Session Type: Poster Session C
Session Time: 9:00AM-11:00AM
Background/Purpose: To develop a novel, fine-grained MRI osteoarthritis (OA) severity score, based on cartilage, osteophytes and meniscus (OA-COM) scores and to predict the progression to different OA severity states using OA-COM scores as the outcome and clinical variables as predictors in a population-based 7-year longitudinal cohort study.
Methods: A population-based cohort, age 40-79, was assessed at baseline, 3- and 7-year follow-up using clinical assessments, x-ray and MRI. MRI OA severity score was based on the sum of scores for cartilage, osteophytes and menisci, measured semi-quantitatively at 6, 8 and 6 sites, respectively, using 0-3 grading for cartilage and osteophytes and 0-2 grading for menisci, for a total score ranging from 0-54, where 0 = no MRI OA and 54 = severe MRI OA. In order to anchor the OA-COM score at interpretable points representing different levels of disease severity, we fit logistic regression models using baseline OA-COM score to predict each baseline Kellgren Lawrence (KL) grade in a subset of data including only KL grades at or one point below the predicted grade. For each model, we produced ROC curves and for each cut point we computed sensitivity, specificity, positive and negative predictive values and sum of squares (SS) of those statistics. The optimal threshold for OA-COM scores in predicting a given KL grade was selected using the cut points with the highest SS, adjusted to attain equal spacing between cut points. To predict progression to different OA severity states, we developed logistic regression models for progression at or above the given OA-COM score cut point from baseline to 7-year follow-up, using the subset of data of patients who could progress; that is, with OA-COM score below the given cut point at baseline. Predictors in multivariable models were selected using forward selection based on best AIC at each step. In order to obtain results that were population-representative, sample weights were used for all analyses.
Results: Of 122 subjects, at baseline, 39.6% had radiographic OA (KL grade ≥2), mean age 55.5, female sex 55.7%. OA-COM score cut points were 12, 18, 24 and 30 for KL grades 1 to 4, respectively. Clinical predictor variables and odds ratios for each OA-COM cut point are shown in the Table 1. BMI was predictive of progression to all OA-COM scores over 7 years. Effusion was a strong predictor for progression to an OA-COM score of 12, 18 and 24, while malalignment and coarse crepitus predicted progression to higher OA-COM scores of 24 and 30.
Conclusion: We have developed a novel MRI OA severity score, the OA-COM score, using cartilage, osteophytes and menisci. OA-COM scores of 12, 18, 24 and 30 represent an MRI score that is equivalent to KL grades 1 to 4, while offering fine-grain differentiation of OA states between KL grades. In prediction modeling, we found that effusion, malalignment and crepitus, as well as age, sex and BMI, were predictive of progression to different states of OA severity over 7 years in this population-based cohort.
To cite this abstract in AMA style:Cibere J, Guermazi A, Nicolaou S, Kopec J, Esdaile J, Wong H, Singer J, Thorne A, Sayre E. Prediction Model for Progression to Different Find-Grained MRI-Based Osteoarthritis Severity States: The Vancouver Longitudinal Study of Early Knee Osteoarthritis (VALSEKO) [abstract]. Arthritis Rheumatol. 2020; 72 (suppl 10). https://acrabstracts.org/abstract/prediction-model-for-progression-to-different-find-grained-mri-based-osteoarthritis-severity-states-the-vancouver-longitudinal-study-of-early-knee-osteoarthritis-valseko/. Accessed January 25, 2022.
« Back to ACR Convergence 2020
ACR Meeting Abstracts - https://acrabstracts.org/abstract/prediction-model-for-progression-to-different-find-grained-mri-based-osteoarthritis-severity-states-the-vancouver-longitudinal-study-of-early-knee-osteoarthritis-valseko/