Comparing the distribution of Hodgkin lymphoma subtypes and survival outcomes between Latin American and US patients Open Access

Classical Hodgkin lymphoma (HL) is a neoplasm with a complex epidemiology pattern.¹ Among its 4 morphologic subtypes, mixed cellularity HL (MCHL) has been associated with Epstein-Barr virus (EBV) infection.² Previous studies suggest that EBV-associated HL is more frequent in resource-limited settings.^3,4 However, existing data primarily derive from single-center studies with limited geographical representation and often lack comparisons of the distribution of HL subtypes across populations.^3-6 Although Hispanic individuals exhibit a higher frequency of EBV-associated HL than White populations,⁷ the absence of lymphoma registries in Latin America complicates direct comparisons for specific subtypes. Consequently, it remains unclear whether the distribution of HL subtypes varies across world regions.

The significant advancements in therapy for HL have led to survival improvements in high-income countries, but outcomes in Latin America remain poorly understood. In the United States, 5-year net survival rates exceed 90% among predominantly non-Hispanic White patients.^8,9 Similarly, a previous study across European countries reported a 10-year relative survival rate for HL of 83% in nations with high health care expenditure, compared with 75% in those with lower expenditure.¹⁰ Survival outcomes are challenging to assess in Latin America, partly due to the lack of high-quality cancer registries¹¹ and studies broadly grouping lymphoma subtypes with varying prognoses as a unique category, limiting meaningful comparisons of HL survival outcomes with other regions.¹² The persistent limited access to timely or novel treatment approaches and fragile health care systems in Latin America^13,14 may negatively affect the survival of patients with HL. Whether survival among Latin American patients with HL is similar to those managed in other developed world regions remains unknown.

Previous studies in the United States have identified the negative impact of some social determinants of health (SDOHs) on HL survival, such as having public or no health insurance or low neighborhood socioeconomic status,^15,16 but few efforts have been conducted in other settings. Because SDOHs may be influenced by cultural or social paradigms, it remains to be determined whether SDOHs identified in high-income countries, such as the United States, apply to developing countries. Additionally, because MCHL has been associated with worse outcomes,¹⁷ it is unclear whether survival differences across regions may be related to the distribution of HL subtypes. To address these gaps, we conducted an international pooled analysis comparing the distribution of HL subtypes, clinical features, and outcomes between Latin American and US patients. Additionally, we explored the effect of insurance status on survival estimates in Latin America only. Understanding differences in clinical presentation, survival, and the role of SDOH across world regions is essential to improve HL outcomes globally.

We conducted a retrospective cohort study among patients aged ≥16 years with newly diagnosed classical HL from 2010 to 2020 in Latin America and the United States. Data for Latin America were obtained from the Grupo de Estudio Latinoamericano de Linfoproliferativos (GELL; in Spanish) across 8 centers in 6 countries (Argentina, n = 255; Cuba, n = 63; Paraguay, n = 25; Peru, n = 316; Uruguay, n = 89; and Venezuela, n = 70). In each of the GELL centers, all patients were included consecutively, were managed in academic centers, and clinical information was collected retrospectively in standardized forms. Because of the largely centralized care in Latin America, most patients are often referred for diagnosis and management. Data for the United States were extracted from 17 population-based registries of the Surveillance, Epidemiology, and End Results (SEER) program (metropolitan area of Atlanta, Greater California Regional, San Francisco-Oakland, San Jose-Monterey, Los Angeles, Greater Georgia, Rural Georgia, Louisiana, Kentucky, New Jersey, New Mexico, Seattle-Puget Sound, Connecticut, Hawaii, Iowa, Alaska Natives, and Utah). In SEER, we used the third edition of the International Classification of Diseases for Oncology codes to identify nodular sclerosis HL (NSHL; 9663/3, 9664/3, 9665/3, and 9667/3), lymphocyte-rich HL (LRHL; 9651/3), MCHL (9652/3), lymphocyte-depleted HL (LDHL; 9653/3, 9654/3, and 9655/3), and classical HL, not otherwise specified (9650/3, 9661/3, and 9662/3). Follow-up was available through 2023 in GELL and through 2021 in SEER.

We initially compared HL subtypes and clinical variables with data collected at diagnosis. For survival analysis, we included patients with available follow-up data and those whose first course of treatment was chemotherapy with or without radiotherapy. Figure 1 details the exclusion criteria.

US patients were stratified into US non-Hispanic Asian (US Asian), US non-Hispanic Black (US Black), US Hispanic, and US non-Hispanic White (US White) patients because of the previously described survival differences.^18-20 Cancer stage was classified using Ann Arbor staging in GELL, and it was identified in SEER using the Ann Arbor and SEER Summary Stage classification variables. Given the lack of clinical variables in SEER (eg, bulky disease or erythrocyte sedimentation rate), we could not stratify patients into risk subgroups.² Insurance status (public vs private) was the only SDOH factor evaluated.

The outcomes of interest were the prevalence of HL subtypes and overall survival (OS). Prevalence was estimated among patients with available data on classical HL subtypes. OS was defined from diagnosis to death from any cause.

We performed a propensity score matching because of differences in clinical variables between Latin America and US race and ethnic subgroups in the unmatched survival cohort. Latin American patients had a higher percentage of advanced-stage disease (Ann Arbor III-IV; P < .001) and MCHL (P < .001; supplemental Table 1). To reduce confounding and facilitate subgroup analyses, we propensity matched Latin American patients to each US race and ethnic group. This matching process involved four 1:1 matches using the nearest neighborhood method. Matching variables included age group, biological sex, year of diagnosis, cancer stage, and HL subtypes. All matched variables between Latin American and combined US patients had a standard mean difference <0.01. However, a standard mean difference >0.01 was observed between Latin American and US Asian patients for those aged 16 to 39 years and those with MCHL and classical HL not otherwise specified (supplemental Figure 1).

We used the χ² test to compare categorical variables and the Kruskal-Wallis test to compare age. The prevalence of HL subtypes from the GELL cohort was evaluated within each country. We estimated 95% confidence intervals (CIs) for multinomial proportions using the methods of Sison and Glaz. To assess whether estimates differ from those of US White individuals, we fitted multivariable Poisson regression models for each subtype, adjusting for age and sex. The log of the population size was used as an offset. We report risk ratios (RRs) and 95% CIs. Venezuela only reported NSHL and was excluded from this analysis.

Survival analyses were performed in the unmatched and matched cohorts. The reverse Kaplan-Meier method was used to estimate the median follow-up. Survival probabilities were analyzed using the Kaplan-Meier method and the log-rank test. Given the propensity nature of the matched cohort, we fitted univariable Cox regression models. Findings from the Cox regression models are reported with hazard ratios (HRs) and 95% CIs. We compared OS between Latin American and combined US patients, but we also explored outcomes by race and ethnicity stratification. For survival curves, pairwise comparisons were performed using the Benjamini-Hochberg adjustment for P values only in the matched cohort. In the Cox models, US White patients were used as the reference category because of the larger sample size, relatively higher 5-year OS observed in the matched analysis, and their frequent use as a reference group in previous studies comparing outcomes across race and ethnic groups. Subgroup analyses involved comparing OS by sex, age group, cancer stage, and HL subtype stratification. We estimated a z score to evaluate whether HRs from the unmatched and matched analyses differ.

The effect of insurance (public vs private) was evaluated in the Latin American cohort only because we lack access to the insurance variable in the SEER data. We excluded 1 patient due to missing insurance data and performed a complete case analysis. We fitted a multivariable Cox regression model adjusting for age, sex, cancer stage, and HL subtypes. We excluded 1 patient from this analysis due to missing insurance data. P values <.05 were considered statistically significant. All analyses were performed in R (version 4.4.1).

The institutional review board of each Latin American participating center approved the development of this study. The SEER program provides deidentified and publicly available data, and its use does not require review by an institutional review board.

We analyzed data at diagnosis from 818 Latin American, 1141 US Asian, 2066 US Black, 3647 US Hispanic, and 11 761 US White patients (Table 1). In the entire cohort, the median age at diagnosis was 38 years (range, 16-91), most patients (53%) were aged 16 to 39 years, and we observed a male predominance (54%). Cancer stage I to II was diagnosed in 54% of all patients.

Male predominance was similar across races and ethnicities (52%-55%; P = .703). The median age at diagnosis was lowest among US Asian patients (32 years) and relatively similar for other groups, ranging from 36 to 39 years (P < .001). Cancer stage III to IV was more frequently diagnosed among Latin American (55%), US Black (54%), and US Hispanic patients (50%) than US White (43%) and US Asian patients (42%; P < .001). Chemotherapy with or without radiotherapy was administered to 95% of Latin American patients compared with 84% to 88% across races and ethnicities. The most common chemotherapy regimen used in 96% of Latin American patients was ABVD (doxorubicin, bleomycin, vinblastine, and dacarbazine).

Of 13 152 patients with available data on HL subtypes, NSHL (79%) was the most common, followed by MCHL (16%), LRHL (4%), and LDHL (1%; Table 1). The proportion of classical HL not otherwise specified was lower in the Latin American cohort (12%; range, 32%-38%; P < .001). Of available classical HL subtypes, MCHL was more frequently identified among Latin American (26%) and US Hispanic patients (19%), whereas the percentage ranged from 13% to 15% for other races and ethnicities (P < .001). NSHL remained the most common subtype across all groups (68%-81%).

We estimated the percentage across the included countries to further evaluate the distribution of HL subtypes in the Latin American region (Figure 2). A higher proportion of MCHL was identified in Peru (47%; 95% CI, 40-54; RR, 3.44; 95% CI, 2.81-4.17), Cuba (46%; 95% CI, 35-60; RR, 3.39; 95% CI, 2.29-4.80), and among US Hispanic patients (19%; 95% CI, 17-20; RR, 1.33; 95% CI, 1.19-1.48) than US White individuals. Paraguay had a statistically nonsignificant higher proportion of MCHL (20%; 95% CI, 8-38; RR, 1.77; 95% CI, 0.64-3.83; supplemental Table 2). The distribution of MCHL and NSHL in Argentina and Uruguay was similar to that of US White patients. Only US Black individuals had a higher percentage of LDHL than US White individuals (2%; 95% CI, 0-4; RR, 1.86; 95% CI, 1.13-2.94). No difference in the distribution of LRHL was observed across groups.

In the unmatched survival cohort (636 Latin American and 15 797 US patients), the median follow-up was 62 months (95% CI, 61-64). The 5-year OS of 87% (95% CI, 84-90) among Latin American patients was similar to the 88% (95% CI, 87-88) for all US patients (univariable HR, 1.06; 95% CI, 0.84-1.34; Figure 3A; supplemental Table 3). Analyses by race and ethnicity suggest a similar 5-year OS between Latin American and US White patients (87% vs 88%; univariable HR, 1.13; 95% CI, 0.89-1.43; Figure 3C; supplemental Table 4).

After propensity score matching, 3180 patients were available for analysis. With a median follow-up of 61 months (95% CI, 59-63 months), Latin American patients experienced a similar 5-year OS to all US patients (87% vs 87%; univariable HR, 1.01; 95% CI, 0.78-1.30; Figure 3B; supplemental Table 3). Similarly, we found a comparable 5-year OS between Latin American and US White patients (87% vs 90%; HR, 1.30; 95% CI, 0.91-1.84; Figure 3D; supplemental Table 4). Pairwise comparisons between Latin American patients and US race and ethnic groups did not identify a statistical difference in 5-year OS (supplemental Table 4).

Subgroup analyses were generally similar between Latin American and US patients in the unmatched and matched analyses (supplemental Figure 2; supplemental Table 3), except for a worse survival among Latin American patients aged 16 to 39 years than all US patients (matched analysis, 5-year OS, 92% vs 95%; HR, 1.71; 95% CI, 1.05-2.80). Matched analysis by race and ethnicity stratification identified worse outcomes among Latin American patients aged 16 to 39 years than US White patients (matched analysis, 5-year OS, 92% vs 98%; HR, 3.22; 95% CI, 1.38-7.50) and male patients (matched analysis, 5-year OS, 84% vs 90%; HR, 1.65; 95% CI, 1.04-2.63; supplemental Table 4). No statistically significant difference in survival was observed between Latin American and US White patients (supplemental Table 4) across other subgroups.

Five-year OS among Latin American patients with public health insurance was comparable to those with private health insurance (88% vs 86%; P = .500; Figure 4). After adjusting for age, sex, cancer stage, and HL subtypes, no difference in mortality risk was observed (HR, 1.00; 95% CI, 0.57-1.71).

The findings of this pooled retrospective study may suggest a distinct distribution of HL subtypes among Latin American patients from a hospital-based cohort, marked by a higher prevalence of MCHL than US White patients from a population-based registry. The prevalence of MCHL remained high for Peru, Cuba, and US Hispanic patients after country stratification, but Argentina and Uruguay had a similar distribution of HL subtypes compared with US White patients. Although Latin American patients had a higher frequency of poor clinical features (MCHL and advanced-stage disease), survival was comparable to US patients. Our findings remained consistent in the unmatched and matched analyses and after race and ethnicity stratification. Insurance status did not affect survival in the Latin American cohort. These findings suggest the need for prospective global etiologic and outcome research, particularly to validate our findings, understand the differential effect of environmental and host factors associated with disease development, and the risk of late effects and long-term outcomes among survivors in Latin America.

The heterogeneity in the prevalence of MCHL across Latin American countries may reflect the interaction of unknown population-specific environmental and host factors influencing HL development. The higher prevalence of MCHL in selected Latin American countries in our study (ie, Peru and Cuba) supports previous research indicating a greater prevalence of EBV-associated HL in the region.^3,4 For instance, the attributable fraction of EBV to HL is ∼60% in Latin America, compared with 32% in North America.⁴ However, the similarity in HL subtype distribution between certain Latin American countries (eg, Argentina and Uruguay) and US White patients in our findings is unclear. Similar to our findings, distinct lymphoma patterns within Latin America have been suggested in previous hospital-based studies, which reported a higher frequency of T-cell lymphomas and virus-associated lymphomas in Peru and Guatemala, whereas Argentina and Chile showed distributions of non-HL subtypes similar to those in North America or Western Europe.^21,22 Our findings extend these observations to HL subtypes, emphasizing the heterogeneity of lymphoma distribution within Latin America and the need to explore the role of environmental and host factors in HL development. However, caution should be exercised in interpreting our results because the Latin American cohort comprises hospital-based data, and population-based inference may be difficult to extrapolate. Further prospective research is needed to determine the prevalence of EBV status across HL subtypes at diagnosis among Latin American patients.

The higher percentage of advanced-stage HL at diagnosis observed in our hospital-based cohort likely reflects the difficulties of Latin American health care systems in detecting disease at earlier stages. Although this finding may be influenced by including national referral centers in our cohort, cancer care in Latin America remains largely centralized.¹⁴ Additionally, the centralization of care may contribute to patient overload, overburdening the centers’ resources and potentially delaying staging. Without high-quality cancer registries in the region,¹³ it is challenging to draw definitive conclusions regarding the distribution of advanced-stage HL across countries. Previous observations from retrospective hospital-based studies in non-HL and multiple myeloma also suggest a higher percentage of advanced-stage disease in Latin American cohorts.^23,24 Advanced disease at diagnosis represents a significant barrier to improving survival outcomes in Latin America, because novel immunotherapies for advanced-stage HL are costly and difficult to distribute within the region.¹³ The OS advantage of brentuximab vedotin over ABVD alone published in 2022²⁵ and the encouraging outcomes on progression-free survival with nivolumab published in 2024²⁶ among patients with advanced-stage HL suggest that survival outcomes will likely improve in high-income countries, but lack of access to these therapies in Latin America will possibly introduce survival differences in the future. Because our study only included the period from 2010 to 2020, continuing surveillance of outcomes between Latin American countries and high-income nations is warranted.

The overall comparable survival between Latin American and US patients highlights the effectiveness of standard first-line HL treatment approaches and opens the path for further research into survivorship outcomes in Latin America. Current treatment paradigms in HL have shifted toward reducing toxicities while maintaining high cure rates.² HL survivors are at increased risk of developing late effects, such as cardiotoxicity, pulmonary complications, second primary neoplasms, or reduced quality of life.²⁷ Few efforts have been conducted to identify the risk of these complications in Latin America. The retrospective nature of our database posed challenges, because quality of life, late effects, toxicity grading, and chemotherapy or radiotherapy dose information were not consistently captured in medical records. Future prospective studies should address these gaps and quantify survivorship outcomes in Latin American populations, considering the availability of therapies and the interplay of environmental factors unique to the region.

The lower survival rate among Latin American young adults and males than US patients in subgroup analyses is concerning and warrants further investigation. Previously identified factors affecting young adults with lymphoma, including age-related inequities in timely diagnosis and treatment, lack of support networks, or challenges in navigating the health care system,²⁸ may likely affect young adults in Latin America more profoundly. The inferior survival among Latin American males was only observed when we compared outcomes with US White patients. Although the reasons for this observation are unclear, possible factors may include cultural behaviors toward accessing the health care system, a more significant socioeconomic burden centralized in young males, or biological factors affecting response to treatment or the development of adverse effects. Future research is needed to elucidate the factors behind our exploratory findings.

The lack of an effect of insurance status on survival outcomes in the Latin American cohort may indicate a geographical variability of the impact of SDOHs on survival. This finding contrasts a previous study conducted in the United States, suggesting worse outcomes among patients with HL covered by public insurance than private insurance.¹⁶ However, it is relevant to note that most patients with public health insurance in our Latin American cohort were from a single country (ie, Peru), limiting the generalizability of this observation. Additionally, previous studies have identified variability in SDOHs between high- and low-income countries, affecting financial toxicity or treatment delay outcomes in cancer populations.^29,30 Further studies are needed to determine whether a given SDOH may have differential effects on survivorship outcomes across geographical locations to effectively design interventions and improve outcomes.

Our study has limitations to address. We compared hospital-based data with a population-based registry, and selection bias is possible for prevalence and survival analyses. However, high-quality cancer registries are lacking in Latin America, and existing resources (eg, World Health Organization) generally do not provide morphologic HL subtype stratification, limiting analysis and interpretation of results. Similarly, although outcomes in the United States may have been influenced by the type of center (ie, academic vs community), the optimal OS for HL with current first-line treatment options likely did not significantly affect outcomes. Additionally, patients in Latin America are primarily managed in a few large academic cancer centers, and less specialized centers often refer to these high-complexity hospitals, suggesting that the centralized cancer care in Latin America allows for capturing most patients diagnosed with HL. The propensity score matching limited the generalizability of survival outcomes between US race and ethnicity subgroups; thus, we mainly focused on comparing Latin American and US patients. We did not estimate relative survival because of the hospital-based nature of our data. Our data did not capture any additional SDOH, suggesting an opportunity to extend our data collection practices. Clinical, treatment, recurrence, or progression data were unavailable in SEER. The absence of certain variables in one or both data sources, including detailed treatment, clinical risk subgroups, comorbidities, extranodal presentation, household income, or educational status, limited further matching. Despite these limitations, our study likely provides a comprehensive attempt to compare HL subtypes, clinical features, and survival outcomes across regions while acknowledging the use of a hospital-based cohort.

This international study suggests a distinct distribution of HL subtypes among Latin American and Hispanic populations, characterized by a higher prevalence of MCHL, underscoring the need to investigate the role of environmental and host factors in disease etiology. The comparable survival between Latin American and US patients and the lack of effect of health insurance on outcomes in Latin American patients provides an opportunity to understand better the long-term risk of treatments and the differential impact of SDOH on survivorship outcomes in different world regions to develop effective interventions. Future prospective studies are warranted to validate our findings and address these unmet needs to improve the outcomes of patients with HL globally.

The authors thank the Amos Medical Faculty Development Program from the American Society of Hematology for providing publication fee support.

Contribution: B.V. contributed to conceptualization, methodology, visualization, and formal analysis and wrote the original draft; A.M.C.G., S.G.R., G.R.S., M.E.E.A., A.v.G., A.R.Q., F.W., M.O., J.F.V., D.C., B.B., M.A.T., V.I., C.O., and R.O.R.-J. contributed resources; and all authors interpreted the findings, reviewed and edited the manuscript, and agreed upon the final version of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Bryan Valcarcel, Department of Lymphoma and Myeloma, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Rockville, TX 20850; email: [email protected].