OBJECTIVE: To review validation studies of the Edinburgh Postnatal Depression Scale (EPDS). METHOD: A systematic search was performed in Medline and the Science Citation Index Expanded (ISI) from the period 1987-2000. For sensitivity and specificity of the EPDS presented in each study, 95% confidence intervals were estimated. Positive and negative predictive values were estimated assuming prevalences of postpartum depression ranging from 5% to 20%. RESULTS: Eighteen validation studies were identified. The study design varied between studies. The sensitivity and specificity estimates also varied: 65-100% and 49-100%, respectively. The confidence intervals were estimated to be wide. Our estimates suggest a lower positive predictive value in a normal population than in the validation study samples. CONCLUSION: Most studies show a high sensitivity of the EPDS. Because of the differences in study design and large confidence intervals, uncertainty remains regarding the comparability between the sensitivity and specificity estimates of the different EPDS versions.