KNEE B-SCORE SHAPE FROM COMPUTED TOMOGRAPHY IS ASSOCIATED WITH SUBCHONDRAL BONE ATTENUATION AND MARGINAL CORTICAL BONE THICKNESS

Abstract

INTRODUCTION

Change in shape of the distal femur demonstrated with MRI is an established biomarker for structural OA progression in clinical trials. This “B-score” has been ported across to CT with minimal bias, which brings the opportunity to include 3-D evaluation of periarticular bone distribution in shape analysis by combining statistical shape modelling (SSM) and cortical bone mapping (CBM).

OBJECTIVE

To look for significant relationships between 3-D knee shape and bone distribution with CT.

METHODS

This exploratory analysis was performed ancillary to the LOSEIT trial evaluating the efficacy of liraglutide in inducing and maintaining weight loss and pain relief in overweight patients with knee OA. After exclusions, 133 participants were included, 65 from the placebo group, 68 from the liraglutide group. All had baseline CT (140kV) and weight-bearing radiographs of both knees. Both knees were segmented from the CT data for CBM using Stradview followed by registration of canonical objects to femurs and tibias using wxRegSurf. SSM was performed on combined femur and tibia registrations using MATLAB 2024a. Index knee data were taken from each participant. Generalized estimating equation (GEE) analysis looked for associations of the first 10 shape modes with KLG controlling for age, sex and mass using Bonferroni correction. 3-D cortical thickness (CTh) and subcortical trabecular attenuation (TA) maps were transferred to the canonical objects. SPM analysis was performed using the MATLAB Surfstat toolbox to establish dependence of CTh and TA distribution on shape controlling for age, sex, mass and KLG.

RESULTS

Study participants were 89 females and 44 males with mean +/- SD age of 59.6 +/- 9.2 yrs, mass 93.3 +/- 16.7 kg and an index knee breakdown of KLG1 = 19, KLG2 = 57, KLG3 = 57. GEE showed shape mode 2 (SM2) was the only mode significantly associated with KLG with an odds ratio of 1.43 (1.28-1.59 95% CI, P<<0.05) for each SD of the mode (Fig. 1, * = P<0.05). Subjective visualization showed substantial similarities of SM2 to the B-score, namely increased femoral articular surface area with marginal articular prominence and narrowing of the intercondylar distance (Fig. 2, +/- 3xSD of the mode). SPM showed subchondral TA was significantly dependent on SM2 across nearly all the femoral articular surface (P<0.05), showing up to 40 HU drop for each increase in SD (Fig 2). Small zones of marginal articular bone at the lateral tibiofemoral compartment showed significant CTh dependence on the shape mode (P<0.05) with an increase of up to 0.2 mm for each SD increase (Fig. 2), but the association was limited to this compartment. In the tibia, this combined shape mode represented peaked widening of the tibial plateau rim, with significant dependence of TA in the posterior lateral tibial plateau (-20 HU per SD increase) and CTh around the medial plateau margin (+0.1 mm per SD increase).

CONCLUSION

Our shape mode 2 was a visual correlate of the B-score and significantly associated with KLG. 3-D analysis demonstrated significantly lower trabecular attenuation in femoral subchondral bone with increasing shape mode along with greater cortical thickness at the joint margins. This distribution suggests that bony remodeling of the articular surface with more advanced structural disease not only involves changes in bone shape, but also widespread subchondral trabecular density loss (as opposed to focal sclerosis) and increased bone thickness at joint margins consistent with osteophytosis. Combining these 3-D bone parameters with shape may therefore be of value in developing future predictive models.