Background: The mammographic grading system used by BreastScreen Australia accredited programs requires needle biopsy of lesions with indeterminate features (grade 3). We wished to determine the predictors of malignancy for these common lesions, so as to reduce surgery for benign lesions. Design: Grade 3 masses assessed between Jan 1996-Dec 2005 are included. New or changing lesions were identified through film review. Imaging, demographic and final outcome data were tabulated. Statistical analysis was performed to determine the predictors of malignancy. Results: 1181 lesions, including 623 new or changing masses are assessed. Of these 98 (8.3%) were malignant. Malignancy was found in 3% first round masses versus 13% of new or changing lesions. Withholding needle biopsy results, interval change (OR 2.85, p = 0.0001), increasing age (p = 0.0001) and diameter were independent predictors of malignancy, lesion diameter having an inverse relationship with malignancy.Once needle biopsy results were included, this parameter became the sole predictor of malignancy as other factors lost significance in a multivariate model. Conclusion: Surgical biopsy is not indicated in well-sampled grade 3 masses with benign needle biopsy results. Surgery is valuable for diagnosis or treatment of the remaining such masses without benign needle biopsy results.