Pain and Radiographic Outcomes After Arthroscopic Treatment Versus Autologous Adipose-Derived Stromal Vascular Fraction Injection for Kellgren–Lawrence Grade 2–3 Knee Osteoarthritis: A Propensity Score–Matched Retrospective Cohort Study

Background and Objectives: Direct comparative evidence between arthroscopic treatment and autologous adipose-derived stromal vascular fraction (SVF) injection for moderate knee osteoarthritis (OA) is limited. This study compared pain and radiographic outcomes after these treatments in patients with Kellgren–Lawrence (K–L) grade 2–3 knee OA. Materials and Methods: This retrospective cohort study included 441 patients who underwent arthroscopic treatment or received intra-articular SVF injection. Propensity scores were estimated using 13 baseline covariates: age, sex, body mass index, baseline visual analog scale (VAS) score, symptom duration, coronal alignment, K–L grade, mechanical symptoms, meniscal tear, cartilage grade, effusion, synovitis, and loose body. We performed 1:1 optimal matching without replacement using a caliper of 0.20 standard deviations of the logit of the propensity score. The matched cohort comprised 203 patients per group. Longitudinal VAS scores through 12 months were assessed using generalized estimating equations. Radiographic progression and conversion to total knee arthroplasty (TKA) were analyzed with methods accounting for matched pairs. Results: All post-match covariates achieved an absolute standardized mean difference below 0.10. The treatment-by-time interaction for VAS score was not significant (p = 0.213). Mean VAS scores were slightly lower after SVF injection at 6 months (between-group difference, 0.22 points; 95% CI, 0.04–0.41; p = 0.018) and 12 months (0.26 points; 95% CI, 0.06–0.45; p = 0.009), where positive differences indicate higher scores after arthroscopic treatment. Radiographic progression occurred in 56 of 203 patients (27.6%) after arthroscopic treatment and 49 of 203 (24.1%) after SVF injection (paired risk difference, −3.4 percentage points; 95% CI, −12.3 to 5.4; exact McNemar p = 0.515). The adjusted odds of a higher 12-month K–L grade did not differ significantly between groups (common odds ratio for SVF versus arthroscopy, 0.85; 95% CI, 0.55–1.30; p = 0.451). TKA was performed in 25 patients (12.3%) after arthroscopic treatment and 13 (6.4%) after SVF injection (exact McNemar p = 0.058). Conclusions: After expanded propensity score matching, the overall 12-month pain trajectories and radiographic outcomes did not differ significantly between arthroscopic treatment and SVF injection. Although the SVF group had statistically lower mean VAS scores at 6 and 12 months, the absolute differences were small and unlikely to be clinically important. No statistically significant difference in TKA conversion was demonstrated. This exploratory result is strictly hypothesis-generating and requires prospective confirmation. These findings do not establish the superiority of either treatment. Because treatment selection was based on individualized surgeon–patient discussion and patient preference, unmeasured confounding may persist despite propensity score matching, precluding causal conclusions regarding comparative treatment effects.