Breast self-examination (BSE) is widely recommended for breast cancer prevention. Following recent controversy over the efficacy of mammography, it may be seen as an alternative. We present a meta-analysis of the effect of regular BSE on breast cancer mortality. From a search of the medical literature, 20 observational studies and three clinical trials were identified that reported on breast cancer death rates or rates of advanced breast cancer (a marker of death) according to BSE practice. A lower risk of mortality or advanced breast cancer was only found in studies of women with breast cancer who reported practising BSE before diagnosis (mortality: pooled relative risk 0.64, 95% CI 0.56-0.73; advanced cancer, pooled relative risk 0.60, 95% CI 0.46-0.80). The results are probably due to bias and confounding. There was no difference in death rate in studies on women who detected their cancer during an examination (pooled relative risk 0.90, 95% CI 0.72-1.12). None of the trials of BSE training (in which most women reported practising it regularly) showed lower mortality in the BSE group (pooled relative risk 1.01, 95% CI 0.92-1.12). They did show that BSE is associated with considerably more women seeking medical advice and having biopsies. Regular BSE is not an effective method of reducing breast cancer mortality.