STATIN USE AND DECREASED BONE MARROW LESION BURDEN: A LONGITUDINAL DEEP-LEARNING QUANTITATIVE ANALYSIS FROM OSTEOARTHRITIS INITIATIVE

Abstract

INTRODUCTION

Subchondral bone marrow lesions (BMLs) are a risk factor for knee OA outcomes and deep-learning (DL) methods can help in automated segmentation and risk prediction.

OBJECTIVE

To determine the association between statin use and longitudinal changes in knee MRI-detected BML volume.

METHODS

Using the Osteoarthritis Initiative (OAI) cohort, we classified participants’ knees into two categories: statin users (those who used statins from baseline to the fourth year of the cohort) and non-users. We employed a 1:1 ratio propensity score (PS) matching method, adjusting for factors including age, sex, race, BMI, smoking, alcohol use, physical activity, KL grade, abdominal obesity, diabetes mellitus, and cardiovascular diseases. We measured quantitative BML volume using a validated deep learning (DL) algorithm, applied to baseline, year-2, and year-4 intermediate-weighted knee MRIs. The outcome was determined by the differences in the 4-year BML volume change between statin users and non-users.

RESULTS

After adjusting for potential confounders, 3206 knees were included (1603 statin users:1603 non-user; 64.1 ± 8.5 years old, female/male ratio: 1.1). Multilevel linear mixed-effect regression model showed that statin use is associated a less degree of increase in BML volume over 4 years (time-treatment interaction estimate, 95% confidence interval (CI): -4.24 mm³/year, -7.26 to -1.22, P = 0.005).

CONCLUSION

Continues statin use is linked to a reduction in the worsening of BML, a known risk factor for the onset and progression of knee OA.