Establishment and validation of a nomogram for predicting preterm birth in intrahepatic cholestasis during pregnancy: a retrospective study

Abstract

Objective: This study aimed to develop and evaluate a nomogram for predicting preterm birth in patients with intrahepatic cholestasis of pregnancy (ICP), with a view to assisting clinical management and intervention.

Methods: This retrospective observational study included 257 pregnant women with ICP from Sichuan Provincial People's Hospital between January 1, 2022 and July 30, 2024. The routine clinical and laboratory information of these patients were also collected. We used the least absolute shrinkage and selection operator (LASSO) and multivariable logistic regression analysis to investigate the association between clinical and laboratory data and preterm birth in ICP patients. A nomogram was developed to predict the likelihood of preterm birth in ICP patients. The prediction accuracy of the model was evaluated by consistency index (C-index), receiver operating characteristic (ROC) curve, area under the curve (AUC), and calibration curve. Decision curve analysis (DCA) was used to evaluate its applicability in clinical practice.

Results: Among the 257 ICP patients, 56 (21.79%) were diagnosed with preterm birth. Cases were randomly divided into a training set (154 cases) and a test set (103 cases). A nomogram was developed to predict preterm birth in ICP patients based on height, twin pregnancy (TP), gestational age at diagnosis (GA at diagnosis), and total bile acid level (TBA) at diagnosis. The calibration curve of the training set was close to the diagonal (C-index = 0.864), and the calibration curve of the test set was also close to the diagonal (C-index = 0.835). These results indicate that the model has a good consistency. The AUC of the training group and the test group were 0.864 and 0.836, respectively, indicating the good accuracy of the model. The DCA reveals that this nomogram could be applied to clinical practice.

Conclusion: The combination of TBA level, TP, height and GA at diagnosis is an effective model for identifying preterm birth in ICP patients. These results will help guide the clinical management and treatment of patients with ICP, thereby reducing maternal and infant safety issues caused by preterm birth.

Keywords: Intrahepatic cholestasis of pregnancy; Nomogram; Predictive model; Preterm birth; Validation.

Fig. 1

Q-Q Plots of Continuous Variables After Normality Transformation. This figure presents the Q-Q plots of continuous variables after a normality assessment using the KS test. Each subplot displays the Q-Q plot for a specific variable, where the x-axis represents the theoretical quantiles and the y-axis shows the sample quantiles. The red line indicates the theoretical reference line for a normal distribution, and the black dots illustrate the observed data distribution. The plots suggest that the distributions of the transformed variables approximate normality, meeting the prerequisites for further analysis

Fig. 2

LASSO Regression Analysis. (A) The optimal penalty parameter λ (0.007) retained eight variables with non-zero coefficients: TP (1.98), FGR (0.83), GA at diagnosis (-1.50), AST_log (1.05), DBil_log (0.07), Height_log (-1.62), TBA_bc (2.61), and GH (0.78). (B) The cross-validation curve indicates that the optimal λ corresponds to the minimum binomial deviance, striking a balance between model complexity and predictive performance

Fig. 3

Flowchart of this study. A total of 257 patients with ICP were included in this study based on the inclusion criteria. We used LASSO logistic regression and multivariate analysis to identify significant predictive factors and establish a nomogram. The training dataset (n = 154) was utilized to estimate the predictive model for preterm birth in ICP patients, while the test dataset comprised 103 patients. We evaluated the nomogram using the AUC, C-index, calibration curves, and DCA

Fig. 4

Nomogram for estimating preterm birth in ICP patients. At the top, points represent the contribution score of each variable to the predicted outcome, ranging from 0 to 100. TP is a binary variable with values of 0 or 1, each value corresponding to a specific score, indicating its impact on the prediction. GA at diagnosis is divided into three ranges: ≥13 weeks and ≤ 33 weeks is set as 0, >33 weeks and ≤ 36 weeks as 1, and >36 weeks as 2, with each range contributing a specific score to the prediction. TBA_bc and Height_log are continuous variables, where Height_log = ln(Height Value) and TBA_bc = TBA^λ-1/λ (with the optimal parameter λ = -1.0972). By summing the scores of all variables, the Total Points can be calculated, ranging from 0 to 350. These total points are then used to determine the Linear Predictor value, which ranges from − 6 to 5, and subsequently, the Diagnosis Rate, which represents the probability of the event occurring, ranging from 0.01 (1%) to 0.99 (99%)

Fig. 5

ROC curves for the training and test sets. (A) AUC for the training set. (B) AUC for the test set. The model demonstrates strong discriminative ability and robust generalization, effectively predicting the risk of preterm birth in ICP patients

Fig. 6

Calibration curves for the training and test sets. (A) Calibration curve for the training set. (B) Calibration curve for the test set. These curves demonstrate good agreement between the predicted and actual outcomes, with low prediction errors, indicating the model’s accuracy and stability in predicting the risk of preterm birth in patients with ICP

Fig. 7

Analysis of DCA for the training and test sets. The blue line represents the net benefit for ICP patients not deemed to be at risk of preterm birth in the training set; the red line represents the net benefit for ICP patients not deemed to be at risk of preterm birth in the test set. The gray diagonal line represents the net benefit of treating all patients as if they were at risk of preterm birth. The further the blue and red lines are from the gray line, the greater the benefit that the model provides in accurately predicting preterm birth