Background/Purpose:

Bone marrow lesions (BML) relate to pain and prognosis in knee osteoarthritis.  The Knee Inflammation MRI Scoring System (KIMRISS) offers simple template-assisted binary BML scoring via a web-based interface. We tested whether KIMRISS was more reliable and responsive to change than the commonly used MRI Osteoarthritis Knee Score (MOAKS) system, particularly for naïve readers.  

Methods:

The KIMRISS web-based overlay system separates the knee into multiple small regions, each scored 0 (no BML) or 1 (BML present).  The reader scrolls through sagittal fluid-sensitive MRI slices and mouse-clicks (or contacts a touch-screen) each region in which BML is present.  Scores are tabulated and exported automatically.  In contrast, MOAKS BML grading is performed in 9 larger knee regions based on the number of lesions present and percentages of the region containing edema and cystic change.

Exercise 1 assessed reliability, in observational data from the Osteoarthritis Initiative.  We had 40 subjects at baseline and 1 year follow-up (80 MRI).  Expert readings were performed by two radiologists (6 and 11 years experience) familiar with KIMRISS, and consensus score computed as the average of these readings. Two radiology residents naïve to KIMRISS and MOAKS reviewed a 55-slide presentation describing KIMRISS, three scored reference cases, and a published manuscript describing MOAKS.  These readers then scored the same data by KIMRISS and MOAKS.  All readers were blinded to scan time point (baseline vs. follow-up). 

Exercise 2 assessed responsiveness, using data from an open-label pilot study testing adalimumab (Humira) knee OA therapy. Two blinded experienced readers scored BML by MOAKS and KIMRISS at enrolment and 12 weeks post-therapy in all 16 patients (32 MRI).   

Reliability statistics were calculated using intra-class correlation coefficients (ICC), with differences significant at p<0.05 when the 95% confidence intervals of ICC did not overlap.  To assess responsiveness in Exercise 2, we computed standardized response means (SRM).

Results:

Subjects were primarily older adults (average 62 and 58 years for Exercises 1 and 2 respectively), and female (80% and 65%), with moderate OA (K-L grade 3 or 4 in 73% and 75%). 

In Exercise 1, reliability of KIMRISS and MOAKS was high and similar for baseline BML (ICC 0.89-0.98 vs 0.85-0.89).  However, changes in BML were significantly more reliably detected by KIMRISS than MOAKS, whether comparing expert readings (ICC 0.82 vs. 0.53) or comparing naïve readings to expert consensus (ICC 0.87-0.92 KIMRISS vs. 0.32-0.51 MOAKS). 

In Exercise 2, decrease in BML was visually obvious post treatment in most patients.  This decrease in BML was statistically significant when scored by KIMRISS (mean change from 37.3 to 29.8, SRM=-0.69, p=0.015), but not by MOAKS (6.25 to 6.06, SRM=-0.12, p=0.63). 

Conclusion:

Comparing two semi-quantitative MRI BML scoring systems, KIMRISS was more responsive to change in BML than MOAKS, and detected change with better reliability than MOAKS for expert and naïve readers.  KIMRISS scoring may be useful in observational and therapeutic trials for knee OA.

« Back to 2015 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/can-bone-marrow-lesions-be-scored-more-reliably-and-responsively-in-knee-osteoarthritis-using-a-web-based-overlay-system-kimriss-than-by-standard-mri-osteoarthritis-knee-scoring-moaks-data/