Prospective evaluation of the risk of pre-eclampsia using logistic regression analysis

To calculate the risk of developing pre-eclampsia (PET) in a consecutive series of low-risk women at 18-24 weeks' gestation, using recently published logistic regression models.

This was a prospective study, with complete follow-up, in a consecutive series of unselected low-risk singleton pregnancies. Uterine artery pulsatility index as well as a combination of maternal factors were recorded at 18-24 weeks' gestation. The distribution of the estimated risks for the 16 PET patients was compared with that obtained for 136 women who had a normal pregnancy, as assessed by routine testing. A receiver-operating characteristics (ROC) curve was plotted to evaluate the detection rate at fixed false-positive rates (FPRs) of 5%, 10% and 20% and the corresponding odds cut-offs.

Just 1/16 (6.2%) women with PET developed the disease before the 34(th) week of gestation. Using the 'All PET' logistic regression model, for 16 PET cases the overall median odds was 1 : 1454, higher compared with that of 1 : 41635 estimated for controls. Using the 'PET >or= 34 weeks' model, the median odds of the 15 women who developed PET late was 1 : 3405, compared with 1 : 40785 for controls. In the case of PET before 34 weeks, the risk was 1 : 426373 vs. 1 : 4159823126 estimated for controls ('PET < 34 weeks' model). Detection rates for the All PET model were 18%, 50% and 62% at a FPR of 5%, 10% and 20%, respectively. For the PET >or= 34 weeks model these detection rates were 6%, 46% and 60%, respectively.

Even though the individual odds estimation is too low to represent the real risk of PET, the recently published logistic regression models detected more than 60% of PET at a FPR of 20% for both All PET and PET >or= 34 weeks models. Using these models in clinical practice does not seem to give any significant improvement over Doppler alone in the prediction of PET, but the use of a PET-specific odds instead of an actual Doppler value alone seems to be useful for clinical management.