Ozone injections reduce pain in knee osteoarthritis: a systematic review and meta-analysis

Arias-Vázquez PI, Castillo-Avila RG, Hernández-Gil K, Guzzardo MN, Guzzardo DR · Medical Gas Research

7 randomized trials and 409 patients, ozone against intra-articular corticosteroid. Pain favoured ozone at every horizon: short term SMD -0.30 (95 percent CI -0.53 to -0.08, P = 0.009, I-squared 0 percent), medium term SMD -0.93 (95 percent CI -1.57 to -0.29, P = 0.005, I-squared 83 percent), long term SMD -1.05 (95 percent CI -2.03 to -0.07, P = 0.04, I-squared 87 percent). Removing the one high-risk trial with the largest effect brought the medium-term estimate down to SMD -0.61 (95 percent CI -1.02 to -0.19, I-squared 48 percent) and it stayed significant. Function is where the picture changes: no difference in the short term (SMD -0.02, 95 percent CI -0.40 to 0.37, P = 0.93) and none in the long term (SMD -0.87, 95 percent CI -2.15 to 0.40, P = 0.18), with a medium-term advantage for ozone (SMD -0.93, 95 percent CI -1.70 to -0.16, P = 0.02) that survived sensitivity analysis at SMD -0.52. The authors grade short-term pain as moderate evidence and medium-term function as low.

Population

7 controlled clinical trials comparing intra-articular ozone with intra-articular corticosteroid in knee osteoarthritis, 409 participants

Intervention

Intra-articular ozone injection

Comparator

Intra-articular corticosteroid injection

Limitations

No included trial was at low risk of bias: on the Cochrane RoB-1.0 scale four were unclear and three were high, with the problems concentrated in selection, performance and detection bias. Heterogeneity was severe wherever the effect was largest, at I-squared 83 to 92 percent for every medium- and long-term estimate, and only the short-term pain analysis was homogeneous. The trials pooled here differ substantially in what they injected: ozone concentrations run from 10 to 35 micrograms per millilitre and the corticosteroid arms use methylprednisolone, triamcinolone or betamethasone at different doses and schedules. Only three trials contributed long-term data. The comparison is against a drug rather than against placebo, so a pooled advantage over corticosteroid is not evidence of an effect over no injection at all. No structural or imaging endpoint appears anywhere in the review. The authors declare no conflicts of interest.