Associations between hormones, metabolic markers, and bone mass in perimenopausal and postmenopausal women

Characteristics of the participants

In the study population, 17.68% were perimenopausal women, with 60% belonging to the normal bone mass group, a significantly higher proportion compared to those in the osteopenia and osteoporosis groups. E2 levels in perimenopausal women were notably higher than in postmenopausal women, while FSH and LH levels were significantly lower. Additionally, perimenopausal women exhibited significantly higher bone densities in the femoral neck and total hip regions compared to postmenopausal women. However, no significant differences were observed between perimenopausal and postmenopausal women in other bone metabolism parameters or metabolite parameters (Fig. 1).

Fig. 1figure 1

Clinical characteristics of participants and comparison between perimenopausal and postmenopausal women. a Proportion of perimenopausal and postmenopausal women in different bone mass groups. b Comparison of sex hormone levels in perimenopausal and postmenopausal women. c Comparison of bone metabolism parameters. d Comparison of ion levels. e Comparison of metabolite parameters. f Comparison of bone density. Variables were assessed with Mann–Whitney U test. Asterisks indicate significant differences between groups (*p < 0.05; **p < 0.01; and ***p < 0.001)

Table 1 describes the main characteristics of all studied subjects. Age, years since menopause, and BMI were significantly different between women with normal bone mass and those with osteopenia and osteoporosis. As expected, women with normal bone mass (median age:51, quartile: 47–55) were younger than women with osteopenia (median age: 59, quartile: 53–63) and osteoporosis (median age: 66, quartile: 62–69). BMI was lower in women with osteopenia (25.33 ± 3.99 kg/m2) and osteoporosis (23.77 ± 3.56 kg/m2) than in women with normal bone mass (26.42 ± 4.08 kg/m2). There was no significant difference in the prevalence of diabetes between women with osteopenia (57.38%) and osteoporosis (56.52%). However, the prevalence of diabetes in the normal bone mass group (37.78%) was significantly lower than that in the osteopenia and the osteoporosis group.

Table 1 Clinical, hormonal, and metabolic markers’ characteristics of 198 perimenopausal and postmenopausal women according to bone mass statusComparison of hormones and metabolic markers

Bone metabolism indexes β-CTX and P1NP were lower in women with normal bone mass compared with those with osteopenia and osteoporosis. Metabolic markers TG, TC, LDL-C, and GLU were significantly lower in women with normal bone mass than in those with bone mass loss. Mg showed a statistically significant difference between the different groups, but considered to have no clinical significance (Table 1).

For hormone levels, women with normal bone mass exhibited significantly lower concentrations of FSH and LH compared to those with osteopenia and osteoporosis. Conversely, women with normal bone mass demonstrated significantly higher levels of E2 and T relative to those with osteopenia and osteoporosis (Table 1).

Association of hormones and metabolic markers with BMD

Spearman correlation analysis showed that age, years since menopause, GLU, FSH, and LH had significant inverse correlations with femoral neck and total hip BMD. Conversely, BMI, E2, and T had significant positive correlations with femoral neck and total hip BMD (Fig. 2; Table S1). After adjusting for age, the partial correlation analysis revealed that FSH (r = − 0.308, p < 0.001) had significant negative correlations with femoral neck BMD. E2 (r = 0.244, p = 0.001) and T (r = 0.226, p = 0.002) had significant positive correlations with femoral neck BMD. For total hip BMD, the partial correlation analysis showed significant negative correlations with years since menopause (r = − 0.259, p = 0.006) and FSH (r = − 0.257, p < 0.001). Additionally, E2 (r = 0.215, p = 0.003) had a significant positive correlation with total hip BMD (Fig. 2; Table S2).

Fig. 2figure 2

Correlations between hormones and metabolic markers with BMD studied by Spearman correlation analysis and partial correlation analysis. a The color map of Spearman correlation coefficients (r) and p values calculated from Spearman correlation analysis. b The color map of partial correlation coefficients (r) and p values calculated from partial correlation analysis after adjusting for age. Yellow color indicates a positive correlation, while purple color indicates a negative correlation. All the p values are color coded based on the scale on the right

Since the T-score of BMD has a better evaluation effect on bone loss in patients, we further studied correlations between T-score of femoral neck BMD and hormones and metabolic markers. Spearman correlation analysis displayed significant inverse correlations with age, years since menopause, FSH, and LH. In addition, BMI, E2, and T were significantly positively correlated with the T-score of femoral neck BMD. After adjusting for age, T-score of the femoral neck BMD still had significant negative correlations with years since menopause, FSH, and LH. (Fig. 2; Tables S1 and S2).

Risk factors for bone mass loss

To explore possible risk factors for bone mass loss in the studied perimenopausal and postmenopausal women, we performed univariate and multivariate binary logistic regression analysis. The occurrence of bone mass loss (osteopenia or osteoporosis) was set up as the dependent variable. Univariate analysis showed that age, years since menopause, the prevalence of diabetes and the levels of β-CTX, P1NP, TC, LDL, GLU, UA, FSH, and LH were risk factors and BMI, 25-(OH)-D, phosphate, E2, and T were protective factors for bone loss in the studied women (Table 2). Then we performed multivariate binary logistic regression analysis including all variables which had significant influence of bone loss identified by univariate logistic regression analysis, except for years since menopause, because there was collinearity between menopausal time and age. This model revealed that age (odds ratio [OR] 1.223; 95% confidence interval [CI] 1.106–1.372; p < 0.001), GLU (OR 1.848; 95% CI 1.116–3.059; p = 0.017), and FSH (OR 1.089; 95% CI 1.003–1.182; p = 0.042) were risk factors for bone mass loss. No significant effect of other variables enrolled in this model was detected (Table 2).

Table 2 Univariate and multivariate binary logistic regression analysis to identify risk factors for bone mass lossBone mass loss prediction model

Using ROC curve analysis, the optimal cutoff values of risk factors in the prediction of bone mass loss (osteopenia or osteoporosis) occurrence were identified (Fig. 3; Table S3). Age (AUC = 0.884, 95% CI 0.833–0.935) and FSH (AUC = 0.824, 95% CI 0.760–0.888) exhibited significant predictive power for bone mass loss, and the optimal cutoff value for age was 55 years (sensitivity 85.6%, specificity 77.8%) and for FSH 9.26 mIU/mL (sensitivity 88.9%, specificity 64.4%). GLU (AUC = 0.683, 95% CI 0.599–0.768) showed moderate discrimination capability of bone mass loss, and the optimal cutoff point was 6.14 mmol/L (sensitivity 75.2%, specificity 53.3%).

Fig. 3figure 3

ROC curves for predicting bone mass loss by using age and levels of FSH and GLU, either independently or in combination. The vertical axis represents sensitivity, and the horizontal axis represents 1-specificity

To evaluate the predictive value of the combined use of these three risk factors for bone mass loss, we constructed a multivariate logistic regression model including age, GLU, and FSH and obtained a regression formula as follows: ln(p/1 − p) = − 14.773 + 0.212 * age + 0.401 * glu + 0.042 * FSH, where p represents the probability of bone mass loss. Hosmer–Lemeshow test showed a good fit of this model (χ2 = 3.272, p = 0.916), Table S4. The combined use of all three risk factors by multivariate logistic regression showed a significant prediction power (AUC = 0.930, 95% CI 0.893–0.967) for bone mass loss in perimenopausal and postmenopausal women. A probability of 0.695 (sensitivity 89.5%, specificity 84.4%) was considered to be the optimal cutoff point for this prediction model (Fig. 3; Table S3).

Comments (0)

No login
gif