OBJECTIVES: Counseling pregnant women on the hazards of drugs requires valid data. No study has definitively established the baseline risk for spontaneous (SA) or therapeutic (TA) pregnancy termination in pregnant women taking antidepressants. Our objectives were to estimate the baseline risk for SAs and TAs and determine whether antidepressants increase those risks. METHODS: We used a random effects meta-analytic model. MEDLINE, EMBASE, Healthstar, and Cochrane databases were searched by 2 independent reviewers for cohort studies published from 1966 to the present (2002) reporting rates of spontaneous and therapeutic abortions in women taking antidepressants in therapeutic doses to treat any depression. Up to 10% of patients could have comorbidities, but could not have exposure to other known abortifacients. Data were extracted by two reviewers; all differences were resolved through consensus. Rates, risk differences, and risk ratios, along with 95% confidence intervals (CI95%) were summarized across studies. Sub-analyses were done by class (TCAs, SSRIs, and dual action agents). RESULTS: We identified 15 potential studies; 6 prospective cohort studies with 11 treatment arms provided extractable data for SA and 6 for TA. All of those studies matched exposed and comparison groups on important confounders such as age, smoking, and alcohol consumption. The baseline risks (CI 95%) were 8.7% (7.5%-9.9%, n=2033) and 6.0% (3.3%-8.7%, n=2033) for SA and TA, respectively. Antidepressants had baseline risks of 12.4% (10.8%-14.1%) and 8.5% (6.0%-11.1%) and were associated with increases in risk of 3.9% (1.9%-6.0%) and 2.9% (0.5%-5.3%), respectively. RRs was 1.52 (1.22-1.89, n=3567) for SA and 1.47 (1.14-1.89, n=3552) for TA. No differences were found among antidepressant classes. CONCLUSIONS: Maternal exposure to antidepressants is associated with a small but significant increased risk for both spontaneous and therapeutic abortions. This risk must be considered when counseling patients. Further research is required to rule out depression or other indications as confounders.