Gender Specific Predictive Performance and Optimal Threshold of Anthropometric Indices for the Prediction of Hypertension among a Ghanaian Population in Kumasi

Abstract

Despite the extensive use of anthropometric indices in the risk prediction of hypertension, there is lack of consensus on the type and the optimal threshold to be used. This current work evaluates the optimal threshold points, discriminative power and comparative performance as discriminators of hypertension for ten (10) anthropometric indices. Standard methods were used to measure weight, height, waist circumference, hip circumference for calculating BMI, PI, WHR, WhtR, CI, AVI, and BAI. Gender specific predictive performance of anthropometric indices were assessed using the area under the curve (AUC) of the receiver operator characteristic curves (ROC). In this Ghanaian cohort, body mass index, waist-to-hip ratio and ponderal index were poor predictors of hypertension. Among the commonly used anthropometric measures, waist circumference at a lower threshold (>75 cm and 80 cm for females and males respectively) than the currently recommended cutoffs was the index of choice for the prediction of hypertension, however significant improvement in prediction was achieved with the use of conicity index (>1.08 female, >1.05 male).