Clinical Endocrinology (2005)
, 131–138doi: 10.1111/j.1365-2265.2005.02312.x
© 2005 Blackwell Publishing Ltd
O R I G I N A L A R T I C L E
Blackwell Publishing, Ltd.
The relationship between serum resistin, leptin, adiponectin,
ghrelin levels and bone mineral density in middle-aged men
Ki Won Oh*, Won Young Lee‡, Eun Jung Rhee‡, Ki Hyun Baek§, Kun Ho Yoon§, Moo Il Kang§,
Eun Joo Yun†, Cheol Young Park*, Sung Hee Ihm*, Moon Gi Choi*, Hyung Joon Yoo* and Sung Woo Park*
Sungkyunkwan University School of Medicine, Seoul, Korea and
of Korea, College of Medicine, Seoul, Korea
*Internal Medicine and †Radiology, Hallym University, Chunchon, Korea,
‡Department of Internal Medicine,
Department of Internal Medicine, The Catholic University
Hormonal factors such as sex hormones, insulin, leptin and adiponectin
are thought to play a role in the mechanisms controlling the associa-
tion of body weight and fat mass with bone mass. However, contra-
dictory results have been reported for the association between serum
adipocytokines and bone mineral density (BMD). We therefore
examined whether the serum adipocytokine and ghrelin levels,
markers of fat metabolism, are associated with BMD in male adults.
Patients and measurements
For 80 male adults (average age
6·4 years; average body mass index (BMI) 24·4
the correlations between serum resistin, leptin, adiponectin and
ghrelin levels with BMD were investigated.
Among the adipocytokines, serum resistin levels were
negatively correlated with lumbar spine BMD (
After adjustment was made for age and BMI, log-transformed serum
leptin showed a significant negative correlation with lumbar spine
BMD, which was not seen on bivariate analysis (
Femoral neck BMD was marginally associated only with serum
adiponectin levels (
r = –0·226, P = 0·062). In multiple regression
analyses, among the adipokines, only resistin was a significant deter-
minant of lumbar spine BMD, although the variance was small
= 0·256). Serum ghrelin levels were not correlated with the BMD
of either body site.
Serum resistin level showed a significant negative
correlation with lumbar spine BMD, although the variance was
small. Further studies are needed to elucidate the role of adipocy-
tokines in bone metabolism.
Body weight is a significant predictor of bone mass.
r = – 0·237, P = 0·05).
r = – 0·237, P = 0·039).
(Received 9 October 2004; returned for revision 11 November 2004;
finally revised 29 March 2005; accepted 25 May 2005)
Body weight has been considered as a strong predictor of bone mass
for both males and females. Although the precise mechanism has not
been specifically clarified, hormonal factors such as sex hormones,
insulin, leptin and adiponectin have been postulated as being associated
with mechanisms between fat mass and bone mineral density (BMD).
Oestrogen is produced by the aromatization of androgen in adipose
tissue, and even in males oestrogen seems to be the major effector
for changes in bone metabolism.
to be associated with BMD, and it has been speculated that the
linkage of obesity and BMD might allow insulin to have a direct
mitogenic effect on the osteoblasts.
Leptin is the product of the ob gene that is primarily released by
white adipose tissue, and it is strongly correlated with fat mass.
Leptin receptors have been discovered in human bone marrow
stromal cells and are known to induce the differentiation of stromal
cells to osteoblasts.
The reduction in femoral length and weight has
been reported in obese leptin receptor-deficient (fa/fa) rats, and the
peripheral infusion of leptin to ovariectomized rats prevented their
further loss of bone.
By contrast, in leptin-deficient mice (ob/ob)
and leptin receptor-deficient mice (db/db), increases in bone formation
and bone mass have been reported, although the intracerebroven-
tricular infusion of leptin caused bone loss.
studies on the role of leptin in bone metabolism have shown con-
and in previous studies performed on women,
the relationship between serum leptin levels and BMD was not made
A recent study has suggested that the effect of leptin sig-
nalling on bone might differ significantly between different skeletal
and this might partially explain the contradictory results
of research on the association of leptin and bone metabolism.
Adiponectin is a recently discovered adipocytokine, an adipose-
specific collagen-like protein, highly expressed in adipocytes.
Adiponectin is negatively correlated with body mass index (BMI),
fat mass, fasting insulin levels and insulin resistance indices, and
it has anti-hyperglycaemic, anti-atherogenic and anti-inflammatory
Serum insulin levels are reported
Ki Won Oh and Won Young Lee contributed equally to this work and all
authors should be considered as first authors.
These results were presented in abstract form at the 26th Annual Meeting of
the American Society for Bone and Mineral Research (ASBMR).
Correspondence: Ki Won Oh, Division of Endocrinology and Metabolism,
Department of Internal Medicine, Sacred Heart Hospital, Hallym University,
#896 Pyungchon-dong, Dongan-gu, Anyang-city, Kyungki-do, Korea.
Tel.: + 82 31 380 3700; Fax: + 82 31 386 2269;
K. W. Oh et al.
© 2005 Blackwell Publishing Ltd, Clinical Endocrinology ,
of serum adiponectin levels and BMD, a clear association of adiponectin
with bone metabolism has not yet been elucidated.
Resistin is an adipocytokine that was discovered recently while
screening for substances that are down-regulated in response to
insulin-sensitizing anti-diabetic drugs.
in glucose homeostasis, insulin resistance and inflammation,
very few studies have reported on the correlation of resistin to bone
Ghrelin is an endogenous ligand of the growth hormone secret-
agogue receptor, and it increases fat mass by stimulating appetite and
food intake, and by reducing fat utilization.
reflect changes in body weight and fat mass, and serum ghrelin levels
are decreased with obesity and increased after diet-induced weight
However, in the cases of gastric bypass surgery-induced
weight loss, secretion of ghrelin was shown to be impaired, bone
turnover was accelerated, bone resorption was increased and bone
mass was decreased.
Although there might be an association
between ghrelin and bone, their precise relationship has not yet been
Therefore, to examine the relationship between serum adipocy-
tokines and bone metabolism in men, we analysed the correlation
of serum resistin, leptin, adiponectin and ghrelin levels with bone
metabolism in middle-aged men.
Although there is a report on the inverse association
Resistin seems to play a role
Serum ghrelin levels
Subjects and methods
Eighty Korean men (mean age 54·5
were selected from among the subjects who participated in the
annual health screening programme at MizMedi Hospital in Seoul,
Korea, from January 1 to December 31, 2002. A total of 561 men
participated during the period and we excluded those with diabetes
mellitus, hyperthyroidism, pituitary disease, hypogonadism, chronic
liver disease or chronic renal disease. Those men who were taking
medicine capable of affecting bone mineral metabolism such as ad-
renal and thyroid drugs, sex hormones, bisphosphonate, calcitonin,
diuretics, vitamin D or calcium supplements were also excluded
from the study, and the subjects were finally selected from those men
who were over 40 years old. All the subjects received BMD evalua-
tions. We measured height, weight and systolic and diastolic blood
pressures. The BMI was calculated as weight (kg) divided by
height (m) squared, and this was used as an index of overall
adiposity. Waist circumference was measured midway between
the lowest rib and the iliac crest, and hip circumference was taken
over the widest part of the gluteal region. The waist to hip ratio
(WHR) was used as the measure of central obesity. The Institu-
tional Review Board of MizMedi Hospital approved the protocol
used in this study, and informed consent was obtained from all the
6·4 years; range 42–70 years)
Measurements of serum adipocytokine levels
Blood samples were taken after an overnight fast. The serum was
separated and stored at
C until further use. The fasting blood
samples of the study population were analysed as follows. The serum
leptin level was measured by radioimmunoassay (RIA, LINCO
Research, USA), adiponectin was measured by enzyme-linked
immunosorbent assay (ELISA, B-Bridge International, Japan),
resistin was measured by enzyme immunoassay (Phoenix Pharma-
ceutical, USA), and ghrelin was measured by RIA (LINCO Research).
The intra-assay coefficient of variation (CV) for leptin was 3·4–8·3%
and the interassay CV was 3·0–6·2%; the intra-assay CV for adiponectin
was 4·6–5·8% and the interassay CV was 3·2–7·3%; the intra-assay
CV for resistin was lower than 5% and the interassay CV was lower
than 14%; the intra-assay CV for ghrelin was 6·5–9·5% and the inter-
assay CV was 9·6–16·2%.
Measurements of serum fasting glucose, insulin, lipid
profiles and sex hormone levels
After 12 h of fasting, the serum fasting blood glucose, total cholesterol,
triglyceride, high density lipoprotein cholesterol (HDL-C) and low
density lipoprotein cholesterol (LDL-C) levels were checked. The
hexokinase method was used to measure the blood glucose levels
and an enzymatic colorimetric test was used to measure the total
cholesterol and triglyceride levels. The selective inhibition method
was used to measure the level of HDL-C and the Friedewald equation
was used to calculate the level of LDL-C.
Serum insulin, oestradiol and total testosterone in fasting blood
samples were measured as follows. Serum insulin levels were
measured by immunoradiometric assay (RIABEAD II, Abbott,
Japan). Serum oestradiol levels were measured by chemiluminescent
competitive immunoassay (ADVIA Centaur Estradiol Assay, Bayer,
USA). Serum total testosterone levels were measured by RIA (Coat-
A-Count Total Testosterone, Diagnostic Products, USA). The
intra-assay CV for insulin was 1·2–1·9% and the interassay CV was
1·4–3·3%; the intra-assay CV for oestradiol was 4·0–12·1% and the
interassay CV was 4·5–8·1%; the interassay CV for total testosterone
Measurements of bone turnover markers
In assessing the biochemical markers of bone turnover, serum cross-
linked C-telopeptides of type I collagen (ICTP) was used as the bone
resorption marker and serum osteocalcin was used as the bone
formation marker. Serum ICTP levels (Telopeptide ICTP, Orion
Diagnostica, Finland) and serum osteocalcin levels (N-tact osteo SP,
INCSTAR, Stillwater, MN, USA) were measured by RIA. Regarding
the accuracy of the biochemical markers of bone turnover, the
intra-assay CV for the serum ICTP was 3·6% and interassay CV was
10·7%; the intra-assay CV for serum osteocalcin was 5·4% and the
interassay CV was 7·7%.
Bone mineral densitometry
Dual-energy X-ray absorptiometry (DEXA, XR-series, Norland,
USA) was used to measure the BMD of the lumbar spine and femoral
neck. Lumbar spine BMD was measured from the second to the
fourth lumbar vertebra and the femoral neck BMD was measured
from the left femoral neck. Regarding the precision of the lumbar
Adipocytokines and BMD in middle-aged men
© 2005 Blackwell Publishing Ltd, Clinical Endocrinology ,
spine BMD, the CV was 1·0%. The CV for the femoral neck BMD
was 1·2%. Osteoporosis and osteopenia were diagnosed according
to the World Health Organization definition: a
2·5 was diagnosed as osteoporosis and a
diagnosed as osteopenia. The T -values were calculated based on the
BMD of normal Korean males.
T -value lower than
-value lower than T
Data are presented as means
0·05 are considered as statistically significant. Evaluation of normality
was performed with the Shapiro–Wilk test and log-transformations
were performed for the serum leptin levels, which did not follow
a normal distribution. Correlation analyses between the serum
adipocytokines, ghrelin levels and anthropometric data, biochemical
results and BMD were performed with Pearson’s correlation test
and partial correlation analyses were performed to adjust for age
and BMI. Multiple regression analyses were used to determine the
predictors of BMD among the adipokines, and the variables that had
significant association with BMD on the bivariate correlation analyses
were selected as independent variables. The model selection was
carried out by the Backward method. The probability of F was used
to select the variables to be included in the model, the variables with
P -values less than 0·05 were entered and variables with
larger than 0·05 were removed from the model.
standard deviation. P -values less than
The general characteristics of the study subjects are presented in
Table 1. Among all the subjects, 19 (23·8%) had normal weights
(BMI < 23 kg/m
), 30 (37·4%) were overweight (23
and 31 (38·8%) were obese (BMI
Asia-Pacific criteria for obesity.
Based on the lumbar spine BMD, 5·0% of the subjects were diagnosed
as osteoporotic, 22·5% osteopenic, and 72·5% were diagnosed as
normal, and based on the femoral neck BMD, 3·8% of the subjects
were diagnosed as osteoporotic, 36·2% osteopenic, and 60·0% were
diagnosed as normal.
BMI < 25 kg/m
) according to the
Bivariate correlation analyses of serum adipocytokines and
ghrelin levels with anthropometrical and biochemical
Log-transformed serum leptin levels [log(leptin)] were positively
correlated to BMI, WHR, blood pressure, and fasting plasma glucose,
insulin, serum total cholesterol, triglyceride, LDL-C and serum
oestradiol levels (Fig. 1, Table 2).
Serum adiponectin levels were negatively correlated to BMI
and WHR, and positively correlated to HDL-C (Fig. 1, Table 2).
Serum resistin levels were positively correlated to LDL-C. Serum
ghrelin levels were negatively correlated to age, WHR and the serum
triglyceride level (Table 2).
None of the bone turnover markers showed a significant association
with the adipocytokines levels or the ghrelin levels, although the
serum calcium levels were significantly correlated to the log(leptin)
On the bivariate correlation analyses performed for the adipocy-
tokines and ghrelin, only adiponectin and resistin showed significant
negative correlation (r =
0·333, P = 0·005).
Table 1. General characteristic of the subjects
Value (mean ± SD)
Body mass index (kg/m2)
Serum leptin (ng/ml)
Serum adiponectin (µg/ml)
Serum resistin (ng/ml)
Serum ghrelin (pg/ml)
Fasting plasma glucose (mmol/l)
Serum total cholesterol (mmol/l)
Serum triglyceride (mmol/l)
Serum HDL cholesterol (mmol/l)
Serum LDL cholesterol (mmol/l)
Serum creatinine (µmol/l)
Serum calcium (mmol/l)
Serum ALP (U/l)
Serum osteocalcin (µg/l)
Serum ICTP (µg/l)
Serum testosterone (nmol/l)
Serum oestradiol (pmol/l)
Lumbar spine BMD (g/cm2)
Lumbar spine T-score
Femoral neck BMD (g/cm2)
Femoral neck T-score
54·5 ± 6·4
24·4 ± 2·5
2·9 ± 1·6
7·1 ± 3·9
79·7 ± 13·2
208·0 ± 99·7
5·2 ± 1·0
5·1 ± 1·1
1·7 ± 1·3
1·3 ± 0·3
3·0 ± 0·8
69 ± 15
2·3 ± 0·1
64·7 ± 40·5
5·1 ± 2·4
2·0 ± 1·3
19·9 ± 6·4
112·7 ± 63·1
1·00 ± 0·15
−0·4 ± 1·2
0·88 ± 0·15
−0·7 ± 1·2
HDL, high density lipoprotein; LDL, low density lipoprotein; ALP, alkaline
phosphatase; ICTP, cross-linked C-telopeptides of type I collagen; BMD,
bone mineral density.
Fig. 1 Correlation between serum adipocytokine levels and body mass index.
K. W. Oh et al.
© 2005 Blackwell Publishing Ltd, Clinical Endocrinology,
Bivariate correlation analyses of BMD with serum
adipocytokine levels, anthropometric and biochemical
Lumbar spine BMD was negatively correlated to age and positively
correlated to BMI, weight and serum triglyceride level (Table 3).
Femoral neck BMD was negatively correlated to age, and positively
correlated to BMI, weight, waist circumference, and fasting plasma
glucose (Table 3).
Serum resistin levels were inversely correlated to the lumbar spine
BMD (r =
0·237, P = 0·050), and the correlation was still significant
even after adjusting for age and BMI (
On bivariate correlation, serum log(leptin) levels were not correlated
with lumbar spine BMD or femoral neck BMD. However, after
adjusting for age and BMI, log(leptin) showed significant negative
correlation with lumbar spine BMD (
adiponectin nor ghrelin levels showed significant correlations with
the lumbar spine or femoral neck BMD, even after adjustments were
made for age and BMI (Table 3).
Serum oestradiol level showed a tendency for positive correlation
with the lumbar spine BMD, and the serum total testosterone level
showed negative correlations with lumbar spine and femoral neck
BMD on the bivariate correlation analyses (Table 3). This correlation
became nonsignificant after adjusting for other confounding vari-
ables in multiple regression analysis (Table 4).
0·308, P = 0·011).
0·237, P = 0·039). Neither
Multiple regression analyses assigning BMD as a
Multiple regression analyses were performed to clarify the determ-
inants of BMD after adjusting for the confounding variables. When
including age, BMI, triglyceride, oestradiol, testosterone, log(leptin)
and resistin levels as the independent variables, age, BMI and resistin
Table 2. Bivariate correlation analyses between serum adipocytokine levels with anthropometric and laboratory variables
Body mass index
Fasting plasma glucose
− − − −0·318
− − − −0·260
− − − −0·224
Correlation analyses were performed using Pearson’s correlation test.
HDL, high density lipoprotein; LDL, low density lipoprotein; ALP, alkaline phosphatase; ICTP, cross-linked C-telopeptides of type I collagen.
Table 3. Bivariate correlation analyses between bone mineral density with
serum adipocytokine levels and anthropometric and laboratory variables
Lumbar spine BMDFemoral neck BMD
Fasting plasma glucose
− − − −0·319
− − − −0·360
− − − −0·237
− − − −0·348
− − − −0·331
Correlation analyses were performed using Pearson’s correlation test.
*R = −0·237, P = 0·039 after adjustment for age and BMI.
**No correlations were seen even after adjustment for age and BMI with
partial correlation analyses.
†R = −0·308, P = 0·011 after adjustment for age and BMI.
BMI, body mass index; log(leptin), log-transformed serum leptin; BMD,
bone mineral density; HDL, high density lipoprotein; LDL, low density
lipoprotein; ALP, alkaline phosphatase; ICTP, cross-linked C-telopeptides of
type I collagen.
Adipocytokines and BMD in middle-aged men
© 2005 Blackwell Publishing Ltd, Clinical Endocrinology, 63, 131–138
were the significant determinants for lumbar spine BMD, although
the variance was small (Table 4). When including age, BMI, fasting
blood glucose, testosterone and serum adiponectin levels as the
independent variables, only age, BMI and fasting blood glucose
levels were significant determinants of femoral neck BMD (data not
In our study, multiple regression analysis showed that serum leptin
was not a significant determinant of BMD at either the lumbar or
vertebral sites. However, when adjustments were made for age
and BMI, a negative correlation emerged between lumbar BMD and
serum leptin, in line with a large study of 2761 men by Ruhl and
Everhart.51 The role of serum leptin in bone metabolism can only
be elucidated by studies in which the confounding variables, which
might independently influence both BMD and serum leptin, such
as BMI, sex hormone levels and insulin, are excluded. A further
potential source of error of discrepancies in the various studies that
have examined the relationship between serum leptin and BMD is
a differential effect of leptin on different skeletal regions, as proposed
recently by Hamrick et al.31 More studies are required to fully elucidate
the effect of leptin on bone metabolism in humans.
Our data demonstrate the marginal association of serum adiponectin
levels with femoral neck BMD. However, in multiple linear
regression analyses, adiponectin was not a determinant for BMD.
Few studies have examined the association between serum adiponectin
levels and BMD. In one, involving a small number of diabetic
patients, an inverse correlation was found,36 but in another, which
examined healthy perimenopausal women, no significant correla-
tion was detected.52 Adiponectin reportedly enhances the expression
of cyclooxygenase-2 (COX-2) in adipose tissue,53 and in osteoblasts
COX-2 can enhance differentiation.54,55 Although the association
between adiponectin and BMD was not significant in our study,
additional studies are necessary to examine the possibility that
adiponectin has an effect on bone formation through COX-2.
Our data have demonstrated for the first time an inverse correlation
between serum resistin levels in humans and lumbar BMD and
multiple regression analysis has confirmed that resistin was a significant
determinant of lumbar BMD. A number of studies have shown that
expression of resistin is regulated by peroxisome proliferator-
activated receptor gamma (PPARγ)56–58 and it is possible that the
correlation of serum resistin and BMD reflects the important effect
of PPARγ on differentiation of mesenchyme. Bone marrow mesen-
chymal cells are a common precursor of both osteoblasts and
adipocytes.59 Differentiation into the osteoblast lineage is inhibited
by PPARγ, while differentiation into adipocytes is stimulated.60,61
Several studies have shown that thiazolidenediones, agonists of
PPARγ, reduce bone turnover in patients with diabetes.62,63 These
studies and our current observations suggest a need for further
investigations of the potential for PPARγ and thiazolidenediones to
influence BMD in humans.
In our study, no association between serum ghrelin levels and
BMD was detected. No study has reported the direct association of
serum ghrelin levels and BMD, and studies concerning the effect of
ghrelin on osteoblasts or osteoclasts are scarce. In animal studies,
gastrectomy has been reported to induce osteopenia by reducing
ghrelin secretion.64 In humans, after gastric bypass surgery, a decrease
in serum ghrelin concentration, an increase in bone resorption
and a decrease in bone mass have been detected.49 In recent work
by Maccarinelli et al., significant stimulatory effects of ghrelin on
osteoblast proliferation and differentiation in rat calvaria cells were
observed.65 This raises the possibility that ghrelin might have a direct
effect on bone metabolism, despite the lack of evidence from this
No significant correlation was observed between serum total
testosterone level and BMD in our study. In previous studies in male
subjects, serum total testosterone level failed to show any correlation
with BMD in most of the studies that measured total testosterone
levels,7,66,67 although some of the studies that measured free testoster-
one levels reported a positive correlation with BMD.8,68 In the liter-
ature, total testosterone seems not to be correlated with BMD,
consistent with our current results. However, total serum testoster-
one levels are determined to a large extent by circulating SHBG levels,
which are dependent on several factors, the most important association
being with body fat.69 Although initial analysis showed a negative
correlation between total serum testosterone and BMD, it seems
likely that the loss of total serum testosterone as a determinant of
BMD is a result of a greater independent effect of body fat increase
on SHBG. Further analysis of free serum testosterone would be
relevant but was not possible in this current study.
There are several limitations in our study. First, the lifestyle of
the subjects, such as smoking and alcohol drinking habits, were not
considered as confounding factors. Smoking has been reported to
be negatively correlated with BMD changes in middle-aged men in
several studies, and the influence of alcohol on bone is probably
dependent upon age and amount of alcohol consumed.70–73 Second,
we did not measure total body fat with a DEXA scan and used BMI
as the only marker for body fat. In addition, we acknowledge that
the measured adipocytokines made only small contributions to the
total variance seen for BMD. Nevertheless, the serum resistin levels
were negatively associated with the lumbar spine BMD on multiple
Table 4. Multiple regression analysis of the variables; lumbar spine BMD as
the dependent variable
Body mass index
BMD, bone mineral density; log(leptin), log-transformed serum leptin.
Multiple regression analysis was performed with the Backward method and
when the probability of F was applied, the variables with P-values less than
0·05 were entered in the model and the variables with P-values larger than
0·05 were eliminated from the model.