Background
Osteoporosis (OP) is a systemic skeletal disorder characterized by low bone mass and deteriorated microarchitecture of bone tissue. The osteoporotic bone is fragile and susceptible to fracture [
1]. Postmenopausal osteoporosis (PMO) results from estrogen deficiency after menopause and is a major type of primary osteoporosis [
2]. PMO may lead to fragility fracture and is one of the most disabling consequences of postmenopausal women. Early identification and effective therapeutic monitoring to PMO is necessary to reduce the public health burden of this pervasive disease [
3].
Osteocalcin (OC), also known as the bone Gla protein (BGP), is a 5.8 kDa, hydroxyapatite-binding protein that could be synthesized by osteoblasts, odontoblasts and hypertrophic chondrocytes [
4] and is the most abundant non-collagenous protein found in bone matrix. On the one hand, OC molecules can be released directly into blood after osteoblastic synthesis during bone formation; on the other hand it can also enter the circulation from osteoclastic bone matrix degradation during bone resorption. Therefore, circulatory OC may come from both bone formation and bone resorption [
5,
6] and serum osteocalcin (sOC) level may theoretically increase in PMO, which is characterized by high bone turnover status with both increased bone formation and bone resorption.
Although serum osteocalcin (sOC) has been widely used as a bone turnover marker to indicate the high bone turnover status in PMO [
7‐
9], the change of sOC level in PMO cases compared to controls remains controversial. Some of the previous studies have found a higher sOC level in PMO patients than age-matched controls, which is consistent with the high bone turnover status in PMO cases [
10‐
12]. Meanwhile, there are also studies reporting the same or even lower sOC level in PMO cases than controls. In 2012, Biver et al. [
13] performed a meta-analysis to compare several bone turnover markers between osteoporotic cases and controls and found no significant difference in sOC, which is so far the only pooled evidence of sOC change in OP. However, they included all types of osteoporosis where bone metabolic patterns might be quite heterogeneous. More studies have emerged since then and the pooled evidence of sOC change in PMO cases compared to controls is needed.
To obtain the evidence, we performed a meta-analysis of the available case-control studies to compare the sOC level between PMO patients and age-matched postmenopausal controls.
Methods
We followed the Preferred Reporting Item for Systemic Review and Meta-analysis (PRISMA) [
14,
15] and performed this meta-analysis based on a protocol, according to the recommendations of the Cochrane Collaboration.
Data sources and search strategies
We searched PubMed and Cochrane Library online for all the available case-control studies published up to (12 October 2018), without restriction to regions and languages. Detailed strategy was: (“biological markers”[Title/Abstract] OR “turnover markers”[Title/Abstract] OR “osteocalcin”[Title/Abstract] OR “BGP”[Title/Abstract]) AND (“postmenopausal osteoporosis”[Title/Abstract] OR “type I osteoporosis”[Title/Abstract]). Conventional searches were supplemented by manual searches of the reference lists of all the relevant studies, review articles and conference abstracts.
Inclusion and exclusion criteria
Studies were included if they met the following inclusion criteria: 1.case-control study design; 2.patients in the case group were diagnosed of PMO, according to the diagnostic criteria recommended by WHO [
16]; 3.participants in the control group were age and sex matched with cases; 4.sOC level was reported. Studies were excluded if they met the following exclusion criteria: 1.secondary osteoporosis, or under any circumstance that bone turnover or sOC level might be affected (e.g., hyperparathyroidism, diabetes mellitus, liver disease, renal insufficiency, anti-osteoporotic therapy, and long-term corticosteroid therapy); 2.studies including males or non-postmenopausal females.
Study selection and data extraction
Two literature reviewers evaluated the eligibility of potential titles and abstracts independently. Included studies were reassessed as full text by inclusion and exclusion criteria. Disagreement was solved by discussion. Further adjudication of a third reviewer was performed if the disagreement remained. The following data were then extracted from each eligible study: the first author’s name, year of publication, demographic information (age, region, and number of people in case and control groups), sampling condition, manufacturer of the OC assay kit, target fragment of the sOC molecule and sOC level.
Quality assessment of the included studies
Assessment to the risk of bias of case-control studies was performed using the Newcastle-Ottawa Scale (NOS) [
17], as recommended by Cochrane Collaboration. A score of 0–9 was allocated to each study. Studies achieving six or more points were considered of high quality, which is essential for a high-level pooled evidence. Again, two reviewers independently evaluated the included studies and disagreement was solved by discussion. Further adjudication of a third reviewer was performed if the disagreement remained.
Statistical analysis
The pooled sOC level was compared between cases and controls. Statistical analysis was performed using Review Manager (RevMan 5.3) software. First, heterogeneity of the included studies was tested. Heterogeneous data between studies was indicated by p ≤ 0.10 (or I2 ≥ 50%), and homogeneous data was indicated by p > 0.10 (or I2 < 50%). A Fixed Effect model was used for homogeneous data, whereas a Random Effects model was used for heterogeneous data. Continuous variables of sOC level were reported with 95% confidence interval (95% CI) of the mean difference (MD) and p value. The test of overall effect with a p < 0.05 was considered to be statistically significant. Collected data were carefully inputted, and then rechecked by two reviewers respectively.
Subgroup analysis according to different fragments of the sOC molecules and different regions of the study population was also performed, with the same method mentioned above. Heterogeneous data between subgroups was indicated by p ≤ 0.10 (or I2 ≥ 50%) and homogeneous data between subgroups was indicated by p > 0.10 (or I2 < 50%).
Discussion
Serum OC has been widely applied to indicate bone turnover status in PMO management with several advantages such as bone specificity, non-invasiveness and dynamic-response. However, updated knowledge of OC challenges its traditional role as a bone turnover indicator in vivo, and studies have reported conflicting results of the change of sOC level in PMO cases.
PMO presents a high bone turnover rate with both increased bone resorption and formation [
10‐
12]. Accordingly, most of the included studies (six out of ten) in our meta-analysis reported a significantly increased sOC level in PMO patients. Yet, there were also three studies reporting the same sOC level between cases and controls and even one study reporting a decreased sOC level in PMO cases.
The subsequent meta-analysis of all the included studies revealed only a trend of increase in sOC level in PMO cases compared to controls [95%CI: (− 1.49, 5.16), p = 0.28], which is not statistically significant. Subgroup analysis according to different regions and sOC fragments also revealed the same sOC level between cases and controls except in the European subgroup, where a significantly increased sOC level was observed in PMO patients.
By sensitivity analysis, we found that the pooled result was significantly influenced by Al-Daghri’s study [
27], in which a significant decrease of sOC level was reported in PMO patients. We analyzed the Asian subgroup and found that after removal of Al Daghri’s study, the increase of pooled sOC level in Asian PMO cases compared to Asian controls became significant (from 0.98 to 0.03), just consistent with the result found in European subgourp. Therefore, it is possible that the different results reported in the subgroup analysis was not due to regional differences, but rather resulted from the effect of Al Daghri’s study on the Asian subgroup. However, every included study counts and we cannot ignore any study meeting the inclusion criteria yet with unexpected result. According to previous studies, osteocalcin, once undercarboxylated (unOC), had been found to act as a hormone [
28] and had a board spectrum of interactions with glucose metabolism [
29], fertility [
30] and even aging [
31]. In Al-Daghri’s study, the exact reason for the decreased sOC level in PMO cases remained unknown, but interactions of OC with glucose metabolism might have an effect on the result since fasting glucose levels in cases were slightly higher than controls and reached over 7.0 mmol/L (PMO vs control: 7.2 ± 3.6 vs 6.5 ± 3.2 mmol/L), indicating possible energy metabolism disorder in PMO group.
The major result of our current meta-analysis indicated no significant difference in the pooled sOC level between PMO cases and controls, which was consistent with the previous relevant meta-analysis reporting the same sOC level in osteoporotic and healthy populations in 2012 [
13]. However, they included both primary (both type I and type II) and secondary osteoporosis, which might have introduced confounding factors into the analysis. In contrast, we compared only PMO cases with age and sex matched controls, in which the bone metabolic pattern was homogeneous and well clarified. Moreover, subgroup analyses according to the potential confounding factors (region of the study population and sOC fragment) were further performed. Therefore, our analysis was more specific and reliable to PMO. Besides, we have also compared the sOC levels in male population with and without OP in a previous meta-analysis and found no significant difference either [
32].
However, some limitations should be mentioned in this meta-analysis. First, all the included studies were case-control studies due to the design of our research, and the level of evidence was lower than randomized controlled trials (RCTs). Second, significant heterogeneity existed between studies, which could still not be eliminated by subgroup analysis. The heterogeneity between studies might probably come from: 1. variations in the target sOC fragments evaluated by different assay kits; 2. variations in the experimental conditions [
33,
34]. The intact OC molecule is rapidly degraded in serum, generating variable segments, thus limiting its utility in lab tests [
35,
36]. Therefore, new techniques for sOC evaluation as well as standardization of methodologies and experimental conditions are urged for better application of sOC. Third, due to limited information from the included studies, physical activity levels could not be considered for variable control in our analysis, which could also have an effect on bone remodeling. However, there might not be a significant difference in physical activity levels in the senile postmenopausal populations between PMO cases and controls as long as the age and postmenopausal status were matched. Last but not least, the carboxylation status of sOC was not mentioned in any of the included studies. As a matter of fact, both OC and unOC in the circulation have been reported to correlate with each other [
37,
38] and with bone quality [
39,
40] but most of the conventional sOC assay kits are designed for total sOC while unOC can only be evaluated with special techniques [
41,
42]. The result of this meta-analysis might not be significantly influenced by the carboxylation status of sOC as the same OC assay kit was adopted in PMO and control groups within each individual study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.