Table 4.
Multilevel linear regression of BMD at LS with regard to the study factors
| Unadjusted analysis | Multiple adjusted analysisa | Fully adjusted analysisb | ||||
|---|---|---|---|---|---|---|
| LS | Coeff. (95% CI) | P | Coeff. (95% CI) | P | Coeff. (95% CI) | P |
| DHEAS (μmol/L) | 0.72 (0.35–1.06) | <0.0001 | 0.25 (−0.015 to 0.48) | 0.063 | 0.29 (−0.038 to 0.56) | 0.078 |
| Time (year) | −0.46 (−0.60 to −0.31) | <0.0001 | −0.45 (−0.62 to −0.31) | <0.0001 | −0.47 (−0.69 to −0.29) | <0.0001 |
| Time2 (year2) | 0.038 (0.026–0.049) | <0.0001 | 0.035 (0.026–0.042) | <0.0001 | 0.039 (0.028–0.047) | <0.0001 |
| DHEAS * time | −0.037 (−0.052 to −0.023) | <0.0001 | −0.035 (−0.050 to −0.022) | <0.0001 | −0.033 (−0.053 to −0.017) | <0.0001 |
| Baseline age (year) | −0.63 (−0.82 to −0.47) | <0.0001 | −0.51 (−0.77 to −0.30) | <0.0001 | ||
| BMI (kg/m2) | 0.44 (0.38−0.49) | <0.0001 | 0.55 (0.50–0.60) | <0.0001 | ||
| HRT | 1.52 (1.25–1.76) | <0.0001 | 1.69 (1.36–1.95) | <0.0001 | ||
| Estradiol (pmol/L) | 0.022 (0.013–0.029) | <0.0001 | ||||
The βs are interpreted as percentage change in BMD at lumbar spine per unit increase in the study factors listed in the table. For example: −0.46 for time variable represents a 0.46% decrease in BMD at the lumbar spine per year during the follow-up period
aAdjusted for baseline age, time-varying (BMI and HRT)
bFurther adjusted for estradiol
DHEAS dehydroepiandrosterone sulfate, LS lumbar spine BMD, BMI body mass index, HRT hormone-replacement therapy, DHEAS * time baseline DHEAS and time interaction, Time 2 quadratic form of time, BMD bone mineral density