Inflammatory markers and the risk of hip fracture: The women's health initiative
Cytokines play a major role in bone remodeling in vitro and in animal models, with evidence supporting the involvement of inflammatory markers in the pathogenesis of osteoporosis. However, less is known about the longitudinal association of inflammatory markers with hip fracture. We tested whether high receptor levels of proinflammatory cytokines are associated with an increased risk of hip fracture in older women. We used a nested case-control study design from the Women's Health Initiative Observational Study (WHI-OS) and selected 400 cases with physician-adjudicated incident hip fractures and 400 controls matched on age, race, and date of blood draw. Participants were chosen from 39,795 postmenopausal women without previous hip fractures, not using estrogens or other bone-active therapies. Incident hip fractures (median follow-up 7.1 years) were verified by review of radiographs and confirmed by blinded central adjudicators. Hip fractures with a pathological cause were excluded. In multivariable models, the risk of hip fracture for subjects with the highest levels of inflammatory markers (quartile 4) compared with those with lower levels (quartiles 1, 2, and 3) was 1.43 (95% confidence interval [CI], 0.98–2.07) for interleukin-6 (IL-6) soluble receptor (SR), 1.40 (95% CI, 0.97–2.03) for tumor necrosis factor (TNF) SR1, and 1.56 (95% CI, 1.09–2.22) for TNF SR2. In subjects with all three markers in the highest quartile, the risk ratio of fracture was 2.76 (95% CI, 1.22–6.25) in comparison with subjects with 0 or 1 elevated marker(s) (p trend = 0.018). Elevated levels of inflammatory markers for all three cytokine-soluble receptors were associated with an increased risk of hip fractures in older women. Future clinical trials should test whether interventions to decrease inflammatory marker levels reduces hip fractures. © 2012 American Society for Bone and Mineral Research.
Elevated levels of proinflammatory markers (ie, cytokines) have been shown to be associated with an increased risk of adverse outcomes, including type 2 diabetes,1 mortality,2 declines in both physical3 and cognitive function,4 dementia,5 and cardiovascular disease (CVD).6, 7 Interleukin-6 (IL-6), interleukin-1 (IL-1), and tumor necrosis factor-alpha (TNF-α) are cytokines that play a major role in bone remodeling, with several in vitro and rodent studies showing the involvement of inflammatory markers in the pathogenesis of osteoporosis.8, 9 Proinflammatory markers have been shown to act on mesenchymal stem cells and osteoclast precursors to enhance osteoclast-mediated bone resorption. In the first physiological pathway, these cytokines bind to stromal cells and increase the expression of receptor activator of NF-κB ligand (RANKL) and macrophage-colony stimulating factor (M-CSF) and decrease osteoprotegerin (OPG) production, resulting in increased activation of osteoclasts.10 In the second physiological pathway, estrogen deficiency results in cytokine-mediated osteoclast activation.11, 12
The association between proinflammatory markers and hip fractures is uncertain. A prior prospective study showed that elevated inflammatory markers are a risk factor for incident fractures.13 However, this prior study included all nontraumatic fractures (n = 156) and did not have enough power to assess this association for hip fractures (n = 39). Hip fractures contribute the greatest to morbidity and mortality among all other osteoporotic fractures.14 The 1-year mortality rate after a hip fracture in women is estimated to range from 17% to 22%.15, 16
We conducted a nested case-control study from the Women's Health Initiative Observational Study (WHI-OS) among 400 cases with physician-adjudicated incident hip fractures and 400 controls matched on age, race, and date of blood draw. We tested whether high receptor levels of proinflammatory cytokines are associated with an increased risk of hip fracture in older women. We focused specifically on the soluble receptors for inflammatory markers as opposed to the markers themselves for the following reasons: in our prior study13 this association was particularly strong for the soluble receptors of TNF; in addition, increases in TNF-α and IL-6 are usually transient, whereas elevations of soluble receptors for these cytokines appear to be more constant.17 Prior research suggests that antigens may induce shedding of soluble cytokine receptors in an attempt to weaken the inflammatory response. Thus, elevated levels of soluble receptors may represent a more prolonged or severe inflammatory state.18, 19
Subjects and Methods
The WHI-OS is a prospective cohort study that enrolled 93,676 women aged 50 to 79 years from 1994 to 1998 at 40 U.S. clinical centers.20 Women were eligible if they were postmenopausal, unlikely to move or die within 3 years, not enrolled in the WHI Clinical Trials, and not currently participating in any other clinical trial. The study was approved by Human Subjects Review Committees at each participating institution, and all participants provided written informed consent.
Follow-up and outcome ascertainment
Women were sent questionnaires annually to report the occurrence of any hospitalization and a wide variety of outcomes, including fractures. Follow-up time for hip fractures ranged from 0.7 to 9.3 years as of August 2004 with a median duration of 7.1 years. At that time, 3.7% of participants had withdrawn or were lost to follow-up and 5.3% had died. Hip fractures were verified by review of radiology reports and confirmed by blinded central adjudicators.21 Hip fractures with a possible or confirmed pathological cause resulting from bone tumors, Paget's disease, bone and joint prosthesis, or surgical manipulation were excluded.
Nested case-control study design
The present analyses use a nested case-control design within the prospective design of the WHI-OS. Participants were excluded if they had a prior history of hip fracture at baseline, were currently taking hormones or had taken them up to 1 year prior to enrollment, or at baseline were taking androgens, selective estrogen receptor modulators, anti-estrogens, or other osteoporosis treatments (bisphosphonates, calcitonin). Women without sufficient serum stored or with unknown ethnicity were also excluded, leaving a final study group of 39,795 eligible participants. From the eligible women, a total of 404 incident hip fractures occurred. We randomly selected 400 incident hip fractures to comprise the case group. From the remaining without hip fractures, one control per case was selected with individual matching by age at screening (± 1 year), race/ethnicity, and date of blood draw (± 120 days).
Current use of prescription and over-the-counter medications was recorded by clinic interviewers by direct inspection of containers. Prescription names were entered into the WHI database and assigned drug codes using Medispan software (First DataBank, Inc., San Bruno, CA, USA).
Vitamin and mineral supplements, including usual current supplement doses of elemental calcium and Vitamin D preparations, taken at least twice weekly for the prior 2 weeks, were entered directly from information on container labels as described in the previous paragraph. Dietary intakes of calcium and Vitamin D were also assessed using a semiquantitative food frequency questionnaire.22 Total calcium and Vitamin D intake was defined as the sum of diet and supplements.
Questionnaires ascertained information on date of birth, race/ethnicity, history of fracture after age 55 years, parental history of hip fracture, diabetes treatment, rheumatoid arthritis (RA), smoking history, self-rated health status, alcohol consumption, corticosteroid use, nonsteroidal anti-inflammatory drug (NSAID) use ≥2 years, and total number of falls since last follow-up. Physical activity was classified on the basis of frequency and duration of walking and mild, moderate, and strenuous recreational activities in the prior week. Kilocalories of energy expended in a week was calculated as the metabolic equivalent (MET) score (kcal hours/week/kg).23 Physical function was measured using the 10-item Rand-36 physical function scale (0–100), with higher scores indicating better physical function.24 We compared women with a score >90 versus ≤90; this cutoff corresponds to the median score. A frailty score was computed and included self-reported muscle weakness, impaired walking, exhaustion, low physical activity, and unintended weight loss between baseline and 3 years of follow-up.25
Weight was measured on a balance beam scale with the participant dressed in indoor clothing without shoes. Height was measured using a wall-mounted stadiometer. Body mass index (BMI) was calculated as weight (kg)/height2 (m2).
A 12-hour fasting blood sample was obtained at the baseline visit, processed, and stored at −80°C according to strict quality control procedures.26 Stored serum samples were sent to testing laboratories where laboratory personnel were blinded to case-control status for all measurements.
Soluble receptors of IL-6 (IL-6 SR) and TNF (TNF SR1 and TNF SR2) were measured in duplicate with ELISA kits (R&D Systems, Minneapolis, MN, USA) at the University of Vermont. The detectable limits for the IL-6 SR (using the DR600 kit), TNF SR I (using the DRT100kit), and TNF SR II (using the DRT200 kit) were 6.5, 3.0, and 1.0 pg/mL, respectively. The interassay coefficients of variation (CVs) of IL-6 SR, TNF SR1, and TNF SR2 were 12.5% to 14.8%, 6.7% to 10%, and 5.6% to 6.2%, respectively. Sex steroid hormones were measured at the Reproductive Endocrine Research Laboratory at the University of Southern California (USC), a WHI-designated core laboratory. Estradiol and testosterone concentrations were quantified using sensitive and specific radioimmunoassays (RIAs) following organic solvent extraction and celite column partition chromatography.27–30 For estradiol, the intraassay and interassay CVs were 7.9% and 8% to 12%, respectively and for testosterone, 6% and 10% to 12%, respectively. Bioavailable hormone concentrations were calculated using mass action equations.31–33 Sex hormone binding globulin (SHBG) was measured using a solid-phase two-site chemiluminescent immunoassay.34 The intraassay and interassay CVs were 4.1% to 7.7% and 5.8% to 13%, respectively. Serum levels of cystatin-C were measured with the Dade Behring BN-II nephelometer and Dade Behring reagents (Ramsey, MN, USA) using a particle enhanced immunonephelometric assay at Medical Research Laboratories International (Highland Heights, KY, USA). Serum cross-linked C-telopeptide (CTX) and amino-terminal propeptide of type I procollagen (PINP) were measured by immunoassay (Synarc Inc., Lyon, France). Serum 25-hydroxvitamin D [25(OH)D] was measured by using radioimmunoassay with DiaSorin reagents (Diasorin, Stillwater, MN, USA). The sensitivity of the 25(OH)D assay was 1.5 ng/mL. Interassay CVs were 11.7%, 10.5%, 8.6%, and 12.5% at 5.6, 22.7, 33.0, and 49 ng/mL of 25(OH)D.
Baseline characteristics were compared between hip fracture cases and matched controls, using McNemar's test for categorical variables and paired t tests for continuous data. We reported the median and interquartile range (IQR) for variables that were not normally distributed, and performed nonparametric analyses using the Wilcoxon signed-rank test. We assigned cytokine soluble receptor concentrations to quartile categories based on the distribution within the controls. We hypothesized that the cases would be disproportionately in the group with higher cytokine levels. The reason for this is that controls provide the expected concentration of inflammatory markers in the population that gave rise to the cases. The complexity and interrelatedness of cytokines involved makes it unlikely that one biomarker would capture all of the risk information. Therefore, a composite measure of inflammation that combines the number of soluble cytokine receptors in the highest quartile for IL-6 and TNF-α was used to determine hip fracture risk. High levels of two or more inflammatory markers more likely represent systemic inflammation than a high level of just one inflammatory marker.35, 36 This composite measure was predefined based on our prior work.13 To further assess the potential for confounding, participant characteristics were compared across a number of inflammatory markers in the highest quartile. The dose-response associations for the number of high inflammatory markers and participants characteristics were evaluated using the Jonckheere Terpstra and Cochrane-Armitage tests of trend, and by treating number of high inflammatory markers as a continuous variable.
For multivariable models, the associations were assessed using conditional logistic regression models to account for the matched case-control design. The odds ratio was used as an approximation of the risk ratio, based on the relative rarity of the outcome incident hip fractures. To examine the impact of these biomarkers individually, we compared women with the lowest cytokine receptor concentrations quartile 1 (Q1) to women with higher concentrations (quartiles 2, 3, and 4), and tested for dose-response relationships. Women in the top quartile (Q4) of cytokine soluble receptors appeared to be the most at risk for hip fracture; and thus were compared to all other women (Q123). Associations were then examined with adjustment for BMI, parental history of hip fracture, previous fractures, self-reported health, treated diabetes, RA, physical activity, smoking, alcohol use, total calcium and vitamin D intake, NSAID use, and corticosteroid use. Further multivariable models compared women with two or three inflammatory markers in the highest quartile to women with ≤1 inflammatory markers in the highest quartile. To investigate mechanisms by which inflammatory markers might be associated with hip fractures, we added the following variables individually to the base model to determine if they mediated this association: frailty score, RAND-36 physical function scale, number of falls, sex-steroid hormones, cystatin-C, bone turnover (CTX and PINP), and 25(OH)D. We then adjusted for all variables simultaneously (except for frailty, which we hypothesized would be correlated with measures of physical function because both rely on the RAND Short Form-36 physical function scale). We also determined the associations between potential mediators and hip fracture adjusted for the base analysis and inflammatory marker levels. This analysis was performed to better understand the directionality (augmentation or attenuation) of potential mediation. The variance inflation factor (VIF) was used to assess multicollinearity in multivariable models. VIF values were <2.5 for all independent variables, indicating that multicollinearity was likely not present in this study.
The vast majority of hip fractures occurred among whites. Thus, a secondary analysis examining our hypothesis was performed among whites only.
The mean age of the subjects was 71 ± 6.2 years and 95% were white (Table 1). Hip fracture cases had significantly lower BMI and physical activity. They were more likely to report corticosteroid use and current smoking compared to controls. In addition, serum levels of 25(OH)D, bioavailable estradiol, and bioavailable testosterone were significantly lower among cases. Conversely, serum cystatin-C levels were significantly higher among cases. TNF SR2 (p = 0.04) concentrations were significantly higher among cases versus controls. IL-6 SR (p = 0.28) and TNF SR1 (p = 0.07) concentrations did not differ between cases and controls.
Table 1. Characteristics by Case-Control Status and Across Number of High Inflammatory Markers in the Control Group
|White, n (%)||380 (95.0)||380 (95.0)||>0.99||196 (92.9)||98 (97.0)||61 (96.8)||24 (100)||0.58|
|Age at baseline >70 years, n (%)||132 (33.0)||132 (33.0)||>0.99||129 (61.1)||68 (67.3)||50 (79.4)||20 (80.3)||<0.01|
|BMI, kg/m2, n (%)|| || ||<0.01|| || || || || |
| <25||144 (36.1)||193 (48.6)|| ||87 (41.4)||36 (35.6)||14 (22.2)||6 (25.0)||<0.01|
| 25–30||150 (37.6)||127 (32.0)|| ||80 (38.1)||38 (37.6)||24 (38.1)||8 (33.3)|| |
| ≥30||105 (26.3)||77 (19.4)|| ||43 (20.5)||27 (26.7)||25 (39.7)||10 (41.7)|| |
|Corticosteroid use, n (%)||4 (1.0)||16 (4.0)||0.01||2 (1.0)||0 (0.0)||2 (3.2)||0 (0.0)||0.59|
|NSAID use ≥2 years, n (%)||81 (21.7)||83 (20.7)||0.73||43 (20.4)||22 (21.8)||18 (28.6)||4 (16.7)||0.44|
|RAND 36–Physical functioning >90, n (%)||117 (30.1)||84 (21.8)||0.01||70 (34.3)||30 (30.0)||12 (19.7)||4 (17.4)||0.02|
|Frailty, n (%)||66 (16.5)||89 (22.3)||0.04||30 (14.2)||16 (15.8)||12 (19.1)||8 (33.3)||0.07|
|General health status, fair/poor, n (%)||42 (10.6)||61 (15.3)||0.05||17 (8.3)||13 (13.0)||8 (12.9)||4 (17.4)||<0.01|
|Treated diabetes, n (%)||19 (4.8)||24 (6.0)||0.43||6 (2.8)||5 (5.0)||5 (7.9)||3 (12.5)||0.02|
|RA, n (%)||23 (5.8)||28 (7.0)||0.47||12 (5.7)||4 (4.0)||6 (9.5)||1 (4.2)||0.68|
|Alcohol lifetime consumption, n (%)|| || ||0.61|| || || || ||0.09|
| Non-drinker||70 (17.6)||58 (14.6)|| ||34 (16.2)||13 (12.9)||17 (27.9)||6 (25.0)|| |
| Past drinker||80 (20.2)||89 (22.4)|| ||44 (21.0)||17 (16.8)||15 (24.6)||4 (16.7)|| |
| <1 drink per day||205 (51.6)||212 (53.3)|| ||107 (51.0)||56 (55.5)||28 (45.9)||13 (54.2)|| |
| ≥1 drinks per day||42 (10.6)||39 (9.8)|| ||25 (11.9)||15 (14.8)||1 (1.6)||1(4.2)|| |
|Current smoker, n (%)||10 (2.5)||36 (9.1)||<0.01||8 (3.9)||1 (1.0)||0 (0)||1 (4.2)||0.51|
|History of fracture at age ≥55 years, n (%)||82 (20.5)||96 (24.0)||0.24||44 (20.9)||19 (18.8)||12 (19.1)||7 (29.2)||0.73|
|Parental history of hip fracture, n (%)||64 (16.0)||80 (20.0)||0.14||31 (14.7)||18 (17.8)||13 (20.6)||2 (8.3)||0.76|
|Physical activity, MET, h/week, median (IQR)||11 (4–19)||7 (2–15)||<0.01||12 (4–20)||11 (4–23)||5 (1–15)||5 (1–11)||<0.01|
|Total number of falls at follow-up, median (IQR)||2 (0–4)||1 (1–3)||0.82||1 (0–4)||2 (1–4)||2 (0–4)||1.5 (0.5–3)||0.89|
|Total vitamin D intake, IU/d, median (IQR)||321 (127–561)||310 (120–541)||0.47||284 (112–560)||452 (140–561)||427 (188–656)||104 (172–454)||0.29|
|Serum 25(OH)D, ng/mL (mean ± SD)||23.9 ± 7.2||22.4 ± 8.1||0.01||23.8 ± 6.9||24.2 ± 6.3||24.4 ± 7.5||21.3 ± 8.2||0.51|
|Total calcium intake, mg/d (mean ± SD)||1167 ± 684||1072 ± 694||0.05||1167 ± 718||1197 ± 667||1216 ± 620||925 ± 600||0.51|
|TNF SR1, pg/mL, median (IQR)||1567 (1370–1838)||1623 (1376–1956)||0.07||—||—||—||—||—|
|TNF SR2, pg/mL, median (IQR)||2490 (2114–2844)||2551 (2158–3062)||0.04||—||—||—||—||—|
|IL-6 SR, pg/mL, median (IQR)||39223 (31013–47714)||40046 (30625–50302)||0.28||—||—||—||—||—|
|Bioavailable estradiol, pg/mL (mean ± SD)||7.5 ± 4.5||6.6 ± 4.3||<0.01||7.1 ± 4.3||6.8 ± 3.4||9.1 ± 5.2||11.1 ± 6.7||<0.01|
|Bioavailable testosterone, pg/mL (mean ± SD)||12.6 ± 7.0||10.9 ± 6.3||<0.01||12.4 ± 7.0||13.0 ± 7.8||11.8 ± 4.9||14.6 ± 7.5||0.52|
|SHBG, µg/dL (mean ± SD)||1.6 ± 0.8||1.8 ± 0.9||<0.01||1.6 ± 0.7||1.6 ± 0.9||1.5 ± 0.7||1.4 ± 0.6||0.09|
|Cystatin-C, mg/L (mean ± SD)||1.06 ± 0.2||1.10 ± 0.3||0.02||0.97 ± 0.14||1.03 ± 0.17||1.29 ± 0.36||1.34 ± 0.30||<0.01|
|PINP, ng/mL (mean ± SD)||49.6 ± 23.7||51.0 ± 23.0||0.42||49.1 ± 25.9||48.4 ± 17.3||53.1 ± 25.8||49.4 ± 21.1||0.49|
|CTX, ng/mL (mean ± SD)||0.41 ± 0.19||0.45 ± 0.22||0.02||0.41 ± 0.19||0.40 ± 0.15||0.44 ± 0.23||0.42 ± 0.20||0.47|
Participant characteristics varied by number of high inflammatory markers in the controls only (Table 1). Whites were more likely to have a greater number of high inflammatory markers than other ethnicities. A higher number of high inflammatory markers was positively (p trend < 0.05) associated with older age, higher BMI, and greater levels of bioavailable estradiol and serum cystatin-C. The positive association between bioavailable estradiol and number of high inflammatory markers was also independent of BMI (p trend = 0.003) (data not shown). There was also an inverse association for number of high inflammatory markers with higher physical activity and better self-reported health. SHBG levels decreased as the number of high inflammatory markers increased; however, this association was not significant (p trend = 0.09). Bone resorption marker levels, serum 25(OH)D, and bioavailable testosterone levels did not vary by number of high inflammatory markers.
Association of quartiles of inflammatory markers with hip fractures
There was a lack of a dose-response relationship between increasing quartiles of soluble cytokine receptors and hip fracture risk (Table 2). In addition, women in Q4 of cytokine-soluble receptor concentrations were compared to all other women in the cohort. In the unadjusted models, women in Q4 of IL-6 SR had 1.53 (95% CI, 1.10–2.14) times the risk of incident hip fracture compared to women in the lower IL-6 SR quartiles. This association was slightly attenuated and no longer significant in the multivariable model, relative risk (RR) = 1.43 (95% CI, 0.98–2.07). The association between TNF SR2 and hip fractures remained significant in the multivariable model (RR, 1.56; 95% CI, 1.09–2.22). There was no association between TNF SR1 and incident hip fractures in the multivariable model (RR, 1.40; 95% CI, 0.97–2.03).
Table 2. WHI Risk Ratios (95% CIs) of Hip Fracture According to Quartiles of Cytokine-Soluble Receptor Concentrations Among the Controls
| Unadjusted (n pairs = 399)||1.00||0.74 (0.48–1.15)||0.75 (0.50–1.11)||1.26 (0.84–1.89)||0.282||1.53 (1.10–2.14)|
| MV modelc (n pairs = 363)||1.00||0.62 (0.37–1.02)||0.64 (0.41–1.01)||1.04 (0.65–1.66)||0.812||1.43 (0.98–2.07)|
| Unadjusted (n pairs = 396)||1.00||0.94 (0.54–1.35)||1.08 (0.72–1.61)||1.31 (0.88–1.94)||0.088||1.33 (0.98–2.09)|
| MV modelc (n pairs = 360)||1.00||1.00 (0.63–1.59)||1.14 (0.72–1.81)||1.48 (0.91–2.40)||0.085||1.40 (0.97–2.03)|
| Unadjusted (n pairs = 398)||1.00||1.09 (0.73–1.64)||1.00 (0.66–1.51)||1.57 (1.06–2.33)||0.033||1.53 (1.12–2.09)|
| MV modelc (n pairs = 362)||1.00||1.09 (0.69–1.74)||0.96 (0.60–1.56)||1.57 (0.98–2.54)||0.076||1.56 (1.09–2.22)|
Number of high inflammatory markers and hip fracture
The risk of incident hip fracture was highest among women with 3 “high” levels (quartile 4) of inflammatory markers (Table 3). In the base analysis, women with two or three high inflammatory markers had 41% (95% CI, −11% to 124%) and 161% (95% CI, 41%–381%) increased risk of incident hip fracture compared to women with zero or one level (p trend = 0.002), respectively. Adjustment for potential mediators one at a time resulted in small attenuations and some augmentations of the association between inflammatory markers and incident hip fractures. After adjusting for estradiol, the increased risk of fracture for women with three high inflammatory markers compared to women with one or zero high inflammatory markers went from 161% to 200%. The most notable attenuation (decrease in 40 percentage points) in hip fracture risk occurred after adjusting for cystatin-C. Women with three high inflammatory markers had a 176% (95% CI, 22%–525%) increased risk of hip fracture compared to women with zero or one high inflammatory marker(s) in the final summary model after adjusting for variables in the base model and all potential mediators. There was also a positive linear trend (p trend = 0.018) between the number of high inflammatory markers and hip fractures in this model.
Table 3. Risk Ratios (95% CIs) of Hip Fracture, According to Number of High Inflammatory Markers
|Crude analysis (n pairs = 394)||1.00 (ref)||1.25 (0.84–1.84)||2.42 (1.43–4.09)||0.001|
|Base analysisa (n pairs = 358)||1.00 (ref)||1.41 (0.89–2.24)||2.61 (1.41–4.81)||0.001|
|Base analysisa + frailty score (n pairs = 358)||1.00 (ref)||1.44 (0.90–2.30)||2.55 (1.38–4.71)||0.002|
|Base analysisa + RAND 36–Physical Functioning (n pairs = 344)||1.00 (ref)||1.40 (0.87–2.23)||2.62 (1.41–4.85)||0.002|
|Base analysisa + total number of falls at follow-up (n pairs = 333)||1.00 (ref)||1.44 (0.89–2.32)||2.79 (1.46–5.31)||0.001|
|Base analysisa + bioavailable estradiol (n pairs = 348)||1.00 (ref)||1.53 (0.95–2.48)||3.00 (1.59–5.67)||<0.001|
|Base analysis + bioavailable testosterone (n pairs = 355)||1.00 (ref)||1.37 (0.86–2.20)||2.86 (1.51–5.41)||0.001|
|Base analysisa + SHBG (n pairs = 356)||1.00 (ref)||1.40 (0.87–2.24)||2.81 (1.48–5.35)||0.002|
|Base analysisa + cystatin-C (n pairs = 353)||1.00 (ref)||1.31 (0.77–2.20)||2.21 (1.12–4.36)||0.024|
|Base analysisa + PINP (n pairs = 345)||1.00 (ref)||1.48 (0.92–2.40)||2.80 (1.49–5.27)||0.001|
|Base analysisa + CTX (n pairs = 348)||1.00 (ref)||1.40 (0.87–2.26)||2.44 (1.32–4.53)||0.003|
|Base analysisa + 25(OH)D (n pairs = 357)||1.00 (ref)||1.41 (0.88–2.25)||2.61 (1.41–4.83)||0.002|
|Summary multivariable modelb (n pairs = 290)||1.00 (ref)||1.36 (0.74–2.52)||2.76 (1.22–6.25)||0.018|
Mediators and hip fracture
The associations between potential mediators and hip fracture incidence adjusted for the base analysis and inflammatory marker levels are shown in Table 4. SHBG and bioavailable testosterone concentrations were significantly associated with hip fracture (RR, 1.41; 95% CI, 1.13–1.75 and RR, 0.96; 95% CI, 0.94–0.99, respectively). Frailty, physical function, falls, bioavailable estradiol, cystatin-C, bone turnover markers, and 25(OH)D were not significantly associated with hip fracture.
Table 4. Association Between Potential Mediators and Hip Fracture Adjusted for the Base Analysis and Number of High Inflammatory Markers
|Frailty (n pairs = 358)||1.36||0.86–2.16|
|RAND 36–Physical Functioning >90 (n pairs = 344)||0.65||0.43–1.01|
|Total number of falls at follow-up (n pairs = 333)||0.97||0.92–1.02|
|Bioavailable estradiol, pg/mL (n pairs = 348)||0.96||0.92–1.00|
|Bioavailable testosterone, pg/mL (n pairs = 355)||0.97||0.94–0.99|
|SHBG, µg/dL (n pairs = 356)||1.41||1.13–1.75|
|Cystatin-C, ng/mL (n pairs = 353)||1.49||0.63–3.54|
|PINP, ng/mL (n pairs = 345)||1.00||0.99–1.01|
|CTX, ng/mL (n pairs = 348)||1.48||0.63–3.50|
|25(OH)D, ng/mL (n pairs = 357)||0.97||0.95–1.00|
Among whites only, inflammatory marker levels were a stronger predictor of hip fracture when compared to the entire cohort. Women with two or three high inflammatory marker levels had 220% (95% CI, 42%–623%) and 79% (95% CI, –1% to 222%) increased risk of hip fracture compared to women with zero or one high inflammatory marker(s) in the final summary model (p trend = 0.003), respectively.
In this prospective, nested-case-control study, we found that women in the highest quartile for all three of IL-6 SR, TNF SR1, and TNF SR2 had over two times the risk of incident hip fracture compared to women with one or zero inflammatory makers in the highest quartile. This risk is roughly equivalent to the risk associated with a 1 SD decrease in bone mineral density (BMD).37 These associations were independent of BMI, self-reported health, physical activity, parental history of fracture, history of fracture, treated diabetes, RA, calcium and vitamin D intake, NSAID and corticosteroid use, frailty, physical function, falls, sex hormones, cystatin-C, bone turnover markers, and 25(OH)D. These findings extend our previous findings13 on all clinical fractures to hip fractures, the most devastating consequence of osteoporosis.
Adjustment for potential mediators primarily augmented the association between number of inflammatory marker in the top quartile and incident hip fractures. We initially hypothesized that adjusting for bioavailable estradiol would attenuate this association because of data showing estrogens oppose the action of cytokines.38 However, in our cohort there was a positive association between bioavailable estradiol and number of inflammatory markers in the highest quartile independent of BMI. Similarly, estradiol levels have been shown to be positively correlated with proinflammatory markers in older women.39, 40 However, contrary to the association of inflammatory makers with hip fracture; bioavailable estradiol was lower among those with hip fractures compared to controls. Thus, negative confounding41 occurred after adjustment for bioavailable estradiol in consequence of the directionality of these associations. Conversely, cystatin-C (a marker for poor renal function) strongly attenuated the association between inflammatory markers and hip fracture. We observed a positive association between cystatin-C levels and number of inflammatory markers in the highest quartile. Several prospective cohort studies have identified an association between inflammatory makers and decline in kidney function.42–44 Though the biological mechanism has not been established, several hypotheses exist. In vivo studies have shown that glomerular injury can be induced directly by TNF-α,45, 46 or mediated by immune cells (ie, monocytes and macrophages).42 Conversely, reduced renal function may result in an increase of inflammatory markers in the blood.47 In this scenario, cystatin-C would not be in the causal pathway, and thus would likely be a confounder of the association between inflammatory markers and fracture. Poor renal function has also been identified as a risk factor for hip fractures in older women.48–50 In our study, cystatin-C concentrations were higher in cases versus controls. The directionality of these associations influenced the attenuation41 as a result of adjusting for cystatin-C.
There is an increased understanding and recognition of the role of the immune system in the development of osteoporosis.51 Multiple cytokines (proinflammatory and anti-inflammatory) and hormones interact to regulate osteoblast and osteoclast differentiation and activity. The balance in these systems plays an important role in the regulation of osteoblasts and osteoclasts. In addition, several longitudinal studies among older women have found an association between high levels of inflammatory makers and increased bone loss.52–55 However, in our analyses of the Health Aging and Body Composition Study (Health ABC) cohort, the association between inflammation and fractures was independent of BMD.13
TNF-α stimulates osteoclast differentiation in vitro and in vivo.56, 57 This can be accomplished indirectly through suppression of OPG expression and stimulation of receptor activator of NF-κB (RANK) in mesenchymal cells.10 TNF-α has also been shown to activate osteoclast precursors directly by acting synergistically with RANKL. This direct mechanism occurs as a result of estrogen deficiency leading to a marked increase of TNF-α.11 In our study, the increased risk of hip fracture was greater for those in the highest quartile for TNF SR2 (56%) and TNF SR1 (40%) compared to other participants. Our previous study found similar risks between soluble receptors for TNF-α and incident fractures.34 This suggests that the role of these biological markers in fracture etiology may be similar for hip and other types of fractures.
IL-6 may influence bone loss and osteoporosis.51 IL-6 is stimulated in response to parathyroid hormone (PTH) and other cytokines including TNF-α.58 IL-6 SR may enhance the biological activity of IL-6. In cell culture, IL-6 only stimulated osteoclastogenesis in the presence of IL-6 SR.59 Also, in transgenic mice, IL-6 SR may bind to IL-6 and increase its biological activity. In our study, the association between serum IL-6 SR and hip fractures was considerable, but not significant. The risk of hip fracture was 43% more likely among participants in the highest IL-6 SR quartile compared to other subjects, but it was not statistically significant.
Our study has a number of strengths. We examined multiple markers of inflammation in relation to incident hip fractures, the most serious consequence of osteoporosis. We also adjusted for many potential confounders, eliminated hormone users from analysis, and explored several mechanisms of potential mediation underlying this association in order to focus more carefully on this group. There were several limitations in our study. First, BMD was only measured in three WHI clinics, thus we were unable to account for it in our analysis. However, the association of inflammatory markers with incident fractures was independent of BMD in our previous analyses.13 Also, hip fractures are a rare outcome in our study population, affecting approximately 1.01% of women, with an annual risk of about 0.14%. This may reflect their relatively young age (age 50–79 years) at enrollment. Among women in a top cytokine-soluble receptor quartile, there was an estimated 50% increase in risk, compared to all other women. Therefore, the absolute risk may have only increased from approximately 0.14% per year to around 0.22% for women in a top inflammatory marker quartile. Third, the adverse effect of inflammation on bone resorption may be exacerbated in states of estrogen deficiency, for instance what is observed after menopause. Therefore, our results are primarily generalizable to postmenopausal white women and cannot be extrapolated to premenopausal women or men. Also, the design of our study was a prospective nested case-control study. We identified all reported incident hip fracture cases that occurred over the follow-up period. We then chose a control matched on age, race, and blood draw. Because hip fractures are more common at advanced ages and among whites, the characteristics of the women differ slightly from the entire WHI cohort because the characteristics were driven by their positive hip fracture history. Furthermore, we measured cytokine-soluble receptor concentrations in the serum; however, these levels may differ in the bone microenvironment and over time. Serum assays may not reflect local cytokine-soluble receptor levels. In addition, several covariates (ie, physical activity and dietary and supplementary intake of calcium) were measured using self-report; therefore, misclassification as a consequence of recall bias is possible. Moreover, as in most epidemiologic cohort studies, we initially relied on self-report of all hip fractures. Medical records were then obtained to confirm the hip fracture by central adjudication review at the WHI coordinating center. Hip fractures are serious events; it is doubtful a participant would forget to report it. Nonetheless, perhaps in the care of proxies, hip fractures may be unreported. This type of misclassification would likely be nondifferential and bias our findings to the null. Finally, residual confounding due to unmeasured factors is a component of all observational studies. For instance, accounting for frailty or health status in an analysis may be limited by self-report.
In summary, elevated levels of inflammatory markers for all three cytokine-soluble receptors were associated with an increased risk of hip fractures in older women. The association was strongest when we combined all inflammatory makers into a composite variable, suggesting that inflammatory burden may be an important biologic factor. Future clinical trials should test whether interventions to decrease inflammatory marker levels reduces hip fractures. Moreover, inhibition of RANKL with RANKL inhibitors could also potentially block the adverse effects of inflammation on bone resorption.
AOY has received research support and/or honoraria from Eli Lilly & Co., Amgen, and GlaxoSmithKline; she has also received consulting fees from Merck & Co. JAC has received research support and consulting fees from Novartis Pharmaceuticals. AZL serves as a consultant to Procter & Gamble and receives research grant support from Pfizer and the Alliance for Better Bone Health. RDJ has received research support from and is on the speaker's bureau for Procter & Gamble Pharmaceuticals, has received research and conference support from Novartis, and has received an honorarium as a Continuing Medical Education speaker for Aventis/Alliance for Better Bone Health. DCB has received research support from Novartis Pharmaceuticals, Amgen, Proctor & Gamble Pharmaceutical Co., and Merck & Co., Inc. RBW serves on clinical trial safety monitoring boards for Novartis Pharmaceuticals and Merck & Co., Inc. MAA, KEB, RB, MED, NCG, and JW-W report no competing interests.
The sponsor (National Heart, Lung and Blood Institute [NHLBI]) has played a role in the design and analyses of the WHI. We would also like to acknowledge the WHI Investigators. For a list of the WHI Investigators, see the Appendix, available at www.annals.org. WHI study protocols are available at http://www.whi.org/about/investigators.php. The WHI program is funded by the NHLBI, National Institutes of Health, U.S. Department of Health and Human Services. Additional support for these analyses was provided by US Public Health Service Research grants: AR053105 and AR048919.
Authors' roles: Study design: JAC, KEB, MED. Study conduct: JAC, MED. Data collection: JAC, MED. Data analysis: KEB, RB, AOY. Data interpretation: JAC, KEB, RB, MED, AOY, JWW, NCG, AZL, RDJ, RBW, DCB, MAA. Drafting manuscript: KEB, JAC. Revising manuscript content: JAC, KEB, RB, MED, AOY, JWW, NCG, AZL, RDJ, RBW, DCB, MAA. Approving final version of manuscript: JAC, KEB, RB, MED, AOY, JWW, NCG, AZL, RDJ, RBW, DCB, MAA. KEB takes responsibility for the integrity of the data analysis.