Background Biomarkers of cartilage metabolism have prognostic potential.
Objective To examine whether serum cartilage biomarkers, cartilage oligomeric matrix protein (COMP), N-propeptide of type IIA procollagen (PIIANP), type II collagen breakdown product (collagen type-II cleavage (C2C)) predict cartilage volume loss and knee joint replacement.
Methods 117 subjects with knee osteoarthritis (OA) had MRI at baseline and 2 years. Cartilage biomarkers were measured at baseline. Change in knee cartilage volume over 2 years and knee joint replacement over 4 years was determined. The population was divided into subgroups with high or low cartilage biomarkers (based on biomarker levels greater than or equal to, or less than, the mean, respectively). The relationships between biomarkers and outcome measures were examined in the whole population, and separately in marker subgroups.
Results The relationship between cartilage biomarkers and cartilage volume loss was not linear across the whole population. In the low (regression coefficient B=–9.7, 95% CI –0.01 to 0.003, p=0.01), but not high (B=–0.46, 95% CI –8.9 to 8.0, p=0.92) COMP subgroup, COMP was significantly associated with a reduced rate of medial cartilage volume loss (p for difference between groups=0.05). Similarly, in the low (B=–8.2, 95% CI –12.9 to –3.5, p=0.001) but not high (B=2.6, 95% CI –3.3 to 8.5, p=0.38) PIIANP subgroup, PIIANP was associated with a significantly reduced rate of medial volume cartilage loss (p for difference=0.003). C2C was not significantly associated with rate of cartilage volume loss. PIIANP was associated with a reduced risk of joint replacement (odds ratio (OR)=0.28, 95% CI 0.10 to 0.93, p=0.04).
Conclusion Cartilage biomarkers may be used to identify subgroups among those with clinical knee OA in whom disease progresses at different rates. This may facilitate our understanding of the pathogenesis of disease and allow us to differentiate phenotypes of disease within a heterogeneous knee OA population.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Osteoarthritis (OA) is a major cause of pain and long-term disability, resulting in extensive use of healthcare resources. The societal and economic burden of this disease is only expected to increase owing to increased longevity and prevalence of obesity. Despite the significant burden of OA, the ability to predict structural change and identify those patients who will progress to the extent of requiring joint replacement is limited. Identifying better predictors of progression would allow more effective allocation of scarce healthcare resources, and reduce the overall burden of this disease.
Loss of articular cartilage is a marker of the severity of OA.1 Biomarkers that reflect changes in cartilage metabolism may have prognostic potential. Since type II collagen is a major component of the cartilage matrix, there is increasing interest in biomarkers of type II collagen turnover.2,–,5 Such biomarkers include the N-propeptide of type IIA procollagen (PIIANP)6 which reflects synthesis of articular cartilage type IIA collagen, and collagen type-II cleavage (C2C), a product of type II collagen breakdown.7 Although it has been suggested that type II collagen synthesis and degradation is altered in OA,8,–,10 previous longitudinal studies examining the relationship between biomarkers of type II collagen turnover and disease progression have provided inconsistent results,2 3 or were unable to identify significant relationships.4 5 It may be that using a ratio of type II collagen synthesis to degradation is beneficial; indeed it has been shown to be more sensitive for investigating disease.2 4 11
Cartilage oligomeric matrix protein (COMP) is an important component of the articular cartilage extracellular matrix, which may also be expressed in other joint tissues, including ligaments, tendons, menisci and synovium.12 13 Alterations in metabolism may affect cartilage integrity and stability.14,–,16 Although it has been suggested that elevated levels of COMP reflect synovitis,17 and the presence and severity of OA,18 19 previous longitudinal studies examining the relationship between COMP and OA progression have provided conflicting results.5 20,–,22
Thus the aim of this study was to examine the relationship between biomarkers of cartilage metabolism and clinical outcome measures of joint structure. This study is particularly novel as very few studies have examined the relationship between cartilage biomarkers and cartilage volume loss assessed by MRI, as well as a joint replacement, a clinically important outcome.
Patients and methods
One hundred and thirty-two subjects aged >40 years with knee OA were recruited by advertising through local newspapers and the Victorian branch of the Arthritis Foundation of Australia and in collaboration with general practitioners, rheumatologists and orthopaedic surgeons as described.23 The study was approved by the ethics committee of the Alfred and Caulfield hospitals in Melbourne, Australia. All patients gave informed consent.
All subjects fulfilled the American College of Rheumatology clinical and radiographic criteria for knee OA23, with pain at baseline in at least one pain dimension of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score >20% and osteophytes present within the knee. Subjects were excluded if any other form of arthritis was present, if there was any contraindication to MRI, if a total knee replacement was planned, or if they were unable to cooperate with study requirements.
At baseline, weight was measured to the nearest 0.1 kg (shoes, socks and bulky clothing removed) using a single pair of electronic scales. Height was measured to the nearest 0.1 cm (shoes and socks removed) using a stadiometer. Body mass index (weight/height2 in kg/m2) was calculated. Knee pain (0–500), stiffness (0–200) and function (0–1700) were assessed with the WOMAC score at baseline, where zero represents no symptoms.24 A list of drugs was collected. Only one subject reported bisphosphonate treatment, and three had been receiving glucosamine for more than 6 months.
Each subject had an MRI performed on the symptomatic knee at baseline, and the mean±SD time for follow-up scan was 1.95±0.21 years later. Where both knees were symptomatic, the knee with least severe radiographic change was imaged to minimise loss to follow-up. Knees were imaged as previously described.25 26 (See online supplementary file 1 for further details.)
Cartilage volume measurement
Tibial cartilage volume was determined by image processing on an independent workstation using Osiris (University of Geneva), as previously described.23 The coefficients of variation (CV) for cartilage volume measures were 3.4% for medial tibial, and 2.0% for lateral tibial cartilage.25 (See supplementary file 2 for further details.)
Bone area measurement
Medial and lateral tibial plateau cross-sectional areas were directly measured from images using Osiris (University of Geneva) as described. The CVs for bone size measures were 2.3% for medial, 2.4% for lateral tibial plateau area.
Measurement of biomarkers
At the time of the first MRI, blood samples were obtained by direct venepuncture and centrifuged (1000 g at 4°C for 10 min) within 30 min of blood sampling. Serum was split into aliquots and frozen upright at −80°C.
All samples were thawed at 4°C, aliquoted, relabelled and refrozen. Before assaying, samples were defrosted at room temperature for 4 h; previously analyte stability had been assessed over 24 h and found to be acceptable at –80°C, 4°C and room temperature.
All ELISAs were run using duplicate wells for each sample, and the mean biomarker level was reported in each case.
Cartilage formation was assessed by measurement of PIIANP (Synarc, Lyon, France). Cartilage degradation was assessed by measurement of serum levels of C2C (C2C ELISA, IBEX) and serum levels of intact or fragmented COMP (Wielisa Human COMP, Wieslab, Lund, Sweden) were assessed. The assays for COMP and C2C were run at Alderley Park, UK. The assay for PIIANP was run by the service provider Synarc.
Biomarker quality assessment
A quality assessment of the kits was completed before assessment of the samples to ensure robustness of the data. This was undertaken using serum from eight male/female healthy volunteers/donors aged 30–60 years and included an assessment of data drift across the plate, lower and upper limit of measurement, linearity of the response and spiking accuracy (recovery values fall within a range of 70–130%). All intra-and inter- assay CVs were within acceptable limits (<25%). The intra- and interassay CVs for each of the biomarkers were; COMP (5.4%, 5.1%), C2C (3.6%, 16.2%) and PIIANP (<10%),6 respectively.
All kits met approval when healthy volunteer samples were analysed. All monitored participant's sample data points fell within the low to middle range of the standard curves and none of the samples approached the upper limit of measurement for any of the analytes. The lower limit of measurement was also recorded for each assay and data below this limit were set to the lower limit of measurement for the analysis.
Ratios of biomarker of cartilage synthesis to cartilage degradation
Ratios of biomarker of cartilage synthesis to cartilage degradation were obtained by dividing the biomarker of cartilage synthesis (PIIANP) by the biomarker of cartilage degradation (C2C) and the log ratios were used in the analysis.
Unpaired two-sample t tests were used to compare means in baseline characteristics between completers and those lost to follow up, the Mann–Whitney U test for comparison of medians and the χ2 test for comparison of proportions. The Kolmogorov–Smirnov statistic was used to test whether the data were normally distributed (a p value <0.05 was considered significantly different from a normal distribution). PIIANP displayed a skewed distribution; this information was log transformed, which resulted in the data approximating the normal distribution. The log transformed data was subsequently used.
Annual percentage change in cartilage volume was computed as 100 × (initial volume − follow-up volume)/(initial volume × time between scans). The distribution of this outcome variable approximated the normal distribution.
Linear and multiple linear regression models were used to examine the rate of cartilage volume loss and biomarker levels. In the results, a positive regression coefficient indicates that for every unit increase in biomarker, there was an associated increased rate of cartilage volume loss. A negative regression coefficient indicates that for every unit increase in biomarker, there was an associated reduction in the rate of cartilage volume loss.
Logistic regression was used to examine the relationship between baseline cartilage biomarkers and risk of knee joint replacement over 4 years. A p value <0.05 (two-tailed) was regarded as statistically significant. All analyses were performed using the SPSS statistical package (standard version 14.0, SPSS, Chicago, Illinois, USA).
Of the original 132 participants, 117 (89%) completed follow-up. Differences in the baseline characteristics are shown in table 1. There were no significant differences between subjects with 4-year data in joint replacement (n=114) and those who were lost to follow-up (data not shown).
The average annual rate of medial and lateral cartilage volume loss was 3.7±4.7% (mean±SD; 95% CI 2.7% to 4.8%) and 4.4±4.7% (mean±SD; 95% CI 3.4% to 5.5%) of initial cartilage. The difference in the rate of loss was not statistically significant (p=0.50).
Cross-sectional relationship baseline cartilage biomarkers with baseline knee cartilage volume
COMP and PIIANP were significantly associated with reduced medial cartilage volume after adjusting for potential confounders (table 2). C2C was associated with increased lateral cartilage volume. PIIANP:C2C was not significantly associated with cartilage volume (data not shown).
Relationship between baseline cartilage biomarkers and change in knee cartilage volume over 2 years
We initially examined the relationship between cartilage biomarkers and rate of cartilage volume loss in the whole population. This relationship was not linear with a change in gradient at approximately the mean for each of the biomarkers. As the data for these biomarkers were normally distributed and the mean and median values were similar, we used the mean to divide the populations into those with high biomarker levels (greater than or equal to the mean) and those with low biomarker levels (less than the mean).
The relationship between baseline cartilage biomarkers and rate of cartilage volume loss in subgroups with high and low biomarker levels are presented in table 3. The independent samples z test was used to compare the rate of cartilage volume loss between high and low cartilage biomarker subgroups.
The relationship between COMP and cartilage volume loss was not linear across the whole population, but rather there appeared to be two subgroups. In the low (less than mean) COMP subgroup, for every unit increase in COMP, there was an associated reduced rate of medial cartilage volume loss. However, there was no significant association between COMP and loss of cartilage volume in the high (greater than or equal to the mean) COMP subgroup. The difference in rate of cartilage volume loss between high and low COMP subgroups was statistically significant (p=0.05) (figure 1A).
Similarly, the relationship between PIIANP and loss of cartilage volume was not linear across the whole population. In the low (less than the mean) PIIANP subgroup, for every unit increase in PIIANP, there was an associated reduced rate of medial cartilage volume loss. However, there was no significant association between PIIANP and loss of cartilage volume in the high (greater than or equal to the mean) PIIANP subgroup. The difference in rate of cartilage volume loss between high and low PIIANP subgroups was statistically significant (p=0.003) (figure 1B). There was only slight agreement between low COMP and low PIIANP subgroups, and this result was not statistically significant (κ value 0.08, p=0.34). There was no significant association between C2C and medial cartilage volume loss, and there was no significant difference in rate of loss between high and low C2C subgroups (p=0.18) (figure 1C). Similar results were obtained when the subject receiving bisphosphonate treatment, and the subjects receiving glucosamine were excluded from the analysis (data not shown).
No significant associations were found between cartilage biomarkers and rate of cartilage volume loss in the lateral compartment, and in the rate of loss between cartilage biomarker subgroups (data not shown).
Relationship between baseline cartilage biomarkers with risk of total knee joint replacement over 4 years
Eighteen subjects underwent joint replacement over the study period. Using logistic regression there was a trend for baseline PIIANP to be associated with reduced risk of total knee joint replacement over 4 years in the whole population, which was significant after adjusting for potential confounders (table 4). Although the number of joint replacements over the study period was modest, this relationship seemed to be stronger in the low (p=0.02) than in the high (p=0.94) PIIANP subgroup (data not shown). When the ratio was considered, a similar association was obtained (data not shown). However, this relationship was attributable to the strong relationship between PIIANP and the ratio PIIANP:C2C (r=0.92, p<0.001).
In this well-characterised population with knee OA, we found that the relationship between serum biomarkers COMP and PIIANP and cartilage volume loss was not linear across the population. In the low (less than the mean) but not high (greater than or equal to the mean) subgroups, COMP and PIIANP were associated with reduced medial cartilage volume loss. PIIANP was also associated with a significantly reduced risk of knee joint replacement. These results suggest that it is possible to use cartilage biomarkers to identify subgroups within an OA population who progress at different rates.
Previous cross-sectional studies found that biomarkers are associated with the presence and severity of OA. COMP has been shown to be significantly higher in subjects with symptomatic, radiographic knee OA than in controls,18 19 27 increased in the early stages of disease,28 but decreased in advanced disease.17 29 Our cross-sectional data support these observations as increased COMP was associated with less medial cartilage volume, and is consistent with previous work showing that PIIANP levels are reduced in subjects with knee OA.2 6
Previous longitudinal studies examining the relationship between COMP and progression of OA have provided conflicting results. Higher baseline COMP predicted increased progression over 5 years in 115 patients and over 3 years in 48 patients.20 21 These studies used heterogeneous summed outcomes to assess disease progression. In contrast, studies of 57 subjects with clinical knee OA22 and 38 subjects with chronic knee pain,30 did not identify any relationship between COMP and radiographic progression over 5 years, or early radiographic changes over 3 years, respectively.
Results from studies using MRI defined outcomes are similarly inconclusive. One study5 demonstrated a detrimental effect of COMP on cartilage loss over 30 months in 137 subjects with symptomatic and clinical knee OA. Another study, showed no relationship between COMP and cartilage volume.31 The discrepancies in the results of these studies may be explained by different measures of cartilage integrity. Hunter et al5 used semiquantitative scoring of cartilage loss. It is also possible that previous studies may have been underpowered to detect a relationship.22 30
Few studies have examined the relationship between PIIANP and progression of knee OA. A previous longitudinal study demonstrated that subjects with lower levels of PIIANP at baseline had higher risk of progression over 1 year as assessed by radiography.2 Another study which compared OA progressors, who had either a reduction in joint space width or knee joint replacement over a 5-year period, with non-progressors did not identify any significant differences in PIIANP between the two groups.3 The authors suggested that the study may have been underpowered, with few progressors. Sharif et al3 did show that among progressors, mean PIIANP levels increased significantly over the 5-year period. Taken together, these results suggest that PIIANP reflects cartilage repair, which is consistent with the results of our study demonstrating that PIIANP was associated with reduced cartilage volume loss and reduced risk of knee joint replacement.
This study has shown that cartilage biomarkers may be used to identify different subgroups within an OA population and highlights that the relationship between these biomarkers and cartilage volume loss is non-linear across the whole population. In those with low (less than the mean) levels of COMP and PIIANP, we observed the expected relationships with reduced cartilage volume loss.32,–,34 However, in those with high (greater than or equal to the mean) levels of COMP and PIIANP we found no significant relationship with cartilage volume loss. We found that there was a statistically significant difference in the rate of cartilage volume loss between those in the high PIIANP and COMP and low PIIANP and COMP subgroups, respectively, supporting the notion that these subgroups are behaving differently. The non-linear relationship we observed may be explained by a ceiling effect, whereby these cartilage biomarkers contribute more at lower levels—and higher levels do not increase the effect. Ceiling effects have been documented previously in varied conditions.35,–,37
Consistent with previous studies, we did not identify a significant relationship between C2C and OA progression.4 5 The isolated cross-sectional relationship between C2C and cartilage volume in the lateral compartment is interesting but further work will be required to clarify whether C2C has the potential to predict disease progression in an OA population. Previous studies have suggested that considering an imbalance between cartilage synthesis and degradation may be useful for predicting progression.2 4 5 11 In our study, the ratio was not useful, as the results were driven by the strong linear relationship between PIIANP and the ratio PIIANP:C2C.
Our study has a number of potential limitations. We were unable to account for multiple joint involvement. Nevertheless, we were able to identify biologically plausible and consistent relationships between cartilage biomarkers and two independent outcome measures. The significant effects obtained in this study were mostly found in the medial rather than the lateral compartment, providing biological plausibility as higher axial loads pass through the medial compartment, which affects cartilage metabolism.38,–,40 We did not evaluate all known biomarkers of cartilage metabolism, and study of these may provide alternative hypotheses. We did not adjust for multiple testing as this study was viewed as hypothesis generating. Further work will be required to determine distinct cut-off points for progression, and larger studies may be required to examine a combination of biomarkers in further subgroups. Further studies will also be required to examine the relationship between change in cartilage biomarkers over time and the risk of knee joint replacement.
In this study, COMP and PIIANP were associated with a reduced rate of cartilage volume loss in the low, but not high, marker subgroups, respectively. PIIANP was also associated with a reduced risk of knee joint replacement. The results of this study suggest that cartilage biomarkers may be used to identify subgroups within a knee OA population with different rates of progression of disease. This may facilitate our understanding of the pathogenesis of disease and highlights the heterogeneity of knee OA.
The authors would like to acknowledge the following people: at AstraZeneca who performed the biomarker assays (Jacqueline Caddy, Sharon Crosby, Joanne Wayne) and for supporting statistical analysis (Euan Macpherson); at DEPM, Judy Hankin who recruited study participants. The authors would especially like to thank the study participants who made this study possible.
Funding This work was supported by NHMRC project grant 980914. PAB is the recipient of an Australian postgraduate association scholarship. AEW is the recipient of an NHMRC career development award (545876).
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Alfred and Caulfield Hospitals in Melbourne, Australia.
Provenance and peer review Not commissioned; externally peer reviewed.