This is, to our knowledge, the first study that reports on causes of bone marrow lesions and documents a genetic contribution to both the prevalence and severity of bone marrow lesions in subchondral knee bone. The heritability estimates were reduced by a small amount after adjustment for chondral defects and ROA, suggesting that they share common genetic mechanisms to only a limited degree. The heritability estimate for knee alignment was zero, suggesting that it is not a heritable trait. Bone marrow lesions were also associated with some structural change within the knee and have some risk factors in common with osteoarthritis.
MRI-defined bone marrow lesions were first described by Wilson and colleagues [20
] in patients with debilitating knee and hip pain. Felson and colleagues [2
] documented its clinical relevance to pain in OA of the knee. Sower and colleagues [8
] reported that women with bone marrow lesions and full-thickness chondral defects accompanied by adjacent subchondral cortical bone defects were significantly more likely than others to have painful OA of the knee. In a recent study of an older population [10
], we demonstrated that ROA was not independently associated with knee pain but MRI-defined bone marrow lesions were associated with knee pain independently of ROA and chondral defects, suggesting an independent effect and wider clinical relevance. However, both the pathology and causes of MRI-defined bone marrow lesions are unknown. Felson and colleagues [6
] reported that medial bone marrow lesions were more likely in OA patients with varus limbs, whereas lateral lesions were seen mostly in those with valgus limbs. Malalignment mediated 37 to 53% of the association between bone marrow lesions and progression of OA of the knee, suggesting that knee alignment may have a role in the occurrence of bone marrow lesions.
The current study is the first to document a significant genetic contribution, suggesting that further studies to identify specific gene(s) responsible for the development of bone marrow lesions might shed light on the prevention and management of knee pain. The heritability estimate was high for prevalent bone marrow lesions and independent of other factors including knee pain, knee alignment, chondral defects, and ROA, suggesting that they are under independent genetic control, with at most a small shared genetic component. However, the inability to estimate the standard error for the prevalence heritability estimates indicates that the results are not robust, possibly reflecting relative limitations of the program we used for dichotomous traits in comparison with continuous traits [21
]. It is likely that the true heritability is substantially lower.
In comparison with prevalent bone marrow lesions, the heritability estimate for severity of bone marrow lesions was lower, but with a smaller standard error. The estimate again remained significant after adjustment for other factors including knee pain, muscle strength and knee alignment, suggesting that they are not under common genetic control. However, the estimate was reduced by 8 to 9% after adjustment for chondral defects and ROA, suggesting that they share common genetic mechanisms to a limited degree.
In contrast to this, but consistent with previous reports [7
], was our observation that bone marrow lesions coexist with chondral defects and ROA of the knee, suggesting that they have environmental factors in common. Significant correlations between bone marrow lesions, age and BMI in the current study support this, although the increased prevalence in males suggests a possible role for trauma. However, in contrast to other reports [6
], we did not find a significant association between knee alignment and bone marrow lesions, possibly because of a low prevalence of ROA in this sample. Further studies with independent samples are needed to confirm these results and confirm whether bone marrow lesions independently predict cartilage loss as chondral defects do [23
The current study has several potential limitations. First, there is controversy about the ideal study design for estimating the heritability of disease. The twin model is often used but has been criticized as overestimating heritability because of the assumption of similar shared environments between monozygotic and dizygotic twins. This has been documented for bone mineral density [24
] but not for osteoarthritis. Family studies such as the present one may be more likely to represent true heritability but make it more difficult to assess the contribution of shared environment. However, before this study, little was known about environmental effects on bone marrow lesions and we adjusted for all significant covariates in the analysis, so the results do not support a strong shared environmental contribution.
Second, the choice of subjects who are at all at higher risk of disease may bias the heritability estimates and limit the generalizability of the results to the general population. However, it is most likely that this bias will act to decrease estimates by decreasing genetic heterogeneity in comparison with an unselected sample.
Third, the bone marrow lesions were assessed in only one plane and the scoring system may not differentiate between various sizes of lesions in sagittal plane. However, most lesions are spherical, which suggests that they will have the same anteroposterior and lateral dimensions and would be strongly correlated with a volumetric scoring system based on mathematical principles [22
]. Measurement error in the assessment of bone marrow lesions may have reduced the estimates. However, the method had high intra-observer reproducibility and we used a single observer for all readings, suggesting that this is not of major concern.
Fourth, using baseline X-ray measurements may not be appropriate because there was a two-year gap between the X-ray and MRI measurements. However, there is little radiographic change over this time frame and within-subject correlation for X-ray changes is very high, suggesting that this is not a big concern.
Fifth, bone marrow lesions in this sample were generally mild with grade 1 lesions accounting for 40% of the total prevalence, raising a concern of clinical relevance. However, these lesions have been associated with knee pain [2
], suggesting that they are still clinically relevant.
Last, a clear elucidation of the nature of MRI-defined bone marrow lesions is uncertain. In a histological study of specimens taken from end-stage knees undergoing total joint replacement, Zanetti and colleagues [25
] reported histological evidence of fibrosis, marrow necrosis and abnormal trabeculae for MRI-defined bone marrow lesions.