What are the uptake and survival outcomes following immune checkpoint inhibitor (ICI) therapy among patients with advanced solid cancers who are traditionally ineligible for clinical trials because of poor performance status or organ dysfunction?
In this cohort study of 34 131 patients with cancer, trial ineligibility was associated with a greater likelihood of ICI monotherapy use compared with non-ICI therapy. Among trial-ineligible patients, there were no overall survival differences between treatment with ICI monotherapy, ICI combination therapy, and non-ICI therapy.
The study results suggest that positive results for ICI therapy in phase 3 trials may not translate to patients with poor performance status or organ dysfunction who are traditionally ineligible for such trials.
Immune checkpoint inhibitors (ICIs) are part of standard of care for patients with many advanced solid tumors. Patients with poor performance status or organ dysfunction are traditionally ineligible to partake in pivotal randomized clinical trials of ICIs.
To assess ICI use and survival outcomes among patients with advanced cancers who are traditionally trial ineligible based on poor performance status or organ dysfunction.
Design, Setting, and Participants
This retrospective cohort study was conducted in 280 predominantly community oncology practices in the US and included 34 131 patients (9318 [27.3%] trial ineligible) who initiated first-line systemic therapy from January 2014 through December 2019 for newly diagnosed metastatic or recurrent nontargetable non–small cell lung, urothelial cell, renal cell, or hepatocellular carcinoma. Data analysis was performed from December 1, 2019, to June 1, 2021.
Trial ineligibility (Eastern Cooperative Oncology Group performance status ≥2 or the presence of kidney or liver dysfunction); first-line systemic therapy.
Main Outcomes and Measures
The association between trial ineligibility and ICI monotherapy uptake was assessed using inverse probability–weighted (IPW) logistic regressions. The comparative survival outcomes following ICI and non-ICI therapy among trial-ineligible patients were assessed using treatment IPW survival analyses. Because we observed nonproportional hazards, we reported 12-month and 36-month restricted mean survival times (RMSTs) and time-varying hazard ratios (HRs) of less than 6 months and 6 months or greater.
Among the overall cohort (n = 34 131), the median (IQR) age was 70 (62-77) years; 23 586 (69%) were White individuals, and 14 478 (42%) were women. Over the study period, the proportion of patients receiving ICI monotherapy increased from 0% to 30.2% among trial-ineligible patients and 0.1% to 19.4% among trial-eligible patients. Trial ineligibility was associated with increased ICI monotherapy use (IPW-adjusted odds ratio compared with non-ICI therapy, 1.8; 95% CI, 1.7-1.9). Among trial-ineligible patients, there were no overall survival differences between ICI monotherapy, ICI combination therapy, and non-ICI therapy at 12 months (RMST, 7.8 vs 7.7 vs 8.1 months) or 36 months (RMST, 15.0 vs 13.9 vs 14.4 months). Compared with non-ICI therapy, ICI monotherapy showed evidence of early harm (IPW-adjusted HR within 6 months, 1.2; 95% CI, 1.1-1.2) but late benefit (adjusted HR among patients who survived 6 months, 0.8; 95% CI, 0.7-0.8).
Conclusions and Relevance
In this cohort study, compared with trial-eligible patients, trial-ineligible patients with advanced cancers preferentially received first-line ICI therapy. A survival difference was not detected between ICI and non-ICI therapies among trial-ineligible patients. Positive results for ICI in phase 3 trials may not translate to this vulnerable population.
Since 2015, several immune checkpoint inhibitors (ICIs) have received first-line approval as monotherapy or in combination with other systemic therapy for treating advanced solid tumors.1,2 Use of ICIs is associated with improved survival in many advanced cancers, including non–small cell lung (NSCLC), urothelial cell (UCC), renal cell (RCC), and hepatocellular (HCC) carcinoma. This survival benefit is driven by durable long-term survival in a subset of patients, including patients with high expression of programmed cell death ligand 1 (PD-L1) or tumor mutational burden.3-8 Generally, ICI therapy has a favorable adverse effect profile in these cancers.9-12 In contrast, non-ICI therapies, including cytotoxic chemotherapies or antiangiogenic therapy, are associated with high rates of toxic effects and treatment-related discontinuation.13-19
Because of their favorable toxic effect profile, ICIs may be preferentially used rather than non-ICI therapies for patients with poor performance status (PS) or liver or kidney dysfunction.20-22 However, patients with poor PS or organ dysfunction are largely excluded from all pivotal randomized clinical trials of ICIs, despite comprising 50% of patients with advanced cancer.21-24 Use of ICIs in trial-ineligible patients has been termed desperation oncology in the lay press.25 Thus, it is critical to understand the magnitude of use and effectiveness of ICIs in this vulnerable population.
In this retrospective cohort study, we assessed the use and comparative overall survival outcomes following first-line ICI and non-ICI therapy among trial-ineligible patients with newly diagnosed advanced solid tumors. We studied 4 representative metastatic solid tumors with established indications for first-line ICI and non-ICI treatments: NSCLC, UCC, RCC, and HCC. We hypothesized that trial-ineligible patients would be more likely to receive ICI monotherapy than trial-eligible patients and that ICI monotherapy and combination therapy would not be associated with a survival benefit compared with non-ICI therapy among trial-ineligible patients.
The study was approved by the University of Pennsylvania institutional review board with a waiver of informed consent owing to the use of deidentified retrospective data. The study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
We used data from the nationwide, deidentified longitudinal electronic health record (EHR)–derived Flatiron Health database. Data originated from approximately 280 US cancer clinics (approximately 800 sites of care), most being community oncology settings. Deidentified patient-level data included structured and unstructured data, such as clinician notes and pathology reports, all of which were curated via technology-enabled abstraction, as described previously.26,27 Data from unstructured EHR-derived digital documents were manually reviewed by centrally managed and trained medical record abstractors using specifically defined abstraction protocols.28-31
The study sample selected patients with a new diagnosis of metastatic or recurrent NSCLC, UCC, RCC, and HCC. Patients were 18 years or older and initiated treatment with first-line systemic therapy for advanced disease between January 1, 2014, and December 31, 2019. We excluded patients who had no evidence of structured EHR activity within 90 days of diagnosis or who received a diagnosis after December 31, 2019. We also excluded patients who did not receive systemic therapy, had a targetable driver variation, received first-line agents as part of a randomized clinical trial, received first-line agents that were not listed in National Comprehensive Cancer Network guidelines for systemic therapy of advanced NSCLC, UCC, RCC, or HCC, or were missing demographic data (eFigure 1 in the Supplement).32-35
Exposures and Variable Definitions
To first assess ICI use, the exposure of interest was trial ineligibility. To define trial ineligibility, we used common exclusion criteria from phase 3 clinical trials of ICI agents: Eastern Cooperative Oncology Group (ECOG) PS of 2 or greater,36,37 or the presence of organ dysfunction. Organ dysfunction was adapted from the National Cancer Institute’s Common Terminology Criteria for Adverse Events, version 5,38 and defined as having a diagnosis code for liver dysfunction (eTable 1 in the Supplement), estimated glomerular filtration rate of less than 30 mL/min/1.73 m2, or total bilirubin levels of less than 3 mg/dL (to convert to μmol/L, multiply by 17.104). Trial ineligibility criteria were defined using a 60-day lookback window from the date of first-line therapy initiation. A patient who met either criterion was classified as trial ineligible; all other patients were classified as trial eligible.
To assess the comparative overall survival among trial-ineligible patients, the exposure of interest was treatment, which was categorized as ICI monotherapy, ICI combination therapy, or non-ICI therapy (eTable 2 in the Supplement). Immune checkpoint inhibitor monotherapy was defined as 1 or more ICIs without non-ICI therapy. Immune checkpoint inhibitor combination therapy was defined as 1 or more ICIs in combination with non-ICI therapy.
To assess ICI use, the primary outcome was receipt of first-line ICI monotherapy. To assess ICI comparative survival outcomes, the primary outcome was overall survival (OS) from the initiation of first-line treatment to date of death.39 Follow-up was terminated at the first of the following dates: date of death, December 31, 2019, or last structured EHR activity (eg, laboratory draw, chemotherapy administration, or outpatient clinical encounter).
We selected the following 9 covariates as potentially contributing to ICI use and OS for use in inverse probability–weighted (IPW) analyses: cancer type (NSCLC, UCC, RCC, and HCC), age at metastatic diagnosis, sex, self-reported race and ethnicity, year of diagnosis, insurance type, practice type (community vs academic), use of opioid pain medication, and use of corticosteroids at the time of diagnosis but before treatment. The latter 2 covariates are surrogates of symptomatic or high-volume disease, and early corticosteroid use is associated with poorer outcomes while receiving ICI therapy.40
We used descriptive statistics to compare cohort characteristics between trial-eligible and trial-ineligible groups and among all treatment groups. We used multiple imputation to impute missing PS and laboratory data (see the eMethods in the Supplement for details on imputation).41 To address systematic differences between trial-eligible and trial-ineligible cohorts and between all treatment groups, we used IPW analyses. We used generalized boosted models to fit 2 propensity score models using all 9 covariates.42,43 The first model predicted the probability of a patient being trial ineligible; the second predicted the probability of a patient receiving ICI monotherapy, ICI combination therapy, or non-ICI therapy. The postweighting balance in covariates was evaluated using standardized mean differences (SMDs); SMDs greater than 0.1 indicated a covariate imbalance.44 To test the hypothesis that first-line ICI monotherapy is used more often for trial-ineligible compared with trial-eligible patients, we used an IPW multivariable logistic regression to estimate the odds of receiving ICI monotherapy compared with non-ICI therapy. To assess the heterogeneity of this association, secondary analyses fit separate IPW models for each of the 4 cancer subcohorts.
To test the hypothesis that, among trial-ineligible patients, there was no survival difference between treatment groups, we used IPW-adjusted Kaplan-Meier curves; IPW-adjusted hazard ratios (HRs) with a time-varying coefficient, which allowed for different effects before and after 6 months; and restricted mean survival times (RMSTs) at 12 and 36 months to compare median OS from first-line treatment initiation (eMethods in the Supplement).45 In a sensitivity analysis to assess the degree of potential unmeasured confounding, we calculated E-values based on point estimates before and after 6 months to quantify the minimum strength of association (on the risk ratio scale) between a confounder and the exposure and outcome needed to nullify the observed exposure-outcome association.46 In another sensitivity analysis to identify whether associations differed by cohort definition criteria, we repeated survival analyses that defined trial ineligibility as an ECOG PS of 2 or greater or organ dysfunction. Finally, to further account for granular differences in PS, we conducted a sensitivity analysis using ECOG PS (0-4) as a continuous covariate in the propensity score.
In an exploratory analysis, we repeated survival analyses among individuals with advanced NSCLC who had a recorded PD-L1 expression level within 60 days of first-line therapy initiation, which was stratified by PD-L1 test status (high vs low expression). Programmed cell death ligand 1 expression was assessed by immunohistochemistry; high vs low expression was defined using a threshold of PD-L1 staining of 50% or more of tumor cells or a PD-L1–stained tumor-infiltrating immune cells covering 10% or more of the tumor area. All hypothesis tests were 2-sided, α = .05, and analyses were conducted between December 1, 2019, and June 1, 2021, using R (version 4.0.2; R Foundation).
Baseline Characteristics of the Total Cancer Cohort
The overall cohort (n = 34 131) consisted of 26 904 patients (78.8%) with advanced NSCLC, 3382 (9.9%) with UCC, 3323 (9.7%) with RCC, and 522 (1.5%) with HCC. The median (IQR) age was 70 (62-77) years; 23 586 (69%) were White individuals, and 14 478 (42%) were women. A total of 9318 patients (27.3%) were trial ineligible, of whom 6570 (71%) had an ECOG PS of 2 or greater and 3416 (37%) had organ dysfunction (eTable 3 in the Supplement). Before imputation, PS was coded for 17 340 patients (50.8%); there were no appreciable differences in other baseline characteristics before or after imputation (eTable 4 in the Supplement). Unweighted baseline characteristics were similar between trial-eligible and trial-ineligible individuals (SMD <0.10), with 3 exceptions: trial-ineligible patients were older (median [IQR] age, 71 [63-78] vs 68 [61-75] years), less likely to have NSCLC (76% vs 80%), and less likely to have commercial insurance (33% vs 38%). After IPW, baseline characteristics were similar between trial-eligible and trial-ineligible individuals.
ICI Treatment Use in Trial-Eligible and Trial-Ineligible Patients
In the total cohort, 4961 patients (14.5%) received first-line ICI monotherapy, 3982 (11.7%) received ICI combination therapy, and 25 188 (73.8%) received non-ICI therapy. Over time, the proportion of patients who received ICI monotherapy increased from 0.0% to 30.2% in trial-ineligible individuals and 0.1% to 19.4% in trial-eligible individuals (eFigure 2 in the Supplement). By the last quarter of 2019, 109 (23.9%) patients with NSCLC, 25 (52.9%) patients with UCC, 42 (55.8%) patients with RCC, and 1 (13.3%) patient with HCC who were trial ineligible received first-line ICI-monotherapy (Figure 1). Trial ineligibility was associated with greater use of ICI monotherapy compared with non-ICI therapy (IPW-adjusted odds ratio [aOR], 1.80; 95% CI, 1.70-1.92). The magnitude of the association was strongest among patients with NSCLC (aOR, 1.95; 95% CI, 1.80-2.10) and UCC (aOR, 2.48; 95% CI, 2.11-2.92), while a significant association was not found among patients with RCC (aOR, 1.11; 95% CI, 0.87-1.43) and HCC (aOR, 1.78; 95% CI, 0.97-3.27) (eTable 5 in the Supplement).
ICI Treatment Effectiveness in Trial-Ineligible Patients
In the trial-ineligible cohort, the median (IQR) follow-up was 7.8 (3.4-16.5) months. During follow-up, there were 6726 deaths (1331 in the ICI monotherapy group, 503 ICI in the combination therapy group, and 4912 in the non-ICI group). After IPW, baseline characteristics were similar between treatment groups (eTable 6 in the Supplement). Inverse probability weighting–adjusted Kaplan-Meier curves are illustrated in the total cohort in Figure 2. Median OS was 9.72 months for ICI monotherapy, 9.33 months for ICI combination therapy, and 8.75 months for non-ICI therapy (Table). During the first 6 months after treatment, ICI monotherapy (IPW-adjusted HR [aHR], 1.19; 95% CI, 1.11-1.27) and ICI combination therapy (aHR, 1.14; 95% CI, 1.08-1.22) were associated with higher mortality compared with non-ICI therapy. In contrast, among patients who survived 6 months after treatment initiation, subsequent mortality was lower for ICI monotherapy (aHR, 0.80; 95% CI, 0.74-0.84) and similar for ICI combination therapy (aHR, 1.0; 95% CI, 0.89-1.13). The RMSTs were similar among ICI monotherapy, ICI combination therapy, and non-ICI therapy groups at 12 months (aHR, 7.8; 95% CI, 7.7-7.9 vs 7.8; 95% CI, 7.6-7.9 vs 8.1; 95% CI, 8.0-8.2 months, respectively) and 36 months (aHR, 15.0; 95% CI, 14.5-15.4 vs 13.9; 95% CI, 13.1-14.7 vs 14.4; 95% CI, 14.1-14.8 months, respectively). For trial-ineligible patients with NSCLC, UCC, and HCC, ICI monotherapy was associated with higher mortality during the first 6 months after treatment (aHR, 1.13; 95% CI, 1.05-1.23; 1.63; 95% CI, 1.36-1.95; and 2.53; 95% CI, 1.75-3.65, respectively) but lower subsequent mortality among patients who survived 6 months (aHR, 0.77; 95% CI, 0.72-0.82; 0.77; 95% CI, 0.64-0.92; and 0.89; 95% CI, 0.46-1.71, respectively) compared with non-ICI therapy (eTable 7 in the Supplement). In HCC, ICI combination therapy was associated with higher mortality before (aHR, 2.47; 95% CI, 1.67-3.66) and after (aHR, 11.2; 95% CI, 1.34-93.36) 6 months. In RCC, there were no significant differences in mortality between treatment groups. The E-values (relative risk) for the point estimates for mortality were 1.51 (<6 months) and 1.92 (≥6 months) for ICI monotherapy and 1.49 (<6 months) and 1.16 (≥6 months) for ICI combination therapy. The RMSTs were similar among treatment groups at 12 and 36 months when using alternate definitions of trial ineligibility (eTable 8 in the Supplement) and incorporating ECOG PS values in the propensity score (eTable 9 in the Supplement).
Exploratory Stratified Analyses
Among patients with NSCLC who had a reported PD-L1 expression (n = 3459), 2498 (72%) had PD-L1 high-expressing tumors and 961 (28%) had PD-L1 low-expressing tumors. Among patients with high and low PD-L1 expression, ICI monotherapy was associated with higher early mortality compared with non-ICI therapy during the first 6 months after therapy (aHR, 1.83; 95% CI, 1.36-2.45 and 2.09; 95% CI, 1.32-3.31, respectively). There was no benefit for ICI monotherapy among patients who survived longer than 6 months after therapy (aHR, 0.90; 95% CI, 0.68-1.19 and 0.68; 95% CI, 0.33-1.42, respectively) (eTable 10 in the Supplement).
In this population-based study of patients with advanced solid tumors, overall use of first-line ICI monotherapy among patients who were traditionally ineligible for phase 3 ICI trials increased substantially over time, even in the absence of prospective evidence of benefit. Despite increased use of ICI, we did not find a survival benefit associated with ICI monotherapy or combination therapy among trial-ineligible patients. Among trial-ineligible patients, ICI monotherapy and combination therapy were associated with a 14% to 19% greater hazard of death during the first 6 months after ICI initiation. In contrast, ICI monotherapy was associated with a 20% lower hazard of death among those who survived 6 months after initiation. Time-dependent ICI outcomes were observed for trial-ineligible patients with NSCLC, UCC, and HCC; in contrast, ICI outcomes were roughly constant throughout follow-up for patients with RCC. Collectively, these results suggest that phase 3 trials that show a survival benefit for ICI therapy may not generalize to patients with poor PS or organ dysfunction; future prospective trials are needed to confirm these findings.
There are 3 main findings worth emphasis. First, to our knowledge, this is one of the first studies to show that ICI monotherapy was preferentially used in trial-ineligible populations. These findings were particularly striking for patients with UCC and NSCLC (IPW-adjusted odds ratios, 2.0-2.5), for whom the non-ICI standard of care is platinum-based chemotherapy, compared with patients with HCC and RCC, for whom the non-ICI standard is oral anti–vascular endothelial growth factor therapy. Concerns about the tolerability of platinum-based chemotherapy may be associated with preferential use of first-line ICI in trial-ineligible patients, despite sparse evidence of ICI benefit in this population.22,24
Second, we did not find evidence of a survival benefit of ICI monotherapy or combination therapy in trial-ineligible populations, with similar OS at 12 and 36 months between ICI and non-ICI regimens. Much of the enthusiasm regarding the use of ICIs has been generated by the approximately 15% of individuals who have durable long-term response and/or survival.47 In contrast to published prospective and retrospective studies of ICI, this study’s trial-ineligible cohort had a median OS of fewer than 10 months, and nearly 40% of trial-ineligible patients who received treatment died within 6 months. We did not observe a survival advantage for ICI monotherapy or combination therapy among trial-ineligible individuals given the high proportion who died within 1 year.
Third, ICI combination therapy was associated with no long-term survival benefit and potential early harm (aHR, 1.14) among trial-ineligible patients. This finding stands in contrast to published phase 3 trials.3,5,6 Shorter survival for ICI combination therapy was particularly notable among patients with HCC. One potential mechanism for this observation may be that ICI combination therapy exacerbates life-threatening adverse effects from ICI and non-ICI therapy. Our findings regarding ICI combination therapy should be interpreted with caution given heterogeneity among cancer types (notably in RCC). However, ICI and chemotherapy adverse effects are exacerbated in patients with poor PS and organ dysfunction.48-50 Thus, our findings argue for cautious use of ICI combination therapy in trial-ineligible patients.
Although we were unable to control for PD-L1 expression in our primary analysis because of high levels of missing data, PD-L1 expression status did not moderate the overall findings in stratified analyses. This is surprising, given that high tumor PD-L1 expression in trial participants is associated with good ICI responses.51 These findings should be interpreted with caution, as PD-L1 stratified analyses may be underpowered.
Our propensity score–adjusted retrospective study suggests that the efficacy of ICIs alone or in combination is worse among trial-ineligible patients than that seen in phase 3 trials. There may be unanticipated harms with ICI use in the 30% of patients who are traditionally excluded from randomized clinical trials.52 Following on specialty society recommendations, phase 3 trials of ICI should allow enrollment of patients with poor PS and/or organ dysfunction to clarify treatment strategies for this population.53
While there was no OS decrement associated with ICIs in trial-ineligible populations, there are consequences to its use. Use of ICIs for patients with poor PS is associated with lower hospice enrollment, more inpatient deaths, and more treatment during the last month of life.20,48,54-58 It is critical to ensure that vulnerable, trial-ineligible patients are not exposed to non–evidence-based therapies that could cause harm and contradict patient goals.59
There are several limitations to this analysis. First, the cohort was limited to patients who received systemic therapy, as data from patients who do not receive systemic therapy is suboptimally captured in real-world data sets. Thus, we cannot make conclusions about the relative efficacy of any systemic therapy vs no systemic therapy in trial-ineligible individuals. Second, given that E-values were modest (1.51 [<6 months] and 1.92 [≥6 months]), there may be residual confounding because of unobserved covariates that are associated with the decision to use first-line immunotherapy. Such unobserved covariates include brain metastases, histology, burden of metastatic disease, and International Metastatic RCC Database Consortium criteria (for RCC). Additionally, there may be subtle differences in cohort composition between cancers over time, given that US Food and Drug Administration approvals for ICI therapy in these cancers occurred at different points between 2015 and 2017. However, residual confounding would only fully explain observed treatment effects if the unmeasured confounder was associated with receipt of ICI monotherapy, increased early mortality, and decreased late mortality. Because the direction of the association reverses, an unmeasured confounder would either need to be time varying or have a time-dependent effect that switched direction midway through follow-up. Given this time dependency, such an unmeasured confounder is unlikely, although future work should incorporate more robust clinical data. Third, ECOG PS was missing for nearly half of the cohort. However, the distributions of all observed covariates other than practice type were similar between patients with coded and noncoded ECOG PS. Fourth, the study cohort was primarily seen in community-based oncology practices; thus, our results may not generalize to academic practices. Fifth, certain criteria, such as kidney dysfunction in RCC, are not associated with poorer outcomes and are not routinely exclusion criterion in ICI clinical trials; however, our results for RCC were consistent in several subgroup and sensitivity analyses.
In this retrospective cohort study, trial-ineligible individuals had a nearly 2-fold greater odds of receiving ICI monotherapy compared with trial-eligible individuals. Among trial-ineligible individuals, ICI therapy had no OS benefit and the potential for early harm. These findings must be validated in prospective phase 3 studies before changing clinical practice. Clinicians who care for patients with poor PS or organ dysfunction should be cautious about ICI use and carefully weigh expected survival gains against the potential for early mortality and adverse effects.
Accepted for Publication: July 26, 2021.
Published Online: November 4, 2021. doi:10.1001/jamaoncol.2021.4971
Corresponding Author: Ravi B. Parikh, MD, MPP, University of Pennsylvania, 423 Guardian Dr, Blockley 1102, Philadelphia, PA 19104 (email@example.com).
Author Contributions: Drs Parikh and Min 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. Drs Hubbard, Long, and Mamtani contributed equally to this manuscript.
Concept and design: Parikh, Gross, Long, Mamtani.
Acquisition, analysis, or interpretation of data: Parikh, Min, Wileyto, Riaz, Cohen, Hubbard, Long, Mamtani.
Drafting of the manuscript: Parikh, Min, Wileyto.
Critical revision of the manuscript for important intellectual content: Parikh, Riaz, Gross, Cohen, Hubbard, Long, Mamtani.
Statistical analysis: Parikh, Min, Wileyto, Hubbard, Long.
Obtained funding: Parikh.
Administrative, technical, or material support: Parikh, Gross.
Conflict of Interest Disclosures: Dr Parikh reported nonfinancial support from Flatiron Health, grants from Humana, personal fees and equity from GNS Healthcare Inc and Onc.AI, and personal fees from the Cancer Study Group and Nanology outside the submitted work. Dr Gross reported personal fees from Flatiron and grants from the National Comprehensive Cancer Network (Pfizer/AstraZeneca), Johnson & Johnson, and Genentech outside the submitted work. Dr Hubbard reported grants from the National Institutes of Health (NIH) during the conduct of the study as well as grants from Merck, Pfizer, and Johnson & Johnson outside the submitted work. Dr Long reported grants from NIH during the conduct of the study and grants from Pfizer and Bayer outside the submitted work. Dr Mamtani reported personal fees from Flatiron Health and Astellas/Seattle Genetics and grants from Merck outside the submitted work. No other disclosures were reported.
Funding/Support: This study was supported by the Conquer Cancer Foundation Young Investigator Award (Dr Parikh), grant K08 CA263541 from the National Cancer Institute (Dr Parikh), grant P30 CA016520 from the National Cancer Institute (Drs Min, Long, and Mamtani), and the Catholic Medical Center Research Foundation (Dr Min).
Role of the Funder/Sponsor: The funding organizations 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.
et al. Speed of adoption of immune checkpoint inhibitors of programmed cell death 1 protein and comparison of patient ages in clinical practice vs pivotal clinical trials. JAMA Oncol
. 2018;4(8):e180798. doi:10.1001/jamaoncol.2018.0798PubMedGoogle Scholar
et al. Five-year survival and correlates among patients with advanced melanoma, renal cell carcinoma, or non–small cell lung cancer treated with nivolumab. JAMA Oncol
. 2019;5(10):1411-1420. doi:10.1001/jamaoncol.2019.2187PubMedGoogle ScholarCrossref
P. Tolerability of sorafenib in the treatment of hepatocellular carcinoma (HCC) in patients with Child A and B liver cirrhosis. J Clin Oncol
. 2009;27(15S):e15593. doi:10.1200/jco.2009.27.15_suppl.e15593Google Scholar
B, de Marinis
F. Immunotherapy in non–small-cell lung cancer patients with performance status 2: clinical decision making with scant evidence. J Clin Oncol
. 2019;37(22):1863-1867. doi:10.1200/JCO.18.02118PubMedGoogle ScholarCrossref
R. Immunotherapy in older adults with advanced cancers: implications for clinical decision-making and future research. Am Soc Clin Oncol Educ Book
. 2018;38(38):400-414. doi:10.1200/EDBK_201435PubMedGoogle ScholarCrossref
et al. Model-assisted cohort selection with bias analysis for generating large-scale cohorts from the EHR for oncology research. arXiv
. Published online January 13, 2020. https://arxiv.org/abs/2001.09765Google Scholar
SS. Comparison of population characteristics in real-world clinical oncology databases in the US: Flatiron Health, SEER, and NPCR. medRxiv
. Published online May 30, 2020. doi:10.1101/2020.03.16.20037143Google Scholar
et al. Generating real-world tumor burden endpoints from electronic health record data: comparison of RECIST, radiology-anchored, and clinician-anchored approaches for abstracting real-world progression in non–small cell lung cancer. Adv Ther
. 2019;36(8):2122-2136. doi:10.1007/s12325-019-00970-1PubMedGoogle ScholarCrossref
et al. Association of broad-based genomic sequencing with survival among patients with advanced non–small cell lung cancer in the community oncology setting. JAMA
. 2018;320(5):469-477. doi:10.1001/jama.2018.9824PubMedGoogle ScholarCrossref
et al. Association of patient characteristics and tumor genomics with clinical outcomes among patients with non–small cell lung cancer using a clinicogenomic database. JAMA
. 2019;321(14):1391-1399. doi:10.1001/jama.2019.3241PubMedGoogle ScholarCrossref
et al. A simplified frailty scale predicts outcomes in transplant-ineligible patients with newly diagnosed multiple myeloma treated in the FIRST (MM-020) trial. Leukemia
. 2020;34(1):224-233. doi:10.1038/s41375-019-0539-0PubMedGoogle ScholarCrossref
LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med
. 2013;32(19):3388-3414. doi:10.1002/sim.5753PubMedGoogle ScholarCrossref
et al. Impact of performance status on treatment outcomes: a real-world study of advanced urothelial cancer treated with immune checkpoint inhibitors. Cancer
. 2020;126(6):1208-1216. doi:10.1002/cncr.32645PubMedGoogle ScholarCrossref
E, van den Bos
J, de Glas
NA. Efficacy and adverse events of immunotherapy with checkpoint inhibitors in older patients with cancer. Drugs Aging
. 2019;36(10):927-938. doi:10.1007/s40266-019-00697-2PubMedGoogle ScholarCrossref
et al. Broadening eligibility criteria to make clinical trials more representative: American Society of Clinical Oncology and Friends of Cancer Research joint research statement. J Clin Oncol
. 2017;35(33):3737-3744. doi:10.1200/JCO.2017.73.7916PubMedGoogle ScholarCrossref
IJ. Real-world use and survival outcomes of immune checkpoint inhibitors in older adults with non-small cell lung cancer. Cancer
. 2020;126(5):978-985. doi:10.1002/cncr.32624PubMedGoogle ScholarCrossref
L. Immune checkpoint inhibitor use near the end of life is associated with poor performance status, lower hospice enrollment, and dying in the hospital. Am J Hosp Palliat Care
. 2020;37(3):179-184. doi:10.1177/1049909119862785PubMedGoogle ScholarCrossref
L. Immune checkpoint inhibitor use near the end of life: a single-center retrospective study. J Palliat Med
. 2020;23(7):977-979. doi:10.1089/jpm.2019.0383PubMedGoogle ScholarCrossref
et al. Safety, efficacy, and patient-reported health-related quality of life and symptom burden with nivolumab in patients with advanced non–small cell lung cancer, including patients aged 70 years or older or with poor performance status (CheckMate 153). J Thorac Oncol
. 2019;14(9):1628-1639. doi:10.1016/j.jtho.2019.05.010PubMedGoogle ScholarCrossref