Vacuum-assisted biopsy is an image-guided technique introduced in 1995 that is thought to be superior to 14G automated-needle biopsy for the evaluation of non-palpable breast lesions. However, prospective randomised studies evaluating its accuracy are unavailable. We conducted a critical review of the currently available literature on the accuracy of vacuum-assisted biopsy and compared it with published data on the accuracy of 14G automated-needle biopsy. The diagnostic performance of vacuum-assisted biopsy was evaluated by reviewing all available English-language literature published in Medline between 1995 and November 2001. Four independent reviewers used standard forms to extract the data. Twenty-two published studies were included. High-risk and DCIS underestimate rates, as well as the miss-rate of cancer, were assessed. High-risk and DCIS underestimate rates for 11G vacuum biopsy were 16% (95% Confidence Interval (CI) 12-20%) and 11% (95% CI 9-12%), respectively, and both were lower than the rates reported for 14G automated-needle biopsy (40% (95% CI 26%;56%) and 15% (95% CI 8%;26%), respectively). Due to incomplete follow-up of the benign lesions, it was impossible to calculate the miss-rates and the sensitivity rate. The results of this review indicate that vacuum-assisted biopsy can decrease the high-risk and DCIS underestimate rates, but it is unclear whether it can also decrease the miss-rates of cancer.