Osteoporosis(OP) is a condition resulting in an increased risk of skeletal fractures due to a reduction in the density of bone tissue [1]. and it is common in postmenopausal women. About 30%of postmenopausal women in Asian countries have OP [2]. It is believed that OP serves as the biggest trigger for osteoporotic fracture, and it can elevate the occurrence of fractures, for every 10% decrease in bone mineral density (BMD), the risk of fracture will increase by 2–3 times [3]. Research findings that more than 15% of postmenopausal women may have a hip fracture, and 50% of postmenopausal women may have a lifelong osteoporosis fracture [4]. Therefore, know about the influencing factors of OP in postmenopausal women are of great significance to the public health.

The occurrence of OP is related to gender, age, hormones, etc. [5]. Among them, hormones are believed to be closely related to OP, and thus gained great attention. In postmenopausal women, the decreased in estrogen levels also leads to changes in other hormones, which can significantly increase the incidence of OP [6]. Previous studies have illustrated that the occurrence of OP is related to thyroid hormone. Abe et al. [7] performed the experiments on the TSHR knockout mouse model and found that the TSH level of the TSHR knockout homozygous mice increased, but the BMD level decreased. Exogenous thyroid hormone supplementation did not reverse bone mineral density loss, suggesting that thyroid stimulating hormone (TSH) may be an independent risk factor for OP.

However, we found that there are a lot of controversies about the relationship between serum TSH and BMD of postmenopausal women, based on the published articles or evidence. Cui Xinjie et al. [8] found that serum TSH and lumbar BMD in postmenopausal women with normal thyroid function was positively correlated. However it was reported that serum TSH in postmenopausal women is negatively correlated with lumbar vertebral BMD by Wang Yi [9]. Besides, Yin Fei et al. [10].did not discover the statistical correlation between the serum TSH and the total hip BMD of postmenopausal. There is still no meta-analysis on the exact relationship between serum TSH levels and OP in postmenopausal women with normal thyroid function. Thus, the present systematic review attempted to identify the Relationship between Normal Range of serum thyroid-stimulating hormone and bone mineral density in the postmenopausal women.

Materials and methods

Articles searching strategy

In this study, a total 6 databases were retrieved, we are proficient in English and Chinese, so we searched articles in English in Cochrane Library, PubMed and Web of Science. For Chinese, we searched in CNKI, Wanfang and VIP, and the retrieval time of each database was from database construction to November 30, 2020. See details in Fig. 1.

The key words mainly included the following: postmenopausal women, older women, thyroid-stimulating hormone (TSH), BMD, osteoporosis. Taking Pubmed as an example, the specific retrieval formula is as follows: ((((Postmenopausal women) OR Older women)) AND ((thyroid-stimulating hormone) OR TSH)) AND ((BMD) OR Osteoporosis). When the same article is published in both Chinese and English, we choose the English article for analysis and select according to the inclusion and exclusion criteria.

Inclusion criteria

① Observational studies in both Chinese and English version; ② Research variables are TSH and OP; ③ Outcome index is the correlation relationship and the effective index is r, OR, mean; ④ The research objects in the original articles are postmenopausal women; ⑤ Articles are either in Chinese or English.

Exclusion criteria

① Review articles; ② Spearman correlation coefficient articles; ③ Studies with incomplete data or offer no way to extract the calculated r, OR; ④ Articles unrelated to correlation coefficient between TSH and BMD; ⑤ Abnormal thyroid function.

Statistical analyses

The correlation coefficient of less than 0.5 does not observe the normal distribution. As it moves to more than 0.5, the result remains the same, hence Fisher proposed to use the “Fisher’ Z transformation” formula for conversion, which adapts the correlation coefficient r into a normally distributed variable Z. Since this study adopted meta-analysis based on the Pearson’s correlation coefficient, the “Fisher’ Z transformation” [11] formula could be used for conversion. The specific conversion formula is as follows:

$$Fisher{^{\prime}}\,\,Z{\rm{ }} = {\rm{ 0}}{\rm{.5}} \times {{1 + r} \over {1 - r}}$$
(1)
$$SE\, = \,\sqrt {\frac{1}{{n - 1}}}$$
(2)
$$Summary{\rm{ }} = {{{e^{2Z}} - 1} \over {{e^{2Z}} + 1}}(Z{\rm{ }}is{\rm{ }}summary{\rm{ }}Fisher{^{\prime}}\,\,Z{\rm{ }}value)$$
(3)

Our study applied Revman5.3 software for statistical analysis, and P < 0.05 was considered statistically significant. EXCEL2010 and Stata12.0 software was used to calculate converted data, draw a dose-response diagram. The above-mentioned formula was adopted to convert the data taking correlation coefficient r as the outcome variable, and thus the Fisher’ Z and standard error (SE) were obtained. Secondly, the Revman5.3 software was applied to perform the inverse variance method, to obtain the summary Fisher’ Z value. Finally, we adopted the formula ③ to gain the combined effect summary r of the correlation coefficient. All of these steps were taken to evaluate the correlation between serum TSH and BMD. In terms of the studies on the relationship between TSH and the risk of OP, most of the articles included only reported the effect size and its 95% CI, and most of them were adjusted for confounding factors. Therefore, the Revman5.3 software can calculate SE and the logarithm of OR (log OR) and then its combined effects observed. Two independent reviewers performed the data extraction, and a third reviewer was consulted for any uncertainties.

Risk of Bias across Studies

Publication Bias

We used a funnel plot and Egger’ s test to evaluate whether there was a publication bias in the included articles.

Sensitivity analysis

The Stata12.0 software was applied for sensitivity analysis. The Chi-square test was performed withα = 0.05 as the significance level; when P < 0.05, the difference was considered statistically significant.

We mainly used the Review Manger 5.3 and Stata12.0 software for data analysis. Heterogeneity is divided into two degrees according to I2, I2 < 50% is low heterogeneity that is acceptable, I2 ≥ 50% is high heterogeneity, andα = 0.05 was applied as the significance level for hypothesis testing of heterogeneity I2. When P < 0.05, I2 ≥ 50%, indicating heterogeneity among multiple studies, the combined effect of OR and its 95% confidence interval is estimated by the random-effects models, and when P > 0.05, I2 < 50%, indicating homogeneity among multiple studies, the fixed-effect model was used to estimate the combined effect and its 95% confidence interval. When the heterogeneity is high, subgroup analysis will be conducted according to the source of the study objects, the detection site of bone mineral density, and the use of anti-osteoporosis drugs in order to find the source of heterogeneity.

Fig. 1
figure 1

Search process and results

Full size image
Table 1 Basic characteristics of the included articles
Full size table

The method recommended by the Agency for Health care Research and Quality (AHRQ) is adopted in evaluating the quality of the cross-sectional studies. It contains 11 items, with a maximum score of 11 points. Articles scored 0–3 points, 4–7 points, and 8–11 points are classified into low quality, medium quality, and high-quality respectively (Additional file 3).19 articles were included in this study are of medium quality.

The quality assessment was independently conducted by the first author, and the second author checked and collated the results in detail. Discussed and resolved any disagreement with the third author.

Results

Basic information of the included articles

This study included 19 articles with 23,960 subjects, and the publication time ranged from 2006 to 2020. Among which, 12 articles were published in China, 3 in South Korea, 1 in the United States, Italy, Israel, and Turkey respectively (Table 1).

The relationship between serum TSH and BMD in postmenopausal women based on the Pearson correlation coefficient

The effect sizes were all Pearson correlation coefficients. The value of the correlation coefficient r in the previous studies was converted by the above-mentioned formula, and shown in Table 2.

We found that TSH was positively correlated with BMD, Fisher’ Z = 0.16, 95% CI (0.00, 0.32), Z = 1.98, P = 0.05 (Fig. 2). The final combined effect value r of TSH and BMD was 0.158.

Table 2 Analytical values obtained from data conversion
Full size table
Fig. 2
figure 2

Meta-analysis of correlation between serum TSH and BMD.

Full size image
Fig. 3
figure 3

Comparison of BMD between high-level/low-level TSH group and control group

Full size image

BMD comparison among TSH groups with different levels in postmenopausal women with normal thyroid function

In original studies, the TSH was trisected according to the Tri -sectional quantiles, namely low, medium, and high. The cutoff value was incorporated into the previous group. We took the middle-level TSH group was used as the control in this study, to analyze the difference in BMD between the high-level TSH group and the low-level TSH group. The BMD of the high-level TSH group was higher than that of the control group, with SMD of 0.22, 95% CI (0.08, 0.35), P = 0.001. The BMD of the low-level TSH group was statistically lower than the control group, SMD at -0.31, 95% CI (-0.44, -0.18), P < 0.001 (Fig. 3).

The relationship between TSH and osteoporosis

Multivariate logistic regression could determine the frequency of osteoporosis in different groups with different TSH levels after adjusting confounding factors (age, BMI, BMD, usage of anti-osteoporosis drugs), combine the effect size OR included in the article.

Fig. 4
figure 4

Meta-analysis of correlation between serum TSH and OP.

Full size image

We found that the risk of osteoporosis for low level TSH was 1.76 times, 95% CI (1.27, 2.45) of that high level TSH (Fig. 4). This indicates that low level TSH will increase the dependence of OP.

Results of subgroup analysis

A subgroup analysis of the source-based research subjects with OR as the measurement index, found that low level TSH group in the community facing higher risk of osteoporosis, OR = 1.89, 95%CI (1.43, 2.49), P < 0.01. In addition, whether anti-osteoporosis drugs were taken or not, low level TSH increased the risk of osteoporosis, OR at [1.84, 95%CI (1.05, 3.22), P = 0.03] and [1.74, 95%CI (1.08, 2.82), P = 0.02] respectively. The results of subgroup analysis of studies with the Pearson correlation coefficient as the outcome index could be found from Table 3. The subjects took calcium and other anti-osteoporosis drugs, summary Fisher’ Z = 0.14, 95%CI (0.02, 0.26), P = 0.03, which could affect the relationship between TSH and BMD. BMD detection site, whether the subjects suffered diabetes or not etc, had no influence on the relationship between TSH and BMD (Table 4).

Table 3 Subgroup analysis to determine the influencing factors of the relationship between TSH level and OP
Full size table
Table 4 Subgroup analysis to determine the influencing factors of the relationship between TSH level and BMD
Full size table

The dose-response relationship between different TSH levels and osteoporosis

It could be seen in additional file 4 that OR of different TSH levels and OP, including 5 research [12, 16, 17, 21, 27] to reflect the dose-response relationship, each study used the high-level group as a control to analyzed the risk of osteoporosis in the different levels of TSH group within the normal range. The TSH level in the table was the average level. We employed Stata12.0 software to draw a dose-response diagram.

Fig. 5
figure 5

The dose-response relationship of osteoporosis at different TSH levels

Full size image

According to the dose-response relationship, we knew that even when TSH was within the normal range, TSH = 2.5mIU/L, the risk of osteoporosis kept in a high level. When it was lower or higher than 2.5mIU/L, the development of OP could be diminished, and the risk of osteoporosis gradually decreased with the increasing TSH level (Fig. 5).

Sensitivity analysis

After eliminating included articles one by one, it was found that with Wang Yi [9] Summary Fisher’ Z value at -2.65, and 95% CI (-2.84, -2.46), the conclusion of the study was totally opposite. Therefore, this article was excluded in the subsequent analysis (Additional file 5). In the research of the relationship between TSH and osteoporosis based on the OR value, the results were relatively stable after screening each studies (Fig. 6).

Fig. 6
figure 6

Sensitivity analysis results based on OR value

Full size image

Publication bias

Publication bias was assessed using Egger’s test. Taking effect as the publication bias of Pearson correlation coefficient, it can be seen that the overall sample size included in the study was large and bilateral symmetric, and the publication bias with effect size as OR value is found to be P = 0.120, 95%CI (-0.663, 4.745). Therefore, no single article appears to influence the combined results (Fig. 7).

Fig. 7
figure 7

Funnel plot for the relationship between TSH and BMD (A); the relationship between TSH and osteoporosis (B)

Full size image

Discussion

Overt hypothyroidism is known to lower bone turnover by reducing both osteoclastic bone resorption and osteoblastic activity. These changes in bone metabolism would result in an increase in bone mineralization [28]. It is reported that the lifetime risk of osteoporosis in women is 40%-50% [29]. There were about 30% postmenopausal women who had osteoporosis in China [30]. However the relationship between serum TSH and BMD in postmenopausal women was still controversial, this study conducted meta-analysis and found that postmenopausal women’ s serum TSH was positively correlated with BMD, r = 0.158. The risk of osteoporosis in postmenopausal women with low-level TSH was 1.76 times of those with high-level TSH, 95% CI (1.27, 2.45). The dose-response relationship showed that when TSH was above 2.5mIu/L, the incidence of osteoporosis tended to be decreased. These results provided a theoretical basis for the prevention and treatment of osteoporosis in postmenopausal women.

Studies have demonstrated that serum TSH acts independently from thyroid hormone in bone metabolism. Wang Jiadan et al. [22]. found that the fluctuation of TSH level in the reference range in postmenopausal women with normal thyroid function may have a certain impact on the BMD of femoral neck, total hip and ward triangle. In addition, the TSH level may be an independent influencing factor of BMD in femoral neck and ward triangle, and those with low TSH level have a higher risk of osteopenia. TSH gets involved in bone turnover mainly through the following aspects: Firstly, TSH can promote osteoblast to secrete OPG, which can competitively inhibit the binding of RANK and RANKL, thus curbing the differentiation of osteoclast precursor cells into osteoclast. Secondly, TSH can activate protein kinase C6 in osteoblast to up-regulate frizzled and Wnt5a in non-canonical pathways, and induce osteoblast differentiation [31], which leaded to increasing level of BMD. This is consistent with the changing trend of TSH and BMD in the study of Wang Xiaodong [32]. The dose-response relationship showed that if the TSH level was below 2.5mIu/L, the risk of osteoporosis gradually increased, but TSH was up to 2.5mIu/L, it decreases with the increase of TSH level. BMD of the high-level TSH group was higher than that of the low-level TSH group, comparing with the BMD of the medium-level TSH group. It demonstrated that TSH was elevated when the BMD increased. Thirdly, TSH inhibits TNF-α and the proliferation and differentiation of osteoclast by binding to TSHR expressed on the surface of osteoclast [31]. In addition, TSH inhibits the binding of RANK and RANKL by directly inhibiting RANKL [33], thereby containing osteoclast precursor cells differentiate into osteoclast and curbing the formation of osteoclast [34] (Fig. 8), and reducing bone resorption.

Fig. 8
figure 8

The mechanism of TSH involved in bone metabolism

Full size image

From the Subgroup analysis it was found that the low-level TSH population from the community had a higher rate of osteoporosis than the patients from hospitals. Whether the anti-osteoporosis drugs were taken or not usually had no influence on the heterogeneity. According to subgroup analysis results targeting the different body part of BMD measurement and whether the subjects suffer from diabetes, we found that the relationship between TSH and BMD constant.

This study has a few limitations. First is that the articles included in this study are from cross-sectional survey and only conducted one measurement of TSH and BMD. Second, the value of BMD is not continuously measured as TSH changes. It will be more convincing that TSH and BMD be measured at a certain interval with multiple measurements. However, the concerned studies were tested for publication bias, and the results illustrated that there was no bias. Sensitivity analysis indicated that Wang Yi’s article [8] was highly sensitive, so it was excluded in the analysis. In future, the controlled trial should be conducted to clarify the relationship between TSH and BMD in postmenopausal women. At the same time, in vitro and in vivo experiments ought to be carried out to further explore how TSH promotes the secretion of OPG by osteoblast, and the specific signaling pathways and molecules that play a role in inhibiting the proliferation and differentiation of osteoclast.

Conclusions

In summary, the serum TSH in the normal range of postmenopausal women was positively correlated with BMD, and high level (> 2.5mIu/L) within the normal range is helpful to decrease the risk of osteoporosis in postmenopausal women.