Digital support for exercise improves physical function in knee osteoarthritis patientsDifferent Exercise Styles Show Varying Results for Knee Osteoarthritis

Previous reviews have commonly pooled heterogeneous digital technologies, leaving the relative effects of different delivery approaches, feedback mechanisms, and levels of professional involvement unclear. This study compared seven exercise and self-management strategies for WOMAC pain, WOMAC physical function, and retention at the end of the intervention in people with knee osteoarthritis. Seven databases were searched through July 2026 for randomised controlled trials involving adults with knee osteoarthritis. Interventions were classified into seven nodes according to exercise delivery, feedback mechanisms, and professional involvement. Frequentist random-effects network meta-analyses were conducted, with WOMAC outcomes expressed as Hedges' g and retention as risk ratios. Risk of bias and certainty of evidence were assessed using RoB 2 and GRADE. Twenty-one RCTs involving 2,250 participants were included. The pain, physical function, and retention networks comprised 8, 12, and 20 trials, respectively. For physical function, D1 vs. C1 (g = 0.47, 95% CI 0.31 to 0.64) and D2 vs. C2 (g = 0.47, 95% CI 0.22 to 0.72) showed improvements, with moderate-certainty evidence. The estimate for D3 vs. C3 was imprecise (g = 0.48, 95% CI −0.13 to 1.10; low-certainty evidence). All three pain comparisons favoured the digitally supported nodes (g = 0.15, 0.21, and 0.50), but all 95% CIs included the null value, with low-certainty evidence. Retention differences for D2 vs. C2 and D3 vs. C3 were likely to be small, whereas local inconsistency was identified for D1 vs. C1. Comparison-specific network estimates favoured D1 over C1 and D2 over C2 for patient-reported physical function, but these contrasts used different comparators and do not establish superiority of digital delivery itself or the independent value of professional support. The incremental value of synchronous or hybrid professional support over face-to-face supervised exercise remains uncertain. Incremental effects on WOMAC pain remain unclear, and retention estimates were generally close to the null, with local inconsistency for D1 vs. C1. Current evidence does not support prioritising synchronous or hybrid professional support solely on the basis of rankings or assuming a linear relationship between digital-support intensity and treatment effects. www.crd.york.ac.uk/prospero, identifier CRD420261467021.