In a prospective cohort study, 3135 pregnant women were followed to evaluate the association of caffeine consumption during pregnancy with late first- and second-trimester spontaneous abortion. Almost 80% of pregnant women used some caffeine; among users the average daily intake was 99.3 mg from all sources. Sources of caffeine were nonherbal tea (used by 49.4% of women), coffee (41.2%), colas (35.0%), and drugs (6.6%). In all, 28% of pregnant women consumed greater than or equal to 151 mg of caffeine daily, and these "moderate-to-heavy" caffeine users were significantly more likely to experience late first- or second-trimester spontaneous abortion when compared with nonusers and light users (0 to 150 mg). Demographic characteristics, reproductive and medical history, contraceptive use, and smoking and drinking habits were taken into consideration. The adjusted relative risk of miscarriage after moderate-to-heavy caffeine consumption was 1.73 (p = 0.03). Light caffeine use (1 to 150 mg daily) was associated with increased risk for spontaneous abortion only among women who aborted in their last pregnancy (RR = 4.18, p = 0.04). Replicative studies are necessary before the association of caffeine with spontaneous abortion can be confirmed.