1520 Background: It is critical for oncologists to be aware of unbiased and interpretable cancer risks (i.e., penetrance) in carriers with germline pathogenic variants in cancer susceptibility genes. However, relevant literature is large and varies significantly in study design, patient ascertainment, and types of risk estimates reported. This heterogeneity can cause inconsistent conclusions between studies and create barriers for clinicians to understand and apply them in practice. To further understand the current literature, we assessed penetrance studies associated with non-BRCA breast cancer susceptibility genes based on study design and ascertainment adjustment. Methods: We used a validated natural language processing-based abstract classifier to identify all penetrance studies regarding eleven genes: ATM, BARD1, CDH1, CHEK2, NBN, NF1, PALB2, PTEN, RECQL, STK11, and TP53. Relevant studies were then manually annotated as “with ascertainment adjustment” if a study was based on: (1) a general population; (2) a pedigree analysis or a family-based study with appropriate ascertainment adjustment; or (3) a hospital-based study or a panel testing analysis with well-matched cases and controls. Results: A total of 49 penetrance studies were identified, with a median of nine studies for each gene (range: 4-16). The case-control study was the dominant study type, accounting for over 80% in five genes, 50% in two genes, and 18% to 43% in the other four genes. The proportion of studies with ascertainment adjustment was generally low (mean: 33%) and varied widely between different genes (7% to 80%). Contradictory breast cancer risks (no increased risk vs. significantly increased risk) were found in eight genes (73%) (Table). The most common ascertainment bias identified was a case-control study with cases (patients) who had a strong family history but using general population controls. Conclusions: Ascertainment bias is common in penetrance studies, but few studies adjust for it appropriately. Clinicians should be aware of this issue, and new methods are warranted to select unbiased risk estimates, synthesize them, and provide the accurate general-population penetrance. [Table: see text]