Osteoarthritis (OA) is a heterogeneous, complex joint disorder of uncertain etiology. It is the most common joint disease, affecting ∼10% of the population over the age of 50 years, and epidemiologic studies suggest that the knee is the most important and most common site of OA, causing pain and disability (1, 2). Over the last two decades, there has been considerable research interest in joint tissue metabolism both in health and in OA. Animal studies have shown that there are subtle biochemical changes in articular cartilage and other joint tissues well before any clinical or radiologic evidence of joint destruction (3, 4). Our own studies in OA patients have shown that synovial fluid concentrations of structural proteins of bone, such as osteocalcin, are associated with bone remodeling in OA (5) and that increased serum hyaluronan (a possible marker of synovial inflammation) predicts loss of cartilage in patients with OA of the knee joint (6). Moreover, a rise in serum concentrations of cartilage oligomeric matrix protein (COMP) during the first year of followup was also found to be associated with disease progression in knee OA (7).
COMP is an important cartilage matrix macromolecule and is the largest matrix protein in articular cartilage after collagen and proteoglycan. COMP is a pentameric molecule with a molecular weight of ∼500 kd and belongs to the thrombospondin family (8). Initially, it was thought to be cartilage-specific, but recent studies have shown the presence of COMP in various other joint tissues, including menisci, ligaments (9), tendons (10), and the synovium (11). However, the concentrations of COMP in these tissues are very low compared with those in cartilage (9). The physiologic role of COMP is still unknown, although COMP has been shown to interact with chondrocytes (12) and is thought to be involved in endochondral ossification (13). More recently, divalent cation–dependent interactions between COMP and types I, II, and IX collagen were demonstrated, suggesting that such interactions increase the stability of the collagen network (14, 15). Mutation of the COMP gene leads to various types of dysplasia and premature development of OA (15, 16). Thus, COMP is an important cartilage matrix macromolecule, and its metabolism could have profound effects on the structure and integrity of cartilage.
COMP had been investigated widely as a biomarker for OA and other joint diseases. However, of the many reports of serum COMP concentrations in OA (7, 9, 17–22), only a few have documented the changes in serum COMP levels over time (7, 23, 24). Cross-sectional studies determined the association of COMP with the presence and severity of disease, while longitudinal studies examined changes in serum COMP levels over 1 year and 3 years, respectively, in relatively small cohorts of OA patients. In 1997, we recruited 135 patients with predominantly medial tibiofemoral OA to monitor the natural progression of the disease sequentially (every 6 months) over 5 years using clinical criteria, imaging (radiography, scintigraphy, and dual x-ray absorptiometry), and serum biomarkers. The hypothesis being tested is that progression of OA is not linear and that there is a relationship between progression over time and changes in serum COMP levels over time. We report here longitudinal measurements of serum COMP in relation to disease outcome both during the course of the study and at 5 years of followup.
- Top of page
- PATIENTS AND METHODS
From the initial 135 patients, radiographs obtained at study entry and at the 5-year followup were available for 115. The demographic, clinical, and radiographic details of all the patients, including those lost to followup, are summarized in Table 1. There were no significant differences in these variables at baseline in patients who were lost to followup and those who completed the 5-year study. There was no association between serum COMP levels and the sex of the patients. The mean ± SD concentrations of COMP in men and women were 13.58 ± 2.94 units/liter and 12.67 ± 3.65 units/liter, respectively (P = 0.17). However, at baseline, the serum COMP levels were significantly higher in patients over the age of 60 years compared with those under the age of 60 years (13.58 ± 2.43 versus 12.08 ± 2.43 units/liter; P = 0.034).
Table 1. Baseline demographic, clinical, and radiographic features of the study patients*
| ||All patients (n = 135)||Patients who completed the study (n = 115)||Patients who were lost to followup (n = 20)|
|Demographic variables|| || || |
| Age, mean ± SD years||63.8 ± 9.6||63.6 ± 9.7||65.15 ± 9.2|
| No. of males/females||62/73||52/63||10/10|
| Body mass index, mean ± SD kg/m2||29.6 ± 5.5||29.6 ± 5.2||29.4 ± 7.2|
| No. of patients with Heberden's nodes||48||42||6|
| No. of patients taking NSAIDs||35||32||3|
|Clinical variables for most painful knee, mean ± SD|| || || |
| Pain score by VAS||51.5 ± 28.8||50.9 ± 28.1||56.1 ± 34.5|
| WOMAC score|| || || |
| Pain||7.5 ± 4.0||7.2 ± 3.9||9.5 ± 4.7|
| Stiffness||3.6 ± 1.8||3.5 ± 1.7||4.3 ± 2.0|
| Function||24.8 ± 12.8||23.8 ± 12.0||32.9 ± 16.8|
|Radiographic variables|| || || |
| Minimum medial compartment TF joint space width in most painful knee, mean ± SD mm||3.2 ± 1.9||3.2 ± 1.9||3.3 ± 2.0|
| Overall K/L grade in worst knee, % of patients|| || || |
| Grade 0||30||25||5|
| Grade 1||13||12||1|
| Grade 2||11||7||4|
| Grade 3||75||67||8|
| Grade 4||4||3||1|
| No. with K/L grade ≥2 in contralateral knee||56||50||6|
|Reason for loss to followup, no. of patients|| || || |
| Moved away or declined followup||–||–||14|
Of the 115 patients who completed the study, 37 experienced OA progression: 22 by the TKR criterion and 15 by the ≥2 mm of tibiofemoral joint space narrowing criterion. In addition to study radiographs, preoperative radiographs obtained before TKR surgery were available for 19 patients, and all had ≥2 mm of tibiofemoral joint space narrowing. The mean ± SD ages of the progressors and nonprogressors were 64.2 ± 7.8 years and 63.3 ± 10.6 years, respectively (P = 0.642), and the percentages of women were 19 ± 51% and 44 ± 56%, respectively (P = 0.690). Patellofemoral joint space narrowing was present at baseline in 34% of patients and was slightly less common in those who experienced OA progression in the tibiofemoral joint (26%). The mean baseline serum COMP concentration was significantly higher in the progressors compared with the nonprogressors (14.12 ± 3.39 units/liter versus 12.62 ± 3.25 units/liter; P < 0.036).
Serum COMP levels varied widely both between patients and within patients over the study period. To illustrate these variations, changes in serum COMP levels during the 5-year study period for a random subset of 25 of the 115 patients who completed the study are shown in Figure 1. At first inspection, it appears that a number of patients have abnormally high variations in COMP levels (e.g., patients A and B). Analysis of the variation in COMP measurements in individual patients revealed these to be patients who had undergone TKR.
Figure 1. Changes in serum levels of cartilage oligomeric matrix protein (COMP) over the entire study period in 25 randomly selected osteoarthritis patients. COMP levels showed wide variations over time in some patients, which was due to total knee replacement (TKR). For example, patients A and B underwent TKR surgery between months 12–18 and months 36–42, respectively.
Download figure to PowerPoint
Serum COMP concentrations in 16 of the patients who underwent TKR and had appropriately timed blood samples were therefore oriented around the time of surgery, and the standardized values during the months before and after surgery are shown in Figure 2. There was a substantial rise in serum COMP levels after surgery, persisting for up to 12 months following surgery. When data were analyzed omitting serum COMP measurements following TKR, the variations between patients approximated a normal distribution, and there were no patients with abnormally high variations in COMP levels. Accordingly, in order to avoid bias in COMP measurements that may be due to TKR, postoperative COMP levels were excluded from all subsequent analyses.
Figure 2. Standardized serum cartilage oligomeric matrix protein (COMP) levels in relation to total knee replacement (TKR) surgery in the 16 patients whose COMP levels were measured before and after surgery. Values are the mean and 95% confidence interval. N = number of patients.
Download figure to PowerPoint
The AUC for the COMP values during the entire study period (excluding the visits after surgery) in the progressors and nonprogressors are shown in Figure 3. Although there was substantial overlap between the 2 groups, the mean values differed significantly (P < 0.003), and further analysis showed that COMP levels in the progressors was higher at all visits compared with those in the nonprogressors (see Figure 4).
Figure 3. Averages of all cartilage oligomeric matrix protein (COMP) levels measured (area under the curve [AUC] per month) during the entire study period in patients whose osteoarthritis (OA) had progressed and in those whose OA had not progressed at the 5-year followup assessment. To allow for the effect of surgery on serum COMP concentrations, the AUC was calculated for each patient for the duration of the study and for patients who underwent total knee replacement surgery up to the point of the surgery. The AUC was then expressed as an average for the number of months of followup. Horizontal bars show the mean.
Download figure to PowerPoint
Figure 4. Sequential changes in mean cartilage oligomeric matrix protein (COMP) levels in patients whose osteoarthritis (OA) had progressed (n = 37) and in those whose OA had not progressed (n = 78) at the 5-year followup assessment. Differences between the nonprogressors and progressors were statistically significant for 4 of the 10 visits (∗ = P < 0.05). Values are the mean ± SEM.
Download figure to PowerPoint
The concentrations of serum COMP at study entry and the 5-year followup and that AUC values for COMP in relation to different types of progression are shown in Table 2. At baseline, the mean ± SD COMP levels in the TKR group were lower than those in the group with ≥2 mm of tibiofemoral joint space narrowing (13.68 ± 2.43 versus 14.62 ± 4.27 units/liter), but the differences between the TKR group and the nonprogressors were not significant at study entry or at the 5-year followup.
Table 2. Serum COMP concentrations at baseline, the 5-year followup, and the AUC per month in relation to the radiographic progression status of the patients at 5 years*
| ||COMP levels in nonprogressors (n = 78)||Progressors (n = 37)||≥2-mm reduction in TF joint width (n = 15)||Total knee replacement (n = 22)|
|COMP level||P||COMP level||P||COMP level||P|
|Baseline||12.62 ± 3.25||14.12 ± 3.39||<0.038||14.62 ± 4.27||0.045||13.68 ± 2.43||0.21|
|5-year followup||12.39 ± 3.58||14.21 ± 2.08||0.07||14.21 ± 2.17||0.085||13.84 ± 3.94||0.20|
|Mean of all 10 visits (AUC per month)||10.82 ± 2.71||12.52 ± 2.71||<0.003||12.71 ± 2.47||0.014||12.37 ± 2.94||0.030|
To further examine the patterns of serum COMP changes in relation to OA progression, we assigned the patients who progressed during the study period to 1 of 3 groups. Group A patients were those whose OA progressed during the first 2 years (n = 13; 10 by the TKR criterion and 3 by the ≥2 mm of tibiofemoral joint space narrowing criterion). Group B patients were those whose OA progressed between years 2 and 3 (n = 6; 3 by TKR and 3 by ≥2 mm of tibiofemoral joint space narrowing). Group C patients were those whose OA progressed between years 3 and 5 (n = 18; 9 by TKR and 9 by ≥2 mm of tibiofemoral joint space narrowing). Since the period of progression was very short in group B (1 year) and there were only 6 patients in the group, the data are shown only for the 2 larger groups.
As shown in Figure 5, the mean serum COMP levels in patients whose OA progressed over the first 2 years of study (group A) increased from month 6 to month 18 before declining. In contrast, the mean serum COMP levels in patients whose OA progressed from years 3–5 (group C) declined from month 0 to month 12 before increasing during the radiographically identified progression period (36–60 months). Group A patients also showed a rise in serum COMP levels toward the end of the study. This can be attributed to 4 of the 8 patients who initially progressed by the ≥2 mm of tibiofemoral joint space narrowing criterion, who then progressed again by the same criterion toward the end of the study.
Figure 5. Mean serum cartilage oligomeric matrix protein (COMP) levels in patients whose osteoarthritis (OA) had progressed between years 0 and 2 (group A; n = 13) and between years 3 and 5 (group C; n = 18) of the 5-year study period. In both groups, there was a rise in serum COMP levels during the period of OA progression. Four of the patients in group A whose OA had initially progressed according to the criterion of ≥2 mm of narrowing of the tibiofemoral joint space also experienced OA progression according to the same criterion during the last 24 months of followup.
Download figure to PowerPoint
Figure 6 shows the probability of progression by the 2 criteria for different levels of serum COMP and their frequency of occurrence in this cohort of OA patients. For these data, the mean of the baseline and 6-month COMP concentrations was used in order to make allowance for the variation in COMP levels. The probability of progression increased when the level of COMP increased. The binary logistic model showed that the probability of progression increased by 15% when the serum COMP concentration increased by 1 unit/liter. Thus, a patient with a COMP level (mean of 2 readings) of 20 units/liter has an ∼30% chance of progression by the TKR criterion and a further 20% chance of progression by the ≥2 mm of tibiofemoral joint space narrowing criterion over the following 4.5 years.
Figure 6. Probability of osteoarthritis progression according to total knee replacement (TKR) surgery (○) (n = 22) and according to a ≥2-mm reduction of the tibiofemoral joint space width or TKR surgery (•) (n = 37). To compare the different types of progression, an ordered logistic regression model was performed. The model consisted of 2 equations, since the probabilities modeled are cumulative over the lower-ordered category. The probability of progression was calculated as 1/[1 + e(ak – bx)], where ak is the intercept on the y-axis, b is the coefficient of the exposure variables, and x is the mean of the exposure variable (cartilage oligomeric matrix protein [COMP]) at visits 1 and 2. The chart on the right shows the frequency of occurrence of the mean baseline and 6-month COMP concentrations in the study population.
Download figure to PowerPoint
- Top of page
- PATIENTS AND METHODS
This cohort of patients with predominantly medial tibiofemoral OA was established to determine the sequential change in serum COMP levels (and other biomarkers) and its relationship to radiographic changes over 5 years. During the 5-year study period, only 20 patients (15%) were lost to followup. There were no significant differences at study entry in the demographic, clinical, or radiographic findings in these patients compared with the 2 groups of patients (progressors and nonprogressors) who completed the 5-year study. Accordingly, this is a suitable cohort of patients with which to investigate the relationship between serum COMP levels measured sequentially every 6 months and radiographic progression over 5 years.
We found that serum COMP levels increased in the period following TKR. This was unexpected but was consistent across all patients who had undergone TKR. It is not clear why this should be so, but there may be an increased production or degradation of COMP in the contralateral knee joint or an increased release of COMP from other tissues, such as ligaments, tendons, and capsule, of the TKR joint. Further investigation of this finding and exploration of other biomarkers following surgery will be an important priority. In order to avoid bias in the interpretation of other results, patients were excluded from further analyses after they underwent TKR; thus, the remaining conclusions are valid whatever the reason for this response to surgery.
Patients whose OA progressed over 5 years had higher serum COMP levels at baseline and throughout the study period compared with those whose OA did not progress. Moreover, the probability of progression increased as the serum COMP value increased, suggesting that COMP levels could be used to identify patients who are at risk of OA progression. There was a large overlap in COMP concentrations between progressors and nonprogressors (Figure 3). However, the data presented in this study suggest that increasing serum COMP concentrations means greater risk of progression of tibiofemoral knee OA (see Figure 6).
Three earlier and smaller studies (46, 38, and 48 patients, respectively) have shown that serum COMP levels are associated with disease progression in knee OA (7, 19, 24). The first two studies (7, 19) reported that a rise in serum COMP levels over 1 and 3 years was associated with progressive joint damage, but baseline levels were not associated with progression over that time period. The third study (24) found that patients who progressed by 2 Kellgren/Lawrence grades (sum of both knees) over 3 years had significantly higher serum COMP levels at baseline and at study end. The data reported in the current study are based on a larger cohort of patients (n = 115) with a well-defined intraarticular distribution of their knee OA and for whom sequential radiographs and measurements of COMP levels were available for the entire study period. Thus, it was possible to relate any temporal variation in COMP levels to the time of OA progression. If progression of knee OA is linear, it would be expected that temporal variations in COMP levels would not be related to progression. If knee OA progression is cyclic, then COMP levels should decrease after periods of progression and increase prior to progression. The results are consistent with the hypothesis that progression of knee OA is phasic.
At study entry, the group that progressed between year 3 and year 5 had high COMP levels that declined before increasing again prior to the period of progression, whereas COMP levels in the group that progressed early in the study (between year 0 and year 2) increased initially, declined toward and after the end of the period of progression, and rose again at the end of the study. It may be that one group of patients had been progressing just prior to study entry and the other was about to progress following completion of the study. However, in the absence of any progression data prior to entry or beyond 60 months, we cannot say conclusively that the data presented show cyclical progression in individual patients. Nevertheless, sequential measurements of COMP levels in blood may identify patients who are likely to experience OA progression within the next year or two. Such patients may then be treated more effectively by targeting the treatment to the time of rising serum COMP levels, and they may also provide a particularly suitable group in which to test new treatments.
COMP is an important structural molecule of the articular cartilage, and degradation of this protein is therefore likely to lead to reduced/absent interaction with chondrocytes, collagens, and other matrix proteins, resulting ultimately in loss of cartilage. It has been shown that synovial fluid from patients with inflammatory arthritis degrade COMP into smaller fragments, and these low molecular weight fragments (50–70 kd) of COMP have been reported to be 3 times higher in OA compared with healthy controls (9). The assay used in the present investigation detects low molecular weight COMP fragments, although it is not known whether these fragments are from mature degraded COMP molecules or newly formed COMP fragments. Nevertheless, since serum COMP levels were persistently high throughout the 5-year study period in patients with progressive joint damage, serum concentrations of COMP may be reflecting active degradation and loss of articular cartilage.
Findings of a recent study suggest that there may be specific proteases that regulate the degradation of COMP in cartilage (29). Thus, chondrocyte death may lead to the release of these and other enzymes that promote degradation of COMP and loss of cartilage. Osteoarthritic cartilage is hypocellular (30), and this is likely to reduce the ability of the cartilage to maintain the matrix integrity. Hypocellularity in OA may be explained by chondrocyte death by apoptosis, which is, on average, 3 times higher in OA cartilage compared with age- and sex-matched controls (31). Therefore, a high rate of apoptosis in OA cartilage may contribute to the degradation of COMP, as well as to loss of interaction between chondrocytes, COMP, and other matrix macromolecules. Interestingly, various dysplasias have been associated with mutations of the COMP gene (15, 16), and a recent study has shown that in the growth plate of patients with pseudoachondroplasia (a condition characterized by short stature, shortening of the long bones, and early-onset OA), there is an increased rate of chondrocyte death by apoptosis compared with that in control patients who had undergone surgery for conditions not associated with pseudochondroplasia (72% versus 45%), and it is associated with decreased cartilage growth (32).
COMP has been investigated widely, and it is likely that serum COMP derives from the index joint, since there is a good positive correlation between serum COMP levels and joint destruction in animal models of arthritis (33, 34) as well as between serum and synovial fluid levels in humans with OA (9, 35). Moreover, the observation that abnormal findings on scintigraphy scans of OA knee joints are positively correlated with serum COMP levels (7, 19) also suggests that serum COMP is derived from the arthritic joint. Some serum COMP is likely to be derived from nonarticular sources and/or asymptomatic joints. However, the concentrations of COMP in nonarticular tissues are very low compared with those in the cartilage and meniscus (9); therefore, articular chondrocytes are probably the main source of COMP in the joints.
In a recent study, Jordan et al (36) reported that serum COMP is associated with both age and ethnicity. In the current study, all patients were Caucasian, and serum COMP levels were not associated with sex, but some associations were found with age. These data are consistent with those of several previous reports (9, 21). The study by Clark et al (21) found that serum COMP levels are significantly higher in patients who were over the age of 65 years compared with patients who were ages 45–55 years. In our study, 56 patients were over the age of 60 years, and they had significantly higher serum COMP levels compared with those in patients younger than 60 years.
In a large population-based study using radiographically defined controls, Clark et al (21) demonstrated that serum COMP levels are, on average, higher in patients with early OA than in unaffected individuals and that they were associated with OA severity as well as with the number of joints involved. However, those authors concluded that due to overlap between the OA patients and the unaffected individuals, COMP cannot be used as a diagnostic marker. The data presented here show that serum COMP is associated with the progression status of the OA patients and that sequential measurements can identify periods when patients' COMP levels change markedly. Thus, this biomarker may help to target treatment during times when patients are most likely to have cartilage destruction. The ability to identify patients whose OA is likely to progress over the coming years would help define the molecular basis and pathogenesis of OA and lead to the development of screening methods for early diagnosis and, therefore, more effective treatment.
In conclusion, the data presented here show that serum concentrations of COMP in patients with knee OA vary widely both between and within patients over time. The concentrations of serum COMP increased markedly during the year following TKR and were higher in those over the age of 60 years, but were unrelated to sex. The patients with progressive joint damage had significantly higher serum COMP levels at study entry, and the high levels persisted over the 5-year period of study. Moreover, the serum COMP concentration was increased during the period of clear radiographic progression and loss of cartilage. Although large between-subject variation precludes the use of individual values to predict progression with confidence, the mean of the two COMP values obtained 6 months apart provided an indication of likely progression over the next few years, and this might help to target treatment in the individual patient and to select cohorts of patients who are particularly suitable for trials of new disease-modifying drugs.