Association of circulating osteocalcin with cardiovascular disease and intermediate cardiovascular phenotypes: systematic review and meta-analysis

Abstract

Objectives. Circulating osteocalcin (OC), a marker which is central in bone mineralization, may be involved in the atherosclerotic process and influence the risk of developing cardiovascular disease (CVD). We conducted a systematic review and meta-analysis of published observational evidence, to assess and quantify the associations of circulating OC (total, undercarboxylated, and carboxylated OC) with cardiovascular outcomes (clinical CVD endpoints and intermediate cardiovascular phenotypes).

Design. Relevant studies were identified in a literature search of MEDLINE, EMBASE, and reference lists of relevant studies to March 2019. Mean differences and risk ratios with 95% CIs were aggregated using random-effects models.

Results. Thirty-three observational studies (prospective and retrospective cohort, case-control, and cross-sectional) with data on 21,021 unique participants were eligible. The pooled risk ratio in a comparison of extreme fourths of total OC levels was 0.98 (95% CI 0.89, 1.08) for composite CVD. Circulating total OC levels were significantly lower in patients with cardiovascular conditions compared with those without these conditions −2.58 ng/ml (95% CI −3.85, −1.32; p < .001). Prospective and cross-sectional data showed significant inverse associations between total OC and traits such as aortic or coronary calcification, coronary atherosclerosis or calcification, carotid intima-media thickness, and plaque score. There was limited data on carboxylated and undercarboxylated OC, with no evidence of associations.

Conclusion. Observational evidence generally supports inverse associations of circulating total OC with risk of atherosclerotic outcomes and CVD endpoints; however, the data were mostly based on cross-sectional evaluations. Large-scale prospective data are needed.

Keywords:

• osteocalcin
• cardiovascular disease
• coronary heart disease
• atherosclerosis
• carotid intima-media thickness
• meta-analysis

Introduction

Cardiovascular disease (CVD) is still the leading cause of global mortality [1]. By 2030, an estimated 23.6 million people will die from CVD [2]. Major risk factors for CVD include a history of diabetes, high blood pressure, raised blood lipids, as well as smoking status [3]. Though these established risk factors explain a large proportion of the risk of CVD, its pathogenesis is still not fully established as it appears other additional factors may be involved. Osteocalcin (OC), a bone matrix protein mainly expressed by osteoblasts [4] and used as a biochemical indicator of bone formation [5], has been proposed as a potential biomarker of CVD risk. Osteocalcin is secreted by osteoblasts in a fully carboxylated form, which is then decarboxylated to a more active biological form [4,6,7]. Circulating blood levels of total OC comprises both carboxylated and undercarboxylated OC. Serum OC has been investigated as a hormone which plays a role in regulating glucose metabolism and fat mass; it promotes insulin sensitivity and secretion by pancreatic β-cells, as well as increases proliferation of the β-cells and stimulates energy metabolism [6]. Reduced circulating levels of OC have been consistently linked with high blood glucose levels, insulin resistance, and type 2 diabetes [6,8–10]. Indeed, it has been discussed that OC may hold potential for the prevention, delay and treatment of obesity and metabolic disorders such as diabetes [8]. Emerging evidence also suggests circulating OC may be involved in CVD development, but the data are sparse and conflicting. A number of studies have shown reduced levels of circulating OC to be associated with adverse cardiovascular outcomes such as arterial calcification, carotid atherosclerosis, increased carotid intima-media thickness (CIMT), and CVD; [11–13] whereas others have not demonstrated any relationships [14,15]. In this context, we performed a systematic review and meta-analysis of all available published observational evidence to clarify and quantify the extent of potential associations of circulating OC (total, uncarboxylated, and carboxylated OC) with (i) clinical CVD outcomes as well as all-cause mortality; and (ii) intermediate cardiovascular traits. We also sought to identify gaps in the existing evidence.

Methods

Data sources and search strategy

This review was conducted using a predefined protocol and in accordance with PRISMA and MOOSE guidelines [16,17]. We searched MEDLINE and EMBASE up to 22 March 2019. The computer-based searches combined terms related to the exposure (e.g. “osteocalcin”) and outcomes (e.g. “cardiovascular disease”, “coronary heart disease”, “atherosclerosis”, “carotid intima-media thickness”, “mortality”) in humans, without any language restriction. Reference lists of selected studies and relevant reviews on the topic were manually scanned for additional publications missed by the original search. Full details on the search strategy are presented in Supplementary appendix 1.

Eligibility criteria

We systematically searched for observational cohort, case-control, or cross-sectional population-based studies that had reported on associations of circulating levels of OC (total OC, undercarboxylated OC, and carboxylated OC) with (i) CVD-related outcomes [composite CVD, coronary heart disease (CHD), stroke, congestive heart failure (CHF), or all-cause mortality]; and (ii) intermediate cardiovascular traits [carotid atherosclerosis, aortic atherosclerosis, coronary artery calcification (CAC), carotid intima-media thickness (CIMT), abdominal aortic calcification (AAC)].

Data extraction and quality assessment

The titles and abstracts of all articles identified by the broad literature search were assessed independently by two reviewers (SKK and SS). Studies that did not meet inclusion criteria were discarded. Full text of selected articles were retrieved and assessed to determine if they met the inclusion criteria. Those studies which met the inclusion criteria were included in the review and data was extracted independently by two reviewers (SKK and SS) using a standard data extraction form. The quality of the studies was assessed independently by both reviewers

When available, data were extracted on: publication date; study design; geographical location; population source; year of baseline survey; sample population; mean/median age at baseline; duration of follow-up (for cohort studies); type of OC; mean/median levels of OC; type of outcome; and reported risk estimates. In the case of multiple publications involving the same study, data were extracted from the most up-to-date study or study with the most comprehensive information was abstracted. We also corresponded with study investigators to provide missing information where relevant. Any discrepancies regarding eligibility of an article were discussed, and consensus reached with a third reviewer (K.K.). For cohort and case-control studies, study quality was assessed based on the nine-star Newcastle–Ottawa Scale (NOS) [18] using three pre-defined domains namely: selection of participants (population representativeness), comparability (adjustment for confounders), and ascertainment of outcomes of interest. The NOS assigns a maximum of four points for selection, two points for comparability, and three points for outcome. Nine points on the NOS reflects the highest study quality. For cross-sectional studies, quality was evaluated using the NOS modified for cross-sectional studies (Supplementary appendix 2(10)). A score of 8 reflected the highest study quality.

Statistical analysis

Mean differences for continuous outcomes and risk ratios for categorical outcomes were used as summary measures across studies. A narrative synthesis was performed for studies that could not be pooled. For data reported as medians, standard errors, ranges, and 95% confidence intervals (CIs), means and standard deviations were estimated using methods as described by Hozo and colleagues [19]. To enable a consistent approach to the meta-analysis, enhance comparability and interpretation of the findings, units of measurements were converted where appropriate and reported study-specific risk ratios (per-unit or standard deviation change, quintiles, or other groupings) were also transformed to involve comparisons between the top quartile and bottom quartile of each study population’s baseline distribution of OC levels, using standard statistical methods. [20,21] described previously. [22–24] To account for the effect of between-study heterogeneity anticipated, we used the inverse variance weighted method to combine summary measures using random-effects models [25]. We evaluated for publication bias using funnel plots and Egger’s regression symmetry tests [26]. All tests were two-tailed and p-values of .05 or less were considered significant. STATA release 14 (Stata Corp, College Station, Texas, USA) was used for all statistical analyses.

Results

Study identification and selection

165 potentially relevant citations were identified. 40 articles were selected for full text evaluation after an initial screen based on titles and abstracts. After detailed assessments, 8 articles were excluded because (i) the outcomes were not relevant to review (n = 5); (ii) the exposure was not relevant (n = 2); and (ii) one article used the same population sample as another study included in the review. The remaining 32 articles based on 33 unique observational studies met all inclusion criteria and were included in the review (Figure 1; Supplementary appendix 3).

Figure 1. Selection of studies included in the meta-analysis.

Study characteristics and study quality

Table 1 shows a summary of the key characteristics of the included. 21,021 unique participants were included in this review. However, not all studies provided relevant data that could be included in the quantitative synthesis. The majority of studies (n = 19) were conducted in Asia (China, Japan, and South Korea); with 10 in Europe (Austria, France, Germany, Italy, The Netherlands, and Spain); 2 in the Pacific (Australia); and 2 in North America (USA). The mean/median baseline age of participants ranged from 49 to 77 years. The majority of studies (n = 20) were cross-sectional in design; 9 prospective cohorts; 2 case-controls; 1 retrospective cohort and 1 prospective case-control. The average follow-up for cohort studies ranged from 0.5 to 10.0 years. For studies that provided relevant data on their recruitment processes, majority of studies (n = 22) reported recruiting patients from healthcare settings, with 7 studies reporting recruitment from a population register. There was considerable variability in study populations which included healthy and community-dwelling participants, post-menopausal women, and patients with pre-existing conditions such as type 2 diabetes, chronic kidney disease, essential hypertension and those at high cardiovascular risk. Among cohort, case-control, and cross-sectional studies, quality score ranged from 3 to 8.

Table 1. Characteristics of studies included in review.

Lead Author, Publication Date	Name of study or source of participants	Location	Study design	Baseline population	Year of baseline survey	Baseline age range (years)	% male	Follow up (years)	Outcome (s)	Total participants	Covariates adjusted for	Study quality
Jie, 1995	Population-based	Netherlands	Cross-sectional	Post-menopausal	1990	66.6*	0.0	NA	Aortic atheroclerosis	113	NA	4
Schleithoff, 2003	Health setting	Germany	Case-control	CHF patients and controls	2000–2001	66.2–71.5*	100.0	NA	CHF	42	NA	4
Montalcini, 2004	Hospital clinic	Italy	Cross-sectional	Post-menopausal women	NR	45–75	0.0	NA	Carotid atherosclerosis	157	Age, hypertension, hyperlipidemia, type 2 diabetes, obesity, smoker	6
Pennisi, 2004	Hospital setting	Italy	Case-control	Atherosclerotic plaques and/or vascular calcification	2003	62.7*	NR	NA	Carotid/femoral atherosclerosis	66	NA	5
Kanazawa, 2009	Hospital	Japan	Cross-sectional	Type 2 diabetes	NR	50–83 50–84	100.0 0.0	NA	CIMT	179 149	Age, duration of diabetes, BMI, serum creatinine, systolic blood pressure, LDL-C, HDL-C, HbA1C, and Brinkmann index	5
Okura, 2010	NR	Japan	Cross-sectional	Essential hypertension	NR	61.0*	58.0	NA	CAC	92	NR	4
Parker, 2010	SOF	USA	Cross-sectional	Community-dwelling postmenopausal	1986–1988	≥ 65	0.0	NA	AAC	363	Age, body mass index, systolic and diastolic blood pressure, diabetes, hypertension, smoking history, eGFR, CRP, estrogen use, intact parathyroid hormone, bone-specific alkaline phosphatase, and 25-hydroxyvitamin D; hip and spine bone mineral density	6
Zhang, 2010	NR	China	Cross-sectional	Subjects referred for coronary angiography	2005–2007	39–85	64.9	NA	CHD	461	Age, gender, BMI, smoking, alcohol, family history of CHD, LDL-C, HDL-C, TG, type 2 diabetes	6
Bao, 2011	Hospital	China	Cross-sectional	Coronary angiography patients	2008–2009	64.9*	100.0	NA	Coronary artherosclerosis index	181	Age, BMI, fasting insulin, HbA1c, HOMA-IR	6
Goliasch, 2011	Hospital setting	Austria	Case-control	MI patients with controls	2004–2008	≤40	88.7	NA	NA	302	NA	8
Kanazawa, 2011	NR	Japan	Prospective cohort	Type 2 diabetes	NR	64.5*	56.0	0.5	Plaque score	50	Duration of diabetes, Brinkman Index	5
Kim, 2012	SBMS	South Korea	Cross-sectional	Subjects referred for QCT	2007–2010	64.7*	0.0	NA	Aortic atheroclerosis	769	NA	4
Reyes-Garcia, 2012	Endocrinology Unit	Spain	Cross-sectional	Type 2 diabetes	2006–2007	57.8*	55.1	NA	CHD	78	Age, gender, BMI, hypt, smoking, DM duration, aortic calcifications, abnormal IMT, carotid plaques	6
Yeap, 2012	HIMS	Australia	Prospective cohort	Community-dwelling elderly men	2001–2004	77.0*	100.0	5.2	CVD, all-cause mortality	3,542	Age, WHR, conventional CVD risk factors	8
Confavreux, 2013	MINOS	France	Prospective cohort	Elderly men	1995–1996	51–85	100.0	10	AAC, all-cause mortality	774	Age, diabetes, hip to waist ratio, hypertension, smoking status, alcohol intake, serum phosphate, low HDL-cholesterol, AAC score, physical activity, and 25OHD	7
Lerchbaum, 2013	LURIC	Germany	Prospective cohort	Suspected CAD	1997–2000	58.72	100.0	7.7	CVD mortality, all-cause mortality	2,271	Age, BMI, smoking, WHR, CETP, hs-CRP, homocysteine, interleukin-6, HDL cholesterol, LDL cholesterol, triglycerides, prevalent metabolic syndrome	8
Ogawa-Furuya, 2013	Hospital	Japan	Cross-sectional	Type 2 diabetes	2005–2011	64.0*	54.1	NA	AAC	218	Age, BMI, creatinine, LDL-C, radial BMD, smoking, duration of DM, HbA1c, HOMA-IR, BAP, UNTx	6
Sheng, 2013	Shanghai downtown residential areas	China	Cross-sectional	Type 2 diabetes	NR	>50	46.8	NA	Carotid atherosclerosis, CIMT	817	Age, sex, smoking, alcohol, family history of diabetes, WC, HOMA-IR, BMI, WHR, TG, TC, HDL-C, LDL-C, CRP, treatments	6
Yamashita, 2013	Outpatient dialysis center	Japan	Prospective cohort	Maintenance haemodialysis patients	2005	20–85	61.1	5.0	CVD	126	Age, sex, time on dialysis, previous CVD, diabetes, serum Calcium x phosphate product, serum parathyroid hormone levels	7
Yang, 2013	SOS	China	Cross-sectional	Post-menopausal women	NR	41–78	0.0	NA	CIMT	1,319	Age, years since menopause, BMI, WC, SBP, DBP, HOMA-IR, TG, HDL-C, CRP, smoking, treatment, family history of CVD	7
Holvik, 2014	LASA	Netherlands	Prospective cohort	Elderly participants	1995–1996	65–88	48.8	4.1	CVD, ACS	832	Age, BMI, SBP, TG, TC, HDL-C, fructosamine	8
Lerchbaum, 2014	LURIC	Germany	Prospective cohort	Suspected CAD	1997–2000	58.72	0.0	7.7	CVD mortality, all-cause mortality	986	Age, BMI, menopause, prevalent CAD, HRT, vitamin K antagonist use	8
Ma, 2014	Shangai Changfeng study	China	Cross-sectional	Middle-aged and elderly	2009–2012	61.3*	100.0	NA	Carotid atherosclerosis	1,077	Age, FPG, PPG, BMI, WHR, smoking, SBP, DBP, TG, HDL-C, LDL-C, HOMA-IR, HOMA-B	7
Choi, 2015	Hospital	South Korea	Cross-sectional	Asymptomatic patients	2010–2012	53.5	70.4	NA	CAC	162	Age, body mass index, smoking (menopause), hypertension, diabetes, systolic blood pressure, HOMA2-IR, triglycerides, HDL cholesterol and lumbar BMD	6
Hwang, 2015	HPC	Korea	Prospective cohort	Middle-aged men	1997	40–78	100.0	8.7	Composite CVD, CHD, stroke	1,290	Age, BMI, current smoking, LDL-C, diabetes, hypertension, serum creatinine	7
Maser, 2015	Diabetes and Metabolic Research Center	USA	Cross-sectional	Type 2 diabetes	NR	≥ 18	42.0	NA	CAC	50	NA	5
Yeap, 2015	HIMS	Australia	Prospective cohort	Community-based men	2001–2004	70–89	100.0	NA	CHD, stroke	3,384	Education, smoking, BMI, WHR, hypertension, dyslipidemia, diabetes, creatinine, vitamin D, prevalent CVD, cancer	7
Zhang, 2015	Hospital	China	Cross-sectional	Non-dialysis with CKD	NR	50–75	62.9	NA	Carotid atherosclerosis	240	Age, sex, BMI, smoking history, MABP, eGFR, therapeutic medication use, and FBG, TC, TG, LDL-C, HDL-C, and hs-CRP levels	6
Kim, 2016	Hospital	South Korea	Cross-sectional	Patients who underwent CABG	2012–2013	60.7*	100.0	NA	Coronary artery stenosis	122	NA	5
Deng, 2018	Hospital	China	Cross-sectional	Patients who underwent screening for carotid atherosclerosis	2014–2016	40–60	100.0	NA	Carotid atherosclerosis	335	NA	3
Ling, 2018	Hospital	China	Cross-sectional	Post-menopausal women with CAD and controls	2015–2016	66.0*	0.0	NA	CHD, coronary atherosclerosis	227	Age, BMI, smoking, hypertension, diabetes, LDL-C, HDL-C, TG, eGFR, statin use and high-sensitivity CRP	5
Zhang, 2018	Hospital	China	Retrospective cohort	Coronary angiography patients	2008–2009	65.5*	66.4	4.4	CVD	247	Age, BMI, WC, SBP, DBP, FPG, 2hPG, HOMA-IR, TG, HDL-C, LDL-C, CRP, current smoker	6

AAC: abdominal aortic calcification; ACS: aortic calcification score; BMD: bone mineral density; BMI: body mass index; CABG: coronary artery bypass graft; CAC: coronary artery calcification; CAD: coronary artery disease; CHD: coronary heart disease; CHF: congestive heart failure; CKD: chronic kidney disease; CIMT: carotid artery intima-media thickness; CRP: C-reactive protein; CVD: cardiovascular disease; DBP: diastolic blood pressure; eGFR: estimated glomerular filtration rate; FPG: fasting plasma glucose; HDL-C: high-density lipoprotein cholesterol; HIMS: Health In Men Study; HOMA-B: homeostasis model assessment of beta cell function; HOMA-IR: homeostasis model assessment of insulin resistance; HPC: Health Promotion Center; LASA: Longitudinal Aging Study Amsterdam; LDL-C: low-density lipoprotein cholesterol; MI: myocardial infarction; NA: not applicable; NR: not reported; SBMS: Severance Bone Metabolism study; SBP: systolic blood pressure; SOF: Study of Osteoporotic Fractures; SOS: Shangai Obesity Study; TC: total cholesterol; TG: triglycerides; WC: waist circumference; WHR: waist-to-hip ratio.

Association of circulating osteocalcin with cardiovascular outcomes

Cohort analysis The pooled risk ratio of 5 cohort studies (3481 participants and 686 events) for composite CVD risk comparing individuals in the top versus bottom fourths of circulating total OC levels, adjusted for several established cardiovascular risk factors was 0.98 (95% CI 0.89, 1.08; p = .669) (Figure 2). The prospective corresponding risk ratios for CHD and stroke based on results of a single study were 1.06 (95% CI 0.38, 2.93) and 0.65 (95% CI 0.29, 1.46) respectively.

Figure 2. Association of circulating total osteocalcin with cardiovascular outcomes and all-cause mortality. CI: confidence interval (bars); OC: osteocalcin.

In pooled analysis of 2 prospective studies (1,760 participants and 378 events), the multivariate-adjusted risk ratio for all-cause mortality was 0.33 (95% CI 0.03, 3.51; p = .355) comparing individuals in the top versus bottom fourths of circulating total OC levels. Four studies could not be included in the meta-analysis because of differences in the exposure categories. In analysis adjusted for several conventional risk factors, Yeap and colleagues reported a U-shaped relationship of circulating total OC with CVD mortality and all-cause mortality in men – the risk being increased at both ends of the distribution of OC levels [27]. In a follow-up of the same study that evaluated CHD and stroke outcomes, total OC was not associated with MI or stroke events [15]. In the Ludwigshafen Risk and Cardiovascular Health (LURIC) prospective cohort study [28], findings reported in men showed a U-shaped association of total OC with fatal CVD and all-cause mortality - with increased risk at both ends of the distribution of OC levels. In a prospective case-control study of 102 MI patients and 200 control subjects, total OC was demonstrated to be associated with premature MI at one-year follow-up [29].

In the single prospective cohort study that evaluated the associations between undercarboxylated OC and cardiovascular endpoints, there was no evidence of an association of undercarboxylated OC with CHD or stroke [15].

Cross-sectional analysis Comparing individuals in the top versus bottom fourths of circulating total OC levels, the multivariate-adjusted risk ratio for CHD in pooled analysis of 3 cross-sectional studies (766 participants and 383 events) was 1.20 (95% CI 0.32, 4.50; p = .789).

Association of serum total osteocalcin with intermediate cardiovascular traits

Cohort analysis The pooled risk ratio for aortic calcification in pooled analysis of 2 prospective cohort studies was 0.87 (95% CI 0.76, 0.99; p = .030) comparing individuals in the top versus bottom fourths of circulating total OC levels. In a prospective cohort study that monitored changes in total OC and plaque score in Japanese patients with type 2 diabetes, changes in circulating total OC was inversely associated with changes in plaque score from baseline in analysis adjusted for atherosclerotic risk factors (β = −0.30; p = .047) [30].

Cross-sectional analysis In pooled analysis of 8 studies (3356 participants), the multivariate-adjusted risk ratio for vascular atherosclerosis or calcification when comparing individuals in the top versus bottom fourths of circulating total OC levels was 0.67 (95% CI 0.37, 1.22; p = .187) (Figure 3). The risk ratio for vascular atherosclerosis or calcification when comparing individuals in the top versus bottom fourths of circulating undercarboxylated OC levels in pooled analysis of 4 studies (762 participants) was 0.76 (95% CI 0.13, 4.63; p = .767) (Figure 3).

Figure 3. Associations of circulating total and undercarboxylated osteocalcin with intermediate cardiovascular traits. AAC: abdominal aortic calcification; CAC: coronary artery calcification; CI: confidence interval (bars); OC: osteocalcin.

Five studies could not be included in the meta-analysis. Kanazawa and colleagues in their assessment of the associations of serum total OC with atherosclerosis parameters in patients with type 2 diabetes, total OC was significantly and inversely correlated with CIMT in men (r = −0.181; p = .023), but not in women (r = 0.022; p = .803) [31]. In a sample of 817 men and postmenopausal women with type 2 diabetes, serum total OC was independently and inversely associated with CIMT (β = −0.181; r = −0.187; p < .001) [12]. Yang et al. also demonstrated a significant and inverse association between serum total OC and CIMT (β = −0.117; r = −0.107; p < .01) [13]. Kim and colleagues also reported total OC to be significantly and inversely correlated with aortic calcification as measured by the aortic calcium score (r = −0.238; p < .001) [11]. An independent and inverse correlation (β = −0.497; p = .003) was also demonstrated between serum total OC and coronary atherosclerosis index in the study by Bao et al. [32].

Circulating osteocalcin levels in patients with cardiovascular outcomes and traits compared with controls

The pooled random-effects mean difference across 9 studies showed significantly lower circulating levels of total OC −2.58 ng/ml (95% CI −3.85, −1.32; p < .001) in patients with cardiovascular conditions compared to subjects without cardiovascular conditions (Figure 4). In pooled analysis of 4 studies, there was no significant difference in circulating undercarboxylated OC levels comparing subjects with and without cardiovascular conditions. In a case-control study, circulating carboxylated OC level was significantly lower comparing subjects with and without CAD −0.50 ng/ml (95% CI −0.71, −0.29; p < .001).

Figure 4. Mean differences in circulating osteocalcin levels comparing subjects with cardiovascular conditions and their respective controls. CAC: coronary artery calcification; CAD: coronary artery disease; CHD: coronary heart disease; CHF: congestive heart failure; CI: confidence interval (bars).

Discussion

Summary of findings

Using a systematic and meta-analytical approach, we have summarized all available observational studies that have assessed the associations of circulating levels of OC (total, undercarboxylated, and carboxylated OC) with clinical CVD endpoints and intermediate CVD traits, in an attempt to address the uncertainties in the evidence. In the analysis of hard CVD outcomes, pooled analysis of prospective data showed no evidence of statistically significant associations of total OC with composite CVD, CHD, or stroke endpoints. However, two prospective studies conducted in men demonstrated a U-shaped association of total OC with fatal CVD and all-cause mortality [27,28]. Both low and high levels of total OC were associated with an increased risk of these outcomes. In the evaluation of intermediate cardiovascular traits, prospective data showed significant inverse associations between total OC and traits such as aortic calcification and plaque score. Though the evidence from cross-sectional studies was inconsistent, majority of studies reported inverse, independent, and significant correlations between total OC and outcomes such as CIMT, aortic calcification, and coronary calcification or atherosclerosis. Circulating total OC was significantly lower in patients with cardiovascular conditions compared with those without these conditions. The data was limited for undercarboxylated and carboxylated OC and generally no significant evidence of associations were demonstrated with the outcomes assessed.

Interpretation of findings

The current evidence does not conclusively support a role of circulating OC (particularly total OC) in the pathophysiology of adverse cardiovascular lesions as well as CVD risk. Though some of the findings show that reduced levels of circulating total OC are associated with greater pathological cardiovascular changes, there was evidence to suggest that increased levels of total OC might also be associated with increased risk of these adverse outcomes. Beyond its well-established pro-osteoblastic functions [33], OC has been demonstrated to have endocrine functions [6]. Consistent evidence shows that reduced circulating levels of OC is associated with increased risk of adverse metabolic outcomes such as type 2 diabetes and metabolic syndrome (MetS), which is via its role in influencing insulin secretion, glucose metabolism, insulin sensitivity, fat mass, beta cell proliferation, and energy expenditure [6,34,35]. Though the mechanistic evidence has mostly involved the use of animal models, recent epidemiological and genetic studies suggest the multiple aspects of the biology of OC are similar for both humans and rodents [36,37]. However, the mechanistic evidence linking OC with atherosclerotic CVD is unclear compared with metabolic conditions such as type 2 diabetes; a number of pathways have been proposed. Conditions such as insulin resistance, metabolic syndrome, non-alcoholic fatty liver disease, and type 2 diabetes, which are closed linked to circulating OC, accelerate the progression of atherosclerotic lesions and are strongly linked to the development of atherosclerosis [12]. Osteocalcin may be involved in the calcification process at arterial and valvular sites, leading to reduced elasticity and compliance of the vasculature [38,39]. Osteocalcin may also play a direct role in the atherosclerotic process, as osteoclast-like cells have been identified in atherosclerotic lesions [39]. Whether the different forms of OC may play different roles in the pathophysiology of adverse cardiovascular outcomes is not clear, as the evidence has mostly been limited to circulating total OC. However, there is evidence from animal models suggesting thatundercarboxylated OC, the more active form of OC, may be of more importance in regulating glucose and energy metabolism [6,34,40] and therefore could be implicated in the development of adverse metabolic outcomes. This is not well established as it is not even certain if undercarboxylated OC might be the active form in humans [36].

Implications of findings

The current findings provide further insight on the role of OC beyond its pro-osteoblastic properties as well as its role in glucose and energy metabolism. The clinical use of circulating OC assays has mainly been for monitoring and assessing the effectiveness of antiresorptive therapy in patients with bone conditions characterised by elevated OC levels such as osteoporosis and Paget’s disease [41].

Though some of the findings were inconsistent which impeded meaningful interpretation, the overall results of the quantitative synthesis of the available data does not underscore a potentially protective role of increased circulating total OC levels on the risk of atherosclerotic lesions and CVD outcomes. The inconsistencies in the findings could be related to the small sample sizes employed by majority of studies, different populations used, different OC assay methods employed, and the fact that circulating OC levels are influenced by ethnicity, gender and menopausal status [42]. There is some suggestion from our results that there may be a more evident association in males. In addition to highlighting the lack of consistent evidence on the cardiovascular effects of OC, this review has also identified several gaps in the literature. Compared with the number of cross-sectional study designs (n = 20), only 10 observational cohort studies were identified to have evaluated the associations and these were limited by small sample sizes and populations with pre-existing disease. Furthermore, the majority of studies used assays for total OC, and therefore the role of other forms of OC is far from clear. This observation could be attributed to the fact that circulating total OC can be conveniently measured in large studies using automated immunoassays [43], while assays for undercarboxylated OC are more cumbersome, labour intensive, or less precise [44]. Large-scale prospective studies conducted in general population settings are needed to address the existing research gaps. Thus far, the potential of strategies that raise concentrations of circulating total OC and their value in the prevention or treatment of vascular diseases is not convincing enough. This is a topic which deserves further investigation.

Strengths and limitations

The strengths and limitations of this systematic meta-analysis merit careful consideration. The notable strengths included the comprehensive search strategy which yielded several published studies on the topic. Overall, this review involved over 21,000 participants and evaluated a wide-range of atherosclerotic and CVD outcomes and their relationships with the different forms of circulating OC. We were able to transform reported risk estimates from majority of contributing studies to a consistent comparison (mean differences and top versus bottom fourths) to allow a consistent combination of estimates across studies, therefore obtaining a reliable estimate of the magnitude of the association and enhancing interpretation of the overall findings. We also conducted a detailed quality assessment of eligible studies. Limitations included the inability to fully examine the impact of adjustment for potential confounding factors, because the review was based on variably adjusted data reported in the published literature; heterogeneity could not be explored because of the limited number of studies available for pooling; and some outcomes and risk estimates were inconsistent between studies and therefore could not be pooled.

In conclusion, aggregate observational data do not generally support inverse and independent associations of circulating total OC with risk of atherosclerotic outcomes and CVD endpoints; however, the data were mostly based on cross-sectional evaluations. There are gaps in the existing literature and large-scale prospective cohort studies are needed to explore the nature and potential magnitude of any association of circulating OC with cardiovascular outcomes.

Conflicts of interest/disclosures

KK has acted as a consultant and speaker for Novartis, Novo Nordisk, Sanofi-Aventis, Lilly and Merck Sharp & Dohme. He has received grants in support of investigator and investigator-initiated trials from Novartis, Novo Nordisk, Sanofi-Aventis, Lilly, Pfizer, Boehringer Ingelheim and Merck Sharp & Dohme.

SS has acted as consultant, advisory board member and speaker for Novo Nordisk, Amgen, Sanofi-Aventis, Lilly, Merck Sharp & Dohme, Boehringer Ingelheim, AstraZeneca and Janssen, NAPP and Novartis. He has received research grants Jansen.

SKK has nothing to declare.

Acknowledgements

We thank Rebeca Reyes-Garcia, MD, PhD, Bone Metabolic Unit (RETICEF), Endocrinology Division, Hospital Universitario San Cecilio, Av. Dr. Oloriz 16, 18012 Granada, Spain for readily providing data on request.

Additional information

Funding

S. Seidu, and K. Khunti acknowledge support from the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care – East Midlands (NIHR CLAHRC – EM), the Leicester Clinical Trials Unit and the NIHR Leicester-Loughborough Diet, Lifestyle and Physical Activity Biomedical Research Unit, which is a partnership between University Hospitals of Leicester NHS Trust, Loughborough University and the University of Leicester. S.K. Kunutsor is supported by the NIHR Biomedical Research Centre at the University Hospitals Bristol NHS Foundation Trust and the University of Bristol. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.