This study used a simple and novel method of measuring approximate BML volume that can be easily employed by researchers and clinicians. The technique demonstrated that many knee regions experienced small longitudinal changes in regional approximate BML volume over 2 years (± 2.5 cm3
) but some knees experienced large longitudinal changes. In agreement with previous findings, baseline regional approximate BML volume and hyaline cartilage full-thickness lesion area (baseline and longitudinal change) were associated [2
]. The new BML measurement technique is sensitive to small changes in BML size over 2 years and, in agreement with previous literature, positively associated with increased cartilage damage.
To our knowledge, this is the first full-length publication reporting the longitudinal relationship between 3-dimensional quantitative measurements of BML size and hyaline cartilage among knees with OA. Several methods have been deployed to quantify BML size but they have various limitations. One method uses manual segmentation of all MR image slices that display a BML. This method is time consuming and usually requires proprietary software [29
]. Therefore, a segmentation technique is difficult to deploy researchers working with large cohorts. A less labor-intensive method that has been used is to take the measurement of the greatest diameter of a lesion. One potential downside to this approach is that it does not account for the three-dimensional nature of these lesions [7
]. An alternative method that is not very labor intensive approximates BML volume with three linear measurements; specifically, the coronal and sagittal images with the largest area of signal abnormality are identified and the largest three dimensions from these two images [37
]. While this appears to be a good compromise between function and practicality, this method does not take into account for the possibility that the greatest width of a BML may change as you view adjacent coronal or sagittal images (Figure ). By measuring the greatest diameter across images the current technique attempts to account for the potential limitations that a single image may not adequately describe the 3-dimensional size of an irregularly shaped BML.
Figure 4 Measurements of approximate BML volume based on only the coronal and sagittal images with the largest area of signal abnormality may misrepresent the greatest diameter of a BML. Image 10, image with the largest BML area, would traditionally be measured (more ...)
Baseline regional approximate BML volume is positively associated with baseline and longitudinal changes in full thickness cartilage lesion area. A positive correlation between semiquantitative assessments of BML size and cartilage damage have been demonstrated in cross-sectional [4
] and longitudinal studies [2
]. Previous reports regarding the within-region association between BML size and cartilage volume, the most common quantitative measure of hyaline cartilage, are conflicting with some showing statistically significant associations [7
] and others showing non-significant associations [2
]. The current correlations concur with previously reported associations between longitudinal change of BML size (greatest diameter) and cartilage volume (r
= -0.03 to -0.40)[7
]. Furthermore, one article, which distinguished between full thickness cartilage lesions and normal or partial thickness lesions, noted that only 7% of knees (11/149 knees) with normal cartilage or partial thickness defect had a BML greater than 1 cm in size while 33% of knees (16/48 knees) with a full thickness defect had a BML greater than 1 cm [10
]. These data support the current findings that BML size may have a strong association with full thickness cartilage lesion area.
Baseline regional approximate BML volume was statistically associated with baseline and longitudinal changes in full thickness cartilage lesion area but not cartilage thickness or volume. The similar findings between cartilage thickness and volume are not unexpected since the parameters are based on similar cartilage data (e.g. cartilage thickness determines cartilage volume). Although the design of this study did not allow us to determine why regional BML volumes had stronger correlations with full thickness cartilage lesion area compared to cartilage volume or thickness, we speculate that BMLs are an early and sensitive biomarker of localized degeneration in response to altered loading. Furthermore, BMLs may also contribute to localized cartilage loss leading to full thickness cartilage lesions. A second hypothesis is that BMLs have a localized influence on cartilage loss and full thickness cartilage lesion area is a more localized measure of hyaline cartilage loss than the other two measures. These hypotheses are supported by the associations between baseline BML size and longitudinal changes in full thickness cartilage lesion area. A third hypothesis is that the association between BML size and cartilage volume and thickness is impeded because OA progression defined by cartilage thickness and volume is not always linear. Some patients experience increased cartilage thickness with KL grade 2 (23% of knees) and KL grade 3 (18%) while the majority lose cartilage thickness and volume [39
]. In contrast, full thickness cartilage lesion area progresses in a linear pattern, which supports its higher correlations with BML size.
Exploratory analyses across regions demonstrated statistically significant associations including relations between baseline cartilage and regional approximate BML volume change. When assessing predictors of longitudinal cartilage loss or BML volume change it is important to consider other regions. Altered structure in one region of the index compartment is likely to alter the loading in both index regions. Inter-regional associations between BML and cartilage parameters warrant its own research. Few studies have evaluated inter-region relationships between BMLs and cartilage parameters but there is previous research to support these relationships. Statistically significant inter-regional correlations between baseline BML size (semi-quantitative) and cartilage volume change have been reported [28
]. When accounting for variables that predict disease progression (e.g. cartilage loss, BML progression) it may be important to consider the influence of other regions throughout the joint.
Our study showed that small changes in BML size were common but that there is a large variation among knees with OA. This corroborates findings from other investigators who also found that among OA knees with BMLs, there was a large variability of longitudinal BML size change [7
]. In a small sample (n
= 14), investigators using semi-automated BML segmentation showed that 43% increased BML volume (change greater than 5%), 36% decreased BML volume, and 21% did not change (± 5% change)[36
]. Small longitudinal changes with large variability has also been reported with assessments of maximum BML diameter (i.e. coefficient of variation [CV] = 719%, 4,360%)[7
While the findings of the current study are interesting, there are several limitations. Approximate BML volume over-estimates the true BML volume but may be strongly correlated to the true BML volume. Furthermore, the method may not detect subtle changes within the baseline approximate volume.
Our data support the construct validity of the new methodology. This technique is intended to be a simple and efficient method of measuring 3-dimensional BML size that can be done with standard MR imaging viewing software. Future research should determine the relationship between true BML volume determined through BML segmentation and approximate BML volume using this technique.