Screening for pre-eclampsia early in pregnancy: performance of a multivariable model combining clinical characteristics and biochemical markers

Objective To investigate the performance of a multivariable model combining a priori clinical characteristics and biomarkers to detect, early in pregnancy, women at higher risk of developing pre-eclampsia (PE). Design Nested case-control study. Setting University medical centre, Quebec, Canada (CHU de Quebec). Population A total of 7929 pregnant women recruited between 10 and 18 weeks of gestation. In all, 350 developed hypertensive disorders of pregnancy (HDP)- of which 139 had PE, comprising 68 with severe PE and 47 with preterm PE-and were matched with two women with a normal pregnancy. Methods We selected a priori clinical characteristics and promising markers to create multivariable logistic regression models: body mass index (BMI), mean arterial pressure (MAP), placental growth factor, soluble Fms-like tyrosine kinase-1, pregnancy-associated plasma protein A and inhibin A. Main outcome measures PE, severe PE, preterm PE, HDP. Results At false-positive rates of 5 and 10%, the estimated detection rates were between 15% (5-29%) and 32% (25-39%), and between 39% (19-59%) and 50% (34-66%), respectively. Considering the low prevalence of PE in this population, the positive predictive values were 7% (5-9%) to 10% (7-13%) for PE and 2% (1-4%) to 4% (3-6%) in the preterm and severe PE subgroups. The multivariable model yielded areas under the receiver operating characteristics curves (AUC) between 0.72 (0.61-0.81) and 0.78 (0.68-0.88). When only BMI and MAP were included in the model, the AUC were similar to those of the a priori model. Conclusions In a population with a low prevalence of preterm PE, a multivariable risk algorithm using an a priori combination of clinical characteristics and biochemical markers did not reach a performance justifying clinical implementation as screening test early in pregnancy.