A performance review of novel adiposity indices for assessing insulin resistance in a pediatric Latino population

INTRODUCTION: Body mass index (BMI) percentile or BMI adjusted for age and sex is the most common anthropometric index to monitor and assess obesity in children. However, the ability of BMI to accurately predict insulin resistance (IR) in youth is debated. Determining the best method to noninvasively measure IR in the pediatric population is especially important due to the growing prevalence of type 2 diabetes mellitus (T2DM), which is more likely to develop in people with IR. Therefore, this study analyzed the performance of BMI against newer anthropometric indices in assessing IR in a pediatric Latino identifying sample. METHODS: We studied 127 pediatric Latino participants from the Arizona Insulin Resistance (AIR) registry and performed linear regression analyses between various measures of IR and adiposity indices, including body mass index (BMI), triponderal mass index (TMI), body adiposity index (BAI), pediatric body adiposity index (pBAI), a body shape index (ABSI), abdominal volume index (AVI), waist to height ratio (WtHR) and waist to hip ratio (WHR). Log transformations of each index adjusted for age and sex and IR were used for the linear regressions. Additionally, we generated receiver operating characteristics (ROC) from logistic regressions between HOMA-IR and HOMA2IR against the same indices. RESULTS: Using the homeostatic assessment of insulin resistance (HOMA-IR), HOMA2IR, the quantitative insulin-sensitivity check index (QUICKI), fasting serum insulin, and FPG/FSI to measure IR, we showed that BMI adjusted for age and sex performs similarly to many of the newer indices in our sample. The correlation coefficients for pBAI [R2: 0.27, 95% confidence interval: 0.88–1.81, p < 0.001] and BMI [R2: 0.27, 95% confidence interval: 0.92–1.92, p < 0.001] were the highest for HOMA-IR. Similarly, pBAI [R2: 0.29, 95% confidence interval: 0.88–1.72, p < 0.001] and BMI [R2: 0.29, 95% confidence interval: 0.93–1.83, p < 0.001] were the highest for HOMA2IR. A similar trend was observed with QUICKI, FSI, and FPG/FSI. ABSI had the lowest R(2) value for all measures of IR. Area under the curve (AUC) values for the receiver operating characteristics (ROC) for HOMA-IR and HOMA2IR support these conclusions. CONCLUSIONS: BMI adjusted for age and sex, despite its usage and simplicity, still stacks up well against newer indices in our Latino sample. Testing these indices across larger samples is necessary to generalize these findings and translate performance to adults.