OBJECTIVE To determine the effectiveness of intrapartum penicillin prophylaxis in preventing early-onset group B streptococcal (GBS) infection in neonates of women whose birth canals are colonized by group B streptococci. DATA SOURCES Articles published between 1966 and 1992 identified from MEDLINE, EMBASE, the Science Citation Index and the Oxford Perinatal Database; the bibliographies of primary studies, textbooks and review articles and published abstracts from major conferences and symposia. DATA SELECTION Studies were selected if four criteria were met: (a) the target population was intrapartum women and neonates, (b) the intervention was penicillin prophylaxis, (c) invasive early-onset GBS infection was an outcome measure, and (d) the studies were controlled trials or cohort studies. Seven primary studies were identified, four of which were randomized controlled trials. DATA EXTRACTION Explicit methodologic criteria were used by two of the authors to assess independently the study quality; one of the reviewers was blind as to author, institution and journal. The baseline characteristics of the population, intervention and outcome were summarized twice and checked for accuracy by two of the authors. DATA SYNTHESIS Five of the studies showed a trend toward a beneficial effect of penicillin prophylaxis, and two showed a statistically significant effect. The pooled odds ratio indicated a 30-fold reduction (95% confidence interval 0.0013 to 0.17) in the incidence of early-onset GBS infection with intrapartum penicillin prophylaxis. Subgroup analyses did not change these results. The magnitude of improvement observed did not differ between women with prenatal risk factors (premature rupture of the membranes and premature labour) and those without these risk factors. CONCLUSIONS There is accumulative evidence that intrapartum penicillin prophylaxis is effective in preventing early-onset GBS infection. Such therapy is beneficial to women whose birth canals are colonized with group B streptococci. Further studies are needed to determine the optimum timing and method of detecting vaginal colonization during pregnancy.