Characteristics of Real‐World Metastatic Non‐Small Cell Lung Cancer Patients Treated with Nivolumab and Pembrolizumab During the Year Following Approval

This study evaluated real‐world usage patterns of the PD‐1 inhibitors nivolumab and pembrolizumab in metastatic non‐small cell lung cancer patients in the first year following regulatory approval of these therapies in this indication. The analysis revealed how the real‐world cohort differed from the clinical trial cohorts, which may inform the design of future studies to understand outcomes of underrepresented populations.

Abstract

Background.

Evidence from cancer clinical trials can be difficult to generalize to real‐world patient populations, but can be complemented by real‐world evidence to optimize personalization of care. Further, real‐world usage patterns of programmed cell death protein 1 (PD‐1) inhibitors following approval can inform future studies of subpopulations underrepresented in clinical trials.

Materials and Methods.

We performed a multicenter analysis using electronic health record data collected during routine care of patients treated in community cancer care clinics in the Flatiron Health network. Real‐world metastatic non‐small cell lung cancer (NSCLC) patients who received nivolumab or pembrolizumab in the metastatic setting (n = 1,344) were selected from a starting random sample of 55,969 NSCLC patients with two or more documented visits from January 1, 2011, through March 31, 2016. The primary study outcome measurement was demographic and treatment characteristics of the cohort.

Results.

Median age at PD‐1 inhibitor initiation was 69 years (interquartile range 61–75). Patients were 56% male, 88% smokers, 65% nonsquamous histology, and 64% diagnosed at stage IV. Of 1,344 patients, 112 (8%) were tested for programmed death‐ligand 1 expression. Overall, 50% received nivolumab or pembrolizumab in the second line, with a substantial proportion of third and later line use that began to decline in Q4 2015.

Conclusion.

During the year following U.S. regulatory approval of PD‐1 inhibitors for treatment of NSCLC, real‐world patients receiving nivolumab or pembrolizumab were older at treatment initiation and more had smoking history relative to clinical trial cohorts. Studies of outcomes in underrepresented subgroups are needed to inform real‐world treatment decisions.

Implications for Practice.

Evidence gathered in conventional clinical trials used to assess safety and efficacy of new therapies is not necessarily generalizable to real‐world patients receiving these drugs following regulatory approval. Real‐world evidence derived from electronic health record data can yield complementary evidence to enable optimal clinical decisions. Examined here is a cohort of programmed cell death protein 1 inhibitor‐treated metastatic non‐small cell lung cancer patients in the first year following regulatory approval of these therapies in this indication. The analysis revealed how the real‐world cohort differed from the clinical trial cohorts, which will inform which patients are underrepresented and warrant additional studies.

Introduction

Conventional cancer clinical trials have been the cornerstone of evidence generation for assessment of the safety and efficacy of therapies, and approval decisions by regulatory authorities. Standard practices such as narrowly defined inclusion criteria are among the methods used to increase internal validity of these studies, often at the expense of external validity and generalizability. As a consequence, following regulatory approval, characteristics of cancer patients treated in the real world can be very different from those in clinical trials [1].

Real‐world evidence (RWE) can be used as a complementary evidence source for traditional clinical trials to expand external validity and generate additional evidence for optimal management of individual patients at the point of care [2]. Information collected as a routine part of medical care and stored in electronic health records (EHRs) is one source of real‐world data that, when of adequate quality and completeness, can be a reliable data source for generating RWE.

The absence of generalizable evidence for tailoring therapies to the individual characteristics of patients can be a barrier to the practice of precision oncology, especially in cases in which multiple competing therapies and lack of validated predictive markers can complicate evidence‐based treatment decisions. Even when there is a strong predictive biomarker, sequencing of therapies in the real‐world setting is rarely informed by conventional clinical trials due to impracticalities of conducting such studies. Better information about the use, effectiveness, and safety of treatments in real‐world populations can help clinicians, regulators, and payers make informed decisions about delivering evidence‐based patient care.

In this retrospective observational study, we analyzed the use of nivolumab and pembrolizumab in a U.S. community practice cohort of metastatic non‐small cell lung cancer (NSCLC) patients following U.S. Food and Drug Administration (FDA) approval through March 2016. In March 2015, FDA approved nivolumab [3] for treatment of patients with metastatic squamous NSCLC with progression on or after platinum‐based chemotherapy [4], [5] and, subsequently, nonsquamous NSCLC in October 2015 [6, 7]. Pembrolizumab was also approved [8] in October 2015 for second‐line treatment of patients with metastatic programmed death‐ligand 1 (PD‐L1)‐positive NSCLC, as determined by an FDA‐approved test [9], [10] (and a year later for first‐line treatment in this setting [11]). Given the differences in the approved indications of these therapies, our study aimed to provide insights into clinical decisions associated with the use of pembrolizumab and nivolumab for the treatment of patients with metastatic NSCLC (mNSCLC) in the real‐world setting, including patient selection and PD‐L1 testing strategy. Widespread use of this class of drugs further presents an opportunity to compare characteristics of patients in clinical trial cohorts with patients who received these drugs following approval in routine community oncology practice. Beginning to define these differences is a useful first step toward optimizing the use of PD‐1 inhibitors.

Our analyses focused on descriptive characteristics, including age, race, line of therapy, and treatment sequence. This study is part of an ongoing effort to investigate the connection between internal and external validity of conventional cancer clinical trials and to identify methods based on RWE to optimize clinical decisions regarding administration of PD‐1 inhibitors to improve patient outcomes.

Materials and Methods

Database Description

Deidentified data from the Flatiron Health longitudinal EHR database were used for this study. At the time of dataset generation, the database represented a diverse group of over 260 community cancer clinics, ranging from small practices to large multicenter practices, with more than 1.6 million active U.S. cancer patients. Data were gathered in a manner that was agnostic to the source EHR, processed and harmonized centrally by Flatiron Health, and stored in a secure, Health Insurance Portability and Accountability Act of 1996‐compliant manner. Patients were selected based on review of unstructured data, which results in a more accurate cohort selection approach [12] compared with traditional ICD‐code‐based methods. Identified patients were generally similar to population‐based registries such as Surveillance, Epidemiology, and End Results Program and National Program of Cancer Registries (unpublished data).

To prepare EHR data for analysis, aggregate records were processed as follows. Structured data (e.g., laboratory test results, information on prescribed drugs) were harmonized and normalized to a standard ontology. Unstructured data (e.g., radiology reports, pathology reports, medical care notes, some biomarker tests) were extracted from EHR‐based digital documents via technology‐enabled chart abstraction. Every data point sourced from unstructured documents was manually reviewed by trained chart abstractors (clinical oncology nurses and tumor registrars, with oversight from oncologists). These processed data were deidentified with third‐party statistical verification and stored in a separate analytic database.

Institutional Review Board (IRB) approval of the study protocol was obtained. Informed consent was waived by the IRB, as this was a noninterventional study using deidentified, routinely collected data.

Data Quality Control

Rigorous quality control (QC) for structured and unstructured data was conducted, including duplicate chart abstraction of at least 10% of critical abstracted variables. QC was performed on all key variables related to patient demographics, disease characteristics, and treatment decisions. Inter‐rater reliability scores were calculated to ensure the integrity of data abstraction.

Patients

Patients with NSCLC with at least two visits to a community‐based cancer clinic in the Flatiron network, treated with nivolumab or pembrolizumab, from January 1, 2011, to March 31, 2016, were included in this study. All patients were confirmed, via review of unstructured data including pathology reports, to have been diagnosed with mNSCLC, or early stage NSCLC with a recurrence or progression, diagnosed on or after January 1, 2011 (Fig. 1).

Figure 1.

Patient selection diagram.

Abbreviations: ICD, International Classification of Diseases; NSCLC, non‐small cell lung cancer.

Exclusion criteria included incomplete historical treatment data within the Flatiron Health database (i.e., patients whose advanced diagnosis date was more than 90 days before they entered the Flatiron Health Database cohort).

Patients not eligible for the study cohort (n = 21,741; Fig. 1) comprised the following: (a) patients who passed away or were referred to hospice prior to FDA approval of the first checkpoint inhibitor for NSCLC (note that the database includes data for 4 years prior to FDA approval); (b) patients with >90 day gap between their advanced diagnosis date and their earliest visit (2,889 patients as noted in the cohort selection diagram); (c) patients who were receiving a different therapy and/or maintenance therapy, and may have received a PD‐1 inhibitor after the study window; and (d) patients whose physician chose not to try a checkpoint inhibitor.

Statistical Analysis

Patient demographic and clinical characteristics were collected from structured and unstructured data to describe the patient population. Characteristics included age at PD‐1 inhibitor initiation, sex, race, region, stage at NSCLC diagnosis, smoking status, histology, genetic biomarker testing at or prior to PD‐1 inhibitor initiation, duration of follow‐up, and visit frequency. Visits included any interaction with the oncology clinic, including treatment, lab, radiology, office, or other visit types.

Lines of therapy (LOTs) were derived based on prespecified algorithms according to antineoplastic usage recorded in the EHR, and indexed to the metastatic diagnosis date. Administration of therapy was defined as uncancelled medication orders and documented administrations. A “regimen” was defined as the name of a particular combination of drugs given in a single line and included all the drugs administered in that line. A line started with the initiation of a new regimen and ended when the patient switched to a subsequent treatment regimen; a gap in drug episodes of more than 120 days or the end of follow‐up also signaled the end of an LOT. The first treatment regimen was designated as the first LOT, and each subsequent change to a new treatment for any reason led to an incremental ordered increase in the LOT (second, third, etc.).

PD‐L1 expression data were abstracted from EHR biomarker testing reports, pathology reports, and oncology clinic visit notes by trained chart abstractors, with full traceability back to source documentation. Abstractors collected relevant testing dates, test‐type information, and PD‐L1 expression result data (including overall status and detailed staining data) for each PD‐L1 testing event. Data were abstracted exactly as reported in each chart; abstractors did not derive or interpret test results when the lab did not provide a clear interpretation. Values for PD‐L1 expression testing results were recorded as positive, negative, equivocal, no interpretation given, or unsuccessful/indeterminate test. Hence, documentation of “positive” on a PD‐L1 expression assay was according to the final results presented in the test report and signaled the information on which the treating clinician acted.

We conducted descriptive analyses on the data generated for this study. Categorical variables were reported as frequency and percentage, and continuous variables were reported as median and interquartile range. No tests of statistical significance were planned or performed. We compared patient and disease characteristics based on the real‐world cohort data reported here with those reported in cited nivolumab and pembrolizumab registrational trials.

Results

Real‐World Cohort Description

Characteristics of the selected cohort of 1,344 patients are described in Table 1. Median age at PD‐1 inhibitor initiation was 69 years, with the majority of patients aged over 65; 27% were aged 75 or older. Patients were more frequently male, distributed throughout the country, and predominantly smokers. Most patients were diagnosed at stage IV (64%) and had nonsquamous histology (65%). This population was generally similar to the overall cohort of advanced NSCLC patients in Flatiron's national database, with more patients with squamous histology and slightly more smokers.

Table 1. Characteristics of a cohort of 1,344 metastatic NSCLC patients who received nivolumab or pembrolizumab in the metastatic setting in U.S. community practices.

Age at PD‐1 inhibitor initiation ranged from 32 to 85, and ages over 85 were rolled down to 85 to prevent reidentification.

^aThis is defined as the first order or administration of nivolumab or pembrolizumab.

^bBiomarker status on or before the start of the first PD‐1 inhibitor line of therapy. In cases in which a patient had multiple tests for a particular biomarker, the result of the most recent successful test prior to the start of PD‐1 inhibitor therapy is displayed.

^cPD‐L1 status captures the interpretation provided in the test report, which is influenced by the reference range for that specific PD‐L1 test. Tests with no explicit interpretation or an equivocal result given in the report are grouped into “unsuccessful/indeterminate test.”

^dAmong 872 patients with nonsquamous histology, 134 (15.4%) patients were not tested for EGFR mutations, and 167 (19.2%) were not tested for ALK rearrangements.

^eStructured follow‐up time is calculated from the relevant time‐point for each patient until their last structured activity (i.e., most recent visit or administration).

^fVisits included treatment, lab, radiology, office, or other visit types.

^gPD‐1 activity as defined by orders or administrations.

^hOnly among patients with >1 order or administration of a PD‐1 inhibitor.

Abbreviations: EGFR, epidermal growth factor receptor; IQR, interquartile range; NOS, not otherwise specified; NSCLC, non‐small cell lung cancer; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1.

The biomarkers assessed in this cohort were epidermal growth factor receptor (EGFR), ALK, and PD‐L1. ALK and EGFR testing rates were 61% and 64%, respectively, with 1.47% of patients carrying an ALK rearrangement and 8.37% being EGFR mutation‐positive. Of patients not tested for EGFR mutations, 334 patients had squamous histology, and of those not tested for ALK rearrangement status, 344 had squamous histology. In this cohort, 112 patients (8%) were tested for PD‐L1 expression, and of those, 64 patients were tested with a lab‐developed test (57% of those tested and 5% of the overall cohort), 27 patients were tested with one of two FDA‐approved assays, and 21 patients were tested with an assay of unknown type. Of the tested patients, 49% were PD‐L1 “positive.”

Median follow‐up time from metastatic diagnosis was 10.6 months, with 2.3 months follow‐up from PD‐1 inhibitor initiation. Median visit frequency from metastatic diagnosis to the end of follow‐up was 3.0 visits per month (note that visits included treatment, lab, radiology, office, or other visit types, a maximum of one visit per day). During the PD‐1 inhibitor administration period (calculated only for patients with >1 administration; n = 1,146), median visit frequency was 3.2 visits per month during the administration period and 3.3 visits per month from the start of administration to the end of follow‐up.

Exploratory Comparison with Clinical Trial Populations

To determine how patients treated with PD‐1 inhibitors differ from those enrolled in relevant registrational trials, we compared available information from nivolumab and pembrolizumab clinical trial cohorts with patients in our community practice cohort treated with PD‐1 inhibitors, as well as with a previously published real‐world cohort of advanced NSCLC patients treated with any regimen [12] (Table 2). Patient and disease characteristics with small differences between real‐world and clinical trial populations were race and smoking status for most comparisons (no statistical comparisons were performed).

Table 2. Comparison of clinical trial cohorts to community practice cohorts.

^aPatients older than 85 had their age adjusted to 85 in the Flatiron dataset for deidentification purposes.

Abbreviations: —, not applicable; EGFR, epidermal growth factor receptor; NOS, not otherwise specified; NSCLC, non‐small cell lung cancer.

Real‐World PD‐1 Inhibitor Usage Patterns

Following approval of nivolumab in March 2015 for treatment of patients with metastatic squamous NSCLC in the second line, there was a striking increase in its use over a short time (Fig. 2A).

Figure 2.

Programmed cell death protein 1 (PD‐1) inhibitor usage over time in a cohort of 1,344 metastatic non‐small cell lung cancer patients. (A): Cumulative number of patients who had started nivolumab or pembrolizumab, by month. Patients who received both drugs in different lines (n = 4) were counted in both strata. (B): Lines of therapy in which the PD‐1 inhibitor was used, by quarter. Numbers in plots indicate the number of patients.

For this cohort, we determined whether nivolumab or pembrolizumab was the first PD‐1 inhibitor administered, in any line, and whether it was used alone or in combination with another drug. In this analysis through March 2016, 93% of patients received nivolumab alone as the first PD‐1 inhibitor, with an additional 3.4% receiving nivolumab along with another therapy (Table 3). In comparison, 3.2% received pembrolizumab alone, and one patient received pembrolizumab in combination with another therapy. Most of the therapies were received in the second line (50%), indexed to metastatic diagnosis date.

Table 3. Earliest line with nivolumab or pembrolizumab in the cohort of 1,344 metastatic non‐small cell lung cancer patients.

Abbreviations: L, line; PD‐1, programmed cell death protein 1.

Breaking down by the drugs administered in the earliest PD‐1 inhibitor line of therapy, the overall pattern was essentially replicated, with nivolumab as the only PD‐1 inhibitor administered in over 90% of patients in each line (Table 3). Overall, 22 patients (1.6%) received a PD‐1 inhibitor in multiple lines, and 4 of these patients had administrations of both nivolumab and pembrolizumab.

PD‐1 Inhibitor Therapy in Context of Other Treatments

To understand PD‐1 inhibitor usage in the context of treatment sequencing, we evaluated treatments in LOTs immediately following and immediately prior to earliest PD‐1 inhibitor administrations (Table 4 and Table 5). In the LOT immediately prior to the earliest PD‐1 inhibitors, platinum doublet therapy was most frequently used (43.1%), with 19.0% receiving platinum doublet plus bevacizumab and another 19.2% receiving single‐agent chemotherapies (Table 5). Single‐agent chemotherapies were most frequently used following the first use of PD‐1 inhibitors in 42.7% of patients, and the distribution of other treatments was much more diverse (Table 4).

Table 4. Treatments in line of therapy immediately following earliest PD‐1 inhibitora (n = 199).

^aAbsence of an LOT following the earliest PD‐1 inhibitor line does not necessarily mean that no therapy was subsequently received, but rather that the patient may not have been followed long enough to receive another therapy.

Abbreviations: EGFR, epidermal growth factor receptor; LOT, line of therapy; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1; TKIs, tyrosine kinase inhibitors.

Table 5. Treatments in line of therapy immediately prior to earliest line containing a programmed cell death protein 1 inhibitor, received in a second or later line (n = 1,117).

Abbreviations: EGFR, epidermal growth factor receptor; TKIs, tyrosine kinase inhibitors.

PD‐1 Inhibitor Administrations by Line of Therapy over Time

Given the rapid uptake in PD‐1 inhibitor use after approval, we determined how treatment patterns changed over time. Following the March 2015 approval, PD‐1 inhibitor use in the first and second lines after metastatic diagnosis increased consistently in each quarter of the period analyzed (Fig. 2B). Use initially increased at a similar pace for all lines, but later‐line use declined after 2015, with the use decline in fifth and subsequent lines beginning even earlier, in Q3 2015.

Discussion

This retrospective study examined the use of PD‐1 inhibitors in the real world setting following US regulatory approval of the first of these therapies. Our analysis of the characteristics of real‐world patients who received these drugs compared with the clinical trial cohorts that led to the approval of these drugs revealed a higher median age at treatment initiation and more patients over age 75 in the real‐world population. The majority of patients in all cohorts were white, but in the clinical trial cohorts, the percentage of Asian patients exceeded that of black patients, whereas the opposite was true in the real‐world cohort. The percentage of patients with smoking history in the real‐world cohort was higher than in all but one nivolumab registrational trial. For the pembrolizumab clinical trials, which were not specific to squamous and nonsquamous histology populations as was the case for nivolumab, the percentage of patients with squamous cell histology was lower than in the real‐world population (17%–19% vs. 32%), likely owing to the earlier approval of nivolumab for squamous NSCLC. Among 872 patients in this cohort with nonsquamous histology, 134 (15.4%) patients were not tested for EGFR mutations, and 167 (19.2%) were not tested for ALK rearrangements. Reasons for lack of testing could include inadequate available tumor tissue for testing, physician or patient choice, missing documentation, and abstractor error.

Beyond the differences compared with the clinical trial populations described above, there were notable differences between this treated cohort and the advanced NSCLC patients in the Flatiron Health Analytic database. In particular, there were more women in the overall population of NSCLC patients and more minorities. These observations may have implications for real‐world treatment decisions, and safety implications. Because of these differences, understanding safety and efficacy profiles in various subgroups can help generate additional evidence for optimal management of individual patients in the real world (e.g., elderly patients).

The analysis here captured use of these PD‐1 inhibitors as they first entered clinical use. Nivolumab's approval for squamous mNSCLC led to rapid initial adoption and a continued increase in use following its approval for nonsquamous patients. Following regulatory approval of pembrolizumab independent of histology but with PD‐L1 expression in October 2015, pembrolizumab began entering clinical regimens; however, pembrolizumab administrations were a small fraction of nivolumab usage, and in no line exceeded 5% of PD‐1 inhibitor use in this population.

PD‐L1 testing is required only for pembrolizumab, as nivolumab was approved regardless of PD‐L1 expression. This explains the initial low rate of PD‐L1 testing in this real‐world cohort with 3% pembrolizumab‐treated patients, and is confirmed by the higher rate of testing of patients who received pembrolizumab. When considering PD‐L1 testing prior to the start of PD‐1 inhibitor, we analyzed 1,348 different patient scenarios, which included data for 4 patients who received nivolumab and pembrolizumab in different lines. When the report designated that the assay was positive (n = 58), patients were prescribed nivolumab (n = 33) versus pembrolizumab (n = 25). When the report designated that the assay was negative (n = 37), patients were prescribed nivolumab exclusively. The latter is consistent with approvals, but the treatment pattern for PD‐L1 negative patients reflects the fact that 97% of patients in this real‐world cohort were treated with nivolumab, and pembrolizumab was approved 7 months after the first nivolumab approval.

Use of PD‐1 inhibitors by line evolved during the period analyzed. When PD‐1 inhibitors entered clinical use following approval, they were used in the early treatment setting (i.e., lines of therapy) as well as for patients in the late stages of disease. As lines of therapy were indexed to metastatic diagnosis in this work, some of the patients who received immunotherapy in the first‐line setting may have received chemotherapy in the advanced setting, which is consistent with the label. In interpreting patterns of use, we thus distinguish between the early PD‐1 inhibitor treatment setting (here, first and second line) versus third and later line use, likely representing salvage treatment. In this cohort, many patients diagnosed with mNSCLC well before the first nivolumab approval for NSCLC received a PD‐1 inhibitor late in their treatment (i.e., in third and later lines of therapy) once these drugs became available. Over time, there was an expected continued increase in routine use in early lines, but the later‐line use leveled off and declined by the end of the period analyzed here, as there were fewer late‐line PD‐1 inhibitor‐naive patients remaining. This aligns with an expected pattern of pent‐up demand for a novel approved therapy among patients who are late in illness without other treatment options; these patients receive treatment, and then the pattern of care settles to be more in line with approved indications. This initial period of equilibration needs to be considered when calculating the real‐world impact of a newly approved therapy on overall survival.

The limitations of our study arise from the complexities of extracting clinically relevant data using current EHR standards that are largely designed for oncologists treating patients, tracking billing, and managing clinical care. Although the strict QC procedures implemented in this study served to maximize data integrity, these real‐world data lacked complete information on Eastern Cooperative Oncology Group performance status and comorbidities—information that is not routinely documented at the point of care. Furthermore, assignment of lines of therapy was contingent on the availability of appropriate underlying EHR data triggering prespecified LOT definitions. PD‐L1 expression status, when available, was according to the final results in the test report. Although this represents the information on which the clinician acted, PD‐L1 “positivity” reflects a range of expression levels; any future consideration of outcomes in this subpopulation will warrant additional consideration. Lastly, this was an analysis of treatment patterns in the community setting only; because treatment patterns in academic settings may vary, a more complete picture of novel therapy adoption and early use would encompass data for academic sites as well.

Conclusion

The results of our study show the value of retrospective examination of real‐world data in gaining new insights into patterns of oncology care and characteristics of mNSCLC patients underrepresented in conventional clinical trials with PD‐1 inhibitors. Additional follow‐up is needed to understand whether the treatment patterns and the differences in patient populations we observed have any impact on clinical outcomes such as mortality.

Author Contributions

Conception/design: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Provision of study material or patients: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Collection and/or assembly of data: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Data analysis and interpretation: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Manuscript writing: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Final approval of manuscript: Sean Khozin, Amy P. Abernethy, Nathan C. Nussbaum, Jizu Zhi, Melissa D. Curtis, Melisa Tucker, Shannon E. Lee, David E. Light, Anala Gossai, Rachael A. Sorg, Aracelis Z. Torres, Payal Patel, Gideon Michael Blumenthal, Richard Pazdur

Disclosures

Amy P. Abernethy: Flatiron Health (E), Roche/Genentech (H), Bristol‐Myers Squibb (C/A); Melissa D. Curtis: Flatiron Health (E); Melisa Tucker: Flatiron Health (E); Shannon E. Lee: Flatiron Health (E); David E. Light: Flatiron Health (E); Anala Gossai: Flatiron Health (E); Rachael A. Sorg: Flatiron Health (E); Aracelis Z. Torres: Flatiron Health (E); Payal Patel: Flatiron Health (E). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board