Kinta D Schott Avaani Bhalla Emma Armstrong Ryan G N Seltzer Floris C Wardenaar ^*

• Arizona State University, Tempe, United States

Introduction: This cross-sectional study aimed to cross-validate an earlier developed algorithmbased screener and explore additional potential predictors for whether athletes will use thirdparty-tested (TPT) supplements.Methods: To justify the initial model behind the supplement safety screener (S3) algorithm which predicts whether athletes will use TPT supplements, a cross-validation was performed using this independent dataset based on responses of a large group of collegiate NCAA DI athletes. Additionally, explorative modeling using stepwise logistic regression was used to identify new predictors for TPT supplement use to create and evaluate a new model for future use.Results: The S3 algorithm was applied to the responses from n=662 athletes using supplements (age: 20±1.5 years, 50% female, from >24 sports) confirming that the algorithm identifies consistent and inconsistent TPT users (χ2 (1)=15.95, p < 0.001), with a moderate area under the curve (AUC, 0.67) and a moderate specificity (68%), but low sensitivity (51%). Explorative modeling identified two new variables: TPT logo recognition, and having at least one name, image, likeness (NIL) deal that may help to predict TPT supplement use.Discussion: Strong relationships between risk groupings and product use outcomes toward TPT supplement use were identified for the athletes screened. The S3 screener showed high sensitivity for identifying student-athletes inconsistently using TPT supplements, but a low specificity, lacking the ability to place less risky athletes into low-risk quadrants. The exploratory modeling, identifying TPT logo recognition and having a NIL deal, further strengthens our knowledge on predictors for consistent TPT supplement use.