Diagnostic Performance of ACR TI-RADS and Sonographic Predictors of Malignancy in Pediatric Thyroid Nodules.

OBJECTIVES

Pediatric thyroid nodules are uncommon but carry a substantially higher risk of malignancy than adult nodules. Accurate risk stratification is crucial, as even modest false-negative rates can have significant clinical implications. Adult-based Thyroid Imaging Reporting and Data System (TI-RADS) is widely used for sonographic risk assessment. However, its performance in children remains unclear.

METHODS

A retrospective cohort study of 99 pediatric patients surgically treated for a thyroid nodule. Preoperative sonographic characteristics and TI-RADS category were assessed, and associations with malignancy were analyzed. Diagnostic performance metrics were calculated, and Srivatsa PED TI-RADS biopsy thresholds were externally validated.

RESULTS

Of 99 patients, 60 (60.6%) had malignant pathology. Malignancy increased across TI-RADS categories, reaching 100% in TR5; however, rule-out performance was limited, with a negative predictive value of 74.4% for TR1-TR3 and an overall missed malignancy rate of 18.3%. Overall sensitivity and specificity were 81.7% and 82.1%, with an area under the receiver operating characteristic curve of 0.85 (95% CI, 0.78-0.92). Only irregular margins (OR 13.3; P = .03) and microcalcifications (OR 5.9; P < .001) were independently associated with malignancy. Srivatsa PED TI-RADS increased biopsy eligibility (62.9% vs 53.6%) and improved sensitivity (77.6% vs 69.0%), with lower specificity (59.0% vs 69.2%).

CONCLUSIONS

In this tertiary-care, surgically managed pediatric cohort, American College of Radiology TI-RADS showed reasonable discrimination but limited rule-out performance in TR1-TR3. Srivatsa PED TI-RADS improved sensitivity at the cost of lower specificity. Findings may not generalize to unselected pediatric populations. Irregular margins and microcalcifications were the strongest predictors and may inform pediatric-specific risk stratification. Prospective validation in unselected cohorts is warranted.