Singh et al's meta-analysis of 13 randomized trials of pioglitazone or rosiglitazone lasting for at least one year. There were 130 pneumonias among the 8,163 treated patients, and 100 among the 9,464 controls, for a relative risk of 1.40 for the thiazolidinedione group (CI 1.1-1.8, p=0.01). Pioglitazone showed more of a signal with a significant RR of 1.63 (while the rosiglitazone group's confidence interval crossed 1.0). Authors estimate that treating 55 diabetic patients with a TZD for 3.7 years would cause one additional pneumonia. Thorax 2011;66:383-388.