A clinically applicable nomogram for predicting recurrent spontaneous abortion using coagulation and autoimmune markers

Abstract

Objective: To develop and validate a risk prediction model for recurrent spontaneous abortion (RSA) by integrating thrombus markers with anticardiolipin antibody (ACA) and antinuclear antibody (ANA) indicators. Methodology: Clinical data of 159 RSA patients (June 2021–June 2022) and 178 healthy controls from Rui'an People's Hospital were retrospectively analysed. Clinical and laboratory parameters, including coagulation markers and autoantibodies, were analysed. Predictor selection was performed using a consensus approach combining univariate logistic regression, LASSO regression, and Random Forest feature importance. Multivariate logistic regression identified independent predictors. Machine learning models were developed and evaluated using a 70:30 train-test split with 5-fold cross-validation. Model performance was assessed using AUC, sensitivity, specificity, and calibration metrics. Results: Six independent predictors were identified: Protein S, Protein C, antithrombin-III (AT-III), fibrinogen (FIB), ACA-IgG, and ANA status. Reduced levels of natural anticoagulants (Protein S, Protein C, AT-III) were associated with increased RSA risk, while elevated FIB and ACA-IgG were significant risk factors. The Random Forest model demonstrated the best performance with an AUC of 0.765 (95% CI: 0.665–0.851), sensitivity of 60.4%, and specificity of 75.9%. The model showed good calibration (Brier score 0.199). A nomogram based on logistic regression achieved an AUC of 0.700. Conclusion: A combined biomarker model integrating coagulation and immune parameters provides moderate-to-good predictive performance for RSA. This clinically applicable tool may facilitate early risk stratification and individualized management, although external validation is required before routine implementation.