The time-effect relationship of intra-articular ozone injection for knee osteoarthritis: a systematic review and meta-analysis

Chen F, Tao Y, Deng J, Zhang L, Yu L, Yang Z, Zhang Y, Zhang C · Frontiers in Pain Research

16 randomized trials and 1,172 participants, pooled by follow-up time rather than overall, which is what makes it readable. WOMAC total fell at 1 month (SMD -0.63, 95 percent CI -1.17 to -0.10) and at 2 months WOMAC pain (SMD -1.39, 95 percent CI -1.96 to -0.81), WOMAC function (SMD -1.27, 95 percent CI -2.06 to -0.49) and VAS pain (SMD -1.25, 95 percent CI -2.10 to -0.40) all improved. VAS pain at 1 month did not reach significance (SMD -0.66, 95 percent CI -1.54 to 0.22). At 3 months no outcome differed. At 6 and 12 months the pooled estimates run the other way, with the ozone arm scoring worse than control on WOMAC total at 12 months (SMD 0.63, 95 percent CI 0.29 to 0.98) and on VAS at 12 months (SMD 1.74, 95 percent CI 0.73 to 2.75). The authors interpret that reversal themselves and do not read it as harm: the long-term controls were almost all hyaluronic acid or platelet-rich plasma, treatments with sustained effect, and when they removed active comparators the one placebo-controlled trial available at 6 months favoured ozone strongly (SMD -1.79, 95 percent CI -2.52 to -1.06). GRADE was moderate to low throughout.

Population

16 randomized controlled trials of intra-articular ozone in knee osteoarthritis, 1,172 participants, searched to 3 December 2025

Intervention

Intra-articular ozone injection

Comparator

Active controls including hyaluronic acid, and placebo where trials used one

Limitations

Heterogeneity exceeded I-squared 50 percent in every analysis, and while subgroup analysis resolved most of it for WOMAC it could not for VAS. Publication bias was not assessed at all: fewer than 10 studies contributed at every time point, so funnel plots and Egger's test were judged unusable, and the authors flag the absence as a limitation of their own work. They also name the small-study effect directly, at a mean of about 73 participants per trial, and warn that unpublished negative short-term trials would change the 1-to-2-month conclusion, which currently rests on placebo-controlled studies that all favoured ozone. No included trial measured imaging progression, biomarkers or long-term safety. At 12 months no placebo-controlled study existed for either outcome, and at 6 months only one did, so the long-term comparison is ozone against active treatment rather than against nothing. The authors declare no commercial or financial relationships.