A total of 37 blinded RCTs (7,661 patients) were included in the review. The study sample size ranged from 10 to 1,575 patients. The mean length of follow-up was three weeks.
Diclofenac versus non-active comparator: Compared with non-active comparator, topical diclofenac was associated with a statistically significantly greater rate of: all adverse events (RR 1.11, 95% CI 1.02 to 1.20; 20 RCTs); withdrawals due to all adverse events (RR 1.65, 95% CI 1.10 to 2.48; 19 RCTs); local adverse events (RR 1.40, 95% CI 1.15 to 1.70; 19 RCTs); and withdrawals due to local adverse events (RR 2.37, 95% CI 1.22 to 4.63; 18 RCTs). There was a marginally statistically significant difference in physician rated tolerance favouring topical diclofenac (RR 1.04, 95% CI 1.00 to 1.09; four RCTs) but there was no significant difference in patient rated tolerance.
Diclofenac gel versus active comparator: Compared with active topical comparator, topical diclofenac gel was associated with a statistically significantly lower rate of all adverse events (RR 0.53, 95% CI 0.32 to 0.89; 11 RCTs), but there was no difference in terms of withdrawals due to all adverse events, local adverse events, withdrawals due to local adverse events, physician rated or patient rated tolerance. Compared with active oral comparator, topical diclofenac was associated with statistically significantly more local adverse events (RR 8.38, 95% CI 5.08 to 13.85; 3 RCTs) and withdrawals due to local adverse events (RR 31.0, 95% CI 4.25 to 225.80; 2 RCTs), but there was no statistically significant difference in all adverse events or withdrawals due to all adverse events.
There was no evidence of statistical heterogeneity in the analyses, apart from withdrawals due to all adverse events in the diclofenac versus active oral comparator (Ι²=74%). Subgroup analysis indicated a difference in the rates of adverse events depending on the formulation of diclofenac (highest with solution) but meta-regression did not support this finding. Meta-regression indicated that older age was a predictor of local adverse events with topical diclofenac, and studies with longer follow-up were also associated with local adverse events. There was no evidence of publication bias.