Use of Predictive Modeling to Tailor Molecular Testing Utilization for Thyroid Nodules

Area of Science:

• Endocrinology
• Oncology
• Radiology

Background:

• Cytologically indeterminate thyroid nodules require further risk assessment.
• Current risk stratification systems may not fully utilize ultrasound features for molecular testing decisions.

Purpose of the Study:

• To evaluate the utility of molecular testing in indeterminate thyroid nodules when guided by an algorithm incorporating ultrasound-based risk of malignancy (ROM) estimation.
• To assess the impact of a novel algorithm on reducing unnecessary molecular testing.

Main Methods:

• Retrospective chart review of thyroid nodules undergoing molecular testing.
• Utilized a Malignancy Risk Estimation System with ultrasound findings and various risk stratification guidelines (Korean, French, AACE, ATA).
• Developed a novel algorithm integrating Korean TI-RADS with ATA and NCCN guidelines.

Main Results:

• Korean TI-RADS demonstrated the highest performance (AUC = 0.83) in the study population.
• Positive molecular test results correlated with higher ROM across all assessed models (P < .05).
• The developed algorithm could have prevented 38% of unnecessary tests on benign/low-ROM nodules.

Conclusions:

• An algorithm integrating pre- and posttest probabilities can effectively guide molecular testing for indeterminate thyroid nodules.
• This strategy may significantly reduce the number of benign or low-risk nodules subjected to molecular testing.