Exp Clin Endocrinol Diabetes 2013; 121(09): 515-520
DOI: 10.1055/s-0033-1351289
Article
© J. A. Barth Verlag in Georg Thieme Verlag KG Stuttgart · New York

Structural Vascular Disease in Africans: Performance of Ethnic-specific Waist Circumference Cut Points using Logistic Regression and Neural Network Analyses: The SABPA Study

J. Botha
1  Physical Activity Sport and Recreation (PhASRec), North-West University, Potchefstroom, South Africa
,
J. H. de Ridder
1  Physical Activity Sport and Recreation (PhASRec), North-West University, Potchefstroom, South Africa
,
J. C. Potgieter
2  School for Psychosocial Behavioural Science, North-West University, Potchefstroom, South Africa
,
H. S. Steyn
3  Statistical Consultation Service, North-West University, Potchefstroom, South Africa
,
L. Malan
4  Hypertension in Africa Research Team (HART), North-West University, Potchefstroom, South Africa
› Author Affiliations
Further Information

Publication History

received 14 November 2012
first decision 28 June 2013

accepted 10 July 2013

Publication Date:
09 August 2013 (eFirst)

Abstract

A recently proposed model for waist circumference cut points (RPWC), driven by increased blood pressure, was demonstrated in an African population. We therefore aimed to validate the RPWC by comparing the RPWC and the Joint Statement Consensus (JSC) models via Logistic Regression (LR) and Neural Networks (NN) analyses. Urban African gender groups (N=171) were stratified according to the JSC and RPWC cut point models. Ultrasound carotid intima media thickness (CIMT), blood pressure (BP) and fasting bloods (glucose, high density lipoprotein (HDL) and triglycerides) were obtained in a well-controlled setting. The RPWC male model (LR ROC AUC: 0.71, NN ROC AUC: 0.71) was practically equal to the JSC model (LR ROC AUC: 0.71, NN ROC AUC: 0.69) to predict structural vascular ­disease. Similarly, the female RPWC model (LR ROC AUC: 0.84, NN ROC AUC: 0.82) and JSC model (LR ROC AUC: 0.82, NN ROC AUC: 0.81) equally predicted CIMT as surrogate marker for structural vascular disease. Odds ratios supported validity where prediction of CIMT revealed ­clinical ­significance, well over 1, for both the JSC and RPWC models in African males and females (OR 3.75–13.98). In conclusion, the proposed RPWC model was substantially validated utilizing linear and non-linear analyses. We therefore propose ethnic-specific WC cut points (African males, ≥90 cm; -females, ≥98 cm) to predict a surrogate marker for structural vascular disease.