Evaluation of prediction models for early-onset preeclampsia in the first trimester: a systematic review
10.3760/cma.j.issn.1007-9408.2018.04.023
- VernacularTitle: 早孕期不同预测模型预测早发型子痫前期效果的系统评价
- Author:
Xiaoyu DUAN
1
;
Biyun XU
2
;
Ziqing NAN
3
;
Yali HU
1
Author Information
1. Medical School of Southeast University, Nanjing 210009, China
2. Scientific Research Department, Nanjing Drum Tower Hospital, Nanjing 210008, China
3. Department of Obstetrics and Gynecology, Nanjing Drum Tower Hospital, Nanjing 210008, China
- Publication Type:Journal Article
- Keywords:
Pregnancy trimester, first;
Preeclampsia;
Risk factors;
Forecasting;
Logistic models
- From:
Chinese Journal of Perinatal Medicine
2018;21(4):266-275
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the characteristics and performance of various prediction models for early-onset preeclampsia, and to provide a reference for further study of preeclampsia prediction methods.
Methods:(1) Databases of PubMed, Medline, Embase, China National Knowledge Infrastructure (CNKI) and Wanfang Database were searched since their inception to October 2016. Studies in models for predicting early-onset preeclampsia during the first trimester were included. (2) Two reviewers independently checked potentially eligible articles, assessed risk of bias and extracted data. (3) Subtotals for the performance of different models were created and their properties were analyzed. Differences between simple models (based upon high risk factors such as demographic figures, medical history and family history, etc) and complex models (based upon blood pressure, uterine artery Doppler and biomarkers) were compared by analyzing forest plot created by SAS 9.4.
Results:(1) Seventeen studies met the inclusion criteria were screened out, including nine prospective cohort studies, two case-control studies and six nested case-control studies. A total of 76 436 gravidas from tendifferent populations were assessed by the established models in these studies. (2) The area under the curve (AUC) of 13 simple models ranged from 0.64 to 0.81 with the sensitivity of 21%-60% when the false positive rate (FPR) was 10%. The AUC of 17 complex models ranged from 0.77 to 0.98 and the sensitivity was between 48.0% and 95.2% at a fixed FPR of 10%. (3) Compared with the simple models, the best complex models could ensure a promotion of 0.171 (range from 0.060 to 0.245) in median AUC, and a promotion of 40.8% (16.0% to 52.2%) in sensitivity at a FPR of 10%. Based on the simple models, additional mean arterial pressure (MAP) would increase the AUC and sensitivity by 0.092 (0.079 to 0.104) and 28.7% (16.2% to 55.0%), respectively, while additional uterine artery pulsatility index (UtA-PI) would bring an increase of 0.106 (0 to 0.137) and 31.8% (-1.0% to 41.9%), respectively. Moreover, when both MAP and UtA-PI were included into the simple models, the AUC and sensitivity would increase by 0.157 (0.094 to 0.218) and 31.6% (12.0% to 52.2%).
Conclusions:Complex prediction models perform better than simple models in prediction of early-onset preeclampsia. However, further confirmation is required in different population.
