A predictive model of pregnancy loss using pre-pregnancy endocrine and immunological parameters in women with abnormal glucose/lipid metabolism and previous pregnancy loss

Abstract

Objective: To investigate the clinical and endocrine risk factors for pregnancy loss in women with abnormal glucose/lipid metabolism and a history of pregnancy loss, and to develop a predictive model to assess the risk of pregnancy loss in these women's subsequent pregnancies.

Methods: Patients with a history of pregnancy loss who had abnormal glucose/lipid metabolism were retrospectively included in this study, and their pre-pregnancy baseline and clinical characteristics were collected. A predictive nomogram was constructed based on the results of the multivariable logistic regression model analysis, and its calibration and discriminatory capabilities were evaluated. The internal validation was then performed and the net benefits were assessed by the clinical decision curve.

Results: The predictive model was eventually incorporated eight variables, including maternal age, previous pregnancy losses, anticardiolipin antibody (aCL) IgG, aCL IgM, thyroid peroxidase antibody, complement 4, free thyroxine and total cholesterol. The area under the curve (AUC) of the nomogram was 0.709, and Chi-square value and P value of the Hosmer-Lemeshow test were 12.786 and 0.119, respectively, indicating that the nomogram had a satisfactory calibration and discriminatory performance. The validation cohort showed a similar result for the discrimination of the nomogram (AUC = 0.715). The clinical decision curve demonstrated the nomogram had good positive net benefits.

Conclusions: This is the first study to predict the risks of subsequent pregnancy loss in women with abnormal glucose/lipid metabolism and history of pregnancy loss using pre-pregnancy clinical and endocrine parameters. This predictive nomogram may provide clinicians assistance to personalize the management of subsequent pregnancies in these patients.

Keywords: Abnormal glucose/lipid metabolism; Nomogram; Pre-pregnancy endocrine parameters; Predictive model; Previous pregnancy loss.

Fig. 1

Flow chart for patient selection in the training cohort

Fig. 2

The predictive nomogram for pregnancy loss risk in women with a history of pregnancy loss. The eight factors are incorporated into the final predictive model, including maternal age, previous pregnancy losses (PPL), anticardiolipin antibody (aCL) IgG and aCL IgM, thyroid peroxidase antibody (TPO-Ab), total cholesterol (TC), free thyroxine (FT-4), and complement 4 (C4). Total scores are calculated by aligning the dots on each numbered row. For instance (expressed in red), the total score of this No.5 patient was 0.159, indicating that her pregnancy loss probability was 17.4%

Fig. 3

ROC curve and calibration curves of the nomogram. A ROC curve evaluating the discrimination of this prediction model for pregnancy loss probability in the training cohort, with 0.709 of the AUC value. B ROC curve evaluating the discrimination of this prediction model in the validation cohort, with 0.715 of the AUC value

Fig. 4

Clinical decision curve and clinical impact curves of the nomogram. A Calibration curve assessing the association between the predicted value and actual rates of pregnancy loss risk among these patients. B Clinical decision curve evaluating the net benefits of the predictive model for pregnancy loss risk in the patients with a history of pregnancy loss