Purpose: In a recently completed study in women with gestational hypertension (GH) or mild preeclampsia (PE) at term we found that induction of labor prevented complications but did not increase the caesarean section rate (Koopmans et al., Lancet 2009). However, despite this finding certain women may still be at increased risk of a caesarean section (CS) when labor has to be induced. We tried to identify patients at high risk of caesarean section using a prediction model.
Method: We used data from our RCT in patients with a singleton pregnancy with a fetus in cephalic position between 36 and 41 weeks of gestation, complicated by GH or mild PE, comparing caesarean section with other modes of delivery. We considered predictive factors from medical history, clinical characteristics (including parameters of cervical ripening obtained by vaginal examination) and laboratory findings. Missing data were completed by multiple imputation. Using multivariable logistic regression analysis with a p-value < 0.175 for inclusion, we constructed a prediction model for caesarean section risk at two stages: before delivery and during delivery. The predictive capacity of our models was examined with receiver-operating-characteristic (ROC) analysis and calibration plots based on predicted risks averaged across the imputations.
Result: Of the 756 women included, 126 (17%) delivered by caesarean section. The multivariable model for caesarean section risk before delivery had an ROC of 0.833 and included ethnicity, parity, previous abortion, gestational age, type of antibiotic treatment, type of analgesia, and degree of cervical dilatation as predictors. Of the laboratory findings uric acid and the degree of proteinuria were included. The model for CS risk during delivery additionally included a parameter indicating clinical deterioration of GH (based on blood pressure, proteinuria or HELLP syndrome). The ROC increased to 0.844. Hosmer-Lemeshow tests and calibration plots indicated that calibration of both models was acceptable. Overfitting will be estimated by bootstrapping.
Conclusion: In women with GH or mild PE at term, the risk of caesarean section strongly depended on characteristics including ethnicity, parity and deterioration of disease. The identified predictors could help to identify women at high risk of caesarean section and could be used by clinicians in their decision making.