Sixty-three RCTs with 29,423 patients were included in the review; sample sizes ranged from 20 to 1,399. Nineteen trials had adequate allocation concealment, but only five were unblinded. The risk of bias from selective reporting was generally low. Follow-up ranged from 24 weeks up to 156 weeks. Completion rates varied, ranging from 49% to 99% in those on active treatment and 5% to 100% in control groups.
A total of 211 patients (0.72%) developed a malignancy during the trial periods; 118 of these were solid tumours and 48 were skin cancer. The remainder were haematological or unspecified. Rates were very low in all groups: biologic agent plus methotrexate (OR 0.77%,95% CI 0.65% to 0.92%); biologic agent monotherapy (OR 0.64%, 95% CI 0.42% to 0.95%) and controls (OR 0.66%, 95% CI 0.52 to 0.84%).
There was no statistically significant difference in the total occurrence of malignancy between TNF biologic response modifiers alone versus placebo or traditional disease modifying therapy (OR 0.98, 95% CI 0.51 to 1.19; 13 RCTs) or TNF plus methotrexate versus either placebo or active comparators (OR 1.5, 95% CI 0.95 to 2.3; 29 RCTs). There were no significant differences for individual agents. There was a doubling of risk of lymphoma and a five-fold increase in other haematologic malignancies with TNF inhibitors compared to controls, but this was not statistically significant. Sensitivity analysis excluding trials that reported zero malignancy meant that a statistically significantly higher risk of malignancy was identified for TNF inhibitors plus methotrexate versus controls at 52 weeks (OR 2.0, 95% CI 1.1 to 3.8).
The only statistically significant finding from comparisons of total malignancy between non-TNF inhibitors with comparators was a reduced incidence at 24 weeks for anakinra plus methotrexate compared to methotrexate alone (OR 0.11, 95% CI 0.03 to 0.45, three RCTs). Pooled estimates for all non-TNF inhibitors were not presented.
There was no evidence of publication bias.