Intra-articular corticosteroid for knee osteoarthritis

Jüni P, Hari R, Rutjes AW, et al. · The Cochrane Database of Systematic Reviews

27 trials and 1,767 participants, corticosteroid against sham injection or no treatment. Pain: SMD -0.40 (95 percent CI -0.58 to -0.22), which the reviewers back-translate to 1.0 cm on a 10-cm visual analogue scale and a number needed to treat of 8 (95 percent CI 6 to 13), with I-squared 68 percent. Stratified by follow-up the benefit decays on a visible curve: SMD -0.48 at 1 to 2 weeks, -0.41 at 4 to 6 weeks, -0.22 at 13 weeks (95 percent CI -0.44 to 0.00), and -0.07 at 26 weeks (95 percent CI -0.25 to 0.11). Function: SMD -0.33 (95 percent CI -0.56 to -0.09), number needed to treat 10, with no evidence of an effect by 13 weeks. Quality of life showed no effect (SMD -0.01, 95 percent CI -0.30 to 0.28), and neither did joint space narrowing (SMD -0.02, 95 percent CI -0.49 to 0.46) — the one structural endpoint in the review, and it is flat. Adverse events, withdrawals for adverse events and serious adverse events all favoured corticosteroid numerically, with every confidence interval crossing the null.

Population

27 trials, 1,767 participants with knee OA

Intervention

Intra-articular corticosteroid injection

Comparator

Sham injection or no treatment

Limitations

Cochrane graded the quality of evidence low for every outcome, for inconsistent effect estimates, imprecise pooled estimates and high or unclear risk of bias in most trials. The review finds small-study effects inside its own analysis: treatment effects were lower in trials randomising at least 50 participants per group (P = 0.05 for pain, P = 0.023 for function), lower again at 100 per group (P = 0.013), lower in unpublished trials (P = 0.023), and funnel plot inspection suggested asymmetry for pain (coefficient -1.21) and more strongly for function (coefficient -4.07, 95 percent CI -8.08 to -0.05). Heterogeneity ran at I-squared 62 to 69 percent up to 13 weeks. The authors conclude that whether there is a clinically important benefit at one to six weeks remains unclear, and record that the single trial describing adequate measures to minimise bias found no benefit.