The authors determined the association between maternal haemoglobin concentration measured at <28 weeks' gestation and late fetal death at >or=28 weeks' gestation (stillbirth). Data were derived from the National Maternal and Infant Health Survey--a nationally representative survey of US deliveries in 1988. Analysis was restricted to women with a singleton live birth (n = 4,199) or a stillbirth (n = 1,375) for whom maternal prenatal care, haemoglobin, smoking status and gestational age data were available. Haemoglobin concentrations during first and second trimesters, respectively, were classified as mild (10.0 to <11.0 and 9.5 to <10.5 g dL(-1)) or moderate (9.0 to <10.0 and 8.5 to <9.5 g dL(-1)) anaemia, or high haemoglobin (>or=14.6 g dL(-1) in either trimester). Hazard ratios (HR) and 95% confidence intervals (CI) for stillbirth were derived from discrete proportional hazards regression models after adjusting for confounders. Stillbirth was not associated with mild anaemia or high haemoglobin in either the first or second trimester of pregnancy. Moderate anaemia measured before 28 weeks' gestation was significantly associated with an increased risk of stillbirth among non-black women (adjusted HR: 4.4; 95% CI: 1.02, 19.01). Moderate anaemia was not associated with stillbirths among black women. Further investigation regarding causal mechanisms for this association is warranted.