Table 1 of univariate analysis showed that mean values of weight, height, and BMI were significantly lower (p < 0.0001) in women with osteoporosis, while their mean age was significantly higher compared to normal control subjects (68.2 vs. 61.3 years, respectively, p < 0.0001). The mean values of BMI for both groups were above 30 kg/m2, which indicated that osteoporosis and control subjects were obese. No significant differences were detected between mean values of calcium, 25(OH)D, PTH, and bone turnover markers (CTX1 and PINP) in women with osteoporosis compared to controls. Due to the fact that there was a significant difference in the mean age between control and osteoporosis groups, the differences between the indicated markers (vitamin D, Ca, PTH, CTX1, and PINP) were reevaluated after taking a subset of the control group with a mean age that matches the mean age of the osteoporosis group. Similar results were obtained that ruled out the differences between the indicated markers being due to age (data not shown).
Table 2 showed that 14.1% of the study subjects had severe vitamin D deficiency (< 10 ng/ml 25(OH)D with mean ± SD 8.51 ± 1.28 ng/ml), whereas 71.8% had vitamin D insufficiency or borderline [25(OH)D 10 to < 20 ng/ml with mean ± SD 13.26 ± 2.20 ng/ml]. Subjects with normal (vitamin D sufficient) represented 14.1% [25(OH)D ≥ 20 ng/ml with mean ± SD 24.55 ± 3.60 ng/ml] [7, 39]. Chi-square analysis showed no statistically significant difference in the prevalence of vitamin D groups between osteoporosis and control subjects. Table 2 also showed the levels of calcium and PTH in subjects in the three vitamin D groups. ANOVA showed that mean serum calcium levels were significantly higher in subjects with vitamin D sufficiency compared to serum calcium levels in subjects with severe vitamin D deficiency or vitamin D insufficiency (p = 0.022). However, it should be indicated that serum calcium levels in the three groups were within the normal range. When PTH levels were compared in the three groups, ANOVA showed that mean values of PTH levels were significantly lower in subjects with vitamin D sufficiency (p < 0.0001) compared to vitamin D insufficiency or severe vitamin D deficiency subjects. The mean values of PTH were not statistically different between severe vitamin D deficiency and vitamin D insufficiency subjects. Approximately, 27% of recruited subject had previous accident that caused broken bones with no difference between osteoporotic and control subjects.
A scatter plot showing the relationship between vitamin D and PTH is depicted in Fig. 1. Vitamin D was negatively correlated with PTH in all subjects (r = −0.295, p < 0.01) as well as in osteoporosis subjects (r = −0.30, p < 0.01).
Table 3 showed the Pearson correlation coefficients of the parameters measured in the tested subjects. Serum 25(OH)D was significantly negatively correlated with PTH, weight, and BMI and positively correlated with calcium. No statistically significant correlations were detected between vitamin D and BMD at the three indicated sites (total hip, femoral neck, and lumbar L1–L4 spine), as well as with age, height, bone resorption (CTX1), and bone formation (PINP) markers. Age was negatively correlated with BMD (total hip, femoral neck, and lumbar L1–L4 spine), height, weight, and BMI. Pearson correlation showed the markers for bone resorption (CTX1) and bone formation (PINP) were positively correlated (r = 0.529, p < 0.01). This correlation between bone-remodeling markers was very strong in subjects with osteoporosis (r = 0.887, p < 0.000) whereas no correlation was seen in normal control subjects (r = 0.03, p = 0.74). Among control subjects, PINP was negatively correlated with femoral neck BMD (r = −0.199, p = 0.015) and height (r = −0.173, p = 0.034) and positively correlated with BMI (r = 0.190, p = 0.021). Calcium levels were positively correlated with vitamin D levels and BMD of total hip, whereas calcium was negatively correlated with PTH. These correlations, despite being significant, were not strong. Body mass index was positively correlated with BMD (in total hip, femoral neck, and lumbar L1–L4 spine) and PTH and negatively correlated with vitamin D.
Correlations within the osteoporosis group were also investigated. No significant correlations were seen between calcium and vitamin D, calcium and PTH, BMI and vitamin D, BMI and PTH, BMI and femoral neck BMD, BMI and age, and height and BMD at the three sites. Lumbar spine BMD was positively correlated with vitamin D (r = 0.193, p = 0.033), negatively correlated with PTH (r = −0.198, p = 0.030), and not correlated with height and age.
To determine which variables could predict the BMD level when controlling for the effect of others at each of the tested locations, multiple regression was performed on all variables (Table 4). The results showed that irrespective of the location, both age and weight of the participants appeared to be significant determinants of BMD. Weight appeared to have a positive medium effect on BMD at all locations, whereas age was a negative predictor that lost some of its correlation strength with BMD in the lumbar spine. All other variables did not show any association with BMD at any tested location, except for vitamin D, which appeared to have an effect on the lumbar spine. The models in Table 4 suggested that the variables entered in the analysis had the strongest predictive effect on total hip BMD (R
2 = 0.34) but less predictive effect on BMD of femoral neck (R
2 = 0.29) or lumbar spine (R
2 = 0.225).
To test which variable has a significant effect that could be a determinant on the probability that the tested subject would become osteoporotic, binary logistic regression was carried out (Table 5). The results showed that age and obesity status of the subject could be important determinants of osteoporosis. Age progression appeared as a risk factor with odds ratio (OR) = 1.1 (1.05–1.15). Obesity status was protective against osteoporosis development with a borderline effect for being overweight (OR = 0.11, p = 0.053) and a very clear effect for being obese (OR = 0.05, p = 0.007). The protective effect of obesity on the development of osteoporosis was only seen in the total hip of obese subjects (OR = 0.195, p = 0.04) but not in overweight subjects. Among all models tested, the overall osteoporotic status model was the strongest in terms of its predictive power and joint effect of the significant variables on the change in the status compared to individual sites. Age was the only significant predictor at all sights and the only significant predictor for osteoporotic development in the femoral neck and lumbar spine.