Nine RCTs (n=3,316) were included in the review (plus 80 individuals excluded from analysis). Data for analysis comprised 2,244 patients who received etanercept (2,484 person-years of follow-up) and 1,072 who received control therapies (1,051 person-years of follow-up).
Twenty six patients treated with etanercept developed malignancies (incidence rate of 10.47 per 1,000 person-years). Seven patients given control treatments developed malignancies (incidence rate of 6.66 per 1,000 person-years). The proportional hazards model found no statistically significant difference between the patient groups (HR 1.84, 95% CI 0.79 to 4.28). The aggregate analysis also did not show a statistically significant effect of treatment (OR 1.93, 95% CI 0.85 to 4.38). There was no evidence of statistically significant heterogeneity between studies (I2=0.0%); use of random effects models did not materially affect the results of the analysis.
None of the other sensitivity or stratified analyses revealed significant effects of the variables investigated. A power calculation showed that to detect a doubling of risk with 80% power and a 5% significance level a minimum of 9,305 patients would be required; the probability of the present meta-analysis detecting such a difference was 39%.