therapy for borderline resectable and unresectable pancreatic cancer, and retrospective studies showing benefit for resectable disease, prospective evaluation of neoadjuvant therapy for resectable disease is limited.6-8
With this background, we aimed to study the feasibility and clinical utility of neoadjuvant and adjuvant (perioperative) chemotherapy for patients with resectable pancreatic cancer. The present prospective clinical trial assessed the feasibility, safety, and efficacy of the 2 multi-agent chemotherapy regimens for this disease, modified FOLFIRINOX (mFOLFIRINOX; oxaliplatin, irinotecan, and fluorouracil) and gemcitabine/nab-paclitaxel. To our knowledge, this is the first randomized trial of perioperative chemotherapy in the National Clinical Trials Network in the United States, and here we report the primary end point results.
The present study was a randomized phase 2 clinical trial of perioperative chemotherapy for resectable pancreatic adenocarcinoma. Resectable disease was defined using cross-sectional imaging (contrast-enhanced computed tomography or magnetic resonance imaging scans of the chest, abdomen, and pelvis) obtained within 28 days prior to registration. The following eligibility criteria had to be met: no interface of the tumor with the celiac, common hepatic, or superior mesenteric arteries (and, if present, variants); less than 180° interface between tumor and vessel wall of the portal or superior mesenteric veins; patent portal vein/splenic vein confluence; and no metastases (no visceral lesions, no lymphadenopathy outside the surgical basin). For tumors of the body and tail of the pancreas, interface with the splenic artery and splenic vein of any degree was considered resectable disease. A formal surgical consultation to confirm resectability was required within 21 days of registration to the study. Other key eligibility criteria were confirmed histologic or cytologic diagnosis of pancreatic adenocarcinoma (histologies other than adenocarcinoma, or mixed histologies, were excluded); measurable disease on imaging, per Response Evaluation Criteria in Solid Tumors (RECIST) 1.1; no prior therapy for the index cancer; age between 18 years and 75 years; and Zubrod Performance Score (PS) of 0 or 1 assessed within 28 days of registration. Normal bone marrow, hepatic, and renal function were required, as assessed by standard laboratory criteria. The study was approved by the central institutional review board of the National Cancer Institute, and then by local institutional review boards of all participating sites. All patients provided written informed consent prior to enrollment.
Baseline imaging studies were uploaded to a central server, and a specialized abdominal radiologist blinded to treatment assignments performed reviews using a 6-point checklist (visible pancreatic mass; measurable disease; absence of arterial interface; venous interface of less than or equal to 180°; patent portal-splenic confluence; and absence of metastatic disease, including lymphadenopathy outside the surgical basin). Patients not meeting all criteria were deemed ineligible; however, since this determination was not in real time, enrolling sites were not notified of this status and all patients were allowed to proceed on protocol-defined therapy. After a temporary closure for planned interim analysis to assess safety of neoadjuvant therapy, an institutional checklist, identical to the one used by the central radiologist, was mandated for completion by the enrolling site radiologist to minimize the enrollment of ineligible patients.
After registration to the study, patients were randomized in equal proportions to one of the 2 treatment arms with stratification for Zubrod PS 0 vs 1. Randomization was performed centrally by the SWOG Operations Office, after patient registration. Allocations were not blinded. Patients in arm 1 were treated with mFOLFIRINOX: oxaliplatin, 85 mg/m2, followed by irinotecan, 180 mg/m2, followed by fluorouracil, 2400 mg/m2, infused via a chemotherapy pump over 46 hours. This treatment was administered every 2 weeks, for a total of 6 neoadjuvant doses and 6 adjuvant doses. Patients in arm 2 were treated with nab-paclitaxel 125 mg/m2, followed by gemcitabine 1000 mg/m2. This treatment was administered every week, 3 weeks on and 1 week off, for a total of 9 neoadjuvant doses and 9 adjuvant doses.
In each arm, supportive treatments, including steroids, antiemetics, and antidiarrheals, both intravenous and oral, were permitted per the treating physician’s discretion. A prescribed dose modification schema was used for common treatment-related toxic adverse effects: doses of each regimen could be reduced by 2 stepwise levels for grade 3 or higher toxic effects. Missed doses were not given later, to prevent prolongation of chemotherapy duration, which could lead to disparate delays in surgery and follow up. Hematopoietic growth factors were not used prophylactically; their use was allowed if neutropenia occurred.
After completion of neoadjuvant chemotherapy, patients underwent repeated cross-sectional imaging. In the absence of disease progression (by RECIST 1.1), patients were taken to surgical resection within 4-8 weeks following the last dose of neoadjuvant chemotherapy. Patients were to start adjuvant chemotherapy within 4 to 12 weeks after resection. The final neoadjuvant chemotherapy doses were used as the initial adjuvant chemotherapy doses.
The primary outcome was 2-year overall survival (OS), calculated from date of randomization to time of death from any cause. Secondary outcomes included toxicity and overall resection rate; among patients with resection, secondary outcomes were R0 resection rate, pathologic response rates, and disease-free survival (DFS) from the time of resection. At each treatment visit, chemotherapy toxicities were assessed using Common Terminology Criteria for Adverse Events version 5.0. Surgical and pathologic reports were centrally reviewed by a study surgical oncologist. Pathologic response was graded using the College of American Pathologists criteria, as follows: 0, complete response with no residual tumor; 1, moderate response with minimal residual cancer (single cells or small groups of cancer cells); 2, minimal response with residual cancer outgrown by fibrosis; 3, poor or no response with no definite response identified (minimal or no tumor kill; extensive residual cancer).9 Pathologic assessments were performed locally. Progression and recurrence of disease were assessed by cross-sectional imaging every 3 months following completion of treatment, with the first post-treatment scan within 2 weeks of completion of all therapy, until disease progression, or last follow-up. For OS and DFS, censoring occurred at the date of last contact. Postoperative complications were also assessed and noted. All resected surgical specimens were banked in a central biorepository for future correlative research.
This study used a randomized phase 2 pick-the-winner design, with minimum activity requirements, to address the objective of choosing a chemotherapy regimen for further study in resectable pancreatic adenocarcinoma. For each arm, the observed 2-year OS was first to be compared to the null hypothesis of 40%, assuming a 58% alternative hypothesis, 88% power, and a 1-sided significance of 0.05. If rates in both arms met this threshold, then a sample size of 100 patients (50 per arm) would provide a 90% probability of selecting the better regimen with an OS hazard ratio of at least 1.4. A total sample size of 112 was planned to allow for approximately 10% ineligible patients.
The primary analysis of OS was conducted in all eligible and evaluable patients according to the intent-to-treat principle. Probabilities of OS were estimated using the Kaplan-Meier method. The observed 2-year OS for each arm was compared with the null hypothesis of 40% using a log-rank test. Statistical differences in efficacy event rates between treatment arms were not assessed because the 2-year OS threshold was not met in either arm. Overall response rate (ORR) was defined as confirmed and unconfirmed complete and partial response. Eligible patients who received at least 1 dose of any drug were included in the safety analysis. Adverse event rates were compared across treatment arms using Fisher exact test; P values were 2-sided with significance level of 5%. Analyses were carried out using SAS version 9.4 (SAS Institute) and R version 3.6.1 (R Project).
From October 2015 to April 2018, a total of 147 patients were enrolled to this study. In November 2016, the study was halted temporarily for the preplanned interim analysis; given rapid accrual, 76 patients had been enrolled by then. At the interim analysis (planned for safety), in response to an observed ineligibility rate of 33%, enrollment was increased to approximately 150 patients; the study reopened to accrual in January 2017 and closed to accrual in April 2018.
One patient was ineligible due to presence of distal bile duct cancer at surgical pathologic evaluation. Of the 146 patients who were reviewed by central radiology, 43 patients (29%) were found to be ineligible for at least 1 reason (eTable 1 in Supplement 1): 15 (35%) had venous interface 180° or greater, 22 (51%) had any arterial interface, and 28 (65%) had suspicion of distant disease; some patients met multiple ineligibility criteria. One patient withdrew study consent prior to any therapy and was therefore not evaluable. Of the 102 eligible and evaluable patients, 44 patients (43%) had no tumor interface with either vein; of the remaining 58, 37 had tumor interface with the portal vein and 51 had tumor interface with the superior mesenteric vein. Of 102 patients, 55 were randomized to the mFOLFIRINOX arm and 47 to the gemcitabine/nab-paclitaxel arm. Baseline characteristics of eligible and evaluable patients are shown in Table 1.
The course of patients through the study is shown in Figure 1. In arms 1 and 2, respectively, 53 of 55 (96%) and 45 of 47 (96%) of patients started neoadjuvant chemotherapy; 46 (84%) and 40 (85%) patients, respectively, completed neoadjuvant chemotherapy, and 40 (73%) and 33 (70%) underwent resection. Neoadjuvant chemotherapy toxic effects, clinical deterioration, and patient refusal (n = 8) were the major reasons for not undergoing surgical resection in the mFOLFIRINOX arm; these (n = 5) as well as disease progression (n = 7) were the major reasons for patients not undergoing resection in the gemcitabine/nab-paclitaxel arm. A total of 40 patients in arm 1 and 33 in arm 2 underwent resection; of these, 31 in arm 1 (78%; 56% overall) and 26 in arm 2 (79%; 55% overall) started adjuvant chemotherapy, and 27 (68%; 49% overall) and 19 (58%; 40% overall) completed all treatment in arms 1 and 2, respectively. After surgery, treatment toxicity and patient refusal, disease progression, and treating physician’s choice were the reasons for not starting adjuvant chemotherapy. (The trial protocol is presented in Supplement 2.)
Rates of the most frequently observed grade 3 and 4 adverse events, hematologic toxicities, fatigue, diarrhea, nausea, and neuropathy, were as expected (eTables 2, 3, and 4 in Supplement 1). In the neoadjuvant setting, a somewhat higher proportion in the gemcitabine/nab-paclitaxel vs mFOLFIRINOX arm experienced neutropenia (27% vs 19%), whereas a slightly higher proportion of patients in the mFOLFIRINOX vs gemcitabine/nab-paclitaxel arm had diarrhea (11% vs 4%); neither of these differences was statistically significant. Surgery-related grade 3 or 4 adverse events were few; the most common events were anemia (14% vs 3%) and anorexia (5% vs 0%) in the mFOLFIRINOX vs gemcitabine/nab-paclitaxel arms, respectively. Detailed surgical outcomes have been published previously.10 Rates of grade 3 or 4 adverse events were lower in the adjuvant vs neoadjuvant setting. A statistically significantly higher proportion of patients in the gemcitabine/nab-paclitaxel arm experienced neutropenia (27% vs 0%, P = .002), and a higher proportion in the mFOLFIRINOX arm experienced neuropathy (16% vs 4%, P = .21).
Neither arm’s 2-year OS estimate was statistically significantly higher than the prespecified threshold of 40%. The estimated 2-year OS was 47% (95% CI, 31%-61%; P = .15) for mFOLFIRINOX and 48% (95% CI, 31%-63%; P = .14) for gemcitabine/nab-paclitaxel, with median OS of 23.2 months (95% CI, 17.6-45.9 months) and 23.6 months (95% CI, 17.8-31.7 months), respectively (Figure 2). Overall response rate was assessed for all 102 eligible patients; there were no statistically significant differences between mFOLFIRINOX and gemcitabine/nab-paclitaxel arms: 5 (9%) vs 10 (21%), P = .15, respectively. Among the 40 patients in the mFOLFIRINOX arm and 33 patients in the gemcitabine/nab-paclitaxel arm undergoing surgical resection: R0 resection was achieved in 34 (85%) and 28 (85%) patients, node-negative resection in 16 (40%) and 15 (45%) patients, and pathologic complete or major response in 10 (25%) and 14 (42%) patients, and estimates of median disease-free survival from resection were 10.9 and 14.2 months, respectively (Table 2).
In this trial of perioperative chemotherapy for resectable pancreatic adenocarcinoma, we have established the feasibility of developing this treatment approach in patients with resectable pancreatic cancer and demonstrated significant interest from multidisciplinary medical and surgical oncology teams as evidenced by the rapid accrual. Neither arm met the prespecified overall survival threshold, but we demonstrated that this approach resulted in acceptable safety and resectability rates. Furthermore, this trial has established the outcome metrics for perioperative chemotherapy in a study that included numerous academic and community centers, with a resection rate of 72% and, among those resected, an R0 resection rate of 85%. In addition, this study demonstrated large interclinician variability in critical assessment of resectability based on established radiologic criteria.
Neoadjuvant chemotherapy has been tested in several small institutional studies, and meta-analyses of these studies suggest benefit compared with up-front resection followed by adjuvant therapy.11,12 Several larger multi-institutional prospective trials have also been published. The PACT-15 study8 enrolled 88 evaluable resectable pancreatic cancer patients and randomized them to 1 of 3 arms: surgery followed by adjuvant gemcitabine, surgery followed by adjuvant PEXG (cisplatin, epirubicin, capecitabine, gemcitabine), or perioperative PEXG. The resection rates in the 3 arms, respectively, were 85%, 90%, and 84%. Median OS was 20.4, 26.4, and 38.2 months, respectively. The PREOPANC trial13 enrolled patients with resectable or borderline resectable disease, randomizing 246 patients to either surgery followed by adjuvant gemcitabine, or neoadjuvant gemcitabine-based chemoradiation followed by resection followed by adjuvant gemcitabine. The resection rates were 72% and 61%, and median OS was 14.3 and 16 months, for the immediate-surgery and neoadjuvant treatment arms, respectively. These results are somewhat disparate: PACT-158 suggests a benefit from neoadjuvant therapy, while PREOPANC13 does not. Although the latter trial enrolled borderline resectable patients as well, the biology is no different. The present study was not intended to be an adjuvant vs neoadjuvant comparison. (In addition, radiation was not used in the present study.) With this context, what we demonstrate most clearly is that resection can be safely achieved in more than 70% of patients. Generally, patients fail to reach resection due to physiological difficulties (inability to tolerate neoadjuvant chemotherapy) or aggressive disease characteristics (disease progression despite neoadjuvant chemotherapy). These are among the most commonly cited reasons for neoadjuvant therapy—it helps select out patients who are unlikely to benefit from a major intra-abdominal operation. Overall, however, fewer than 50% of patients completed all therapy, largely due to failure to receive adjuvant chemotherapy. Surgical complications were low; adjuvant therapy was hampered by patient refusal, disease progression, and toxic effects. Encouraging results were seen in the PRODIGE-24 trial,14 with a median overall survival of 54 months with adjuvant modified FOLFIRINOX. However, it is critical to recognize that these patients represent a highly selected subgroup. They were patients who did not progress prior to resection (even up-front resection arms of PACT-158 and PREOPANC13 show many failures prior to resection), were able to complete resection without intraoperative findings of metastatic disease, were able to recover from surgery to a performance score of 0 or 1, and were further selected for favorable biology by postoperative levels of cancer antigen 19-9. (In the present study, tumor markers were not followed, and were not used to make any patient selection or treatment decision.) A substantial proportion of patients with potentially resectable pancreatic cancer will not qualify for enrollment in such adjuvant trials but can be candidates for neoadjuvant/perioperative therapy trials. In this study, 16 of 73 patients (22%) who underwent surgery were unable to start adjuvant chemotherapy. In the PRODIGE-24 study,14 only 66% of patients were able to complete all planned adjuvant chemotherapy. In the present study, 86 (88%) of 98 patients who started neoadjuvant therapy were able to complete it, highlighting improved delivery preoperatively. These findings collectively raise the question of whether total neoadjuvant therapy should be considered as an approach in this setting in future clinical trials.
To our knowledge, this is the first time the 2 frontline chemotherapy regimens for pancreatic cancer have been tested in 1 study. The present study was not designed for a between-arm comparison; nonetheless, the 2 arms resulted in very similar outcomes. Prior to this study, cross-trial comparisons in metastatic disease suggested that FOLFIRINOX might be more efficacious, at the cost of increased toxic effects.15,16 While the statistical design of this study was not built for a head-to-head comparison, the results for the 2 arms are similar with respect to efficacy. One exception is a numerically higher proportion of patients in the gemcitabine/nab-paclitaxel arm achieving a complete or major pathologic response, which was perhaps offset by a slightly lower proportion of patients completing all therapy in that arm. That latter point also establishes the fact that the two regimens are equally toxic when administered per protocol—hematologic toxic effects were higher, in fact, with the gemcitabine/nab-paclitaxel arm. It is to be noted here that the use of growth factors was not allowed at the initiation of therapy, as they were recommended only in the setting of neutropenia.
Another key finding from this study is the challenge in delineating resectability in this disease. The definition of resectability has evolved over time, and for this clinical trial we used the Intergroup criteria.17 These criteria define resectability on the basis of tumor–vessel wall interface, that is, the degrees of circumference of the respective vessel wall in contact with solid tumor. The new approach, moving away from subjective descriptors such as involvement, abutment, and encasement and toward objective geometric criteria, is endorsed by multiple specialty organizations.18-20 Nonetheless, while the debate on the exact definition is beyond the scope of the present study, there is consensus that resectability definition is associated with outcomes, and should be approached meticulously.21,22 We describe a systematic real-world evaluation. In 43 of 147 (29%) cases enrolled in our study with well-defined radiologic criteria, upon retrospective central radiology review, the criteria were not met. At our interim analysis, 25 (33%) of the first 76 patients enrolled had ineligible disease per central radiology review. Therefore, we revised the protocol to increase the sample size accordingly. We also added a requirement for a radiology checklist to be signed off by the site radiologist to minimize ineligible cases. Still, 18 (26%) of the subsequent 70 patients centrally reviewed had ineligible disease, showing only a marginal improvement. The reasons for this are conjectural but include inadequate multidisciplinary review, a desire to provide the patient the benefit of doubt, and the absence of a simple and rapid method to adjudicate metastatic disease without invasive procedures. Regardless, these findings highlight the need for better dissemination of radiologic resectability criteria, and to have real-time central radiology reviews in neoadjuvant treatment trials for patients with resectable pancreatic cancer, to ensure that a homogeneous population is being evaluated.
In summary, the present study did not demonstrate an improved OS with perioperative chemotherapy, compared with historical data from adjuvant trials in resectable pancreatic cancer. We demonstrate the feasibility of multidisciplinary treatment using perioperative chemotherapy for patients with resectable pancreatic cancer. We establish key clinical outcome metrics in this setting, using the 2 most active chemotherapy regimens for pancreatic cancer. We also reveal the challenges in quality control that serve as lessons for the future conduct of trials in this setting. These findings serve as a platform for studying novel therapies for patients with neoadjuvant/perioperative resectable pancreatic cancer in future trials.
Accepted for Publication: October 28, 2020.
Published Online: January 21, 2021. doi:10.1001/jamaoncol.2020.7328
Correction: This article was corrected on September 23, 2021, to correct an error in the Visual Abstract.
Corresponding Author: Davendra P. S. Sohal, MD, MPH, Division of Hematology and Oncology, University of Cincinnati, 3125 Eden Ave, ML 0562, Vontz Center, Room 1302, Cincinnati, OH 45267 (firstname.lastname@example.org).
Author Contributions: Dr Guthrie and Ms Duong had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Sohal, Ahmad, Gandhi, Beg, Guthrie, Lowy, Philip, Hochster.
Acquisition, analysis, or interpretation of data: Sohal, Duong, Ahmad, Beg, Wang-Gillam, Wade, Chiorean, Guthrie, Lowy, Philip, Hochster.
Drafting of the manuscript: Sohal, Duong, Ahmad, Chiorean, Guthrie, Hochster.
Critical revision of the manuscript for important intellectual content: Sohal, Ahmad, Gandhi, Beg, Wang-Gillam, Wade, Chiorean, Guthrie, Lowy, Philip, Hochster.
Statistical analysis: Sohal, Duong, Beg, Wang-Gillam, Guthrie.
Obtained funding: Sohal, Hochster.
Administrative, technical, or material support: Beg, Wade, Chiorean, Lowy, Hochster.
Supervision: Sohal, Ahmad, Beg, Guthrie, Philip, Hochster.
Other—central radiology review for eligibility: Gandhi.
Conflict of Interest Disclosures: Dr Sohal reported personal fees from Perthera, Incyte, and Ability Pharma. Dr Beg reported personal fees from AstraZeneca, Merck, Array, and Ipsen during the conduct of the study. Dr Chiorean reported personal fees from Celgene during the conduct of the study; and personal fees from Array, Ipsen, Legend, and Sobi, and grants from Boehringer Ingelheim, Halozyme, Merck, Stemline, Roche, MacroGenics, Fibrogen, Rafael, and Clovis outside the submitted work. Dr Lowy reported personal fees from HUYA, Merck, and Rafael, and grants from Tanabe Mitsubishi and Syros outside the submitted work. Dr Philip reported personal fees from Celgene honoraria outside the submitted work. No other disclosures were reported.
Funding/Support: National Institutes of Health, National Cancer Institute grants CA180888, CA180819, CA180820, CA180821, CA189830, CA180801, CA189953, CA189957, CA239767, CA189821, CA189972, CA233230, CA189858, CA189958, CA189822, CA189848, CA189971, CA13612, CA189873, CA189856, CA180798, CA189861, and CA189954.
Role of the Funder/Sponsor: The National Institutes of Health had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Data Sharing Statement: See Supplement 3.