Article Text
Abstract
Objective: Bone marrow lesions (BML) are important in established knee osteoarthritis, predicting pain and progression of disease. Whether BML are also associated with longitudinal changes in knee structure in an asymptomatic population is unknown.
Methods: 148 healthy pain-free women in middle age with no history of knee injury or clinical knee osteoarthritis who had a magnetic resonance imaging (MRI) scan performed on their dominant knee at baseline, had another MRI 2 years later to assess whether having a BML present at baseline affected change in tibiofemoral cartilage defects and tibial cartilage volume.
Results: BML were present in 14.9% of women at baseline. The risk of progression of total tibiofemoral cartilage defects was significantly higher when a very large BML was present (odds ratio 5.55, 95% CI 1.04 to 29.6) compared with when no BML was present, after adjusting for potential confounders. In the lateral compartment, the rate of cartilage volume loss was significantly greater when a BML was present after adjusting for confounders (regression coefficient 39.2 mm3, 95% CI 11.1 to 67.2, p = 0.007).
Conclusions: In healthy women without pain at baseline, large BML were associated with both progression of cartilage defects in the whole tibiofemoral joint and more rapid lateral tibial cartilage loss. These data suggest that the relationship between BML and knee cartilage in healthy women is similar to that described in established osteoarthritis. It is possible that BML may predict an increased risk of knee osteoarthritis and facilitate the identification of novel interventions to prevent disease.
Statistics from Altmetric.com
Osteoarthritis is a complex disease, involving all the tissues of the affected joint. Whereas the tissue of origin of osteoarthritis is controversial, subchondral bone appears important in early disease.1 Both human and animal studies have suggested that bone changes may precede cartilage damage.2 3 Once osteoarthritis is established, increased subchondral bone metabolism has been linked to disease progression.4 This relationship may be mediated by local cytokine release, as in established osteoarthritis bone explants are able to affect cartilage metabolism.5
Bone marrow lesions (BML), visible using magnetic resonance imaging (MRI), have been recognised as a feature of knee osteoarthritis.6 7 In established knee osteoarthritis, these are associated with pain and an increased likelihood of cartilage loss.6–8 They are present in approximately 10% of healthy middle-aged individuals without knee pain or a history of significant knee injury, and have been attributed to biomechanical factors or systemic factors such as diet.7 9–12 Two cross-sectional studies in different populations suggested that BML are associated with age, height and body mass index (BMI), all of which are risk factors for knee osteoarthritis.10 11 BML have also been consistently associated with an increased prevalence and severity of cartilage defects and metaphyseal expansion, both characteristics of the increasing severity of knee osteoarthritis.10 11 Although longitudinal studies in osteoarthritis suggest that BML are linked with increased cartilage loss, there are no similar studies in asymptomatic individuals. Therefore, whether BML predict structural change, which in turn may progress to knee osteoarthritis in a pain-free population, is unknown.
The aim of this study was thus to examine, in a population of women in midlife, with no knee pain and thus no clinical knee osteoarthritis, the relationship between BML and changes in knee cartilage defects and volume over 2 years.
METHODS
This was a prospective cohort study in which eligible participants aged 40–67 years were recruited to examine factors affecting knee cartilage from a longitudinal cohort study of normative hormone levels in well women.13 These women were initially recruited from a database established from the electoral roll in Victoria, Australia, between April 2002 and August 2003.13 Participants underwent a baseline MRI scan on their dominant knee between October 2003 and August 2004 and a follow-up MRI scan on the same knee approximately 2 years later.14 Women who had experienced significant knee pain (ie, pain requiring any intervention by a health professional, medication or necessitating non-weight-bearing therapy) or a knee injury in the past 5 years that necessitated treatment or required rest for more than one day, or who had a contraindication to undergoing MRI, or who were unlikely to be available to complete the full 2-year study were excluded.14 Knee radiographs were not obtained. The study was approved by the Alfred Hospital and Monash University Human Research Ethics Committees. All participants gave written informed consent.
Anthropometric data
Each participant’s height and weight was measured at the time of the original study (2002–3). BMI was calculated from these data (weight (kg)/height2 (m2)).
MRI and the measurement of cartilage volume, defects, bone area and BML
MRI
An MRI of the dominant knee (defined as the lower limb from which the subject stepped off when initiating gait) was performed between October 2003 and August 2004 and approximately 2 years later.14 Knees were imaged in the sagittal plane on a 1.5-T whole-body magnetic resonance unit (Philips) using a commercial transmit–receive extremity coil as described.14
Assessment of BML
BML were defined as areas of increased signal intensity adjacent to subcortical bone present in either the medial or lateral distal femur or proximal tibia T2-weighted images.6 Two trained observers, blinded to patient characteristics, together assessed the presence of lesions for each subject.6 The presence of BML was determined. A lesion was defined as “large” if it appeared on two or more adjacent slices and encompassed at least one quarter of the width of the tibial or femoral cartilage being examined from coronal images, comparable to the previously described “grade 2” BML by Felson et al.6 Lesions were further classified as “very large” if they appeared on three or more slices, comparable to the lesions described as “grade 3” by Felson et al.6 The reproducibility for the determination of BML was assessed using 60 randomly selected knee MRI (κ value 0.88, p<0.001). If an individual had more than one BML underlying a cartilage plate or within the knee, the BML of the highest grade was used for analysis.
Assessment of cartilage defects
Cartilage defects were graded on the magnetic resonance images in the medial and lateral tibial and femoral cartilages using a validated classification system.15–17 The cartilage defect score for a cartilage plate was defined by the most severe cartilage defect present, graded as follows: grade 0, normal cartilage; grade 1, focal blistering and intracartilaginous low-signal intensity area with an intact surface and bottom; grade 2, irregularities on the surface or bottom and loss of thickness of less than 50%; grade 3, deep ulceration with loss of thickness of more than 50%; grade 4, full-thickness cartilage wear with exposure of subchondral bone. A cartilage defect had to be present in two or more consecutive slices. The most severe cartilage defect present in the cartilage plate was used for analysis. The cartilage defect score for a compartment was calculated by summing the grades of the most severe cartilage defect in the tibial and femoral cartilage plates in that compartment. The medial and lateral scores were summed to obtain the total tibiofemoral cartilage defect score. Cartilage defect scores were read blind to sequence. Intraobserver reliability (expressed as intraclass correlation coefficient) was 0.90 for the medial tibiofemoral compartment and 0.89 for the lateral tibiofemoral compartment.3 Change in cartilage defects was classified as to whether the defect score progressed (ie, increased score), regressed (ie, reduction in score) or remained stable (ie, no change in score).
Cartilage volume measurement
The volumes of the medial and lateral tibial cartilage plates were measured using image processing on an independent workstation using the software program Osiris (University of Geneva). A trained observer read each MRI blinded to the timing of images. Independent measures of volume were made in a blinded fashion by a second trained observer.18 19 The coefficients of variation for the medial and lateral cartilage volume measures were 3.4% and 2.0%, respectively.18 19 Annual change in cartilage volume was calculated as (follow-up cartilage volume subtracted from initial cartilage volume) divided by the period of time between MRI scans.18
Bone area measurement
Medial and lateral cross-sectional areas of tibial plateau were determined by creating an isotropic volume from the input images that were reformatted in the axial plane. Areas were directly measured from these images. Coefficients of variation for the medial and lateral tibial plateau areas were 2.3% and 2.4%.18
Statistical methods
Baseline characteristics were compared between subjects in whom BML (at least large or very large) were present and absent, using unpaired t-tests for continuous variables and Somer’s d-test for ordinal variables. Ordinal logistic regression was used to determine whether the presence or absence of a BML predicted cartilage defect progression, stability or regression. Multiple logistic regression was used to determine the odds of cartilage defect progression versus regression/stability, and of cartilage defect regression versus stability/progression depending on whether a BML was present in the compartment being examined. The distribution of annual change in cartilage volume was examined for normality. Once confirmed, linear regression was used to determine whether there were differences in change in cartilage volume in subjects with and without large or very large BML in the medial and lateral tibiofemoral compartments, respectively, and in the total tibiofemoral joint after adjusting for potential confounding. A p value less than 0.05 (two-tailed) was regarded as statistically significant. All analyses were performed using the SPSS statistical package (version 15.0.0).
RESULTS
Of 176 women recruited to this study, 148 women completed the follow-up (84.1%). Twenty-eight women did not undergo a second MRI because of death (n = 1), migration (n = 1), knee injury (n = 4), surgery (n = 1), withdrawal of consent (n = 3) and being unable to be contacted (n = 18). Apart from being younger (mean age 49.5 years (SD 6.3), p = 0.02 for difference) the characteristics of these women were similar to those who completed the study.
Table 1 provides the baseline characteristics of the 148 women who completed the study. BML were present in 22 women (14.9%), including eight “very large” BML. There were 11 BML in the medial compartment, of which three were “very large”. In the lateral compartment, there were 11 BML, of which five were “very large”. Subjects with “large” BML had a significantly larger lateral tibial plateau area (p = 0.03), total tibial plateau area (p = 0.04) and higher cartilage defect scores in all compartments (p⩽0.001) than those without BML.
Are BML associated with the progression or regression of cartilage defects?
The number of knees in which the total cartilage defect score progressed, remained stable or regressed according to whether a “large” or “very large” BML was present at baseline is presented in table 2.
“Very large” BML were associated with an increased risk of progression of total tibiofemoral cartilage defects (odds ratio (OR) 5.55, 95% CI 1.04 to 29.6, p = 0.045) after accounting for age, BMI, initial defect score and bone area (table 3). Although large BML were associated with a tendency towards the total tibiofemoral cartilage defect score to progress (OR 2.89, 95% CI 0.96 to 8.68, p = 0.06), after accounting for the same factors, the strength of this finding weakened. The medial and lateral compartments were examined separately. When a “very large” BML was present in the medial compartment, cartilage defects in that compartment were more likely to progress (OR 31.4, 95% CI 1.60 to 607, p = 0.02; table 3). Similar results were obtained when large BML were excluded from the analysis (results not shown). These significant effects were independent of the initial cartilage defect score, which was the most significant determinant of the risk of progression of cartilage defects.
Whether BML were associated with the risk of cartilage defects improving (regressing) was examined. “Large” BML were associated with a tendency for cartilage defects to be less likely to improve (regress), but rather to remain stable or progress in the total tibiofemoral compartment (OR 0.27, 95% CI 0.05 to 1.60, p = 0.15). Although the direction of all point estimates suggested a protective effect in which there were sufficient BML to examine this question, these did not reach statistical significance (data not shown).
Are BML associated with change in cartilage volume?
The relationship between BML and annual change in cartilage volume in the medial, lateral and total tibial cartilage was examined (table 4). In the lateral compartment, univariate analysis suggested “large” BML were associated with a higher loss of cartilage volume (p = 0.02; table 4), such that when a “large” BML was present at baseline, cartilage loss occurred at a rate of 35.8 mm3/year higher than when no BML was present. After adjusting for age, BMI, initial cartilage volume and bone area, the significance of this difference increased (regression coefficient, r = 39.2 mm3/year, 95% CI 11.1 to 67.2, p = 0.007; table 4). This trend persisted when only “very large” BML were examined, although the magnitude of difference and significance was diminished (r = 29.1 mm3/year, 95% CI −12.8 to –70.9, p = 0.17). No significant effect of BML (either “large” or “very large”) on annual change in medial and total tibial cartilage was found (table 4). Similar results were obtained when “large” BML were excluded from the analysis (results not shown).
DISCUSSION
This study examined the relationship between BML and changes in knee cartilage in healthy women. In these women, when BML were present at baseline, tibiofemoral cartilage defects tended to be more likely to develop or become more severe. This relationship was stronger when only larger BML were considered. Also, when a BML was present in the lateral compartment, more rapid loss of lateral tibial cartilage was seen.
This is the first longitudinal study examining whether, in pain-free knees, BML are associated with a change in cartilage. In individuals with osteoarthritis, BML have been demonstrated to predict increased progression of cartilage defects and loss of cartilage volume.7 8 Increases in BML size have been associated with changes in urinary markers of cartilage metabolism.20 Our results related to cartilage defects were also stronger when a larger BML was present. In a healthy population, without knee pain, change in cartilage defects may be more sensitive than change in cartilage volume, accounting for the weaker relationship seen between BML and cartilage volume. These data thus suggest that the relationship between BML and cartilage changes is similar in knees with established symptomatic osteoarthritis as well as in knees without clinical knee osteoarthritis or pain. This supports the hypothesis that the structural changes in knee osteoarthritis may be viewed as a continuum, extending across the spectrum from the healthy joint to one with established knee osteoarthritis. Conversely, these data provide evidence against the construct that BML in asymptomatic women represents a different process to that seen in knee osteoarthritis.12
Whether BML directly cause changes in articular cartilage or whether a common exposure such as microtrauma or obesity affects both BML and change in cartilage is unknown. Nevertheless, BML may act as a marker for an increased risk of progression of cartilage defects in a pain-free population. Identifying factors that affect the progression of cartilage defects in a healthy population is important because cartilage defects have been proposed as an early marker of knee osteoarthritis;21 their presence predicts more rapid loss of articular cartilage in healthy individuals and predicts joint replacement in those with osteoarthritis.15 17 Whereas there is a tendency for defects to progress in a healthy population, approximately one third resolve.22 23 Identifying factors that prevent defect progression and/or promote defect resolution will be important in preventing knee osteoarthritis. However, to date, only lower BMI has been associated with the resolution of cartilage defects.22 Age and higher BMI have been associated with defect progression.22 23 These data suggest that BML, especially very large ones, are associated with an increased likelihood of cartilage defects progressing. It may be useful to identify factors that predict BML and affect their resolution, especially before the development of established knee osteoarthritis. The natural history of BML in a pain-free population is currently unknown; it is possible that by modifying this, the natural history of cartilage defects will also be affected, either directly or indirectly.
The association of BML with the progression of cartilage defects in a healthy pain-free population thus requires further study. In osteoarthritis, the presence and change in the size of BML is associated with parallel changes in serum markers of cartilage breakdown.20 It is possible that BML may be a useful therapeutic target in the prevention of knee osteoarthritis; by affecting the prevalence of BML we may be able to reduce or delay the onset of osteoarthritis and diminish the progression of cartilage defects. This suggests that a systemic approach may be feasible, with data suggesting that the presence of BML is susceptible to systemic factors such as medications and dietary intake. For example, the use of antiresorptive therapy (bisphosphonates and oestrogen therapy) in postmenopausal women has been shown to be protective of BML.24 More recent data suggest that dietary manipulation may also be helpful, given the recent findings that BML are more likely to be present when dietary vitamin C and fruit intake is low, and monounsaturated fatty acid, polyunsaturated acid and n-6 polyunsaturated acid intake is high.9 25
This study has a number of limitations. As only women were examined, these findings may not be generalisable to men. However, osteoarthritis is more prevalent in women and significant gender effects have been described in the knee structure. This may have improved our ability to demonstrate relationships. Although we cannot exclude radiographic osteoarthritis, none of the women had knee pain. These data would thus be generalisable to the population for whom the primary prevention of osteoarthritis would apply. The power of this study to show an effect was limited by the low prevalence of BML (14.9%). Therefore, although we were able to demonstrate significant relationships, albeit with wide confidence intervals, between BML and progression in cartilage defects in the medial and total tibial cartilages and change in lateral cartilage volume, we may have been unable to identify weaker relationships. It is emerging that a change in cartilage defects may be seen earlier than changes in cartilage volume in a healthy population. Therefore, an increased study duration may have enabled us to detect a stronger relationship between a change in cartilage volume and the presence of BML. Similarly, the low number of BML limited our power to show effects in both compartments. For example, we showed a significant relationship between BML and the progression of defects in the medial compartment only and between BML and lateral tibial cartilage loss. Although it is possible that these differences may be due to unmeasured differences (eg, biomechanical factors) between the compartments, it is also possible that with more cases we would have been able to demonstrate similar results throughout the knee. Despite this, the findings tended to be consistent, whether large or very large BML were examined as the predictor of change, and showed the same direction of effect in all cartilages examined. In this study we were not able to measure knee alignment, which may have affected our findings. However, recent work suggests that the relationship between BML and progression persisted after accounting for alignment.7 The main strength of this study is that it examines a healthy population of women longitudinally, allowing the relationship between BML and cartilage pathology to be examined in the early or prediseased state. We have also used a very conservative method for grading BML and cartilage defects, such that changes were required on at least two adjacent magnetic resonance images. This ensured that only large lesions were included, in contrast to some previous studies that required an abnormality to be present in one slice only.26 27
These data suggest that BML in pain-free knees, especially larger BML, are associated with the progression of cartilage defects and possibly cartilage volume loss in a healthy population. This suggests that BML have predictive validity, and may be useful as a target for the prevention of knee osteoarthritis. Given that recent data suggest that BML are affected by systemic factors, such as medications and diet, this raises the possibility that a systemic intervention, such as dietary manipulation, may be successful in reducing the prevalence of knee osteoarthritis. Even a small delay in the onset of clinical disease would have a substantial impact on the morbidity related to osteoarthritis, the most common form of arthritis worldwide.
Acknowledgments
The authors appreciate the assistance of Roy Morgan Research Australia in the conduct of this research. They would especially like to thank the study participants who made this study possible.
REFERENCES
Footnotes
Competing interests: None.
Funding: This work was supported by grants from the National Health and Medical Research Council of Australia (grant numbers 219279 and 334267). SRD is an NHMRC principal research fellow (490938). AEW, FH and YW are the recipients of NHMRC public health (Australia) fellowships (317840, 418961 and 465142, respectively). MD-T is the recipient of an Australian postgraduate award.
Ethics approval: The study was approved by the Alfred Hospital and Monash University Human Research Ethics Committees.
Patient consent: Obtained.