Using radiography to assess the efficacy of a disease-modifying osteoarthritis (OA) drug on joint structure is challenging. Subchondral bone marrow abnormalities determined by magnetic resonance imaging (MRI) and urinary excretion of C-terminal crosslinking telopeptide of type II collagen (CTX-II) have recently been shown to be predictors of radiographic progression in patients with knee OA, suggesting that these may represent valuable biomarkers with increased sensitivity compared with findings on radiography. The aims of this investigation were to analyze, in patients with knee OA, whether the values associated with these 2 OA biomarkers can change within 3 months, and to investigate the relationships between bone marrow abnormalities and CTX-II.
Knee MRI scans were obtained in 377 patients with painful knee OA (76% women, mean age 63 years, mean disease duration 6.6 years) at both baseline and 3 months. The femoral and tibial condyles and the patella were divided into 8 sites for the scoring of bone marrow abnormalities. A bone marrow abnormality was defined as an area of increased signal on T2-weighted images of the subchondral bone. All scans were reviewed centrally and scored by a single trained radiologist using a validated 4-point scoring method. Fasting urine and serum samples were also collected from all patients at baseline, month 1, month 2, and month 3, in order to measure the levels of urinary CTX-II and serum CTX-I, a biochemical marker of bone resorption.
At baseline, 82% of patients had MRI evidence of bone marrow abnormalities. Bone marrow abnormality scores correlated significantly with CTX-II levels (P < 0.0001). Within 3 months, the bone marrow abnormality score decreased in 37 patients (9.8%), increased in 71 patients (18.8%), and did not change in the majority of patients (71.4%). Patients with baseline urinary CTX-II levels in the highest tertile had a relative risk of 2.4 (95% confidence interval 1.1–5.0) of worsening bone marrow abnormalities at 3 months compared with patients with levels in the lowest tertile, after adjustment for age, sex, and body mass index. In patients who showed a decrease in the bone marrow abnormality score at 3 months, urinary CTX-II levels decreased significantly (mean −75 ng/mmole creatinine), whereas levels increased (mean +23 ng/mmole creatinine) in patients showing an increase in the bone marrow abnormality score (P = 0.01 between the 2 groups). No significant association between bone marrow abnormalities and serum CTX-I was observed.
In patients with painful knee OA, bone marrow abnormalities on MRI can change within only 3 months in ∼30% of patients. Reduction in the extent of bone marrow abnormalities is associated with a decrease in cartilage degradation.
Osteoarthritis (OA) of the knee frequently affects the elderly population and is characterized by cartilage loss, osteophytes, and subchondral sclerosis and cysts (1). Radiography is currently the only method accepted by regulatory agencies for assessing progression of OA in phase III clinical trials. However, radiographs have several limitations. They provide a direct measure of bone alterations but only an indirect index of cartilage loss via measurement of the distance between opposing articular cortices, described as the joint space width. By the time OA is demonstrable radiographically, often significant joint damage is already present, with or without accompanying pain. Finally, because changes in joint space width are small compared with the precision error of radiography, at least 1–2 years are usually required to assess accurately the progression of joint damage or the reduction in damage with treatment. In order to develop effective disease-modifying OA drugs, new diagnostic techniques with improved sensitivity compared with that of radiography are urgently needed, particularly in the proof-of-concept stage of drug development.
Two alternative techniques have recently emerged as promising tools for assessing progression of joint damage in OA. These include magnetic resonance imaging (MRI) (2) and molecular markers of joint tissue turnover (3–6). Several recent studies have shown that MRI provides accurate and precise evaluation of the key joint structures, including articular cartilage, osteophytes, bone marrow, synovium, ligaments, and menisci, although most of these studies focused on cartilage assessment (7–13). Although cartilage loss is a cardinal feature of OA, abnormalities in subchondral bone have also been shown to provide useful information. Studies using bone scintigraphy, for example, have demonstrated that increased radiotracer uptake is associated with pain and an increased risk of progression of radiographic knee OA (14).
Subchondral bone abnormalities have also been demonstrated on MR images of OA joints (15–17). This MRI feature presents as areas of high signal intensity with ill-defined, feathery margins in the subarticular marrow space on fat-suppressed T2-weighted or intermediate-weighted spin-echo images or on STIR images. Because signal from fat is suppressed on each of these pulse sequences, the entire signal in the image is attributable to water (extracellular and/or intracellular). The appearance is identical to that seen in trauma, osteomyelitis, acute avascular necrosis, and transient osteoporosis, or in the bone marrow edema syndrome. Accordingly, this feature is commonly referred to as bone marrow edema. Histologic findings in bone specimens obtained from patients undergoing total knee replacement surgery include areas containing trabecular necrosis and occasional fibrosis but not overt edema (18). As a result, the term “bone marrow edema” has been contested by some investigators. MRI is more sensitive for the detection of water than are the methods used in those studies. However, because of these controversies, in the current report we will refer to this MRI feature as “bone marrow abnormality.”
The exact cause of bone marrow abnormalities in OA is not known. Possibilities include pulsion of joint fluid into the marrow space through defects in the articular surface, inflammation in reaction to cartilage breakdown products or other factors in intruded joint fluid, or microtraumatic changes associated with mechanical overloading. The association of bone marrow abnormalities with joint malalignment and local cartilage loss (16) supports the latter possibility, but definitive evidence is still lacking. Recently, it has been shown that bone marrow abnormality in patients with knee OA is associated with pain (15) and is predictive of progression of joint damage, as evidenced by articular cartilage loss assessed by radiography (16) or MRI (17).
A variety of molecular markers for joint tissue turnover have been described, and these markers differ in terms of their tissue specificity and sensitivity in detecting alterations in OA (3–6). Among these, urinary excretion of specific fragments of C-terminal crosslinking telopeptide of type II collagen (CTX-II) has been shown to be associated with progression of joint damage in OA (19–21). However, no study has yet investigated the relationship between these imaging and molecular markers in OA, particularly longitudinally.
The aim of this investigation was to examine the relationship between bone marrow abnormalities and urinary CTX-II levels in a large cohort of patients with knee OA participating in a randomized proof-of-concept clinical trial.
PATIENTS AND METHODS
Patients with knee OA.
Five hundred thirty-seven patients (76% women, mean ± SD age 62.5 ± 8.1 years, mean ± SD disease duration 6.6 ± 6.0 years) were enrolled in a randomized, placebo-controlled, multicenter, multinational, double-blind clinical trial of an investigational drug involving 31 sites in 9 countries (Czech Republic, Finland, Germany, Lithuania, Norway, Poland, Slovak Republic, Slovenia, and Ukraine). Inclusion criteria were as follows: adult (male or female) older than age 50 years; presence of clinically confirmed knee OA for at least 3 months prior to enrollment; pain in or around the knee most of the time, with less than 30 minutes of morning stiffness and the presence of crepitus on active motion; radiographically confirmed osteophytes of the knee (grade 2 or 3 according to the Kellgren/Lawrence scale); and a total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (22) score of ≥50 mm on a 100-mm visual analog scale (VAS).
Patients were excluded from the study if they had a history of an acute inflammatory joint disease, an acute major trauma, treatment with any investigational drug within the 3 months before enrollment, or treatment with alkylating agents, growth factors, biologic agents, or immunosuppressants; intake of any OA-directed medication within 1 week prior to inclusion or intake of oral steroids (>10 mg/day) within 4 weeks before the start of the study; treatment with intraarticular corticosteroids or hyaluronic acid injections within the previous 2 months; advanced disease as defined by radiographic grade 3 joint space narrowing, using visual scoring according to the Osteoarthritis Research Society International atlas (23); or joint replacement surgery planned within the next 6 months. If both of the patient's knees fulfilled the inclusion criteria, the worst joint was considered as the signal knee and was followed up throughout the study. If both knees were equally affected, the right knee was chosen as the signal knee. All patients had normal renal function as documented by normal creatinine clearance. In this analysis, we studied the 377 patients for whom MR images and urine samples were available at both baseline and 3 months. As shown in Table 1, the baseline characteristics of these patients were very similar to those of the total study population.
Table 1. Baseline characteristics of patients with knee osteoarthritis*
Study population (n = 377)
Whole cohort (n = 537)
Except where indicated otherwise, values are the mean ± SD. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; VAS = visual analog scale; BMA = bone marrow abnormality; CTX-II = C-terminal crosslinking telopeptide of type II collagen.
62.5 ± 8.1
62.0 ± 8.1
Body mass index, kg/m2
30.2 ± 4.8
30.3 ± 4.8
Sex, % women
Disease duration, median (25th, 75th percentiles) months
57 (19, 47)
61 (28, 114)
WOMAC score (100-mm VAS)
69.0 ± 10.2
69.6 ± 10.7
65.5 ± 13.0
66.4 ± 13.5
71.1 ± 13.8
71.2 ± 14.7
69.8 ± 10.5
70.3 ± 10.9
BMA score, median (25th, 75th percentiles)
4 (1, 7)
Medial tibiofemoral compartment
1 (0, 5)
Urinary CTX-II, median (25th, 75th percentiles) ng/mmole creatinine
432 (299, 617)
Serum CTX-I, median (25th, 75th percentiles) pg/ml
347 (245, 519)
MRI of the knee.
Each subject underwent MRI of the signal knee (as defined above) at baseline and at 3 months. Several different MRI systems were used to acquire the images, ranging from a 0.2T extremity scanner to a 1.5T whole-body scanner. Every protocol included either a fat-suppressed, T2-weighted pulse sequence or a STIR sequence in the sagittal plane to assess bone marrow abnormalities. All images were transferred to a Unix Workstation (Hewlett Packard, Palo Alto, CA) and were scored for bone marrow abnormalities on a scale of 0 to 3, according to the previously described whole-organ MRI score (WORMS) system (24), by a single radiologist with extensive experience in using WORMS and who was blinded to the chronology of the MR images. Scores were entered directly into an electronic database. In a prior study (25), the intraclass correlation coefficient was 0.93 for bone marrow abnormality scoring by the radiologist who evaluated the images in this study.
Urinary CTX-II and serum CTX-I measurements.
Fasting second-void morning urine samples and fasting serum specimens were collected at baseline, month 1, month 2, and month 3, and were stored at a temperature below −70°C until analyzed. Urinary CTX-II was measured by a competitive enzyme-linked immunosorbent assay based on a mouse monoclonal antibody raised against the EKGPDP linear 6–amino acid epitope of the type II collagen C-telopeptide (CartiLaps; Nordic Bioscience, Herlev, Denmark) (26). This sequence is found exclusively in type II collagen and not in the other collagens, including type I, or other structural proteins. Intraassay and interassay coefficients of variation were <8% and <15%, respectively. Urinary CTX-II measurements were corrected for urinary creatinine as measured by a standard colorimetric assay.
Serum CTX-I, a specific biochemical marker of bone resorption, was measured on the Elecsys 2010 automated analyzer (Roche Diagnostics, Mannheim, Germany) using β-CrossLaps serum reagents (Roche Diagnostics). This assay is specific for crosslinked β isomerized type I collagen C-telopeptide fragments and uses 2 monoclonal antibodies, each of which recognizes the Glu-Lys-Ala-His-Asp-Gly-Gly-Arg peptide. The intraassay and interassay coefficients of variation are <6%, and the sensitivity is 10 pg/ml (27). All measurements were performed in duplicate in a centralized specialty laboratory (Molecular Marker Services, Synarc, Lyon, France).
Unless specified otherwise, all data are expressed as the median (25%, 75% percentile), because both bone marrow abnormalities on MRI and biochemical markers were not normally distributed. The significance of changes in bone marrow abnormality scores between baseline and 3 months was assessed using the nonparametric Wilcoxon signed rank test for paired data. Changes in urinary CTX-II and serum CTX-I levels were calculated from the individual regression slopes of values at baseline, month 1, month 2, and month 3 over time. Relationships between bone marrow abnormalities, the urinary CTX-II level, the serum CTX-I level, and their respective changes from baseline were assessed by Spearman's rank correlation coefficients. The relationship between baseline urinary CTX-II levels and the risk of worsening bone marrow abnormalities was assessed by logistic regression analyses after adjustment for sex, age, and body mass index (BMI). All statistical analyses were performed on SAS version 10.0 for Windows (Chicago, IL).
Three hundred seventy-seven patients for whom MR images and biologic samples were available at baseline and followup were included in this study. Eighty-two percent of patients had MRI evidence of bone marrow abnormalities (score ≥1) in the knee, and 70% of these patients had bone marrow abnormalities in the medial tibiofemoral compartment. At baseline, urinary CTX-II levels (but not serum CTX-I levels) in this population, which comprised mainly postmenopausal women, were, on average, 106% higher than those reported recently in untreated postmenopausal women with no knee OA (27, 28). The bone marrow abnormality score, urinary CTX-II levels, and serum CTX-I levels did not significantly correlate with the WOMAC total score or subscores, except for a slight association of the bone marrow abnormality score for the medial tibiofemoral compartment with the WOMAC score for stiffness (r = 0.12, P = 0.026).
At baseline, urinary CTX-II levels correlated significantly with the bone marrow abnormality score (r = 0.29 and r = 0.23 [P < 0.0001] for the total score and the medial tibiofemoral compartment score, respectively). Patients with a total bone marrow abnormality score of >8 had, on average, 45% higher CTX-II levels than did patients with a bone marrow abnormality score of 0. Similarly, in patients with CTX-II levels in the highest tertile, the median total score for bone marrow abnormalities was 5, compared with a score of 2 in patients with CTX-II levels in the lowest tertile (Figure 1).
Three months after the start of this longitudinal study, there was a significant change in the total bone marrow abnormality score (P = 0.0015), although the median change was equal to 0 (from −10 to +7) in the whole population. As shown in Figure 2, 37 patients (9.8%) demonstrated a decrease in the total bone marrow abnormality score, whereas 71 patients (18.8%) showed an increase; the majority of patients (71.4%) showed no change in the bone marrow abnormality score. Only 2 of the 37 patients who demonstrated a decrease in the total bone marrow abnormality score showed complete normalization (i.e., a score of 0). As shown in Table 2, patients who demonstrated a decrease in the total score for bone marrow abnormalities did not differ in terms of demographics, WOMAC scores, and the total bone marrow abnormality score at baseline from those who showed worsening of the bone marrow abnormality score.
Table 2. Baseline characteristics of patients with knee osteoarthritis who demonstrated a decrease or an increase in the BMA score over 3 months*
Subjects whose total BMA score decreased (n = 37)
Subjects whose total BMA score increased (n = 71)
There was no significant difference (P > 0.30) between groups for any parameter. BMA = bone marrow abnormality; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; VAS = visual analog scale.
Age, mean ± SD years
64.8 ± 8.1
64.0 ± 6.3
Body mass index, mean ± SD kg/m2
30.3 ± 4.5
31.5 ± 4.3
Sex, % women
Disease duration, median (25th, 75th percentiles) months
65 (25, 117)
57 (31, 106)
WOMAC score, median (25th, 75th percentiles) on 100-mm VAS
70.3 (62.4, 77.9)
69.7 (62.3, 75.0)
69.2 (59.8, 76.3)
67.0 (57.8, 75.6)
75.2 (70.3, 82.3)
74.5 (64.0, 81.3)
70.9 (68.0, 78.8)
70.1 (61.8, 76.1)
Total BMA score, median (25th, 75th percentiles)
6.0 (4.0, 8.5)
5.0 (2.3, 7.8)
We then analyzed the relationships between both urinary CTX-II and serum CTX-I levels and the respective changes in the total bone marrow abnormality score. As shown in Table 3, after adjustment for sex, age, and BMI, there was a significant association between increased urinary CTX-II levels at baseline and the risk of a worsening total bone marrow abnormality score at 3 months. There was also a significant positive correlation between changes in the urinary CTX-II level at 3 months and changes in the bone marrow abnormality score at 3 months, with patients showing a decrease in the total bone marrow abnormality score also showing a significant decrease in the urinary CTX-II level, and patients showing an increase in the bone marrow abnormality score showing a sustained increase in CTX-II values (Figure 3). In contrast, we observed no significant association between either the baseline or change in the serum CTX-I level and baseline or change in bone marrow abnormality score (data not shown).
Table 3. Increased baseline levels of urinary CTX-II and risk of worsening total BMA score*
Baseline CTX-II level
Worsening of BMA at 3 months
% of patients
OR (95% CI)
The odds ratio (OR) was adjusted for sex, age, and body mass index. CTX-II = C-terminal crosslinking telopeptide of type II collagen; BMA = bone marrow abnormality; 95% CI = 95% confidence interval.
In this large population of well-characterized patients with painful OA of the knee, we showed that bone marrow abnormalities (as measured by MRI) and urinary CTX-II, which represent 2 candidate biomarkers of disease progression, are highly prevalent, can change significantly within 3 months, and are positively interrelated.
Eighty-two percent of patients in this cohort had evidence of bone marrow abnormalities when all knee joint surfaces were considered. This value is similar to that reported by Felson et al (15), who observed that 77.5% of 351 patients with painful knee OA (of similar age to the current cohort) had bone marrow abnormalities, and similar to that reported more recently by Sowers et al (29) in a smaller cohort of patients with knee OA and pain, 73% of whom had bone marrow abnormalities. Using the same MRI sequence with a 1.5T scanner and the same WORMS method of scoring abnormalities, Carbone et al recently observed that the prevalence of bone marrow abnormalities was 60% among postmenopausal women in the Health, Aging and Body Composition Study, but that only 37% of these women had evidence of radiographic knee OA (25). Thus, both our study and others indicate that subchondral bone marrow abnormalities are highly prevalent in the knees of patients with OA, especially in those presenting with knee pain. We did not, however, find any significant association between the bone marrow abnormality score and the severity of pain as assessed by the WOMAC score. To be included in our study, all patients had to have pain and a total WOMAC score of ≥50 mm. This high and narrow range of pain scores may explain the lack of significant relationships between bone marrow abnormalities and pain observed in our study. Felson et al (15) also reported no association between bone marrow abnormalities and the severity of pain as assessed by the WOMAC index when they restricted their analyses to subjects with knee pain.
In this study, we observed increased urinary CTX-II levels in patients with knee OA. This observation was consistent with previous reports (19–21, 26, 28). Because urinary CTX-II levels reflect the degradation of cartilage type II collagen in all body joints, including the hands and spine (28), which are frequently affected by degenerative diseases in the elderly, it is likely that the increased CTX-II levels observed in this study do not reflect alterations in only the signal knee.
To our knowledge, this is the first report of short-term changes in bone marrow abnormalities in patients with knee OA. We observed that a total of 29% of patients with bone marrow abnormalities at baseline showed a change in the bone marrow abnormality score over the 3-month study period. The bone marrow abnormality score improved in 9.8% of these patients, although only 0.5% of patients experienced complete resolution, and in 18.8% of patients the bone marrow abnormality score increased. In a smaller group of 70 patients with symptomatic knee OA, using the same WORMS bone marrow abnormality scoring system, Hunter et al (17) observed that over an 18-month period, 5% of patients had a decrease in the bone marrow abnormality score, whereas 40% showed an increase in the bone marrow abnormality score. Moreover, the severity of the bone marrow abnormality score at baseline in these patients, as well as the change in the bone marrow abnormality score over the course of the study, correlated with the progression of articular cartilage loss in these patients, based on the WORMS cartilage score.
In a recent cross-sectional study of 540 women ranging in age from 69 years to 81 years, Carbone et al (25) reported a lower bone marrow abnormality score, based on the WORMS method, in subjects treated with estrogen replacement or bisphosphonates (residronate, alendronate) than in those not receiving antiresorptive bone therapy, even after adjusting for age, race, study site, BMI, nonsteroidal antiinflammatory drugs, thiazides, calcium supplements, hip bone mineral density, smoking, knee extensor strength, and the use of walking devices. The limitations of a cross-sectional study design notwithstanding, these results raise the possibility that bone marrow abnormalities can respond to therapy. This ability to predict articular cartilage loss in OA, coupled with a capacity for very rapid change and even the potential for a therapeutic response, makes bone marrow abnormality an extremely intriguing MRI marker for proof-of-concept studies of disease-modifying OA drugs.
It might be argued that the observed changes in the bone marrow abnormality score may be, in part, attributable to the phenomenon of regression to the mean, which is based on the greater probability that the scores in knees with a low score at baseline will increase rather than decrease over time, and conversely that scores in knees with a high score at baseline will decrease rather than increase over time. This explanation, however, appears unlikely, because the baseline median score was similar in patients showing an increase in the bone marrow abnormality score and those showing a decrease in the bone marrow abnormality score.
Additional OA biomarkers are molecular markers of bone and cartilage degradation. In this study, we analyzed serum CTX-I as a specific biochemical marker of bone resorption and urinary CTX-II as an index of cartilage degradation. We found that both baseline levels and changes in urinary CTX-II levels (but not serum CTX-I levels) were related to a higher baseline bone marrow abnormality score and/or an increased risk of worsening bone marrow abnormalities over 3 months, associating for the first time these 2 early predictors of long-term cartilage loss in OA. Interestingly, none of the other risk factors for disease progression, including age, female sex, BMI, or severity of pain, could predict which patients experienced increased or decreased bone marrow abnormalities.
It may appear surprising that bone marrow abnormalities were associated with CTX-II and not with serum CTX-I, a systemic biochemical marker of bone resorption. Because the levels of bone markers in serum or urine mainly reflect the overall rate of skeletal turnover, which can be affected by different conditions independent of OA, they are likely to lack sensitivity to detect focal abnormalities of subchondral bone metabolism, a factor that may explain the inconsistent and somewhat disappointing data generated between markers of bone turnover and joint damage in OA (3, 4). In contrast, several studies have shown a consistent association between increased baseline levels of urinary CTX-II and more rapid progression of radiographic knee or hip OA over the subsequent 1–6 years (19–21). Additionally, we previously showed that decreased urinary CTX-II levels over 3 months in patients with early rheumatoid arthritis who were treated with an active combination of prednisolone, methotrexate, and sulfasalazine were associated with slower radiographic progression over the subsequent 5 years (30), suggesting that early changes in this marker may be related to long-term progression, although this hypothesis needs to be confirmed in OA.
Limitations of this study include the lack of assessment of angular knee malalignment; in other studies (16, 17) knee malalignment has been shown to explain at least part of the association between bone marrow abnormalities and cartilage loss in knee OA. We did not assess cartilage damage by MRI and thus could not confirm the relationship between this feature of OA and urinary CTX-II levels and bone marrow abnormalities. However, the duration of the study was probably too short to permit significant changes in the morphologic parameters of cartilage to be demonstrated on MRI. Also, the status of other potential sources of CTX-II in the body, such as the hips, shoulders, and other diarthroidial joints, was not evaluated in this study and could have confounded the results.
In conclusion, this longitudinal study demonstrates that both bone marrow abnormalities and elevated urinary CTX-II levels are highly prevalent in patients with knee OA and can change in a significant proportion of these patients within only 3 months. Moreover, changes in these 2 early predictors of rapid disease progression in patients with knee OA are significantly correlated, suggesting construct validity of their use in clinical trials. Further evaluation of the utility of these 2 biomarkers for clinical studies of disease-modifying OA drugs is recommended.