Although observational studies have often claimed high success rates in women undergoing cervical cerclage for an incompetent cervix, randomized trials have not consistently favored this procedure. This multicenter, randomized, controlled trial sought to determine whether patient selection for cerclage could be improved by transvaginal sonographic measurements of cervical length. More than 47,000 women were offered scanning at 11 to 14 and 22 to 24 weeks gestation. Of 470 women whose cervix measured 15 mm or less in length, 253 entered the trial and were randomly assigned to either cervical cerclage using a Shirodkar suture or expectant management. The major outcome was the frequency of delivery before 33 completed weeks gestation. Cervical cerclage did not significantly lower the rates of early preterm delivery, neonatal mortality, or neonatal morbidity compared with expectant management. The 2 groups had similar proportions of spontaneous delivery or delivery after preterm prelabor rupture of membranes. Mean birth weight was 204 g more in the cerclage group, but this arose from a 1-week difference in gestational age at delivery. More than twice as many women in the cerclage group received antibiotics, most often because of prolonged prophylaxis and symptomatic vaginal discharge. More women in the cerclage group received tocolytic agents immediately after the procedure or because of threatened preterm labor. Approximately one third of women required emergency removal of the suture, most often because of preterm labor or premature labor or rupture of membranes. In this trial, placing a cervical suture in women with a short cervix did not lower the risk of early preterm delivery in a substantial proportion of cases. Routinely measuring cervical length at 22 to 24 weeks gestation by transvaginal sonography does identify women who are at high risk of early preterm delivery.