ARTHRITIS & RHEUMATISM
Vol. 58, No. 12, December 2008, pp 3854–3864
© 2008, American College of Rheumatology
Inverse Association of General Joint Hypermobility With
Hand and Knee Osteoarthritis and
Serum Cartilage Oligomeric Matrix Protein Levels
Hsiang-Cheng Chen,1Svati H. Shah,2Yi-Ju Li,2Thomas V. Stabler,2Joanne M. Jordan,3
and Virginia Byers Kraus2
Objective. Extensive joint hypermobility, lower
serum cartilage oligomeric matrix protein (COMP)
levels, and early-onset osteoarthritis (OA) are pheno-
types of inherited pseudoachondroplasia and multiple
epiphyseal dysplasia. However, few studies have evalu-
ated the association between articular hypermobility
and primary OA. We undertook the present study to
evaluate this association and to test the hypothesis that
COMP levels are associated with hypermobility in pa-
tients with OA and individuals without OA.
Methods. Two separate cohorts were available for
analysis, the CARRIAGE (CARolinas Region Interac-
tion of Aging Genes and Environment) extended family
and a subset of the GOGO (Genetics of Generalized
Osteoarthritis) sibpair cohort. In the CARRIAGE fam-
ily, we performed hand and knee examinations and
hypermobility evaluations (Beighton criteria) and ob-
tained sera for measurement of COMP and hyaluronan
(HA). Data on COMP and HA levels and extensive joint
radiographic and hypermobility data were also avail-
able for the GOGO cohort.
Results. The prevalence of hypermobility was 13%
in the CARRIAGE family and 5% in the GOGO cohort.
In the CARRIAGE family, hypermobility was associated
with a significantly reduced prevalence of hand (espe-
cially proximal interphalangeal joint) and knee OA and
lower mean serum COMP levels, both in the total cohort
and in non-hand-OA subgroups. These results were
further validated in the GOGO subsets without radio-
graphic OA, in which hypermobility was also associated
with a significantly reduced mean serum COMP level
(P < 0.0001 adjusted for age). Serum HA levels did not
differ in relation to hypermobility in either cohort.
Conclusion. The present results indicate that
there is an inverse relationship between hypermobility
and hand and knee OA, and that hypermobility is
associated with lower serum COMP levels. Genetic
variations of the COMP gene may account for some
subgroups of benign joint hypermobility.
Osteoarthritis (OA) is a complex multifactorial
disorder that often results in chronic disability and is
associated with various risk factors (1), including older
age, obesity, female sex, muscle weakness, joint mal-
alignment, and genetic predisposition (2). Joint hyper-
mobility due to ligamentous laxity has empirically been
regarded to be a risk factor for OA (3), although the few
studies reported to date have had conflicting results
(Table 1). The true prevalence and risk of musculoskel-
etal disorders associated with hypermobility is unknown.
Joint hypermobility, estimated to affect ?5–25%
of the population depending on age, sex, and race (4),
occurs both as a lone benign trait and as a manifestation
of a variety of severe but rare heritable disorders,
including Marfan’s syndrome (1 in 12,000), Ehlers-
Supported by the National Institute on Aging, NIH (Claude
D. Pepper Older Americans Independence Centers program grant
2-P60AG11268), the Mary Duke Biddle Foundation, the Trent Foun-
dation, and GlaxoSmithKline. Dr. Chen’s work was supported by a
student grant from the Taiwanese government.
1Hsiang-Cheng Chen, MD: Duke University Medical Center,
Durham, North Carolina, and Tri-Service General Hospital and
National Defense Medical Center, Taipei, Taiwan;
MD, MHS, Yi-Ju Li, PhD, Thomas V. Stabler, MS, Virginia Byers
Kraus, MD, PhD: Duke University Medical Center, Durham, North
Carolina;3Joanne M. Jordan, MD, MPH: University of North Caro-
lina at Chapel Hill.
Dr. Kraus is a licensee of Phase Bioscience Inc. (PhaseBio),
Research Triangle Park, North Carolina, for a long-acting intraartic-
Address correspondence and reprint requests to Virginia
Byers Kraus, MD, PhD, Box 3416 Duke University Medical Center,
Durham, NC 27710. E-mail: firstname.lastname@example.org.
Submitted for publication February 15, 2008; accepted in
revised form August 24, 2008.
2Svati H. Shah,
Danlos syndrome (1 in 5,000), osteogenesis imperfecta
(1 in 100,000), pseudoachondroplasia (?1 in 200,000 in
the US), and some forms of multiple epiphyseal dyspla-
sia (1 in 10,000) (5,6). The latter conditions, pseudo-
achondroplasia and multiple epiphyseal dysplasia, are
characterized by prominent joint laxity and variable
short stature, short extremities, and early-onset OA,
especially in the hips and knees, and are due to muta-
tions in genes encoding for cartilage oligomeric matrix
protein (COMP) (7). During the ascertainment of the
CARRIAGE (CARolinas Region Interaction of Aging
Genes and Environment) family, we noted joint hyper-
mobility in many members. We therefore investigated
the association of joint hypermobility with clinical OA
phenotypes and with levels of the OA-related serum
biomarkers COMP and hyaluronan (HA). We then
validated our results in a larger family-based study, the
GOGO (Genetics of Generalized Osteoarthritis) study.
PATIENTS AND METHODS
Study populations. The CARRIAGE family study is a
prospective family-based longitudinal study of the interactions
between aging, genetic susceptibility, and environmental risk
pertaining to the development of several age-related chronic
diseases, including OA, cardiovascular disease, and eye dis-
eases (glaucoma and macular degeneration). This family came
to be studied in the context of health fairs we were requested
to conduct at several large family reunions. The extended
family described here is of mixed African American and
American Indian ancestry and is one of the most extensively
pedigreed existing families in the US, comprising 9 generations
with 3,357 pedigreed members, and originating from one
founder born in the 1700s (8). Ascertainment of 350 family
members was conducted during 3 family reunions between
2002 and 2006. Ascertainment included physician-performed
examinations for hand OA and joint hypermobility (n ? 287)
and knee OA (n ? 120). Height and weight were measured
and general medical history ascertained, including a query or
examination for blue sclerae (suggestive of osteogenesis im-
perfecta). The majority of the participants (n ? 278) also
consented to having blood drawn. For purposes of the present
analyses, we excluded 2 participants with known clinical rheu-
matoid arthritis and subjects younger than 25 years of age (n ?
5), to avoid potential confounding of the biomarker measures
by cartilage growth plate metabolism (9,10). After these exclu-
sions, full clinical data on the hand were available for a total of
280 participants, and both full clinical data on the hand and
biomarker data were available for 271 participants; 115 of the
latter group also had clinical knee examination data. Written
informed consent was obtained from each participant, and the
study was conducted with the approval of the Duke Institu-
tional Review Board.
The GOGO cohort is a large sample of Caucasian
sibling pairs and nuclear family members ascertained through
a collaborative consortium of 7 sites in the US and UK (11).
Full biomarker data, Beighton scores (12), and hand, knee, and
hip radiographic OA data were available on 708 individuals
from 2 of the sites (Duke University and University of North
Carolina at Chapel Hill). There were no individuals in either
the CARRIAGE family or the GOGO cohort with excessively
short stature or radiographic features (in the case of the
Summary of previous studies of OA and articular hypermobility*
Author, year (ref.)
Country of origin
or ethnic origin
of subjectsStudy design
Relationship of joint
hypermobility and OA
Scott et al, 1979 (36) UK Clinical population; age-
matched OA controls
9/100 (9) Beighton score ?4
and left MCP2
Beighton score ?5
OA increased in general;†
hand OA decreased;
knee OA increased
OA in 12 of 20 patients with
CMC1 joint OA increased;
DIP and PIP joint OA
DIP and PIP joint OA
Bridges et al, 1992 (3)US Clinical observation20/130 (15)
Jonsson et al, 1995 (34)IcelandPatients with established
19/100 (19) Beighton score ?4
Jonsson et al, 1996 (35)Iceland Female patients with
established thumb base
OA sibpair family study
17/50 (34)Beighton score ?4
Dolan et al, 2003 (33) UK79/716 (11)Beighton score ?1 Knee OA decreased
Kraus et al, 2004 (32)Caucasian (UK
39/1,043 (3.7)Beighton score ?4 PIP joint OA decreased
Present study Extended family–based
36/280 (13) Beighton score ?4 PIP joint and knee OA
decreased; serum COMP
*MCP2 ? second metacarpophalangeal; CMC1 ? first carpometacarpal; DIP ? distal interphalangeal; PIP ? proximal interphalangeal; COMP ?
cartilage oligomeric matrix protein.
† Type of osteoarthritis (OA) not reported.
JOINT HYPERMOBILITY, HAND AND KNEE OA, AND COMP LEVELS3855
GOGO cohort) suggestive of pseudoachondroplasia or multi-
ple epiphyseal dysplasia, nor was there anyone with blue
Survey of joint symptoms in the CARRIAGE family.
Information on self-reported joint symptoms was obtained by
asking, “Which of the following joints have bothered you in the
last year?” The surveyed joint sites included hands, knees, hips,
spine (neck, upper, lower back), ankles, shoulders, elbows,
wrists, and big toes.
Definitions of clinical hand and knee OA outcomes in
the CARRIAGE family. Findings of OA in the CARRIAGE
family, including clinical hand OA defined by a modification of
the American College of Rheumatology (ACR) criteria (13),
clinical hand OA defined by the GOGO criteria (11,14), and
clinical knee OA by ACR criteria (15), are listed and described
in Table 2. A non-hand-OA subgroup in the CARRIAGE
family was defined on the basis of not meeting the modified
ACR criteria or GOGO criteria for clinical hand OA.
Definitions of non-OA subgroups in the GOGO cohort.
The non-OA subgroups in the GOGO cohort were defined
using the available clinical data on the hand, as well as on the
basis of Kellgren/Lawrence grading of radiographic OA of the
hands, knees, or hips (16) (grade ?2 considered non-OA)
Beighton criteria for hypermobility. Hypermobility
was determined according to the criteria described in 1973 by
Beighton et al (12). Patients are graded on a 0–9 scale based on
their ability to achieve the following: (a) passive dorsiflexion of
the fifth finger ?90°; (b) passive apposition of the thumb to the
forearm; (c) hyperextension of the elbow ?10°; (d) hyperex-
tension of the knee ?10°; and (e) ability to rest the palms flat
on the floor with straight knees. Beighton scores were analyzed
as continuous traits and as binary traits; for binary trait
analyses, patients were considered to have hypermobility if
they scored ?4 of a possible 9 points (12). The Beighton
criteria have been shown to have high intra- and interrater
Biomarker analysis. Serum was isolated, aliquoted,
and stored at ?80°C within 4 hours of blood collection, until
biomarker analyses were performed. Serum biomarker assays
were performed in duplicate on each sample, and analyses
were repeated as necessary for samples with a coefficient of
variation (CV) of ?15%. COMP was measured with an
in-house sandwich enzyme-linked immunosorbent assay
Summary of investigations and definitions*
No. with outcome
(n ? 280)
(n ? 708)
Hand OA (modified
1) Hard tissue enlargement of ?2 of 10 selected joints;
2) hard tissue enlargement of ?2 DIP joints; 3) ?3
swollen MCP joints
1) ?3 joints with bone enlargement (of DIP or PIP
joints) or CMC1 joint squaring, and 2 of the 3 joints
involving the same joint group; 2) bone enlargement
of ?1 DIP joint of digits 2–5; 3) bilateral hand
involvement; 4) ?3 swollen MCP joints
Lacks hand OA by modified ACR criteria
Hand OA (GOGO
No hand OA (modified
No hand OA (GOGO
No DIP OA/no PIP OA/
no CMC1 OA
Lacks hand OA by GOGO criteria 228 ND
Lacks radiographic OA (K/L grade ?2 bilaterally in all
joints of the given group)
No knee OA/no hip OA/
no knee or hip OA
Lacks radiographic OA (K/L grade ?2 in joint group
Knee OA (ACR clinical
Knee pain and a minimum of 3 of 6 other features:
1) age ?50 years, 2) morning stiffness ?30 minutes,
3) crepitus, 4) bone tenderness, 5) bone
enlargement, 6) absence of palpable warmth
Beighton criteria treated as a continuous variable (0–9)
or as a dichotomous variable (non-hypermobility ?4;
Measured by ELISA or binding protein assay
Serum COMP and HA271 708
* For this study, the American College of Rheumatology (ACR) criteria for hand OA were modified in that symptoms and deformity were not
included. CARRIAGE ? CARolinas Region Interaction of Aging Genes and Environment; GOGO ? Genetics of Generalized Osteoarthritis;
ND ? not determined; K/L ? Kellgren/Lawrence; HA ? hyaluronan; ELISA ? enzyme-linked immunosorbent assay (see Table 1 for other
3856 CHEN ET AL
(ELISA) as previously described (18), using monoclonal anti-
bodies 17-C10 (epitope in the epidermal growth factor–like
domain) and 16F12 (epitope in the NH2-terminal domain)
against human COMP. The minimum detection limit is 120
ng/ml. Intra- and interassay CVs were ?5.8% and 8.7%,
respectively. HA was measured with an enzyme-linked binding
protein assay (Corgenix, Westminster, CO). The assay uses
enzyme-conjugated HA binding protein from bovine cartilage
to specifically capture HA from human serum. The minimum
detection limit is 10 ng/ml. Intra- and interassay CVs were
?4.7% and 7.0%, respectively.
Statistical analysis. The chi-square test was used to
assess the significance of the prevalence of joint symptoms and
prevalence of OA by hypermobility status. The Mann-Whitney
U test was used to assess the significance of the mean number
of OA-affected joints according to hypermobility status. Serum
biomarker concentrations were logarithmically transformed to
meet requirements of normality for parametric statistical
analyses. Two-sample t-tests were used to evaluate differences
in mean biomarker concentrations between the hypermobility
and non-hypermobility groups. One-way analysis of variance
with Tukey’s multiple comparison test was used to evaluate the
relationship between Beighton scores and concentrations of
biomarkers. A generalized estimating equation (GEE) was
used to control for dependency due to familial clustering of
CARRIAGE family members and GOGO nuclear families
(SAS version 9.1; SAS Institute, Cary, NC). For the CAR-
RIAGE family, we classified individuals into 8 clusters based
on their relationship to 8 members descended from the
founder of the CARRIAGE family. For the GOGO cohort,
individuals were clustered by family. All analyses were adjusted
for age, and additional adjustment for body mass index (BMI)
was performed for all analyses involving prevalence of clinical
or radiographic OA. Adjustment for hand OA status was
included in the logistic regression analysis of hypermobility
and COMP levels in the CARRIAGE family. Analyses of
hypermobility and COMP were performed in the full sample of
GOGO patients from Duke University and the University of
North Carolina and in subgroups without radiographic OA. P
values less than or equal to 0.05 were considered significant.
Hypermobility and joint symptoms in the CAR-
RIAGE family. Joint hypermobility (Beighton score ?4)
was present in 36 (12.9%) of the 280 CARRIAGE
family participants examined. The hypermobility group
and the non-hypermobility group did not differ signifi-
cantly in age or BMI (Table 3). The age distribution of
the 115 participants in the subgroup with knee examina-
tion data was similar to that of the group as a whole, with
the hypermobility group having a slightly, but nonsignifi-
cantly, younger mean age (mean ? SD 53.5 ? 14.2 years
in the hypermobility group and 58.3 ? 14.7 years in the
non-hypermobility group; P ? 0.15). The hypermobility
group had a female predominance compared with the
non-hypermobility group. The prevalence of hand and
knee joint symptoms was lower in the hypermobility
group, but the prevalence of symptoms in the other
joints was similar (Table 3).
Hypermobility and OA in the CARRIAGE family.
Compared with the group without hypermobility, the
hypermobility group in the CARRIAGE family had a
lower prevalence of hand OA by the modified ACR and
GOGO criteria, and a lower prevalence of knee OA by
participants (n ? 280)*
Demographic characteristics of the CARRIAGE (CARolinas Region Interaction of Aging Genes and Environment) family study
(Beighton score ?4)
(n ? 36)
(Beighton score ?4)
(n ? 244)P†
Age, mean ? SD years
BMI, mean ? SD kg/m2
Self-reported joint symptoms
53.27 ? 14.38
29.67 ? 6.09
56.36 ? 14.64
31.23 ? 6.67
* Except where indicated otherwise, values are the number (%) (based on a total group of 25 in the hypermobility group and 224 in the
non-hypermobility group with available joint symptom data).
† By unpaired t-test for evaluation of age and body mass index (BMI); by likelihood ratio chi-square test for evaluation of self-reported joint
JOINT HYPERMOBILITY, HAND AND KNEE OA, AND COMP LEVELS3857
the ACR criteria (Table 4). By logistic regression,
hypermobility in the CARRIAGE family was found to
be associated with a decreased likelihood ratio of hand
OA (P ? 0.02 for hand OA according to the modified
ACR criteria; P ? 0.008 for hand OA according to the
GOGO criteria), which remained significant after ad-
justment for age and BMI (for OA by the modified ACR
criteria, P ? 0.024 adjusted for BMI, P ? 0.043 adjusted
for age, P ? 0.047 adjusted for age and BMI; for OA by
the GOGO criteria, P ? 0.009 adjusted for BMI, P ?
0.018 adjusted for age, P ? 0.02 adjusted for age and
BMI) (Table 4). Compared with the non-hypermobility
group, the hypermobility group exhibited significantly
fewer proximal interphalangeal (PIP) joints clinically
affected with OA (P ? 0.005), and a nonsignificant
reduction in the prevalence of OA of the distal inter-
phalangeal (DIP) and first carpometacarpal (CMC1)
joints. Taking all 3 hand joint groups into consideration,
hypermobility was significantly inversely associated with
the number of OA-affected joints (Table 4). Hypermo-
bility was also associated with a decreased likelihood
ratio of knee OA (P ? 0.02 [P ? 0.035 adjusted for BMI,
P ? 0.058 adjusted for age, P ? 0.068 adjusted for age
Hypermobility and biomarkers in the CAR-
RIAGE family. Mean ? SD ln serum COMP levels
decreased significantly with increasing hypermobility
(7.48 ? 0.42 for subjects with a Beighton score of 0,
7.37 ? 0.42 for those with a Beighton score of 1–3,
7.29 ? 0.43 for those with a Beighton of score ?4 (P ?
0.034 after adjustment for age) (Figure 1a). The mean ln
serum COMP level was significantly lower in the hyper-
mobility group versus the non-hypermobility group (P ?
0.05) (Figure 1b); however, mean ln serum HA levels did
not differ significantly between the 2 groups (Figure 1c).
After adjustment for age and hand OA status, the mean
ln serum COMP level was marginally significantly lower
in the hypermobility group compared with the non-
hypermobility group (for hand OA by GOGO criteria,
P ? 0.069; for hand OA by modified ACR criteria, P ?
To more clearly define the association of hyper-
mobility and serum COMP independent of hand OA
status, we analyzed the relationship in the non-hand-OA
subgroup of the CARRIAGE family. The mean ln
serum COMP level was also significantly lower in the
non-hand-OA subgroup with hypermobility compared
with the members without hypermobility. When hand
OA was excluded on the basis of clinical GOGO criteria,
mean ? SD ln serum COMP levels were 7.26 ? 0.43 in
the hypermobility group versus 7.42 ? 0.43 in the
non-hypermobility group (P ? 0.053 adjusted for age);
when hand OA was excluded on the basis of the
modified ACR criteria, ln serum COMP levels were
7.27 ? 0.42 in the hypermobility group versus 7.42 ?
0.41 in the non-hypermobility group (P ? 0.049 adjusted
for age) (Figures 1d and 1e). When familial clusters were
taken into account with the GEE analysis, the inverse
association between hypermobility (using Beighton
score as a continuous covariate) and serum COMP level
Clinical OA status by hypermobility status in the CARRIAGE family*
(Beighton score ?4)
(Beighton score ?4)P‡
Age- and BMI-
Hand (GOGO criteria) (n ? 280)
GOGO criterion 1
GOGO criterion 2
GOGO criterion 3
Hand (modified ACR criteria) (n ? 280)
ACR criterion 1
ACR criterion 2
Knee (ACR criteria) (n ? 115)
Joint site involvement (n ? 280), mean ? SD
No. of joints with OA
DIP ? PIP ? CMC1 joints
0.22 ? 0.63
0.14 ? 0.49
0.028 ? 0.17
0.39 ? 0.99
0.60 ? 1.35
1.01 ? 1.97
0.14 ? 0.48
1.75 ? 3.12
* In the studies of hand OA, n ? 36 and 244 in the hypermobility group and the non-hypermobility group, respectively; in the studies of knee OA,
n ? 25 and 90 in the hypermobility group and the non-hypermobility group, respectively. Except where indicated otherwise, values are the number
(%). CARRIAGE ? CARolinas Region Interaction of Aging Genes and Environment (see Table 1 for other definitions).
† See Table 2 for description of the modified American College of Rheumatology (ACR) criteria and the Genetics of Generalized Osteoarthritis
‡ By likelihood ratio chi-square test for OA definition; by Mann-Whitney U test for joint site involvement.
3858CHEN ET AL
remained highly significant (P ? 0.0035 adjusted for
age). When hypermobility was analyzed as a binary trait
(Beighton score ?4 or ?4), the results were consistent
but not significant (P ? 0.082 adjusted for age).
Hypermobility and biomarkers in the GOGO
cohort. Joint hypermobility was present in 36 (5%) of
the 708 GOGO study participants. Consistent with the
CARRIAGE family results, the mean ? SD ln serum
COMP levels in the GOGO cohort also decreased
significantly with increasing hypermobility (6.94 ? 0.53
in subjects with a Beighton score of 0, 6.72 ? 0.53 in
subjects with a Beighton score of 1–3, and 6.63 ? 0.54 in
subjects with a Beighton score of ?4) (P ? 0.0001
adjusted for age). The mean ln serum COMP level was
consistently lower in the hypermobility group compared
with the non-hypermobility group (6.60 ? 0.61 versus
6.91 ? 0.53; P ? 0.0009) (6.63 ? 0.54 versus 6.90 ? 0.53;
P ? 0.004 adjusted for age). In contrast, before and after
controlling for age, ln serum HA levels were similar in
the 2 groups (mean ? SD 3.44 ? 0.88 versus 3.63 ? 0.85;
P ? 0.2) (3.63 ? 0.81 versus 3.62 ? 0.80; P ? 0.9
adjusted for age).
To evaluate the possibility that OA itself might
cause an apparent diminution in manifestations of hy-
permobility due to loss of joint range of motion, we
repeated the analyses in the individuals without radio-
graphic OA. For each joint site, the mean ln serum
COMP level was significantly lower in the group with
Figure 1. Relationship between serum biomarker levels and joint hypermobility in the total cohort and the subgroups without hand osteoarthritis
(OA) of the CARRIAGE (CARolinas Region Interaction of Aging Genes and Environment) family. a, Log-transformed serum cartilage oligomeric
matrix protein (COMP) levels in the total cohort (after adjustment for age) by hypermobility status, defined by dividing Beighton scores into 3 groups
(no hypermobility [score 0], moderate hypermobility [score 1–3], and high hypermobility [score ?4]). b, Log-transformed serum COMP levels in the
total cohort by hypermobility status. c, Log-transformed serum hyaluronan (HA) levels in the total cohort by hypermobility status. d and e,
Log-transformed serum COMP levels by hypermobility status in the subjects without clinical hand OA as assessed by the GOGO (Genetics of
Generalized Osteoarthritis) study criteria (d) (n ? 228 [34 with and 194 without joint hypermobility]) or the modified American College of
Rheumatology (ACR) criteria (e) (n ? 233 [34 with and 199 without joint hypermobility]). P values were calculated by one-way analysis of variance
(ANOVA) with Tukey’s multiple comparison (a) or by 2-sample t-test (b–e). Data are presented as box plots, where the boxes represent the 25th
to 75th percentiles, the lines within the boxes represent the mean, and the lines outside the boxes represent the minimum and maximum values. NS ?
JOINT HYPERMOBILITY, HAND AND KNEE OA, AND COMP LEVELS3859
joint hypermobility (Figure 2). Results on the 251 indi-
viduals who had no radiographic evidence of either hip
OA or knee OA were similar, with lower serum COMP
levels in the hypermobility group (P ? 0.002 adjusted for
age). In contrast, mean ln serum HA levels did not differ
on the basis of hypermobility status in the non–
radiographic OA groups (data not shown). Similar to
findings in the CARRIAGE family data set, GEE
analysis of data from the GOGO cohort revealed a
highly significant inverse association between hypermo-
bility (using Beighton score as a continuous covariable)
and serum COMP level (P ? 0.0001 adjusted for age).
When hypermobility was analyzed as a binary trait (?4
or ?4), the inverse association with serum COMP level
was also apparent (P ? 0.01 adjusted for age).
The present results provide evidence of a rela-
tionship between serum COMP levels and general joint
hypermobility. COMP is a 524-kd homopentameric non-
collagenous glycoprotein that is derived from cartilage
and also found in ligaments and tendons (19). Recent in
vitro studies have shown that COMP can interact with
types I, II, and IX collagen, fibronectin, and all matrilins
(20–23) and that COMP can bind to types I, II, and IX
collagen with high affinity (24). Interestingly, some
autosomal-dominant osteochondrodysplasias (pseudo-
achondroplasia and some cases of multiple epiphyseal
dysplasia) are caused by COMP mutations that interfere
with normal extracellular matrix assembly, which is
thought to contribute to the development of the disease
phenotypes (25,26). Pronounced hypermobility and low
serum COMP levels are features of these osteochondro-
Low serum COMP levels may result from reten-
tion of mutant COMP within the rough endoplasmic
reticulum of chondrocytes and tendon cells (29). How-
ever, not all of the COMP-associated chondrodysplasias
Figure 2. Log-transformed COMP levels by hypermobility status in the non-OA subgroups of the GOGO cohort. a, Subjects without hand OA as
assessed by the ACR criteria (n ? 167). b, Subjects without radiographic OA of the distal interphalangeal (DIP) joints (n ? 77). c, Subjects without
radiographic OA of the proximal interphalangeal (PIP) joints (n ? 135). d, Subjects without radiographic OA of the first carpometacarpal (CMC1)
joints (n ? 333). e, Subjects without radiographic OA of the knees (n ? 374). f, Subjects without radiographic OA of the hips (n ? 399).
Hypermobility was defined as a Beighton score of ?4. P values (adjusted for age) were calculated by 2-sample t-test with equal variance. Data are
presented as box plots, where the boxes represent the 25th to 75th percentiles, the lines within the boxes represent the mean, and the lines outside
the boxes represent the minimum and maximum values. K/L ? Kellgren/Lawrence (see Figure 1 for other definitions).
3860 CHEN ET AL
appear to be storage diseases (25,26); therefore, other
as-yet-undefined mechanisms, such as altered COMP
protein or RNA synthesis or stability, may account for
low serum COMP levels in these chondrodysplasias. By
analogy, genetic variation within the COMP gene might
influence both serum COMP levels and ligamentous
structure leading to articular hypermobility phenotypes
in the CARRIAGE family and GOGO cohort. Of note,
in a recent study of an Icelandic cohort of 331 subjects,
Jonsson et al found linkage (logarithm of odds score 3.8)
of joint hypermobility (?90° dorsiflexion of either fifth
finger) to chromosome 19P 13.3, which is within 16 Mb
of the COMP gene (30). Also, Hakim and colleagues
have reported autosomal-dominant inheritance of be-
nign joint hypermobility affecting female twins (31).
We also demonstrated that generalized articular
hypermobility is inversely associated with clinical hand
(PIP joint) OA and possibly also knee OA. This confirms
and extends our previous results showing that, in the
GOGO cohort, hypermobility was associated with a
lower prevalence of PIP joint OA and possibly meta-
carpophalangeal (MCP) joint OA (32). The present
study showed that, after accounting for age, the preva-
lence of PIP joint and knee OA was lower among
subjects with joint hypermobility, with a similar trend
observed for DIP and CMC1 joint OA. In the previously
reported study of hypermobility in the GOGO cohort,
no conclusions could be drawn regarding hypermobility
and DIP joints, because study inclusion required the
presence of OA in at least 1 DIP joint in the proband
and 1 sibling (11,32). No such inclusion criteria were
used in the CARRIAGE family study, and we observed
a trend toward fewer OA-affected DIP joints in associ-
ation with hypermobility. It is possible that with a larger
sample size or radiographic phenotyping, the inverse
relationship of hypermobility and OA of the DIP, MCP,
and CMC1 joints might be further validated.
Our results are also consistent with those of a
recent community-based study of postmenopausal
women, showing a reduced risk of radiographic knee OA
among those with joint hypermobility (33). Preliminary
data from another study, in a cohort of Icelandic subjects
(n ? 1,839) with a 31% prevalence of any hypermobility,
have also shown a reduction in clinical knee OA in
association with hypermobility (P ? 0.04 by chi-square
analysis) (Jonsson H: personal communication). More-
over, in 2 separate studies, Jonsson and colleagues found
a reduced frequency of OA of the hand IP joints in
association with hypermobility (34,35). In contrast, some
studies have demonstrated a higher prevalence of OA in
association with hypermobility. Decades ago, 2 studies
showed a higher prevalence of OA among individuals
with joint hypermobility, from groups of highly selected
patients referred for clinical evaluation (3,36); however,
in 1 of those studies, different effects were observed for
the knee (increased OA in association with hypermobil-
ity) compared with the hand (decreased OA in associa-
tion with hypermobility) (36). Jonsson et al also reported
more frequent CMC1 joint OA in association with
hypermobility (34). Thus, the type and strength of the
effect of hypermobility on OA susceptibility may differ
by joint group.
Although several previous studies have empha-
sized an association of the benign joint hypermobility
syndrome with musculoskeletal symptoms (37–39), our
study showed that CARRIAGE family members with
joint hypermobility had a lower prevalence of self-
reported symptoms in their hand and knee joints, com-
pared with participants without hypermobility. More-
over, in studies of children and adolescents, it has been
found that not all subjects deemed to have hypermobility
have a history of musculoskeletal symptoms and disor-
ders or develop them later in life (40). In accordance
with our results, Larsson et al also found a lower
prevalence of hand symptoms in musicians with lax
fingers performing repetitive fine hand movements,
compared with their peers with less flexibility (41). The
effect of hypermobility on symptoms may be specific to
particular joint groups since hypermobility appeared to
increase low back symptoms in timpani players (41).
Thus, again, joint group is one possible factor explaining
some of the differences in associations between
hypermobility-related symptoms and OA risk.
There are many other possible factors that might
explain the conflicting reports on associations of hyper-
mobility, osteoarthritis, and musculoskeletal symptoms.
Murray has suggested that hypermobility alone may not
account for musculoskeletal syndromes but that other
cofactors, such as obesity, sedentary lifestyle, or joint
overuse, may be important moderators of the symptoms
and outcomes of hypermobility (40). Murray also sug-
gested that there may be a high-risk subgroup of chil-
dren in whom hypermobility would be underrecognized
when hypermobility is defined using the typical Beighton
score thresholds for adults (40). Mechanical joint forces
may vary on a joint-specific basis due to ligamentous
laxity. It is possible that individuals with hypermobility
may moderate their activity due to pain or joint insta-
bility, which may reduce the risk of OA.
Finally, individuals with hypermobility represent
both a phenotypically and a genotypically heterogeneous
group. In addition to pseudoachondroplasia and multi-
JOINT HYPERMOBILITY, HAND AND KNEE OA, AND COMP LEVELS3861
ple epiphyseal dysplasia (COMP mutations), hypermo-
bility in variable degrees is associated with other genetic
syndromes including, among others, Marfan’s syndrome
(fibrillin mutations) and Ehlers-Danlos syndrome (mu-
tations of COL1A1, COL1A2, COL3A1, COL5A1,
COL5A2, ADAMTS-2, and tenascin XB) (42). In these
conditions, factors other than hypermobility may ac-
count for joint symptoms and OA, but failure to recog-
nize these difficult-to-diagnose underlying conditions
may contribute to the general tendency to ascribe joint
symptoms and OA to the presence of hypermobility.
These issues require further exploration in future studies.
A strength of the present study is that all family
members were invited to participate and included, inde-
pendent of hypermobility status or signs or symptoms of
musculoskeletal disorders. Althouth it is possible that
the healthier family members may have been more likely
to attend the family reunions, we thus avoided the
common selection bias of most other studies related to
hypermobility, which have relied on clinic-based popu-
lations with a high prevalence of joint symptoms. These
family data may therefore be more representative of the
Although we obtained comparable results in 2
separate cohorts, some shortcomings of the present
study remain. The study was limited by its cross-sectional
nature, and therefore, we cannot completely rule out the
possibility that OA masks the manifestations of hyper-
mobility, although OA seldom affects the wrists or fifth
MCP joints that contribute to the Beighton score. How-
ever, we believe that the inverse association of hyper-
mobility and OA in our study was not due to waning
joint laxity with age and OA, because the negative
association persisted after adjustment for age. We con-
sidered the possibility that lower COMP levels in the
hypermobility group might be a manifestation of
younger age and fewer OA-affected joints since serum
COMP levels are positively associated with severity of
radiographic OA (18,43). Therefore, we also analyzed
the non-OA participants to evaluate the association of
hypermobility with levels of both biomarkers (COMP
and HA). Serum COMP levels were consistently lower
in association with hypermobility in both cohorts of
non-OA subjects. In contrast, levels of HA, a marker of
joint tissue turnover in OA (44), were unchanged in
association with hypermobility. These findings thus dem-
onstrate that the association of hypermobility with lower
prevalence of OA and lower serum COMP levels was
not a result of age or of hypermobility being masked by
To address the possibility that COMP fragmen-
tation was a cause of lower serum COMP levels in
subjects with hypermobility, we performed additional
sandwich ELISAs and Western blot analyses (data not
shown) on sera from a test subset of 9 individuals (some
with and some without hypermobility, some with and
some without OA, some with low and some with high
COMP levels), using 3 different monoclonal antibodies
to the COMP amino-, middle, and carboxy-terminus as
described previously (45) (reducing gels for antibodies
16F12 and 12C4; nonreducing gel for antibody 17-C10).
Specifically, we evaluated these sera for evidence of
COMP fragments of 50–90 kd, as described previously in
a report of a study using polyclonal antibodies against
human COMP (46). There were no 50–90-kd COMP
fragments in the serum of any of the test subjects, but
COMP fragments of this size were readily discerned in
the positive control (EDTA cartilage extract). Findings
of ELISAs using combinations of these 3 anti-COMP
antibodies also supported the notion that full-length
COMP was present in the serum (data not shown). Thus,
the lower serum COMP levels associated with hypermo-
bility were due to absolute lower levels of full-length
COMP in the serum, rather than targeted degradation.
In summary, we report evidence of an inverse
relationship between serum COMP levels and general
joint hypermobility. In addition, in the extended CAR-
RIAGE family of mixed African American and Ameri-
can Indian heritage, we have replicated the results from
the GOGO Caucasian sibpair cohort, in which we dem-
onstrated that general articular hypermobility was in-
versely associated with OA of the PIP joints. Our
findings also suggest the possibility of an inverse associ-
ation between joint hypermobility and knee OA, al-
though this result needs to be validated in a larger
cohort. We hypothesize that hypermobility may decrease
biomechanical strain on joints, and this in turn would
confer protection against hand and knee OA. Hypermo-
bility is a heritable trait associated with lower serum
COMP levels and COMP mutations in some osteochon-
drodystrophies. The present results suggest that genetic
variation within the COMP gene may be a candidate to
account for benign joint hypermobility, a condition
whose etiology is as yet unknown. The results also
suggest that the extent of hypermobility might serve as a
quantitative trait for identifying alleles that are protec-
tive against OA.
We would like to thank Dr. William Kraus for funding
assistance, Dr. Vladimir Vilim for the kind gift of the 16F12/
3862 CHEN ET AL
17C10 monoclonal antibodies, Norine Hall, Carol Haynes, and
Sarah Nelson for database management, and everyone who
made the family reunions possible.
Dr. Kraus had full access to all of the data in the study and
takes responsibility for the integrity of the data and the accuracy of the
Study design. Chen, Shah, Kraus.
Acquisition of data. Chen, Shah, Stabler, Kraus.
Analysis and interpretation of data. Chen, Shah, Li, Jordan, Kraus.
Manuscript preparation. Chen, Shah, Li, Jordan, Kraus.
Statistical analysis. Chen, Li, Kraus.
1. Jordan JM, Kraus VB, Hochberg MC. Genetics of osteoarthritis.
Curr Rheumatol Rep 2004;6:7–13.
2. Felson DT, Lawrence RC, Dieppe PA, Hirsch R, Helmick CG,
Jordan JM, et al. Osteoarthritis: new insights. Part 1: the disease
and its risk factors. Ann Intern Med 2000;133:635–46.
3. Bridges AJ, Smith E, Reid J. Joint hypermobility in adults referred
to rheumatology clinics. Ann Rheum Dis 1992;51:793–6.
4. Grahame R. Joint hypermobility and genetic collagen disorders:
are they related? Arch Dis Child 1999;80:188–91.
5. Bravo JF, Wolff C. Clinical study of hereditary disorders of
connective tissues in a Chilean population: joint hypermobility
syndrome and vascular Ehlers-Danlos syndrome. Arthritis Rheum
6. Briggs MD, Chapman KL. Pseudoachondroplasia and multiple
epiphyseal dysplasia: mutation review, molecular interactions, and
genotype to phenotype correlations. Hum Mutat 2002;19:465–78.
7. Briggs MD, Hoffman SM, King LM, Olsen AS, Mohrenweiser H,
Leroy JG, et al. Pseudoachondroplasia and multiple epiphyseal
dysplasia due to mutations in the cartilage oligomeric matrix
protein gene. Nat Genet 1995;10:330–6.
8. Chen HC, Shah S, Stabler TV, Li YJ, Kraus VB. Biomarkers
associated with clinical phenotypes of hand osteoarthritis in a large
multigenerational family: the CARRIAGE family study. Osteoar-
thritis Cartilage 2008;16:1054–9.
9. Urakami T, Manki A, Inoue T, Oda M, Tanaka H, Morishima T.
Clinical significance of decreased serum concentration of cartilage
oligomeric matrix protein in systemic juvenile idiopathic arthritis.
J Rheumatol 2006;33:996–1000.
10. Thonar E. Serum keratan sulfate concentration as a measure of
the catabolism of cartilage proteoglycans. In: Brandt K, editor.
Cartilage changes in osteoarthritis. Indianapolis: Indiana Univer-
sity School of Medicine; 1990. p. 105–8.
11. Kraus VB, Jordan JM, Doherty M, Wilson AG, Moskowitz R,
Hochberg M, et al. The Genetics of Generalized Osteoarthritis
(GOGO) study: study design and evaluation of osteoarthritis
phenotypes. Osteoarthritis Cartilage 2007;15:120–7.
12. Beighton P, Solomon L, Soskolne CL. Articular mobility in an
African population. Ann Rheum Dis 1973;32:413–8.
13. Altman R, Alarcon G, Appelrouth D, Bloch D, Borenstein D,
Brandt K, et al. The American College of Rheumatology criteria
for the classification and reporting of osteoarthritis of the hand.
Arthritis Rheum 1990;33:1601–10.
14. Kraus VB, Varju G, Li Y, Martin E, Jordan JM, Renner JB, et al.
Assessment of the usefulness of clinical hand examination for
determination of OA affected status for genetic studies [abstract].
Arthritis Rheum 2001;44 Suppl 9:S140.
15. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et
al. Development of criteria for the classification and reporting of
osteoarthritis: classification of osteoarthritis of the knee. Arthritis
16. Kellgren JH, Lawrence JS. Radiological assessment of osteo-
arthrosis. Ann Rheum Dis 1957;16:494–502.
17. Boyle KL, Witt P, Riegger-Krugh C. Intrarater and interrater
reliability of the Beighton and Horan Joint Mobility Index. J Athl
18. Vilim V, Olejarova M, Machacek S, Gatterova J, Kraus VB,
Pavelka K. Serum levels of cartilage oligomeric matrix protein
(COMP) correlate with radiographic progression of knee osteoar-
thritis. Osteoarthritis Cartilage 2002;10:707–13.
19. Muller G, Michel A, Altenburg E. COMP (cartilage oligomeric
matrix protein) is synthesized in ligament, tendon, meniscus, and
articular cartilage. Connect Tissue Res 1998;39:233–44.
20. Mann HH, Ozbek S, Engel J, Paulsson M, Wagener R. Interac-
tions between the cartilage oligomeric matrix protein and matri-
lins: implications for matrix assembly and the pathogenesis of
chondrodysplasias. J Biol Chem 2004;279:25294–8.
21. Di Cesare PE, Chen FS, Moergelin M, Carlson CS, Leslie MP,
Perris R, et al. Matrix-matrix interaction of cartilage oligomeric
matrix protein and fibronectin. Matrix Biol 2002;21:461–70.
22. Holden P, Meadows RS, Chapman KL, Grant ME, Kadler KE,
Briggs MD. Cartilage oligomeric matrix protein interacts with type
IX collagen, and disruptions to these interactions identify a
pathogenetic mechanism in a bone dysplasia family. J Biol Chem
23. Rosenberg K, Olsson H, Morgelin M, Heinegard D. Cartilage
oligomeric matrix protein shows high affinity zinc-dependent
interaction with triple helical collagen. J Biol Chem 1998;273:
24. Thur J, Rosenberg K, Nitsche DP, Pihlajamaa T, Ala-Kokko L,
Heinegard D, et al. Mutations in cartilage oligomeric matrix
protein causing pseudoachondroplasia and multiple epiphyseal
dysplasia affect binding of calcium and collagen I, II, and IX. J Biol
25. Schmitz M, Becker A, Schmitz A, Weirich C, Paulsson M, Zaucke
F, et al. Disruption of extracellular matrix structure may cause
pseudoachondroplasia phenotypes in the absence of impaired
cartilage oligomeric matrix protein secretion. J Biol Chem 2006;
26. Chen TL, Stevens JW, Cole WG, Hecht JT, Vertel BM. Cell-type
specific trafficking of expressed mutant COMP in a cell culture
model for PSACH. Matrix Biol 2004;23:433–44.
27. Mabuchi A, Momohara S, Ohashi H, Takatori Y, Haga N,
Nishimura G, et al. Circulating COMP is decreased in pseudoa-
chondroplasia and multiple epiphyseal dysplasia patients carrying
COMP mutations. Am J Med Genet A 2004;129A:35–8.
28. Tufan AC, Satiroglu-Tufan NL, Jackson GC, Semerci CN, Solak S,
Yagci B. Serum or plasma cartilage oligomeric matrix protein con-
centration as a diagnostic marker in pseudoachondroplasia: differen-
tial diagnosis of a family. Eur J Hum Genet 2007;15:1023–8.
29. Weirich C, Keene DR, Kirsch K, Heil M, Neumann E, Dinser R.
Expression of PSACH-associated mutant COMP in tendon fibro-
blasts leads to increased apoptotic cell death irrespective of the
secretory characteristics of mutant COMP. Matrix Biol 2007;26:
30. Jonsson H, Ingvarsson T, Hauksson V, Petursson H, Kristjansson
K, Stefansson K. Linkage analysis for hand hypermobility suggests
a susceptibility gene on chromosome 19P [abstract]. Osteoarthritis
Cartilage 2007:15 Suppl C:C162.
31. Hakim AJ, Cherkas LF, Grahame R, Spector TD, MacGregor AJ.
The genetic epidemiology of joint hypermobility: a population
study of female twins. Arthritis Rheum 2004;50:2640–4.
32. Kraus VB, Li YJ, Martin ER, Jordan JM, Renner JB, Doherty M,
et al. Articular hypermobility is a protective factor for hand
osteoarthritis. Arthritis Rheum 2004;50:2178–83.
33. Dolan AL, Hart DJ, Doyle DV, Grahame R, Spector TD. The
JOINT HYPERMOBILITY, HAND AND KNEE OA, AND COMP LEVELS3863
relationship of joint hypermobility, bone mineral density, and Download full-text
osteoarthritis in the general population: the Chingford Study.
J Rheumatol 2003;30:799–803.
34. Jonsson H, Valtysdottir ST. Hypermobility features in patients
with hand osteoarthritis. Osteoarthritis Cartilage 1995;3:1–5.
35. Jonsson H, Valtysdottir ST, Kjartansson O, Brekkan A. Hypermo-
bility associated with osteoarthritis of the thumb base: a clinical
and radiological subset of hand osteoarthritis. Ann Rheum Dis
36. Scott D, Bird H, Wright V. Joint laxity leading to osteoarthrosis.
Rheumatol Rehabil 1979;18:167–9.
37. Adib N, Davies K, Grahame R, Woo P, Murray KJ. Joint
hypermobility syndrome in childhood: a not so benign multisystem
disorder? Rheumatology (Oxford) 2005;44:744–50.
38. Remvig L, Jensen DV, Ward RC. Epidemiology of general joint
hypermobility and basis for the proposed criteria for benign joint
hypermobility syndrome: review of the literature. J Rheumatol
39. Ferrell WR, Tennant N, Baxendale RH, Kusel M, Sturrock RD.
Musculoskeletal reflex function in the joint hypermobility syn-
drome. Arthritis Rheum 2007;57:1329–33.
40. Murray KJ. Hypermobility disorders in children and adolescents.
Best Pract Res Clin Rheumatol 2006;20:329–51.
41. Larsson LG, Baum J, Mudholkar GS, Kollia GD. Benefits and
disadvantages of joint hypermobility among musicians. N Engl
J Med 1993;329:1079–82.
42. Wikipedia contributors. Ehlers-Danlos syndrome. In: Wikipedia,
The Free Encyclopedia; 2008. Resource number 216287626 (ac-
cessed 12 June 2008).
43. Kelman A, Lui L, Yao W, Krumme A, Nevitt M, Lane NE.
Association of higher levels of serum cartilage oligomeric matrix
protein and N-telopeptide crosslinks with the development of
radiographic hip osteoarthritis in elderly women. Arthritis Rheum
44. Elliott AL, Kraus VB, Luta G, Stabler T, Renner JB, Woodard J,
et al. Serum hyaluronan levels and radiographic knee and hip
osteoarthritis in African Americans and Caucasians in the
Johnston County Osteoarthritis Project. Arthritis Rheum 2005;52:
45. Vilim V, Lenz ME, Vytasek R, Masuda K, Pavelka K, Kuettner
KE, et al. Characterization of monoclonal antibodies recognizing
different fragments of cartilage oligomeric matrix protein in
human body fluids. Arch Biochem Biophys 1997;341:8–16.
46. Neidhart M, Hauser N, Paulsson M, DiCesare PE, Michel BA,
Hauselmann HJ. Small fragments of cartilage oligomeric matrix
protein in synovial fluid and serum as markers for cartilage
degradation. Br J Rheumatol 1997;36:1151–60.
3864 CHEN ET AL