Abstract
BACKGROUND: Whether baseline self‑reported physical activity (PA) is associated with the risk of primary knee replacement (KR) for osteoarthritis (OA) remains uncertain, particularly across radiographic severity. METHODS: In 2,534 Osteoarthritis Initiative (OAI) participants with radiographic knee OA (RKOA; Kellgren and Lawrence (KL) ≥ 2) who completed the Physical Activity Scale for the Elderly (PASE) at baseline, we examined the association between PASE tertiles (low, moderate, high) and OA‑related primary KR over 108 months. We modeled OA‑KR as the event of interest and non‑OA primary KR and pre‑KR death as competing risks. We fit both cause‑specific and subdistribution hazard models, adjusted for demographics, BMI, race/education, economic status (income categories), WOMAC pain, knee alignment, bilateral involvement, and prior knee injury/surgery. Because KL grade violated the proportional hazards assumption and interacted with PASE (P for interaction < 0.001), analyses were stratified by KL grade (2 vs. > 2). PASE was self‑reported and cross‑sectional (baseline only). RESULTS: Over nine years, 372 participants underwent primary OA‑KR. The effect of physical activity on OA-KR risk was strongly modified by baseline radiographic severity. In subdistribution hazard models, participants with KL grade 2 and moderate physical activity had a substantially lower risk of OA-KR compared to those with low activity (adjusted HR 0.38, 95% CI 0.20-0.73). In contrast, among those with KL grade > 2, moderate activity was associated with a higher risk (adjusted HR 1.46, 95% CI 1.08-1.97). In both strata, high levels of activity conferred no evidence of benefit or harm (KL grade 2: HR 0.86, 95%CI 0.49-1.50; KL grade > 2: HR 1.33, 95%CI 0.94-1.88). Cause-specific models yielded similar results. CONCLUSION: The association between baseline self‑reported PA and OA‑KR depends on radiographic severity. For individuals with KL grade 2, moderate PA was associated with lower KR risk; for KL grade > 2, moderate PA was associated with higher KR risk. Clinically, PA advice may be framed by severity: encourage moderate-intensity, low-impact activity and strength/neuromuscular training in earlier disease, and emphasize load management (low-impact modes, pacing, symptom monitoring) and weight management in advanced disease within shared decision-making. Because PASE is self‑reported and measured once, causal inference is limited; longitudinal and wearable‑based assessments are needed.