Forty-two randomised controlled trials (52,516 patients) were included in the review. All of the included trials reported blinding of patients and personnel. Randomisation methods and allocation concealment were both assessed as being adequately reported in 24 trials. Withdrawal rates (range from 17% in the tiotropium Soft Mist Inhaler group to 33% in ICS and placebo groups) were reported in 41 trials. Twenty-eight trials adequately reported loss to follow-up. Six trials described objective adjudication of cause of death.
Direct comparison meta-analysis: Compared with placebo, tiotropium Soft Mist Inhaler demonstrated a significantly increased risk of all-cause death (OR 1.49, 95% CI 1.05 to 2.11; three trials; Ι²=15.3%). Risk of all-cause death was significantly increased for tiotropium HandiHaler over LABA-ICS (OR 1.81, 95% CI 1.07 to 3.05; one trial).
Comparisons with placebo revealed a significantly increased risk of death from cardiovascular causes with tiotropium Soft Mist Inhaler (OR 1.96, 95% CI 1.07 to 3.60; two trials; Ι²=0%). Sensitivity analysis using the Mantel-Haenszel method did not differ substantially from the initial analysis (Peto method). No evidence of publication bias was found.
Mixed treatment comparison (fixed-effect) meta-analysis: Tiotropium Soft Mist Inhaler was associated with a universally increased risk of all-cause death compared with placebo (OR 1.51, 95% CrI 1.06 to 2.19), tiotropium HandiHaler (OR 1.65, 95% CrI 1.13 to 2.43), LABA (OR 1.63, 95% CrI 1.10 to 2.44) and LABA-ICS (OR 1.90, 95% CrI 1.28 to 2.86).
Risk of death from cardiovascular causes was significantly increased for tiotropium Soft Mist Inhaler compared with placebo (OR 2.07, 95% CrI 1.09 to 4.16), tiotropium HandiHaler (OR 2.38, 95% CrI 1.20 to 4.99), LABA (OR 3.04, 95% CrI 1.48 to 6.55), LABA-ICS (OR 2.79, 95% CrI 1.37 to 6.02) and ICS (OR 2.39, 95% CrI 1.18 to 5.12). No moderate or substantial between-study heterogeneity was shown.
In the random-effects model, tiotropium Soft Mist Inhaler was consistently shown to have an increased risk of all-cause death compared with any comparator and demonstrated an increased risk of death from cardiovascular causes in comparison with LABA-ICS.
Further results were reported fully in the paper.