Insurance status impacts overall survival in Burkitt lymphoma

Abstract

The impact of insurance status on clinical outcomes in Burkitt (BL) and plasmablastic (PBL) lymphomas remains unknown. We used the National Cancer Database to examine insurance status’ effect on overall survival (OS) in adults diagnosed with these lymphomas between 2004 and 2014. BL patients with private insurance had significantly better OS compared to those without. In patients aged <65 years, hazard ratios were 1.4 for uninsured status (95% confidence interval 1.2–1.7), 1.2 for Medicaid (95% CI 1.0–1.4), and 1.5 for Medicare (95% CI 1.2–1.9). For patients aged >65, hazard ratio for uninsured status was 8.4 (95% CI 2.5–28.3). Conversely, underinsured PBL patients experienced no difference in OS. Thus, expanding insurance-related access to care may improve survival in BL, for which curative therapy exists, but not PBL, where more effective therapies are needed. Our findings add to mounting evidence that adequate health insurance is particularly important for patients with curable cancers.

Introduction

A patient’s insurance status represents a major factor in the utilization of cancer therapies and outcomes in the United States.[1–3] Uninsured and Medicaid-insured cancer patients are more likely to be diagnosed at an advanced stage than those with private insurance.[1] Additionally, patients without private insurance experience longer lag time between cancer-specific symptom onset and diagnosis when compared to those who are privately insured.[4] Similar disparities have been described in non-Hodgkin lymphoma (NHL), the most common hematologic malignancy in the U.S., with an estimated 125,850 cases diagnosed in 2016.[5] According to studies of the National Cancer Database (NCDB), NHL patients with private insurance were up to 8 times more likely than those with Medicaid or no insurance to receive immunotherapy such as the monoclonal CD20 antibody rituximab, [6,7] a treatment known to improve survival for many NHL patients.[8,9] Thus, patient insurance status can influence survival disparities at multiple levels, as disease stage and treatment are important prognostic factors.[10,11] For patients with diffuse large B-cell (DLBCL) and Hodgkin lymphoma—two aggressive, but curable, lymphoid malignancies—Medicaid and no insurance have both been associated with inferior survival when compared to private insurance;[12,13] however, limited data exist regarding the impact of insurance status on other aggressive lymphoma subtypes.

Burkitt lymphoma (BL) is an aggressive NHL with an estimated 1480 new cases diagnosed in the US in 2016.[5] BL is simultaneously one of the most aggressive lymphomas, with a tumor volume doubling time of just 24 hours, and one of the most curable, with clinical trials showing 3-year survival rates over 80%.[14,15] However, a nationwide cohort study estimated 3-year OS at just 56%, indicating a significant discrepancy between clinical trial and “real-world” outcomes.[15] Plasmablastic lymphoma (PBL) is a rare, aggressive form of NHL with an estimated 180 cases diagnosed in 2016.[5] Like BL, PBL often occurs in HIV-positive patients and is commonly associated with Epstein Barr virus (EBV) infection. Untreated, PBL has an OS of just 3–4 months.[16] Prognosis remains poor even with intensive therapy, with a median OS of 8 months.[17]

Delay in therapy and failure to receive adequate treatment can have serious implications for aggressive lymphomas; thus, it is essential to understand the factors that lead to poor outcomes. Despite evidence that a patient’s insurance status is an important prognostic consideration in other aggressive lymphomas, literature is scarce on the impact of insurance status on BL outcomes. To address this issue, we investigated the effect of insurance status on survival in adults with BL, compared the impact of insurance status on BL outcomes to that seen in PBL, and examined the interaction between insurance status, HIV status, race, and socioeconomic status (SES) on BL survival.

Methods

Data Source

We used data from the NCDB, a nationwide, hospital-based cancer registry jointly sponsored by the American Cancer Society and American College of Surgeons. The NCDB contains 34 million historical records and captures 70% of newly diagnosed cancer cases in the U.S., obtaining data from more than 1,500 Commission on Cancer (CoC)-accredited facilities. These facilities report patients’ vital status and date of death to the NCDB annually.[18]

Study Population

Patients were identified using the third edition of the International Classification of Disease for Oncology (ICD-O-3) histology codes 9687 for BL and 9735 for PBL, following the International Lymphoma Epidemiology Consortium (InterLymph) hierarchy and the 2008 World Health Organization classification of lymphoid neoplasms. Patients were eligible for the study if age was >18 years old, diagnosis occurred between 2004–2014, and BL or PBL represented the primary tumor. Patients whose reporting facility was not CoC-accredited in the follow-up years for vital status from 2004–2014, or whose primary payor was unknown, missing, or non-Medicare/Medicaid government insurance (e.g., Veterans Administration or Indian Health Services), were excluded (Figure 1).

Figure 1. CONSORT diagram.

depicting Burkitt lymphoma case selection process. A similar process was also executed for cases of plasmablastic lymphoma.

Study Variables

Insurance status at diagnosis was grouped into the following primary payer categories: private insurance, no insurance, Medicaid, and Medicare. Race/ethnicity was categorized into white, Hispanic, black, and other. We used the proportion of adults without a high school diploma (HSD) according to zip code of residence as a regional measure of SES.[19,20] This data was obtained from the 2012 American Community Survey and classified into categories of <7%, 7 – 12.9%, 13–20.9%, and >21%, as described previously.[21] HIV status was recorded as positive, negative, or unknown. Disease stage was defined according to staging groups from the American Joint Committee on Cancer’s Cancer Staging Manual and separated into early (I/II) and advanced stages (III/IV).[22] Pre-existing medical conditions and/or complications were recorded, and a Charlson-Deyo comorbidity score was calculated.[23] Initial treatment status was defined as treatment (chemotherapy ± immunotherapy) or no treatment. Overall survival was calculated in months as time from date of diagnosis to either the date of death, date of last contact, or December 31, 2014, whichever occurred first.

Statistical Analysis

Chi-squared tests were used to compare the sociodemographic and clinical characteristics of the study cohort by insurance status. Since patients aged 65 years and older are eligible for Medicare coverage, all further analyses were stratified into groups based on age < or ≥ 65 years. Only two patients aged ≥ 65 years with PBL had no insurance, so no meaningful analysis could be performed for this subset. Kaplan-Meier survival curves were stratified by insurance status and log-rank tests were performed. The Cox proportional hazard assumption was checked for all variables. Univariable Cox proportional hazard models were generated to describe the unadjusted associations for the covariables, and multivariable Cox proportional hazard models were generated to estimate the hazard ratio (HR) associated with insurance status when adjusted for other covariables. Models were fitted using stepwise addition of variables; variables were included if significance criteria of 0.10 was met. The following covariates were considered for inclusion: sex, race, education level, HIV status, presence of B symptoms (fever, night sweats, and/or weight loss), stage, Charlson-Deyo comorbidity score, initial treatment status, and time from diagnosis to treatment. We then performed the same analyses for PBL.

Additional analyses were performed by generating multivariable Cox proportional hazard models stratified based on age < / ≥ 65 years to estimate HRs for BL survival associated with insurance status, HIV status, race, and education level, as well as their interactions. Interactions were included in the model if significance criteria of 0.05 was met using a Wald test. Propensity score-matched Cox proportional hazard modeling was also performed to control for observed confounding factors. To further assess the interaction between insurance status and HIV status, the resulting models were stratified by HIV status. Among BL patients ≥ 65, no HIV-positive patients were uninsured, so no meaningful analysis could be performed on this subset. Sensitivity analysis including only HIV-negative patients in each age group was performed to evaluate uncontrolled and unknown confounders in this patient population.

All statistical analyses were performed using R version 3.3.2 software (R Project for Statistical Computing). The threshold for statistical significance was set at α = 0.05.

Results

Burkitt lymphoma

We identified 7073 patients diagnosed with BL between 2004–2014, of whom 51.4% had private insurance, 13.2% had Medicaid, 27.7% had Medicare, and 7.7% had no insurance. Of the 5235 BL patients aged < 65, 65.0% had private insurance, 17.2% had Medicaid, 7.6% had Medicare, and 10.2% had no insurance. Of the 1838 BL patients aged ≥ 65, 12.9% had private insurance, 1.5% had Medicaid, 85% had Medicare, and 0.65% had no insurance.

Through 2014, 3392 (48%) patients with BL had died. Uninsured and Medicaid-insured patients were more likely to be Hispanic or black, have lower SES, have B symptoms, be HIV-positive, and have a comorbidity score ≥ 2 when compared with privately insured patients. Medicare patients were less likely to be male, have zero comorbidities, be HIV-positive, and receive systemic treatment compared to privately insured patients (Table 1).

Table 1.

Differences by insurance status in demographic and disease characteristics of patients with Burkitt lymphoma in the National Cancer Database (NCDB), 2004–2014.

Variable	Total (N=7073)	Private (%)	Uninsured (%)	Medicaid (%)	Medicare (%)	p-value
Age at diagnosis
18–64	5235	93.5	97.8	97	20.2	<0.0001
65–90	1838	6.5	2.2	3	79.8
Sex
Male	5099	75.2	78.6	73.4	64	<0.0001
Female	1974	24.8	21.4	26.6	36
Race/ethnicity
White	4837	73.2	46.4	47.8	75.3	<0.0001
Hispanic	764	8.4	26.9	19.7	6.6
Black	786	9	19.6	23.8	6.7
Other	637	8.7	7.1	7.9	10.6
Unknown	49	0.7	0	0.8	0.8
Percent without high school diploma
< 7	1668	28.8	12.1	12.7	22.3	<0.0001
7–12.9	2091	31.7	22.5	21.5	31.4
13–20.9	1799	22.8	29.3	31.4	26.4
>21	1370	14.7	33.8	32.9	17.6
Unknown	145	2	2.4	1.6	2.2
B symptoms
Yes	2541	35	41.5	41.9	33.2	<0.0001
No	3629	53.2	48.1	43.6	52.4
Unknown	903	11.8	10.4	14.5	14.4
HIV status
Positive	1760	24.7	37.7	50.4	9.6	<0.0001
Negative	3360	48.9	42.4	32.1	53.7
Unknown	1953	26.4	19.9	17.5	36.7
Stage
I/II	2044	30.5	28.7	23.6	28.6	0.0003
III/IV	4267	59.3	59	67.1	59.4
Unknown	762	10.3	12.2	9.2	12.1
Comorbidity score
0	4986	75.4	70.4	59.5	66.6	<0.0001
1	909	10.2	8.8	8.6	21
2+	1178	14.4	20.8	31.9	12.4
Initial treatment
Systemic	6122	90.1	86.1	88.9	78.9	<0.0001
None	838	8.1	12.8	9	20
Days between diagnosis and treatment
0–14	4523	66.8	63.8	64.6	58.5	0.2132
15–30	1160	16.7	17.6	16.4	15.5
>30	618	8.3	8.6	8.6	9.6
Unknown	772	8.2	10.1	10.4	16.4

For BL patients aged < 65 years, OS was 73.6% at 1 year, 66.6% at 3 years, and 63.9% at 5 years. For BL patients aged < 65 years with private insurance, Medicaid, Medicare, and no insurance, 5-year OS was 67.4%, 60.6%, 37.9%, and 58.2%, respectively (Figure 2A). Uninsured, Medicaid-insured, and Medicare-insured BL patients aged < 65 years had a significantly increased risk of death compared to patients with private insurance (Table 2). Being uninsured, Medicaid-insured, or Medicare-insured remained predictive of inferior OS even after controlling for sociodemographic and clinical factors. Lower SES, advanced stage, HIV-positive status, B symptoms, lack of treatment, and having 1 or more comorbidities were significant, independent predictors of survival in BL patients aged < 65. In propensity score-matched analysis, lack of private insurance remained a significant predictor of overall survival in these patients (Supplemental Table 1). In multivariable sensitivity analysis that included only HIV-negative patients younger than 65, Medicare remained significant in its impact on overall survival (HR 1.98, 95% CI [confidence interval] 1.5–2.6), with a trend towards inferior survival for uninsured patients (HR 1.3, 95% CI 1–1.7).

Figure 2. Overall survival by insurance status for patients with Burkitt lymphoma (BL) or plasmablastic lymphoma (PBL) in the NCDB, 2004–2014.

(A) BL patients aged < 65. (B) BL patients aged ≥ 65. (C) PBL patients aged <65. (D) PBL patients aged ≥ 65. P-values obtained by log-rank test.

Table 2.

Cox regression models of predictors of mortality in Burkitt lymphoma patients aged < 65 in the NCDB, 2004–2014.

	Univariable		Multivariable
Variable	HR (95% CI)	p-value	HR (95% CI)	p-value
Insurance status
Private	---		---
Uninsured	1.52 (1.3,1.7)	<0.0001	1.44 (1.2,1.7)	0.0003
Medicaid	1.55 (1.4,1.7)	<0.0001	1.22 (1,1.4)	0.0192
Medicare	2.06 (1.8,2.4)	<0.0001	1.53 (1.2,1.9)	<0.0001
Sex
Men	---
Women	0.88 (0.8,1)	0.0182
Race/ethnicity
White	---		---
Hispanic	1.01 (0.9,1.2)	0.9355	0.75 (0.6,0.9)	0.0052
Black	1.48 (1.3,1.7)	<0.0001	1.01 (0.9,1.2)	0.8967
Other	0.84 (0.7,1)	0.0473	0.93 (0.7,1.2)	0.5854
Percent without HSD
< 7	---		---
7–12.9	1.21 (1.1,1.4)	0.0047	1.1 (0.9,1.3)	0.2955
13–20.9	1.63 (1.4,1.9)	<0.0001	1.35 (1.1,1.6)	0.0012
>21	1.81 (1.6,2.1)	<0.0001	1.39 (1.1,1.7)	0.001
B symptoms
No	1.00 (ref)		---
Yes	1.52 (1.4,1.7)	<0.0001	1.19 (1,1.3)	0.0072
HIV Status
Negative	---		---
Positive	1.69 (1.5,1.9)	<0.0001	1.28 (1.1,1.5)	0.0014
Stage
I/II	---		---
III/IV	2.29 (2,2.6)	<0.0001	2.16 (1.8,2.5)	<0.0001
Comorbidity score
0	---		---
1	1.68 (1.5,1.9)	<0.0001	1.4 (1.2,1.7)	0.0006
2+	1.81 (1.6,2)	<0.0001	1.14 (1,1.3)	0.1193
Initial treatment
Systemic	---		---
Other	2.95 (2.6,3.4)	<0.0001	3.19 (1.2,8.6)	0.0226
Days from diagnosis to treatment
0–14	---		---
15–30	0.76 (0.7,0.9)	<0.0001	0.81 (0.7,0.9)	0.0076
>30	0.68 (0.6,0.8)	<0.0001	0.88 (0.7,1.1)	0.17

Abbreviations: hazard ratio (HR); confidence interval (CI); HSD (high school diploma); reference (---).

For patients with BL aged ≥ 65 years, OS was 45.7% at 1 year, 38.6% at 3 years, and 33.6% at 5 years. The 5-year OS rates for patients aged ≥ 65 years with private insurance, Medicaid, Medicare, and no insurance were 45.0%, 24.5%, 32.3%, and 21.4%, respectively (Kaplan-Meier curves shown in Figure 2B). Medicare-insured patients had a significantly increased risk of death when compared to those with private insurance (Table 3). Controlling for sociodemographic and clinical factors, being uninsured remained a significant predictor of worse OS. Advanced stage, B symptoms, having ≥ 2 comorbidities, and treatment type were significant, independent predictors of survival in BL patients aged ≥ 65 years. In propensity score-matched analysis for this group, insurance categories were limited to either private insurance or Medicare since the small number of patients with either Medicaid or no insurance precluded adequately balanced groups, but Medicare remained a significant predictor of overall survival in these patients (Supplemental Table 1). In multivariable sensitivity analysis that included only HIV-negative patients older than 65, lack of insurance remained significant in its impact on overall survival (HR 6.59, 95% CI 2–21.7).

Table 3.

Cox regression models of predictors of mortality in Burkitt lymphoma patients aged ≥ 65 in the NCDB, 2004–2014.

	Univariable		Multivariable
Variable	HR (95% CI)	p-value	HR (95% CI)	p-value
Insurance Status
Private	---		---
Uninsured	1.88 (0.9,3.8)	0.0828	8.38 (2.5,28.3)	0.0006
Medicaid	1.3 (0.8,2.1)	0.2856	0.75 (0.3,1.8)	0.5264
Medicare	1.3 (1.1,1.5)	0.0043	1.29 (0.9,1.8)	0.1211
Sex
Men	---
Women	0.98 (0.9,1.1)	0.7344
Race/ethnicity
White	---
Hispanic	0.9 (0.7,1.1)	0.3972
Black	0.94 (0.7,1.3)	0.6808
Other	1.04 (0.9,1.2)	0.6192
Percent without HSD
< 7	---		---
7–12.9	0.95 (0.8,1.1)	0.4844	0.8 (0.6,1)	0.0006
13–20.9	1.21 (1,1.4)	0.0151	0.9 (0.7,1.2)	0.0006
>21	1.04 (0.9,1.2)	0.6988	0.72 (0.5,1)	0.0006
HIV status
Negative	---		---
Positive	1.3 (0.9,1.9)	0.1584	0.96 (0.5,1.7)	0.8793
B symptoms
No	---		---
Yes	1.3 (1.2,1.5)	<0.0001	1.34 (1.1,1.7)	0.0059
Stage
I/II	---		---
III/IV	1.65 (1.4,1.9)	<0.0001	1.63 (1.3,2.1)	<0.0001
Comorbidity score
0	---		---
1	1.19 (1,1.4)	0.0098	0.98 (0.8,1.2)	0.8667
2+	1.9 (1.6,2.3)	<0.0001	1.67 (1.2,2.3)	0.0021
Initial treatment
Systemic	---		---
Other	4.91 (4.3,5.6)	<0.0001	7.46 (3.2,17.5)	<0.0001
Days from diagnosis to treatment
0–14	---		---
15–30	0.66 (0.6,0.8)	<0.0001	0.74 (0.6,1)	0.0183
>30	0.56 (0.5,0.7)	<0.0001	0.69 (0.5,0.9)	0.0116

Abbreviations: hazard ratio (HR); confidence interval (CI); HSD (high school diploma); reference (---).

Plasmablastic lymphoma

We identified 475 patients with PBL diagnosed between 2004–2014 in the NCDB, of whom 32.8% had private insurance, 21.3% had Medicaid, 38.3% had Medicare, and 7.5% had no insurance. Of the 314 patients with PBL aged < 65, 44.0% had private insurance, 30.2% had Medicaid, 14.6% had Medicare, and 10.8% had no insurance. Of the 161 patients with PBL aged ≥ 65, 10.6% had private insurance, 3.7% had Medicaid, 84.5% had Medicare, and 1.2% had no insurance (Supplemental Table1).

For patients with PBL aged < 65 years, OS was 48.8% at 2 years. For patients with PBL ≥ 65 years, OS was 37.7% at 2 years. For both age groups, uninsured, Medicaid- and Medicare-insured PBL patients had similar OS when compared to patients with private insurance (Figure 2C & 2D). In multivariable models that included sociodemographic, prognostic, and treatment factors, insurance status was not associated with differential survival, regardless of age group. Advanced stage was a significant, independent predictor of worse OS in adults with PBL (Supplemental Tables 1 and 2). Presence of B symptoms and lack of treatment were also associated with worse OS for PBL patients aged ≥ 65 (Supplemental Table 2).

Interaction between insurance, HIV status, race and socioeconomic status

Interaction between insurance and HIV statuses for BL patients aged < 65 years met the Wald test significance criteria and was included in the model. Specifically, interaction between Medicare and HIV status was significant. Performing analyses stratified on HIV status, the HR comparing Medicare to private insurance was much higher among HIV-negative patients (2.4, CI 1.9, 3.0) than among their HIV-positive counterparts (1.4, CI 1.1, 1.7). Meanwhile, uninsured and Medicaid-insured patients had similar HRs in comparison to privately insured among HIV-negative patients (1.3 and 1.2, respectively) and HIV-positive (1.4 and 1.2, respectively). Interactions between insurance status and HIV status as well as between HIV status and race met the Wald test significance criteria for patients aged ≥ 65 and were included the model. Specifically, interactions between Medicaid insurance and HIV-positive status (HR 26.2) as well as between black race and HIV-positive status (HR 4.4) were significant.

Discussion

To the authors’ knowledge, this is the first nationwide hospital-based US study to examine the relationship between insurance status and outcomes for BL or PBL and the first study performed using the NCDB for these malignancies. Among BL patients aged < 65 years, those with no insurance, Medicaid, or Medicare had significantly worse OS compared to those with private insurance after controlling for sociodemographic, clinical, and treatment factors. Among BL patients aged ≥ 65, uninsured patients had worse OS compared to those with private insurance. Conversely, among patients with PBL, no difference was observed in OS by insurance status. As the government’s role in providing insurance continues to be a focus of national political discourse, we must understand the impact insurance status has on oncologic outcomes. Our study adds to evidence that adequate health insurance is of particular importance for patients with curable cancers such as BL. This concept is highlighted by our analyses in PBL, in which insurance status does not impact survival, likely because novel treatments are needed to improve its dismal outcomes.

Underinsured BL patients (i.e., those with no insurance, Medicaid, or Medicare) aged < 65 were more likely to have B symptoms, present at an advanced stage, be HIV-positive, have multiple comorbidities, and have low SES, which may explain some, though not all, of the observed difference in OS. Patients with Medicaid were more likely to present at an advanced stage than the other groups. This could reflect uninsured patients delaying diagnosis by waiting to qualify for Medicaid before seeking medical care, leading to worse-appearing outcomes than expected.[24] Underinsured BL patients aged < 65 years were less likely to receive chemotherapy than those with private insurance. This is a concerning finding, since effective intensive treatment substantially improves BL outcomes.[25–27] Fittingly, we found lack of systemic therapy was associated with the highest HR (3.19, CI 1.2, 8.6). For a chemo-sensitive disease like BL, outcome disparities may reflect shortcomings of the US healthcare system. Health policy can aid in expanding access to care and reducing survival disparities in BL. Restructuring the US healthcare system to reduce the percentage of uninsured or underinsured patients, particularly with passage of the Affordable Care Act in 2010, has improved care accessibility.[28,29] Increased coverage has led to earlier oncologic diagnosis and more effective treatment [30,31] as well as to improvements in mortality, with the largest improvements seen in ‘healthcare-amenable’ conditions like curable cancer.[32,33] For BL patients ≥ 65 years, Medicare patients resembled those who were privately insured, although they were more likely to have comorbidities, which may have contributed to the observed difference in outcomes.

In addition to our findings on insurance status, our study contributes new information regarding BL prognostic factors in a nationwide registry. We identified a comorbidity score ≥ 2 and presence of B symptoms as predictors of worse OS for adults of all ages with BL, and identified having a comorbidity score of 1 as a predictor of worse outcomes for BL patients aged <65 years. Few studies have examined nationwide predictors of survival for BL since adding rituximab to standard treatment regimens, limiting the ability to counsel patients on prognosis and risk-stratify for ongoing research. Given the heterogenous outcomes for BL at the population level, better understanding the factors driving disparities is necessary. Castillo et al. proposed a risk-stratification model based on race, age, and clinical stage.[34] Our study corroborates the importance of stage, as well as receipt of chemotherapy, as prognostic factors for BL. Insurance status, B symptoms, and comorbidity score should be evaluated for incorporation in future prognostic indices for BL.

Our results confirm advanced stage and newly identified comorbidity score ≥ 2 as predictors of worse outcomes in PBL.[35] Despite overall poor outcomes in PBL, systemic treatment remains associated with improved survival. HIV-positive status loses significance in the multivariable model when removing treatment, suggesting a possible interaction between the variables; however, sample size provided inadequate power to assess this.

As a retrospective, registry-based study, our study has several limitations. First, the findings may not be fully generalizable to the public, as patients treated at CoC-accredited hospitals may underrepresent the most underserved patients. Second, we were unable to completely assess the relationship between insurance status, treatment, and outcome. Rituximab was coded as chemotherapy rather than immunotherapy for most years included in our study, and the NCDB does not provide specific treatment regimens, a possible mediating factor between insurance status and outcomes. Third, the NCDB lacks data on other potential confounders, such as health literacy and individual-level SES. Furthermore, insurance status is recorded in the NCDB at time of diagnosis, so we were unable to assess the impact of changes in insurance status over time, or coverage from multiple sources.

Prior studies implicate access to care as a crucial factor for outcomes in BL. Costa et al. identified an outcome gap between population-based survival and survival observed in clinical trials for BL, likely from improved access to appropriate care experienced by clinical trial participants.[15] Mukhtar et al. found that survival disparities exist for black adults aged 40–70 with BL when compared with white counterparts, but not in comparisons of pediatric or elderly patients.[36] This could be explained by disparities in access to care, suggested to be a greater issue for adults than for children or elderly, due to increased variation in SES and access to insurance in that age group.[36–39] Black BL patients aged < 65 years had significantly worse OS in our univariate model, but not in our multivariable model, suggesting that insurance status, SES, and HIV status – not included in the analysis by Mukhtar et al.– may explain some of these survival disparities.[36] This was confirmed by removing insurance status, education level, and HIV status from our multivariable model, after which black patients showed significantly worse OS (HR 1.20, p=0.028).

Evidence remains mixed on the impact of HIV on BL outcomes.[40,41] Our results suggest that HIV-positive status may contribute to worse outcomes for BL patients aged < 65. HIV-positive patients aged <65 with Medicare had better outcomes relative to private insurance, which was not observed in their HIV-negative counterparts. This interaction may reflect substantial comorbidities in HIV-negative patients aged <65, since such patients would be likely to qualify for Medicare through another severe illness, injury or disability before reaching age 65. Our results also highlight Medicare-related issues concerning both coverage and care for HIV-positive adults. All HIV-positive patients diagnosed with BL should qualify for Medicare; however, just 9% of HIV-positive patients in our cohort received Medicare, while 40% remained uninsured or had Medicaid coverage. Further, HIV-positive patients with Medicare (HR 1.36, CI 1.07–1.73) have similar outcomes in comparison with those with no insurance (HR 1.35, CI 1.09–1.69), and worse outcomes than those with Medicaid (HR 1.22, CI 1.03–1.43).

Flowers and Nastoupil proposed a framework for understanding disparities in DLBCL outcomes, through interactions between biological, environmental, individual and social factors, that can be similarly applied to BL.[42] HIV is strongly associated with BL, as well as with social and environmental factors. Our analysis underscores the importance of this interplay, as we identified significant interactions between HIV status, Medicaid insurance and black race among BL patients aged ≥ 65. Our results suggest that elderly HIV-positive BL patients with Medicaid or of black race have worse outcomes than their HIV-negative counterparts, relative to private insurance or white race, respectively. A large-scale population-based dataset that includes clinical, pathologic, and outcome information is essential to better describe the interplay between sociodemographic, biologic, and clinical factors in their impact on prognosis. Additional research on barriers to care and its interaction with biological and clinical factors is necessary to further elucidate their impact on survival disparities in BL.[42]

Although BL and PBL share many characteristics, our results suggest that expanding access to care is unlikely to improve outcomes in PBL in the same way that it would in BL. The disparate impact of insurance status on survival for these malignancies is likely due to differences in the availability of effective therapeutic strategies. Although a single standard front-line therapy does not exist for BL, several intensive regimens have demonstrated excellent response rates (>85%) and 3-year OS (> 75%), including CODOX-M/IVAC, R-HyperCVAD, and dose-adjusted R-EPOCH.[25–27,43,44] Addition of rituximab has further improved 5-year relative survival in BL.[15] Immunotherapies like rituximab are a driver of disparities in lymphoma outcomes, with underinsured, black, and low-SES patients less like to receive it as treatment.[6] Inequity in immunotherapy usage may contribute to the inferior survival observed in underinsured and low-SES BL patients and should be the focus of future studies. Meanwhile, median OS in PBL remains in the range of a few months, even with intensive chemotherapy regimens.[17] Since PBL is a CD20-negative neoplasm, the improvement in outcomes seen in the rituximab era with other NHL has not translated to survival gains. Novel treatment strategies are desperately needed to improve PBL outcomes.

We have identified insurance status as an important predictor of outcomes in BL, and our findings suggest that expanded access to care may alleviate survival disparities for BL patients. Further examination of the relationship between practice patterns in BL and outcomes can enlighten public policies to make oncologic care more accessible and its delivery more effective. Future research must focus on the impact of other barriers of access to care, such as cost or availability of healthcare services, as well as the effect of health policy changes, on survival in Burkitt lymphoma.