Predicting body mass index using arm span: an alternative approach for anthropometric studies

Abstract

Body Mass Index (BMI) is a widely used anthropometric indicator for assessing body composition and nutritional status. However, its reliance on stature can pose challenges, particularly in populations where accurate height measurement is difficult, such as the elderly and individuals with musculoskeletal impairments. Arm span has been proposed as a reliable alternative to stature for BMI calculation. This study aims to evaluate the effectiveness of arm span as a surrogate for stature in BMI estimation among young adult females in Kolkata, India. It also examines the correlation between stature-based BMI and arm-span-based BMI, assessing their predictive accuracy for body weight. A total of 100 female graduate students (aged 20–23 years) participated in this cross-sectional study. Height, arm span, and weight were measured following standard anthropometric protocols. Statistical analyses included correlation analysis, linear regression modelling, and receiver operating characteristics (ROC) analysis, with significance set at p < 0.05. The correlation between BMI calculated from arm span and BMI from stature was strong (R² = 0.918), indicating that both methods yield comparable results. Regression models demonstrated that arm span and arm-span-based BMI explained 38.5% of weight variation (R² = 0.385), while stature and stature-based BMI explained 40.2% (R² = 0.402). Additionally, arm span was a strong predictor of stature (R² = 0.698). ROC analysis confirmed the high predictive power of BMI classification using arm span (AUC = 0.92, accuracy = 91.67%). The findings support the use of arm span as a viable alternative to stature in BMI estimation. Although stature-based BMI demonstrated slightly higher predictive accuracy, arm-span-based BMI provided comparable results, making it a practical substitute in populations where height measurement is challenging. Further research incorporating additional anthropometric variables may enhance predictive accuracy.