Bone marrow lesions (BML), detected by magnetic resonance imaging (MRI) have been implicated in the pathogenesis of knee osteoarthritis.1^–3 They have been shown to be present in both symptomatic2^–5 and asymptomatic populations.6 7 The presence of BML has been associated with structural changes in the knee, including joint space narrowing,8 loss of cartilage1 4 5 and increased prevalence and severity of cartilage defects.6 7 9 Although BML have been associated with knee pain, the data are conflicting, whereas a number of studies report a relationship between BML and pain,2 9^–13 others show no such association.3 4 14 15 Both mechanical factors such as such as trauma,16^–19 knee malalignment,8 and increased weight6 as well as systemic factors such as osteoprotective medications20 and nutritional factors21 22 have been shown to affect the risk of BML.

Little is known about the natural history of BML. Most previous studies have focussed on symptomatic populations with established clinical knee osteoarthritis rather than asymptomatic populations.4 5 13 15 23 In subjects with symptomatic knee osteoarthritis, the data suggest that BML are unlikely to resolve, with one study suggesting that 99% of BML either remained the same or increased over 15 or 30 months5 and another finding that only 10% resolved over 2 years.15 Similarly, in a population with chronic knee pain, which included subjects with and without radiographic osteoarthritis, only 22% completely resolved over 2 years.23

There are no data on the natural history of BML in asymptomatic populations with no clinical knee osteoarthritis or of factors that may effect the development of BML. The aim of this paper was to examine the natural history of BML in a healthy population, free of knee pain at baseline, and to identify factors associated with the incidence and progression of BML over 2 years and whether incident BML are associated with the development of pain.

METHODS

Subjects

Subjects were recruited from the Melbourne Collaborative Cohort Study, a prospective cohort study of community-based people, aged 40–69 years, established to examine the role of lifestyle and genetic factors in the risk of cancer and chronic diseases from middle age and beyond, as described.7 Subjects were excluded if they had: osteoarthritis, as defined by the American College of Rheumatology clinical criteria;24 current or past knee disease; a history of knee pain in the past 5 years lasting for more than 24 h; a previous knee injury requiring non-weight-bearing treatment for more than 24 h or surgery (including arthroscopy); or a history of any arthritis diagnosed by a medical practitioner or contraindication to MRI, as described.25 The study was approved by the Human Research Ethics Committee of the Cancer Council of Victoria and Monash University Standing Committee on Ethics in Research Involving Humans. All participants gave written informed consent.

Data collection

Study participants completed a questionnaire that included information on their demographics at baseline and at the 2-year follow-up. Weight was measured to the nearest 0.1 kg (shoes, socks and bulky clothing removed) using a single pair of electronic scales. Height was measured to the nearest 0.1 cm (shoes and socks removed) using a stadiometer. From these data, body mass index (BMI; weight/height kg/m²) was calculated. The development of pain in the knee was assessed at the 2-year follow-up by asking the question “have you had any pain in your knee in the past 12 months, yes or no?”, for those who answered yes, the duration of pain was also determined by asking whether this pain lasted for “less than 24 h”, “more than 24 h but less than half a month” or “more than half a month” in the past 12 months.

Magnetic resonance imaging

An MRI of the dominant knee of each subject (defined as the lower limb from which the subject stepped off when initiating gait) was performed at baseline and approximately 2 years later.6 Knees were imaged in the coronal plane on a 1.5-T whole-body magnetic resonance unit (Philips Medical Systems, Eindhoven, The Netherlands) using a commercial transmit–receive extremity coil. The following sequence and parameters were used: fat saturated, fast spin echo three-dimensional, T2-weighted (2200 ms, 20/80 ms/90° repetition time/echo time/flip angle) with a slice thickness of 3 mm, a 0.3 interslice gap, one excitation, a field of view of 11–12 cm and a matrix of 256 × 128 pixels.6

Bone marrow lesions

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 assessed on coronal T2-weighted fat-saturated images.26 Two trained observers (MD and AW), who were blinded to patient characteristics, as well as the sequence of images, together assessed the presence of lesions for each subject. The presence or absence of a 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. This is comparable to the previously described “grade 2” BML by Felson et al.2 Lesions were further classified as “very large” if they appeared on three or more slices. This is comparable to the previously described “grade 3” by Felson et al.2 Fig 1 shows a “large” (A) and “very large” (B) BML. The reproducibility for determination of the BML was assessed using 60 randomly selected knee MRI (κ value 0.88, p<0.001). If a person had more than one BML underlying a cartilage plate, the BML of the highest grade was used for analysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Coronal T2 images of a “large” (A) and “very large” (B) bone marrow lesion.

Statistical analysis

Descriptive statistics for the characteristics of the study subjects were tabulated. Independent sample t tests were performed to compare means and χ² analysis to compare proportions. The relationship between risk factors, incident pain and incident BML was assessed by binary logistic regression. A p value of less than 0.05 (two-tailed) was regarded as statistically significant. All analyses were performed using the SPSS statistical package (standard version 14.0).

RESULTS

Of the 297 pain-free subjects imaged at baseline, 271 (90%) participants completed the 2-year follow-up. The reasons for the loss to follow-up of 26 subjects were: death (three); withdrawal for health reasons (four); withdrawal of consent (10); ineligible for follow-up (pacemakers) (four) and inability to be contacted (five). The only significant difference between those who completed follow-up and those who were lost to follow-up was that those lost to follow-up were slightly heavier (p = 0.01). All participants were pain free at baseline.

Incidence of BML

A total of 234 of the 271 participants did not have a BML at baseline. Over 2 years, 33 knees developed new BML (14% of knees). Those who developed a BML weighed more (p = 0.02), had a higher BMI (p = 0.01) and reported a higher proportion of incident pain (p = 0.05) (table 1). Of the 33 BML that developed over the study, 20 (61%) were located within the medial and 13 (39%) within the lateral compartment. One person developed a BML in both the medial and lateral compartments. Of the 33 incident BML, 13 were graded as “very large”.

Factors associated with developing BML over 2 years are presented in table 2. Although age, gender and height were not significantly associated with developing a BML over 2 years, increased BMI was significantly associated with developing a BML both before (p = 0.002) and after (p = 0.001) adjusting for confounders. For every unit increase in BMI, there was an associated 15% increased likelihood of developing a BML (p = 0.001) and a 20% increased likelihood of developing a “very large” BML (p = 0.001), after adjusting for confounders.

We examined whether self-reported knee pain at follow-up differed in those subjects who developed a BML over 2 years compared with those who did not (table 3). Among people with no BML in the knee at baseline, incident knee pain was reported in 45% of people who developed a BML compared with 29% who remained BML free. Self-reported knee pain, assessed by the question “have you had any pain in your knee in the past 12 months, yes or no?”, was positively associated with incident BML over 2 years, with people who developed a “very large” BML being 4.2 (95% CI 1.2 to 15.1) times more likely to report knee pain than people who did not develop a “very large” BML, after adjusting for confounders (p = 0.03). Furthermore, developing an incident BML over 2 years was positively associated with the development of knee pain lasting more than half a month (OR 3.5, 95% CI 1.21 to 10.0, p = 0.02) in univariate analyses. The development of a “very large” incident BML was also significantly associated with the development of knee pain lasting more than half a month in the past 12 months in univariate analysis (OR 6.5, 95% CI 1.8 to 23.9, p = 0.005) and after adjusting for confounders (OR 5.0, 95% CI 1.11 to 23.3, p = 0.04).

Natural history of prevalent BML

Thirty-seven of the 271 participants who completed follow-up had a BML at baseline. Twenty-five (67%) BML were located within the medial and 14 (38%) within the lateral compartment. Two people had a BML in both compartments. Of the 37 BML that were present at baseline, 11 were graded as “very large”, with six located within the medial and five within the lateral compartment. Of the 37 BML that were present at baseline, 20 (54%) persisted at follow-up and 17 (46%) completely resolved. There were no significant differences in age (p = 0.7), gender (p = 0.98) and BMI (p = 0.50) between people who had a BML at baseline that resolved compared with people whose BML persisted over the 2-year study period (table 1). Furthermore, there were no significant associations between risk factors such as age (p = 0.8), gender (p = 0.98), weight (p = 0.4) and BMI (p = 0.6) and BML persisting over 2 years. There were no associations with the development of pain.

DISCUSSION

In this population of healthy participants with no clinical knee osteoarthritis, BML developed in 14% of people over 2 years. Increased BMI was a risk factor for incident BML, and incident BML were associated with the development of knee pain in a population in which all participants were free of pain at the beginning of the study. Approximately half of the BML present at baseline resolved over the 2-year study period.

This is the first study to report the natural history of BML and risk factors associated with the incidence of BML in a healthy, pain-free population. Incident BML have previously been reported in cohorts mixed in relation to radiographic osteoarthritis diagnosis and/or pain symptoms.5 15 23 In a population with chronic knee pain, in which 80% had radiographic knee osteoarthritis, new lesions developed in 21% of people over 2 years.15 Similarly, among patients with primary knee osteoarthritis, new BML developed over 15 or 30 months in 20% of knees.5 In a small study of 47 people with painful knees with (88%) or without (12%) osteoarthritis, eight new BML were identified over 2 years.23 Our finding that BML developed in 14% of people over 2 years is lower than has been described in symptomatic populations. Furthermore, approximately half the BML present at baseline resolved over 2 years, which is higher than has previously been shown in subjects with osteoarthritis and/or knee pain.4 5 15 23 These data suggest that, not only are BML less common in asymptomatic subjects with no clinical knee osteoarthritis, but they are more likely to resolve and the rate of development is lower.

In this study, we found that incident BML were associated with developing knee pain. The association between pain and BML is conflicting.3 4 14 15 To our knowledge only one previous study has reported an association between incident BML and incident pain and this was in a population of people who either had knee osteoarthritis or were at high risk of developing osteoarthritis.13 In that study, incident BML were found to be more common in the knees of people with incident pain over 15 months compared with those with no incident knee pain.13 This current study has extended these findings by observing a consistent relationship between incident BML and the development of pain in a healthy, middle-aged population. While subchondral bone has been suggested as a contributor to knee pain,27 the exact causes of knee pain remain unclear. It is known that the marrow of subchondral bone is richly innervated with nociceptive pain fibres.28 Taken together with the findings of our study, this suggests that the role of bone in knee pain may be partly mediated via BML development.

We found that increased BMI was a risk factor for incident BML. This has not previously been examined in a longitudinal study. However, these findings are consistent with data from a cross-sectional study of asymptomatic middle-aged women, which showed that the prevalence of BML was associated with increased body weight and BMI.6 A similar association was also found in a cross-sectional study examining a population containing people with and without radiographic osteoarthritis.10

The findings of this study suggest that BML occur in healthy populations but are less common and are more likely to resolve compared with osteoarthritis/symptomatic populations.9 In this current study, the incidence of BML was also associated with developing new knee pain, mirroring previous observations in osteoarthritis/symptomatic populations. This suggests that BML play a role in knee pain in osteoarthritis possibly as a continuum from a normal to a clinically diseased joint. Furthermore, in this study obesity was associated with developing new BML. Both mechanical factors, such as knee malalignment8 and increased weight,6 have been shown to be risk factors for BML prevalence. Knee alignment was not examined in this study and it may be that, in part, the effect of obesity as a risk factor for BML acts via malalignment. However, systemic factors such as nutritional factors21 22 also affect the risk of BML so a metabolic effect of obesity may also be possible. Given that BML have also been shown to be associated with structural changes associated with the progression of osteoarthritis,1 4^–9 it may be that by preventing their development in healthy populations it could help to reduce the incidence and burden of knee pain.

This study has a number of potential limitations. First, the study examined a healthy, community-based population selected on the criteria of no knee pain or injury, and therefore the results are not generalisable to symptomatic populations or people who have injured their knees. However, the findings of our study can be generalised to populations that may be targeted for primary prevention. Second, we did not obtain radiographs of the knees, so that subjects may have had asymptomatic, radiographic osteoarthritis. However, we used the American College of Rheumatology clinical criteria of osteoarthritis to determine the status of knees, and individuals with significant knee injury in the past, pain at baseline, knee surgery or a medical diagnosis of any other type of arthritis were excluded.24 It is possible that the findings reported of a lower rate of development and persistence compared with symptomatic/radiographic osteoarthritis populations reflect the low prevalence of undetected radiographic osteoarthritis that may be present in this population. However, it could be argued that the development of radiological osteoarthritis is a step along the spectrum of changes from the normal to the osteoarthritis knee. BML incidence and/or persistence may represent one such step in the pathway because it has been shown that 10% of knee cartilage is lost by the time the first radiological changes of osteoarthritis can be identified.29 Finally, due to the small number of prevalent BML at baseline, our analysis of persistent BML may not have been sufficiently powered to detect any weak associations between age, gender, BMI and symptoms.

The findings of this study suggest that BML develop in healthy populations at a lower rate than has been reported in osteoarthritis populations, and that approximately half of the BML present at baseline resolved. Increased weight and BMI were risk factors for incident BML. Furthermore, the incidence of BML was positively associated with the development of pain. These data suggest that in middle-aged people with no clinical knee osteoarthritis, BML are reversible and may provide a target for interventions aimed at symptoms and the prevention of knee osteoarthritis.