To determine the efficacy of periabortal antibiotics in preventing postabortal upper genital tract infection using data from published trials. We performed a literature search of all studies published from January 1966 to September 1, 1994, using MEDLINE, and we manually searched bibliographies of published articles. MEDLINE search terms included: abortion, infection, prophylaxis, antibiotics, pelvic inflammatory disease (PID), and suction curettage. Randomized, controlled trials comparing antibiotics with placebo in women undergoing suction curettage abortion before 16 weeks' gestation were identified. Data were extracted independently by two reviewers, one of whom was blinded to journal, year of publication, authors, and institution. Data from 12 studies were combined using meta-analytic techniques based on a fixed-effects model. The overall summary relative risk (RR) estimate for developing postabortal upper genital tract infection in women receiving antibiotic therapy compared with those receiving placebo was 0.58 (95% confidence interval [CI] 0.47-0.71). Of high-risk women, those with a history of PID had a summary RR estimate of 0.56 (95% CI 0.37-0.84); women with a positive chlamydia culture at abortion had a summary RR estimate of 0.38 (95% CI 0.15-0.92). Of low-risk women, those with no reported history of PID had a summary RR estimate of 0.65 (95% CI 0.47-0.90); in women with a negative chlamydia culture, the summary RR estimate was 0.63 (95% CI 0.42-0.97). The lowest summary RR estimate was among women drawn from populations with a low incidence (5-6%) of postabortal infection (summary RR estimate 0.22, 95% CI 0.11-0.42). The overall 42% decreased risk of infection in women given periabortal antibiotics is similar to the risk reduction demonstrable when only studies published before 1985 are combined (summary RR estimate 0.63, 95% CI 0.44-0.89). Our meta-analysis revealed a substantial protective effect of antibiotics in all subgroups of women undergoing therapeutic abortion, even women in low-risk groups. No more placebo-controlled trials should be performed, because women assigned to placebo are exposed to preventable risk. Routine use of periabortal antibiotics in the United States may prevent up to half of all cases of postabortal infections.