PURPOSE:To evaluate whether cerclage in twins reduces the rate of spontaneous preterm birth <32 weeks when compared to expectant management. METHODS:This is a retrospective cohort study of twin pregnancies with the following indications for cerclage from two institutions: history of prior preterm birth, ultrasound-identified short cervix ≤2.5 cm, and cervical dilation ≥1.0 cm. The "cerclage" cohort received a cerclage from a single provider at a single institution from 2003-2016. The "no cerclage" group included all patients with similar indications that were expectantly managed from 2010-2015, at a second institution where cerclages are routinely not performed in twin pregnancies. The primary outcome was the rate of spontaneous preterm birth at <32 weeks. Secondary outcomes were the rates of spontaneous and overall (including medically indicated) preterm births at <32 weeks, < 34 weeks, and <36 weeks, chorioamnionitis, birth weight, and neonatal mortality within 30 days of life. We also performed a planned subgroup analysis stratified by cerclage indication. RESULTS:In all, 135 women were included in two cohorts: cerclage (n = 96) or no cerclage (n = 39). The rates of spontaneous preterm birth <32 weeks were 10.4% (n = 10) with cerclage versus 28.2% (n = 11) without cerclage (OR 0.23, CI 0.08-0.70, p = .017). After adjusting for cerclage indication, clinical history, age, chorionicity, insurance type, race, BMI, in-vitro fertilization, and multifetal reduction, there remained a significant reduction in the cerclage group of spontaneous preterm birth <32 weeks (adjusted odds ratio (aOR) 0.24, CI 0.06-0.90, p = .035), spontaneous preterm birth <36 weeks (aOR 0.34, CI 0.04-0.81, p = .013) as well as in overall preterm birth <32 weeks (aOR 0.31, CI 0.1-0.86, p = .018), and overall preterm birth <36 weeks (aOR 0.37, CI 0.10-0.84, p = .030). When stratified by short cervix or cervical dilation in the cerclage versus no cerclage groups, there was a significant decrease in spontaneous preterm birth <32 weeks in the cerclage group with cervical dilation (11.1 versus 41.2%, p = .01) but not in the cerclage group with short cervix only, even for cervical length <1.5 cm. Pregnancy latency was 91 days in the cerclage group versus 57 days in the no cerclage group (p = .001), with a median gestational age at delivery of 35 versus 32 weeks (p = .002). There was no increase in chorioamnionitis in the cerclage group. Furthermore, there was a significant increase in birth weight (median 2278 versus 1665 g, p < .001) and decrease in perinatal death <30 days (1.6 versus 12.9%, p = .001). CONCLUSIONS:Cerclage in twin pregnancies significantly decreased the rate of spontaneous preterm birth <32 weeks compared to expectant management. However, when stratified by cerclage indication, this decrease in primary outcome only remained significant in the group with cervical dilation.