Prospective evaluation of ultrasound and biochemical-based multivariable models for the prediction of late pre-eclampsia.

Objective: Prospective assignment at 11 + 0 to 13 + 6  weeks of risk for late pre-eclampsia (PE) using eight logistic regression-based statistical models.

Methods: Five hundred and fifty-four pregnancies. Uterine artery pulsatility index, parity, body mass index, mean arterial pressure, pregnancy-associated plasma protein-A, free β-human chorionic gonadotrophin and maternal age, were combined to obtain 'a posteriori risk of PE'.

Results: We observed 39 cases (7%) of late PE. There were 12 cases of severe PE and 27 of mild PE. According to the models used, the estimated detection rate ranged from 38.5% to 84.6% with a false-positive rate of 10%. The median risk ratio (estimated median risk of PE in affected pregnancies divided by estimated risk of PE in unaffected pregnancies) ranged between 1.66 and 7.61. The most reproducible biochemical-based model was a mixed model encompassing maternal history and pregnancy-associated plasma protein-A.

Conclusion: Some of the multivariable models drawn from the literature accurately predicted the late PE occurrence. The failure of some models may be because of the population in question not bearing several of the risk factors used to generate the models proposed. An effective combined screening at first trimester for late PE seems possible.