Body mass index (BMI) and waist circumference (WC) are often used to predict the risk of developing type II diabetes. With recent developments in three-dimensional optical (3DO) imaging, more advanced statistical methods for capturing body shape variance are possible. The current study aimed to compare an advanced 3DO analysis approach to conventional BMI and WC estimates for predicting common biomarkers used to assess the risk of developing diabetes.
As part of the Shape Up! Study, 154 healthy adults (89 women) underwent whole body 3DO surface scans and blood testing. Optical scans were spatially registered using a standardized 60,000 vertex template. Shape variation was described using principal component (PC) analysis for vertex locations. Regression analyses were completed with fasting glucose, insulin, HOMA-IR, and percent hemoglobin A1C as dependent variables. PCs that explained 95% of the sample’s shape variation were used as independent variables and compared to models using BMI and WC as independent variables. Analysis for men and women were completed separately.
The R^2’s for PC models predicting fasting glucose, insulin, HOMA-IR, and percent hemoglobin A1C were 0.49, 0.40, 0.41, and 0.46, respectively for men and 0.32, 0.33, 0.28, and 0.30 for women (all p<0.0001). The R^2’s for PC models were larger than the corresponding BMI and WC models for fasting glucose, insulin, HOMA-IR, and percent hemoglobin A1C (0.40, 0.26, 0.29, and 0.39, respectively for men; 0.25, 0.07, 0.11, and 0.25 for women; all p<0.05).
Equations created from PC models of three-dimensional whole body shape improve diabetes risk prediction compared to conventional BMI and WC estimates. Since 3DO information can be rapidly, inexpensively, and safely collected in adults and children, these observations suggest this approach may have value as a research and clinical tool.