SEER-22 Study Finds Mycosis Fungoides Survival Gaps

Key Takeaways

• In a U.S. SEER-22 cohort of 13,694 patients with mycosis fungoides, 5-year overall survival was 85.8%, with lower estimates in Black and American Indian/Alaska Native patients and the highest estimate in Asian/Pacific Islander patients.
• After adjustment for age, sex, diagnosis year, socioeconomic context, and stage, Black patients had worse survival than White patients, while Asian/Pacific Islander patients had better survival.
• Lower county-level median household income independently tracked with worse survival, and a registry-based LASSO-Cox model retained external discrimination in SEER-8; however, the authors report that standard Cox regression performed similarly and that machine-learning methods offered limited incremental benefit from the available registry variables.

Mycosis fungoides (MF), the most common cutaneous T-cell lymphoma (CTCL), often has an indolent course but shows wide variation in outcomes beyond disease extent alone. In a population-based study of racial and socioeconomic inequities, investigators analyzed Surveillance, Epidemiology, and End Results (SEER) Research Limited-Field Data, 22 Registries for patients with MF identified by International Classification of Diseases for Oncology, Third Edition (ICD-O-3) code 9700/3 and diagnosed from 2000 through 2021. After excluding death-certificate-only and autopsy-only cases and second-or-later reportable primaries, the analytic cohort included 13,694 patients; overall survival was the endpoint, assessed with Kaplan-Meier methods and sequential Cox models, with a prespecified comparison of 2000-2010 versus 2011-2021 and county-level median household income used as an area-level measure of socioeconomic status (SES). A separate prognostic-modeling program used a non-overlapping SEER-22 development cohort with external validation in SEER-8.

Overall survival across the cohort was 96.8% at 1 year, 90.8% at 3 years, and 85.8% at 5 years. At 5 years, survival was 85.3% for White patients, 82.7% for Black patients, 92.9% for Asian/Pacific Islander (API) patients, and 82.1% for American Indian/Alaska Native (AI/AN) patients, with significant separation across racial groups. In the fully adjusted model, Black patients had higher mortality than White patients, with a hazard ratio of 1.79 (95% CI 1.63-1.97), while API patients retained a survival advantage.

Five-year overall survival improved from 84.4% in 2000-2010 to 87.1% in 2011-2021, but the race-by-era interaction was not significant. Lower county-level income also remained independently associated with worse survival; for example, 11.2% of Black patients versus 4.8% of White patients were in the lowest income category. In a separate 10,650-patient modeling cohort, LASSO-Cox retained external discrimination in 4,753 SEER-8 patients with a Harrell C-index of 0.748, while the boosted model transported less well across registries.

Because this was an observational registry analysis, the findings do not establish causal effects of race, SES, treatment, diagnosis era, or policy changes on survival. SEER did not include treatment-specific details, comorbidity burden, insurance continuity, or healthcare-access measures, and county-level income reflects contextual conditions rather than an individual patient's own income. Combined Summary Stage is also broader than clinical tumor-node-metastasis-blood staging, and small subgroup sizes with possible race or stage misclassification may have limited precision, especially for AI/AN patients and some API estimates.

Clinician Questions

How was socioeconomic status defined in this mycosis fungoides survival analysis?

Socioeconomic status was represented by county-level median household income linked to residence, derived from the 2017-2021 American Community Survey and grouped into 6 SEER-defined categories, with the highest-income category used as the reference. This was an area-level contextual measure rather than an individual patient's own income or personal socioeconomic circumstances.

Which data elements were used to build the mycosis fungoides survival prediction models?

The mycosis fungoides survival models used routinely collected registry variables: age, sex, race or ethnicity group, SEER Summary Stage, county-level income group, rural-urban residence, and diagnosis period. Those models were developed in a non-overlapping SEER-22 cohort and then tested externally in SEER-8.