Exploring FNA Thresholds for Thyroid Nodules: A Five-Year Comparison of C-TIRADS and Kwak TI-RADS

Background : Accurate differentiation between malignant and benign thyroid nodules is essential. Although ultrasound serves as the primary diagnostic modality, there is a paucity of longitudinal data concerning nodule progression and limited validation of the Chinese Thyroid Imaging Reporting and Data System (C-TIRADS). This underscores the necessity for evidence-based follow-up strategies. The present study sought to assess the developmental trends of thyroid nodules over an extended follow-up period and to evaluate the diagnostic efficacy of the C-TIRADS. Methods: This was a single-center retrospective longitudinal cohort study. We included patients who underwent thyroid ultrasound between 2019 and 2024 with at least one follow-up examination. Exclusions were: (1) prior thyroid malignancy; (2) incomplete baseline data; (3) follow-up <12 months without pathology/cytology. In the primary analysis, the nodule was the statistical unit. Correlation among multiple nodules within the same patient was addressed by clustering at the patient level. A sensitivity analysis at the patient level retained only the largest or most suspicious nodule per patient. Two radiologists (≥5 years’ experience) independently graded nodules using C-TIRADS and Kwak Thyroid Imaging Reporting and Data System (Kwak TI-RADS) while blinded to the reference standard. Disagreements were adjudicated by a senior radiologist. Inter-reader agreement was reported using the weighted kappa (κ). The reference standard was surgical histopathology whenever available. For non-surgical cases, cytology Bethesda V/VI was considered malignant; nodules with ≥24 months of stable imaging (no upgrade or suspicious progression) were considered benign. Indeterminate cases were excluded from the primary endpoint analysis. We reported the verification proportion and performed two sensitivity analyses: (1) surgery-only subset; (2) including nodules deemed benign after ≥24-month stable follow-up. For predefined thresholds (e.g., C-TIRADS ≥4B, Kwak ≥4), 2×2 tables were used to derive sensitivity, specificity, PPV, and NPV with 95% CIs. Area Under the Curves (AUCs) were estimated and compared using the DeLong method. Because nodules were clustered within patients, GEE with an exchangeable correlation structure or cluster-robust standard errors were applied. Temporal changes in sonographic features were assessed with trend tests. To estimate clinical impact, we counted FNA recommendations at each threshold, the reduction in FNAs, and missed cancers, and constructed decision-curve analyses to quantify net benefit. Missing data were handled by complete-case analysis. Ultrasonographic data from patients diagnosed with thyroid nodules in 2019 and monitored consistently until 2024 were analyzed to evaluate longitudinal changes in nodule characteristics. A retrospective analysis of 272 surgically confirmed cases (173 malignant and 99 benign) was performed to compare the diagnostic efficacy of he C-TIRADS with the Kwak TI-RADS. Results: 1.Throughout the follow-up period, the nodules identified in the study participants were predominantly multiple and bilaterally distributed across the thyroid lobes, with their prevalence increasing annually. The majority of these nodules were round or oval in shape, with diameters less than 10 mm. However, there was a gradual increase in the proportion of nodules exceeding 10 mm in diameter, as well as those that were oval or irregularly shaped. Ultrasound characteristics such as microcalcifications, border invasion, a taller-than-wide shape (aspect ratio ≥1), and internal vascularity were largely absent in most nodules. Nonetheless, there was an upward trend in the proportion of nodules exhibiting these features. Nodules characterized by internal hypoechogenicity and lacking posterior acoustic attenuation constituted a significant proportion and did not demonstrate a notable trend of change over time. 2. The AUC for the C-TIRADS classification in the diagnosis of nodules was determined to be 0.784. A C-TIRADS category of ≥4B was identified as the optimal cutoff value for the diagnosis of malignant nodules, yielding sensitivity, specificity, positive predictive value, negative predictive value, and accuracy rates of 73.99%, 77.78%, 85.33%, 63.11%, and 75.37%, respectively. Furthermore, no statistically significant difference (P > 0.05) was observed in the diagnostic efficacy between the C-TIRADS classification and the Kwak TI-RADS classification for thyroid nodules( Table 2 ). Conclusion: Microcalcifications, irregular margins, a taller-than-wide morphology, and internal vascularity can be indicative of malignant transformation during extended monitoring. The C-TIRADS demonstrates diagnostic accuracy comparable to that of the Kwak TI-RADS, thereby reinforcing its clinical utility in the management of thyroid nodules.