Impaired arterial function associated with thinning of cortical bone in rheumatoid arthritis

Authors


Abstract

Objective

To evaluate the association between peripheral arterial function and cortical bone thickness in rheumatoid arthritis (RA).

Methods

In a cross-sectional study, we measured the combined cortical thickness (CCT) of the second metacarpal bone from hand radiographs, and the ankle-to-arm systolic blood pressure ratio, also known as ankle–brachial index (ABI), in RA patients. We evaluated the association between the 2 using multinomial logistic regression.

Results

We obtained CCT and ABI measurements in 588 RA patients. The mean ± SD CCT was 3.62 ± 1.16 mm. The proportion of patients with ≥1 ABI value ≤0.9, indicating obstructed lower limb arteries, increased from 18 (9.2%) of 191 patients in the highest CCT tertile to 25 (12.5%) of 200 in the middle CCT tertile to 38 (19.2%) of 198 in the lowest CCT tertile (P for trend 0.005). We noted a similar pattern for ABI values >1.3, indicative of arterial incompressibility (frequencies in high, middle, and low CCT tertiles were 4.7%, 9.5%, and 19.9%, respectively; P for trend ≤0.001). These trends remained significant after multivariable adjustment for potential confounders. After adjustment for the manifestations of RA and cumulative glucocorticoid dose, the association between CCT and arterial obstruction remained significant, but that with arterial incompressibility weakened considerably.

Conclusion

There is an association between metacarpal cortical bone thinning and obstruction or incompressibility of the peripheral arteries in RA. The association with incompressibility may be mediated by systemic inflammation and/or glucocorticoids, but that with obstruction is independent of a wide array of potential confounders. Clinicians should be alert to the possibility of impaired arterial function RA patients with thinned metacarpal cortical bone.

INTRODUCTION

Osteoporosis and atherosclerosis are 2 important side effects of rheumatoid arthritis (RA). Thinning of the cancellous bone near the joints, known as periarticular osteopenia, is a radiographic sign typical of RA, distinct from the cortical erosions that are characteristic of RA (1). Patients with RA also lose substance from the axial skeleton, a phenomenon promoted by inflammation, disuse, and glucocorticoids (2–5). Less attention has been directed to the cortical diaphyseal bone, which also seems to be thinned at an accelerated rate in patients with RA. As is the case with the periarticular cancellous bone, cortical diaphyseal bone loss correlates with joint erosions and joint space narrowing (6), systemic inflammation, and glucocorticoid exposure (7).

RA is also associated with an increased susceptibility to atherosclerotic cardiovascular (CV) comorbidity (8–10). We and others have observed an association between systemic inflammation markers in RA and the carotid intima-media thickness or plaque (2, 11–15). In addition, we have reported that patients with RA have an increased frequency of lower limb arterial incompressibility (16). Incompressibility is the physiologic correlate of medial arterial calcification (17, 18), also known as Mönckeberg's sclerosis (19).

Because the frequency of both atherosclerosis and osteoporosis is increased in patients with RA, some co-occurrence of the 2 is to be expected. The fact that the 2 processes are also associated with aging further raises this expectation (20). Nevertheless, the 2 conditions seem to occur in association more often than can be explained by age alone (21, 22). Recent evidence suggests that osteoporosis and atherosclerosis share pathophysiologic links (23). Naturally, this has raised the possibility that similar therapies could be used to treat both disorders (24).

Despite the high frequency of CV disease and osteoporosis in RA, information about the extent to which they coexist in patients with RA is sparse (25). In the present study, we examined the association between cortical bone thickness and impaired arterial function in a sample of patients with RA. We aimed to isolate the association between the 2 processes from the influence of factors that could confound the association, including CV risk factors, inflammation markers, and exposure to glucocorticoids.

PATIENTS AND METHODS

We performed a cross-sectional analysis of data collected from a cohort of patients with RA. We have described the recruitment and characteristics of the patients in a previous publication (26). Briefly, from 1996 to 2001, we enrolled consecutive patients who met the 1987 American College of Rheumatology (formerly the American Rheumatism Association) RA classification criteria (27) into ÓRALE (Outcome of Rheumatoid Arthritis Longitudinal Evaluation), a study of the disablement process in RA (28). We enrolled patients from 6 rheumatology clinics in San Antonio, Texas: a county-funded clinic; a Veterans Administration clinic; a private, university-based faculty practice; a community-based, 7-rheumatologist private practice; an army clinic; and an air force clinic. All patients resided in Bexar County, Texas, or nearby communities. Our institutional review board approved the study, and all patients gave written, informed consent to participate in the study. A physician or a research nurse trained to perform the study procedures examined all patients and reviewed the medical records. The study protocol included an anteroposterior plain radiograph of the hands and wrists.

We ascertained age, sex, and race/ethnicity by self-report (29). We recorded height and weight and calculated body mass index (BMI; weight in kilograms/height in meters squared). We recorded dates of symptom onset and diagnosis of RA, confirmed by the medical records (27). One of 2 physician examiners, both trained in joint examination techniques, assessed 48 joints in each patient for joint tenderness or pain on motion, swelling, or deformity, and recorded the presence or absence of subcutaneous nodules (30, 31). Interrater reliability of the joint examination was 0.98 for the tender joint count, 0.92 for the swollen joint count, and 0.99 for the deformed joint count.

We ascertained the cumulative glucocorticoid dose as described previously (32). Briefly, at the baseline visit and each of the followup visits, we asked patients to bring a list of their medications, and we reviewed pharmacy and medical records. We asked patients whether they were receiving glucocorticoids, and if so we asked for the date when the glucocorticoids were first prescribed and the dose currently in use. We estimated the cumulative oral glucocorticoid dose by multiplying the current daily dose by the number of days since glucocorticoids were initiated. For the few patients who received glucocorticoids on alternate days or on other nondaily schedules, we averaged the dose over the number of days to obtain the daily amount. We considered patients who had received glucocorticoids less often than monthly, orally, or by injection as not having received glucocorticoids. At each followup visit, we calculated the cumulative dose since the previous visit. The total cumulative glucocorticoid dose is presented as the summed dose over each interval, expressed in prednisone equivalents.

We ascertained CV risk factors as described in an earlier publication (15). We measured CV risk factors at multiple times from the recruitment of patients into the study of disablement in RA until the arterial measurements were performed. For hypertension, we averaged blood pressure measurements obtained at each study visit. For obesity, we measured height and weight at the time of the arterial assessments, to calculate the current BMI. Diabetes mellitus and hypercholesterolemia were ascertained from a medical record and medication review, supplemented by self-report. We considered these conditions present if a physician recorded them in the medical record, or if a patient had taken antidiabetic or lipid-lowering medications during the course of the study. In addition, we considered patients to have hypercholesterolemia if their fasting plasma cholesterol measured during a study visit was ever ≥200 mg/dl, or if they received lipid-lowering drugs. We considered an alternative definition of hypercholesterolemia, adding to the above definition persons whose low-density lipoprotein (LDL) cholesterol was ≥160 mg/dl, or persons with ≥2 CV risk factors and LDL ≥130 mg/dl. Persons who had ever smoked cigarettes were classified as current smokers if they continued to smoke and former smokers if they had quit in the past. We considered women to be postmenopausal if they had undergone surgical menopause; if they were receiving hormone replacement therapy; if a physician recorded them as being postmenopausal in the medical records; or if they were ≥55 years of age, an age at which ≥95% of US women are postmenopausal (33, 34).

We classified socioeconomic status (SES) according to Nam and Powers (35), using years of education, inflation-adjusted monthly household income, and current or past occupation. The 3 variables were used to calculate an SES score on an ascending 0–100 scale (35).

We measured the combined cortical thickness (CCT) as reported in a previous publication (7). Briefly, we digitized anteroposterior hand radiographs using an HP ScanJet 4c scanner with a transparent media adaptor (Hewlett Packard, Palo Alto, CA). After contrast enhancement and magnification × 2, one author (JFR), acting as the sole reader, used the Image Tool analysis software package version 2.3 (Dental Diagnostic Science, University of Texas Health Science Center at San Antonio, TX) to measure the CCT. We marked the medullary and cortical margins at the midpoint of the left and right second metacarpal bones. We measured the distance between the marks in millimeters, averaging both sides for analyses. We estimated the test–retest reliability of the CCT measurements. More than 1 year after the initial set of images were measured, we randomly selected 45 patients for remeasurement of the original radiograph. In 3 of the 45 patients, the radiographs were incomplete or blurry, leaving 42 available for measurement. The intraclass correlation between the original and repeat measurements was 0.93 (95% confidence interval [95% CI] 0.88–0.97) (36).

Our study protocol did not include standardized measurement of the bone mineral density (BMD) using a clinically approved technique such as dual x-ray absorptiometry (DXA). However, we were interested in estimating the diagnostic performance of the CCT as a surrogate osteoporosis classifier. To accomplish this, we reviewed patients' medical records seeking DXA measurements of BMD obtained within 3 years of the hand radiographs, as part of patients' clinical care. From these DXA measurements, we used the T scores from the spine, femoral neck, or total hip to define osteoporosis. For the purpose of evaluating the diagnostic performance of the CCT, we considered a patient to have osteoporosis if any of the DXA-obtained T scores was less than −2.5.

We assessed arterial function by means of the ankle–brachial index (ABI), also known as the ankle-to-arm systolic blood pressure ratio (37), as described in an earlier publication (16). Briefly, we used a Doppler apparatus (model 841-A; Parks Medical Electronics, Aloha, OR) and a standard blood pressure cuff to detect the systolic blood pressure wave at the dorsalis pedis and posterior tibial arteries on each lower limb. Thus, each patient could have up to 4 arterial measurements. We divided each lower limb artery systolic pressure by the average of the right and left brachial systolic pressures. The normal ABI lies between 0.91 and 1.3; ABI values ≤0.9 suggest the artery is obstructed, and ABI values >1.30 indicate incompressibility (16).

Erythrocyte sedimentation rate (ESR) and titers for serum rheumatoid factor (RF) were measured in each hospital laboratory. Five of the centers used a manual Westergren ESR technique, and the sixth center used an automated analyzer (Diesse Diagnostic Senese, Milan, Italy) (38, 39). Latex agglutination was used to measure RF in 5 hospitals, and an enzyme-linked immunosorbent assay technique was used in the sixth (40). In each hospital, a positive RF was defined as a titer occurring in <5% of normal patients. Most patients with negative results for RF were tested on several occasions.

Statistical analysis.

We examined tabulations and frequency distributions of the key study variables. We estimated the accuracy of the CCT in classifying DXA-defined osteoporosis using the area under the receiver operating characteristic (ROC) curve of the CCT (41). To obtain a CCT cutoff value to use as a surrogate osteoporosis classifier, we chose the CCT value with the highest sensitivity and specificity in identifying DXA-defined osteoporosis. To examine the association between the CCT and the ABI, we categorized the CCT into tertiles, and tabulated the frequency of arterial obstruction and incompressibility in each tertile. We also tabulated the frequency of osteoporosis, defined as described above from the CCT, among patients with obstructed, incompressible, or normal lower limb arteries. For both of these tabulations, we used binomial 95% CIs. To adjust our estimates of the CCT-ABI association for the possibility of confounding by other variables that could influence either or both the CCT and ABI, we used multinomial logistic regression (42). In these multivariable models, the ABI was used as a 3-level dependent variable, with the normal ABI as the referent category. We tested sequential models that included age at RA onset, RA duration, sex, ethnic group, CV risk factors, menopausal status, manifestations of RA, inflammatory markers, RF, and cumulative glucocorticoid exposure as covariates. All P values are 2-sided to ≤10−3, without modification for multiple testing. We used the Stata 8 Statistical Software package (StataCorp, College Station, TX) for all analyses.

RESULTS

We enrolled 779 patients in the ÓRALE study from 1996 to 2001. A plain radiograph of the hands and wrists became part of the evaluation protocol in 1999. Thus, hand radiographs were obtained for 731 patients (94%). CCT measurements were performed in 649 patients (89% of those with radiographs). By February 2000, when we began to perform ABI measurements on members of the ÓRALE cohort, 66 patients had died and 32 had moved away from the San Antonio area before we could schedule an appointment. This left 681 patients still eligible for measurement of the ABI. Of these, we could not establish contact with 17, and 19 declined participation. The ABI was obtained for 644 patients (95% of those eligible). The CCT measurement was not obtained for all available radiographs. Thus, a total of 588 patients had both an ABI measurement and a CCT measurement on a radiograph obtained within 3 years of the ABI. With the exception of the CCT, all variables used in this analysis were ascertained at the time of the ABI measurement, from 2000 through 2003.

The demographic and clinical characteristics of the 588 patients who had an ABI and a CCT measurement are provided in Table 1. Members of the ÓRALE cohort who did not have the CCT and the ABI measurements did not differ significantly in age or sex from those who did.

Table 1. Characteristics of 588 patients with RA at the time of measurement of ankle–brachial index*
 Ankle–brachial index
All patients<0.9 (obstructed)0.9–1.3 (normal)>1.3 (incompressible)
  • *

    Values are the number (percentage) unless otherwise indicated. RA = rheumatoid arthritis; SES = socioeconomic status (measured using the Nam and Powers scale [35]); ESR = erythrocyte sedimentation rate.

  • P ≤ 0.001 versus normal ankle–brachial index.

  • P ≤ 0.05 versus normal ankle–brachial index.

  • §

    P ≤ 0.01 versus normal ankle–brachial index.

No. of patients 8044563
Age, mean ± SD years58.8 ± 12.462.9 ± 11.356.9 ± 12.366.2 ± 10.0
Age at RA onset, mean ± SD years44.6 ± 13.549.9 ± 13.743.3 ± 13.0473 ± 14.1
RA duration, mean ± SD years14.1 ± 10.312.9 ± 8.213.7 ± 12.818.8 ± 15.5
Women416 (70.8)45 (56)336 (76)35 (55)
White194 (33.0)33 (41)141 (32)20 (32)
Lower tertile SES203 (35.0)31 (39)148 (33)24 (38)
Middle tertile SES186 (31.6)28 (35)137 (31)21 (33)
Upper tertile SES199 (33.8)21 (26)160 (29)18 (29)
Diabetes117 (20)26 (33)76 (17)15 (24)
Hypertension169 (29)26 (32)124 (28)19 (30)
Hypercholesterolemia319 (54)53 (66)230 (52)36 (57)
Former smoker224 (41)43 (54)168 (38)31 (49)
Current smoker110 (19)23 (29)80 (18)7 (11)
Body mass index, mean ± SD kg/m229.1 ± 6.329.1 ± 6.229.3 ± 6.527.8 ± 4.6
Tender joint count, mean ± SD14.2 ± 1314.5 ± 12.914.1 ± 13.013.9 ± 13.4
Swollen joint count, mean ± SD4.2 ± 4.94.7 ± 5.34.1 ± 3.64.6 ± 5.1
Deformed joint count, mean ± SD15.8 ± 12.117.5 ± 11.814.3 ± 11.524.7 ± 13.0
Nodules279 (47)46 (58)196 (44)37 (59)
Rheumatoid factor496 (85)70 (88)374 (84)52 (83)
ESR, mean ± SD mm/hour41.2 ± 26.446.7 ± 30.239.3 ± 37.047.5 ± 40.1
C-reactive protein, mean ± SD mg/dl16.3 ± 23.717.3 ± 25.115.3 ± 13.222.2 ± 14.7
HLA–DRB1 shared epitope423 (72)64 (81)312 (70)47 (75)
Cumulative glucocorticoid dose, mean ± SD gm10.0 ± 15.39.3 ± 13.89.4 ± 13.915.5 ± 24.2§

The mean ± SD CCT was 3.6 ± 1.2 mm. The mean CCT did not differ significantly between patients with and without ABI measurements (3.6 ± 1.2 mm in the 588 patients with ABI versus 3.8 ± 1.2 mm in the 61 patients without ABI; P = 0.4).

The 588 patients who had both ABI and CCT measurements contributed 2,347 lower limb arteries to the analysis. Two patients contributed only 2 arteries because they each had an amputated lower limb. Another patient contributed only 3 arteries because a skin ulcer prevented us from obtaining the systolic pressure measurement from 1 artery. The remaining 585 patients contributed 4 arteries each. The ABI was ≤0.9, indicating obstruction, in 202 arteries (8.6%). The ABI exceeded 1.3 in 145 arteries (6.2%), indicating incompressibility. The remaining 2,000 arteries (85%) were considered normal because the ABI was higher than 0.9, but not higher than 1.3.

On a patient-level basis, 86 patients (15%) had ≥1 obstructed artery, and 77 (13%) had at least 1 incompressible artery. There were 2 patients who had 1 incompressible and 1 obstructed artery. A third patient had 3 obstructed arteries and 1 incompressible artery. In all other patients with abnormal arteries, all arteries had the same type of abnormality. For the person-based analysis, we considered the 3 patients with mixed abnormalities to have obstruction.

We stratified the CCT distribution into tertiles. We were interested in seeing whether the proportion of patients with obstructed or incompressible arteries varied according to the CCT stratum. The result of this analysis is shown in Figure 1. We noted significant variation in the proportion of both types of abnormal arteries between the 3 CCT tertiles, with the lowest proportion of both incompressible and obstructed arteries occurring among the patients in the highest CCT tertile (Figure 1).

Figure 1.

Proportion of persons with incompressible or obstructed lower limb arteries, according to tertile of combined cortical thickness (CCT). There were significant trends for the proportion of patients with either incompressibility or obstruction to increase, associated with lower CCT tertiles (P ≤ 0.001 for incompressibility and P = 0.005 for obstruction). The associations remained significant after multivariable adjustment for age, sex, cardiovascular risk factors, and menopausal status. 95% CI = 95% confidence interval.

We also examined the ABI-CCT association from the perspective of the ABI. The mean CCT varied significantly among the ABI categories (P ≤ 0.001 by one-way analysis of variance). Patients with normal lower limb arteries had the highest mean ± SD CCT (3.8 ± 1.2 mm). Patients with obstructed arteries had a mean ± SD CCT of 3.3 ± 1.2 mm (P = 0.003 compared with patients with normal arteries). Among the patients with incompressible arteries, the mean CCT was 3.0 ± 1.2 mm (P ≤ 0.001).

A CCT value of 2.9 mm provided the best cutoff for DXA-defined osteoporosis (i.e., a T score less than −2.5). Values at or below this cutoff provided a sensitivity of 83% and a specificity of 76% in identifying osteoporosis diagnoses from a DXA. The area under the ROC curve for the CCT in identifying DXA-defined osteoporosis was 0.84, a diagnostic accuracy that compares favorably with other methods of estimating osteoporosis from peripheral sites (43–45). We found that the proportion of patients with a CCT <2.9 mm varied significantly between patients with normal lower limb arteries and those with obstructed or incompressible arteries (Figure 2).

Figure 2.

Proportion of patients with osteoporosis, as defined by a combined cortical thickness (CCT) ≤ 2.9 mm, according to the type of arterial abnormality present. A significantly higher proportion of patients with obstructed or incompressible lower limb arteries had osteoporosis (P ≤ 0.001). This association remained significant after adjustment for age, sex, cardiovascular risk factors, and postmenopausal status. 95% CI = 95% confidence interval.

The number of obstructed or incompressible arteries could range from 0 to 4 in each patient. We found a significant association between the CCT and the number of affected arteries: as the number of affected arteries increased, the CCT decreased (Table 2). This pattern was accentuated when we considered each artery simply as normal or abnormal.

Table 2. CCT of the second metacarpal, according to the number of incompressible or obstructed lower limb arteries*
 No. of patientsCCT (mm)P
  • *

    Values are the mean ± SD unless otherwise indicated. CCT = combined cortical thickness of the second metacarpal bone at the mid-diaphysis.

No. of incompressible  arteries   
 05233.7 ± 1.1Referent
 1303.0 ± 1.00.001
 293.5 ± 1.20.7
 3112.8 ± 1.30.01
 4162.6 ± 1.4≤0.001
 P for trend ≤0.001 
No. of obstructed arteries   
 05083.7 ± 1.2Referent
 1263.5 ± 1.20.5
 2133.3 ± 1.40.3
 3173.4 ± 1.20.3
 4253.1 ± 1.20.02
 P for trend 0.006 

Because the association we observed between the ABI and the CCT could be due to the influence of a confounder, it was important to examine the association between these 2 variables and other variables that could confound the association. An important variable that could influence both bone and arteries was the cumulative glucocorticoid dose. In bivariate analyses, the cumulative glucocorticoid dose was associated with arterial incompressibility, with an odds ratio (OR) of 1.17 per 10 gm of glucocorticoids (95% CI 1.02–1.35). However, cumulative glucocorticoids were not significantly associated with arterial obstruction (OR 0.95, 95% CI 0.80–1,14). Glucocorticoids were significantly associated with the CCT, with every 10 gm of glucocorticoids being associated with a 0.15-mm loss of CCT (95% CI 0.09–0.21). These associations were independent of age and sex.

We also conducted a multivariable analysis of the ABI-CCT association, adjusting for additional potential confounders. We accomplished this using multinomial logistic regression, with the categorized ABI as the dependent variable. We chose this regression technique because the categorized ABI is a 3-level variable with the natural referent category in the middle of the distribution (i.e., obstructed/normal/incompressible). Table 3 shows the OR associated with a 1 SD difference in CCT, adjusting for 4 groups of covariates: age and sex, CV risk factors, menopausal status, and RA manifestations and cumulative glucocorticoids. This analysis demonstrates that the association between CCT and arterial obstruction remained after adjustment for all the covariates we tested. In the case of incompressibility, the strength of the association between CCT and incompressibility diminished considerably upon adding the RA manifestations to the model, and the statistical significance of the association was lost upon adjustment for the cumulative glucocorticoid dose (Table 3).

Table 3. Lower limb obstruction or incompressibility associated with lower CCT in 588 patients with RA, adjusting for potential confounders*
Model covariatesOR-associated lower CCT (95% CI)
ObstructionIncompressibility
  • *

    CCT = combined cortical thickness; RA = rheumatoid arthritis; OR = odds ratio; 95% CI = 95% confidence interval; SES = socioeconomic status; CV = cardiovascular.

  • OR per CCT that is lower by 1 SD.

  • Covariates were grouped as follows: CV risk factors = diabetes mellitus, systolic blood pressure, cigarette smoking, hypercholesterolemia, body mass index, menopausal status; RA manifestations = tender, swollen, and deformed joint counts, subcutaneous nodules, erythrocyte sedimentation rate, rheumatoid factor positivity.

None1.50 (1.17–1.92)2.07 (1.56–2.76)
Age, sex, SES1.56 (1.17–2.08)2.07 (1.48–2.88)
Age, sex, CV risk factors1.74 (1.26–2.41)2.03 (1.42–2.92)
Age, sex, CV risk factors, RA manifestations1.62 (1.12–2.35)1.51 (1.01–2.27)
Age, sex, CV risk factors, RA manifestations, cumulative glucocorticoids1.62 (1.12–2.35)1.47 (0.97–2.24)

DISCUSSION

We measured the CCT of the second metacarpal bone in a group of patients with RA using conventional hand radiographs (46–48). We related the CCT to the lower limb arterial function, which we assessed using the ABI (37). Due to its noninvasive nature and ease of implementation, ABI has been used extensively in population-based epidemiologic studies, where it has been found to be a significant predictor of cardiovascular mortality (49, 50). In an earlier report, we described evidence that arterial function is more frequently impaired in patients with RA than among age-matched controls (16). Arterial incompressibility, the functional consequence of arterial calcification, was the type of impairment that differed the most between patients with RA and controls (16).

Here, we found that compared with patients whose CCT was in the highest tertile, patients in the middle and lowest tertiles had a progressively higher probability of lower limb artery dysfunction. During our study of these patients with RA, we assembled a comprehensive battery of covariates for statistical adjustment. This allowed us to control for the possibility of confounding, an important consideration in noninterventional studies such as ours. Age may be the most important potential confounder of the association between cortical bone status and arterial dysfunction, because both atherosclerosis and osteoporosis are age-related processes. As shown in Table 3, confounding by age is unlikely to explain the artery-bone association. In fact, age adjustment seemed to strengthen the CCT association with arterial obstruction.

Because of the influence of CV risk factors on the arteries, we adjusted for these risk factors in our multivariable models. This not only did not weaken the association between cortical thinning and arterial obstruction, but strengthened it (Table 3), suggesting that the association occurs through mechanisms that are independent of the CV risk factors.

Among the most striking of our findings is the loss of association between cortical bone thinning and arterial incompressibility upon adjustment for glucocorticoids (Table 3). We previously described a significant association between cumulative glucocorticoids and both arterial incompressibility (32) and CCT loss over time (7). The weakening of the association between the CCT and arterial incompressibility upon glucocorticoid adjustment suggests that bone thinning and incompressibility may share glucocorticoids as an etiology. Such was not the case with arterial obstruction, because neither glucocorticoids nor any of the other adjustment covariates available to us effaced the association between the CCT and arterial obstruction. This suggests that bone loss is associated with arterial obstruction through mechanisms independent of CV risk factors or glucocorticoids.

Our findings are consistent with a recent study in which trabecular bone loss in the radius was associated with arterial pulse wave velocity, a measure of arterial wall stiffness (25). There have been few reports otherwise of association between osteoporosis and arterial function in RA. However, in the general population, the atherosclerosis-osteoporosis association has been well documented. Abdominal aortic calcification in association with vertebral osteoporosis was first reported as far back as the 1960s (51). One previous study used a metacarpal cortical bone measure similar to the one we used (52). Progression of vessel calcification has been associated with loss of metacarpal cortical bone in pre- and postmenopausal women without RA (53). An association between osteoporosis and atherosclerosis has also been described in patients with systemic lupus erythematosus (54, 55).

Arterial calcification, the “skeleton in the closet” of atherosclerosis (22), may share certain mechanisms with calcium deposition in bone (24). Bone components such as osteoprotegerin, osteopontin, and osteocalcin seem to have shared effects in the arterial wall (23, 24). In this respect, it is of interest that osteoprotegerin-deficient mice develop osteoporosis and severe arterial medial calcification (56), while osteoprotegerin-transgenic mice develop osteopetrosis (57).

It is noteworthy that the arterial dysfunction most strongly associated with the CCT was incompressibility, which reflects arterial calcification (17). However, the association weakened upon adjustment for RA manifestations, and was lost upon adjustment for glucocorticoid exposure. This suggests that active rheumatoid inflammation and glucocorticoids may be the link between bone loss and arterial calcification. It should also be noted that the CCT association with arterial obstruction remained even after adjustment for glucocorticoids and other covariates. Further research will be needed to disentangle the mechanism of this independent association.

Our study has limitations. The association we observed could be due to confounding by variables for which we did not adjust. However, residual confounding seems unlikely, given the rich set of adjustment covariates available. Despite potential measurement error, the test–retest reliability of the CCT was excellent. It should be noted that more accurate, automated techniques to measure metacarpal bone are gaining wider use (58, 59). Although the CCT is not currently approved in the US for the diagnosis and management of osteoporosis, our findings suggest that the CCT may perform well in detecting DXA-defined osteoporosis. Finally, our estimate of the cumulative glucocorticoid dose is limited in that we did not attempt to reconstruct increases or reductions in steroid dose prior to recruitment, nor did we have information about intraarticular or other parenteral steroids. A more accurate estimate would likely enhance the association between glucocorticoid exposure, the ABI, and the CCT.

We conclude that cortical bone loss is associated with impaired arterial function in patients with RA. Further research to investigate potential mechanisms of the association would be of considerable interest. Clinicians should be alert to the possibility that patients with thinning of the metacarpal cortical bone may have arterial impairments. Conversely, patients with impaired arterial function may have cortical bone loss. Clinicians who encounter patients with these abnormalities should consider implementing appropriate diagnostic and therapeutic measures.

AUTHOR CONTRIBUTIONS

Dr. del Rincón had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Roldán, Escalante, del Rincón.

Acquisition of data. Roldán, Escalante, del Rincón.

Analysis and interpretation of data. Roldán, Escalante, del Rincón.

Manuscript preparation. Roldán, Escalante, del Rincón.

Statistical analysis. Roldán, Escalante, del Rincón.

Acknowledgements

We thank Samvel Pogosian, MD, for assistance in conducting the study; the Bartter General Clinical Research Center; and the San Antonio rheumatologists who gave us permission to study their patients.

Ancillary