A novel quantitative approach to the measurement of abdominal aortic calcification as applied to the Canadian Multicenter Osteoporosis Study (CaMOS)

Bone 97 (2017) 201–208

Contents lists available at ScienceDirect

Bone journal homepage: www.elsevier.com/locate/bone

Full Length Article

A novel quantitative approach to the measurement of abdominal aortic calcification as applied to the Canadian Multicenter Osteoporosis Study (CaMOS)☆ Mark Grant a, Mandy E. Turner a, Jeremy Murray-Guenther a, Tassos Anastassiades b, Wilma M. Hopman c, Stephen M. Adams a, Paul Jeronimo a, Robert Nolan d, Michael A Adams a, Rachel M. Holden a,b,⁎ a

Department of Biomedical and Molecular Science, Queen's University, Kingston, ON, Canada Department of Medicine, Queen's University, Kingston, ON, Canada c Clinical Research Centre, Kingston General Hospital, Department of Public Health Sciences, Queen's University, Kingston, ON, Canada d Department of Radiology, Queen's University, Kingston, ON, Canada b

a r t i c l e

i n f o

Article history: Received 4 August 2016 Revised 18 January 2017 Accepted 18 January 2017 Available online 19 January 2017 Keywords: DXA Osteoporosis Atherosclerosis Vascular calcification General population studies Radiology Vascular calcification

a b s t r a c t Background and aims: Lateral spine radiographs provide an inexpensive resource for characterizing abdominal aortic calcification (AAC). A widely accepted measurement of AAC is the semi-quantitative technique generated by the Framingham Heart Study (F-AAC-24). We sought to develop an analytical method to quantify ACC (QAAC) on lateral spine radiographs and compare the finding to conventional subjective measurements. Methods: Severity of AAC was quantified by measuring pixel intensities in the user-defined region of the aorta with internal standardization to the vertebral endplates and background calibration to the density of the vertebral body. The association between bone mineral density (BMD) measured by dual energy X-ray absorptiometry (DXA) and AAC measured by QAAC, F-AAC-24 and a modified Framingham score (F-AAC-12) was determined in 110 participants of the Canadian Multicenter Osteoporosis Study (CaMOS). Results: The inter-observer reliability for the QAAC was slightly higher than with the visual and semi-quantitative Framingham method and the pseudo-colored images illustrate the potential to meaningfully resolve severity of calcification. There was a significant negative association between QAAC and BMD measures of the hip and spine. This association remained significant after adjustment for age, sex, estimated glomerular filtration rate, phosphate and hypertension. Significant predictors of F-ACC-12 and 24 included age and hypertension. Conclusions: The QAAC is a reproducible approach to measuring AAC. Whether it is capable of monitoring subtle calcific changes over time requires further study. This technique could be applied to large studies that seek to determine the impact of interventions that modify bone density as a treatment for vascular calcification and cardiovascular disease in the general population. © 2017 Published by Elsevier Inc.

1. Introduction Vascular calcification and osteoporosis are age-related processes that coincide in individuals [1]. Many studies over the past two decades have demonstrated an inverse relationship between bone mineral density (BMD) and measures of vascular calcification [2–9]. In the Framingham Heart Study, women with the greatest loss of BMD over 25 years also demonstrated the most severe progression of abdominal aortic calcification (AAC) measured by a semi-quantitative visual technique [10]. ☆ Kingston Cohort of the CaMOS study. ⁎ Corresponding author at: 3048C Etherington Hall, Queen's University, Kingston, ON K7L 3V6, Canada. E-mail address: [email protected] (R.M. Holden).

http://dx.doi.org/10.1016/j.bone.2017.01.018 8756-3282/© 2017 Published by Elsevier Inc.

Vascular calcification is an actively regulated process that occurs similarly to ossification and results in the pathological deposition of mineral in the intima and media of arteries [11]. Calcification of atherosclerotic plaque in the intima results in vessel occlusion whilst calcification of the medial layer is preferential to the elastic lamina and causes vascular stiffening [3]. Although many risk factors for osteoporosis and arterial calcification are shared (e.g. age, sex), the pathogenesis and mechanisms underlying bone loss and progressive vascular calcification may have biochemical and biological links [12,13]. Lateral spine radiographs provide a widely available and inexpensive resource for characterizing AAC. The most widely accepted measurement of AAC is a semi-quantitative technique generated by the Framingham Heart Study (F-AAC-24) [14]. This visual approach quantifies calcific deposits in the region of the aorta at the level of the first through the fourth lumbar vertebrae. A calcific deposit is considered

202

M. Grant et al. / Bone 97 (2017) 201–208

present if the densities are visible to the eye in the area parallel to the lumbar spine and is then graded on a 0–3 scale depending on the length of the lesion. The focus of this research was to develop a quantitative measurement technique (QAAC) with the objective of characterizing AAC severity. The technique relies on the relationship between density and brightness on x-ray images. Our primary hypothesis was that by measuring the pixel intensity within the region of the abdominal aorta relative to the vertebral body and endplates, the QAAC would quantitatively report AAC. The Canadian Multicenter Osteoporosis Study (CaMOS) is an epidemiological study of risk factors for osteoporosis in Canadians [15–17]. Radiographic and biochemical data at year 10 of this ongoing project from patients enrolled at the Kingston Study Centre, Ontario, Canada were used to examine the relationship between BMD and vascular calcification using QAAC methodology and compared to results obtained using the established semi-quantitative Framingham method [14]. We also hypothesized that the data derived using the QAAC technique, compared to the Framingham technique, would be more reproducible and would be inversely correlated with BMD in participants of the CaMOS study. Therefore, the objectives of the present study were (1) to develop an analytical method to quantify ACC on lateral spine radiographs (QAAC); (2) to compare the intra- and inter-observer reliability of the QAAC and Framingham scoring methods; (3) to determine the association between BMD and AAC measured by QAAC and Framingham and (4) determine the association between AAC scores measured by QAAC and F-AAC24 and demographic and laboratory variables in subjects from the Kingston Study Center of the CaMOS study.

2. Materials and methods 2.1. Patients This study was based on data from the Canadian Multi-centre Osteoporosis Study (CaMOS). Full details regarding the study population, assessments, and procedures have been previously reported [7,15,17–19]. Subjects of the Kingston Study Center of CaMOS were enrolled in the present study if (1) subjects had both a lateral spine X-ray and BMD measurement by dual X-ray absorptiometry (DXA) at year 10 (2); the X-ray had sufficient width to encompass the entire abdominal aorta; and (3) there was no evidence of any overlying radiopaque structures, such as urogenital lithiasis. After applying these inclusion criteria, 112 eligible subjects remained from the 193 subjects still enrolled in year 10 in the Kingston Cohort of the CaMOS study. The study was conducted according to the Declaration of Helsinki and was approved by the Queen's University Human Ethics Research Board.

2.2. Bone mineral density measurement BMD of the lumbar spine, femoral neck, and trochanter were measured by DXA using Hologic (Hologic, Inc., Bedford, MA, USA) or Lunar (Lunar GE Medical Systems, Madison, WI, USA) densitometers. BMD was expressed in g/cm [2].

2.3. Abdominal aortic calcification (AAC) measurement 2.3.1. Framingham method Calcification was assessed visually in the anterior and posterior wall as a proportion of the adjacent vertebrae (L1–L4) [14]. Severity scoring was as follows: 0 = no calcification; 1 = lesion length b one-third; 2 = lesion length between one-third and two-thirds; 3 = lesion length N two-thirds.

2.3.2. QAAC technique 2.3.2.1. Scanning the abdominal aortic region. The severity of AAC was quantified by measuring pixel intensities using a defined region of interest (ROI) that automatically exported all data points along its length as it moved horizontally across the user-defined region of the aorta (Image Pro 6.0 software) (Fig. 1). Without rotating the image, the ROI was aligned parallel to the anterior surface of L3. The length of the ROI was determined by the reader extending it from the superior surface of L3 to just above the superior surface of L4. The ROI ran posterior to anterior across the image and a pre-determined distance of 300 pixels was automatically collected. The QAAC technique is limited to the L3 and L4 aortic regions due to frequent interference by the 11th and 12th rib at L1 and L2. The ROI was then pivoted from L3 to L4 and, in doing so, created a complete but non-overlapping data set (Fig. 2). For the purposes of comparison to the Framingham method, the QAAC was compared to the F-AAC-24 as well as a modified version of the Framingham technique, F-AAC-12, which was generated by limiting the scoring to the L3 and L4 regions. Internal Standardization: All pixel intensities originating from the abdominal aortic region were measured relative to the vertebral endplates of the adjacent vertebrae allowing it to serve as the internal standard for the technique and to minimize the influence of image quality on the data set according to the equation: Fraction of regional standard 0 Pixel intensity′ x : ¼ Average maximal pixel intensity of vertebral endplates ðLxÞ

2.3.2.2. Background calibration. All planar x-ray images possess a baseline brightness that is present throughout the entire image. Since the QAAC technique depends on brightness, this background was corrected for. However, unlike the regional standards, which used only the vertebral endplates, background calibration considered the entire vertebral body. The pixel intensities that resulted from scanning the entire vertebral body demonstrated the range of brightness representative of calcification for that particular region of the image. A cut-off value was determined, such that any pixel intensity in the lower 10th percentile was not considered in the calcification measurement. 2.3.2.3. Scoring. The remaining pixels of the vertebral body were organized in a frequency distribution of pixel intensities, generating 95 bins which defined calcified tissue for that image. A ninety-five binning strategy was chosen over a conventional 256 binning strategy typically used in quantification of calcification by CT to ensure that at least 70% of subjects had N150 pixels per bin, providing reasonable resolution to achieve our aims. The pixel intensities collected from the aortic region were sorted into the above mentioned bins. Only values equal to or greater than the cut-off value fell within the range of brightness that represents calcified tissue, and contribute to the calcification score. To account for the area and the extent of calcified tissues, the relative contribution of each pixel depends on its brightness. Multiplying the number of pixels within a bin by that particular bin number produced a linearly weighted score. For example, the pixels in the first bin, defined as the lowest amount of calcification, were assigned one point per pixel, the pixels in the second bin were assigned two points per pixel, the pixels in the third bin were assigned three points per pixel, and so on until bin 95. The linearly weighted score was divided by the number of pixels in that particular region to generate a scaled score. This prevented larger sections from scoring higher or conversely smaller sections scoring lower. With this scoring method, a finite value was generated for each section of the abdominal aorta corresponding to L3, and L4. In order to generate one value that could describe the overall aortic calcification, the scores from L3 and L4 were summed.

M. Grant et al. / Bone 97 (2017) 201–208

203

Fig. 1. AAC is quantified by measuring pixel intensities. (A) The superior and (B) inferior endplates are individually outlined by the user as to determine the height of the line profile tool and the distance it will travel. An average of the maximal pixel intensities from each horizontal position is calculated to represent the regional standard. (C) The user traces the lumbar vertebrae of interest with a rectangular region of interest (ROI). The ROI defines the height and length by which the line profile tool will run. All data is collected and exported to Microsoft Excel for processing. (D) The user defines a ROI tool anterior to the lumbar vertebrae adjacent the region of interest. The tool will scan across the abdominal aortic region, collecting all pixel intensities and exporting them to Microsoft Excel for further processing.

2.4. Validation of QAAC 2.4.1. Pseudo-colored images In order to demonstrate the regions of the image that the program considered calcified, a visualization method was created that involved overlaying a pseudo coloring of the image. All the pixels that were recorded in the lowest bin were colored blue and the highest populated bin was colored red with the RGB spectrum coloring the bins in between. Each region in the image has a different pixel distribution and therefore different definition of calcification. Therefore, the pseudo coloring was applied to each region. Then these images were overlaid on top of the original image keeping only the region that was analyzed pseudo colored for each overlaid image (Image Pro 6.0 and Adobe Photoshop) to provide a visual representation of pixels that are considered calcified (Fig. 3). 2.5. Reliability study Inter-observer reliability was assessed by having two blinded readers analyze the same 50 images for severity of AAC using both the semi-quantitative technique, and the Framingham technique. In addition, one blinded reader analyzed the same set of 50 images twice, using both techniques, to assess their respective intra-observer reliabilities. 2.6. Evaluation of bias associated with internal standardization method Fig. 2. Internal standardization. The superior and inferior endplates of each L3 and L4 vertebrae serve as internal standards for the abdominal aortic region adjacent. Areas highlighted represent the region of aorta that is eventually scanned by the ROI tool. The colors correspond to the regional standards that define each section. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Since bone radio-density may be a surrogate indicator of BMD, we sought to check that the internal standardization methods and the background elimination process did not bias the assessment of QACC. For each patient the average maximal pixel intensity for each vertebral

204

M. Grant et al. / Bone 97 (2017) 201–208

Fig. 3. Pseudo-coloring of digital images highlights areas of calcification. Digital x-ray images before (top panel) and after (bottom panel) pseudo-coloring applied in subjects with minimal (A–B), moderate (C–D) and severe (E–F) aortic calcification. Areas considered to be calcified by the quantitative technique are represented by a polychromatic spectrum, ranging from blue, the least severe, to red, the most severe. Intermediate calcification is represented by green (low-to-moderate) and yellow (high-to-moderate). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

endplate (L3 and L4) and the 10th percentile of each vertebral body was plotted against BMD (both spine and hip) and Spearman correlations were calculated. 2.7. Statistical analysis All data were analyzed using IBM SPSS software (version 20; Armonk, NY). Descriptive statistics (mean ± SD for continuous data and frequency with percentage for categorical values) were generated for all variables. The Spearman correlation coefficient was used to evaluate the bivariate (unadjusted) association of QAAC and F-AAC-12 and F-AAC-24 with all variables that were continuous in nature; the independent samples t-test or Mann-Whitney U was used for variables that were dichotomous. Multiple linear regression models were then estimated to examine the association between BMD and QAAC (model A), F-AAC-12 (model B) and F-AAC-24 (model C). Variables that were significantly associated with QAAC, F-AAC-12 and F-AAC-24 in the bivariate analysis (p b 0.1) were selected as covariates in the models. Because of co-linearity between BMD measures at different sites, the measures were sequentially added to the regression analysis and the best model was selected. Intra-class correlation coefficient (ICC), as calculated by SPSS, was used to assess the reproducibility of both techniques. 3. Results Fifty images were analyzed by two readers to determine the intraclass correlation coefficients (ICC). Both techniques demonstrated excellent agreement between and within readers. The ICC was marginally higher for the QAAC (Table 1). The Spearman correlation coefficient

between FAAC-12 and FAAC-24 was r = 0.97. Pseudo-colored images were generated for low, medium, and high severity of AAC quantified by the QAAC (Fig. 3). There was no correlation between the average maximal endplate pixel value, the internal standardization value, and lumbar BMD at either the L3 (r = −0.04, p = 0.7) or L4 (r = −0.18, p = 0.09) vertebrae. Similarly, there was no correlation between the 10th percentile of the vertebral body, the background cut off value, and lumbar BMD at either the L3 (r = 0.05, p = 0.6) or L4 (r = −0.14, p = 0.2) vertebrae. Fig. 4 demonstrates these relationships at L3 and L4. The characteristics of the CaMOS participants in this study are presented in Table 2. Of the 112 participants, 79 were female and 11 had diabetes. The cohort was elderly with a mean (SD) age of 68.9 (7.4) years. All but two participants were Caucasian; two were of Asian descent. The prevalence of ACC (score N 0) measured by the Framingham technique was 54%. The median [IQR] score by the QACC technique was 7.3 [4.7, 10.6].

Table 1 Intra-class correlation coefficients (ICC) for intra- and inter-rater agreement for severity of AAC measured by Framingham technique (FAAC) and Quantitative technique (QACC). FAAC

QACC

n

Inter-rater agreement –L2–L4 sum –L3–L4 sum

0.94 0.81

0.96 0.96

39 50

Intra-rater agreement –L2–L4 sum –L3–L4 sum

0.92 0.90

0.98 0.98

39 50

M. Grant et al. / Bone 97 (2017) 201–208

205

Fig. 4. Evaluation of bias. There was no association between the endplate average pixel intensity (internal standardization measure) and lumbar total BMD measured by DXA at either L3 (A) or L4 (B). There was no association between the background calibration (10th percentile of each vertebral body) and lumbar total BMD measured by DXA at either L3 (C) or L4 (D).

The bivariate associations between measures of calcification (F-AAC12, F-AAC-24 and QAAC) and demographic and laboratory variables as well as BMD of the lumbar spine and hip are presented in Table 2. Correlations involving the total scores (lumbar, hip) and their respective t-scores are not necessarily identical as both men and women are included in this analysis. There was a significant negative correlation between higher QAAC score and lower BMD measures of the spine and hip (both T-score and absolute score). Females and those with higher phosphate and lower eGFR had significantly higher QAAC scores. There was a weak correlation between hip total T-score and F-AAC-12 and F-AAC-24 scores but no association with the other BMD measures. Significant predictors of higher F-AAC-12 and F-AAC-24 scores included age, lower eGFR and higher homeostasis model assessment (HOMA) index (F-ACC-12 only). Multiple linear regression modeling revealed significant associations between higher QAAC and lower DXA Hip and lumbar spine T-scores after adjustment for sex, age, eGFR, phosphate and hypertension (Table 3A and B). Female sex and age (lumbar spine only) were also significant predictors of QAAC after adjustment. Independent predictors of F-ACC-12 and 24 included age and hypertension only (Table 3C and D). 4. Discussion Better quantification of the progression of vascular calcification could provide an important rationale for timing the introduction of interventions, determining success of treatments and in anticipating premature morbidity and mortality in populations. In this study, a new index for measuring vascular calcification was assessed, the QAAC, and compared to the current standard Framingham method. The interobserver reliability for the QAAC was slightly higher than with the visual, semi-quantitative Framingham method but the pseudo-colored images illustrate the potential of the quantitative technique to meaningfully resolve the continuum of severity of calcification. In this relatively small group of subjects, there was a statistically significant inverse correlation

between the QAAC and measures of hip and lumbar spine BMD that remained significant after adjustment for age, sex, hypertension, eGFR and phosphate. Whether this finding is clinically relevant requires further study however lumbar spine x-rays may offer a safe, affordable, and widely available modality for the assessment of AAC. Studies using the Framingham method have reported inter- and intra-rater agreement scores ranging from 0.71–0.92, and 0.93–0.98, respectively, depending on the study and on which of the subtypes were applied (i.e. affected segment score, anterior or posterior affected segment score, or antero-posterior score). Therefore the ICC observed in this study for the Framingham method was similar to previously published values [14]. There was significant and negative association between the QAAC and BMD of the hip and spine after adjustment for important co-variates that we did not observe with the Framingham method. Higher AAC scores reported by the Framingham technique have been associated with lower BMD in several [1,9,20] but not all [3] studies yet typically these studies have involved hundreds to thousands of patients. Therefore a present sample size of just over 100 patients is considerably smaller than previous studies that have demonstrated this association. The Framingham method can identify discrete, calcified lesions that are more likely to represent calcified plaque in the intima of either the anterior or posterior wall of the aorta. The predictors of higher Framingham scores in this cohort included traditional risk factors for atherosclerosis such as age, hypertension and elevated HOMA, an index that reflects the presence of the metabolic syndrome [21]. Subjects with diabetes had overall higher F-AAC-12 scores however only 11% of participants enrolled in this study actually had a diagnosis of diabetes. Calcification of the elastic lamina of the media has been linked mechanistically to demineralization changes occurring in bone [13]. The significant, inverse relationship that we observed with the QAAC measurement suggests that this technique may also detect AAC in the media that could be directly linked to changes in bone mineralization. Consistent with this hypothesis, the predictors of higher QAAC scores

206

M. Grant et al. / Bone 97 (2017) 201–208

Table 2 Demographic, laboratory, BMD and calcification measures in study group (n = 112) and bivariate associations between all variables and QACC, F-AAC-12 and F-AAC-24. n = 112

QACC

F-AAC-12

F-AAC-24

N (%)

Mean (SD)

Mean (SD)

Mean (SD)

79 (71) 33 (29)

9.0 (4.4)d 5.4 (3.0)

2.38 (2.72) 3.03 (3.22)

3.06 (3.80) 3.67 (4.39)

13 (11) 99 (89)

7.5 (6.1) 8.0 (4.1)

3.54 (4.01) 2.44 (2.70)

4.77 (5.31) 3.04 (3.75)

34 (30) 78 (70)

8.7 (5.0) 7.6 (4.0)

3.94 (3.34)d 1.97 (2.44)

5.29 (4.94)d 2.35 (3.10)

Demographic variables - continuous Age (years) BMI (kg/m2)

Mean (SD) 68.9 (7.4) 28.4 (4.5)

Spearman's rho 0.2 −0.1

Spearman's rho 0.4c 0.03

Spearman's rho 0.37c 0.04

Laboratory data Creatinine (μmol/l) eGFR (ml/min) Calcium (mmol/L) Phosphate (mmol/L) PTH (pmol/L) 25(OH)D (nmol/L) HOMA

79.1 (20.5) 75.0 (15.9) 2.36 (0.11) 1.09 (0.17) 6.3 (3.0) 75.8 (31.9) 2.83 (2.54)

−0.1 −0.2a 0.01 0.32c −0.04 0.1 −0.2a

0.32c −0.4c 0.11 0.06 0.03 0.12 0.25a

0.30a −0.36c 0.09 0.07 0.04 0.11 0.27

Calcification measures F-AAC-24 F-AAC-12 QACC

Median [IQR] 2 [0,5] 2 [0,4] 7.3 [4.7, 10.6]

0.19 0.21a –

0.97a – 0.21a

– 0.97a 0.19

Bone mineral density Lumbar total BMD (g/cm2) Lumbar total T-score Hip total BMD (g/cm2) Hip total T-score

Mean (SD) 1.02 (0.19) −0.26 (1.5) 0.93 (0.15) −0.68 (1.04)

−0.34c −0.32b −0.47c −0.38c

−0.07 −0.09 −0.17 −0.22a

−0.08 −0.09 −0.17 −0.22a

Demographic variables - categorical Sex Female Male Diabetes Yes No Hypertension Yes No

Spearman's correlation coefficients. eGFR: estimated glomerular filtration rate, PTH: parathyroid hormone, 25(OH)D: 25-hydroxy-vitamin D, HOMA: homeostasis model assessment. a p b 0.05. b p b 0.01. c p b 0.001. d t-test comparing categorical variables p b 0.001.

included female sex and the level of serum phosphate. The association between female sex and the QAAC may not be unexpected as it is well recognized that women are at greater risk for osteoporosis. Abnormal phosphate homeostasis has been consistently linked with calcification in patients with advanced kidney disease where phosphate is acknowledged to be a key signaling molecule in the initiation and progression of the process. Serum phosphate, even within the normal range as evident in this study and others, is emerging as a significant predictor of cardiovascular disease in the non-CKD population [22–30]. The single shared predictor of calcification in the bivariate analysis, regardless of measurement technique, was the level of kidney function. Twenty-one study participants had an eGFR b 60 ml/min/m2 yet the significance of this relationship was unchanged after exclusion of these patients. That is, among those with an eGFR N60 ml/min/m2, there remained a significant correlation between lower eGFR and higher scores on both F-ACC-12 and F-ACC-24 and QACC measures. In a report involving CaMOS subjects, compared to men and women with normal kidney function, individuals with decreased kidney function had a greater decline in their BMD over 5 years of follow-up [31]. This bone loss occurred despite the presence of only modest impairment of kidney function. With reduced kidney function, novel risk factors for bone demineralization and vascular calcification develop and include abnormal phosphate and vitamin D metabolism [32–35]. In a much larger sample of 815 subjects, Figueiredo et al. similarly reported that higher serum phosphate levels (but still within the normal range) and lower hip BMD were independent bone/mineral variables that were associated with elevated AAC measured by the Framingham technique [2]. Small changes in serum phosphate levels in the absence of overt hyperphosphatemia could be a trigger for calcification.

The exact molecular mechanisms underlying the link between vascular calcification and osteoporosis are unknown. However, abnormalities in mineral metabolism could represent a common pathway to both conditions. Levels of calcium, 25(OH)D level or PTH were not associated with measures of calcification in this study. Overall, phosphate was predictive of QAAC in the bivariate analysis but this association was not significant after adjustment for age, sex, bone density, eGFR and hypertension. There are limitations associated with using lateral lumbar radiographs to assess AAC. A lateral radiograph is a 2D projection from a 3D volume and therefore is subject to out of plane tissue attenuation of x-rays which provides a major challenge for achieving precision and rigor between imaging studies particularly in obese subjects. In addition, the focus of lateral lumbar radiograph is usually to image the spine; therefore the entire abdominal aortic region is not always captured. A number of subjects were excluded from our study for this reason. Furthermore, it is possible that overlying structures, such as urolithiasis, could be erroneously included in calcification scores. This is especially true for the QAAC technique because there is little room in the protocol for reader interpretation. Further, the goal of ease-of-use and repeatability afforded by parallelogram ROI shapes that do not require reader interpretation could potentially limit the ability of the technique to develop a continuous calcification map. This could be addressed by using a large single polygon ROI tailored to the individual patient. The QAAC technique only utilizes the L3 and L4 sections of the aorta which represents a less extensive measure then is obtainable using Framingham. A further limitation of the QAAC is the time it takes to generate an AAC severity score. In comparison to the Framingham technique, which takes less than a minute to generate an AAC severity score, the

M. Grant et al. / Bone 97 (2017) 201–208 Table 3 Multi-variable linear regression models of predictors of QAAC, F-AAC 12, and F-AAC 24. 3A: Dependent variable = QACC Individual predictors

95.0% confidence interval for B

B

β

p-Value

Lower bound

Upper bound

0.08 2.73 −0.01 2.74 0.98 −1.14

0.14 0.29 −0.05 0.11 0.11 −0.28 0.26 b0.0001

0.16 0.003 0.60 0.24 0.23 0.002

−0.03 0.97 −0.07 −1.91 −0.64 −1.86

0.19 4.50 0.04 7.40 2.60 −0.43

3B: Dependent variable = QACC Pre-selected predictors variables

Age (per year) Female sex eGFR (ml/min/m2) Phosphate (mM) Hypertension DXA spine T-score

Individual predictors

95.0% confidence interval for B

B

β

p-Value Lower bound

Upper bound

0.12 2.60 −0.01 3.98 0.75 −0.66

0.21 0.27 −0.03 0.16 0.08 −0.23

0.03 0.007 0.79 0.11 0.38 0.013

0.24 4.48 0.05 8.85 2.43 −0.14

Model summary: Adjusted R-square Model significance

0.01 0.71 −0.06 −0.88 −0.93 −1.17

0.23 b0.0001

Age (per year) Female sex eGFR (ml/min/m2) Phosphate (mM) Hypertension DXA hip T-score

None to report. Individual predictors

95.0% confidence interval for B

B

β

p-Value Lower bound

Upper bound

0.09 −0.86 −0.03 1.92 1.36 −0.37

0.22 −0.14 −0.15 0.11 0.22 −0.13

0.03 0.17 0.14 0.24 0.02 0.14

0.16 0.37 0.01 5.18 2.49 0.13

Model summary: Adjusted R-square Model significance

0.01 −2.09 −0.07 −1.34 0.22 −0.87

0.21 b0.0001

3D: Dependent variable = FAAC-24 Pre-selected predictors variables

Age (per year) Female sex eGFR (ml/min/m2) Phosphate (mM) Hypertension DXA hip T-score Model summary: Adjusted R-square Model significance

one-time measures of laboratory markers as well as the lack of data with respect to sex hormones. The QAAC technique is a highly reproducible approach to measuring AAC. Although the technique is complex in design, it is simple to teach and operate for new readers. Pseudo-color imaging offers a unique benefit to the QAAC technique that allows users to easily visualize overall calcification burden. The QAAC technique generates an overall area of calcification and includes pixel saturation to approximate lesion density thereby capable of assessing severity of calcification even when very severe. That is, by measuring pixel intensities, a continuous value, the QAAC technique has greater resolving power. Whether it is capable of monitoring subtle calcific changes over time requires further study. We were careful to ensure that the radio-density was not a surrogate indicator of BMD and confirmed that our internal standardization methods and the background elimination process did not bias the assessment of QACC. In summary, overall there was a significant inverse relationship between bone density and aortic calcification measured by the novel QAAC technique after adjustment for important co-variates. We propose that the QAAC could be applied to observational studies and randomized controlled trials that seek to determine the impact of interventions that modify bone density and determine if such treatments could also modify vascular calcification and cardiovascular disease in a general population in the community. Conflict of interest

3C: Dependent variable = FAAC-12 Pre-selected predictors variables

207

Individual predictors

95.0% confidence interval for B

B

β

p-Value Lower bound

Upper bound

0.12 −0.85 −0.03 2.65 2.20 −0.44

0.23 −0.10 −0.11 0.11 0.25 −0.12

0.02 0.33 0.28 0.25 0.007 0.20

0.23 0.86 0.02 7.18 3.77 0.25

0.01 −2.26 −0.08 −1.88 0.61 −1.14

0.20 b0.0001

eGFR - Estimated glomerular filtration rate, β - standardized coefficient, B - unstandardized coefficient. male = 0, female = 1.

QAAC technique requires approximately 10 min per image although this could be reduced in the future through the use of semiautomated scripts. Finally, a computer with sufficient processing power is required to support the Microsoft Excel Macro program in handling the large amount of data generated by the QAAC technique. Further limitations of this study include its cross-sectional design and

Financial support The authors acknowledge the Canadian Institute for Health Research (CIHR MOP 133419) and the Department of Medicine, Queen's University for providing funds that supported this research author. Author contributions Study design: RMH, MAA and TS. Study conduct: MG, JG, SA, RN, and MAA. Data analysis: RMH, MG, MAA, MT, PJ and WH. Data interpretation: RMH, MG, MAA and TA. Drafting manuscript: RMH, MG, MAA and TA. Approving final version of manuscript: MG, JG, TA, WH, SA, RN, MAA, MT, JG and RMH. RMH takes responsibility for the integrity of the data analysis. Acknowledgements The authors acknowledge Ms. Erin Christilaw for her assistance in reading the radiograph images as well as Dr. Karen Rees-Milton for her assistance with the CaMOS patient data. . References [1] Y.Z. Bagger, H.B. Rasmussen, P. Alexandersen, T. Werge, C. Christiansen, L.B. Tanko, PERF study group: links between cardiovascular disease and osteoporosis in postmenopausal women: serum lipids or atherosclerosis per se? Osteoporos. Int. 18 (2007) 505–512. [2] C.P. Figueiredo, N.M. Rajamannan, J.B. Lopes, V.F. Caparbo, L. Takayama, M.E. Kuroishi, I.S. Oliveira, P.R. Menezes, M. Scazufca, E. Bonfa, R.M. Pereira, Serum phosphate and hip bone mineral density as additional factors for high vascular calcification scores in a community-dwelling: the Sao Paulo Ageing & Health Study (SPAH), Bone 52 (2013) 354–359. [3] E. Flipon, S. Liabeuf, P. Fardellone, R. Mentaverri, T. Ryckelynck, F. Grados, S. Kamel, Z.A. Massy, P. Dargent-Molina, M. Brazier, Is vascular calcification associated with bone mineral density and osteoporotic fractures in ambulatory, elderly women? Osteoporos. Int. 23 (2012) 1533–1539. [4] J.A. Hyder, M.A. Allison, M.H. Criqui, C.M. Wright, Association between systemic calcified atherosclerosis and bone density, Calcif. Tissue Int. 80 (2007) 301–306. [5] J.A. Hyder, M.A. Allison, N. Wong, A. Papa, T.F. Lang, C. Sirlin, S.M. Gapstur, P. Ouyang, J.J. Carr, M.H. Criqui, Association of coronary artery and aortic calcium with lumbar

208

[6]

[7]

[8]

[9]

[10]

[11]

[12] [13] [14]

[15]

[16]

[17]

[18]

M. Grant et al. / Bone 97 (2017) 201–208 bone density: the MESA abdominal aortic calcium study, Am. J. Epidemiol. 169 (2009) 186–194. J.A. Hyder, M.A. Allison, E. Barrett-Connor, R. Detrano, N.D. Wong, C. Sirlin, S.M. Gapstur, P. Ouyang, J.J. Carr, M.H. Criqui, Bone mineral density and atherosclerosis: the multi-ethnic study of atherosclerosis, abdominal aortic calcium study, Atherosclerosis 209 (2010) 283–289. N.E. Jensky, J.A. Hyder, M.A. Allison, N. Wong, V. Aboyans, R.S. Blumenthal, P. Schreiner, J.J. Carr, C.L. Wassel, J.H. Ix, M.H. Criqui, The association of bone density and calcified atherosclerosis is stronger in women without dyslipidemia: the multi-ethnic study of atherosclerosis, J. Bone Miner. Res. 26 (2011) 2702–2709. D. Li, S.S. Mao, B. Khazai, J.A. Hyder, M. Allison, M.C. R, I. de B, J.J. Carr, M.H. Criqui, Y. Gao, M.J. Budoff, Noncontrast cardiac computed tomography image-based vertebral bone mineral density: the multi-ethnic study of atherosclerosis (MESA), Acad. Radiol. 20 (2013) 621–627. M. Naves, M. Rodriguez-Garcia, J.B. Diaz-Lopez, C. Gomez-Alonso, J.B. CannataAndia, Progression of vascular calcifications is associated with greater bone loss and increased bone fractures, Osteoporos. Int. 19 (2008) 1161–1166. D.P. Kiel, L.I. Kauppila, L.A. Cupples, M.T. Hannan, C.J. O'Donnell, P.W. Wilson, Bone loss and the progression of abdominal aortic calcification over a 25 year period: the Framingham heart study[Erratum appears in Calcif Tissue Int. 2004 Feb;74(2): 208] Calcif. Tissue Int. 68 (2001) 271–276. C.M. Giachelli, M.Y. Speer, X. Li, R.M. Rajachar, H. Yang, Regulation of vascular calcification: roles of phosphate and osteopontin, Circ. Res. 96 (2005) 717–722 (Review, 84 refs). C.E. Lampropoulos, I. Papaioannou, D.P. D'Cruz, Osteoporosis—a risk factor for cardiovascular disease?[Review] Nat. Rev. Rheumatol. 8 (2012) 587–598. L.L. Demer, Y. Tintut, Mechanisms linking osteoporosis with cardiovascular calcification, Curr. Osteoporos. Rep. 7 (2009) 42–46. L.I. Kauppila, J.F. Polak, L.A. Cupples, M.T. Hannan, D.P. Kiel, P.W. Wilson, New indices to classify location, severity and progression of calcific lesions in the abdominal aorta: a 25-year follow-up study, Atherosclerosis 132 (1997) 245–250. G. Ioannidis, S. Pallan, A. Papaioannou, M. Mulgund, L. Rios, J. Ma, L. Thabane, K.S. Davison, R.G. Josse, C.S. Kovacs, N. Kreiger, W.P. Olszynski, J.C. Prior, T. Towheed, J.D. Adachi, CaMos research group: glucocorticoids predict 10-year fragility fracture risk in a population-based ambulatory cohort of men and women: Canadian multicentre osteoporosis study (CaMos), Arch. Osteoporos. 9 (2014) 169. W. Zhou, L. Langsetmo, C. Berger, S. Poliquin, N. Kreiger, S.I. Barr, S.M. Kaiser, R.G. Josse, J.C. Prior, T.E. Towheed, T. Anastassiades, K.S. Davison, C.S. Kovacs, D.A. Hanley, E.A. Papadimitropoulos, D. Goltzman, CaMos research group: longitudinal changes in calcium and vitamin D intakes and relationship to bone mineral density in a prospective population-based study: the Canadian multicentre osteoporosis study (CaMos), J. Musculoskelet. Nueronal Interact. 13 (2013) 470–479. J.M. Muir, C. Ye, M. Bhandari, J.D. Adachi, L. Thabane, The effect of regular physical activity on bone mineral density in post-menopausal women aged 75 and over: a retrospective analysis from the Canadian multicentre osteoporosis study, BMC Musculoskelet. Disord. 14 (2013) 253. W.M. Hopman, T. Towheed, T. Anastassiades, A. Tenenhouse, S. Poliquin, C. Berger, L. Joseph, J.P. Brown, T.M. Murray, J.D. Adachi, D.A. Hanley, E. Papadimitropoulos, Canadian normative data for the SF-36 health surveyCanadian Multicentre Osteoporosis Study Research Group CMAJ 163 (2000) 265–271.

[19] N. Kreiger, A. Tenenhouse, L. Joseph, T. Mackenzie, S. Poliquin, J.P. Brown, J.C. Prior, R.S. Rittmasters, The Canadian multicentre osteoporosis study (CaMOS): background, rationale, methods, Can J Aging 18 (3) (2015) 376–387. [20] P.W. Wilson, L.I. Kauppila, C.J. O'Donnell, D.P. Kiel, M. Hannan, J.M. Polak, L.A. Cupples, Abdominal aortic calcific deposits are an important predictor of vascular morbidity and mortality, Circulation 103 (2001) 1529–1534. [21] T.P. Murphy, R. Dhangana, M.J. Pencina, A.M. Zafar, R.B. D'Agostino, Performance of current guidelines for coronary heart disease prevention: optimal use of the Framingham-based risk assessment, Atherosclerosis 216 (2011) 452–457. [22] R.N. Foley, Phosphate levels and cardiovascular disease in the general population[Review] [31 refs] Clin. J. Am. Soc. Nephrol. 4 (2009) 1136–1139. [23] R.N. Foley, A.J. Collins, C.A. Herzog, A. Ishani, P.A. Kalra, Serum phosphate and left ventricular hypertrophy in young adults: the coronary artery risk development in young adults study, Kidney Blood Press. Res. 32 (2009) 37–44. [24] M. Tonelli, F. Sacks, M. Pfeffer, Z. Gao, G. Curhan, Cholesterol And Recurrent Events Trial Investigators: Relation between serum phosphate level and cardiovascular event rate in people with coronary disease[Erratum appears in Circulation. 2007 Dec 4;116(23):e556] Circulation 112 (2005) 2627–2633. [25] M. Tonelli, G. Curhan, M. Pfeffer, F. Sacks, R. Thadhani, M.L. Melamed, N. Wiebe, P. Muntner, Relation between alkaline phosphatase, serum phosphate, and all-cause or cardiovascular mortality, Circulation 120 (2009) 1784–1792. [26] J. Kendrick, M. Chonchol, The role of phosphorus in the development and progression of vascular calcification[Review] Am. J. Kidney Dis. 58 (2011) 826–834. [27] N. Shobeiri, M.A. Adams, R.M. Holden, Phosphate: an old bone molecule but new cardiovascular disease risk factor, Br. J. Clin. Pharmacol. (2013) 3–19. [28] S.J. Onufrak, A. Bellasi, L.J. Shaw, C.A. Herzog, F. Cardarelli, P.W. Wilson, V. Vaccarino, P. Raggi, Phosphorus levels are associated with subclinical atherosclerosis in the general population, Atherosclerosis 199 (2008) 424–431. [29] W. Jahnen-Dechent, A. Heiss, C. Schafer, M. Ketteler, Fetuin-a regulation of calcified matrix metabolism[Review] Circ. Res. 108 (2011) 1494–1509. [30] W. Jahnen-Dechent, C. Schafer, M. Ketteler, M.D. McKee, Mineral chaperones: a role for fetuin-a and osteopontin in the inhibition and regression of pathologic calcification[Review, 70 refs] J. Mol. Med. 86 (2008) 379–389. [31] S.A. Jamal, V.J. Swan, J.P. Brown, D.A. Hanley, J.C. Prior, A. Papaioannou, L. Langsetmo, R.G. Josse, Canadian multicentre osteoporosis study research group: kidney function and rate of bone loss at the hip and spine: the Canadian multicentre osteoporosis study, Am. J. Kidney Dis. 55 (2010) 291–299. [32] G.A. Block, Therapeutic interventions for chronic kidney disease-mineral and bone disorders: focus on mortality[Review] Curr. Opin. Nephrol. Hypertens. 20 (2011) 376–381. [33] N.I. Parikh, S.J. Hwang, M.G. Larson, U. Hoffmann, D. Levy, J.B. Meigs, C.J. O'Donnell, C.S. Fox, Indexes of kidney function and coronary artery and abdominal aortic calcium (from the Framingham offspring study), Am. J. Cardiol. 102 (2008) 440–443. [34] M.C. Foster, S.J. Hwang, M.G. Larson, J.H. Lichtman, N.I. Parikh, R.S. Vasan, D. Levy, C.S. Fox, Overweight, obesity, and the development of stage 3 CKD: the Framingham heart study, Am. J. Kidney Dis. 52 (2008) 39–48. [35] N.I. Parikh, S.J. Hwang, M.G. Larson, D. Levy, C.S. Fox, Chronic kidney disease as a predictor of cardiovascular disease (from the Framingham heart study), Am. J. Cardiol. 102 (2008) 47–53.