Association of osteocalcin, osteoprotegerin, and osteopontin with cardiovascular disease and retinopathy in type 2 diabetes

Novel biomarkers of vascular disease in diabetes could help identify new mechanistic pathways. Osteocalcin, osteoprotegerin, and osteopontin are key molecules involved in bone and vascular calcification processes, both of which are compromised in diabetes. We aimed to evaluate possible associations of osteocalcin, osteoprotegerin, and osteopontin with cardiovascular disease (CVD) and diabetic retinopathy (DR) among people with type 2 diabetes (T2D).

sustained by different molecular and cellular mechanisms, 13 we also hypothesised that different osteokines, based on their different biology, may be differentially associated with these two complications of diabetes. Accordingly, we analysed data from the 'Sapienza University Mortality and Morbidity Event Rate (SUMMER) study in diabetes' to evaluate possible associations between circulating bone metabolism markers and macro-and micro-vascular diabetic complications.

| Study design and population
This is a cross-sectional analysis performed on the baseline data and serum samples collected within the SUMMER study (registered at ClinicalTrials.gov, NCT02311244). This was an observational, prospective, collaborative study aimed at identifying new predictors of all-cause and cardiovascular mortality and morbidity in patients with adult-onset diabetes. 14 SUMMER enroled consecutive individuals, regardless of being newly or previously diagnosed with T2D, from July 2014 to December 2018 in 10 Italian diabetes clinics. Exclusion criteria included severe psychiatric illnesses, end-stage renal disease, renal dialysis, hepatic cirrhosis, active cancer of any type, and chronic treatment with corticosteroids.
At the time of enrolment, demographic data, medical history, and biochemical information as well as blood samples were collected from each participant. Serum samples were subsequently aliquoted and stored at −80°C prior to assay.
At the time the population for this study was selected (March 2019), 2486 SUMMER participants had been screened for the presence of glutamic acid decarboxylase antibodies (GADA). We aimed to analyse a subgroup comprising the first 850 GADA-negative SUMMER participants. However, two participants were excluded because technical issues meant their samples were not available for the measurement of osteocalcin, osteoprotegerin and osteopontin, resulting in a final study cohort of 848 individuals.

| Laboratory assays and data extraction
Commercial BioVendor ELISA kits were used to assess serum concentrations of osteocalcin (Brno, Czech Republic; Catalogue # RIS002R), osteoprotegerin (Brno, Czech Republic; Catalogue # RD194003200), and osteopontin (Brno, Czech Republic; Catalogue # RD191446200R). The osteocalcin inter-assay coefficient of variation is 4.6% with an intra-assay coefficient of variation being 3.9%, the osteoprotegerin inter-assay coefficient of variation is 5.8% with an intra-assay coefficient of variation being 3.5%, and the osteopontin inter-assay coefficient of variation is 3.9% with an intra-assay coefficient of variation being 5.7%.
The following data were retrieved from the SUMMER study database for the 848 selected participants:

| Outcomes
The two outcomes we analysed were history of CVD (defined as history of acute myocardial infarction, coronary artery revascularisation, or stroke) and prevalent DR (defined as a fundus oculi examination performed by an expert ophthalmologist documenting any grade of retinopathy or maculopathy in either eye). Since osteoprotegerin and osteocalcin are excreted in urine, diabetic nephropathy was not investigated as a complication, but eGFR was included as a possible confounder.

| Statistical analysis
Descriptive statistics are presented as frequencies and proportions for categorical variables, and as mean and standard deviation or median and 25th-75th percentiles, as appropriate, for continuous variables. The normal distribution assumption for continuous variables was tested with the Shapiro-Wilk test. Continuous variables with a parametric distribution were compared between groups using Student's t test, and the Kruskal-Wallis test was used for non-parametric variables. Spearman rank test was used to test relationships between continuous variables. Categorical variables were compared with a χ 2 or Fisher's exact test as appropriate.
Standardized mean difference (SMD) between groups was also calculated. Logistic regression models with the study outcomes as dichotomous dependent variables and osteocalcin, osteoprotegerin or osteopontin as main exposures were used to estimate the odds ratios (OR) with 95% confidence intervals (CI) associated with a one standard deviation increase in the serum concentrations (natural log) of the main exposures after adjusting for potential confounders. Since the pathophysiology and the mechanisms of injury of CVD and DR differ, 13 we analysed CVD and DR as separate outcomes. Five separate models were evaluated: (1) adjusted for general confounders (sex, age, smoking habits and BMI); (2) adjusting as for model 1 plus diabetes-related variables The following non-parametric variables were logarithmically transformed (natural log) before entering the model: osteocalcin, osteoprotegerin, osteopontin, age, BMI, age at diagnosis, HbA 1c , uric acid, and eGFR.
Propensity score matching with Kernel matching was also used to test differences in osteokines levels between people with or without CVD or DR matched for unbalanced clinical features (see Supplementary Appendix for details). 16 Associations of each bone biomarker with the two study outcomes were tested at a two-sided alpha-level <0.05 after Bonferroni correction (i.e. after multiplying p-values for the number of tests).
Assuming a distribution of bone biomarkers in the population similar to our previous observations, 6,17 the study was >80% powered to detect a 15% difference in bone biomarkers between groups, at an alpha-level of 0.05. Stata/IC 12.1 (StataCorp) and Prism 9.0 (Graph-Pad Software) were used to perform the statistical analyses and produce graphical representations.

| Ethics
The study was performed in accord with the Declaration of Helsinki.
The study protocol was approved by the coordinating centre's Ethic  University Hospital. Participants signed a written informed consent to participate in the study. 14 3 | RESULTS  Osteoprotegerin levels were higher among females than males (p < 0.001), while no differences in osteocalcin and osteopontin levels by sex were found (Supplementary Table S1). Osteocalcin and osteoprotegerin levels tend to increase with age, age at diagnosis, disease duration, and lower eGFR (Supplementary Table S1). HbA 1c was directly related to osteoprotegerin concentrations, while inverse associations between BMI and osteocalcin and between triglycerides and osteocalcin were found (Supplementary Table S1).

| Cardiovascular disease
People with a history of CVD, compared with those without, were more frequently male, older at diabetes onset, with a longer diabetes duration at enrolment, and less frequently never smokers (

| Diabetic retinopathy
People with DR, compared with those without DR, were older with longer diabetes duration and younger age at diagnosis (

| DISCUSSION
This cross-sectional analysis performed on baseline SUMMER study data shows that, among people with T2D, serum osteocalcin concentrations are positively associated with CVD and that osteoprotegerin and osteopontin concentrations are positively associated with DR. More specifically, for each standard deviation increase in serum osteocalcin concentrations (natural log), we found 32% higher probability of having CVD. Additionally, each standard deviation increase in osteoprotegerin and osteopontin concentrations (natural log) was associated with 25% higher probability of having DR.
These results suggest that osteokines may be involved in pathways directly related to vascular disease, expanding their biological relevance.
Several previous studies have investigated the association of osteocalcin with CVD. 18 However, these studies invariably included a heterogeneous population with low proportion of people with T2D.
Since a complex relation between osteocalcin and diabetes has been described previously, 19,20 our study aimed at validating the association between osteocalcin and CVD in a large contemporary cohort comprising only people with non-autoimmune diabetes (n = 848). In  diabetes. 22 Accordingly, our study attempted to minimise the effect of confounders on the relationship between osteocalcin and CVD by focussing exclusively on a well-characterised cohort of people with T2D.
As with previous observations in people without diabetes 6 and those with type 1 and type 2 diabetes, 7,23 we report higher osteoprotegerin concentrations among SUMMER participants with CVD than their non-CVD counterparts. However, this association disappeared after adjusting for general confounders, suggesting that the relationship between osteoprotegerin and CVD in T2D may be mediated by common cardiovascular risk factors or therapies. On the contrary, we found an independent association between osteoprotegerin and DR, which is in line with previous smaller studies suggesting higher osteoprotegerin concentrations in people with diabetes with proliferative retinopathy than in those with non-proliferative or no retinopathy. 24 A proposed mechanism involves the dysregulation of the osteoprotegerin/RANKL/RANK pathway that leads to inflammation and angiogenesis in proliferative DR. 25 We also observed that a one standard deviation increase Strengths of this study include the multi-institutional contemporary cohort, the large sample size, the centralised serum analysis, and the exclusion of patients testing positive for GADA.
A limitation of our study is that, as with many previous studies investigating the effect of bone biomarkers, our analysis relied on a cross-sectional design that cannot establish a cause-and-effect relationship between bone biomarkers and clinical outcomes. A prospective evaluation of the longitudinal data from the SUMMER study F I G U R E 2 Forest plot of the regression models testing the associations of osteoprotegerin and osteopontin with DR. Odds ratio (OR) and 95% confidence intervals (CI) are given for one standard deviation increase in osteoprotegerin and osteopontin concentrations (natural log). Osteocalcin was not included in the models because it was not associated with DR in the pairwise comparison (see text