Coronary calcification and bone microarchitecture by high-resolution peripheral quantitative computed tomography from the São Paulo Ageing and Health (SPAH) Study

Epidemiological studies reveal a link between osteoporosis and the risk of ischemic cardiovascular disease. We illustrate an association between coronary calcification and bone microarchitecture in older adults based on the SPAH study. This cross-sectional research comprised 256 individuals subjected to cardiac coronary computed tomography angiography (CCTA) for coronary artery calcification (CAC), high-resolution peripheral quantitative computed tomography (HR-pQCT) at the tibia and radius with standardized z score parameters, and dual-energy X-ray absorptiometry (DXA) to evaluate bone status. We used Student’s t test and the Mann–Whitney and Chi-squared tests for comparison of basal measurements. Association analysis was performed using the Poisson regression model with adjustment for CAC and sex. Multivariate analysis revealed different bone variables for predicting CAC in DXA and HR-pQCT scenarios. Although most of the bone parameters are related to vascular calcification, only cortical porosity (Ct.Po) remained uniform by HR-pQCT. Results for were as follows: the tibia—women (exp β = 1.12 (95% CI 1.10–1.13, p < 0.001) and men (exp β = 1.44, 95% CI 1.42–1.46, p < 0.001); the radius—women (exp β = 1.07 (95% CI 1.07–1.08, p < 0.001) and men (exp β = 1.33 (95% CI 1.30–1.37, p < 0.001). These findings suggest an inverse relationship between CAC and cortical bone content, as assessed by HR-pQCT, with higher coronary calcification in individuals older than 65 years.

Atherosclerotic disease and osteoporosis are highly prevalent pathologies in older adults 1,2 . In addition to sharing risk factors 3 , evidence suggests that the diseases have common pathophysiological mechanisms [4][5][6] .
Furthermore, bone loss is associated with cardiovascular disease severity at the time of diagnosis, and the degree of vascular calcification is a significant predictor of bone loss and vertebral and nonvertebral fractures 7,8 .
In this scenario, coronary artery calcification (CAC), as calculated by the Agatston score 9 , has proven to be a powerful tool in personalized cardiovascular event risk assessment 10  www.nature.com/scientificreports/ All participants agreed to and signed the written informed consent form prior to enrollment into the study, thereafter, they were asked to fill out a questionnaire, which captured personal pathological background, current medications, and health behaviors. Specially trained examiners performed anthropometric measurements. Height and weight were measured using a digital scale with a stadiometer (Mettler Toledo, mod. 2096PP, Toledo do Brasil Indústria de Balanças Ltda); all measurements were taken closest to 0.1 kg and 0.1 cm, and each participant's weight in kilograms was divided by the square of their height in meters to calculate body mass index (BMI). Waist circumference measurement was also performed.
Systemic arterial hypertension (SHT) was defined as documented history or treatment with antihypertensive medications; diabetes mellitus (DM) was defined as physician diagnosis and/or use of insulin or oral hypoglycemic agents and dyslipidemia (DLP) as a known but untreated disease or by current use of hypolipidemic drugs. Metabolic syndrome was defined using the recommendations of the third NCEP panel report 27 . A positive smoking history was considered current or previous smoking. A history of acute myocardial infarction (AMI), angina, stroke, carotid disease, osteoporosis, and coffee consumption was self-reported. The degree of physical activity was classified as either low (no housework), moderate (housework, gardening, nonregular walking), or high (regular physical activity outside the usual routine, e.g., walking or dancing at least twice a week for 30 min 28 ). Coffee consumption data collection was based on a previously validated questionnaire 29 : participants were asked if they consumed at least one cup daily (50 mL = 1 cup). Alcohol intake was split into four groups: current drinker, intake in the last 3-4 years, never, or in a social manner. Bone T-score (BT score) ≤ − 2.5 was considered the smallest measurement acquired by DXA for the lumbar spine, total hip, or femoral neck. New fracture was selfreported and considered if present in the last 5 years after the initial entry-study date. To calculate creatinine clearance (CrCl), we used the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation 30 . Laboratory parameters. Blood samples were collected to assess the lipid profile, creatinine, glycemia, insulin, and biochemical bone metabolism profile (calcium, phosphorus, alkaline phosphatase, PTH, and 25-hydroxyvitamin D) and perform ultrasensitive PCR.
Bone remodeling markers were analyzed in 10 mL of fasting blood stored at − 80 °C. The following markers of serum bone metabolism were examined: CTX (C-terminal telopeptide type I collagen) and P1NP (aminoterminal propeptide of type I procollagen) by electrochemiluminescence (Elecsys systems, Roche Diagnostics, Mannheim, Germany). The intra-assay coefficients of variation (CVs) for CTX and P1NP were 2.5 and 2.2%, respectively.
Evaluation of coronary calcification. Image acquisition for coronary calcium evaluation was performed using a 320-detector CT system (Aquilion One ViSION, Toshiba Medical System, Japan), with cardiac synchronization using prospective triggering with a maximum temporal resolution of 125-250 ms as referenced in the literature 31 . The produced slice thickness of 0.5 mm was reconstructed to a 3 mm slice thickness for analysis, X-ray tube peak voltage was fixed at 120 kV, and the tube current was adjusted to the patient's body size, with a range of 250-450 mA. The adaptive iterative dose reduction 3-dimensional algorithm (AIDR 3D) was used for iterative reconstruction 32 , resulting in a radiation dose ranging from 2.6 to 4.0 mSv. CAC values were calculated using the Agatston score 9 at a dedicated workstation (Aquarius workstation, TeraRecon, Inc, San Mateo, CA). The nonenhanced scan protocol was used to analyze the calcium score evaluation. The images were evaluated by one specialist medical doctor with at least 2 years of specific training who was unaware of the patient's clinical and tomographic information.
Evaluation of the distal tibia and radius using HR-pQCT. The distal segment of the tibia and radius were evaluated. The measurement included 110 slices, corresponding to a 9.02 mm section along the axial direction. For the distal radius and tibia, the first slice of the scan was positioned at fixed distances from the reference line of 9.5 and 22.5 mm, respectively 33 . Microarchitecture was measured using a three-dimensional HR-pQCT system (Xtreme CT; Scanco Medical AG). This system allowed simultaneous acquisition of a stack of parallel slices with a resolution of 82 µm (voxel size). The following imaging parameters were used: 60 kVp effective energy, 95 mA X-ray tube current, and 1536 × 1536 matrix. Quality control was monitored through daily examinations of a phantom (Phanton) containing hydroxyapatite rods (HA) incorporated in an equivalent resin fabric (QRM, Moehrendorf, Germany). Daily device calibration was performed to obtain images with good technical quality; parameters were kept stable throughout the study period.
All measurements were obtained using the same scanner, and one biomedical specialist analyzed the data according to guidelines 21 . The quality of the scan image was categorized as follows: perfect (G1), showing a slight (G2); moderate (G3); or unacceptable (G4) degree of movement artifact 34 . As the presence of movement artifacts was unacceptable, for the distal tibia, only 11 exams were repeated due to movement during the exam (talking, coughing).
In addition to standard morphological analysis, automatic segmentation software called "advanced cortical analysis" was used to determine cortical parameters 35,36 .
The HR-pQCT precision measurement from our laboratory, expressed as the coefficient of variation, ranged from 0.25 to 1.16% for the density parameters at the tibia site 37 .
Areal bone mineral density. Areal bone mineral density at the lumbar spine, femoral neck, and total hip were measured using DXA (Hologic QDR 4500, Inc. Bedford, MA, USA), and the results are expressed in g/cm 2 . www.nature.com/scientificreports/ The T-score was applied as a noninvasive tool for osteoporosis diagnosis and as an addendum to the original description by the WHO criteria definition threshold 30,38 . The same DXA scanner was used for all single exams, and the results were evaluated by one biomedical specialist. The least significant changes (LSCs) for BMD measurement were 0.033 g/cm 2 at the lumbar spine, 0.047 g/cm 2 at the femoral neck, and 0.039 g/cm 2 at the total hip. Statistical analysis. Statistical analysis was performed using STATA statistical software, version 12.0 (Stata Corp., College Station, USA). Continuous variables are expressed as medians and interquartile ranges or means ± standard deviations depending on the Kolmogorov-Smirnov test for normality. Continuous variables in the total sample were separated by sex and analyzed for differences with Student's t test if they showed a parametric pattern or the Mann-Whitney test if they did not (as with insulin and RCP values). Linkert scales such as for alcohol consumption and degree of physical activity were also evaluated by the Mann-Whitney test. In the case of binomial variables, a Chi-squared test was performed, and Fisher's adaptation was used as needed. The study's response variable was the calcium score, as expressed by its integer value. This variable was grouped by sex, assumed nonnegative integer values, with an asymmetric distribution on the left. Variables related to HR-pQCT were standardized in z scores using the Formula: z = (observed value − sample mean)/sample standard deviation 39 . T-scores were used to evaluate bone density compared to normal values in a young adult and formed part of the regression model. In univariate analysis, variables with p < 0.20 were used in the regression model with a stepwise variable selection procedure. Analysis of association was performed using a generalized linear regression model with a Poisson probability distribution because of the nonparametric distribution of calcium content. In addition, a layered stratification approach was taken: (1) by sex; and (2) by anatomical site. Thus, there were four groups (radio men, tibia men, radio women and tibia women). All tests were two-tailed, and p values < 0.05 were considered statistically significant.

Results
Anthropometric and clinical characteristics. The baseline patient characteristics regarding anthropometric and clinical variables are shown in Table 1. The mean age was 79.5 ± 4.24 years without sex differences. Women represented the largest group (65%). In them was observed a higher percentage of non-white women compared to men. Waist circumference was similar in both groups. BMI, hypertension, and diabetes were significantly higher in women, and consequently, women had a higher prevalence of metabolic syndrome. Stroke and osteoporosis prevalence was also higher in women. In contrast, alcohol consumption was higher in men. Different grades of physical activity from low to high were similar in both groups. Although the women had more fall histories and worse BT scores, they had similar new fracture events when compared to men. Regarding coronary calcium, median CAC values did not differ significantly between the groups (Fig. 2).
Laboratory characteristics. Lipid indices (CT, HDL, and LDL) and phosphorus were higher in women; carbohydrate metabolism, as represented by glucose and insulin, was similar for both groups. Although plasma creatinine was different between the groups, renal function, as measured by clearance, was similar for men and women ( Table 2).
Medication use. Table 3 shows that thiazide diuretics were the most used medication (and more frequently in women), followed by statins and beta-blockers. Regarding osteoporosis, specific treatment was used more by women than men: almost twice as much for calcium and 25(OH)D and with a higher proportion for bisphosphonate. Unfortunately, neither the specific kind of bisphosphonate nor the dosage of vitamin D was recorded.
Bone mineral density and bone microarchitecture based on CAC . In univariate analysis, total hip aBMD in both sexes and femoral neck aBMD in men were associated with lower CAC. Inversely, lumbar spine aBMD in both sex, and femoral neck aBMD in women, were associated with higher coronary calcification. Also, we found that T-score ≤ − 2.5 SD, was associated with higher coronary calcification in women; however, the association was the opposite in men.
Regarding HR-pQCT parameters, Ct.vBMD and Ct.Po in women at the tibia was negatively and positively associated with CAC, respectively. In men, more variables were associated with higher CAC, namely, Ct.vBMD, CT.Po, Tb.N, and Ct.Th. Overall, trabecular parameters were positively associated with CAC, with a 31-47% chance for each standard deviation of the z score. For the radius, only CT.Po had a positive association with CAC in women, and the other parameters followed an opposite trend. In men, Ct.vBMD, Ct.Th, and CT.Po were positively associated with CAC (Tables 4 and 5).
Multivariate analysis was adjusted for CAC and sex, the dependent variables were DXA and HR-pQCT parameters. Surprisingly, some tibia and radius DXA/HR-pQCT results in multivariate analysis were different between the sexes (Table 6), with paradoxical results, as described below.
Tibia measurements. In women, regarding DXA results, a greater quantity of CACs was associated with BT score ≤ − 2.5 SD. Concerning HR-pQCT, CT.Po showed a positive relationship. In contrast, Ct.vBMD had an inverse relation of 21% for each unit in the standard deviation of the z score.
In men, BT score ≤ − 2.5 SD was significantly associated with lower CACs. Concerning HR-pQCT parameters, only CT.Po was associated with higher CAC, with a positive 44% chance for a unit in the calcium score, Ct.vBMD presented a coefficient opposite to what was expected compared to women. www.nature.com/scientificreports/ Radius measurements. In women, BT score ≤ − 2.5 SD was also significantly associated with higher CACs.
About HR-pQCT parameters, CT.Po and Ct.vBMD were associated with greater coronary calcification. In men, vascular calcification was not related to osteoporosis based only BT score ≤ − 2.5 SD. Concerning CT.Po, it was congruent not only with the radius measurements in women but also with the tibial measurements. However, in men, Ct.vBMD showed an inverse relationship with coronary calcification (Table 6).

Discussion
In this cross-sectional study involving healthy older adults, we observed an association between CAC and HR-pQCT parameters. Our main findings suggest an inverse relationship between calcium content in the coronary arteries and bone microarchitecture.
Previous studies have revealed a relationship between CAC and bone density. Using DXA and multidetector tomography in postmenopausal women, Lee et al. 40 reported an associated odds ratio of 2.79, 95% CI 1.05-6.90) for osteoporosis, with a CAC score above 100 and higher odds of multivessel disease (8.91, 95% CI 1.93-41.22), as adjusted for cardiovascular risk factors and age. Nonetheless, that study did not address intake of bone metabolism drugs. www.nature.com/scientificreports/ Later, a large cohort study with an 8-year follow-up 41 found increased mortality risk and CAC related to BMD. The study used the trabecular thoracic spine as a reference for BMD levels, and after stratifying CAC into four categories, the observed association was more robust in postmenopausal women than men. Interestingly, the authors also explored ethnicity and found a weaker link for Asians in the CAC 1-100 category. We were unable to investigate this link because our population was not ethnically diverse.
Cortical and trabecular bone loss exhibited different behaviors in renal patients. Costa et al. 42 used vertebral tomography and observed that cortical bone loss in mid-term follow-up (24 months), unlike trabecular bone loss, was not associated with aortic calcification. One of the hypotheses to explain this lack of association was  www.nature.com/scientificreports/ that cortical bone has lower porosity and turnover rates than trabecular bone. The study found no effect of boneaffecting medications (diuretics, calcium base phosphate binders, and calcitriol).
Another study evaluated the relationship between BMD (assessed by HR-pQCT) and ischemic heart disease (myocardial infarction and angina) 43 . Interestingly, lower cortical volumetric BMD in the distal radius but not in the tibia was found; the study did not estimate coronary calcium content or medications, which can modify bone turnover.
Other studies, however, showed inverse results between vessel calcification and BMD. Aoyagi et al. 44 evaluated aortic calcification contrasting with different skeletal sites (calcaneus, proximal and distal radius) and failed to   www.nature.com/scientificreports/ observe a relationship in women. Their rationale revolved around hormonal imbalances and a lack of knowledge about vitamin D intake.
To the best of our knowledge, only one study in the literature has compared these two biological parameters using highly accurate imaging methods. However, this study was performed in a specific population with chronic kidney disease on dialysis, in whom lower trabecular and cortical parameters were associated with higher coronary calcification 24 .
Our results showed that higher CT.Po in both sexes and lower Ct.vBMD in women were associated with higher CACs. Furthermore, BT score ≤ − 2.5 SD, as evaluated by DXA, was associated with higher CACs in women.
Regarding HR-pQCT analysis, changes in cortical bone parameters were more greatly influenced by age and sex 45,46 . Mc Donald et al. 47 showed a decline in Ct. BMD with age in both women and men (14% and 17%, respectively), and the increase in CT.Po was three times higher in women than in men. Whitmarsh et al. 48 observed a significant decrease in Tb. BMD (38.1%) and Ct.Th (13%) using micro-CT on femoral head specimens from women undergoing hemiarthroplasty from the 9th to the 10th decade of life. The difference between the sexes likely involves the influence of testosterone supporting periosteal bone expansion and estrogen preventing bone loss 49 . This might explain why our findings were more evident in women.
Moreover, there is evidence of a different response to exogenous and endogenous factors, in addition to the metabolic and structural differences between cortical and trabecular tissue 50 . A study analyzing HR-pQCT reference curves in 450 healthy women aged 20-85 years observed an age-related increase in CT.Po in women, though trabecular parameters remained relatively stable 35 .
A hypothesis that might explain this antagonistic pattern is the existence of an intracortical remodeling transition area that increases porosity by the process of trabecularization of the internal cortex, which may cause underestimation in the measurement of cortical bone and overestimation in that of trabecular tissue 51 . This phenomenon is subject to ex vivo analysis using computed microtomography (μCT), HR-pQCT images from cadaveric specimens, or electron microscopy 52 . A study of postmortem femur specimens obtained from 24 white women aged 29-99 years (mean age 69 years) suggested that cortical thinning occurs owing to the coalescence of cortical pores rather than endocortical expansion 53 .
This behavior is likely amplified by interaction of the similar pathophysiological processes of the two illnesses. Because of the high accuracy of both methods in diagnosis and risk stratification, it may be relevant to continue investigating this association to obtain new prevention and treatment strategies that involve both pathologies, especially for females.
Furthermore, our results reinforce knowledge of the relationship between VC and decreased bone remodeling and have important clinical implications in the design of more specific diagnostic and treatment strategies for both pathologies.
We recognize the limitations of our study, as a cross-sectional design does not allow for inference of cause and effect in the association. The sample size might be considered small regarding the initial referential study, which confers less statistical power. The small number of participants in the group of men may explain those results, and the differences in bone-health medications were lower in this group.
In conclusion, we observed that cortical bone impairment at the tibia and radius, as assessed by HR-pQCT, is associated with increased coronary calcification in older people from the São Paulo Ageing and Health Study cohort. Otherwise, bone mineral density by DXA showed the opposite trend when related to coronary calcification depending on sex.