Author Affiliations: Department of Dermatology, Baylor Research Institute, Dallas, Texas (Drs Ryan and Menter) (firstname.lastname@example.org); and Institute of Health Care Research and Improvement, Dallas (Mr Daoud).
In Reply: Dr Tzellos and colleagues raise several issues that we found challenging to address when conducting the statistical analyses for our study. The limited power of our meta-analysis to conclusively determine an association between anti–interleukin (IL)-12/23 agents and cardiovascular risk was a main focus of our discussion. We would have chosen to perform a more statistically robust time-to-event analysis but were not granted access to patient-level data. The next best option was to perform a meta-analysis of publicly available summarized data. The meta-analysis was conducted using the procedures suggested by the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines.1 The studies were deemed homogenous as identified by the I2 statistic (heterogeneity: χ28 = 1.79, P = .99; I2 = 0%). It has been suggested that the “incorporation of heterogeneity into an estimate of a treatment effect should be a secondary consideration when attempting to produce estimates of effects from sparse data—the primary concern is to discern whether there is any signal of an effect in the data.”2 Thus a fixed-effects method was chosen. We chose to use the Mantel-Haenszel fixed-effects method of the risk difference over the Peto odds ratio method because 16 of 22 studies had zero events. Use of the Peto odds ratio method would have necessitated the removal of these 16 studies from the analysis, resulting in a serious selection bias.
Ryan C, Daoud Y, Menter A. Biologic Therapies for Chronic Plaque Psoriasis and Cardiovascular Events—Reply. JAMA. 2011;306(19):2095-2096. doi:10.1001/jama.2011.1661