Inequalities in CAR T-cell therapy access for US patients with relapsed/refractory DLBCL: a SEER-Medicare data analysis Open Access

Hematologic malignancies, including leukemia, lymphoma, myeloma, myelodysplastic syndromes, and myeloproliferative neoplasms, affect ∼1.6 million people in the United States as of 2023, contributing to >9% of all cancer-related deaths.¹ Despite significant advances in chemotherapy, radiotherapy, and targeted therapies,² the equitable diffusion of these innovations remains a critical challenge, with disparities in access and outcomes persisting across different populations and geographic regions.^3-7 

In recent years, cellular therapies, such as the chimeric antigen receptor (CAR) T cell (CAR-T), has emerged as a transformative treatment for hematologic cancers.^8,9 Tisagenlecleucel was the first CAR-T therapy approved in 2017, and shortly thereafter axicabtagene ciloleucel was approved. Subsequently, other CAR-T therapies, such as brexucabtagene autoleucel, ciltacabtagene autoleucel, idecabtagene vicleucel, and lisocabtagene maraleucel were approved for other indications, increasing treatment options for patients.^10,11 

Nevertheless, there are several barriers to accessing CAR-T therapy in the real world.^12-15 The complexity of manufacturing, delivering, and administering CAR-Ts has resulted in limited access to CAR-Ts because of logistical challenges, high costs of treatment, and treatment administration being limited to predominantly academic medical centers. Furthermore, the need for on-site follow-up care given the adverse event profile of the treatment has imposed additional barriers to which telemedicine could only provide partial answers.^12,16 However, there is relatively little quantitative evidence identifying the specific groups of patients who may face higher barriers to accessing CAR-Ts.^3,17 

This study addresses this knowledge gap by leveraging Surveillance, Epidemiology, and End Results (SEER)–Medicare data to evaluate factors associated with CAR-T use among US patients with relapsed/refractory diffuse large B-cell lymphoma (DLBCL) treated with third-line (3L) or later (3L+) therapies. Specifically, we quantify the impact of geographic distance to authorized treatment centers (ATCs) on access to CAR-T therapy and predict how mitigating these barriers could enhance use and improve survival outcomes. By providing actionable insights, this research aims to inform strategies to address disparities in access and optimize the delivery of CAR-T therapy, ensuring its benefits reach all eligible patients.

This study included individual patient-level data on Medicare beneficiaries diagnosed with DLBCL between 2007 and 2019 using Medicare claims data from 2007 through 2020 and linked to the SEER database (supplemental Methods). SEER data are from large, population-based cancer registries that collect clinical data at the time of tumor diagnosis, whereas Medicare claims include comprehensive treatment history for Medicare beneficiaries with DLBCL.

To better understand additional sources of disadvantage at the community level, the SEER-Medicare individual data were linked to 2 additional data files. The first supplemental data set was the Area Health Resource files, which contain information on population socioeconomic factors such as poverty status, insurance coverage, and racial composition by county; second, facility-level data on hospital characteristics, including size and capacity, and number of employees, were also determined using the Provider of Services files from the Centers for Medicare and Medicaid Services.

Patients aged ≥65 years with a previous diagnosis of DLBCL after 1 January 2007, who received >2 lines of therapy were identified using a previously validated algorithm.¹⁸ Lines of therapy were determined using the cancer treatments set out in the Enhanced Oncology Model 51’s list of initiating therapies, and we used the days between treatments to identify new treatments. Of these, patients were included if they did not die within 3 months of diagnosis and were continuously enrolled in Medicare parts A and B from the diagnosis date to either death or the end of the study period. Because the SEER-Medicare database does not have claims-level data for Medicare Advantage patients, those enrolled in a Medicare Advantage plan during any study month after diagnosis were excluded to ensure that claims from beneficiaries were captured consistently over time. Finally, patients with >2 lines of cancer therapy were included for the primary analysis (Figure 1; see supplemental Methods for lines of therapy algorithm).

There were 3 main outcomes of interest: (1) CAR-T use, (2) distance to 3L treatment facility (of any kind), and (3) life years gained. CAR-T use was the first outcome, which was identified using International Classification of Diseases-10 procedure codes (supplemental Methods). Distance to 3L treatment was calculated based on great-circle distance between the centroids of the patient’s zip code and provider or ATC zip code at which they received 3L treatment. The distance to the nearest 3L treatment was used to understand demographic characteristics associated with longer travel distances to receive treatment. However, this measure of distance does not capture distances that patients may have traveled if CAR-T treatment centers were available to them. We used distance from the patient’s zip code to the nearest ATC that was operational as of the year they received 3L treatment to understand the impact of this distance on the likelihood of CAR-T. Finally, we predicted the impact of reduced distance to the nearest ATC on CAR-T use on life years gained, as identified from clinical trials.¹⁹ 

First, descriptive statistics were used to assess the patient and county-level characteristics of those receiving CAR-T therapy. Covariates we evaluated included patient age, sex, race (White/non-White), Medicare dual eligibility status (whether the patient was ever dual eligible), and the number of unique chronic conditions the patient had after their diagnosis. We also use explanatory variables that describe the beneficiary’s area of residence, including whether the beneficiary lived in a metropolitan area or not (as defined by the US Department of Agriculture’s rural-urban continuum codes), the proportion of the county’s population without health insurance, total population of the county, number of hospital admissions per 10 000 in the county, proportion of the county’s population that is Black, the unemployment rate in the county, and average household income in the county (equation 1 in supplemental 1). The analysis encompassed all patients diagnosed with DLBCL who received any 3L+ cancer therapy.

Second, the factors affecting how far a patient traveled for any 3L+ treatment were estimated using a multivariate regression, with distance traveled for any 3L+ therapy as a function of patient characteristics and socioeconomic status (equation 2 in supplemental 1). This analysis aimed to determine whether certain groups of patients travel shorter or longer distances for treatment.

Third, in order to determine whether distance to the closest ATC affects CAR-T use among patients with 3L+ DLBCL, all else being equal, a multivariate logistic regression was used with probability of CAR-T treatment as a dependent variable and distance as an independent variable (equation 3 in supplemental 1). The great-circle distance was measured, controlling for very large distances (>125 miles). The regression also controlled for patient characteristics (eg, dual eligibility, race/ethnicity, sex, and age). The marginal effect of decreasing distance to the closest ATC was calculated to assess the impact of distance traveled on the probability of receiving CAR-Ts. To ensure robustness, we estimated the impact of the number of ATCs within a 25-mile radius (supplemental Table 7), rather than straight-line distance, on CAR-T use. We also estimated the impact of having zero ATCs within a 25-, 50-, and 75- mile radius on CAR-T use. (see supplemental Tables 8-10 for additional details). In these regressions, we control for patient race (White/non-White), sex, whether the patient lived in a metropolitan area or not, the number of unique chronic conditions the patient has after their diagnosis, patient age, dual eligibility status, and state and year fixed effects.

Fourth, states were categorized into “poor-access” and “better-access” states based on whether the average distance from patients with 3L+ DLBCL to the nearest treatment center in the state was above or below 50 miles. The relationship between distance and probability of receiving CAR-Ts was then extrapolated to examine how CAR-T use would change if the average distance to the nearest ATC was reduced such that all states became better-access states.

Finally, the number of life years gained annually was estimated based on the change in CAR-T use owing to decreased distance to the nearest ATC.^19,20 

Data analyses were conducted using Stata, release 15.0 (StataCorp LLC, College Station, TX).

There were 62 489 patients diagnosed with DLBCL in the SEER-Medicare database from 2007 through 2019 who were aged >65 years and eligible for Medicare. After applying all inclusion and exclusion criteria, there were 15 673 patients treated for DLBCL and 2241 who had received ≥3 lines of therapy between 2017 and 2020 (Figure 1).

Of 2241 patients in the 3L+ sample, 122 (5.4%) received CAR-Ts. Those receiving CAR-Ts had shorter median follow-up time from 3L treatment (3.3 vs 4.9 months) and were more likely to be dual eligible for Medicare and Medicaid at any point during 3L therapy (18.0% vs 14.5%; Table 1). Those receiving CAR-Ts also tended to be diagnosed later in our sample (2017 vs 2014), which is likely because of CAR-Ts first being approved in October 2017, with only clinical trials being a source of access before this. Furthermore, our results show that younger patients are more likely to receive CAR-Ts (age 70 vs 73 years).

This multivariate analysis of CAR-T use on patient and county-level characteristics found use was lower among patients who were relatively younger (OR, 0.893; P < .001) and had fewer comorbidities (OR, 0.904; P = .001; Figure 2). It was also lower among patients from counties with a lower median household income (OR, 1.176; P = .004), a higher share of population with no health insurance (OR, 0.944; P = .010; Figure 2).

We divided our sample from the 14 states included in the SEER-Medicare database into poor-access and better-access states based on a distance of 50 miles from the residence of the average patient to the nearest ATC (as of 2020). This threshold of 50 miles was based on the mean distance to the nearest ATC as of 2020, which was 52.8 miles. Poor-access states included Georgia, Iowa, Idaho, Kentucky, Louisiana, New Mexico, and Utah, whereas better-access states were California, Connecticut, Massachusetts, Michigan, New Jersey, New York, and Washington. Patients living in poor-access states not only lived further from ATCs, but these states tended to also have a lower median household income (Figure 3).

For any 3L therapy, patients with DLBCL who received CAR-Ts traveled a median distance of 38 miles (range, 0-2505) for treatment compared with only 12 miles (range, 0-2701) for those receiving other treatments. Dual-eligible patients (or those with lower income) who received treatment traveled shorter distances for any 3L treatment.

Although patients receiving CAR-Ts lived a median distance of 38 miles from the ATC at which they received treatment, the average patient treated for 3L DLBCL lived a mean distance of 30 miles from the nearest ATC. The probability of an eligible patient receiving CAR-Ts decreased by 6.2% for every 10 miles further from the closest ATC. In other words, closer proximity to CAR-T treatment centers increased the odds of receiving CAR-T. We found that at the average distance to an ATC of 171 miles, the predicted probability of receiving CAR-T was 3.8%. Among patients living within 125 miles to their nearest ATC (the cutoff used in the analysis to control for very large distances), the average distance was 30 miles. The predicted probability of receiving CAR-Ts at 30 miles was 9.1%, controlling for distances of >125 miles. In addition, controlling for various patient-level characteristics and accounting for state- and year-invariant effects, we found that the probability of getting CAR-Ts decreased by 48.9% (P = .035) when there was no ATC within 25 miles of the patient’s postal code of residence. (Figure 4; supplemental Table 8). As an additional robustness check of the control for the number of chronic conditions, we constructed the National Cancer Institute Comorbidity Index and used this to control for patient comorbidities. Although the sample is reduced from 2241 to 1357 patients and the number of beneficiaries receiving CAR-Ts is reduced to 76 because of availability of 12 months of claims before diagnosis, the findings are robust to the use of the National Cancer Institute Comorbidity Index: the probability of an eligible patient receiving CAR-Ts decreased by 6.3% for every 10 miles further from the closest ATC.

If access to CAR-Ts was improved such that the average distance for poor-access states (104.4 miles) was similar to that of the better-access states (34.2 miles), there would be a 2.5 percentage point increase in the probability of receiving CAR-Ts (from 6.6% to 9.1%), which represents a 37.6% (P < .001) relative increase in the total number of patients receiving CAR-Ts. Among the 3 lowest access states (New Mexico, Idaho, and Louisiana), there would be a 277.7% increase in the number of patients receiving CAR-Ts (from 2.4% to 9.1%).

To estimate the potential impact of improved CAR-T access on health outcomes, we used the improved survival rates observed in the ZUMA-1 clinical trial to model the impact of the improved CAR-T use on total life years gained.²⁰ We assumed the 50.5% 2-year survival rate from CAR-T use from the ZUMA-1 clinical trial relative to the 12.0% 2-year survival rate without CAR-Ts based on published studies.²⁰ The incidence of DLBCL in the United States is 4.68 DLBCL cases per 100 000 person-years, resulting in 15 673 DLBCL cases per year.²¹ Of these, 987 (6.3%) received 3L therapy.¹⁸ If CAR-T use increased from 6.6% to 9.1% of eligible patients, this would result in 105 life years gained annually among 3L+ Medicare patients with DLBCL (Figure 5).

Since the groundbreaking approval of the first CAR-T therapy in 2017, the therapeutic landscape has expanded to include 6 approved CAR-T therapies in the United States, targeting diverse malignancies such as DLBCL, follicular lymphoma, and multiple myeloma.¹² Beyond hematologic cancers, these therapies are being explored for nonmalignant diseases and solid tumors, reflecting their transformative potential.^22,23 With >500 active clinical trials for cell and gene therapies,²⁴ providers’ and patients’ treatment options are likely to increase over time. However, the persistent barriers that hinder access to CAR-T therapy for disadvantaged populations threaten to perpetuate inequities across this next generation of cellular and gene-based treatments.

Specifically, this study found that dual-eligible patients and patients living in remote or low-income communities were less likely to receive CAR-Ts, and increased access to CAR-Ts for patients with DLBCL is likely to save ∼10 lives per year in this population segment based on the early CAR-T time period. Given this result, policymakers should think creatively to identify the most cost-effective approaches for improving access to these disadvantaged populations. Addressing barriers related to distance is 1 policy option, which can be achieved by tackling the factors that will make it more feasible to establish a larger number of ATCs, particularly in disadvantaged areas. This approach would require training and resources for these centers to be approved to administer CAR-Ts but would significantly increase patient access.⁴ The use of telemedicine to connect with already established programs may allow for more local care after CAR-Ts at the current ATCs and may also help with training of the community sites as they build programs.¹⁶ Maintaining Foundation for the Accreditation of Cellular Therapy standards²⁵ and reporting to the Center for International Blood and Marrow Transplant Research to comply with the 15 year monitoring required by the US Food and Drug Administration for products using integrating viral vectors²⁶ are important to best practices in standards of care and quality management but can present significant regulatory burdens both financially and in terms of staffing. Lessons learned by established hematopoietic transplant programs while they restructured to include other cellular therapies may serve as guides for new programs.^27,28 

Distance to an ATC is only one part of a larger set of barriers. First, although increasing the number of ATCs would improve access, this must be achieved in a balanced and calculated way, because expansion of CAR-Ts to community-based practices has other important considerations. These practices had lower rates of CAR-T completion, in large part because of disease progression and decline in clinical status before CAR-Ts.²⁹ Second, patients further from an ATC may be less likely to get a referral to a facility because of the longer travel time and need for a reliable caregiver to enable successful CAR-T administration. Numerous initiatives have been aimed at addressing this issue, including the Community Oncologist Patient ID Roundtable, which has developed a standardized framework to enable community oncology care teams in assessing patients for CAR-T consultation.³⁰ Increasingly, nurse navigators can help facilitate both the complexities of the journey for the patient and improve communication between the local teams and the ATC. Medicare coverage for nurse navigation programs may make it easier for ATCs to adopt this model.³¹ 

Although the cost of CAR-T therapy itself is covered by most public and private payers in the United States, there are other costs to consider. Although payer reimbursement for CAR-T administration has improved in recent years, provider health system reimbursement for any accompanying hospital stay needed to administer CAR-Ts may not sufficiently cover hospital inpatient stay costs; the issue of hospital inpatient stay reimbursement is particularly acute for hospitals treating a large number of Medicaid beneficiaries.³² Patients themselves may incur additional costs for travel and lodging, as these costs are not typically covered by insurance. Although our study focused on Medicare beneficiaries, working-age adults may also face costs associated with short-term income loss during treatment; caregivers may face similar income loss.³³ 

Finally, barriers can change and should be reevaluated over time. Early in the availability of CAR-Ts, challenges included knowledge base, insurance coverage, and manufacturing limitations.³⁴ More recently, with the increase in available products and indications, difficulties arise in having enough trained clinical staff, cell therapy laboratory technicians, apheresis capacity, and local housing capacity for the required stay within the US Food and Drug Administration–mandated time frame. With improving access, these issues would be exacerbated given the increased volume, and solutions will be needed in these areas. Two recent developments will likely lead to positive changes. First, the Risk Evaluation and Mitigation Strategies programs for all approved CAR-T products have been modified with the goal of decreasing the burden on the health care system.^35,36 Secondly, retrospective data from a multicenter consortium have shown recently that cytokine release syndrome and immune effector cell–associated neurotoxicity syndrome rates were low after the first 2 weeks.³⁷ These results may allow patient monitoring regulations to be decreased, which could, in turn, reduce housing needs locally at an ATC and shrink the caregiver time burden.³⁸ 

The strengths of this study include its use of claims data to provide detailed information about patient treatment choices for 3L+ therapy, as well as registry-based information on patient tumor type. Zip code level data on patients provides the ability to determine accurately how distance to the nearest treatment facility affects the likelihood of CAR-T receipt. Importantly, this analysis measures the degree to which distance acts as a barrier to care differentially across patients with different social determinants of health.

The study also has several limitations. The data for the SEER-Medicare database used in this analysis were only available through 2020, resulting in a limited follow-up time period since the first CAR-T approvals in 2017. This also potentially led to the underestimation of CAR-T use, because it was subsequently approved for 2L therapy in 2022.³⁹ Second, although the analysis is restricted to Medicare fee-for-service beneficiaries, over half of DLBCL cases are diagnosed after age 65 years.⁴⁰ CAR-T use may be more common among individuals aged <65 years, who may encounter barriers related to having dependents or work commitments, potentially limiting their ability to travel for treatment. However, this study does not investigate the barriers to access faced by this population. Third, the analyses assumed that patients go to their nearest CAR-T facility, when, in practice, this is not always the case.

This study highlights the significant role of geographic and socioeconomic barriers in limiting access to CAR-T therapy for Medicare patients with DLBCL. Patients in low-income and disadvantaged areas, often located further from ATCs, face compounded challenges in accessing this potentially curative therapy. Addressing these disparities through strategies such as expanding the geographic distribution of ATCs and providing logistical support for underserved populations could significantly improve access and survival outcomes.

The authors acknowledge Kathy Batt for her clinical assistance in identifying therapies in this population. The authors acknowledge the efforts of the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.

The collection of cancer incidence data used in this study was supported by the California Department of Public Health pursuant to California Health and Safety Code Section 103885; Centers for Disease Control and Prevention’s National Program of Cancer Registries (cooperative agreement 1NU58DP007156); and the National Cancer Institute (NCI)’s SEER (contract HHSN261201800032I awarded to the University of California, San Francisco; contract HHSN261201800015I awarded to the University of Southern California; and contract HHSN261201800009I awarded to the Public Health Institute). This research was funded by Kite. It was also supported, in part, by a National Institutes of Health (NIH)/NCI Cancer Center support grant P30 CA008748 (M.-A.P. and G.L.S.).

The ideas and opinions expressed herein are those of the author(s) and do not necessarily reflect the opinions of the NIH, State of California, Department of Public Health, the NCI, and the Centers for Disease Control and Prevention or their contractors and subcontractors. The funder made the following contributions to the study: design and conduct of the study; analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The funder had no role in the collection and management of the data.

Contribution: A.P.C. 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; A.P.C., J.T.S., and S.M. were responsible for the data analysis; and all authors contributed to the development of this manuscript.

Conflict-of-interest disclosure: A.P.C., J.T.S., and S.M. are employees of FTI Consulting, Inc, a consulting firm to health care, life sciences, government, and nongovernmental entities. S.V. and K.H., are employees of Kite/Gilead, the sponsor of this study. At the time of the study, A.R.P. was an employee of Kite/Gilead, the sponsor of this study. M.-A.P. reports honoraria from Adicet, Allogene, AlloVir, Caribou Biosciences, Celgene, Bristol Myers Squibb (BMS), Equilium, ExeVir, ImmPACT Bio, Incyte, Karyopharm, Kite/Gilead, Merck, Miltenyi Biotec, MorphoSys, Nektar Therapeutics, Novartis, Omeros, Orca Bio, Sanofi, Syncopation, VectivBio AG, and Vor Biopharma; serves on data safety monitoring boards (DSMBs) for Cidara Therapeutics and Sellas Life Sciences; serves on the scientific advisory board of NexImmune; reports ownership interests in NexImmune, Omeros, and Orca Bio; and has received institutional research support for clinical trials from Allogene, Incyte, Kite/Gilead, Miltenyi Biotec, Nektar Therapeutics, and Novartis. R.T.M. reports serving as consultant for Incyte, Kite/Gilead, and Novartis; receiving research support from Gamida Cell, Kite/Gilead, Orca Bio, and Novartis; serving on scientific advisory committee for Artiva; and participating in a DSMB for Novartis, Century Therapeutics, and Vor Biopharma. G.L.S. has received research funding to the institution from Janssen, Amgen, BMS, Beyond Spring, and GPCR Therapeutics, Inc; and is on the DSMB for ArcellX. L.C.A. is a consultant for Incyte, Kite/Gilead, and BMS.

Correspondence: Andrea P. Chung, FTI Consulting, Inc, 555 12th St NW STE 700, Washington, DC 20004; email: andrea.chung@fticonsulting.com.