|Home | About | Journals | Submit | Contact Us | Français|
To ascertain the predictive role of longitudinally-acquired biochemical measures of cartilage turnover in relation to x-ray-defined knee osteoarthritis (OAK), knee pain and functioning.
This is a feasibility study based on 72 enrollees of the Michigan site of Study of Women's Health Across the Nation, a longitudinal, population-based cohort study with 11 annual visits to characterize health at the mid-life. At visits in 1996, 1998 and 2007, radiographs were evaluated for the presence of OAK (≥2 on the Kellgren and Lawrence (K-L) scale). Knee pain and stiffness were assessed by interview. Functioning was assessed using the Western Ontario and McMaster Universities Osteoarthritis Index. Cartilage oligomeric matrix protein (COMP) and collagen Type II telopeptides (CTX-II) were assayed in serum and urine samples collected on alternate years from 1997 through 2006. We related trajectories of the cartilage biochemical markers from these five time points to OAK severity (no knee OA, K-L score<2; mild knee OA, K-L score=2; moderate/severe OAK, K-L score=3 or 4), pain, stiffness, or functioning, using longitudinal nonlinear mixed modeling.
The 2007 prevalence of x-ray defined OAK was 50% in these 72 women. Upward trajectories of COMP (P=0.02) and CTX-II (P=0.006) were associated with increased OAK severity and body size. COMP trajectories were associated with pain and stiffness, but not functioning. CTX-II trajectories were associated with stiffness scores, but not knee pain or functioning scores.
Multiple, biennial measures of COMP or CTX-II taken over a 10-year period were predictive of subsequent OAK and knee stiffness.
There has been substantial interest in the viability of measuring degradation products of joint tissues, especially of cartilage extracellular matrix. Two biochemical markers, cartilage oligomeric matrix protein (COMP) and crosslinked C-telopeptides of Type II collagen (CTX II) are potentially predictive of the development of knee OA and its attendant pain. However, there are few studies that have prospectively related biomarker measures of cartilage or collagen to x-ray-defined OAK, knee pain, stiffness or functioning in relation to the development of knee OA.
Functionally, COMP binds to type II collagen fibers and stabilizes the articular cartilage collagen fiber network. Serum COMP has been reported to have increasingly greater release into the synovial fluid and serum in patients with OA and rheumatoid arthritis1-7. Sharif et al. demonstrated that serum COMP levels were significantly higher among OA patients with progressive disease compared to those with non-progressive disease3 and that there was an association between changes in COMP levels and radiographic progression in knee OA over 5 years8. COMP has been shown to be positively correlated with three-year changes in joint space width9. Most recently, Hunter et al. demonstrated that a single measure of increased COMP predicted subsequent 30-month follow-up cartilage loss on MRI10.
Urinary CTX-II, a biochemical marker of type II collagen breakdown, has been negatively correlated with joint surface area and was an important predictor of joint damage assessed by radiographs11. Urinary CTX-II was significantly higher in participants with knee OA compared to controls11. More recent work has shown that increased CTX-II measures are related to knee OA progression as assessed with K-L scores12 or joint space width from radiographs13 and cartilage loss from MRI14,15. Though some have suggested that urinary CTX-II should be considered a bone marker rather than a cartilage marker, the fourth and most critical cartilage layer is at the bone interface and elevations of CTX-II may represent the more compromised cartilage16.
Much of the current data with biomarkers is based on a cross-sectional single-time assessment of biomarkers related to x-ray defined knee OA levels or with a single-time baseline measurement of biomarkers related to subsequent x-ray-defined knee OA progression. There is a dearth of studies with multiple longitudinal measurements of the biomarkers related to the incidence or increased severity (assessed with X-ray and/or clinical symptom report) of knee OA in a population-based sample. For that reason, the aims of this feasibility study were (1) to determine the likelihood that longitudinal measures of COMP and CTX-II, measured in specimens collected five times biennially over a 10-year period predicted the incidence of x-ray defined knee osteoarthritis, (2) to determine if longitudinal measures of COMP and CTX-II were related to knee OA severity, and (3) to relate longitudinal measures of COMP and CTX-II to changes in knee pain, knee stiffness, and knee functioning over a 10-year period.
Michigan SWAN is one of seven Study of Women's Health Across the Nation (SWAN) sites. SWAN is a multi-ethnic longitudinal study characterizing the endocrinological, physiological and behavioral changes that occur during the menopausal transition. Michigan SWAN includes a population-based sample identified in two Detroit-area communities of southeast Michigan who met the SWAN inclusion criteria. At baseline, women were aged 42−52, still menstruating, and not using exogenous hormone therapy. The Michigan population-based cohort of 543 women has been followed with annual assessments from baseline (1996−7) through 10 follow-up visits. Retention has been very good; the 10th follow-up visit included data collection from 83% of the still-living 525 participants. Michigan SWAN is the only SWAN site that includes studies of osteoarthritis.
This is a feasibility study using data from the first 72 women with knee x-rays from the Michigan SWAN cohort, examined between February 1 and June 1, 2007 who met the following three inclusion criteria: they had at least 5 biennial serum and urine specimens collected at the time of their annual SWAN visits between baseline and 2007 stored in a local repository; knee x-rays in 1996/7 and 1998/9 as well as 2007; and arthritis-related interview data for describing their knee pain, stiffness and functioning status at the same three time points. This report represents 360 data points (5 per participant, biannually) over a 10-year period. Age, body mass index, and ethnicity of these 72 women were not statistically different than these characteristics in other women of the cohort. The University of Michigan Institutional Review Board approved the study protocol and written informed consent was obtained from each participant.
Knee OA was defined with the Kellgren-Lawrence scale17 in grading structural hard tissue changes on x-ray images. Weight-bearing anterio-posterior radiographs obtained in a semi-flexed position18 were taken of both knees using General Electric radiographic equipment in 1996/7 and 1998/9 (model X-GE MPX-80; General Electric Medical Systems, Milwaukee, WI) and, in 2007, the AXIOM Aristos radiographic system with integrated digital flat detector technology (Erlangen, Germany). Radiographs were evaluated independently by 2 readers, with subsequent consensus reading for discordant scores. The interrater reliability was in excess of 0.85. Knee x-rays were scored for the presence of OA defined by the Kellgren and Lawrence (K-L) scale depicted in the Atlas of Standard Radiographs of Arthritis (0=normal, 1=doubtful OA, 2=minimal OA, 3=moderate OA, and 4=severe OA)17. This scale is based on the degree of osteophyte formation, joint space narrowing, sclerosis, and joint deformity. OA was defined as the presence of at least 1 knee with a grade of 2 or higher. Women with knee replacement (n=1) were classified as having severe OA.
Knee pain was characterized by asking if women had knee pain and if it occurred in more than 15 days of the month prior to interview19. Women were categorized as having no knee pain (0), knee pain for less than 15 days per month (1), or knee pain for more than 15 days per month (2). Stiffness was based on responses to two questions that asked if women had morning knee stiffness and whether that stiffness lasted more than 10 minutes. Women were categorized as having no knee stiffness (0), knee stiffness for less than 10 minutes (1), or knee stiffness for 10 minutes or more (2). Knee functioning was characterized with the 17-item WOMAC index which had been scaled to values between 0−10020,21. Women were apprised of their K-L x-ray scores approximately 6 weeks following their interviews and imaging.
At each annual visit, blood and urine specimens were collected fasting, between 7 a.m. and 11 a.m. and, whenever possible, in days 2−5 of the follicular phase of the menstrual cycle. Cycles became too irregular to index them against menstrual bleeding as women transitioned to the late menopause transition stage, but the specimen collection protocol still included the time of day and fasting requirements. Participants had been at rest for at least 30 minutes prior to blood drawing. Blood specimens were centrifuged, aliquoted, and stored in ultralow freezers where they were maintained without thawing until assay.
Urinary CTX-II was assayed with a competitive enzyme-linked immunosorbent assay (ELISA) based on a monoclonal antibody raised against the EKGPDP linear six-amino acid epitope of the type II collagen C telopeptide (CartiLaps, Nordic Biosciences, Herlev, Denmark)22. Intra-assay coefficients of variation (CV) were 6%, 8% and 10% in the high, medium and low range of measurements, respectively. Serum COMP was measured with a two-site ELISA using native human COMP as the standard and immunogen (COMPTM ELISA kit, AnaMar Medical)11. Intra-assay CVs were 2.6%, 3.3% and 7.1%, in the high, medium and low range of measurements. uCTX-II values were adjusted for urinary creatinine measured by a colormetric assay. All samples from each individual were measured in the same analytical run to avoid inter-assay variability.
Weight and height, measured annually with a calibrated balance beam scale and stadiometer, were used to calculate body mass index (BMI) [weight (kg)/height (m)2]. Obesity was defined as a BMI greater than 30 kg/m2.
Distributions of the measures of the biomarkers and body size were evaluated for presence of outliers and marked deviation from a normal distribution. Log transformations for the analytes and BMI were employed in data analyses; following data analysis, the variables were back-transformed for reporting the results unless otherwise noted.
Simple means and standard errors were used to describe the biomarkers, age, body size, and functioning characteristics according to x-ray defined OAK scores. Frequencies of OAK and pain were tabulated. T-tests were employed to assess the unadjusted differences in groups. Pearson correlations and 95% confidence intervals were use to relate biomarkers to age and BMI while Spearman correlations were used to relate the biomarkers to the WOMAC physical functioning score.
A longitudinal mixed modeling approach was used to evaluate COMP and CTX-II change over time in relation to the OA-related outcomes of interest. The five biennial COMP and CTX-II values from each woman, along with BMI, were incorporated into proportional odds (after testing for and rejecting non-proportional odds model) ordinal logistic mixed models using Proc NLMixed when the dependent variables were knee K-L score, pain score or stiffness score. For these analyses, we had three time points (baseline and follow-up visits 2 and 10) for the knee K-L scores and five time points for the biochemical markers (from follow-up visits 2, 4, 6, 8 and 10). Further, the logCOMP and logCTX-II values were incorporated, along with BMI, into longitudinal mixed models using Proc Mixed where the dependent variable was the continuous measure of physical functioning from WOMAC. The appropriateness of model fits were assessed both graphically and using residual analyses.
To estimate the predictiveness of the various levels of biomarkers’ change over time in relation to x-ray defined OAK, each woman's five values were fitted using mixed models (Proc Mixed) to generate a woman-specific intercept and slope. Then, receiver operator characteristic (ROC) curves were employed to display the sensitivity and 1-specificity, or the true positive rate against the false positive rate, of these slope values for predicting incident OAK as identified in the 2007 x-rays (75% of the 72 women who did not have a K-L score ≥ 2). The c-statistic represents the probability, based on the area under the curve, that those with greater markers change have OAK as compared to those without OAK23. The c-statistic is a goodness-of-fit measure of the ROC curve. Values in excess of .60 indicate discrimination by the markers, but good discrimination is identified when values are in excess of .80.
SAS 9.1 and Macro facilities (SAS Institute, Cary, NC) were used to perform the statistical analyses and plot the findings.
In 1997/8, the prevalence of x-ray defined knee OA (K-L ≥ 2) was 25% with 20% having a K-L score=2 (mild osteoarthritis) and 5% had a K-L score >2 (moderate to severe osteoarthritis) (see Table 1). In 2007, the prevalence of x-ray defined knee OA (K-L ≥ 2) was 50% with 22% of the sample having a K-L score of 2 (mild osteoarthritis) and 28% having a K-L score > 2 (moderate to severe osteoarthritis) (see Table 1). In 2007, the women with severe OA (K-L score 3 or 4) had a mean WOMAC physical functioning score of 15.8 ± 15.2 while the mean scores among those women with mild or no OAK were 10.6 ± 15.8 and 3.1 ± 8.7, respectively. The mean WOMAC score for the group with severe knee OA was statistically significantly higher than the mean WOMAC score for women without OA (P = 0.0003).
Using longitudinal non-linear mixed modeling, the upward trajectories of COMP and CTX-II, adjusted for BMI, were statistically significantly related to the severity of x-ray defined OAK, as shown in Table 2. In the first model, the intercept allowed a comparison of the K-L scores of 0−1 vs. K-L scores of 2−4 and the threshold2 variable compared K-L scores of 0−2 vs. 3−4. This model incorporated a test of the statistically significant contribution of COMP (P = 0.02) or BMI (P < 0.0001) to the odds of observing these relationships. Similarly, the second model incorporated a test of the statistically significant contribution of CTX-II (P = 0.006) or BMI (P < 0.0001) to the odds of observing these relationships.
COMP and CTX-II trajectories, adjusted for BMI, were statistically significantly and positively associated with stiffness scores, P < 0.007 and 0.04, respectively (Table 2). COMP trajectories were associated with increasing pain (P < 0.02), but not functioning scores (Table 2, Table 3). CTX-II trajectories were not associated with knee pain or physical functioning scores (Table 2, Table 3).
We examined the sensitivity and specificity for change over time in both COMP and CTX-II to predict the “gold standard” x-ray-defined incident knee OA (only those women with K-L scores of 0−1 at baseline) and depicted the findings in receiver operator curves shown in Figures 1 and and2.2. Change over time in CTX-II was slightly better than change in COMP over time in predicting x-ray defined incident knee OA, although neither biomarker was highly predictive as shown with the modest c-statistics of .60 for COMP and .64 for CTX-II. Further, it was difficult to identify a rate of change value for either biomarker above which there was a greater likelihood of observing incident knee OA defined by x-ray. Simultaneous inclusion of both biomarkers did not improve the prediction as identified with a c-statistic of the same magnitude.
Our five biennial measures of COMP and CTX-II from specimens collected over a 10-year period demonstrated progressive increases in profiles that were associated with subsequent OAK disease severity, defined from x-rays, and joint stiffness. This suggests that both biomarkers predict the development of knee osteoarthritis in middle-aged women over a 10-year period. We demonstrated that trajectories of COMP and CTX-II over a 10-year period of time were sensitive and specific for the prediction of incident x-ray defined OAK; however, the sensitivities and specificities were modest and did not allow ready identification of rates of change that were likely to result in the identification of incident x-ray defined OAK.
Serum concentrations of COMP have been considered a potentially prognostic indicator of joint damage and used to identify persons at higher risk of OA progression3,9,24,25. Higher levels of COMP have been reported in cross-sectional studies of persons with OAK26, in those with OAK and post-traumatic knee injury compared to controls1 and in differentiating persons with knee OA but with synovitis vs. no synovitis27. While there have been numerous studies of collagen markers, only three studies have related markers measured over time to measures of radiographic OAK3,8,9.
Upward COMP trajectories have not been shown to be highly specific for defining the initiation of OAK nor to defining the area of the joint where compromise is occurring. One potential reason for this lack of specificity may be that COMP expression can be identified in tissues other than articular cartilage including ligaments, tendons, menisci, and synovial membranes25,28 and pathology in these tissues would be more effectively characterized using magnetic resonance imaging than radiographs. A recent study has shown a modest association of COMP with MRI-defined cartilage losses10. This has lead investigators to speculate that COMP reflects inflammation as well as soft tissue turnover14, a speculation supported by the recent report that the metalloproteinase ADAMTS-7 was found to bind directly to and degrade COMP29. However, COMP is not related to C-reactive protein an acute inflammatory protein of liver origin29.
Mazieres et al. 30 identified that patients with painful hip OA in whom uCTX-II values were in the upper tertile of the distribution had a greater risk of progression (OR=3.7, 95% CI 2.5, 5.6) than patients in the lower two tertiles of the distribution. Further, a recent 5-year longitudinal study reported average CTX-II values were greater in progressors compared to non-progressors12. However, Mazzuca31 could not distinguish uCTX-II levels among women who were characterized as progressors vs. non-progressors in a study of obese women participating in a clinical trial of knee OA intervention. This contrast in findings may be related to assay characteristics as these studies did not use the same assay.
Immunochemistry studies of CTX-II in animals and humans indicate that the epitope is not only related to molecules at the cartilage surface but also the bone-to-cartilage interface at the calcified region32,33. As such, excretion of uCTX-II may be greater among those with more severe knee OA that includes cartilage defects that have full penetration to the bone34. Single-time measures of uCTX-II has been associated with severity of knee and hip OA13,30,35.
There are few longitudinal studies evaluating CTX-II change in relation to knee defined OA. Using 4 measures at 6 months intervals in a 2-year trial of bisphosphonate, there was a dose-dependent drop in uCTX-II observed at the 6-month visit that was maintained for the ensuing 18 months36. Further, a 3-month change in CTX-II, along with serum hyaluronic acid, predicted the 1-year change in medial and lateral tibial cartilage thickness assessed with MRI14.
Jung et al.37 identified that a uCTX-II value of 266 ng/mmol creatinine had a specificity of 87.5% and sensitivity of 75.5% for predicting x-ray defined knee OA, using the same assay employed in our study. The mean value of uCTX-II in our sample at baseline was lower than this value; in contrast, the mean value was higher than this cutpoint value 10 years later when the prevalence of x-ray defined OA in our study was 50%. We provide new information that indicates that while CTX-II change over time is significantly associated with x-ray defined knee OA, its predictive capability for incident x-ray defined knee OA is modest. Further, we could not identify an optimal cutpoint above which the rate of uCTX-II had good sensitivity and specificity.
This report has strengths and limitations. This was a study to assess the feasibility of examining markers in relation to the menopause transition hormone and thus includes only women. The relatively small sample size precludes us from testing interactions between the biomarkers and potentially important variables representing change in weight over the 10-year period. Strengths of the study include the longitudinal nature of the study that includes measures of characteristics important in OA including stiffness, joint pain, and knee functioning measured concurrently with the biomarkers. Other strengths include a subsample of women who are in the age range where knee OA is believed to develop and followed sufficiently long to see that development.
In summary, multiple, biennial measures of both sCOMP and uCTX-II measured in specimens collected over a 10-year period were associated with subsequent x-ray defined OAK and stiffness. Rarely are associations with stiffness reported. sCOMP values, but not uCTX-II, were associated with knee pain scores. Neither measure was associated with the WOMAC-derived knee functioning score. However, our evaluation of these biomarkers in relation to incident x-ray defined OAK suggests that their predictability is modest. Thus, both sCOMP and uCTX-II changes over time were associated with the initiation and severity of OAK defined by x-ray, but had modest sensitivity and specificity in predicting incident OAK.
The Study of Women's Health Across the Nation (SWAN) has grant support from the National Institutes of Health, Department of Health and Human Service, through the National Institute on Aging, the National Institute of Nursing Research and the National Institutes of Health Office of Research on Women's Health [NR004061, AG012505, AG012535, AG012531, AG012539, AG012546, AG012553, AG012554, AG012495]. The Michigan SWAN site-specific study of strength and functioning is supported by AG017104.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest statement
None of the authors have any conflicts of interest to disclose.