Objective: The purpose of this study was to develop a dynamic nomogram model to predict the risk of spontaneous preterm birth at <32 weeks in twin pregnancy. Design: A retrospective analysis and multicentre validation study Setting and Population: Women with twin pregnancies followed up in two tertiary medical centres from January 2017 to March 2019. Methods: Data on maternal demographic characteristics, transvaginal cervical length and funneling were extracted. The prediction model was constructed with independent variables determined by logistic regression analyses. The risk score was calculated according to the dynamic nomogram model. Main outcome measures: The risk of spontaneous preterm birth at <32 weeks in twin pregnancy. Results: In total, 1065 twin pregnancies were eligible for the study, of which the data of 764 cases (92 twin preterm cases (<32 weeks) and 672 control cases) were obtained from a tertiary medical centre as the training group and those of 301 cases (36 twin preterm cases (<32 weeks) and 265 control cases) from the other tertiary medical centre as the external validation group. Based on logistic regression analyses, we built a dynamic nomogram model with satisfactory discrimination in both the training group(C-index: 0.856, 95% CI: 0.813-0.899) and external validation group(C-index: 0.808, 95% CI: 0.751-0.865). The restricted cubic splines and ROC curve supported the performance of the prediction model. Conclusions: We developed and validated a dynamic nomogram model to predict the individual probability of preterm birth in twin pregnancy at <32 weeks.