Association between bone mineral density and C-reactive protein in a large population-based sample

Authors

  • Paola de Pablo,

    Corresponding author
    1. University of Birmingham College of Medical and Dental Sciences, City Hospital, Sandwell and West Birmingham Hospitals NHS Trust, and Queen Elizabeth Hospital, University Hospital Birmingham NHS Trust, Birmingham, UK
    • Rheumatology Research Group, Institute of Biomedical Research, 3rd Floor, School of Immunity and Infection, College of Medical and Dental Sciences, University of Birmingham, Vincent Drive, Edgbaston, Birmingham B15 2TT, UK
    Search for more papers by this author
  • Mark S. Cooper,

    1. University of Birmingham College of Medical and Dental Sciences, City Hospital, Sandwell and West Birmingham Hospitals NHS Trust, and Queen Elizabeth Hospital, University Hospital Birmingham NHS Trust, Birmingham, UK
    Search for more papers by this author
  • Christopher D. Buckley

    1. University of Birmingham College of Medical and Dental Sciences, City Hospital, Sandwell and West Birmingham Hospitals NHS Trust, and Queen Elizabeth Hospital, University Hospital Birmingham NHS Trust, Birmingham, UK
    Search for more papers by this author

Abstract

Objective

Several studies suggest that bone mineral density (BMD) is reduced in chronic inflammatory diseases. Higher serum levels of C-reactive protein (CRP) have been associated with lower BMD in women and older adults. However, it is not clear whether this association holds in a representative sample of the general population. The purpose of this study was to examine the relationship between BMD and CRP level in a large representative US population-based sample from the National Health and Nutrition Examination Survey (NHANES).

Methods

We included participants age ≥20 years with BMD (total and subregions) measured by dual x-ray absorptiometry scans and complete information on covariates from NHANES. The association between CRP level and BMD was evaluated using multivariate linear regression models, adjusting for potential confounders and further adjusting for comorbid diseases, medications, and serum vitamin D levels.

Results

The study sample included 10,475 participants (53% Caucasian, 22% Mexican American, 18% African American, and 7% other races). Men had higher BMD and lower CRP concentrations than women. BMD (total body BMD as well as subtotal BMD and BMD of the extremities, ribs, and trunk subregions) was inversely associated with quintiles of CRP concentration both in men and in women in a dose-dependent manner (for total BMD, P for trend < 0.0001 for men, P for trend = 0.0005 for women). The associations were independent of medications, comorbidities, and other potential confounders. The results remained largely unchanged with further adjustment for serum vitamin D levels.

Conclusion

Among men and women in a large representative population-based sample, the CRP level was inversely and independently associated with total BMD in a dose-dependent manner.

Chronic inflammatory diseases are associated with systemic bone loss (1), osteoporosis, and increased risk of nontraumatic fracture. Such diseases include rheumatoid arthritis (RA) (2, 3), ankylosing spondylitis (4), systemic lupus erythematosus (5), inflammatory bowel disease (6, 7), and chronic obstructive pulmonary disease (COPD) (8). In this context, the pathogenic damage to bone is mediated by the interaction of inflammatory cells, cytokines, and bone cells. For example, in RA, generalized bone loss is related to increased osteoclast activity mediated by inflammatory cytokines, such as tumor necrosis factor (TNF), interleukin-1 (IL-1), and IL-6, through the RANKL/RANK/osteoprotegerin system (9, 10). Bone loss reversal has been observed with anti-TNF therapy, with some studies showing increments in spinal and femoral bone mineral density (BMD) (6, 11–17).

While the association between chronic inflammatory diseases and bone loss, osteoporosis, and increased fracture risk is well established, data are scarce regarding whether low-grade inflammation has a similar effect or whether there is a dose-response relationship between inflammation and bone loss in the general population. Previous epidemiologic studies of the association between biomarkers of inflammation and BMD or bone loss have provided inconsistent or even contradictory results, with 2 studies showing an inverse association between C-reactive protein (CRP) level and BMD among Korean women and Southern Tasmanian older adults (18, 19), and other studies showing no independent associations between CRP level and BMD (20–23) among American or Australian women ages ≥65 years (20, 21), healthy American postmenopausal women ages 45–65 years (22), and Italian adults ages 40–79 years (23). Furthermore, it is unclear whether such an association exists in a comprehensive sample of the general population. The objective of the present study was to examine the association between CRP level and BMD, and their dose-response relationship, in a large representative population-based sample.

METHODS

Data source.

We used data from the National Health and Nutrition Examination Survey (NHANES) conducted between 1999 and 2004. This cross-sectional survey used a multistage, stratified probability design to produce a representative sample of the civilian US population (24). NHANES participants were interviewed and underwent a standardized physical examination at the mobile examination centers, where dual x-ray absorptiometry (DXA) scans and blood samples were also obtained. The study sample included individuals age ≥20 years with BMD (total and subregions) measured by DXA scans.

Data collection.

DXA scan.

BMD was measured using whole-body DXA scans (Hologic QDR-4500A fan-beam densitometer with Hologic software version 8.26:a3*). BMD measurements were obtained for the total body, all extremities, head, and trunk. The trunk included left and right ribs, pelvis, thoracic spine, and lumbar spine (25). Subtotal BMD refers to BMD values excluding head values.

CRP level.

CRP concentrations were quantified by high-sensitivity latex-enhanced nephelometry on a BN II nephelometer (Dade Behring). Detailed descriptions of blood collection and processing, quality control and quality assurance, and the laboratory method can be found elsewhere (26).

Covariates.

Demographic information included age, sex, race/ethnicity, level of education, and poverty index. The poverty income ratio was computed as the ratio of family income to the poverty threshold as produced annually by the Census Bureau. NHANES collected information on smoking, physical activity, body measurements, medical conditions, medication use (including diuretics, antihypertensive agents, and over-the-counter analgesics and nonsteroidal antiinflammatory drugs and use of any prescription medications [over the past month]), and blood tests, including serum creatinine, lipids, and vitamin D levels.

Blood pressure was based on the mean of the second and third systolic and diastolic readings from the 3 blood pressure readings obtained in the mobile examination center. Participants who reported current use of antihypertensive or antidiabetic medication (insulin or oral agents) were considered to have high blood pressure or diabetes, respectively (27). Previous physician diagnosis of diabetes, stroke, coronary heart disease, congestive heart failure (CHF), chronic bronchitis (COPD), and arthritis was based on self-report during the household interview. The presence of arthritis was based on the following 2 questions: “Doctor ever said you had arthritis?” and “Which type of arthritis?”

Smoking history was assessed during the household interview for participants ages ≥20 years. Participants were classified as never smokers (<100 cigarettes in their lifetime), former smokers (≥100 lifetime cigarettes, not currently smoking), and current smokers (≥100 lifetime cigarettes, currently smoking). For former smokers, time since cessation (recency, in years) and for former and current smokers, smoking intensity (cigarettes per day) and duration (in years) were calculated from interview responses. Smoking was modeled using a comprehensive smoking index, which simultaneously accounts for intensity, duration, and time since cessation of smoking as previously described (28–31).

Alcohol intake was determined from the food frequency questionnaire administered to participants to assess their usual consumption over the past month (32). Physical activity was defined based on leisure time physical activity (33). Body mass index (BMI) was calculated as weight (kg)/height (m2), and participants were categorized as having normal weight (BMI <25), being overweight (BMI 25–30), or being obese (BMI >30).

Data were collected on prescription medication use during a 1-month period prior to the interview date. Briefly, during the household interview, participants were asked if they had taken a medication in the past month for which they needed a prescription, and those who answered “yes” were asked to show the interviewer the medication containers of all the products used. We included data on the use of thiazide diuretics, bisphosphonates, heparin, steroids (including inhaled glucocorticoids), anticonvulsants (hydantoin, carbamazepine, barbiturates, benzodiazepines), antirheumatic drugs, and statins.

Details about the procedures of all the laboratory tests are reported elsewhere. Serum glucose concentration was measured using an enzymatic reaction (Cobas Mira assay; Roche). Serum creatinine level (mg/dl) was assessed using the Jaffe rate method (kinetic alkaline picrate). Glomerular filtration rate (GFR) was estimated using the abbreviated Modification of Diet in Renal Disease study equation (34).

Measurements of serum 25-hydroxyvitamin D (25[OH]D) were performed in NHANES 2000–2006, at the National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA, using DiaSorin RIA kit 25(OH)D assays. The results were available for NHANES 2001–2006.

Statistical analysis.

Summary statistics are presented as the mean ± SD or the median and interquartile range for continuous measures, and as frequencies for discrete variables. We followed NHANES guidelines for analyses of DXA scan data, which have been described elsewhere (25). The serum concentrations of CRP were modeled as a continuous variable or categorized into sex-specific quintiles. The association between CRP concentrations and BMD was evaluated according to a prespecified analysis plan using multivariate linear regression models adjusting for potential confounders, with total BMD as the dependent variable and CRP as the independent variable. Because men and women differ with regard to CRP and BMD, a priori separate models were fit for men and women. Models were adjusted for age, race/ethnicity, poverty income ratio, smoking, alcohol intake, physical activity, BMI, GFR, menopause, and hormone use (use of birth control pills and oral hormone replacement therapy). A priori, we tested for effect modification using interaction terms with age, menopause, race, BMI, and smoking in models with total BMD adjusting for potential confounders.

Multivariate analyses included further adjustment for comorbidities, including diabetes, hypertension, arthritis, COPD, cardiovascular disease, CHF, and stroke, as well as medication use, including thiazide diuretics, bisphosphonates, heparin, steroids, anticonvulsants, antirheumatic drugs, and statins. We also adjusted for vitamin D status in a sample subgroup or which serum levels of 25(OH)D were available. The analysis was replicated for specific subregions, including all extremities, ribs, trunk, head, thoracic spine, lumbar spine, pelvis, and subtotal BMD. The statistical analysis was performed with Stata 11 software, accounting for the complex survey design.

RESULTS

The final study sample included 10,475 participants who had undergone a DXA scan (Table 1). Of those, 53% were Caucasian, 22% were Mexican American, 18% were African American, and 7% were of other races. Their mean age was 51 years (range 20–85 years). Fifty percent of sample participants were women, 47% of whom were postmenopausal. Compared to women, men had higher levels of physical activity, a higher alcohol intake, and were more likely to be smokers (current or former). In general, men had higher BMD as previously described (35) and lower CRP concentrations than women. In both men and women, BMD at the right side extremities was higher than BMD at the left side extremities (Figure 1).

Table 1. Characteristics of the study sample*
 Men (n = 5,261)Women (n = 5,214)
  • *

    Except where indicated otherwise, values are the number (%). CSI = comprehensive smoking index; BMI = body mass index; METS = metabolic equivalents; GFR = glomerular filtration rate; IQR = interquartile range; CRP = C-reactive protein; COPD = chronic obstructive pulmonary disease; CHF = congestive heart failure; CVD = cardiovascular disease.

  • Self-reported rheumatoid arthritis, osteoarthritis, and other types of arthritis.

  • Hydantoin, carbamazepine, barbiturate, benzodiazepines.

Age, mean ± SD years51 ± 1851 ± 19
Race/ethnicity  
 Caucasian2,837 (54)2,696 (52)
 African American966 (18)983 (19)
 Mexican American1,117 (21)1,159 (22)
 Other341 (7)376 (7)
Poverty income ratio, mean ± SD2.77 ± 1.592.59 ± 1.61
Smoking status  
 Never2,335 (44)3,293 (63)
 Former1,729 (33)1,053 (20)
 Current1,197 (23)868 (17)
CSI, mean ± SD0.76 ± 1.200.54 ± 1.04
Alcohol intake, mean ± SD servings/day0.77 ± 1.800.29 ± 1.08
BMI, mean ± SD kg/m227.95 ± 5.3928.73 ± 6.86
Physical activity, mean ± SD METS83.36 ± 18048.67 ± 102
GFR, median (IQR) ml/minute/1.73 m2 of body surface area89.86 (76.81–104.71)89.83 (74.73–107.32)
CRP, median (IQR) mg/dl0.18 (0.08–0.38)0.28 (0.11–0.63)
Serum vitamin D, median (IQR) ng/ml22 (17–28)21 (15–28)
Comorbidities  
 Stroke188 (3.57)167 (3.20)
 Hypertension1,785 (34.37)1,796 (35.13)
 Arthritis1,177 (22.37)1,652 (31.68)
 COPD214 (4.07)435 (8.34)
 Diabetes mellitus548 (10.42)521 (9.99)
 CHF188 (3.57)144 (2.76)
 CVD537 (10.21)348 (6.67)
Medications  
 Steroids147 (2.79)186 (3.57)
 Bisphosphonates19 (0.36)160 (3.07)
 Thiazides346 (6.58)581 (11.14)
 Anticonvulsants176 (3.35)253 (4.85)
 Statins650 (12.36)526 (10.09)
 Antirheumatic drugs23 (0.44)38 (0.73)
Figure 1.

Bone mineral density (BMD) in the extremities by quintiles of C-reactive protein (CRP) concentration. BMD values were higher in men than in women and higher in the right side extremities than in the left side extremities in both groups. There was a significant inverse dose-dependent association between CRP concentration quintiles and BMD in the extremities, independently of age, race/ethnicity, poverty income ratio, smoking, alcohol intake, physical activity, body mass index, glomerular filtration rate, menopause, and hormone use (in women). Values are the adjusted mean (95% confidence interval).

Among both men and women, we found a significant inverse dose-dependent association between CRP concentration and total body BMD (Table 2). Men with CRP concentrations in the first quintile had a mean total body BMD of 1.186 gm/cm2 (95% confidence interval [95% CI] 1.178–1.195), compared to 1.167 gm/cm2 (95% CI 1.159–1.175) for men with CRP concentrations in the fifth quintile (P for trend < 0.0001) (model 1). Women with CRP concentrations in the first quintile had a mean total body BMD of 1.089 gm/cm2 (95% CI 1.082–1.096), compared to 1.071 gm/cm2 (95% CI 1.062–1.080) for women with CRP concentrations in the fifth quintile (P for trend = 0.0005). No significant interactions were observed between CRP concentration and menopause (P = 0.09), age (P = 0.25 for men, P = 0.22 for women), race/ethnicity (P = 0.42 for men, P = 0.41 for women), BMI (P = 0.08 for men, P = 0.51 for women), and smoking (P = 0.16 for men, P = 0.8 for women).

Table 2. Total BMD by sex and by quintiles of CRP concentration in the study sample*
 Total BMD, adjusted mean (95% CI) gm/cm2
Model 1Model 2Model 3§Model 4
  • *

    CRP = C-reactive protein; 95% CI = 95% confidence interval; IQR = interquartile range.

  • In model 1 (5,261 men, 5,214 women), total bone mineral density (BMD) was adjusted for age, race/ethnicity, poverty income ratio, smoking, alcohol intake, physical activity, body mass index, glomerular filtration rate, and menopause and hormone use (use of birth control pills and oral hormone replacement therapy) in women.

  • In model 2 (5,193 men, 5,113 women), total BMD was further adjusted for comorbidities (diabetes mellitus, arthritis, hypertension, stroke, cardiovascular disease, congestive heart failure, and chronic obstructive pulmonary disease).

  • §

    In model 3 (5,193 men, 5,113 women), total BMD was further adjusted for medications (thiazides, anticonvulsants, bisphosphonates, statins, heparin, antirheumatic drugs, and steroids).

  • In model 4 (3,515 men, 3,449 women), total BMD was further adjusted for serum vitamin D levels.

Men    
 CRP quintile, median (IQR) mg/dl    
  1, 0.04 (0.03–0.05)1.186 (1.178–1.195)1.186 (1.177–1.194)1.185 (1.177–1.194)1.195 (1.186–1.205)
  2, 0.10 (0.08–0.11)1.182 (1.172–1.191)1.181 (1.171–1.191)1.181 (1.171–1.190)1.193 (1.181–1.204)
  3, 0.18 (0.16–0.21)1.182 (1.172–1.192)1.181 (1.170–1.191)1.181 (1.171–1.191)1.194 (1.182–1.207)
  4, 0.33 (0.28–0.38)1.173 (1.164–1.182)1.173 (1.165–1.182)1.173 (1.164–1.182)1.183 (1.173–1.193)
  5, 0.78 (0.58–1.37)1.167 (1.159–1.175)1.167 (1.159–1.174)1.167 (1.160–1.175)1.183 (1.173–1.193)
 P for trend<0.00010.00010.00020.011
Women    
 CRP quintile, median (IQR) mg/dl    
  1, 0.05 (0.03–0.07)1.089 (1.082–1.096)1.088 (1.081–1.094)1.088 (1.081–1.095)1.100 (1.091–1.109)
  2, 0.16 (0.13–0.20)1.079 (1.071–1.087)1.078 (1.070–1.086)1.078 (1.071–1.086)1.088 (1.078–1.097)
  3, 0.31 (0.27–0.36)1.078 (1.071–1.085)1.076 (1.069–1.083)1.076 (1.070–1.083)1.083 (1.074–1.092)
  4, 0.57 (0.49–0.68)1.071 (1.063–1.077)1.070 (1.063–1.077)1.070 (1.062–1.076)1.077 (1.069–1.085)
  5, 1.29 (1.02–1.89)1.071 (1.062–1.080)1.070 (1.062–1.080)1.071 (1.062–1.079)1.077 (1.066–1.088)
 P for trend0.00050.00060.00060.0002

Among men and women, there was a significant inverse dose-dependent association between quintiles of CRP concentration and BMD at all the extremities (P for trend < 0.0001 for men, P for trend < 0.004 for right arm and right leg for women) (Figure 1), ribs (P for trend = 0.001 for men, P for trend < 0.007 for women), trunk (P for trend = 0.03 for men, P for trend < 0.001 for women) (Figure 2), and subtotal BMD (P for trend = 0.00003 for men, P for trend = 0.0008 for women). There was an inverse dose-dependent association between CRP concentration and BMD at the lumbar spine in women (P for trend = 0.005) but not in men (P for trend = 0.73) (Figure 3). There was no association between CRP concentration and BMD at the head, pelvis, and thoracic spine (P > 0.05).

Figure 2.

BMD in the trunk by quintiles of CRP concentration. BMD values were higher in men than in women. There was a significant inverse dose-dependent association between CRP concentration quintiles and BMD in the trunk, independently of age, race/ethnicity, poverty income ratio, smoking, alcohol intake, physical activity, body mass index, glomerular filtration rate, menopause, and hormone use (in women). Values are the adjusted mean (95% confidence interval). See Figure 1 for definitions.

Figure 3.

BMD in the lumbar spine by quintiles of CRP concentration and sex. Triangles represent men; diamonds represent women. BMD values were higher in men than in women. There was a significant inverse dose-dependent association between CRP concentration quintiles and BMD in the lumbar spine for women (P for trend = 0.005) but not for men (P for trend = 0.73), adjusted for age, race/ethnicity, poverty income ratio, smoking, alcohol intake, physical activity, body mass index, glomerular filtration rate, menopause, and hormone use (in women). Values are the adjusted mean (95% confidence interval). See Figure 1 for definitions.

Multivariate analyses with further adjustment for comorbidities and medication use revealed similar results (Table 2). The results remained largely unchanged if participants taking bisphosphonates were excluded. Subgroup analyses with further adjustment for vitamin D levels showed similar results in both groups (P for trend = 0.011 for men, P for trend = 0.0002 for women) (Table 2).

DISCUSSION

Our results indicate that among men and women in a large representative population-based sample, the CRP concentration was inversely and independently associated with BMD in a dose-dependent manner (extremities, ribs, trunk, subtotal BMD, and total body BMD). The associations were independent of comorbidities, medication use, and other potential confounders, including serum vitamin D levels. There were no significant interactions between the CRP concentration and menopause, age, race/ethnicity, BMI, and smoking.

The results of the present study are consistent with those of 2 previous studies demonstrating an inverse association between CRP concentration and BMD in Korean women and Southern Tasmanian older adults (18, 19). Koh et al investigated the association between CRP concentration and BMD measured with DXA at the femoral neck and lumbar spine in 4,693 healthy Korean women ages ≥30 years. Women with osteopenia and/or osteoporosis had higher serum CRP levels than did those with normal BMD after adjusting for age, BMI, and years since menopause. Compared to women in the lowest quintile of CRP concentration, both before and after menopause, those in the highest quintile were more likely to have osteoporosis and/or osteopenia and significantly lower BMD at the femoral neck but not at the lumbar spine (18). This is in contrast to our results of an independent inverse association between lumbar spine BMD and CRP concentration, which may be explained by differences in genetic factors and environmental exposures between the 2 samples, as the present study is based on a representative sample of the US population, while Koh et al recruited healthy Korean women from a health promotion center, raising the possibility of selection bias.

Ding et al evaluated the association between inflammation markers (including CRP, IL-6, and TNFα) and BMD as measured by DXA scans at baseline and 2.9 years later in a sample of 168 older adults. Total change in BMD was associated with baseline levels of all inflammation markers, as well as change in the CRP and IL-6 levels, independent of age, weight, height, smoking, BMD, and comorbidities. Changes in other BMD sites were significantly predicted by IL-6 level (hip and spine) and TNFα level (spine only) (19). However, the sample was small and limited to older adults.

Ganesan et al evaluated the association between the CRP concentration and total hip BMD in 2,807 women age ≥65 years from NHANES-III (20). The CRP concentration was inversely associated with total hip BMD in bivariate analysis, although the association was not significant in multivariate analysis. However, the effect size reported by Ganesan et al (20) was similar to the effect size observed in the present study when the CRP concentration was dichotomized as detectable versus undetectable (results not shown). Thus, our data reported here are consistent with the results of the study by Ganesan et al.

Our results should be contrasted with those of other studies that found no correlation between CRP concentration and BMD (20–23). Of note, other studies showing no association between CRP concentration and BMD used different methods to assess BMD (21–23). For example, a recent study conducted among 184 healthy postmenopausal women showed no association between CRP concentration and trabecular bone measured as volumetric BMD with peripheral quantitative computed tomography (22). An Australian study conducted among 444 women age ≥65 years found no correlation between CRP concentration and BMD as measured by DXA (DPX-L; Lunar) (21). Schett et al observed no association between CRP concentration and BMD assessed with quantitative ultrasound of the heel among 683 participants age 45–79 years in the Bruneck Study; however, an inverse association was observed in a small subgroup of subjects with CRP levels >7.5 mg/liter (23).

Fracture risk is strongly related to BMD. Interestingly, 2 studies showed a positive association between CRP concentration and risk of fracture (21, 23), which was not confirmed in a more recent study (36). The association between CRP concentration and fracture risk was evaluated prospectively in the Bruneck Study in 906 participants, of whom 69 experienced nontraumatic fractures. While the CRP level did not correlate with bone density measures, the adjusted relative risk of a nontraumatic fracture was 9.4 (95% CI 3.6–24.8) in the highest tertile of CRP concentration compared with the lowest tertile (23). Another study showed that for each SD increase in log-transformed CRP concentration, there was a 24–32% increase in fracture risk, depending on the site-specific BMD used in the model, independent of BMD or prevalent fracture, among women age ≥65 years followed up over a median of 5.5 years (21). This finding was not confirmed in a prospective cohort study among older men and women age 70–79 years followed up over a mean of 5.8 years. The incidence rate of fracture increased with increasing levels of cytokines, but the trend was not statistically significant for CRP and IL-6 levels, although a composite measure of elevated inflammation markers (i.e., combining information on CRP, IL-6, and TNFα) identified persons at high risk of fracture (36).

Our results indicate that in the general US population, differences in total body BMD between extreme quintiles of CRP concentration are 2–3%, depending on the site-specific BMD used. These results imply that CRP is linked with risk of fracture in the general population. Several lines of evidence suggest that these differences have clinical and public health relevance. For example, the differences are similar to those seen across vitamin D concentrations (37) or to differences seen after vitamin D and calcium supplementation (38). Note that there was no association between vitamin D levels and CRP concentration in this sample.

BMD may be influenced by skeletal load, which is determined by muscle forces as well as by gravity from weight bearing (39). Distinct skeletal subregions vary in the extent to which they are weight bearing, and, for example, the upper extremities may be considered non–weight bearing. Weight bearing is particularly important in maintaining BMD at the pelvis and legs (40). The results observed for 3 skeletal subregions examined (i.e., head, pelvis, and thoracic spine) do not appear to be related to weight bearing. Further, we included all extremities to compare between weight bearing and non–weight bearing across extremities, given that the dominant side in most people is the right side. While the BMD was higher in the right extremities compared with the left extremities in this population-based sample, the significant inverse association between BMD and CRP concentration was similar for all extremities. With regard to the lumbar spine, there was an inverse association in women and no association in men. This difference might be explained by degenerative changes in the spine with age, such as osteoarthritis with osteophyte formation, which may mask BMD loss (41). The lack of association between BMD and CRP concentration at 3 of the skeletal subregions examined (i.e., head, pelvis, and thoracic spine) is intriguing. This may be related to regional bone characteristics, which may also explain why bone disorders may have tropism for certain regions, for example, the skull in Paget's disease (42). Such differences may also protect against the effect of inflammation on bone metabolism.

Important strengths of the present study include the large population-based sample, which allowed for dose-response analyses with high precision, accounting for many potential confounders of the association, including serum vitamin D concentrations. Furthermore, the estimates are generalizable to the US population. This is a comprehensive study that has several unique strengths, as the sample included individuals age ≥20 years from several racial/ethnic groups, and men and women were evaluated separately. BMD was measured by total body DXA scans, and we analyzed whole-body BMD and BMD of all body skeletal subregions. However, unmeasured confounding cannot be excluded. Also, temporality or causality cannot be established due to the cross-sectional nature of the study design. Nevertheless, the results are robust and suggest that higher CRP levels are associated with lower bone mass in healthy men and women, even with CRP levels below a level normally considered clinically relevant. Most important, our results suggest that in the general population a subclinical inflammatory process, even at low levels, appears to uncouple bone formation from resorption in favor of excessive bone resorption.

In conclusion, among men and women in a large representative population-based sample, the CRP level is inversely and independently associated with BMD (total and most subregions) in a dose-dependent manner. If these results are confirmed in longitudinal studies, strategies to modify systemic inflammation should be tested to see whether they modify the long-term risk of bone loss.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. de Pablo 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 conception and design. De Pablo, Buckley.

Acquisition of data. De Pablo.

Analysis and interpretation of data. De Pablo, Cooper, Buckley.

Ancillary