J Balani1, S L Hyer1, H Shehata2, F Mohareb3. 1. Department of Endocrinology, Epsom and St Helier University Hospitals NHS Trust, Surrey, UK. 2. Department of Maternal Medicine, Epsom and St Helier University Hospitals NHS Trust, Surrey, UK. 3. Department of Bioinformatics, Cranfield University, Cranfield, UK.
Abstract
OBJECTIVE: To develop a model to predict gestational diabetes mellitus incorporating classical and a novel risk factor, visceral fat mass. METHODS: Three hundred two obese non-diabetic pregnant women underwent body composition analysis at booking by bioimpedance analysis. Of this cohort, 72 (24%) developed gestational diabetes mellitus. Principal component analysis was initially performed to identify possible clustering of the gestational diabetes mellitus and non-GDM groups. A machine learning algorithm was then applied to develop a GDM predictive model utilising random forest and decision tree modelling. RESULTS: The predictive model was trained on 227 samples and validated using an independent testing subset of 75 samples where the model achieved a validation prediction accuracy of 77.53%. According to the decision tree developed, visceral fat mass emerged as the most important variable in determining the risk of gestational diabetes mellitus. CONCLUSIONS: We present a model incorporating visceral fat mass, which is a novel risk factor in predicting gestational diabetes mellitus in obese pregnant women.
OBJECTIVE: To develop a model to predict gestational diabetes mellitus incorporating classical and a novel risk factor, visceral fat mass. METHODS: Three hundred two obese non-diabetic pregnant women underwent body composition analysis at booking by bioimpedance analysis. Of this cohort, 72 (24%) developed gestational diabetes mellitus. Principal component analysis was initially performed to identify possible clustering of the gestational diabetes mellitus and non-GDM groups. A machine learning algorithm was then applied to develop a GDM predictive model utilising random forest and decision tree modelling. RESULTS: The predictive model was trained on 227 samples and validated using an independent testing subset of 75 samples where the model achieved a validation prediction accuracy of 77.53%. According to the decision tree developed, visceral fat mass emerged as the most important variable in determining the risk of gestational diabetes mellitus. CONCLUSIONS: We present a model incorporating visceral fat mass, which is a novel risk factor in predicting gestational diabetes mellitus in obese pregnant women.
Entities:
Keywords:
Gestational diabetes; machine learning; obesity; predictive model; principal component analysis; visceral fat mass
Authors: Boyd E Metzger; Thomas A Buchanan; Donald R Coustan; Alberto de Leiva; David B Dunger; David R Hadden; Moshe Hod; John L Kitzmiller; Siri L Kjos; Jeremy N Oats; David J Pettitt; David A Sacks; Christos Zoupas Journal: Diabetes Care Date: 2007-07 Impact factor: 19.112
Authors: Boyd E Metzger; Steven G Gabbe; Bengt Persson; Thomas A Buchanan; Patrick M Catalano; Peter Damm; Alan R Dyer; Moshe Hod; John L Kitzmiller; Lynn P Lowe; H David McIntyre; Jeremy J N Oats; Yasue Omori Journal: J Matern Fetal Neonatal Med Date: 2012-09-07
Authors: Susan Y Chu; William M Callaghan; Shin Y Kim; Christopher H Schmid; Joseph Lau; Lucinda J England; Patricia M Dietz Journal: Diabetes Care Date: 2007-04-06 Impact factor: 19.112
Authors: Carolyn A Chiswick; Rebecca M Reynolds; Fiona C Denison; Sonia A Whyte; Amanda J Drake; David E Newby; Brian R Walker; Shareen Forbes; Gordon D Murray; Siobhan Quenby; Susan Wray; Jane E Norman Journal: BMJ Open Date: 2015-01-14 Impact factor: 2.692
Authors: Annette L Briley; Suzanne Barr; Shirlene Badger; Ruth Bell; Helen Croker; Keith M Godfrey; Bridget Holmes; Tarja I Kinnunen; Scott M Nelson; Eugene Oteng-Ntim; Nashita Patel; Stephen C Robson; Jane Sandall; Thomas Sanders; Naveed Sattar; Paul T Seed; Jane Wardle; Lucilla Poston Journal: BMC Pregnancy Childbirth Date: 2014-02-18 Impact factor: 3.007
Authors: F van Hoorn; Mph Koster; C A Naaktgeboren; F Groenendaal; A Kwee; M Lamain-de Ruiter; A Franx; M N Bekker Journal: BJOG Date: 2020-09-01 Impact factor: 6.531