Performance of LFSPRO prediction in TP53 mutation status for prospectively collected probands.

Abstract

Genetic counseling and testing for germline mutations are essential for identifying individuals at increased risk for cancer. Pathogenic/likely pathogenic (P/LP) variants in TP53 are diagnostic of Li-Fraumeni syndrome (LFS), a highly penetrant disorder with diverse, early-onset tumors. Current clinical guidelines, such as Chompret and classic criteria, provide frameworks for identifying individuals at risk for P/LP TP53 variants; however, genetic counselors often encounter people with features concerning for LFS that do not clearly meet established criteria, creating challenges for risk assessment and testing decisions. We evaluated whether LFSPRO, a Mendelian, family-history-based model that estimates the individual's probability of harboring a deleterious TP53 variant, improves identification of individuals with LFS relative to guideline criteria. In a prospectively collected cohort of 178 probands who underwent clinical genetic counseling and germline TP53 testing, LFSPRO showed superior discrimination compared with Chompret criteria, with higher sensitivity (81% vs. 33%) and specificity (88% vs. 65%) and improved positive predictive values (PPVs: 0.53 vs. 0.14; negative predictive values [NPVs]: 0.96 vs. 0.85). Receiver operating characteristic analysis confirmed strong discriminatory performance (area under the curve [AUC] = 0.88). Calibration analysis using observed-to-expected ratios indicated good agreement between predicted and observed P/LP variant frequencies (observed/expected = 1.07). These findings demonstrate that LFSPRO outperforms traditional guideline-based criteria for identifying individuals with TP53 mutations in real-world clinical settings. By providing quantitative, well-calibrated TP53 P/LP variant probabilities rather than binary classifications, LFSPRO can enhance genetic counseling and support testing decisions, particularly for individuals who do not clearly meet existing criteria.