Evolving prognostic impact of B symptoms in non-Hodgkin lymphoma: A competing-risks analysis of SEER data, 2010–2019

Abstract

Abstract Background: Constitutional B symptoms, including fever, night sweats, and unintentional weight loss, have long been recognized as adverse prognostic indicators in non-Hodgkin lymphoma (NHL). Historically integrated into the Ann Arbor staging system, their presence has been associated with higher tumor burden, systemic inflammation, and inferior survival. However, the therapeutic landscape of NHL has shifted over the past decade, with routine use of anti-CD20 monoclonal antibodies (e.g., rituximab), intensified chemotherapy regimens, and improved supportive care. These advances may have reduced the prognostic burden of B symptoms. We investigated temporal trends in the survival impact of B symptoms using a large, population-based cohort using robust statistical methods. Methods: We analyzed 119,707 adults with first-primary NHL diagnosed between 2010–2019 in the SEER 18 registries, with follow-up through 2022. B symptoms were recorded as present or absent. Patients were categorized into an early era (2010–2014) and a late era (2015–2019). Multivariable Cox proportional hazards models adjusted for age, sex, race/ethnicity, SEER stage, chemotherapy, and radiation were used to estimate hazard ratios (HRs) for overall mortality. An interaction term (B symptoms × era) assessed temporal changes. Secondary analyses included Fine–Gray competing-risks regression for lymphoma-specific death, restricted mean survival time (RMST) differences over 60 months, and a 6-month landmark Cox model to reduce early-death bias. Results: B symptoms were reported in 25.3% of patients in the early era and 27.8% in the late era (p < 0.001). In adjusted Cox models, B symptoms were associated with significantly worse survival in both eras: HR 1.52 (95% CI 1.48–1.57) in the early era and HR 1.72 (95% CI 1.67–1.77) in the late era. However, the interaction term showed a relative decline in hazard over time (HRinteraction = 1.16; 95% CI 1.11–1.19; p = 2.4×10⁻¹³), suggesting attenuation of excess risk. In Fine–Gray competing-risk models, subdistribution HRs for lymphoma-specific mortality decreased from 1.63 (95% CI 1.57–1.69) to an interaction sHR of 1.10 (95% CI 1.05–1.16; p = 6.9×10⁻⁵). RMST analyses showed modest narrowing: mean survival was 46.7 vs 39.2 months in asymptomatic vs. symptomatic patients in the early era (-7.6 months), and 47.4 vs. 40.0 months (-7.4 months) in the late era. Among 97,027 patients alive at 6 months, B symptoms remained associated with increased mortality (HR 1.15; 95% CI 1.09–1.21; p = 9.9×10⁻⁸), confirming persistent, though diminished, risk. Conclusions: In this large, nationally representative NHL cohort, the adverse impact of B symptoms on survival has persisted but reduced over the past decade. This attenuation was consistent across multiple statistical approaches, including time-varying Cox models, competing-risks regression, RMST comparisons, and sensitivity analyses. The reduced prognostic burden may reflect improvements in systemic therappy, supportive care, and overall disease management. Despite this progress, B symptoms remain a statistically and clinically significant marker of worse survival. These findings support updating prognostic tools to evolving outcomes and suggest that symptom-driven decisions, such as surveillance intensity, may warrant reevaluation. Further studies should explore biologic correlates and investigate whether novel biomarkers can enhance traditional symptom-based risk stratification in modern NHL care.