Introduction
There is increasing interest in the role of bone marrow lesions (BMLs), detected by magnetic resonance imaging (MRI), in the pathogenesis of knee osteoarthritis (OA) [
1,
2]. Histological examination of BMLs in knees has reported that they may represent areas of osteonecrosis, oedema, trabecular abnormalities and bony remodeling [
3]. BMLs are present in both symptomatic [
4‐
7] and asymptomatic populations [
8,
9]. Although BMLs are found to be extremely common in OA populations and, once present, are unlikely to resolve [
7,
10,
11], in asymptomatic populations they tend to have a more fluctuating course [
12]. BMLs have most commonly been described in relation to mechanical factors such as trauma [
13‐
16], knee malalignment [
17], and increased body weight [
8]. However, more recently systemic factors such as osteo-protective medications [
18] and nutritional factors [
19,
20] have been reported to affect the risk of BMLs.
Very little is known about the relation between BMLs and other changes in knee structures in asymptomatic, clinically healthy populations. Most previous studies have focussed on symptomatic populations with established knee OA [
6,
10,
11,
21], where BMLs are associated with knee symptoms [
4,
21‐
25] and progression of structural changes including joint space narrowing [
17], loss of cartilage [
6,
26] and increased prevalence and severity of cartilage defects [
23,
27]. More recently in an asymptomatic population, the presence of BMLs at baseline was shown to be associated with longitudinal progression of cartilage defects and loss of cartilage volume [
28] suggesting that BMLs also have a pathogenic role in pre-clinical OA.
The significance of development or resolution of prevalent BMLs has only recently been examined in populations with, or at high risk of, knee OA [
6,
7,
29]. In two of these studies, the majority of BMLs persisted so both could only examine the effect of change in size of the BMLs, had limited ability to examine incident BMLs, and had no power to investigate resolution [
6,
7]. In contrast, for participants of the Multi-centre Osteoarthritis Study (MOST) who either had or were at high risk of OA, approximately 40% of BMLs completely resolved and about one-third of cartilage locations developed new BMLs over 30 months, but no significant association between resolution of BMLs and change in cartilage was seen. In addition, the presence, resolution and progression of the BMLs was observed simultaneously within the same knee suggesting that complete resolution of all BMLs in a knee occurred less frequently. Worsening of BMLs and development of new BMLs was associated with increased cartilage loss compared with where BMLs remained stable [
29]; however, no comparison between knees with incident BMLs and knees that remained free of BMLs was made. Recently, we have shown for asymptomatic populations that BMLs fluctuate with about 50% resolving and about 14% of people developing new ones over two years [
12,
30]. Thus, the aim of this study was to examine the relation between incident BMLs and the resolution of BMLs prevalent at baseline and change in knee cartilage over two years in a cohort of asymptomatic middle-aged adults.
Discussion
In this cohort of asymptomatic middle-aged adults, the development of new BMLs in knees free of BMLs at baseline was associated with the progression of tibiofemoral cartilage defects over two years. In contrast, the resolution of BMLs was associated with reduced loss of medial tibial cartilage volume and a trend towards reduced progression of tibiofemoral cartilage defects.
The relation between incident BMLs and change in cartilage has only recently been examined [
29]. Among elderly participants with or at high risk of knee OA, development of new BMLs was associated with a worsening cartilage score as assessed using the WORMS (Whole organ MRI score) scale compared with knees where a BML remained stable; however, a comparison of cartilage loss with knees that remained BML free was not performed [
29]. Although we did not show a relation between incident BMLs and change in cartilage volume, there was progression of cartilage defects. This may be due to the relative short duration of two years of follow up; in a pain-free population, people are likely to have slower cartilage loss, and also due to the fact that cartilage defects are an earlier and independent marker of cartilage pathology [
37]. We have shown that cartilage defects are present in asymptomatic people with no clinical or radiological OA and to be predictors of cartilage loss in healthy people [
41] and those with OA [
37], independent of initial cartilage volume. Thus, it may be that the relation we have observed between incident BMLs and cartilage defects reflects early cartilage pathology and longer duration of follow up may be needed in order to observe subsequent cartilage volume loss.
In this asymptomatic population we found that the resolution of BMLs over two years was associated with beneficial effects on cartilage as evidenced by reduced loss of tibial cartilage volume and a trend towards reduced progression of tibiofemoral cartilage defects, suggesting this is not simply due to cartilage swelling. Our results are supported by recent observations in OA populations [
6,
7,
29]. For subjects with OA, an increase in size of BML was shown to be associated with increasing C-terminal cross-linking telopeptide of collagen type II levels [
6] and increased cartilage loss [
7,
29]. To our knowledge only one study, the MOST, has examined cartilage changes in knees where BMLs resolved; however, no significant association was observed between resolution of BMLs and change in cartilage [
29]. This may, at least in part, be due to the mixed nature of the MOST population because the purpose of the MOST was to examine a population at high risk of OA. In the MOST population, approximately 12% had symptomatic OA, approximately 24% had symptoms and about one-third had a Kellgren Lawrence score greater than or equal to two and past injury and surgery were not excluded. Therefore, the joints of these participants may already be further along the pathological pathway of structural change from the normal joint to one with OA, where the factors culminating in a BML, and acting on the whole knee, are established. In this situation, the reduction in change of cartilage associated with the resolution of BMLs may be lessened. In contrast, our population was asymptomatic and participants with prior injury or knee surgery were excluded.
There is growing evidence to suggest that BMLs have an important role in the pathogenesis of knee OA. They are common and persistent in symptomatic OA where they are associated with pain and progression of OA [
4,
6,
17,
21‐
26]. Although less common in asymptomatic people, BMLs are also associated with progressive knee cartilage pathology [
28,
42]. In this asymptomatic population with no clinical OA, the development of new BMLs was associated with adverse effects on knee cartilage, while resolution of BMLs was associated with improvement in cartilage. Although it has been suggested that BMLs are largely due to adverse biomechanical factors, we, and other investigators, have shown that systemic factors also affect the risk of BMLs [
18,
20,
43]. It may be that in the observed relation between BMLs and cartilage, factors contributing to the development of BMLs have resulted in impairment of the supply of nutrients and oxygen to the overlying cartilage plate, which may also reduce the strength of the bony support of articular cartilage [
44,
45]. Our data also suggest that this is reversible because resolution of BMLs was associated with reduction in cartilage defects and cartilage loss. Thus identifying factors that reduce the incidence of BMLs and increase their resolution may offer therapeutic targets in the prevention of knee OA.
This study has a number of potential limitations. Firstly, it examined a healthy asymptomatic 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. On the other hand, the findings from our study can be generalised to populations that may be targeted for primary prevention or early treatment of knee OA. Second, we did not obtain radiographs of the knees, so some subjects may have had asymptomatic radiographic OA. However, we used the American College of Rheumatology clinical criteria of OA [
32] to determine the status of knees, and individuals with significant knee injury in the past, pain at baseline, knee surgery or medical diagnosis of any other type of arthritis were excluded. Due to the small number of persistent BMLs we were unable to examine change in BML size. The small number of BMLs may have also reduced our power to detect significant associations and may explain the trends reported. In this study we did not assess knee alignment, which has been shown to be associated with the presence of BMLs [
17]. If malalignment were to be a major determinant of BMLs, we would not expect it to change significantly in a healthy asymptomatic population over a period of only two years, so would expect it to underestimate the relations we observed.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
FC, AW, DE, GG and RO were all involved in the design and implementation of the study including data collection and measurement. MD, AE, AF, YY and FC were involved in the analysis and interpretation of the data. All authors were involved in the manuscript preparation.