Ann Rheum Dis 67:683-688 doi:10.1136/ard.2007.073023
  • Extended report

Correlation between bone lesion changes and cartilage volume loss in patients with osteoarthritis of the knee as assessed by quantitative magnetic resonance imaging over a 24-month period

  1. J-P Raynauld1,
  2. J Martel-Pelletier1,
  3. M-J Berthiaume2,
  4. F Abram3,
  5. D Choquette1,
  6. B Haraoui1,
  7. J F Beary4,
  8. G A Cline4,
  9. J M Meyer4,
  10. J-P Pelletier1
  1. 1
    Osteoarthritis Research Unit, University of Montreal Hospital Centre (CHUM), Notre-Dame Hospital, University of Montreal, Montreal, Quebec, Canada
  2. 2
    Department of Radiology, Maisonneuve-Rosemont Hospital, Montreal, Quebec, Canada
  3. 3
    ArthroVision, Montreal, Quebec, Canada
  4. 4
    Procter & Gamble Pharmaceuticals, Mason, OH, USA
  1. Dr J-P Raynauld, MD, FRCPC, Osteoarthritis Research Unit, University of Montreal Hospital Centre (CHUM), Notre-Dame Hospital, 1560 Sherbrooke Street East, Montreal, Quebec, Canada, H2L 4M1; jp.raynauld{at}
  • Accepted 21 August 2007
  • Published Online First 29 August 2007


Objective: To evaluate in patients with knee osteoarthritis (OA) the size changes in bone oedema and cysts over 24 months, and to contrast these changes with cartilage volume loss using quantitative magnetic resonance imaging.

Methods: 107 patients with knee OA, selected from a large trial evaluating the effect of a bisphosphonate, were analysed by magnetic resonance imaging at baseline and 24 months. Assessments of subchondral bone oedema and cysts, and cartilage volume were done.

Results: At baseline, 86 patients showed the presence of at least one type of bone lesion: 71 oedema, 61 cysts and 51 both. At 24 months, although not statistically significant, the oedema total size change increased by 2.09 (SD 15.03) mm, and the cyst by 1.09 (8.13) mm; mean size change for the oedema was +0.38 (2.18) mm and −0.10 (4.36) mm for the cyst. When analysed according to subregions, an increase was found for the cyst size in the trochlea (+0.67 (2.74) mm, p = 0.02) and in the lateral tibial plateau (+0.15 (0.83) mm, p = 0.09), and for the oedema size in the medial tibial plateau (+1.73 (8.11) mm, p = 0.05). At 24 months, significant correlations were seen between the loss of cartilage volume and oedema size change in the medial condyle (−0.40, p = 0.0001) and the medial tibial plateau (−0.23, p = 0.03), and the changes in cyst size in the medial condyle (−0.29, p = 0.01). A multivariate analysis showed that the oedema size change was strongly and independently associated with medial cartilage volume loss (−0.31, p = 0.0004).

Conclusion: These data demonstrate that bone lesions are prevalent in knee OA. The correlation of the oedema and cyst size increase in the medial compartment over time with a greater loss of cartilage volume in this area underlines the importance of subchondral bone lesions in OA pathophysiology.

With the population ageing, osteoarthritis (OA) is becoming an increasingly common cause of disability.1 2 Magnetic resonance imaging (MRI) allows precise visualisation of joint structures such as cartilage, bone, synovium, ligaments and menisci, as well as the pathological changes in these tissues. Our group3 4 among others57 has recently developed and validated a system capable of quantifying knee cartilage volume using MRI combined with a dedicated software.

A large clinical trial assessing the effects of a bisphosphonate on knee OA structural changes over 24 months was recently completed.8 A longitudinal study of a subset of subjects from this clinical trial was recently reported and demonstrated a significant global cartilage volume loss at 24 months of follow-up.9 Data also showed that rapid disease progression might have been predicted at the outset of the study, based on certain clinical variables: female gender, high body mass index (BMI), higher level of pain and stiffness, and reduced joint mobility. A more rapid disease progression was further predicted by concomitant meniscal damage, mainly in the form of tear and extrusion.10 Bone lesions, although rarely seen on standard radiographs, are frequently observed in MRI exams as an abnormally high signal (hypersignal) within the subchondral bone area.1113 The term bone marrow oedema is usually used when such hypersignal is non-circumscribed, while a bone cyst describes a sharply delineated hypersignal.13 The pathological significance of such hypersignal is not entirely clear as it may not necessarily represent oedema per se, but can also be due to cellular infiltration.14 Bony cysts on the other hand may represent an intrusion of synovial fluid under pressure into the substance of the bone as they usually lie close to the surface,15 probably in a myxomatous phase in early formation as previously suggested.12 Cartilage microfractures, from mechanical stress, may also explain such cyst formation.16 The clinical importance of bone lesions while assessing knee OA with MRI was initially shown by Felson et al17 demonstrating that bone marrow lesions (oedema) were strongly associated with knee pain. Such association was further shown by others13 18 especially if an adjacent cartilage full-thickness defect19 or bone attrition18 was also present. However, it seems that there is no association of knee symptoms with bone cysts as suggested by Hayes et al13 and Torres et al.18 To our knowledge, longitudinal follow-up of these bone lesions seen in knee MRI, either oedema or cyst, was done, and no clear relationship was ever established between the change of the lesion size and the concomitant change in structural cartilage volume over time.

The aim of the present study was to evaluate, from a knee OA patient cohort, the presence and the change in size of bone oedema and cysts over 24 months, to contrast these changes with cartilage volume loss using quantitative MRI and to explore whether these bone changes could be predictors of cartilage volume loss over time.


Patient selection

A subset of 107 patients as previously described9 was selected from 1232 patients enrolled in a large clinical trial evaluating the impact of a bisphosphonate on knee OA.8 These patients, recruited from outpatient rheumatology clinics, fulfilled the American College of Rheumatology criteria for knee OA,20 and had symptomatic disease that required medical treatment in the form of acetaminophen, traditional NSAIDs or selective COX-2 inhibitors. Eligible patients were required to display radiological evidence of OA of the affected knee within 6 months of the outset of the study. Finally, patients had to have a minimum joint space width of the medial compartment between 2 and 4 mm, at least one osteophyte, and a narrower medial compartment compared with the lateral compartment. No patient had sole lateral compartment disease.

Patients were excluded if they had chondrocalcinosis, if their disease was secondary to other conditions, including inflammation, sepsis, metabolic abnormalities and trauma, or if they displayed any contraindication to the use of MRI. Further exclusion factors included a radiological grade IV on the Kellgren–Lawrence scale for the study knee or severe (class IV) functional disability. Patients were permitted to receive simple analgesics or NSAIDs, the regimens of which could be changed according to the preference of the rheumatologist and the clinical course of the patient. A centralised ethics committee approved this study, and each patient gave informed consent.

Briefly, the baseline characteristics of this cohort were largely in line with the demographic and disease characteristics of a general OA population: mean age was 62.4 (7.5) years, 64% were female, average BMI of 30.6 (4.3) kg/m2. The other characteristics of this study population were previously described.9

Clinical evaluation

Patients underwent clinical evaluation at baseline and 24 months. They were first evaluated on the basis of the WOMAC index.21 Its French-Canadian translation has been fully validated.22 In addition, the patients themselves used a visual analogue scale to make a global assessment of their condition (patient global assessment: 0 = very good; 100 = very bad) and to rate the pain they were having that day (patient pain score: 0 = no pain; 100 = most severe pain). Finally, the SF-36 was administered to the patients at each visit.23 A washout of medications was done before clinical evaluation; NSAIDs were discontinued at least 48 h before the investigation, and acetaminophen, 24 h.

Knee magnetic resonance imaging acquisition

High-resolution, three-dimensional MRI was obtained for each patient with OA at baseline and at 24 months, using the commercially available Magnetom Vision 1.5 Tesla machine with integrated knee coil (Siemens, Erlangen, Germany).24 These examinations are optimised three-dimensional FISP acquisitions with fat suppression. All parameters were set to produce images with the highest cartilage contrast, resolution, and signal-to-noise ratio within a reasonable acquisition time.

Magnetic resonance imaging data processing

Cartilage thickness and knee joint volume were measured by two trained blinded readers using a specially developed computer program (Cartiscope; ArthroVision Inc., Montreal, Quebec, Canada).3 24 Difference maps between the acquisitions for patients with OA at baseline and those at 24 months were blinded in terms of which time point was assessed. The change in knee cartilage volume was obtained by subtracting the follow-up volume from the initial volume.4 The change in cartilage volume over time was calculated for the entire knee (global), for each of the knee medial and lateral compartments, and for several cartilage subregions: the trochlea; the medial femoral condyle, the lateral femoral condyle and their summation (femoral condyles); the medial tibial plateau, the lateral tibial plateau and their summation (tibial plateaus). The reproducibility of the method was previously reported.4 The coefficient of variation was excellent with a CV RMS of 2.2% for the global cartilage volume, 1.6% for the medial and 2.6% for the lateral compartment.

Subchondral bone changes (fig 1)

Figure 1 Representations of bone cysts and oedema as seen by magnetic resonance imaging in the human knee femoral condyle. Arrows indicate cysts in the trochlea and patella, and the dotted line the oedema.

The evaluation of subchondral bone changes was performed on the same MRI as those used for the cartilage assessment. Two main types of bone lesion hypersignals were typically observed: a white hazy hypersignal (oedema) and a white sharply delimited hypersignal (cyst). To assess specific bone lesions (oedema or cyst), the reader manually selects the MRI slice that yields the greatest bone lesion size. This maximum size is then measured in millimetres using software cursors applied to the greatest diameter of each bone lesion. This is repeated for each lesion throughout all knee subsections. A reliability study was done according to a two-reader consensus measure in millimetres of a specific lesion size twice at a 6-week interval. Thirty-eight cysts and 41 oedema lesions were randomly selected from 10 patients. The reliability study results of the lesion size measurements were r = 0.96, p<0.0001 for the cysts and r = 0.80, p<0.001 for the oedema (test–retest Spearman correlation).

All lesion size changes were measured comparing baseline with 24 months of follow-up. For each patient, the mean size change of the bone lesions (oedema or cyst) throughout the entire knee was computed as well as the summation of all the lesion size changes. As a bisphosphonate (risedronate) was administered to some patients as per the original protocol,8 the bone lesion sizes were also evaluated according to the original four treatment groups (risedronate 5 mg/day, 15 mg/day, 50 mg/week, or placebo).

Meniscal lesions

The meniscal damage evaluation was also performed using the same MRI used for cartilage assessment. The meniscal damage scoring system was as published,10 and knee menisci were evaluated by an experienced radiologist who was blinded to the time sequences and cartilage volumes. In brief, the proportion of the menisci affected by the tear or extrusion was separately scored using a semi-quantitative scale: 0 = no damage, 1 = one of three areas involved (anterior, middle, posterior horns), 2 = two of three involved, 3 = all three areas involved. The extent of meniscal extrusion on the medial or lateral edges of the femoral tibial joint space, not including the osteophytes, was evaluated for the anterior, middle and posterior horns of the menisci in which 0 = no extrusion, 1 = partial extrusion, and 2 = complete extrusion with no contact with the joint space (severe).

Knee x-rays

The joint space width of the target knee was evaluated at baseline and 24 months at the narrowest point in the medial tibio-femoral compartment,25 allowing for the standardisation of radiographs by positioning the knee in a semi-flexed position under fluoroscopic guidance. Each of the radiographs measured the minimum joint space width in the medial compartment using an automated computerised method.

Statistical analysis

Data were systematically entered into a computerised database using a blinded double-entry procedure, after which descriptive statistics for patient characteristics were tabulated. The bone lesion size (oedema and cyst) and the cartilage volume changes from baseline to 24 months were blindly estimated with a one-sample t-test for each time point. The impact of bisphosphonate upon the bone lesions and upon the cartilage volume loss according to the presence or absence of these bone lesions was assessed using one-way ANOVA and 2×2 comparison of each treatment regimen versus placebo using the two-sample t-test. Changes in bone lesions versus baseline were statistically assessed by a one-sample Student’s t-test. The cartilage volume losses in the subregions of the knee were assessed as mean percentage losses compared with baseline. Univariate correlations between bone lesion changes and cartilage volume loss were assessed by the Spearman correlation test. Multivariate linear regression was used to correct these correlations for confounders and included bone lesion size at baseline, age, gender, BMI, and the meniscal lesions, tear and extrusion, yielding coefficient regressions with corresponding p-values. Correlation and regression coefficients with a negative sign mean greater association with cartilage volume loss. For gender, a negative sign means more cartilage loss for women. All statistical analyses were done using Statistica, version 7 (StatSoft, Tulsa, OK, USA). All tests were two-sided, and p⩽0.05 was considered statistically significant. Analyses were not adjusted for multiple comparisons.


Bone lesions at baseline

Patient baseline characteristics differed according to presence (n = 86) or absence (n = 21) of bone lesion, either oedema or cyst (table 1a). The patients with bone lesions had greater weight and BMI and a trend towards greater age. Similar characteristics were found for the patients having oedema (n = 71; table 1b) where again higher BMI and a trend towards greater weight was demonstrated. Moreover, for the patients with bone cysts (n = 61; table 1c), greater age, BMI and weight were found.

Table 1 Patient characteristics at baseline

Gender, joint symptoms, especially knee pain and function, generic quality of life (SF-36 physical component) and knee joint space width variables were similar at baseline regardless of the presence or absence of bone lesions.

Change in knee bone lesions over time

At 24 months, the mean cyst size change throughout the entire knee showed an increase of 0.38 (2.18) mm (mean (SD)) and mean oedema size change a decrease of 0.10 (4.36) mm (table 2). The summation of all lesion size changes yielded to an overall increment of 2.09 (15.03) mm for oedema size and 1.09 (8.13) mm for cyst size. Due to the variability in bone lesion size changes, none of these reached statistical significance.

Table 2 Bone lesion size changes over 24 months for the entire knee

Change in bone lesion size over time in knee subregions

Interestingly, when the data were analysed according to subregion (table 3), a significant increase in lesion size was found for the mean cyst size in the trochlea with a trend in the lateral tibial plateau, and for the mean oedema size in the medial tibial plateau.

Table 3 Mean bone lesion size changes at 24 months in subregions of the knee

Univariate correlation between bone lesion changes and regional cartilage volume loss

When the data on the bone lesion changes at 24 months in the different subregions were contrasted to the cartilage volume loss (table 4), significant correlations were seen with change in oedema size in the medial compartment, in which cartilage loss in both the medial femoral condyle and the medial tibial plateau were statistically significant. Comparison with cyst size change showed a positive correlation only for the medial femoral condyle. No statistical correlation was seen for any subregion in the lateral compartment.

Table 4 Univariate Spearman correlations

Multivariate linear regression predicting cartilage volume loss at 24 months

A multivariate analysis controlling for confounding factors such as age, gender, BMI, meniscal extrusion and tear, pain and bone lesions at baseline, showed that mean oedema size change was strongly and independently associated with the medial femoral condyle cartilage volume loss at 24 months (table 5a) while such independent association was not seen when looking at the global (table 5b) or the medial compartment (table 5c) cartilage loss. As expected, meniscal tear and severe meniscal extrusion were also correlated with medial femoral condyle cartilage loss. Several multivariate models looking at other cartilage areas, such as the lateral compartment or the central portions of the femoral condyle or tibial plateau, did not reveal significant associations (data not shown).

Table 5 Multivariate linear regression 

Effects of bisphosphonate on bone lesions

The patients included in this study were recruited from a large study looking at the bisphosphonate (risedronate) impact upon knee OA progression.8 The original trial had four treatment arms, placebo, and three risedronate regimens: 5 mg daily, 15 mg daily and 50 mg weekly. Among the patients that had any bone lesions, 24 were allocated to placebo, 22 to 5 mg daily, 17 to 15 mg daily, and 23 to 50 mg weekly. No impact of any of the bisphosphonate regimens was found on the bone lesion changes (oedema, p = 0.53; cysts, p = 0.71). Comparisons between placebo and any of the treatment regimens and among the two types of bone lesions were not statistically significant (data not shown). Interestingly and although not statistically significant, a difference was found for an absence of progression of the mean (SD) cyst and oedema size in the risedronate 50 mg weekly group (cyst, −0.4 (12.3) mm; oedema, −0.3 (12.7) mm) compared with the placebo group where the mean (SD) size increased (cyst, +1.9 (7.6) mm; oedema, +0.4 (14.5) mm) (cyst, p = 0.46; oedema, p = 0.86). The results of cartilage volume loss (global, medial compartment and medial femoral condyle) for the same patients with bone lesions were not different among the four groups of the risedronate study. Patients without any lesions were too few to detect any statistical significance on cartilage volume loss among the four treatment groups (data not shown).


Through the use of MRI on a longitudinal study of 107 subjects with symptomatic OA of the knee, we demonstrated that bony lesions such as oedema or cysts are extremely prevalent (more than 75%), which is in agreement with previous reports.13 17 2628 These data also demonstrate that bone changes are prevalent in knee OA. Moreover, the correlation between the increase in the oedema and cyst size in the medial compartment over time and a greater loss of cartilage volume in this area underlines the likelihood of a role for subchondral bone lesions in OA pathophysiology.

The clinical importance of bone marrow oedema while assessing knee OA with MRI was initially shown by Felson et al17 in a cross-sectional observational study of 401 symptomatic patients, which demonstrated that bone marrow lesions, especially the large ones, were strongly associated with knee pain. Such association was further shown by others13 18 particularly if an adjacent cartilage full-thickness defect19 or bone attrition18 was also present. However, the present study as well as others26 did not corroborate these findings, but agree with the study of Kornatt et al,29 which found that knee effusion and osteophytes were associated with knee OA pain but no other MR findings. This discrepancy between studies may reflect the heterogeneity of the knee OA population in age, disease duration, co-medication and also, importantly, the fact that pain perception is multifactorial. Here, we demonstrated that bone lesions were associated with age, weight and BMI but no other clinical variable. It was not entirely unexpected to find no association of pain with bone oedema, as most patients already had baseline oedema lesions. Alternatively, but not exclusively, it could be that the patients recruited for this study were less symptomatic than those having severe knee pain who had undergone MRI to investigate the pain source as in other studies.

Some studies have examined the quantitative changes in cartilage volume over time in a symptomatic population with knee OA 4 9 3032 or contrasted these quantitative changes with other anatomical structural changes of the knee, such as meniscal damage,10 28 33 34 but few look at the presence of bone marrow oedema at baseline and its association with faster cartilage degradation. Felson et al35 demonstrated such association in which knees with medial bone marrow lesions showed a higher incidence of medial progression than those without lesions. In addition, the same group36 showed that 69% of the knees having bone marrow oedema in the medial compartment also had radiographic progression. They also found a strong association between bone lesions and cartilage loss demonstrated by radiographs with limb misalignment.28 Our recently published study9 demonstrated that the strongest predictors of cartilage loss in patients with knee OA, measured quantitatively with the use of MRI, were the presence of severe meniscal extrusion, severe medial tear, and medial and/or lateral bone oedema along with some clinical variables, providing additional arguments to support the relationship between cartilage volume loss and other anatomical knee changes.

In this study, we demonstrated that the oedema or cyst size changes were more significant in some areas of the knee, a clear indication that OA bone structural progression, like cartilage, is not evenly distributed in this joint. The finding of significant changes in subregions of the knee compared with the entire knee may reflect heterogeneity of the lesion changes from one region to another. For instance, some areas, such as the trochlea, have a greater propensity for cyst size change (table 3), which is in accordance with Hayes’ findings demonstrating the majority of the bone marrow lesions and cartilage defects were mainly located in the patellofemoral compartment of the knee.13 The same is true for oedema size change in the medial tibial plateau.

The correlation of the increase in the oedema and the cyst size with cartilage volume loss was seen in the medial but not the lateral compartment. A potential explanation would point toward an increase in subchondral bone biophysical forces as most trans-joint pressure is encountered in this area when standing and walking (weight-bearing area). Moreover, our study specifically demonstrated the association between cartilage volume loss and the change in bone marrow oedema size, independent of any other clinical variable, especially meniscal damage, also known to be a strong predictor of OA progression. The fact that weight and BMI were closely associated with the presence of bone cysts, and BMI with bone oedema, as suggested by our findings (tables 1b and c) may also support the hypothesis of this greater trans-joint pressure to the bone lesion aetiology.

In our study, we did not show a convincing trend toward a decrease in cyst or oedema lesion size changes at 24 months for patients who received high doses of a bisphosphonate compared with placebo. In their cross-sectional study, however, Carbone et al37 showed that bisphosphonate and oestrogen therapy significantly decreased the prevalence of bone lesions and knee pain. This discrepancy could be explained by the fact that our study was not adequately powered to fully assess the role of an anti-resorptive agent on the bone lesions.

The present study, like any other, has its limitations. Our cohort is representative of the average patient population with typical knee OA seen at a rheumatology outpatient clinic. The use of a bisphosphonate in this study did not significantly impact bone lesion progression; however, as the number of patients within each group was small, no definitive conclusions could be made. It is also possible that bisphosphonate use in the majority of our patients may have tempered the true signal of the bone lesion progression that only a cohort of a large number of patients without such medication could provide. Knee alignment, which could also impact medial knee cartilage loss as suggested by Hunter et al,28 was not assessed in this study. Knee misalignment may indeed create focal stress on the medial compartment, which may in turn trigger the appearance of bone oedema and cartilage loss. This will be further explored in ongoing studies.

Because of the relatively short time span of our study, we also failed to demonstrate a temporal relationship between the appearance of a bone lesion and the subsequent cartilage defect or vice versa. It remains unknown which of the pathological features appears first, which could then establish a pathophysiological link. The same is also true for the relationship between the oedema lesion and the cyst lesion.

We also did not look at a potential association between the bone cysts seen on MRI and those on standard radiographs. It seems, however, that a large number of the cysts seen on MRI are not visible otherwise, as suggested by Ostlere et al.11

The nature of the bone lesions (oedema or cyst) seen on the MR images remains unclear as we did not have access to any pathological specimens; for example, those collected after a total knee replacement. Interesting insightful information from further anatomical and histopathological studies could reveal the nature of the bone lesions and their role in the later stage of OA and its contribution to cartilage degradation and knee symptoms.

In summary, this study showed that subchondral bone lesions, such as oedema and cysts, readily seen in MR acquisitions and easily quantifiable, may help identify subgroups at risk for faster disease progression, which in turn may impact patient selection for eventual treatment with structure-modifying OA drugs. However, understanding the reason for the distribution of these bone changes may be even more important, as they may reflect biomechanical changes that need to be modified, as they may reflect a population that will not readily respond to pharmacological interventions in the absence of biomechanical ones.


We would like to thank Virginia Wallis for her assistance in the manuscript preparation. We also would like to acknowledge Josée Thériault and André Pelletier for their technical expertise. We are grateful to Procter & Gamble Pharmaceuticals (Mason, OH, USA) for supporting part of the study with a grant.


  • Competing interests: None.


Free sample This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of ARD.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.