Validation of a first-trimester screening model for pre-eclampsia in an unselected population

Objective To validate the performance of a previously constructed first-trimester predictive model for pre-eclampsia (PE) in routine care of an unselected population. Methods A validation cohort of 4621 consecutive women attending their routine first-trimester ultrasound examination was used to test a prediction model for PE that had been developed previously in 5170 women. The prediction model included maternal factors, uterine artery Doppler, blood pressure and pregnancy-associated plasma protein-A. Model performance was evaluated using receiver–operating characteristics (ROC) curve analysis and ROC curves from both cohorts were compared unpaired. Results Among the 4203 women included in the final analysis, 169 (4.0%) developed PE, including 141 (3.4%) cases of late-onset PE and 28 (0.7%) cases of early-onset PE. For early-onset PE, the model showed an area under the ROC curve of 0.94 (95% CI, 0.88–0.99), which did not differ significantly (P = 0.37) from that obtained in the construction cohort (0.88 (95% CI, 0.78–0.99)). For late-onset PE, the final model showed an area under the ROC curve of 0.72 (95% CI, 0.66–0.77), which did not differ significantly (P = 0.49) from that obtained in the construction cohort (0.75 (95% CI, 0.67–0.82)). Conclusion The prediction model for PE achieved a similar performance to that obtained in the construction cohort when tested on a subsequent cohort of women, confirming its validity as a predictive model for PE. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.