Table 2.
Multivariable Models of Change in Lumbar Spine (L1–L4) Bone Mineral Density From Baseline to Week 48, by Randomized Treatment Group
| Model | Variables in Model | Estimated Effect Coefficient (Standard Deviation) | P Valuea |
|---|---|---|---|
| 1 | Randomized treatment group (vitamin D3 vs placebo) | 0.491 (0.423) | .246 |
| 2 | Randomized treatment group (vitamin D3 vs placebo) | 0.575 (0.414) | .166 |
| Baseline 25-OHD (ng/mL) | –0.048 (0.021) | .026 | |
| Baseline lumbar spine BMD (g/cm2) | –2.541 (1.321) | .056 | |
| 3 | Randomized treatment group (vitamin D3 vs placebo) | 0.494 (0.418) | .238 |
| Baseline 25-OHD (ng/mL) | –0.044 (0.022) | .049 | |
| Baseline lumbar spine BMD (g/cm2) | –2.406 (1.471) | .103 | |
| Covariatesb |
Abbreviations: 25-OHD, 25-hydroxy vitamin D; BMD, bone mineral density.
a P value from likelihood ratio χ2 test.
bCovariates in the model (continuous variables unless otherwise specified) were as follows: age at enrollment (years); race (Black/African American vs other race); sex at birth (male or female); body mass index (kg/m2); lean body mass (g); efavirenz use ever (yes/no); duration of tenofovir disporoxil fumarate use (months); CD4 at entry (cells/μL); viral load (detectable vs below limit of quantitation); weight-bearing exercise (yes or no by questionnaire at baseline). In the full model, none of these covariates was statistically significantly associated with change in lumbar spine BMD from baseline to week 48 (P > .05).