Increased urinary excretion of C-telopeptides of type II collagen (CTX-II)
predicts cartilage loss over 21 months by MRI
E. B. Dam Ph.D.*, I. Byrjalsen D.M.S.C., M. A. Karsdal Ph.D., P. Qvist Ph.D.
and C. Christiansen Ph.D., D.M.S.C.
Nordic Bioscience A/S, Herlev Hovedgade 207, DK-2730 Herlev, Denmark
Objective: Osteoarthritis (OA) is characterized by increased bone and cartilage metabolism leading to joint damage. The urinary excretion of
C-telopeptides of type II collagen (CTX-II) has earlier predicted progression in radiographic OA (ROA) e useful for participant selection in clin-
ical studies of potential disease modifying OA drugs (DMOADs). We investigated the longitudinal interrelationship between CTX-II and knee
cartilage volume quantified from magnetic resonance imaging (MRI).
Methods: We followed 158 subjects [48% females, 36 with knee ROA at baseline (BL)] for 21 months. The Kellgren and Lawrence (KL) index
and joint space width were assessed from radiographs (acquired load-bearing, semi-flexed). MRI scans were acquired from a 0.18 T Esaote
scanner (40?flip angle (FA), TR 50 ms, TE 16 ms, scan time 10 min, resolution 0.7 mm ?0.7 mm?0.8 mm) and medial tibial and femoral
cartilage volume was quantified. Radiographs and MRI were acquired at BL and follow-up. Fasting morning urine samples (second void)
were collected for BL CTX-II measurement.
Results: CTX-II was 56% higher in ROA subjects (P ¼0.0001). In addition, elevated BL CTX-II was associated with radiographic progression
(by KL or joint space narrowing) although not statistically significant. Contrarily, elevated BL CTX-II predicted longitudinal cartilage loss by MRI
(middle/high tertiles had odds ratios 4.0/3.9, P <0.01) corresponding to 3.1% increased yearly cartilage loss.
Conclusion: Prognostic markers in study selection criteria must ensure that placebo-treated participants progress to enable efficacy demon-
stration. And efficacy markers must allow progression detection within the study period. Our results support applying CTX-II for selection of
high risk subjects and applying the fully automatic MRI-based framework for quantification of cartilage loss.
ª 2008 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
Key words: Collagen type II, Cartilage, Volume, MRI, Prognostic.
Osteoarthritis (OA) is a slow, chronic disease characterized
by degradation of articular cartilage, leading to joint space
narrowing (JSN), loss of joint mobility, pain, and eventually
joint replacement. At present, no documented, effective
treatment is available [recent trials of potential disease
modifying osteoarthritis drugs (DMOADs) have shown no
convincing results1e4], which may both be a result of lack
of understanding of the disease but in addition lack of ap-
propriate means of disease assessment5e7.
Development of structure modifying drugs may be facili-
tated by the availability of appropriate biomarkers, which
can be integrated into all steps in the development pro-
cess7,8. The current gold standard JSN is slow and hetero-
geneous, and traditionally assessed from radiographs.
However, the slow longitudinal progression in JSN e rela-
tive to the variation e does not allow for small sample sizes
which are optimal for short term proof-of-concept phase II
clinical trials. Therefore, to reduce both time and cost of
the drug development process, there is a need for additional
tools such as biochemical markers and advanced imaging
technologies that will predict long term outcomes.
Biochemical markers of protease degraded cartilage matrix
constituents have been the subject of much attention from
many groups of researchers8. Some biochemical markers re-
spond to pathological activities resulting in increased turnover,
Levels of urinary C-telopeptides of type II collagen (CTX-II)
have been reported to be associated with risk of radiographic
disease10,11and elevated CTX-II has been associated with
a more severe structural damage assessed by JSN10.
Currently, structural joint damage is monitored by various
radiograph techniques and other scoring systems (e.g.,
whole organ MRI score (WORMS)12). As cartilage is not vi-
sualized on radiographs, alternative imaging technologies
are under validation and in particular magnetic resonance
imaging (MRI) seems promising. First and foremost, carti-
lage is visible in MRI, and by using 3D scans morphometric
analysis is possible. Several semi-automatic methods for
cartilage quantification have been reported13,14, and re-
cently our group reported a fully automatic computer-based
method for quantification of a range of morphometric pa-
rameters, including cartilage thickness, volume, curva-
ture15, and homogeneity16.
The present study was undertaken to investigate the in-
terrelationship between two biomarkers for assessment of
joint damage in OA: (1) the urinary excretion of CTX-II,
which is released during degradation of articular cartilage,
and (2) a newly developed MRI-based, fully automated car-
tilage volume quantification procedure.
*Address correspondence and reprint requests to: Dr Erik B.
Dam, Nordic Bioscience A/S, Herlev Hovedgade 207, DK-2730
Herlev, Denmark. Tel: 45-44525252; Fax: 45-44525251; E-mail:
Received 18 January 2008; revision accepted 23 July 2008.
Osteoarthritis and Cartilage (2009) 17, 384e389
ª 2008 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
Material and methods
The study population consisted of 159 subjects (48% females) above 21
years of age recruited to include individuals having OA according to Ameri-
can College of Rheumatology (ACR) criteria as well as healthy controls. The
majority of the subjects were invited from address lists to ensure an even dis-
tribution across gender and ages, but the population also contained volun-
teers with known knee problems (and therefore likely to have OA).
Individuals with other diseases affecting the joints (e.g., rheumatoid arthritis,
Paget’s disease, joint fractures, hyperparathyroidism, hyper- and hypothy-
roidism) were excluded as were individuals taking medication known to affect
bone and/or cartilage (e.g., bisphosphonates, vitamin D, hormones, SERMs,
prednisolone, anabolic androgens, and PTH). In addition, subjects with pre-
vious knee joint replacement or presenting any contraindication for MRI ex-
amination were excluded. All participants had radiographs and MRI of both
knees at baseline (BL) visit, and fasting morning urine was collected and
stored at ?20?C until analysis. Participants were invited to attend a follow-
up visit after 21 months with essentially the same investigations. For the
present analysis, 158 participants were available excluding one where a urine
sample was not obtained. For demographic details, see Table I.
The subjects signed informed consent forms. This single-center study
was conducted in accordance with the Helsinki Declaration II and European
Guidelines for Good Clinical Practice17. The study protocol was approved by
the local ethical committee.
standing in a weight bearing position with knees slightly flexed and the feet ro-
tated externally. The SynaFlex (Synarc, USA) was used to fix orientation and
sophalangeal view modified for fixed angle). Radiographs were acquired in the
posterioreanterior (PA) position, while the central beam was displayed directly
to the mid point of the line passing through both popliteal regions.
From the radiographs, the Kellgren and Lawrence (KL) score19was as-
sessed by a trained radiologist. As the study was designed to potentially de-
tect small differences, the BL and follow-up radiographs were examined pair-
wise blinded to the order. Furthermore, the joint space width (JSW) was
measured by manually marking the minimal distance between tibia and fe-
mur within the medial tibial plateau. The inter-scan quantification imprecision
for JSW (estimated from scanerescan pairs with a week in-between) was co-
efficient of variation (CV) 3.3% (root mean squared (RMS) CV 6.2%).
The method has previously been described15,20. Briefly, MRI scans were
acquired from an Esaote C-Span 0.18 T scanner using a modified Turbo 3D
T1 sequence with near-isotropic voxels (40?FA, TR 50 ms, TE 16 ms, scan
time 10 min, resolution 0.7 mm ?0.7 mm ?0.8 mm). During scanning, the
test subjects were lying supine without loading.
Automatic segmentation of the tibial and femoral medial cartilage com-
partments was performed using a voxel classification method based on su-
pervised learning and the total cartilage volume for the medial tibio-femoral
compartment was quantified. The sensitivity and specificity of the automatic
segmentation were 83.9 and 99.9%, respectively, compared to manual seg-
mentation by a radiologist. The inter-scan quantification imprecision for total
cartilage volume (estimated from scanerescan pairs with a week in-be-
tween) was CV 2.6% (RMS CV 3.6%).
Urinary excretion of CTX-II was quantified by the CartiLaps enzyme-
linked immunosorbent assay (ELISA) (Nordic Bioscience, Denmark) using
instructions from the manufacturer. Briefly, the competition ELISA employs
monoclonal antibody MAb F46 recognizing the six amino acid sequence
EKGPDP specific for the CTX-II. The intra- and inter-assay imprecision are
7.8 and 12.2% (RMS CV), respectively (according to the package insert).
The fasting morning (second void) urinary CTX-II concentration was cor-
rected for creatinine levels. To reduce the variability of the CTX-II measure-
ments, BL values were calculated as the mean of two separate
We used three alternative definitions for progression of OA:
Early radiographic progression: Those radiographically healthy at BL
(KL ?1) were divided into progressors/non-progressors depending on
whether any knee had an increase in KL score at follow-up. Note that
some of the progressors thereby remained radiographically healthy (KL
0 at BL and KL 1 at follow-up).
Early JSN: Again, we divided those healthy at BL. We defined JSN pro-
gressors to be above the median JSN of the radiographically healthy
non-progressors (as defined above).
Early cartilage loss: Analogously, we defined increased cartilage loss to be
above the median loss for the healthy non-progressors.
The CTX-II valueswere logarithmically transformed to obtainnormality and
symmetry of variance. The unpaired two-tailed Student t test was used for the
pair-wise comparison of BL characteristics between subjects with knee radio-
graphic OA (ROA) and controls, for assessment of differences between gen-
ders, and for assessment of the difference in JSW and cartilage volume
changes categorized according to the CTX-II levels. The Mantel-Haenszel es-
timate was used to test for significance of odds ratios (ORs).
The potential confounding effects of gender, age, and body mass index
(BMI) were investigated and linear correction was applied where relevant (as
stated below) to both BL and follow-up data.
For all tests P< 0.05 was considered significant. All statistical calcula-
tions were performed using the SAS software package (release 9.1, SAS In-
stitute Inc., Cary, NC, USA) and the Matlab software package (version 7.4.0,
Mathworks Inc., MA, USA).
The demographic data for the study population demon-
strated significant differences at BL among those individ-
uals having radiographic knee OA (ROA) as compared to
those without (radiographic disease is defined as having
a KL> 1 in one or both knees). Females and males with
ROA were about 14 years older, had slightly lower height
of 3e4 cm, had about 6 kg higher body weight, increased
BMI, and increased level of CTX-II (for details, see Table I).
Demographic characteristics of the study population at BL
?ROA (n ¼60)
þROA (n ¼16)
?ROA (n¼ 62)
Values given are mean (SD), andageometric mean (?1 SD range). The level of significance denotes for each gender the difference
between ?ROA and þROA with *P<0.5; **P< 0.01; ***P<0.001.
Osteoarthritis and Cartilage Vol. 17, No. 3
Among the 158 subjects, 76 had a KL score of 0 in both
imum KL of 2, 18 had maximum KL of 3, and one had KL 4.
Out of the 158 subjects being analyzed at BL, 138 (87%)
attended a follow-up visit at 21 months. There were no sig-
nificant differences in age, BMI, gender, JSW, OA state, or
cartilage volume between those that did and did not attend
At BL, we could not detect any associations between
CTX-II and JSW by radiographs (e.g., no linear correlation:
r ¼?0.06, P ¼ 0.5) or between CTX-II and cartilage volume
by MRI (r ¼ ?0.05, P ¼ 0.6). There were also no associa-
tions between CTX-II and gender, age, or BMI.
However, at BL, gender-specific linear correlations were
found between JSW and age and BMI, and gender-specific
correlations between cartilage volume and BMI. For all re-
sults below, we therefore also provide the results where
the data has been corrected accordingly (JSW corrected
for gender, age and BMI; volume for gender and BMI).
BL ASSOCIATIONS OF KNEE ROA AND CTX-II LEVELS
First, it was observed that in this study population, individ-
uals with radiographic disease in the knee had elevated uri-
nary excretion of type II collagen fragments [Fig. 1(a)].
Specifically, CTX-IIin the
(0.262e0.323) mg/mmol creatinine (mean ?1 S.E.M. range)
compared to 0.187 (0.177e0.197) mg/mmol creatinine in
the healthy population (P ¼0.0001).
Secondly, we determined if elevated urinary CTX-II was
indicative of presence of radiographic disease at BL. For
this investigation, the study population was stratified in ter-
tiles according to the urinary CTX-II level [Fig. 1(b)]. The OR
for having ROA in the high tertile compared to the low tertile
was 4.8 (P < 0.001).
ROA groupwas 0.291
PREDICTION OF EARLY RADIOGRAPHIC PROGRESSION
Next, we investigated if BL measurements of CTX-II
could predict structural degeneration during the 21
months follow-up period. Among the 109 subjects without
knee ROA at BL, 19% had an increase in the KL score
over the 21 months follow-up (Table II). Only three sub-
jects in the diseased group had progression, correspond-
ing to 10%.
We evaluated whether BL CTX-II was predictive of KL
progression among those healthy at BL. Although elevated
CTX-II implied a trend toward an elevated risk of longitudi-
nal progression with an OR of 2.3 when comparing the
low and high CTX-II tertile groups, this was not statistically
significant (Fig. 2). The groups with low and elevated CTX-II
were of similar ages (54 and 52 years, respectively,
Likewise, elevated CTX-II at BL was associated with in-
creased progression as determined by JSN (Fig. 3). Specif-
ically, the OR for progression for the middle and high tertiles
were 1.9 and 1.8 (with JSW corrected for gender, age, and
BMI: 2.0 and 1.8, respectively). However, this trend did not
reach statistical significance.
CTX-II WAS PREDICTIVE OF CARTILAGE LOSS DETERMINED
Finally, we compared BL measurements of urinary CTX-II
with changes over the 21 months in quantitative assess-
ments of cartilage volume by MRI. The subjects with BL
CTX-II in the middle and high tertiles had elevated risk for
increased cartilage loss (Fig. 4). Specifically, the OR for
the middle and high tertiles were 4.0 and 3.9 (P < 0.01 for
both). When correcting cartilage volume for gender, age,
and BMI, these ORs were 4.0 and 5.7 (P < 0.01 and
P <0.001, respectively). The groups with low and elevated
cartilage loss were of similar ages (55 and 51 years, re-
spectively, P ¼0.2).
Further, the mean yearly cartilage loss in the middle/
high CTX-II tertiles was 434 mm3/year higher than in the
low tertile (P <0.0001) e corresponding to 3.1% of the
mean cartilage volume. When cartilage volume was cor-
rected as above, the increased cartilage loss was
437 mm3/year or 3.2%.
Fig. 1. Relationship between CTX-II and ROA at BL: urinary excretion of CTX-II in study participants stratified according to the presence
(n ¼ 36) or the absence (n ¼122) of radiographic knee OA (ROA) at BL [Fig. 1(a)]. Individuals with a maximum KL score of 1 in either
knee were designated ?ROA, the remaining grouped in þROA. Subjects with ROA of the knee had BL urinary CTX-II levels of 0.291
(0.262e0.323) mg/mmol creatinine (mean ?1 S.E.M. range) compared to 0.187 (0.177e0.197) mg/mmol creatinine in the healthy population
(P ¼0.0001). The inverse relation was analyzed by the OR for having ROA (KL>1 in at least one knee) at BL when the population was strat-
ified based on the BL urinary CTX-II level [Fig. 1(b)]. The low tertile was used as a fiducial reference with OR¼1. The OR for ROA at BL for the
high tertile was 4.8 (P<0.001).
E. B. Dam et al.: Increased CTX-II predicts cartilage loss
We investigated the diagnostic and prognostic properties
of the CTX-II biomarker targeting urinary excretion of CTX-
II. Firstly, we evaluated the diagnostic ability to distinguish
the group with ROA. Secondly, we evaluated the prognostic
ability by investigating whether the BL CTX-II values pre-
dicted the longitudinal progression in KL score, JSW, and
In this study population, we observed a significant asso-
ciation between the presence of radiographic disease and
urinary CTX-II. This is in agreement with several previous
reports10,11,21e24(for a review of additional collagen type
II markers, see Ref. 25). In patients with ROA of the knee
we found the urinary CTX-II to be 0.291 mg/mmol creatinine,
corresponding to a 56% increase above the healthy con-
trols. Previously, increases of 3710and 126%21have
In contrast, we did not find an association between BL uri-
nary CTX-II and cartilage volume by MRI. Not even in the
healthy population. This suggests that the inter-individual
variation in metabolic activity of the articular cartilage is sig-
nificant. Interestingly, in the healthy sub-group of this popu-
lation with females and males having near-identical mean
ages of 52.5 and 52.9 years, we did not detect any gender
difference in CTX-II levels. In contrast, cartilage volume by
MRI was significantly higher in men than women (data not
shown). This discrepancy could be caused by the correction
of the biochemical index with creatinine, which will correct
for kidney function and muscle mass, in which muscle
mass is the larger determinant in between genders.
PROGRESSION DATA, RADIOGRAPHS
Some longitudinal studies have investigated using BL uri-
nary CTX-II for prediction of long term radiographic progres-
sion in OA10,26e30. In particular, for the Rotterdam Study,
Reijman and coworkers10found that subjects in the highest
Radiographic progression in the study population
Disease statusBL (n) Follow-up (n)Progression (n)
Radiographic disease progression in 138 subjects attending the
follow-up visit at 21 months. Radiographic progression was defined
as having an increase in KL score in one or both knees.
Fig. 2. The OR for longitudinal progression in the level of ROA
when the population was stratified based on the BL urinary CTX-
II level. We included the group of healthy at BL and defined pro-
gression as an increase in KL in either knee. The low CTX-II tertile
was used as a fiducial reference with OR ¼1. The OR for longitu-
dinal progression for the high tertile was 2.3 (not significant).
Fig. 3. The OR for predicting increased longitudinal JSN when the
population was stratified based on the BL urinary CTX-II level. In-
creased JSN was defined as a JSN above the mean JSN for the
group without no ROA at BL (KL ?1) and no longitudinal progres-
sion. The low CTX-II tertile was used as a fiducial reference with
OR ¼1. The OR for longitudinal progression for the middle and
high tertiles were 1.9 and 1.8 (not significant).
Fig. 4. Prediction of longitudinal cartilage loss was quantified by the
OR for elevated cartilage loss when the population is stratified into
tertiles by the BL CTX-II level. Only subjects without ROA at BL
were included. The low CTX-II tertile was used as a fiducial refer-
ence with OR¼1. The OR for longitudinal progression for the mid-
dle and high tertiles were 4.0 and 3.9 (P <0.01 for both).
Osteoarthritis and Cartilage Vol. 17, No. 3
quartile of CTX-II values had an increased risk of disease
progression in knee OA (defined as a JSN ?2 mm) of OR
6.2 (n ¼ 1235). Two of the other studies, both investigating
the associations for subjects in the upper tertile of CTX-II
values, reported a 23% decreased JSW for hip OA subjects
(n ¼ 376, Echodiah30) and an OR of 6.7 for rapidly progres-
sive hip OA (n ¼ 11529). Finally, other studies have also in-
vestigated associations between CTX-II and JSN, e.g.,
Mazie `res et al.26(n ¼ 333, also Echodiah, elevated CTX-II
gives twofold risk of JSN and hip replacement) and Sharif
et al.27(n ¼ 135, mild-to-moderate OA, CTX-II above
mean gives 3.4-fold risk of JSN or knee surgery) supporting
the relationship and Mazzuca et al.28(n ¼120, obese
women with OA) finding no relationship.
In the present study, the association between elevated
CTX-II and radiographic disease progression did not reach
statistical significance. Direct comparisons with the results
from the Echodiah30and Garnero29studies are problematic
since they investigated hip OA. However, one of the struc-
tural endpoints in the reports from the Rotterdam10and
Sharif27studies was JSN in the knee e similar to this study.
In the present study, the study period was relatively short
and the number of individuals having radiographic progres-
sion of knee OA was relatively small, i.e., only 19 individuals
without knee ROA at BL, which could explain the lack of
statistical significance (compared to the Reijman report
where they followed 1235 subjects over 6.6 years and the
Sharif report following 135 subjects for 5 years).
PROGRESSION DATA, MRI
We wanted to investigate if an alternative measure of
structural damage in knee OA could be predicted by eleva-
tion in the urinary CTX-II levels. To our knowledge, investi-
gations of the association of CTX-II scores and longitudinal
cartilage loss from MRI are rare. The first known was a study
by Bruyere et al.31, where they found no prediction of 1-year
cartilage loss by BL CTX-II values. However, they did find
a marginal association between 3-month CTX-II elevation
and 1-year cartilage thinning (comparing highest and lowest
quartiles, P ¼ 0.04, 62 OA subjects).
Recently, a fully automatic knee cartilage volume quanti-
fication framework, including assessment of cartilage vol-
ume, was developed by our group15,20, which was
reported to be highly correlated to manual assessment of
cartilage volume, able to discriminate groups of healthy
and OA subjects, associated with disease severity as deter-
mined by KL score, and quite precise with a mean inter-
scan CV of 2.6%. We therefore wanted to determine if
prediction of joint damage by BL CTX-II was improved using
MRI-based assessment of cartilage loss as the study end-
point. We found that high BL CTX-II reflecting cartilage deg-
radation was indeed significantly associated with a more
accelerated loss of cartilage volume e specifically, the mid-
dle/high tertiles had ORs of 4.0/3.9 compared to the low
tertile (P <0.01, results not reduced by correction for gen-
der, age, and BMI). It is noteworthy, that in the current study
of only 158 subjects of which 21 had early radiographic pro-
gression (13%) it was possible to show a statistical associ-
ation between the biochemical index at BL and the change
in structural readout in just 21 months.
Intuitively, the urinary excretion of the type II collagen
fragments should be more related to cartilage volume
than the radiographic observations of KL score and JSW.
KL score reflects both bone and cartilage morphology of
the joint and JSW is confounded by the 3D to 2D projection
and meniscus influence. This is confirmed by our results.
Despite a clear trend, the association between BL CTX-II
scores and longitudinal radiographic progression were not
statistically significant contrary to what larger, longer stud-
ies have demonstrated; but still our results showed a strong
association with cartilage loss quantified from MRI. This
supports that MRI may indeed be a more sensitive modality
than radiographs for quantifying progression of OA.
LIMITATIONS OF THE STUDY
We focused the investigation of progression of OA to the
early stages. Specifically, we analyzed the sub-population
without radiographic signs of OA at BL (KL ? 1). Therefore,
the conclusions are only valid for progression during the
early stages of OA. A study population with progressed
OA would be needed to validate the findings at later stages
of OA. Furthermore, the relatively small number of subjects
in this study also implies that the findings need to be vali-
dated on other populations.
The cartilage volume measurements were based on an
MRI scanner with a 0.18 T magnet. The use of low-field
MRI is sparsely validated compared to high-field MRI32. In
particular, high-field MRI may allow cartilage volume mea-
surements with higher accuracy and precision (implying
that studies may be conducted with a smaller population).
However, low-field MRI is much cheaper to install, maintain,
and operate. Future studies are needed to evaluate whether
low-field MRI can be a cost-effective alternative to conven-
tional high-field MRI for clinical studies.
MEDICAL INTERVENTION, CLINICAL STUDIES
It is a prerequisite for successful clinical investigations of
chondroprotective drugs, that further structural damage in
the study participants receiving placebo exceeds the level
of detection. Therefore it is essential to identify individuals
who in the follow-up period have accelerated structural
damage in the joint. A recent phase III clinical trial failed
to demonstrate a treatment effect of risedronate to reduce
radiographic progression (decrease in JSW) in patients
with knee OA2. In the study placebo group, 13% had radio-
graphic progression in the 2 year follow-up similar to the
13% with early radiographic progression in our population.
We could therefore speculate that use of the methodology
presented here e a combination of selection of high risk
subjects using CTX-II with quantification of efficacy using
our MRI-based quantification of cartilage loss e would po-
tentially have facilitated a different study outcome. Interest-
ingly, in a sub-group analysis of the risedronate trial, the
biochemical responders actually had significantly reduced
In addition to the relatively strong association between
CTX-II and the longitudinal cartilage loss by MRI demon-
strated in this study, the independence of operator interac-
tion of this fully automatic imaging procedure may facilitate
large scale studies without a potentially problematic single-
reader bottleneck or alternatively added measurement var-
iation implied by multiple readers.
Previous reports have shown that CTX-II is associated
with both the prevalence and the progression of ROA at
the knee; however, this is the first study to show that ele-
vated BL urinary CTX-II is predictive of cartilage loss
E. B. Dam et al.: Increased CTX-II predicts cartilage loss