MON-397 Assessment Of Thyroid Ultrasound As A Screening Tests For Thyroid Cancer In A Community Setting: A Logistic Regression Analysis

Abstract Disclosure: M. Amini: None. E. Shrestha: None. A. Ghimire: None. J. Newman: None. A.B. Silverberg: None. S.G. Albert: None. Introduction: The American College of Radiology Thyroid Imaging, Reporting and Data System (ACR-TI-RADS) stratifies thyroid nodules (TN) with thyroid ultrasound (TUS) to guide fine-needle aspiration (FNA) decisions in evaluating for thyroid cancer (TC). Developed in high-prevalence populations, the ACR guidelines designate abrupt ”cut-points” combining each TI-RADS imaging score with a discrete nodule size for recommending FNA. We evaluated its performance in a low-risk community and explored a logistic regression analysis (LGA) to generate a continuous risk prediction. Methods: Retrospective review of patients aged 18-75, who underwent TUS at a Midwest center (7/1/2023 - 12/31/2024). LGA was applied using the product of TI-RADS score times (x) nodule size in mm for predicting TC risk. A probability logit model was developed to estimate a positive predictive value (PPV) for cancer risk. Patients were stratified by demographics and indication for TUS. Results: 487 patients (371 women, 116 men, 220 Black, 267 White) underwent TUS. 52 were excluded due to loss of follow up, other malignancies or severe comorbidities. 201 patients met ACR criteria for FNA. 13 lesions were found at surgery (papillary thyroid cancer (n=7), follicular thyroid cancer (n=3), non-invasive follicular thyroid neoplasm with papillary like nuclear features (NIFTP, n=1), follicular adenoma (n=1) and oncocytic (Hurthle cell) adenoma (n=1)). The prevalence of TC was 2.9%, substantially lower than the 29% prevalence used to derive current ACR guidelines. Binary LGA of TI-RADS 3 with size suggested that a higher cut point of >60 mm (rather than 25 mm in the ACR criteria) would predict a PPV for TC of >5%. LGA showed a continuous, rather than threshold based, increase in cancer risk with increasing TI-RADS x mm. The predictive value for screening for TC was highest for thyroid nodules demonstrated on imaging studies, and lowest when TUS was obtained for thyroid dysfunction without anatomical abnormalities (p < 0.001). Binary LGA showed that Black patients had significantly lower cancer risk for all TI-RADS x mm levels compared to White patients (p=0.029). No significant differences were observed in TC risk when controlled for sex (p=0.89) or age (p=0.98). Conclusion: The current guidelines for TUS risk stratification for selecting patients for FNA were developed in populations with higher prevalence of thyroid cancer (with prevalence of 23% - 29%). Real-world experience has been documented at lower prevalence rates of 2.6 - 7%. Rather than absolute TC risk demarcations, a continuous risk model using TI-RADS x mm offers a more granular risk assessment that may improve shared decision-making, especially for borderline cases. Race specific variations in TC risk warrant further validation. Integration of local cancer prevalence into electronic health record systems may optimize personalized FNA recommendation. Presentation: Monday, July 14, 2025