Background BI-RADS Category 4 patients have a 2–95% risk for malignancy and are generally recommended for breast biopsy with little discrimination for risk level or distinction between risk of invasive or in situ disease. Our study sought to determine if higher thresholds for biopsy based on stratifying for risk and distinguishing between risk of invasive cancer and DCIS could reduce biopsy rates and increase cancer-to-biopsy yields without missing cancers urgent for resolution. Methods 108 BI-RADS 4 cases with final outcomes data were evaluated from a prospective cohort of 215 consecutive patients seen at a same-day multidisciplinary breast clinic for women with mammograms categorized as BI-RADS 0, 4, or 5 in 2006–07. Final outcomes were determined from pathologic diagnosis or two-year follow-up. Risk estimates (RE) for DCIS and invasive cancer were collected prospectively and re-assessed by a radiologist blinded to outcomes and prior reading assessments. Cases were stratified according to the risk ranges of the BI-RADS 4 subcategories and risk of invasive or in situ disease. Biopsy rates, cancer-to-biopsy yields, and number of malignancies missed were calculated for various thresholds for intervention. Results A ROC curve for invasive cancer risk for the radiologist demonstrated a 98.5% level of accuracy (95% confidence interval [CI]: 96.9%, 100%). 60 cases had some risk for invasive cancer and 48 had some risk for DCIS. There were 14 invasive cancer and 11 DCIS outcomes, 3 of which were high-grade. Pathologic assessment from biopsy or surgery was available for 100 patients. The outcomes of 8 cases were determined by benign two-year follow-up. There are several strategies for intervention that improve biopsy yield and reduce biopsies for benign disease as shown in Table 1. If cases with RE between 2–10% for DCIS or invasive cancer were not biopsied, 23% of biopsies would be avoided and the yield would increase to 30%. If cases with invasive cancer RE between 10–95% and DCIS RE between 50–95% were biopsied, 52% of biopsies would be avoided and the yield would increase to 39%. One invasive ductal carcinoma (3 mm, Grade 2) would be missed, although with six-month follow-up, this would not be a problem. Limitations Small sample size; one radiologist providing RE may not be representative of general mammographic assessment. Conclusion Setting higher biopsy thresholds for BI-RADS 4 lesions can safely reduce biopsy rates and increase biopsy yields. Given evidence suggesting that low/intermediate grade DCIS may be overdiagnosed, distinguishing between DCIS and invasive cancer risk at screening by offering active surveillance as an alternative to biopsy for BI-RADS 4 lesions suspicious for non-high-grade DCIS may be a promising approach for reducing biopsies. This will be prospectively tested in a reader study using several radiology readers in a series of 750 cases in the Athena Breast Health Network. New biopsy thresholds can be set if the results of our study can be validated. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-09-02.