Objective The study was undertaken to test the hypothesis that shortened midtrimester cervical length is more predictive of early (<26 weeks) than later (26-34 weeks) spontaneous preterm birth. Study design This is a secondary analysis of a blinded, multicenter observational study of 183 women with a prior preterm birth. Vaginal sonography was begun at 16 to18 weeks' gestation and scheduled every 2 weeks (maximum 4 scans per patient). Cervical length and any observed dynamic shortening were recorded at each visit to determine the shortest observed cervical length from 16 to 24 weeks' gestation. The shortest cervical length measurements were categorized as less than 25 mm, 25 to 29 mm and 30 mm or greater. The initial cervical length was also compared with the shortest cervical length to categorize patients on the basis of the timing of cervical shortening 30 mm or less. Contingency table, linear regression, and survival analysis were used to analyze the relationship between cervical length groups and spontaneous preterm birth. Results In both the less than 25 mm and 25 to 29 mm groups, the incidence of spontaneous midtrimester birth (<26 weeks) was higher than the incidence of later (26-34 weeks) preterm birth (<25 mm group: 37% vs 19%; 25-29 mm group: 16% vs 3%, respectively) as compared with women with a shortest cervical length 30 mm or greater, who had rates of 1% and 9% respectively (P < .0001). Similarly, women who had an initial cervical length 30 mm or less and those who shortened their cervix to 30 mm or less before 22 weeks were also more likely to experience a midtrimester than later preterm birth, whereas women who shortened their cervix 30 mm or less later (22-24 weeks) or who maintained a cervical length greater than 30 mm had lower rates of midtrimester than later preterm birth (P < .0001). Conclusion Shortened cervical length in the midtrimester preferentially predicts early, as opposed to later, spontaneous preterm birth in high-risk women. © 2004 Elsevier Inc. All rights reserved.