BACKGROUND: The Sloane Project, an audit of UK screen-detected non-invasive carcinomas and atypical hyperplasias of the breast, has accrued over 5000 cases in 5 years; with paired radiological and pathological data for 2564 ductal carcinoma in situ (DCIS) cases at the point of this analysis. We have compared the radiological estimate of DCIS size with the pathological estimate of DCIS size. We have correlated these sizes with histological grade, specimen-handling methods, particularly the use of specimen slice radiographs, and the success or failure of breast-conserving surgery (BCS). METHODS: The Sloane Project database was interrogated to extract information on all patients diagnosed with DCIS with complete radiological and pathological data on the size of DCIS, nuclear grade, specimen handling (with particular reference to specimen radiographs) and whether primary BCS was successful or whether the patient required further conservation surgery or a mastectomy. RESULTS: Of 2564 patients in the study, 2013 (79%) had attempted BCS and 1430 (71%) had a successful single operation. Of the 583 BCS patients who required further surgery, 65% had successful conservation and 97% of them after a single further operation. In successful one-operation BCS patients, there was a close agreement between radiological and pathological DCIS size with radiology tending to marginally overestimate the disease extent. In multiple-operation BCS, radiology underestimated DCIS size in 59% of cases. The agreement between pathological and radiological size of DCIS was poor in mastectomies but was improved by specimen slice radiography, suggesting specimen-handling techniques as a cause. CONCLUSION: In 30% of patients undergoing BCS for DCIS, preoperative imaging underestimates the extent of disease resulting in a requirement for further surgery. This has implications for the further improvement of preoperative imaging and non-operative diagnosis of DCIS so that second operations are reduced to a minimum.