We evaluated the use of multivariate analysis in the prediction of shoulder dystocia (SD). One hundred consecutive cases with SD were matched with 100 controls without dystocia. All patients had term, vaginal delivery. Multivariate analysis was used to identify independent variables significantly related to shoulder dystocia. The regression coefficients for the identified factors were used to calculate a composite score from which receiver operating characteristics (ROC) curves were derived. Birthweight (BW), 1-hour Glucola (GLU), operative vaginal delivery (OVD), and height of fundus (HOF) were related independently to SD. The sensitivity and specificity reached 84 and 80%, respectively, with BW + GLU + OVD. Significant associations persisted when HOF and carbohydrate intolerance were substituted for BW and GLU, respectively. SD is independently associated with BW, GLU, and OVD, and may be predicted with clinically acceptable accuracy using multiple variables. This model may be useful in the design of prospective studies for managing suspected macrosomia.