Table 3. Longitudinal comparison of subchondral BML worsening between human COX2 inhibitor users vs. non-selective NSAID users with MetS-OA.

Subchondral BML Worsening (MOAKS)	COX2 inhibitor users vs. non-selective NSAID users with MetS-OA
	Odds ratio (95% Confidence Interval), p-value, Sample size	Number of events in each group[COX2 inhibitor: NSAID]*
Worsening in number of affected subregions with BML	0.35 (0.13–0.93), p:0.035, N: 88 (44:44)	Improvement, [3:2] No change, [14:9] Worsening, [27:33]
Improvement in number of affected subregions	1.21 (0.19–7.72), p:0.839, N: 88 (44:44)	Yes, [3:2]
Maximum worsening in BML score	0.45 (0.2–0.99), p:0.046, N: 88 (44:44)	No change, [18:9] Worsening by ≤1 grade, [22:27] by ≥2 grades, 4:8
Improvement in the subregions' BML score	1.28 (0.33–5.00), p:0.722, N: 88 (44:44)	Yes, [9:6]

Among the PS-matched COX2 inhibitor users and non-selective NSAID users with MetS-OA, participants with available baseline and 24-month follow-up MRIs were included. A musculoskeletal radiologist read and scored MRIs according to validated MOAKS measures of subchondral BML worsening. Logistic mixed-effect regression models were used for subchondral BML assessments. All analyses were adjusted for the baseline Kellgren-Lawrence (KL) and Osteoarthritis Research Society International medial joint space narrowing (OARSI JSN) grades of knees. Subchondral BML worsening was assessed using standard MOAKS measures. N corresponds to the total number of knees included in each analysis and the number of matched knees of MetS-OA and PTOA participants in the parenthesis.

^* Number of events for each outcome has been shown separately in the brackets for COX2 inhibitor users and non-selective NSAID users.

BML: Bone marrow lesion, MetS: metabolic syndrome, MOAKS: MRI Osteoarthritis Knee Score, NSAID: Non-Steroidal Anti-Inflammatory Drug, OA: osteoarthritis.