Dynamic changes in serum IL-6, TNF-α, and β₂-microglobulin as early predictors of post-treatment relapse in lymphoma: a prospective cohort study

Background Early identification of relapse after first-line chemotherapy remains a major challenge in lymphoma management. Conventional imaging provides limited sensitivity for minimal residual disease and cannot detect early biochemical changes. This prospective cohort study evaluated whether dynamic changes in serum IL-6, TNF- α , and β₂-microglobulin (β₂-MG) can serve as early predictors of relapse. Methods A total of 260 patients with pathologically confirmed lymphoma who completed standard chemotherapy were enrolled and followed longitudinally. Serum IL-6, TNF- α , and β₂-MG were measured at T1 (end of chemotherapy), T2 (3 months post-treatment), and T3 (6 months post-treatment). Logistic regression, ROC analysis, Kaplan–Meier curves, and Cox proportional hazards models were used to assess predictive performance. Results During follow-up, 78 patients (30.0%) experienced relapse. All three biomarkers were significantly elevated in the relapse group at T1, and differences widened at T2 and T3. At T2, IL-6 (adjusted OR = 1.08), β₂-MG (adjusted OR = 1.38), and Ann Arbor stage and IPI score remained independent predictors. β₂-MG exhibited the strongest individual predictive value (AUC = 0.85), while the combined IL-6 + TNF- α + β₂-MG model achieved superior discrimination (AUC = 0.91). Higher biomarker levels were associated with shorter relapse-free survival in Cox models, including high IL-6 (HR = 1.78), TNF- α (HR = 1.52), and β₂-MG (HR = 2.06). Dynamic changes also showed strong predictive value, with ΔIL-6 (OR = 1.21) and Δβ₂-MG (OR = 1.34) indicating early biological progression before radiologic relapse becomes detectable. Conclusion Longitudinal monitoring of IL-6, TNF- α , and β₂-MG identifies early inflammatory trajectories linked to lymphoma relapse. IL-6 and β₂-MG at 3 months post-chemotherapy are independent and robust predictors, and a multi-marker model markedly enhances early-warning performance.