Background Radiography is only able to depict structural joint damage at advanced stages of knee osteoarthritis. Pre-radiographic structural damage is likely to increase the risk of incident radiographic OA (ROA). Subchondral bone marrow lesions (BMLs) expose the joint to increased risk of cartilage loss locally but it is unknown if BMLs increase risk for incident ROA.

Objectives The aim of the study was to assess if presence and severity of subchondral BMLs one year prior to the occurrence of incident ROA (time point “P-1”) and at study enrollment (“baseline”) increase the risk for incident ROA in a nested, matched case-control study in the Osteoarthritis Initiative (OAI) cohort.

Methods Participants were drawn from the OAI including 4796 participants with, or at risk of knee osteoarthritis. We studied 110 knees that developed incident ROA before the 48 month visit based on the following definition: either KL 0 in both knees or KL 0 in one knee and KL 1 in the contralateral knee at baseline. They were each matched with a control knee that did not develop incident ROA, with the same KL grade in both knees at baseline. Matching was further done by gender and age within 5 years. MR images were acquired at four OAI clinical centers using Siemens Trio 3 T scanners. MRIs were read for subchondral BMLs in 14 articular subregions using the semiquantitative MOAKS system. Only BML size, which is scored from 0-3, was considered in the analyses. Analyses were performed on a knee level only considering all 14 articular subregions.

Conditional logistic regression was applied to assess the risk of incident ROA in regard to maximum BML size per knee at P-1 and baseline. In addition, the number of subregions per knee exhibiting BMLs and the summed BML score per knee was analyzed in regard to risk for ROA for the time point P-1 and for baseline.

Results Subjects were on average 58.6 years old (SD ± 8.5), predominantly female (62.8%) and overweight (mean BMI 28.0 SD ± 4.7).

At P-1 risk of incident ROA was marginally increased for knees exhibiting BMLs with a maximum grade of 1 (odds ratio (OR) 1.83 95% confidence intervals [0.94,3.52]) and was significantly increased for BMLs with a maximum grade of ≥2 (OR 3.38 [1.64, 6.98]). Only knees with a maximum grade of ≥2 at baseline showed an increased risk of subsequent ROA (OR 2.18 [1.09,4.38]).

Knees with ≥1 subregions affected by BMLs at P-1 had an increased risk of ROA (for 1 subregion OR 2.83 [1.43, 5.59], for ≥2 subregions OR 2.06 [1.05, 4.02]), while this was not the case for the baseline visit.

A summed BML score of ≥4 increased risk of ROA at P-1 (OR 3.80 [1.37, 10.51] and baseline OR 2.86 [1.01, 8.08]).

Conclusions Presence of any BMLs predicted incident ROA one year later compared with knees without BMLs. Risk for ROA was further increased for knees exhibiting large BMLs and a sum score of 4 or more. The subchodnral bone seems to play an importnant role in OA incidence.

Of note at baseline, only knees with a maximum score of ≥2 exhibited an increased risk of ROA. BMLs seem to be an important predictor of ROA both at baseline and at one year prior to the development of ROA.

Disclosure of Interest F. Roemer Shareholder of: Boston Imaging Core Lab, LLC, Consultant for: Merck Serono, NIH, C. K. Kwoh Consultant for: Novartis, M. Hannon: None Declared, R. Boudreau: None Declared, D. Hunter Grant/research support from: Australia Research Council Future Fellowship, Consultant for: DonJoy, NIH, Stryker, F. Eckstein Shareholder of: Chondrometrics, Z. Wang: None Declared, J. Grago: None Declared, A. Guermazi Shareholder of: Boston Imaging Core Lab, LLC, Consultant for: Astra Zeneca, Genzyme, Novartis, Stryker, Merck Serono