The objective was to determine the association of moderate to heavy caffeine consumption during pregnancy on spontaneous abortion and birth weight in humans. Data sources used included a computerized literature search of MEDLINE (1966-July 1996); EMBASE (1988-November 1996); Psychlit I (1974-1986); Psychlit II (1987-1996); CINAHL (1982-May 1996) and manual search of bibliographies of pertinent articles. Inclusion criteria were: English language research articles; pregnant human females; case control or cohort design; documented quantity of caffeine consumption during pregnancy; control group with minimal or no caffeine consumption (0 to 150 mg caffeine/d); documented data regarding spontaneous abortion and/or fetal growth. The exclusion criteria were: case reports; editorials; review papers. The methods section of each study was examined independently by two blinded investigators with a third investigator adjudicating disagreements. Two independent investigators extracted data onto a standardized form. A third investigator adjudicated discrepancies. We compared a caffeine-exposed group (>150 mg/d) and controls (0 to 150 mg/d), using Mantel-Haenszel pooling. Of the 32 studies meeting inclusion criteria, 12 had extractable data (6 for spontaneous abortion, 7 for low birth weight, 1 common study). Mantel-Haenszel odds ratio (CI95%) was 1.36 (1.29-1.45) for spontaneous abortion in 42,988 pregnancies. The overall risk ratio was 1.51 (1.39-1.63) for low birthweight (<2500 g) in 64,268 pregnancies. Control for confounders such as maternal age, smoking, and ethanol use was not possible. We concluded that there is a small but statistically significant increase in the risks for spontaneous abortion and low birthweight babies in pregnant women consuming >150 mg caffeine per d. A possible contribution to these results of maternal age, smoking, ethanol use, or other confounders could not be excluded.