Background Indeterminate thyroid cytology classified as Bethesda III or IV represents a persistent clinical dilemma, with substantial uncertainty regarding malignancy risk and frequent reliance on diagnostic surgery. Optimizing post–fine-needle aspiration (FNA) management using readily available clinical and ultrasonographic information may reduce unnecessary surgical intervention while preserving oncologic safety. Methods We conducted a single-center retrospective cohort study including 206 adult patients with Bethesda III or IV thyroid nodules evaluated. Clinical characteristics, biochemical parameters, and standardized ultrasound features were extracted. Malignancy was defined by surgical histopathology or a prespecified composite reference standard. Univariable and multivariable logistic regression analyses were performed to identify independent predictors of malignancy. Model discrimination and calibration were assessed using the area under the receiver operating characteristic curve (AUC), Brier score, and calibration slope. The diagnostic yield of repeat FNA was evaluated, and an optimized post-FNA management pathway was developed and compared with observed clinical practice using paired analyses and decision curve analysis. Results Malignancy was confirmed in 87 of 206 nodules (42.2%). In multivariable analysis, lower free thyroxine levels (adjusted odds ratio OR 0.63, 95% confidence interval CI 0.48–0.84), taller-than-wide shape (OR 3.69, 95% CI 1.21–11.27), marked hypoechogenicity (OR 3.12, 95% CI 1.01–9.64), punctate echogenic foci (OR 2.01, 95% CI 1.01–4.89), and suspicious cervical lymph nodes (OR 4.31, 95% CI 1.10–16.90) were independently associated with malignancy. The multivariable model demonstrated good discrimination, with an apparent AUC of 0.875 and an optimism-corrected AUC of 0.798, along with acceptable calibration (Brier score 0.183). Repeat FNA was performed in 73 patients (35.4%), yielding actionable cytologic reclassification in 53.4%, including upgrading to Bethesda V/VI in 9.6%. Compared with observed practice, the optimized pathway reduced the proportion of nodules assigned to surgery (70.9% vs 57.8%) and unnecessary surgery among all nodules (38.3% vs 29.1%), while maintaining similar use of repeat FNA. When restricted to operated nodules, the corresponding rate of unnecessary surgery was 54.1%. Decision curve analysis showed superior net clinical benefit of the optimized pathway across clinically relevant risk thresholds. Conclusions Integrating conventional clinical and ultrasound features provides an effective framework for post-FNA risk stratification in Bethesda III and IV thyroid nodules. An optimized management pathway based on this approach may reduce unnecessary surgery while maintaining appropriate malignancy detection, supporting more individualized and proportionate clinical decision-making.
Wang et al. (Thu,) studied this question.