Table 2.
Multivariate linear regression analysis of factors independently associated with bone mineral density of the hip or spine.
| Characteristic | Femoral neck (β-coefficient)a | P-value | Lumbar spine (β-coefficient)a | P-value |
|---|---|---|---|---|
| Age (per year) | −0.004 | 0.002 | — | — |
| Non-black race | −0.072 | < 0.0001 | −0.073 | < 0.0001 |
| Weight (per kg)b | 0.003 | < 0.0001 | 0.003 | < 0.0001 |
| HIV-infection | −0.012 | 0.05 | −0.04 | 0.02 |
| Low testosterone | −0.03 | 0.01 | −0.05 | 0.002 |
| Ever use of prednisone | — | — | −0.04 | 0.09 |
| Recent (past 5 years) heroin use | — | — | −1.12 | 0.26 |
| Current methadone maintenance treatment | — | — | −2.53 | 0.012 |
β-coefficients less than zero indicate an inverse association between the characteristic and bone mineral density. Because variables were selected for inclusion in each model only if the univariate association was < 0.2, some variables are excluded from each model.
Substitution of body mass index (kg/m2) for weight (kg) did not significantly alter the results of either model, nor alter the β-coefficients or P-values of the other variables in either model.