Bone marrow lesions in knee osteoarthritis change in 6-12 weeks.

OBJECTIVES: Knee osteoarthritis (OA) is thought to be a slowly evolving disease with glacial changes in cartilage morphology necessitating trials of potential treatments lasting 1-2 years with evidence that over 6 months change in cartilage is not detectable. In contrast to cartilage, bone has the capacity to adapt rapidly, such as after fracture. We tested whether bone marrow lesions (BMLs) change in volume in 6 and 12 weeks, suggesting they may provide evidence of short term fluctuations of joint damage.
METHODS: In 62 patients with patellofemoral knee OA (mean age 55.7 years, 59.7% women, mean BMI 31.0), we obtained baseline, 6 and 12 week knee MRIs with contrast enhancement. Of those with BMLs at baseline, we assessed BML volume on the axial proton density fat saturated (FS) images and postcontrast sagittal T1 weighted FS images. We manually segmented BML volumes, testing repeatability of BML volumes in knees remeasured. Using the standard deviation of the difference between repeated measurements to calculate Bland-Altman Limits of Agreement, we determined how much BML volume change represented a change greater than due to chance.
RESULTS: Fifty-two patients had BMLs at baseline. Test-retest reliability for BML volume was high (ICC 0.89, 95% CI 0.80-0.97). All knees showed at least some change in BML volume by 6 and 12 weeks. On the axial view at 6 weeks, 20/49 (40.8%) knees showed BML volume changes greater than the limits of agreement with similar results at 12 weeks. BML changes were evenly divided among knees with enlarging and shrinking BMLs. 63.3% of the knees had more than 50% change in BML volume at either 6 or 12 weeks on the axial view and 48.7% on the sagittal view.
CONCLUSIONS: Knee BML volumes change in several weeks in many persons with knee OA. To the extent that they could be regarded as treatment targets, trials testing BML effects could avoid the usual prolonged structure modification trials.

Introduction

Knee osteoarthritis (OA) is thought to be a slowly evolving disease with glacial changes in cartilage morphology necessitating trials of potential treatments lasting 1–2 years. There is evidence that change in cartilage over 6 months using quantitative cartilage morphometry is not detectable. The slow evolution of structural disease has been thought to be a major impediment to the rapid evaluation of promising treatments and makes it difficult to gauge whether treatments have structural effects on disease.

In contrast to cartilage, bone has the capacity to adapt rapidly. Bone marrow lesions (BMLs) show evidence histologically of bone damage or microfracture and are related to malalignment and to pain. BMLs have been shown to change in volume at 1 year of follow-up, and one study evaluating biochemical markers of cartilage metabolism in which repeated MRI's were acquired and scored semiquantitatively suggested that some knees had a change in BML volumes over 3 months.

There is a long history in medicine of intermediate outcomes being used as treatment targets for disease when the ultimate outcome occurs too infrequently (or too slowly) for trials to use it as an outcome. Examples include focussing on hypertension instead of its dreaded outcome, stroke and bone density instead of fracture. There is already evidence that BMLs are strongly related to outcomes of interest in OA; they increase the risk of cartilage loss; of OA progression by X-ray, and of the development of knee pain. To examine their potential value as outcomes in short term OA trials, it is necessary to determine whether they often change over periods shorter than 1 year, suggesting that, as therapeutic targets, they might permit testing of multiple putative treatments for disease in short term trials.

Using a rigorous quantitative assessment of BML volume and frequent knee MRI's, we sought to evaluate whether BML volumes in a knee changed within 6 or 12 weeks and how often this occurred.

Methods

Patients in this study were enrolled in an ongoing randomised trial of a patellofemoral (PF) brace for the treatment of patellofemoral osteoarthritis (PFOA). As such, all patients had to have evidence of knee pain on most days originating predominantly in the PF joint on physical examination and had to have evidence on imaging studies (X-ray or MRI) or arthroscopy of OA in the PF joint. For X-rays, we required a definite osteophyte and joint space narrowing in the PF joint on skyline or lateral knee radiographs. For MRI and arthroscopy, we required typical changes of OA with at least cartilage loss present in the PF joint. After baseline assessment, subjects underwent contrast enhanced (CE) baseline MRI and then obtained CE MRI's again using the same protocol and same magnet at 6 and 12 weeks. We obtained our MRI's with contrast because BMLs were not the only focus of the study. Subjects underwent a check of their estimated glomerular filtration rate (EGFR) to ensure adequate renal function (EGFR ≥45 cc/min) prior to consent, and the study was approved by Stockport Research Ethics Committee. We studied one knee per person.

MRI methods

Using a 1.5 T Philips MRI, we obtained axial Proton Density Weighted (PDW) Fat Saturated (FS) repetition time (TR) 1500 ms, echo time (TE) 15 ms, Field of View (FOV) 16 cm, 256 × 256 matrix, slice thickness 3 mm with 0.3 mm gap and T1 weighted sagittal postcontrast scans FS TR 500 ms TE 17 ms FOV 16 cm, 384 × 384 matrix slice thickness 3 mm with 0.3 mm gap in all subjects at baseline, 6 and 12 weeks. Contrast agent was Dotarem® at a dose of 0.2 mg/kg. Postcontrast images have been shownto provide similar assessments of BMLs as PDW FS non-CE images. Technicians underwent extensive training from an experienced musculoskeletal radiologist (CEH), in which they were taught how to differentiate BMLs from fat saturation failure, bone necrosis, haematopoietic marrow and other findings. During the process of segmentation, the radiologist reviewed any lesions where questions were raised about the nature and size of the lesion and, in addition, he reviewed the 10% of knees that showed the greatest change either at 6 or 12 weeks to ensure that these changes were, in reality, changes in BML volumes. We asked technicians at iMorphics to manually segment BML volumes in which BML areas on each MRI slice were outlined extending from the subchondral bone plate to the farthest depth of their extension into cancellous bone. Cystic changes within BMLs were included in the BML volumes. We used the axial PD and sagittal postcontrast images to carry out this work. For the axial images only, we focused on the patella and femur only which were consistently visualised on all axial images. For sagittal images, we segmented BMLs in patella, femur and tibia. The areas were then summed to compute BML volume. Sagittal and axial were segmented at different times so that all images from a knee were not analysed together. To optimise our ability to detect changes in BML volumes, segmentations were carried out paired – a specific knee's repeated MRIs were segmented before moving to the next knee. The segmenter was blinded both to the focus of this study and to any treatment. Each subject was randomised to immediate vs delayed treatment so that all subjects would have received treatment by 12 weeks. Study personnel were blinded to randomisation code. While we often confirmed BML changes by comparing findings in the two orthogonal planes, we focus here on findings from each plane separately, since the two did not necessarily reveal the same BML volumes and since the sagittal plane provided the only data on tibial BMLs. Our focus here is on BML volumes in each of the bones, not the volume of each BML, so that we do not report on BMLs which merged or split during follow-up.

To evaluate repeatability of BML volume segmentations, at a time at least 1 month after the initial work, the segmenter was asked to resegment knees, randomly selected from those that had been previously segmented with new ID numbers assigned (we resegmented 24 axial images and 16 sagittal ones).

Using the standard deviation of the difference between repeated measurements to calculate Bland–Altman Limits of Agreement, we determined how much BML volume change represented a change greater than due to chance. We labelled this volume change as the minimum detectable change (MDC), a threshold above which changes would exceed those expected using the Bland–Altman Limits of Agreement approach. This method has been previously described as a particularly conservative boundary for MDC limits. We did this separately for each of the axial and sagittal images. To determine the proportion with BML change greater than MDC, we divided all knees into those with any increase in BML volume vs those with any decrease in volume. To examine the relation of axial and sagittal BML volumes, we carried out a random-effects multiple regression model testing whether the axial volumes predicted sagittal volumes.

Results

A total of 62 subjects were studied (for characteristics of the sample, see Table I). Of these we segmented BMLs at three time points per person for axial view in all patients, and for sagittal postcontrast view in 48 patients. Of these 62, 10 had no BMLs on axial view at baseline. Of the remaining 52 images we rejected axial segmentations for three knees based on quality concerns, leaving 49 subjects in our axial data set. Of the 48 sagittals, three of them were in the group that had no BMLs on the axial view. In addition, we rejected eight segmentations based on quality control concerns, leaving 37 in the sagittal data set.

Table I

Descriptive statistics of the study sample

Statistic (recorded at baseline visit)	Whole sample (N = 62)	Axial sample (N = 59‡)	Sagittal postcontrast sample (N = 40‡)
Age (years), mean (SD)	55.12 (7.84)	54.93 (7.92)	54.25 (7.12)
BMI (kg/m2), mean (SD)	30.97 (5.80)	30.91 (5.80)	30.39 (5.98)
Females, frequency (%)	37 (59.68%)	35 (59.32%)	23 (57.50%)
KOOS pain subscale score∗ (0–100), median (IQR)	47.22 (38.89–58.33)	47.22 (36.11–58.33)	47.22 (36.11–58.33)
Pain on nominated activity VAS† (0–10), median (IQR)	7.20 (5.90–8.30)	7.20 (5.90–8.40)	6.80 (5.90–8.40)
Pain in last week VAS† (0–10), median (IQR)	6.40 (4.60–7.70)	6.30 (4.40–7.50)	6.20 (4.60–7.20)

Scoring for KOOS pain subscale is from 100 (no pain) to 0 (extreme pain).

Scoring for the VAS is from 0 (no pain) to 10 (extreme pain).

Represents the axial and sagittal samples after excluding knees whose segmentations were rejected for quality control reasons but includes knees without BMLs.

Test–retest reliability for the total BML volumes for the scans measured repeatedly showed ICC for axial views = 0.81 and ICC for sagittal views = 0.60. The Bland–Altman 95% limits of agreement for the axial images, assuming a zero-difference mean (no change in BML volume), were −733 mm3 to +733 mm3. For sagittal views, the 95% limits of agreement were −5669 mm3 to +5669 mm3 When we restricted sagittal BMLs to the femur and patella, the same bones as in the axial view, we found a strong relationship between axial and sagittal postcontrast BML volumes (r2 = 0.78, P <0.001), with sagittal volumes predicting slightly higher volumes of BMLs than the axial view.

Axial view

In the 49 knees with BMLs, median BML volume at baseline was 1058 mm3 (IQR 249 mm3–2789 mm3). All of these knees showed at least some change in BML volume by 6 and 12 weeks (see Fig. 1). An example of one knee showing shrinking BML volume on the axial view is shown in Fig. 2. Using the Bland–Altman Limits of Agreement as the threshold for minimal detectable change (MDC) on the axial view, 11 knees showed an increase in BML, and nine a decrease at 6 weeks. Thus, 20/49 knees (40.8%) showed change beyond the MDC at 6 weeks. When we looked at change from baseline to 12 weeks, we found eight knees that had an increase and 11 knees with a decrease beyond the Bland–Altman based MDC threshold (19/49, 38.8%). We note that most knees in our sample had small volumes of BMLs at baseline, with 24/49 (49.0%) having values less than 1000 mm3 at baseline. Reaching MDC using the Bland–Altman Limits of Agreement method would have meant that BML volume in these knees needed to change by almost 100%. To examine a relative (percent) change in BML volume, we evaluated the proportion of knees with more than 50% change in either direction (i.e., either positive or negative) in BML volume from baseline to either 6 and/or 12 weeks (Table II). Of the 49 knees with BMLs at baseline, only 36.7% (18 knees) did not show BML volume change by >50% at either 6 or 12 weeks.

Fig. 1

Change in BML volume on axial view, compared with Bland–Altman limits, at weeks 6 (left) and 12 week (right).

Fig. 2

Example of a patient with a patellar lesion changing over 12 weeks (denoted by white arrow) – axial view.

Table II

Proportion of knees with >50% change (either positive or negative) in total BML volume at 6 or 12 weeks

	Knees (frequency)	Knees (%)
Axial view
Change ≤50%	18	36.73
>50% Change at only 6 weeks	8	16.33
>50% Change at only 12 weeks	8	16.33
>50% Change at both 6 and 12 weeks	15	30.61
Total	49	100.00
Sagittal postcontrast view
Change ≤50%	19	51.35
>50% Change at only 6 weeks	9	24.32
>50% Change at only 12 weeks	4	10.81
>50% Change at both 6 and 12 weeks	5	13.51
Total	37	100.00

Sagittal view

When we examined change in BML volume using the sagittal postcontrast enhanced images, our results were similar although the Bland–Altman limits were considerably wider, meaning that substantial changes in volumes did not sometimes meet these greater thresholds for MDC. When we examined change in BML volume greater than these limits of agreement, we found that 3/37 (8.1%) had change at 6 weeks and that 6/37 had change at 12 weeks (16.2%). When we examined how many of these knees had greater than a 50% change in BML volume (Table II), we found that 19/37 (51.4%) did not show at least 50% change and that >50% change in BML volume occurred at either 6 or 12 weeks or both times in the remaining 48.6% (18/37). Examples of BML changes on the sagittal postcontrast view are shown in Figs. 3 and 4.

Discussion

In many knees with OA, the volume of BMLs changes in 6–12 weeks, and to the extent that these changes can be affected by treatment, this rapid change could be used to evaluate the course of disease or the effect of treatment.

One other group has also noted change in BMLs in less than a year, although not over such a brief time period as in our study. Garnero and colleagues using the WORMS semiquantitative scale reported that 28.6% of knees with OA showed change in BMLs at 3 months. Like ours, these were subjects in an OA clinical trial, although the time course of the MRI's in relation to treatment is unknown, MRI magnets varied from 0.2 T to 1.5 T, and individual subjects may have obtained their scans in different magnets at different times.

Other groups have documented changes in BML volumes over 1 year or more and, like our findings, published results suggest that enlargement of lesions is more common than shrinkage, although there was change seen in both directions.

Volume assessment of BMLs is challenging because they are poorly circumscribed lesions, making it difficult to define their borders. We carefully reviewed the segmentations done for this analysis and rejected several based on concerns that the volumes were incorrectly assessed because of these challenges. However, the volume changes in many of these lesions were unequivocal and were seen in both planes of segmentation and by our radiologist reviewing the images. It is likely that the changes we saw are large enough to be detected also by semiquantitative assessment of BMLs using approaches such as the WORMS scale or others. We have presented preliminary evidence showing a high cross sectional correlation of BML volumes and WORMS BML scores. If BML changes were detected sensitively with semiquantitative scoring, this could shorten image analysis time.

One potential limitation of our work is that we used different imaging approaches to assess BMLs, using a postcontrast sagittal image and a PDW axial one. While these two approaches have been reported to provide the same assessments of BMLs, the postcontrast image showed slightly greater volumes even when restricted to the same regions of the knee.

What changes can we expect on the basis of biology? Since BMLs show evidence of bone trauma on histology, the best parallel might be fracture healing which occurs over weeks. The most common pattern in our group was not healing, however, but rather enlargement of lesions, a phenomenon seen also in the other longitudinal studies in OA. We note that BMLs are strongly related to focal overloading in the joint, usually from malalignment. We are unsure whether the rapid change in volume of these lesions reflects fluctuation in the mechanical environment of the joint or rather evanescent pathology than has not heretofore been described, such as oedema or inflammation which might be expected to fluctuate more than joint mechanics.

Further work is needed to clarify the clinical meaning of these short term changes. For example, while longer term BML changes correlate with fluctuations in knee pain, it is not clear whether short term volume changes in BMLs have similar clinical relevance.

In conclusion, BMLs, lesions visible on MRI in the knee, commonly fluctuate in volume over 6–12 weeks. If these short term volume changes are influenced by treatment, our findings suggest that BMLs would be a convenient target for structure-modifying treatments.

Contributions

Conception and design: DTF, TC, JO, CED.

Acquisition of data: EJM, MC, AG.

Analysis and interpretation of the data: MJP, AG, TC, ML.

Drafting of article and reviewing for critical content: DTF, MJP, EJM, MC, AG, TC, ML, JO, CEH.

Final approval of submitted version: DTF, MJP, EJM, MC, AG, TC, ML, JO, CEH.

Obtaining of funding: DTF, JO.

Administrative, technical support: DTF, TC, ML, JO.

Role of funding source

The funding source (Arthritis Research UK) had no role in study design, collection, analysis or interpretation of data, in writing the manuscript or in submitting the manuscript.

Competing interests

No author has any relevant competing interests or conflicts of interest.