A predictive model for spontaneous preterm birth and delivery within 7 days in women with threatened preterm labor: a...

To develop and internally validate predictive models for spontaneous preterm birth (sPTB) in women with threatened preterm labor (TPL).

This prospective observational study included 264 women with TPL between 24 + 0 and 36 + 0 weeks of gestation. Candidate predictors available at clinical assessment included demographic, obstetric, sonographic, inflammatory, and cervicovaginal biomarker variables. Least absolute shrinkage and selection operator (LASSO) regression with tenfold cross-validation and the one-standard-error criterion was used for variable selection. Model performance was evaluated using receiver operating characteristic analysis, calibration, bootstrap validation, and decision curve analysis.

Eighty-four women (31.8%) experienced sPTB before 37 weeks, and 54 (20.5%) experienced sPTB within 7 days after sampling. Five predictors were retained in both final models: cervical length, quantitative PAMG-1 concentration, C-reactive protein, uterine contraction frequency, and mild vaginal bleeding or spotting. For sPTB before 37 weeks, the apparent AUC was 0.845 (95% CI 0.788-0.903), and the optimism-corrected AUC was 0.830. For 7-day sPTB, the apparent AUC was 0.939 (95% CI 0.900-0.978), and the optimism-corrected AUC was 0.929. The integrated models showed higher discrimination than cervical length alone, PAMG-1 alone, or cervical length plus PAMG-1.

Integrated models combining cervical length, quantitative PAMG-1, C-reactive protein, uterine contraction frequency, and mild vaginal bleeding or spotting showed promising performance for predicting sPTB in women with TPL. Given the modest sample size and internal validation only, these findings should be considered preliminary and require external validation before clinical implementation.