Objective To determine the prevalence of pica during pregnancy, maternal hemoglobin levels at delivery, and the association of pica with 2 adverse pregnancy outcomes: low birth weight and preterm birth. Design A retrospective cohort study was conducted using eligible subjects (n = 281) from a cross-sectional survey (n = 366). Subjects/setting Mothers aged 16 to 30 years with infants younger than 1 year of age who participated in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) were interviewed at 4 WIC clinics in Houston and Prairie View, Tex. Medical records from 31 hospitals were abstracted. Main outcome measures Maternal hemoglobin levels at delivery, mean birth weight, and mean gestational age were compared for women who reported pica and women who did not. Statistical analyses [[Chi].sup.2] Tests, orthogonal t tests, analysis of variance, and regression analysis were used to test the relationships among study variables. Results Pica prevalence categorized by substance was as follows: ice, 53.7%; ice and freezer frost, 14.6%; other substances such as baking soda, baking powder, cornstarch, laundry starch, baby powder, clay, or dirt, 8.2%; and no pica, 23.5%. Women in all 3 pica groups had lower hemoglobin levels (mean[+ or -]+standard deviation, 114[+ or -]12 g/L, 110[+ or -]13 g/L, and 113[+ or -]12 g/L, respectively) at delivery than women who did not report pica (118[+ or -]11 g/L) (P[less than].01 for all pica groups vs no pica). There were no differences in mean birth weight or mean gestational age of infants born to women from the 3 pica groups and the No Pica group. Applications The findings suggest that pica practices are associated with significantly lower maternal hemoglobin levels at delivery but are not associated with pregnancy outcomes. Dietitians should ask pregnant women with anemia about pica and should counsel women who report pica regarding the health risks associated with it.