Arm span increases predictive value of models for prevalent vertebral deformities: The Japanese Population-based Osteoporosis (JPOS) Study

Objective We examined anthropometric indicators to improve predictive ability of asymptomatic vertebral fracture screening models. Study design and setting Data were obtained from the 1996 Japanese Population-based Osteoporosis (JPOS) Study. McCloskey–Kanis criteria diagnosed vertebral deformities on X-ray absorptiometric images in 693 women aged ≥50.The multiple logistic regression model included age, height, weight, postmenopausal status, total hip BMD, and arm span (AS) or sitting height as explanatory variables. Akaike's information criterion (AIC) evaluated model goodness-of-fit. Results Age-adjusted AS and sitting height in subjects with and without vertebral deformities were 147.2 ± 0.6 cm and 148.5 ± 0.2 cm (P = 0.055), 78.5 ± 0.5 cm and 79.9 ± 0.2 cm (P = 0.007), respectively. Every 5-cm increase in AS indicated 1.5-fold increased risk of prevalent vertebral deformity in the model including age, height, weight, postmenopausal status, and BMD. Including the explanatory variable AS in models yielded better predictive accuracy than excluding AS (AIC, 441.7 vs 446.6, respectively). Sitting height did not significantly influence model predictive ability. Conclusion Predictive accuracy of model for vertebral fracture including age, height, weight, postmenopausal status, and BMD improved when AS was added as an explanatory variable. Models to screen for asymptomatic vertebral fractures should include AS.