Improving eligibility criteria for first-line trials for patients with DLBCL using a US-based Delphi-method survey

Diffuse large B-cell lymphoma (DLBCL) is the most common adult non-Hodgkin lymphoma (NHL) and is clinically heterogeneous.^1,2  Although 60% of patients with DLBCL who receive standard first-line treatment with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) will experience a cure,³  the remaining 40% exhibit a median overall survival under 1 year, indicating significant unmet treatment needs for patients in poorest-risk groups.⁴  Recent randomized controlled clinical trials (RCTs) designed to improve first-line therapy for DLBCL have shown consistent negative results^5-11  that may be due in part to time-intensive assessment of eligibility status limiting enrollment of patients who require urgent therapy.¹² 

A key observational study showed that shorter diagnosis-to-treatment interval (DTI) is strongly associated with inferior event-free survival in newly diagnosed DLBCL.¹²  In the discovery cohort (n = 986 patients prospectively enrolled in the University of Iowa and Mayo Clinic Molecular Epidemiology Resource) and validation cohort (n = 1444 patients prospectively enrolled in the Lymphoma Study Association LNH-2003 clinical trials program), the median DTIs were 14 days (range, 0-155 days) and 23 days (range, 0-215 days), respectively. In both cohorts, patients with a DTI <14 days experienced reduced 24-month event-free survival when compared with patients with DTI >14 days. In the multivariable logistic regression model involving both cohorts, longer DTI predicted improvement in 24-month event-free survival, independent of clinical factors in the international prognostic index (IPI), with per-week odds ratios of 0.83 (95% confidence interval, 0.76-0.91; P < .001) and 0.93 (95% confidence interval, 0.88-0.98; P < .001) in the discovery and validation cohorts, respectively, in IPI-adjusted models.

This finding has implications for bias in DLBCL RCTs as patients who require rapid treatment are less likely to undergo the often time-consuming steps necessary to determine trial eligibility. Somewhat paradoxically, this results in reduced trial participation by the very patients who are most in need of improved treatment approaches: those with inferior outcomes on standard therapy. For instance, most recent RCTs for patients with DLBCL have described DTIs much longer than 14 days (eg, the ROBUST trial, which reported a median DTI of 31 days).¹¹  When study populations are enriched for patients with longer DTI despite the use of poor-risk IPI eligibility criteria, we should expect improved survival in the control arms as was seen in US and European patients treated with R-CHOP,¹²  which may contribute to negative study results overall.^5-11  Even the first-line POLARIX trial that met its primary endpoint had a mean DTI of 26 days.¹³ 

Efforts to revise trial enrollment processes and reduce DTIs among study populations have significant potential to limit bias in modern first-line DLBCL RCTs and increase the likelihood of observing true effects of novel study drugs. In the present study, we aimed to identify common and uncommon enrollment criteria from first-line DLBCL RCTs in the R-CHOP era and to develop consensus expert recommendations for modernized eligibility criteria in the interest of enrolling clinically diverse study populations that include patients with poor-risk disease in first-line DLBCL RCTs.

We selected DLBCL RCTs representative of first-line clinical trials in the R-CHOP era. We identified studies that ranged from initial RCTs investigating R-CHOP in the first-line treatment of DLBCL through recent RCTs examining R-CHOP plus targeted therapy. We prioritized RCTs for inclusion based on access to study protocols through prior work¹⁴  or that were available as published supplemental materials. We selected this timeframe in the interest of assessing the full range of eligibility criteria in first-line clinical trials during the R-CHOP era and to capture criteria that may have been included in earlier studies and carried forward into recent trials despite lack of modern clinical indication. We tabulated inclusion and exclusion criteria from analyzed studies using study protocols, study publications, and information available on clinicaltrials.gov and calculated the average number of enrollment criteria per trial. After tabulating enrollment criteria, we further categorized enrollment criteria from selected studies into discrete criterion categories (eg, individual eligibility criteria pertaining to creatinine thresholds were categorized under the criterion category “renal function”). Common criterion categories were defined as those present in greater than or equal to two-thirds of RCTs, moderately common criterion categories were present in greater than or equal to one-third and fewer than two-thirds of RCTs, and uncommon criterion categories were present in fewer than one-third of RCTs.

We conducted a Delphi-method survey^15,16  with a panel of experts in lymphoma clinical trial design to identify consensus-essential and consensus-unnecessary eligibility criteria for first-line DLBCL RCTs. Survey participants were asked to complete a Delphi-method questionnaire regarding recommendations for modernized eligibility criteria with the specified aim of reducing DTI and increasing enrollment of patients with poorest-risk disease while maintaining patient safety in first-line DLBCL RCTs. Invited survey participants included clinical investigators and experts in biostatistics involved in the Lymphoma Epidemiology of Outcomes Cohort Study representing 8 academic medical centers in the United States. We derived eligibility criteria for inclusion in the survey based on common and moderately common eligibility criterion categories as tabulated from the prior DLBCL RCTs in order to focus expert discussion on eligibility criteria that were frequently included in first-line DLBCL RCTs in the R-CHOP era. We omitted criteria from inclusion in the survey that we believed were necessary for patient safety and would not foster discussion during the Delphi-method survey. We asked prospective survey participants to recommend additional pertinent eligibility criteria for inclusion in the survey that were not already represented after tabulation from analysis of the prior first-line DLBCL RCTs. See supplemental Figure 1 for a flowchart illustrating the criterion category selection process for inclusion in the Delphi-method survey.

The survey consisted of 3 rounds and was conducted via e-mail using Google Forms. In the first round, we asked survey participants to rate the importance of each eligibility criterion using a 1 to 9 Likert-style scale (1 = unnecessary for inclusion in future RCTs; 9 = essential). Survey participants were able to see a list of prospective survey respondents but did not have access to current responses of other members of the survey group when completing questionnaires. In addition, the first-round survey requested the total number of years each participant had worked in hematology/oncology. Based on first-round results, we designated each eligibility criterion as essential, unnecessary, unresolved, or in disagreement. Essential criteria had a median Likert value ≥7. Unnecessary criteria had a median value ≤3. Criteria were in disagreement if greater than one-third of respondents selected a value ≥7 and greater than one-third selected a value ≤3. Unresolved criteria showed median values from 4 to 6 and were not in disagreement. Statistical analyses related to survey results included median, range, and interquartile range (IQR) values for number of years’ experience in hematology/oncology; frequency, median, and IQR values for Likert values by eligibility criterion; and percentage of criterion categories designated as essential, unnecessary, unresolved, or in disagreement. Respondents received a personalized summary of round-1 results including median values and IQRs for all included criteria, histograms showing the ranking value distribution for each criterion, indication of the participant’s personal Likert response by criterion, and comments from survey participants for each eligibility criterion. All survey responses including Likert rankings and comments were anonymized in personalized summaries received by survey participants.

In the second round, we invited all round-1 survey participants to reassess eligibility criteria that were unresolved or in disagreement after round 1. Round 2 employed the methods used in round 1 and requested recommendations for threshold values for quantitative criteria. Survey options for quantitative threshold values were selected from threshold values used in the analyzed DLBCL RCTs from the R-CHOP era. The second-round survey also included free-text fields for survey respondents to provide context for their recommended threshold values or to suggest other threshold values. Statistical analyses conducted after round 2 included repeating analyses conducted following the round-1 survey as well as calculating the magnitude of change between round-1 and round-2 Likert ratings by criterion (ie, round-2 Likert value minus round-1 Likert value) by respondent for each respondent who participated in both round 1 and round 2. In addition, we calculated the frequency of quantitative thresholds selected by the expert panel for quantitative eligibility criteria. Based on survey results, we developed preliminary recommendations for eligibility criteria. A final survey asked all round-1 survey participants whether they agreed or disagreed with each preliminary recommendation and to provide additional context or explanation for their response. We developed finalized recommendations for eligibility criteria based on survey results. All authors reviewed our finalized recommendations for consensus, modernized eligibility criteria in first-line DLBCL RCTs.

We performed statistical analyses using R version 3.6.2 or greater.

We selected 19 DLBCL RCTs for analysis that spanned the R-CHOP era (Table 1). Notably, this study was designed and initiated before the authors had access to eligibility criteria from the REMoDL-B¹⁷  and GOYA¹⁸  trials (see “Discussion” and supplemental data for further analysis incorporating these RCTs). Across all trials, we tabulated 451 total enrollment criteria, with an average of 23.7 criteria per study (standard deviation, 6.3; range 14 to 37 criteria). Analysis of the 451 tabulated enrollment criteria revealed 51 discrete criterion categories (Table 2) across the 19 RCTs including 18 common, 11 moderately common, and 22 uncommon criterion categories.

We derived 30 total eligibility criterion categories for inclusion in the survey from common and moderately common criterion categories in the 19 DLBCL RCTs. From 29 common and moderately common criterion categories, 3 criterion categories were expanded for inclusion in the survey as follows: “histology” was expanded to include cell-of-origin (COO) subtype and CD20 positivity; “female reproductive concerns” was expanded to include pregnancy status, breastfeeding status, and contraception or abstinence; and “other neurologic pathology” was expanded to include peripheral neuropathy and history of stroke or intracranial hemorrhage. We omitted 3 criteria that we believed were necessary for patient safety (clinical contraindications to study therapy, known allergy or hypersensitivity reaction to study drugs, and whether the patient is receiving or has received contraindicated therapies that would preclude receiving study therapy). Prior to initiation of the first-round survey, 1 additional eligibility criterion category, “central pathology review prior to enrollment,” was proposed for inclusion in the questionnaire by a prospective survey participant for a total of 31 criterion categories included in the survey.

Among 29 clinical investigators invited to participate in the Delphi-method survey, 17 respondents including physicians and experts in biostatistics representing 8 academic medical centers in the United States completed the round-1 survey for a response rate of 58.6%. The median number of years’ experience in hematology/oncology among respondents was 17 years (IQR, 12 years; range, 3-30 years). The first round examined the previously derived 31 eligibility criterion categories. After the first-round survey, 12 of 31 criteria (39%) qualified as essential, 9 (29%) qualified as unnecessary, 2 (6%) showed disagreement, and 8 (26%) remained unresolved (Figure 1A). Essential criteria with the highest median Likert-style values included pregnancy status, breastfeeding status, and whether the patient was already receiving an investigational drug. Unnecessary criteria with the lowest median rankings included minimum life expectancy, HCV status, and central pathology review prior to enrollment. See supplemental data for results of all survey rounds.

Fifteen of 17 first-round respondents participated in the second round (response rate, 88.2%). Participants reassessed 10 criteria that had shown disagreement or remained unresolved. Based on round-2 results, 1 additional criterion, performance status, was deemed essential (Figure 1B). Although study participants commonly changed their Likert-style rankings between rounds 1 and 2 of the survey as illustrated in round-2 survey results available in supplemental data, the remaining 9 criteria continued to show disagreement or remain unresolved. Survey participants provided recommendations for threshold values for 10 quantitative eligibility criteria including measurements of renal, hepatic, and cardiac function, threshold levels for platelet count and white blood cell count, and recommended dimensions for measurable disease on imaging.

Twelve of 17 first-round participants responded to the final survey regarding preliminary recommendations (response rate, 70.6%). Across all criteria, the average rate of agreement among lymphoma experts in support of preliminary recommendations was 90.4% (standard deviation, 9.7%). Nine preliminary recommendations showed 100% agreement across survey respondents, and 23 of 31 criteria had ≥90% agreement. See supplemental data for data regarding agreement among experts with preliminary recommendations by criterion. Panel experts established final recommendations for all eligibility criteria assessed in prior survey rounds (Table 3).

We refined eligibility criteria for frontline DLBCL RCTs utilizing a Delphi-method survey that incorporated recommendations from a multi-institution panel of lymphoma experts from academic medical centers in the United States. We defined consensus recommendations for 31 eligibility criteria including threshold values for 10 quantitative criteria, and we identified 13 essential and 9 unnecessary criteria from criteria common to DLBCL RCTs in the R-CHOP era. Notably, our survey group liberalized or identified as unnecessary multiple criteria that were frequently included in prior DLBCL RCTs, emphasizing the importance of updating preenrollment assessments that became commonplace over time but are no longer clinically indicated in modern RCTs. We believe that our recommendations for updated threshold values and for eliminating criteria deemed currently unnecessary by the survey group can streamline trial enrollment and shorten DTI. This approach aligns with recommendations from national panels of experts for enhancing enrollment in cancer clinical trials.^19-21  Further, survey results address multiple issues we believe are pertinent to reduce DTI in DLBCL RCTs including eliminating central pathology review prior to enrollment and allowance of prephase corticosteroids or bridging therapy prior to administering study drugs. Final recommendations for eligibility criteria showed strong agreement among survey respondents and are readily applicable in frontline DLBCL trials. It is our hope that the consensus recommendations included in this study will streamline the enrollment process for first-line DLBCL RCTs, shorten DTI for ensuing study populations, and increase the likelihood of identifying true effects of novel drugs for first-line treatment of DLBCL.

Notably, although survey respondents designated 22 eligibility criteria as essential or unnecessary, 9 criteria remained either unresolved or showed disagreement, illustrating the challenges inherent in reaching full consensus among experts in clinical trial design. Review of survey results shows that essential criteria generally pertained to safety of study drugs (eg, pregnancy and breastfeeding status), prognostic factors (eg, IPI score, age at diagnosis, and Ann Arbor stage), and patient-specific clinical factors including renal and hepatic function. Regarding IPI score, it should be noted that the survey group advised inclusion of IPI score or elements of the IPI score in eligibility criteria at this time given that patients with low IPI score have favorable outcomes with standard therapies, allowing for selection of patients in greatest need of trials. However, survey respondents emphasized that a better understanding of DLBCL biology may someday allow us to move beyond IPI score as an eligibility criterion. Unnecessary criteria included comorbid disease states that may warrant evaluation as part of routine clinical practice but are now clinically manageable during trial participation and should not preclude eligibility (eg, HIV, HBV, and HCV status). Similarly, preenrollment steps that would slow enrollment and increase DTI (eg, central pathology review prior to trial participation) were deemed unnecessary. Regarding eligibility criteria that remained unresolved or showed disagreement, the survey group commonly concluded that assessing the importance of these criteria required additional, case-by-case information. The additional information required may depend on the purpose of a given study, with the survey group advising, for example, that assessment of COO subtype should be included as an eligibility criterion only if the study drug targets COO subtypes. Alternately, the additional needed information may relate to the investigational therapy in question; for example, the survey group recommended that inclusion of platelet count thresholds should depend on whether the study drug is known to cause thrombocytopenia. Lastly, the required case-by-case information may pertain to the medical history of the patient being evaluated for inclusion. For example, the survey group concluded that a patient’s underlying cardiac function is pertinent in any study that includes an anthracycline, but assessment of left ventricular ejection fraction for all prospective patients prior to trial participation regardless of clinical indication may unnecessarily increase DTI. Criteria that remained unresolved or showed disagreement also involved factors influencing R-CHOP dose reduction or conditions likely to be exacerbated by R-CHOP toxicity. There was not investigator consensus regarding whether these criteria should be requirements for eligibility or applied using clinical judgment as physicians determine which patients to consider for enrollment. Finalized recommendations for eligibility criteria account for the added nuance of criteria that were unresolved or showed disagreement and can be readily applied to modern first-line DLBCL RCTs.

Multiple recent studies describe the importance of modernizing eligibility criteria for clinical trials in oncology including analyses highlighting criteria pertaining to minimum age threshold,²¹  comorbid organ dysfunction,²²  and HIV status.²³  In addition, recent examinations focused on DLBCL in particular indicate that organ-function based eligibility criteria in DLBCL RCTs limit inclusion of patients with poorest-risk disease^24,25  and that DLBCL trial eligibility criteria have become increasingly restrictive over time, likely limiting recruitment.²⁶  Our work addresses this urgent research gap by updating and streamlining enrollment criteria with important ramifications for inclusion of poor-risk patients in first-line DLBCL RCTs. It is our hope that by focusing on first-line DLBCL RCTs, our study will build on these prior analyses to streamline enrollment specific to first-line DLBCL clinical trials.

Limitations of our study include a round-1 response rate of 58.6% with 17 out of 29 invited experts responding to the initial questionnaire. A systematic review of Delphi-method studies showed that the median number of individuals invited to participate across 80 analyzed Delphi-method studies was 17 prospective participants,¹⁵  indicating that the total number of respondents in our survey was greater than the median number of respondents for studies of this kind. Nonetheless, the round-1 response rate and sample size may limit generalizability of study results. Response rates for subsequent rounds of the survey showed sustained participation among panel experts, with 88.2% and 70.6% of e-mailed survey participants replying to the round-2 and preliminary-recommendations questionnaires, respectively. Additionally, it should be noted that invited respondents in the present study were limited to physicians and experts in biostatistics representing major academic medical centers in the United States. The selection of experts for the survey panel has an important impact on survey results, and a similar study conducted with different stakeholders (eg, representatives of community health centers, ethics committees, or pharmaceutical companies) or conducted in a different region (eg, outside the catchment area of major academic medical centers or outside the United States) may reach different consensus recommendations for eligibility criteria in first-line DLBCL RCTs. However, our survey respondents represent a diverse group of lymphoma clinical trials research experts with considerable depth and breadth of clinical research experience, which achieves a key objective for enhancing enrollment in clinical trials for diverse populations based on prior recommendations.^19,27  The scope of the present study was limited to identifying consensus recommendations for eligibility criteria from expert clinical investigators in the United States, and survey results may not be representative of opinions, perspectives, or local issues relevant to trial design in other areas. Recommendations from additional stakeholders could enrich the discussion regarding modern eligibility criteria in first-line DLBCL RCTs.

Strengths of the present study include the incorporation of expert opinion from clinical investigators with significant experience in first-line DLBCL RCTs to define streamlined essential and unnecessary eligibility criteria. The expert panel included senior clinical investigators; Lymphoma Committee chairs from SWOG, ECOG-ACRIN, and the Alliance; and members and chairs from the National Cancer Institute Lymphoma Steering Committee. Additionally, survey results exhibited high agreement rates for recommendations among lymphoma experts across proposed eligibility criteria (see supplemental data for concordance rates for preliminary criteria prior to development of finalized recommendations). Lastly, incorporation of inclusion and exclusion criteria from 19 DLBCL clinical trials from the R-CHOP era places our methods in context with pivotal trials spanning the relevant history of DLBCL RCTs. We considered whether these methods overemphasized eligibility criteria that were included early in the R-CHOP era but were excluded from more recent studies. When restricting analyzed trials to recent DLBCL RCTs that examined targeted therapies and included R-CHOP in the control arm, categorization of common and moderately common criteria yielded largely similar results when compared with criterion categories incorporating all 19 analyzed studies (supplemental Table 1). We also considered whether inclusion of additional first-line DLBCL RCTs from the R-CHOP era would have altered our results and yielded different survey questions in the Delphi-method survey. We examined inclusion and exclusion criteria from 2 DLBCL RCTs not included in analysis, REMoDL-B¹⁷  and GOYA,¹⁸  and reassessed common, moderately common, and uncommon eligibility criterion categories with the addition of these RCTs (supplemental Table 2). Eligibility criteria from both studies showed considerable overlap with common criteria from the previously analyzed 19 DLBCL RCTs, indicating that our survey was representative of eligibility criteria from first-line DLBCL RCTs in the R-CHOP era. With the REMoDL-B and GOYA studies included, 2 criterion categories, pulmonary function and recent surgical history, changed from uncommon to moderately common criterion categories and thereby would have been included in the Delphi-method survey if these additional studies had been incorporated in analysis. All other common and moderately common criterion categories remained the same with inclusion of the REMoDL-B and GOYA studies, yielding no further changes to the Delphi-method survey. It should be noted that the majority of RCTs included in analysis that incorporated assessment of pulmonary function in enrollment criteria stipulated exclusion based on substantial pulmonary impairment or uncontrolled chronic obstructive pulmonary disease, which were arguably captured in our Delphi-method survey under the criterion category “presence of other significant, uncontrolled, concomitant disease at investigator’s discretion.”

We believe that the recommended eligibility criteria included in this study are ready for application to novel first-line DLBCL RCTs. S1918 (NCT 04799275), a phase 2/3 RCT focused on treatment of newly diagnosed DLBCL and other lymphomas in patients ages 75 or older, is currently recruiting patients and adheres to many recommendations included in the present study, illustrating the potential of finalized recommendations to help guide development of future DLBCL clinical trials. Next steps for this work beyond direct application in modern DLBCL RCTs include combining recommended eligibility criteria with large DLBCL patient data sets to analyze demographic and outcomes data of prospective patient cohorts defined using proposed criteria. A recent study examining the impact of trial eligibility criteria on outcomes in a nationwide cohort of patients with newly diagnosed DLBCL suggests that real-world data can be used to evaluate the impact of eligibility criteria on prospective first-line DLBCL study populations.²⁵  This would enable modification of recommended criteria as needed prior to development of future studies and could further increase the likelihood of enrolling a more clinically diverse patient population that includes poor-risk groups. Additionally, application of recommended eligibility criteria to DLBCL data sets that incorporate patient-level genetic characteristics would allow for anticipation of ensuing genetic cohorts when using finalized eligibility criteria and would facilitate development of future clinical trials incorporating novel targeted therapies. Although we believe that modernizing eligibility criteria is an important step toward increasing enrollment of poor-risk groups, additional research is needed in other avenues that may address DTI and barriers to trial participation including efforts to increase access, engagement of community sites, timely drug delivery, streamlined molecular testing when applicable, selection of time windows for meeting screening eligibility criteria, baseline imaging modalities, and other factors beyond the scope of the present study.

Applying modernized eligibility criteria has important implications for reducing the number and nature of preenrollment evaluations for all patients, enabling participation of patients who require urgent therapy in particular, and increasing generalizability of study results for first-line DLBCL RCTs. Given the identified impact of DTI on enrollment of poor-risk patients in first-line DLBCL trials,¹¹  it is our hope that application of our expert consensus recommendations for eligibility criteria in first-line DLBCL RCTs will facilitate enrollment of a more clinically diverse patient population and increase generalizability of study results while maintaining patient safety.

This work was supported by grants from the National Center for Advancing Translational Sciences of the National Institutes of Health (NIH) (grants TL1TR002382 and UL1TR002378) (R.A.H.). C.R.F. is a Cancer Prevention and Research Institute of Texas (CPRIT) Scholar of Cancer Research. The project described was supported in part by the National Cancer Institute (NCI) (K24 CA208132 award for mentored patient-oriented research in lymphoma) and CPRIT Award RR190079 (C.R.F.). Research reported in this publication was supported in part by the Biostatistics Shared Resource of Winship Cancer Institute of Emory University and NIH/NCI under award number P30CA138292.

Contribution: R.A.H. analyzed results and made tables and figures; S.P.P. and M.J.L. assisted in data collection; J.M.S. assisted in statistical analysis; S.M.A., N.L.B., K.A.B., A.F.C., C.C., J.W.F., P.B.J., B.S.K., J.P.L., B.K.L., I.S.L., P.M., M.J.M., N.M.-S., P.M.R., J.R.W., and C.R.F. developed consensus recommendations and edited the manuscript; J.L.K. wrote and edited the manuscript; and R.A.H. and C.R.F. designed the research and wrote the manuscript.

Conflict-of-interest disclosure: C.C. receives research funding from Gilead, BMS, Genentech, and Verastem. J.W.F. discloses data safety monitoring committee participation for Acerta and Novartis. B.S.K. receives research funding from Genentech, ADC Therapeutics, Abbvie, AstraZeneca, and Beigene and provides consulting for Genentech, ADC Therapeutics, Abbvie, AstraZeneca, BeiGene, Pharmacyclics, Celgene/BMS, TG Therapeutics, Teva, Janssen, and MEI. J.P.L. provided consulting advice to Sutro, Epizyme, BMS/Celgene, Bayer, Gilead/Kite, GenMab, Genentech/Roche, Abbvie, Incyte, Janssen, Eisai, Mustang Bio, and Second Genome, and received research support from Genentech. I.S.L. has served on the advisory boards of Seattle Genetics, Janssen Scientific, and Verastem. P.M. is a consultant for ADCT, AstraZeneca, Bayer, Beigene, BMS, Cellectar, Epizyme, Gilead, Incyte, Janssen, Karyopharm, Morphosys, Regeneron, and Verastem. M.J.M. discloses consulting or advisory roles with MorphoSys, Kite Pharma, and Pfizer and research funding from Celgene, NanoString Technologies, Genentech, and Morphosys. P.M.R. has consultancy for Kite Pharma and research funding from SeaGen and Genentech. J.R.W. discloses consulting work with Novartis, Kite/Gilead, BMS, Morphosys, AstraZeneca, Janssen, Genentech, Iksuda, Umoja, and ADC Therapeutics and research funding from Novartis, Kite/Gilead, BMS, Morphosys, AstraZeneca, Janssen, Genentech, Curus, Unum, and ADC Therapeutics. J.L.K. has participated on advisory boards for Janssen, MorphoSys, TG Therapeutics, and Gamida Cell and received research funding from Oncternal Therapeutics, Viracta Therapeutics, and Atara Biotherapeutics. C.R.F. is a consultant for Abbvie, Bayer, BeiGene, Celgene, Denovo Biopharma, Genentech/Roche, Genmab, Gilead, Karyopharm, Pharmacyclics/Janssen, SeaGen, and Spectrum and has received research funding from 4D, Abbvie, Acerta, Adaptimmune, Allogene, Amgen, Bayer, Celgene, Cellectis, EMD, Gilead, Genentech/Roche, Guardant, Iovance, Janssen Pharmaceutical, Kite, Morphosys, Nektar, Novartis, Pfizer, Pharmacyclics, Sanofi, Takeda, TG Therapeutics, Xencor, Ziopharm, Burroughs Wellcome Fund, Eastern Cooperative Oncology Group, National Cancer Institute, V Foundation, and Cancer Prevention and Research Institute of Texas as the CPRIT Scholar in Cancer Research. N.M.-S. has institutional research support from Genentech/Roche, Bristol Myers Squibb/Celgene, AstraZeneca, Innate Pharmaceuticals, Corvus Pharmaceuticals, C4 Therapeutics, Secura Bio, Verastem, and Daiichi Sankyo and has served as a consultant for Daiichi Sankyo, Secura Bio, Kiowa Hakka Kirin, Ono Pharmaceuticals, C4 Therapeutics, and Karyopharm.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. R.A.H. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Correspondence: R. Andrew Harkins, Department of Internal Medicine, Emory University School of Medicine, 49 Jesse Hill Dr. SE, Suite 491, Atlanta, GA 30303; e-mail: andrew.harkins@emory.edu.