One hundred and twenty-nine RCTs were included (n at least 51,942, exact number not reported or calculable). Sixty-four studies were identified from databases, while an additional 65 RCTs were identified from the FDA archives.
Sixty (96%) of the included studies identified in bibliographic databases were double-blind and 55 (86%) accounted for all randomised patients. The quality of the trials identified from the FDA website was not reported.
ALT/AST more than three times the upper limit of normal: this occurred in 0.29% (95% CI: 0.17, 0.51) of patients taking placebo. There was overlap between the 95% CIs for celecoxib, valdecoxib, meloxicam and naproxen and placebo. The highest rate of elevation was with diclofenac (3.55%, 95% CI: 3.12, 4.03), followed by rofecoxib (1.80%, 95% CI: 1.52, 2.13). There was no overlap between either diclofenac or rofecoxib and other NSAIDs or placebo.
Discontinuation due to hepatic-related adverse events: this occurred in 0.08% (95% CI: 0.02, 0.29) of patients taking placebo. Diclofenac was the only NSAID with higher non-overlapping 95% CI compared with placebo; the rate was 2.17% (95% CI: 1.78, 2.64).
Hepatic-related serious adverse events: these occurred very rarely. The highest rates were seen with naproxen (0.06%, 95% CI: 0.02, 0.15), followed by rofecoxib and diclofenac. All of these 95% CIs overlapped with placebo.
Hepatic-related hospitalisation, death and liver transplant: only 1 of 37,671 patients was hospitalised (with naproxen), giving an overall rate for all NSAIDs of 2.7 per 100,000 patients (95% CI: 0.5, 15). Only 1 of 51,942 patients had a liver-related death, giving an overall rate for all NSAIDs of 1.9 per 100,000 patients (95% CI: 0.3, 11). No liver transplants were reported. There was overlap amongst all treatment groups in the 95% CIs for liver-related hospitalisations and deaths.
The results were similar after analysing the data by type of arthritis. The studies showed a trend towards higher rates of aminotransferase elevations and liver-related discontinuations with longer term use of diclofenac compared with shorter term use, and for higher doses of diclofenac compared with lower doses (the data were reported).