| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Studies |
Departments of Social Medicine and Clinical Epidemiology (C.S.-N.) and Medicine I, Endocrinology, and Metabolism (C.S.-N., H.B., G.L.-B., H.W., M.J.S., R.Z.), University of Heidelberg Medical Center, D-69115 Heidelberg; and Department of Medicine, Kliniken Bergmannsheil, University of Bochum (J.P.), D-44789 Bochum, Germany
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
10 yr) indicated
that the effect of IL-6 on bone loss weakened with increasing distance
from menopause and was no longer significant in women more than 10 yr
after menopause. Among women up to 10 yr past menopause (n = 39),
serum IL-6 was the single most important predictor of femoral bone
loss, accounting for up to 34% of the total variability of change in
BMD. The unadjusted linear model predicted an annual 1.34% (95%
confidence interval, 0.67–2.01) decrease in total hip BMD per log unit
increase in serum IL-6. A similar, although nonsignificant, effect of
serum IL-6 on vertebral bone loss was restricted to women within the
first 6 yr after menopause (n = 18). These epidemiological data
show that serum IL-6 is a predictor of postmenopausal bone loss, and
that the effect appears to be most relevant through the first
postmenopausal decade. Whether these findings reflect pathogenetic
differences between early and postmenopausal bone loss, and whether
serum IL-6 also predicts fracture risk need further elucidation. |
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
During the past decade, considerable evidence has accumulated that one of the pathways by which estrogen may exert a protective effect on the skeleton may be by governing the effect of cytokines on bone remodeling. Estrogen deficiency dramatically alters the dependency of bone cells on several cytokines, including interleukin 6 (IL-6), IL-1, and tumor necrosis factor (TNF) (8, 9, 10). For example, ovariectomy-induced stimulation of osteoclastogenesis in mice can be prevented by neutralizing antibodies against IL-6 (11) or abolishment of IL-6 function by gene knockout (12).
Although these data provide strong evidence for the involvement of cytokines in bone loss due to sex hormone deficiency, most of the results have been derived from in vitro and animal studies. It is therefore still unclear whether this pathophysiological model of bone loss can be fully applied to humans, and whether serum IL-6 concentrations would be useful in predicting future bone loss. To our knowledge, the relationship between serum IL-6 and bone loss has not previously been studied in postmenopausal women. Two studies have examined the cross-sectional relationship of circulating IL-6 levels to bone mineral density (BMD) (13) or markers of bone turnover (14) in women, and no significant association was observed. It is possible that an effect of IL-6 on bone is missed in cross-sectional analysis. Besides, the effect may be modified by menopausal age. Cell culture studies by Pacifici and colleagues (15, 16) and by our group (17) suggest that increased cytokine production in osteoblasts and their precursors may be restricted to the early postmenopausal phase. Against this background, the aim of the present study was to determine the predictive effect of serum IL-6 concentrations on femoral and vertebral bone loss in a population-based, longitudinal study of postmenopausal women and to examine whether the effect persists across the entire range of menopausal age.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The University of Heidelberg Medical Center is one of eight German centers participating in the European Prospective Osteoporosis Study (EPOS) (18). The main objectives of this multicenter study, details of the sampling process, and results of the baseline survey, known as the European Vertebral Osteoporosis Study (EVOS), have been reported previously (19, 20).
In the framework of EPOS, an age- and sex-stratified random sample of birth cohorts between 1910–1940 was drawn from the population registry of the town of Eppelheim in the vicinity of Heidelberg, Germany. Between January 1992 and March 1993, 297 men and 283 women or 58% of eligible and contactable persons who were then 51–82 yr of age participated in the EPOS baseline survey (20, 21, 22, 23). The present study is based on female cohort members who also completed a follow-up examination in 1995–1996 after an average of 3.3 yr (range, 2.6–4.3 yr); follow-up rates from baseline among women were 62% (n = 176). The study was conducted with approval of the ethics committee on clinical investigations at the University of Heidelberg Medical Center and after the participants gave written consent.
IL-6 measures were complete in 166 women. We excluded 7 women considered pre- or perimenopausal because they reported regular menses during the preceding 12 months. Another 22 women were excluded for preexisting conditions or medication known to affect bone loss. Of these, 2 women had chronic bone disease (primary hyperparathyroidism, multiple myeloma), 1 woman had renal failure and was treated with genuine vitamin D, 5 women were receiving long-term oral corticosteroid therapy, and 14 women received treatment with fluoride or bisphosphonates between baseline and follow-up. This left 137 postmenopausal women for the present investigation.
Women receiving hormone replacement therapy (HRT) at baseline (n = 21) and women who had started HRT during follow-up (n = 9) were included for cross-sectional analysis of the association between HRT and serum IL-6. They were excluded, however, from longitudinal analyses of the predictive effect of cytokines on bone loss. As these analyses also required knowledge of the exact menopausal age, the main study population was reduced to 89 postmenopausal women, who had not undergone hysterectomy before menopause and had never used HRT (n = 82) or stopped HRT at least 1 yr before the baseline survey (n = 7).
Interview and postal mailers
A standardized interview based on the EVOS questionnaire (24) was conducted at baseline by trained personnel to assess medical history including risk factors of osteoporosis and, in women, reproductive history. In women, menopausal state was determined from the answer to the question: "are you still having menstrual periods?" and "if not, when was your last menstrual period?" Women who had stopped menstruating for at least 1 yr were considered postmenopausal. Among women without hysterectomy before menopause, the age at menopause was calculated from the reported year of the last menses as the age at the cessation of menses plus 1 yr (25).
Information on HRT and osteotropic drug therapy between baseline and follow-up was obtained by postal mailers. A standardized questionnaire was sent to surviving members of the German EVOS population twice a year between September 1, 1993, and March 1, 1996. Postal follow-up was complete in 84% of the Heidelberg cohort and in 100% of those who came back for a second clinic visit. Participants were asked if they had been taking any prescription drug for osteoporosis during the preceding 6 months and, if so, to write down the trade name of the drug. In addition, women were asked if they had newly started HRT, and a space was provided to fill in the trade name of the preparation as well as the month and the year it was first taken.
At the time of the second bone density measurement, participants were asked to answer a short standardized questionnaire regarding major changes in health status since baseline. They were specifically asked if they had been immobilized for more than 2 weeks, been newly diagnosed by a physician with arthritis or chronic gastrointestinal disease, or received long-term corticosteroid treatment.
Evaluation of spinal x-rays
Standardized lateral x-rays of the thoracic and lumbar spine were taken at baseline and at follow-up according to the EVOS protocol (19). Spinal x-rays were evaluated by a combination of vertebral morphometry and radiological expert reading for differential diagnosis of vertebral deformity as previously described in detail (26). Severe osteophytosis was defined in the presence of at least one grade 3 or grade 4 osteophyte according to the classification by Nathan (27).
Anthropometric measurements
Standardized measurements of height in centimeters and weight in kilograms were taken at baseline and at follow-up in light clothing with shoes removed. A Seca stadiometer calibrated to 1 cm and Seca scales calibrated to 0.1 kg were used for this purpose. The body mass index (BMI) in kilograms per m2 was calculated from concurrent measurements of height and weight, and the change in BMI was calculated as the difference between BMI at follow-up and baseline.
BMD measurements
BMD was measured by dual x-ray absorptiometry using the same equipment (QDR 1000, Hologic, Inc., Waltham, MA) and measurement protocol at baseline and at follow-up. Posterior-anterior scans of the left proximal femur and the lumbar spine were performed according to the guidelines of the European Concerted Medical Action for the Quantitation of Osteoporosis (23, 28, 29). The right proximal femur was scanned in only a few cases with a history of hip replacement on the left side. The average of vertebral bone density measurements at L2–L4 and total femoral BMD were chosen for the present analysis. All BMD measurements were calibrated to the semianthropomorphic European Spine Phantom (ESP) (30) as previously described (23, 29). Absolute bone loss in milligrams per cm2 was calculated as the difference between the follow-up and baseline values; relative bone loss was expressed as the percent change relative to the baseline value.
Quality control procedures included daily measurements of the phantom provided by the manufacturer to assure machine stability. Repeat measurements of the ESP were used to calculate machine precision by dividing the SD at a particular specified density by that density. Based on five independent measurements of the ESP repeated three times at yearly intervals, the mean machine precision was 0.41% (range, 0.24–0.70) at 0.5 g/cm2, 0.63% (range, 0.29–0.97) at 1.0 g/cm2, and 2.03% (range, 1.01–3.04) at 1.5 g/cm2. In vivo precision was calculated from two measurements of femoral and vertebral BMD in nine young healthy volunteers with measurements 4–12 weeks apart. The mean short-term precision was 1.1% (range, 0.2–2.5) for total hip BMD and 2.8% (range, 0.6-6.0) for BMD at the lumbar spine.
Baseline and follow-up BMD scans were submitted to external review and centralized evaluation at the Institut für Funktionsanalyse im Gesundheitswesen GmbH (IFH, Hamburg, Germany) according to a standardized protocol based on the manufacturer’s recommendations (Institut für Funktionsanalyse im Gesundheitswesen GmbH Quality Management Center, Hologic, Scan Acquisition, and Analysis Guidelines). Of 176 women who participated in both surveys, a total of 168 had complete BMD measurements at both skeletal sites. In four cases missing values were due to the subject’s refusal or bilateral hip replacement. The remainder of four BMD scans (two at the lumbar spine and two at the proximal femur) had to be excluded from the analysis because of false positioning or artifacts.
Biochemical assays
Nonfasting morning venous blood samples were drawn at baseline, separated within 3 h after phlebotomy, and aliquoted. A serum sample was immediately processed in a routine laboratory for a chemistry panel including liver enzymes, creatinine, and total alkaline phosphatase. Serum alkaline phosphatase was measured by an automated colorimetric assay as previously described in detail (31). Laboratory upper normal ranges were 170 U/L for serum alkaline phosphatase, 114.9 µmol/L (1.3 mg/dL) for serum creatinine, 15 U/L for aspartate aminotransferase (AST, SGOT), and 18 U/L for alanine aminotransferase (ALT, SGPT). Additional serum and heparinized plasma aliquots were kept frozen at -80 C until assayed for more specific biochemical parameters. Spot urine specimens were protected from light exposure, and aliquots were stored at -30 C.
IL-6 was measured in previously unthawed serum samples using a
highly sensitive amplified commercial ELISA with an alkaline
phosphatase signal amplification system (Quantikine HS, human IL-6
immunoassay, R & D Systems, Inc., Minneapolis, MN). The
assay has a sensitivity of 0.012 IU/mL (0.094 pg/mL). The intraassay
coefficient of variation was 9%, and the interassay coefficient of
variation was 16%. Results were expressed in picograms per mL and
converted to international units per mL using the National Institute of
Biological Standards and Control/WHO IL-6 International
Reference Standard 89/548 (conversion factor = 0.131). In the
subgroup of women less than 10 yr after menopause, human IL-1ß (high
sensitivity), soluble IL-1 receptor type I, soluble IL-1 receptor type
II, IL-1 receptor antagonist, TNF
(high sensitivity), soluble TNF
receptor type I, and soluble TNF receptor type II were later measured
in remaining serum or plasma aliquots by ELISA (R & D Systems, Inc.).
Circulating levels of sex hormones, adrenal androgens, and sex hormone-binding globulin (SHBG) were assayed in previously unthawed plasma aliquots as described previously (22). The detection limit for plasma 17ß-estradiol was 20 pmol/L (5.4 pg/mL).
Biochemical markers of bone metabolism and calciotropic hormones were measured in an endocrinological research laboratory from previously unthawed serum aliquots. The methods of analysis have been previously described in detail (21, 31, 32). Human intact PTH was measured with a two-site luminometric immunoassay (Magic Lite, Bayer Corp., Fernwald, Germany) at an upper normal range of 65 ng/L. The intra- and interassay coefficients of variation were less than 7% and less than 9% for serum bone-specific alkaline phosphatase (S-BAP), less than 10% and less than 15% for serum osteocalcin, less than 5% and less than 12% for 25-hydroxyvitamin D3, less than 6% and less than 11% for human intact PTH, and 10% and 15% for urinary pyridinoline and deoxypyridinoline.
Serum concentrations of TSH were determined with a high-sensitivity TSH-coated tube assay (Johnson & Johnson, Rochester, NY) with a lower normal cut-off of 0.3 mIU/L as previously described (33).
Statistical analysis
SAS software (version 6.12, SAS Institute, Inc., Cary, NC) was used for data analysis. Mean differences in serum IL-6 and other continuous variables between postmenopausal women with and without HRT at baseline or between women within or beyond the first menopausal decade were tested by the general linear model procedure for ANOVA and covariance in unbalanced designs. The strength of cross-sectional associations was assessed by simple and partial Pearson correlation analyses. The distribution of circulating cytokines, SHBG, endogenous sex hormones, and most biochemical markers of bone turnover (serum osteocalcin and urinary cross-links) showed considerable deviation from the normal distribution; hence, these data were transformed to a natural logarithmic scale. For these variables, reported group means from analyses of variance or covariance represent the antilog of mean logarithmic data, which corresponds to the geometric mean in the more intuitive original measurement scale. Estimates from linear regression models can be related to the original measurement scale by taking the antilogarithm of the log units change.
Linear regression techniques were applied to assess the
determinants of serum IL-6 concentrations at baseline and the
predictive effect of serum IL-6 on bone loss. Baseline variables
examined for their cross-sectional relationship to serum IL-6 included
anthropometric variables (age, menopausal age, and BMI), serum
creatinine, behavioral factors (physical activity, cigarette smoking,
and alcohol consumption), comorbidity (elevated liver enzymes,
suppressed TSH, and history of diabetes mellitus), and plasma SHBG and
endogenous sex hormone levels. The estrone/androstenedione ratio was
computed as an index of aromatase activity. These variables as well as
baseline measures of calcium intake, serum levels of calciotropic
hormones, and change in health status between baseline and follow-up
(change in BMI, immobilization >2 weeks, and history of arthritis or
chronic gastrointestinal disease) were examined as confounders of the
effect of serum IL-6 on bone loss. The effect of individual factors was
first examined in univariate linear models. Main factors added to the
multivariable prediction models were serum IL-6, menopausal age, BMI,
serum creatinine, serum intact PTH, and, in separate models, plasma
SHBG and sex hormone levels. Lifestyle variables and comorbidity were
added to multivariable models if they were univariately related to the
outcome variable (serum IL-6 or bone loss) at the P <
0.150 significance level. The homogeneity of the effect of serum IL-6
on bone loss across menopausal age was tested by including a product
term with menopausal age as a continuous or dichotomous (>10
vs. 10 yr) variable. A cut-off of 10 yr was chosen as the
accelerated phase of postmenopausal bone loss is considered to extend
through the first decade past menopause (4).
P < 0.05 was considered statistically significant,
based on two-sided tests.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Women using HRT at baseline (n = 21) had significantly lower
mean serum levels of IL-6 (P < 0.001) than 116
nonusers (Table 1
). This group difference
remained significant after age adjustment (0.15 vs. 0.21
IU/mL; P = 0.013), but was considerably reduced after
adjustment for age, BMI, and serum creatinine (0.16 vs. 0.20
IU/mL; P = 0.083). No difference in mean serum IL-6 was
observed between 103 never users and 13 previous users of HRT (data not
shown). Analyses regarding the predictive effect of serum IL-6 on bone
loss were restricted to 89 women of known menopausal age who did not
receive HRT at baseline or during follow-up. These 89 women did not
significantly differ from all 116 nonusers of HRT with respect to
baseline characteristics (Table 1).
|
. A total of 11 women had vertebral
fractures or severe osteophytosis at the lumbar spine. As both
conditions lead to inaccurate measurements of vertebral BMD, these
women were excluded from analyses involving vertebral bone loss as an
end point.
|
Older postmenopausal women had significantly lower mean serum
25-hydroxyvitamin D concentrations than early postmenopausal women
(Table 3), and the difference increased
after adjustment for age (38.6 vs. 69.0 nmol/L;
P = 0.002). No group differences were observed with
respect to crude mean levels of plasma SHBG and endogenous sex
hormones, biochemical markers of bone turnover, and serum intact PTH.
This was also true after adjusting for age, BMI, and serum creatinine,
except for higher mean adjusted bioavailable testosterone levels (0.20
vs. 0.14 nmol/L; P = 0.040) in early
vs. older postmenopausal women. About 20% of women in each
group had plasma estradiol levels below 36.71 pmol/L (10 pg/mL).
|
Average bone loss at the hip was more pronounced in older than in
younger postmenopausal women, and the reverse was observed for
vertebral bone loss (Table 3
). However, none of these differences was
statistically significant. A quadratic model best described the
relationship between menopausal age and bone loss, suggesting an
initial decline in rates of bone loss with increasing distance from
menopause, and a new rise in bone loss rates in women of older
menopausal ages (Fig. 1, A and B). For
relative annual rates of femoral bone loss, the fitted curve took the
shape of an inverse J (Fig. 1A). Starting from an average annual bone
loss of approximately 0.5%/yr, rates of bone loss slightly declined
during the first 10 yr after menopause and continuously rose thereafter
up to an average rate of about 2.5%/yr. An inversely U-shaped curve
characterized the relationship of vertebral bone loss rates with
menopausal age (Fig. 1B), reflecting that the highest rates of bone
loss at an average of 1%/yr were observed in women within the first
years after menopause and again in late menopause. No significant
linear or curvilinear association was observed between biological age
and femoral or vertebral bone loss.
|
Crude mean serum IL-6 concentrations did not significantly differ
between early and late postmenopausal women (Table 3). As illustrated
in Fig. 2, an effect of menopausal age
group on serum IL-6 was at least in part masked by the fact that
biological age was linearly and positively related to serum IL-6 in
older, but not in younger, postmenopausal women. Among women beyond 10
yr after menopause, age explained 7% of the variability in serum IL-6,
although the effect did not reach statistical significance
(P = 0.061) due to a limited sample size.
|
In multivariable analysis a higher biological age, early
postmenopausal status (within 10 yr after menopause), higher BMI,
higher serum creatinine, and lower serum intact PTH were all
independently related to higher serum IL-6. In addition, the
estrone/androstenedione ratio significantly contributed to the model
(Table 4). Associations with other sex
hormones or SHBG, as evaluated in separate models, were not
significant. We examined whether the relationship between sex hormones
or SHBG with serum IL-6 was modified by menopausal age group. As BMI
was positively related to serum IL-6 as well as to plasma estradiol
levels (r = 0.38; P < 0.001), we also considered
a modifying effect of BMI. Adding the respective product terms to
multivariable regression models revealed no interaction with menopausal
age group. However, the effect of circulating estradiol was modified by
BMI, suggesting that a significant and inverse relationship between
plasma estradiol and serum IL-6 was diminished with increasing BMI
(ß = -1.6930; P = 0.013 per log unit increase
in estradiol and ß = 0.0702; P = 0.006 for the
interaction with BMI). Similar interactions with BMI were observed for
plasma estradiol when used as a group variable (>36.71 vs.
36.71 pmol/L) as well as for circulating plasma estrone and
dehydroepiandrosterone sulfate (data not shown). Plasma bioavailable
testosterone showed no linear relationship to serum IL-6 in
multivariable models with or without interaction terms included.
|
Higher serum IL-6 concentrations were strongly related to higher
femoral bone loss in univariate linear regression models (Table 5). On the other hand, endogenous sex
hormones, including plasma estradiol as a continuous or dichotomous
variable (cut-off, 36.71 pmol/L), were not predictive of bone loss in
univariate or menopausal age- and BMI-adjusted models. However, a
significant protective effect of plasma estradiol, estrone, and
bioavailable testosterone on femoral bone loss was apparent in models
including serum intact PTH and an interaction between PTH and sex
hormones. In models also adjusting for menopausal age and BMI, the
protective effect was most pronounced for bioavailable testosterone
(ß = 21.8471; P = 0.004 per log unit increase in
bioavailable testosterone and ß = -0.5097; P =
0.002 for the interaction with PTH) and estradiol as a dichotomous
variable (ß = 23.3418; P = 0.013 for total
plasma estradiol levels 
36.71 vs. < 36.71 pmol/L and
ß = -0.4787; P = 0.030 for the interaction with
PTH).
|
).
Adding interactions between estrogens and BMI did not change the
results. Bioavailable testosterone in interaction with PTH was the only
sex hormone that remained predictive of femoral bone loss independently
of IL-6.
The IL-6 effect also persisted in multivariable regression
analysis adjusting for menopausal age, BMI, serum intact PTH, and other
factors related to bone loss at the hip in univariate models at the
P < 0.150 significance level. These included an
elevation of liver enzymes at baseline as well as a decrease in BMI and
bone-related morbidity (history of chronic gastrointestinal disease,
arthritis, or immobilization) between baseline and follow-up. Testing
for the homogeneity of the IL-6 effect across menopausal age, a product
term between serum IL-6 and menopausal age was added and significantly
contributed to the model (Table 5). The direction of the interaction
suggests that a significant effect of IL-6 on bone loss fades with
increasing distance from menopause. Apart from serum IL-6, a higher
menopausal age, initial elevation of liver enzymes, and morbidity
during follow-up remained independently predictive of increased bone
loss at the hip (data not shown). A significant interaction between
serum IL-6 and menopausal age was also evident from a separate
multivariable model fitting the regression of femoral bone loss on
menopausal age as a dichotomous variable (>10 vs. 10 yr),
biological age, and additional covariates as described above (ß
= -13.2762; P = 0.002 per log unit increase in IL-6
and ß = 13.6372; P = 0.024 for the product
term). From this model, estimates of the independent effect of serum
IL-6 on femoral bone loss can be calculated as ß = -13.2762
among early and ß = 0.3610 among late postmenopausal women.
Figure 3, A and B, graphically
depicts the univariate relationship between serum IL-6 and absolute
femoral bone loss in early and late postmenopausal women, as derived
from analyses stratified for menopausal age group. In women up to 10 yr
after menopause (Fig. 3A), IL-6 explained 34% of the variability in
absolute bone loss [ß = -12.7650; SE(ß) =
2.9389; P < 0.001]. In contrast, serum IL-6 was not
predictive of femoral bone loss [ß = -5.4754;
SE(ß) = 4.5507; P =
0.235] among women of older menopausal age (Fig. 3B). A test for
difference in slopes of regression lines from univariate models was not
statistically significant (P = 0.197). As shown above,
the difference increased to statistical significance after accounting
for the effect of covariates (ß = -13.2762 vs.
ß = 0.3610; P = 0.024). Results were similar for
relative bone loss at the hip. For example, the unadjusted linear model
predicted an annual 1.34% decrease in total femoral BMD per log unit
increase in serum IL-6 among early postmenopausal women [ß =
-1.3397; SE(ß) = 0.3398;
r2 = 0.30; P < 0.001].
|
Among women more than 10 yr after menopause, factors univariately related to higher bone loss included a higher menopausal age, lower initial femoral BMD, elevation of liver enzymes at baseline, decrease in BMI since baseline, and bone-related morbidity during follow-up. Consistent with our observations in the entire study population, plasma levels of endogenous sex hormones, in particular estrone and bioavailable testosterone, were predictive of bone loss in interaction with serum intact PTH. The small sample size precluded more complex multivariable analyses to assess the independent effect of individual predictors of bone loss in older postmenopausal women.
A linear effect of higher serum IL-6 levels to higher bone loss at the lumbar spine was also restricted to the early postmenopausal phase. Up to 6 yr after menopause, we observed a similar, albeit nonsignificant, effect of serum IL-6 on vertebral bone loss (ß = -13.4085; r2 = 0.16; P = 0.101), but the number of observations in this subgroup was reduced to only 18 women.
Predictive effect of biochemical markers of bone turnover on bone loss
For comparison, we also assessed the predictive effect of markers of bone turnover on bone loss among postmenopausal women. Higher serum S-BAP was the only biochemical marker to predict increased bone loss. In univariate linear regression models, this marker explained 5% and 8% of the total variability of bone loss at the hip or at the lumbar spine respectively. Similar results were obtained from multivariable models controlling for menopausal age, BMI, serum creatinine, change in BMI, comorbidity factors, and seasonal variation. Interaction terms between markers and menopausal age or menopausal age group did not significantly contribute to these models. The significant effect of serum IL-6 on femoral bone loss as well as the interaction with menopausal age group persisted in multivariable models, also adjusting for S-BAP (data not shown).
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
A similar pattern for the effect of serum IL-6 on vertebral bone loss was observed, although the association was considerably weaker and less consistent. This observation seems puzzling, given the fact that the spine is the predominant skeletal site of accelerated bone loss. However, many other associations were also much weaker for the spine than for the femur, pointing, rather, at a measurement problem. This is consistent with findings in other population-based studies, which have suggested that measurements of vertebral BMD changes may be misleading with respect to the overall changes in skeletal BMD (35, 36, 37, 38). Alternatively, an effect of serum IL-6 on cancellous bone may be best seen within very close distance to menopause; it would then be masked in our study due to the small number of women within the first few postmenopausal years.
The observation that the predictive effect of IL-6 on bone loss was restricted to the first 10 yr after menopause confirms results from cell culture studies previously reported by our group and others. In vitro studies of cytokine secretion in circulating human mononuclear cells (15, 16) and human bone marrow cells (17) have shown increased secretion rates of IL-6 and related cytokines in cells from early, but not late, postmenopausal women compared with baseline levels in cells from premenopausal women. The fact that the IL-6 effect on bone loss is modified by menopausal age may explain why an association between serum IL-6 and BMD was not detected in a previous cross-sectional study by other researchers (13). We found a cross-sectional association of initial femoral BMD to serum IL-6 only after stratification for menopausal age group, and the relationship was evident in women of younger, but not older, menopausal age (data not shown).
The precise mechanism for the restriction of the association between IL-6 and bone loss to early menopause remains to be elucidated. There are two possible explanations for the observed differential effect of circulating IL-6 levels on bone loss among early and older postmenopausal women. First, there may be a selective increase in the sensitivity of bone cells toward IL-6 in early menopause. Indeed, recent in vitro studies have demonstrated that ovariectomy up-regulates the expression of the two subunits of the IL-6 receptor in stromal cells in ex vivo murine bone marrow cultures (39). Secondly, elevated serum IL-6 concentrations in older postmenopausal women may be less likely to reflect increased IL-6 secretion in bone due to estrogen deficiency. Serum IL-6 concentrations are known to rise with age, as has been shown in the present analysis and in a previous cross-sectional study of 80 healthy women by others (14). We were able to demonstrate that a higher biological age and early postmenopausal status (within 10 yr after menopause) were significantly and independently related to higher serum IL-6 levels. Although the metabolic pathways underlying the age-related increase in serum IL-6 are still subject to research, recent data point to a possible link with atherosclerosis (40, 41).
This leaves us with the question of what factors may determine individual differences in serum IL-6 concentrations during the early postmenopausal phase. Although current HRT users were found to have lower serum IL-6 than nonusers in the present study and in one previous population-based study (42), a relationship between serum IL-6 and endogenous estrogens was not obviously present among younger or older postmenopausal women. A link between endogenous estrogen activity and serum IL-6 was indicated by the observed inverse association of serum IL-6 to the estrone/androstenedione ratio and to serum intact PTH. However, as previously reported by McKane and colleagues (14), IL-6 and total plasma estradiol in our population were rather positively related in univariate linear models. We found that the association between serum IL-6 and residual estrogen concentrations was modified by BMI, as an inverse relationship between IL-6 and endogenous estrogens was increasingly concealed at higher levels of BMI. Results from a recent population-based study (42) support our observation that obesity, as estimated by BMI, is strongly correlated with higher serum IL-6 concentrations, and that BMI and estrogen status may interact in the determination of circulating IL-6. It remains to be shown how environmental, hormonal, and genetic factors act together in the determination of IL-6 expression and serum IL-6 levels after menopause. The first evidence of an association between bone mass and polymorphisms in the IL-6 gene or related genes is emerging (43, 44).
Differences in residual estradiol, SHBG, or other sex hormone concentrations did not explain the IL-6 effect on bone loss, nor were they related to bone loss in univariate or menopausal age- and BMI-adjusted linear models. Based on the results from a large population-based cohort study of postmenopausal women 65 yr and over, estradiol levels below 18.36 pmol/L (5 pg/mL) may be critical in the prediction of bone loss (2, 3) and osteoporotic fractures (2, 5). We cannot exclude that the immunological assay system we used may have not been sensitive enough to demonstrate a relationship between plasma estradiol and bone loss. On the other hand, plasma estradiol, as both a continuous and a dichotomous variable (cut-off, 36.71 pmol/L), as well as estrone and bioavailable testosterone did show a protective effect on femoral bone loss in interaction with serum intact PTH in our dataset, whereas SHBG demonstrated an inverse association. This supports the concept that there is a later, cytokine-unrelated effect of endogenous estrogens on bone metabolism, perhaps mediated by an effect on extraskeletal calcium homeostasis (4).
As IL-6 is known to exert its osteoclastogenic effects as part
of a complex cytokine network (10), we subsequently
determined serum concentrations of related cytokines (IL-1ß, soluble
IL-1 receptor type I and type II, IL-1 receptor antagonist, TNF, and
soluble TNF receptor type I and type II) from remaining serum or plasma
aliquots in the subset of early postmenopausal women (data not shown).
None of these cytokines was significantly related to bone loss. This
could result from a greater analytical variability or a more pronounced
difference between systemic cytokine levels and those in the local bone
environment.
In the present study biochemical markers of bone turnover were not predictive of bone loss in postmenopausal women, with the exception of serum S-BAP. This is in apparent contrast to previous findings demonstrating a predictive effect of markers of bone resorption, in particular pyridinium cross-links, on fracture risk in elderly women (45, 46). A potential bias toward a younger and healthier population sample, as described below, may have been responsible for underestimation of the ability of these markers to predict bone loss in women of the present study. On the other hand, there is increasing evidence that biochemical markers of bone turnover may be poor predictors of bone loss during the early postmenopausal years (47, 48). As low bone mass is one major risk factor for fragility fractures in the elderly (49), and accelerated bone loss in early postmenopause is believed to be the most important contributing factor to low bone mass in women (1), the ability of IL-6 to specifically predict bone loss during the early postmenopausal years may be of great clinical significance.
The strengths of the present study lie in the population-based setting and the longitudinal design. Furthermore, we were able to analyze the relationship between IL-6 and bone loss in the light of other endogenous and environmental factors known or suspected to be related to age-related bone loss (3, 35, 50, 51, 52).
Our study also has several major limitations. First, the number of women was small, limiting multivariable analyses and the precision of estimates derived from the prediction models. Secondly, as response rates to the initial survey at baseline were only 58%, we cannot rule out that our results were affected by selection bias. Comparisons of age, health and functional status between participants and nonparticipants in the Heidelberg EVOS cohort as well as for all German study centers combined indicated that a selective participation of younger and healthier individuals already occurred at baseline (53). Loss of individuals during follow-up, which is of concern to all longitudinal studies, almost certainly added to this effect. As older, less frail, and chronically ill subjects are most likely to be underrepresented in the study population, it will not be truly representative of all postmenopausal women, and the findings have to be interpreted with care. Bone loss as well as the predictive ability of IL-6, sex hormones, and biochemical markers of bone turnover may have been underestimated, particularly among older postmenopausal women.
In summary, this is the first study to show that serum IL-6 is a predictor of postmenopausal bone loss, and that the effect appears to be most relevant through the first postmenopausal decade. In line with previous results from in vitro and animal studies, these findings support the hypothesis that IL-6 is an important mediator of bone loss during the first accelerated phase of bone loss, whereas other mechanisms may be more relevant to bone loss in later postmenopausal life. However, our correlative data cannot prove a cause-effect relationship. Serum IL-6 could be merely a marker of some related causal pathomechanism. Notably, the IL-6 effect on femoral bone loss was not explained by residual estradiol levels, although there was evidence for an association between lower endogenous estrogen activity and higher IL-6 levels. Further studies are warranted to elucidate the mechanisms underlying our observations and to determine whether serum IL-6 also predicts the risk of osteoporotic fracture.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
* This work was supported by Bundesministerium für Forschung und Technologie (Grants 01KM 9101/6 and 01KM 9304/0) and Deutsche Forschungsgemeinschaft (Grant SCHE 390/3-1).
Received October 16, 2000.
Revised January 12, 2001.
Accepted January 16, 2001.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  M K. Shea, G. E Dallal, B. Dawson-Hughes, J. M Ordovas, C. J O'Donnell, C. M Gundberg, J. W Peterson, and S. L Booth Vitamin K, circulating cytokines, and bone mineral density in older men and women Am. J. Clinical Nutrition, August 1, 2008; 88(2): 356 - 363. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Ding, V. Parameswaran, R. Udayan, J. Burgess, and G. Jones Circulating Levels of Inflammatory Markers Predict Change in Bone Mineral Density and Resorption in Older Adults: A Longitudinal Study J. Clin. Endocrinol. Metab., May 1, 2008; 93(5): 1952 - 1958. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. O'Brien A Possible Role of Recurrent Major Depression in Risk of Fracture Arch Intern Med, November 26, 2007; 167(21): 2370 - 2370. [Full Text] [PDF]
|
|
|
|
 |
|
 M Bustamante, X Nogues, L Mellibovsky, L Agueda, S Jurado, E Caceres, J Blanch, R Carreras, A Diez-Perez, D Grinberg, et al. Polymorphisms in the interleukin-6 receptor gene are associated with bone mineral density and body mass index in Spanish postmenopausal women Eur. J. Endocrinol., November 1, 2007; 157(5): 677 - 684. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Ferrucci and D. Alley Obesity, Disability, and Mortality: A Puzzling Link Arch Intern Med, April 23, 2007; 167(8): 750 - 751. [Full Text] [PDF]
|
|
|
|
 |
|
 G. P. Anderson COPD, asthma and C-reactive protein. Eur. Respir. J., May 1, 2006; 27(5): 874 - 876. [Full Text] [PDF]
|
|
|
|
 |
|
 H.-P. Yu, T. Shimizu, Y.-C. Hsieh, T. Suzuki, M. A. Choudhry, M. G. Schwacha, and I. H. Chaudry Tissue-specific expression of estrogen receptors and their role in the regulation of neutrophil infiltration in various organs following trauma-hemorrhage J. Leukoc. Biol., May 1, 2006; 79(5): 963 - 970. [Abstract] [Full Text] [PDF]
|
, Early
postmenopausal women (