Schett G, Kiechl S, Redlich K, Oberhollenzer F, Weger S, Egger G, Mayr A, Jocher J, Xu Q, Pietschmann P, Teitelbaum S, Smolen J, Willeit J. Soluble RANKL and Risk of Nontraumatic Fracture. JAMA. 2004;291(9):1108-1113. doi:10.1001/jama.291.9.1108
Author Affiliations: Department of Internal Medicine III, Division of Rheumatology (Drs Schett, Redlich, and Smolen), and Department of Pathophysiology (Dr Pietschmann), University of Vienna, Vienna, Austria; Department of Neurology, University Clinic of Innsbruck, Innsbruck, Austria (Drs Kiechl and Willeit); Departments of Internal Medicine (Drs Oberhollenzer, Weger, and Egger), Laboratory Medicine (Dr Mayr), and Radiology (Dr Jocher), Bruneck Hospital, Bruneck, Italy; Department of Cardiological Sciences, St George's Hospital Medical School, London, England (Dr Xu); and Department of Pathology, Washington University School of Medicine, St Louis, Mo (Dr Teitelbaum).
Context The receptor activator of nuclear factor κB ligand (RANKL) is
essential for osteoclast and, possibly, osteoblast activation and may represent
a key link between bone formation and resorption.
Objective To determine the relationship between serum level of RANKL and the risk
of nontraumatic fracture.
Design, Setting, and Participants As part of a prospective population-based study conducted in Bruneck,
Italy, we recorded all fractures that occurred between 1990 and 2000 in 906
participants and classified them as traumatic (n = 115) or nontraumatic (n
= 31). Serum levels of RANKL and osteoprotegerin and characteristics of bone
metabolism and lifestyle were assessed in 1990 and at follow-up in 1995 and
Main Outcome Measure Incident nontraumatic fracture by levels of RANKL.
Results Levels of RANKL did not differ between sexes and were not related to
age, menopausal status, lifestyle characteristics, or data from bone ultrasound
at the heel. However, RANKL emerged as a significant predictor of nontraumatic
fracture. In pooled logistic regression analysis, the relative risks of nontraumatic
fracture in the lowest and middle vs highest tertile for RANKL were 10.0 (95%
confidence interval [CI], 2.3-43.1) and 3.9 (95% CI, 0.8-19.0) (P<.001 for trend), respectively. Patients in the highest-tertile
group had a low risk of fracture even in the presence of other predisposing
factors, whereas women aged 60 years or older in the lowest tertile had a
5-year rate of nontraumatic fracture greater than 7%.
Conclusions A low level of RANKL is an independent predictor of nontraumatic fracture.
This finding is consistent with the hypothesis of an important role of RANKL
in human bone turnover and if confirmed in future investigations may gain
relevance for assessment of fracture risk.
Bone undergoes a continuous remodeling process, allowing optimal adaptation
of microarchitecture to individual demands. Bone resorption triggered by osteoclasts
is physiologically coupled and usually in balance with bone formation, which
is mediated by osteoblasts.1 Recently, essential
physiological interactions between osteoclasts and osteoblasts have been unraveled
and the receptor activator of nuclear factor κB ligand (RANKL) was proposed
as a key player in this scenario because (1) RANKL stimulates osteoclastogenesis
and induces osteoclast activation2- 5;
(2) recent data suggest that RANKL directly activates osteoblasts and triggers
bone formation at concentrations well below those necessary to induce osteoclastogenesis6; (3) RANKL is involved in coupling since its expression
on osteoblasts stimulates osteoclastogenesis2,3;
and (4) juvenile Paget disease, a rare genetic condition of very high bone
turnover, is associated with serum levels of RANKL more than 10 times those
measured in healthy individuals.7
Functionally, RANKL, a protein with structural homology to tumor necrosis
factor, acts through binding to the transmembrane receptor RANK.2- 5,8,9 Competitive
binding to the naturally occurring decoy receptor osteoprotegerin (OPG) blocks
RANKL/RANK interaction at the ligand/receptor level.10,11 RANKL
is expressed by osteoblasts, bone marrow stromal cells, and activated T cells
and occurs in circulation as a soluble molecule,12 making
it suitable for laboratory assessment.
Despite the recent advances in understanding the significance of RANKL
in experimental animal models, relevance in physiological bone metabolism
of humans and potential consequences for bone quality and fracture risk remain
to be clarified. In the current large, prospective study, we investigated
whether RANKL qualifies as a laboratory marker of fracture risk in the general
The Bruneck Study is a prospective population-based survey of the epidemiology
and pathogenesis of atherosclerosis and disorders of the brain and bone.13 The study protocol was reviewed and approved by the
appropriate ethics committees, and all study participants gave written informed
consent. At baseline in 1990, the study population was recruited as a random
sample, stratified according to sex and age, of all inhabitants of Bruneck,
Italy (125 women and 125 men in each of the fifth to eighth decades of age).
A total of 93.6% participated, with data assessment completed in 919 participants.
Reevaluations were performed after 5 years (1995) and 10 years (2000).13 Blood samples for measurement of RANKL and other
parameters were available from 919 (1990), 826 (1995; 96.5% of those alive),
and 700 (2000; 97.7% of those alive) participants, respectively. For the current
analysis, 1 participant with a pathologic fracture due to bone metastasis
and 12 who had experienced nontraumatic fractures before baseline were excluded.
In the remaining 906 participants, determination of clinically apparent nontraumatic
fractures was nearly complete (>99%) between 1990 and 2000.
Lifetime peripheral and clinically apparent vertebral fractures were
carefully recorded for all study participants using the participant's self-report
and a standardized reevaluation of all radiographs ever taken in study participants.
The situation in Bruneck is unique in that (1) the only radiography facility
in the district is located at the hospital and all radiographs were available
for review; (2) it is convenient to perform radiography in virtually all cases
of injury and moderate to severe or long-lasting back pain; and (3) population
mobility in the survey area was extremely low during the study. None of the
reported fractures occurred outside of the Bruneck area. For all radiologically
confirmed fractures, localization, date of occurrence, and associated circumstances
were recorded. Fractures were classified as nontraumatic if resulting from
a fall from standing height or less or manifesting without any trauma.14 Other fractures, especially those of fingers, toes,
skull, face, cervical vertebrae, and chest/sternum, were considered traumatic.15 Vertebral fractures were radiologically defined by
a decrease of at least 20% and at least 4 mm of anterior, medial, or posterior
vertebral height (compared with the posterior margin of the same vertebra
or, if reduced, above adjacent vertebra) in lateral thoracic and lumbar spine
radiographs (segments T4 to L5).16
All lifestyle variables were assessed in 1990, 1995, and 2000. Body
mass index was calculated as weight in kilograms divided by the square of
height in meters. Smoking status and alcohol consumption were recorded as
described previously.13 A physical activity
score was calculated from the scores for work (3 categories) and sports/leisure
activities (0, ≤2, or >2 h/wk).13 Socioeconomic
status was defined by a 3-category scale (low, medium, or high) based on information
about occupational status and educational level of the person with the highest
income in the household.13 Diabetes was diagnosed
according to World Health Organization criteria. Bone ultrasound data (broadband
ultrasonic attenuation and speed of sound) were assessed in 2000 at the heel
using quantitative ultrasound equipment (SAHARA, Hologic Inc, Bedford, Mass).
Blood samples were drawn in 1990, 1995, and 2000 after an overnight
fast and 12 hours of abstinence from smoking.13 Serum
was immediately frozen and stored at −70°C without any cycle of
thawing-freezing until analysis. Serum levels of osteocalcin, parathyroid
hormone, and β-crosslaps were measured by electrochemiluminescence immunoassay
(ECLIA, Roche Diagnostics, Mannheim, Germany). 25-Hydroxyvitamin D was analyzed
by automated chemiluminescence assay (Nichols Advantage, Nichols Institute
Diagnostics, San Juan Capistrano, Calif). Intra-assay coefficients of variation
for osteocalcin, parathyroid hormone, β-crosslaps, and 25-hydroxyvitamin
D testing were low at 0.6%, 1.4%, 3.1%, and 3.7%, respectively. Serum levels
of soluble uncomplexed RANKL were measured by a commercial sandwich enzyme-linked
immunosorbent assay (Biomedica, Vienna, Austria), as described previously.12 All measurements were performed by a single experienced
technician who was unaware of any characteristics of study participants. Chimeric
OPG-Fc protein (R & D Systems, Minneapolis, Minn) was coated on microtiter
plates and used to bind free RANKL in the samples. In a second step, RANKL
captured by OPG was detected by a specific affinity-purified and biotinylated
rabbit antibody (Leinco Technologies, St Louis, Mo) followed by incubation
with streptavidin peroxidase and visualization with tetramethylbenzidine.
Biosynthetic RANKL (Peprotech, Rocky Hill, NJ) diluted in human serum was
used as a standard. Intra-assay and interassay coefficients of variation were
6% and 8%, respectively. The lower detection limit of the test was 0.1 pmol/L.
Osteoprotegerin was measured using a sandwich enzyme immunoassay (R &
D Systems). Recombinant OPG from Research Diagnostics Inc (Flanders, NJ) served
as a standard. Intra-assay and interassay coefficients of variation were less
Person-years of follow-up for each participant were accrued from the
1990 baseline until diagnosis of nontraumatic fracture, death, or August 1,
2000, whichever came first. Participants who had nontraumatic fractures were
censored with respect to subsequent follow-up. Participants were divided into
3 approximately equally sized groups according to tertiles of RANKL. Relative
risks (RRs) were estimated with rate ratios comparing the incidence of nontraumatic
fractures in each tertile with that in the highest (referent) tertile using
pooled logistic regression.17,18 This
technique treated each observation period (1990-1995 and 1995-2000) as a follow-up
substudy in which updated risk factor measurements were used to predict fracture
risk. Observations in both periods were pooled into a single sample. This
approach has been shown to be asymptotically equivalent to the Cox regression
model with time-dependent covariates given short intervals between reevaluations
and low rates of events.17,18 Multivariate
models were adjusted for age (years), sex (men, premenopausal women, or postmenopausal
women), follow-up period (1990-1995 or 1995-2000), socioeconomic status (low,
medium, or high), smoking (cigarettes per day), alcohol consumption (grams
per day), physical activity score, diabetes (no vs yes), body mass index,
creatinine levels (milligrams per deciliter), hormone therapy (no vs yes),
and, facultatively, other types of medication and parameters of bone metabolism.
We performed tests for linear trend by treating the medians in each category
of RANKL as a continuous variable. Regression-adjusted rates of nontraumatic
fractures according to tertiles of RANKL, age, sex, and menopausal status
were calculated with the marginal method of the regression adjustment technique.19 All reported P values are
2-sided; P<.05 was considered statistically significant.
SPSS software version 11.5 (SPSS Inc, Chicago, Ill) was used for all analyses.
During 8087 person-years of follow-up, 31 cases of nontraumatic and
115 cases of traumatic fractures were documented. Baseline distribution of
RANKL and main descriptive characteristics are depicted in Figure 1. Individual levels of RANKL emerged as comparatively stable
over time, as indicated by a high correlation between 1990 and 1995 measurements
(r = 0.63) and a mean absolute difference of 0.077
pmol/L (95% confidence interval [CI], −0.001 to 0.157 pmol/L) (SD, 1.14
pmol/L). A scatterplot is shown in Figure
2. Age, sex and menopausal state were equally distributed among
tertiles of baseline RANKL level (Table
1). Participants with lower levels of RANKL tended to be of lower
socioeconomic status, less active, and more likely to be current smokers and
have diabetes. However, none of these trends achieved significance considering
the performance of multiple comparisons. Furthermore, serum levels of RANKL
were unrelated to bone ultrasound data assessed at the heel and various parameters
of bone metabolism except for OPG (modest inverse relation; P<.001).
The incidence of nontraumatic fractures varied from 0.7 per 1000 person-years
in the highest tertile to 8.1 per 1000 person-year in the lowest tertile of
RANKL (Table 2). In pooled logistic
regression analyses adjusted for age, sex, menopausal status, and period of
follow-up (n = 1712), the 5-year risk of nontraumatic fracture increased with
decreasing tertiles of RANKL (RR, 10.0; 95% CI, 2.3-43.1 in the lowest tertile
and RR, 3.9; 95% CI, 0.8-19.0 in the middle tertile compared with the highest
tertile; P for trend <.001) (Table 2). In multivariate analyses, after simultaneous control for
a variety of demographic and lifestyle variables, creatinine concentration,
and hormone therapy, results did not change appreciably (RR, 9.7; 95% CI,
2.2-42.1 in the lowest tertile and RR, 4.0; 95% CI, 0.8-19.7 in the middle
tertile compared with the highest tertile; P for
trend <.001). In this model, the RR of nontraumatic fractures was 1.4 (95%
CI, 0.9-2.3; P = .15) for a 10-year increase in age
and 4.2 (95% CI, 0.4-47.3; P = .25) and 9.7 (95%
CI, 2.8-33.3; P<.001), respectively, for premenopausal
and postmenopausal women vs men. Additional adjustment for levels of OPG and
other parameters of bone metabolism (Table
2) and concomitant medication had no further effect. As expected,
corresponding Cox regression models with time-dependent covariates yielded
results very similar to those of the pooled logistic regression models (Table 2).
There was no evidence of a differential association between RANKL and
nontraumatic fracture in men vs women (interaction term, P = .66) or in different age groups (interaction term, P = .42), and the findings were consistent in analyses that separately
focused on hip and vertebral fractures. The findings were essentially the
same after exclusion of participants with diabetes (Table 2). Four persons with diabetes had a nontraumatic fracture.
Three of these had a RANKL level in the lowest tertile and 1 in the middle
Next, we calculated regression-adjusted rates of nontraumatic fractures
in subgroups according to age, sex, and menopausal status (Figure 3). In brief, participants in the highest tertile of RANKL
faced a low risk of nontraumatic fracture irrespective of age and sex. However,
fracture risk steeply increased from the highest to the lowest tertile of
RANKL, especially in postmenopausal women.
In contrast with nontraumatic fractures, traumatic fractures were not
related to RANKL level in multivariate pooled logistic regression analysis
(RR in the lowest and middle vs highest tertiles of RANKL, 1.1 [95% CI, 0.7-1.8]
and 1.0 [95% CI, 0.6-1.7], respectively; P for trend
= .72) (see Table 2 for adjustment).
Results were virtually identical after exclusion of participants with diabetes.
The intervals between blood sampling and occurrence of nontraumatic and traumatic
fractures were a mean of 2.5 and 2.8 years, respectively. After a fracture,
levels of RANKL did not change significantly.
As anticipated, the relationship between RANKL and nontraumatic fracture
was independent of serum OPG. Osteoprotegerin per se was associated with risk
of nontraumatic fracture (RR in the lowest and middle vs highest tertiles
of OPG, 0.2 [95% CI, 0.1-0.6] and 0.4 [95% CI, 0.2-0.9]; P for trend = .001) but this relationship disappeared after adjustment
for age (RR, 0.5 [95% CI, 0.2-1.8] and 0.6 [95% CI, 0.2-1.6]; P for trend = .23) and other variables (RR, 0.5 [95% CI, 0.2-1.8] and
0.6 [95% CI, 0.2-1.5]; P for trend = .20) due to
a high correlation between OPG and age (Spearman r =
Our prospective study demonstrates that a low serum level of RANKL is
a highly significant risk predictor for nontraumatic fractures in the general
population independent of age, sex, menopausal status, level of OPG, and lifestyle
characteristics. Study participants in the highest tertile of RANKL faced
a low risk of nontraumatic fractures irrespective of the presence or absence
of other predisposing factors (incidence of <1 per 1000 person-years).
In contrast, women aged 60 years or older with a serum RANKL level in the
lowest tertile showed a regression-adjusted 5-year rate of fracture greater
than 7% (Figure 3). Although RANKL
was significantly associated with fracture risk in all subgroups, absolute
risk differences and, accordingly, predictive relevance were highest in postmenopausal
women (Figure 3). Importantly, the
serum level of RANKL was comparatively stable over time and emerged as unrelated
to bone ultrasound data, which may be regarded as a surrogate of bone mass.
However, the latter observation should be regarded as preliminary because
bone ultrasound data were assessed at the heel and standard radiological measurements
of bone density at predilection sites of osteoporosis were not available.
Many of the previously established risk predictors of nontraumatic fractures
such as advanced age, postmenopausal status, impaired neuromuscular function,
and gene polymorphisms in the estrogen receptor α and collagen type
1 genes are considered to be responsible for loss of bone mass.20- 26 It
is commonly assumed, however, that resistance against fracture also depends
on the quality of bone.27 Although nature and
causality of the relationship between RANKL and fracture risk in our study
are speculative at this time, we believe that RANKL plays a key role in human
bone remodeling and, thus, affects bone quality. Hypothetically, low levels
of RANKL may be associated with a low degree of bone remodeling, unfavorable
bone microarchitecture, and, therefore, enhanced fracture risk. Lack of an
association of RANKL with osteocalcin and β-crosslaps in our population
is not in contradiction to this concept because continuous adaptation of bone
microarchitecture putatively occurs within a physiological range of bone turnover.
In recent years, several laboratory parameters have been identified
which predict or modify the risk of osteoporotic fractures, including β-crosslaps
and the aforementioned gene polymorphisms in the estrogen receptor α
and collagen type 1 genes.25,26,28 However,
no blood test has so far found broad access to the routine estimation of fracture
risk in the general community. RANKL is a promising candidate for closing
this gap, provided that our findings will hold true in future investigations.
The strengths of this study include its representative nature for the
general community (near-complete participation and follow-up), the broad age
range studied (40-89 years), and the high degree of accuracy in assessing
and classifying fractures in the setting of a population study. Several potential
limitations deserve consideration as well. First, clinically inapparent vertebral
fractures were not assessed. However, such events are of little relevance
from a clinical viewpoint, and nonassessment of minor disease phenotypes may
be expected to weaken evident relations rather than to create spurious ones.
Second, the study cohort consisted of whites only and was population-based.
Thus, results should not be extrapolated to other races or chronically ill
or immobilized patients. Third, the number of participants with nontraumatic
fractures in our study was comparatively low (n = 31), and, as a consequence,
95% CIs for risk estimates are broad. It must be acknowledged, however, that P values for the main findings are at a comfortable range
at <.001, data are highly consistent in subgroup and confirmation analyses,
and the lower limit of the 95% CI of fracture risk for the lowest vs highest
tertile group still exceeds 2 (Table 2).
Finally, in this study beginning in 1990, only a few participants were treated
with drugs affecting bone metabolism. Although this may be considered to be
an advantage for data analysis and accuracy, it should be noted that prescription
of these drugs is much more common currently.
In conclusion, our study identifies a low serum level of RANKL as a
novel and significant risk predictor of nontraumatic fractures in the general
community. This finding fits well into the hypothesis of a crucial role of
RANKL in human bone turnover and quality and may gain relevance in the routine
assessment of fracture risk. Clinical application of our findings awaits prior
confirmation in independent population samples. Moreover, standardized thresholds
and the sensitivity and specificity of the test remain to be defined for a
reliable interpretation of RANKL measurements in individual patients.