Age-dependent effects of surgical approach in T3b differentiated thyroid carcinoma: a population-based analysis using machine learning

Current guidelines recommend total thyroidectomy for all T3b differentiated thyroid carcinoma (DTC) with gross strap muscle invasion, yet evidence supporting this universal approach remains limited and conflicting. We analyzed 6,920 T3b DTC patients from the SEER database (2004-2022) who underwent lobectomy (n = 282) or total thyroidectomy (n = 6,638). Overall survival (OS) served as the primary outcome. Random survival forest (RSF) with SHapley Additive exPlanations (SHAP) analysis identified interactions between surgical approach and clinical variables. Cox regression with restricted cubic splines examined age-dependent surgical effects on OS. Neither RSF nor Cox regression (HR 0.96, 95% CI: 0.68-1.37, P = 0.84) showed a significant OS difference between surgical approaches. SHAP interaction analysis revealed significant age-dependent heterogeneity. Restricted cubic splines identified age 45 as the threshold where the surgical benefit converged. In patients <45 years, total thyroidectomy demonstrated significantly superior survival outcomes compared with lobectomy (adjusted HR 3.26, 95% CI: 1.30-8.22, P = 0.012; 10-year survival 95.5 vs 98.4%). Conversely, no difference existed in patients ≥45 years (HR 0.85, 95% CI: 0.58-1.24, P = 0.40). The multiplicative interaction was 0.26 (95% CI: 0.10-0.70, P = 0.008), confirming a negative age-surgery interaction. Although this study demonstrates that surgical approach shows no significant survival difference in T3b DTC, the negative interaction effect between surgical approach and age indicates that the benefit of total thyroidectomy progressively diminishes with advancing age, providing a survival advantage only in patients younger than 45 years. These findings challenge universal total thyroidectomy recommendations and support age-stratified surgical decision-making for T3b DTC.