Objective: To assess the association between maternal caffeine intake and risk of pregnancy loss using a systematic review and meta-analysis. Design: Categorical and dose-response meta-analysis of prospective studies. Setting: Relevant articles were identified by searching MEDLINE and SCOPUS databases through 30 January 2015. Two authors independently extracted information from eligible studies. Random-effects models were used to derive the summary relative risks (RR) and corresponding 95 % CI for specific categories of caffeine consumption and for a continuous association using generalized leastsquares trend estimation. Subjects: A total of 130 456 participants and 3429 cases in fourteen included studies. Results: Compared with the reference category with no or very low caffeine intake, the RR (95 % CI) of pregnancy loss was 1·02 (0·85, 1·24; I 2 = 28·3 %) for low intake (50-149 mg/d), 1·16 (0·94, 1·41; I 2 = 49·6 %) for moderate intake (150-349 mg/d), 1·40 (1·16, 1·68; I 2 = 18·6 %) for high intake (350-699 mg/d) and 1·72 (1·40, 2·13; I 2 = 0·0 %) for very high intake (≥700 mg/d). In the dose-response analysis, each 100 mg/d increment in maternal caffeine intake (~1 cup of coffee) was associated with 7 % (95 % CI 3 %, 12 %) higher risk of pregnancy loss. Our results may have been affected by publication bias, but the association remained significant for the subset of larger studies. Furthermore, adjustment for smoking and pregnancy symptoms may have been incomplete, potentially resulting in residual confounding. Conclusions: Albeit inconclusive, higher maternal caffeine intake was associated with a higher risk of pregnancy loss and adherence to guidelines to avoid high caffeine intake during pregnancy appears prudent.