Introduction Effects of interventions on patient-reported outcomes may be subjective and modulated by patients’ expectations regarding treatment efficacy. The ‘gold standard’ for minimising such biases are double-blind randomised controlled trials. We analysed the effects of tiotropium on health-related quality of life in chronic obstructive pulmonary disease (COPD) in placebo-controlled trials and assessed whether trial design (double-blind versus open-label) is a relevant modifier of the effects of tiotropium.

Methods Trials of ≥6 months’ duration investigating the effect of tiotropium versus placebo on health-related quality of life in COPD (assessed using St George’s Respiratory Questionnaire [SGRQ]) were identified from the Boehringer Ingelheim clinical trial database and by a systematic literature search in MEDLINE, with a cut-off date of 30 November 2011. As a clinical end point, the mean difference between treatment groups in SGRQ total score was assessed. Trials were grouped according to double-blind or open-label design. We performed a network meta-analysis including standard methodology to test for interaction to evaluate whether trial design is a potential modifier of effect size or its direction.

Results We identified 12 trials in which tiotropium had been administered double-blind and three trials with open-label application. The overall effect for mean difference versus placebo in SGRQ total score was -2.98 units (95% confidence interval [CI): -3.49, -2.47). For the double-blind trial subgroup, mean difference versus placebo was -3.20 (95% CI: -3.75, -2.65) compared to -1.67 (95% CI: -3.02, - 0.32) for open-label trials. The p-value for interaction between subgroup and effect on SGRQ total score was 0.04.

Conclusions In patients with COPD, trial design (double-blind versus open-label) was a statistically significant modifier of the effect of inhaled tiotropium on health-related quality of life. The modification was quantitative, resulting in a substantial underestimation of the effect of tiotropium on SGRQ total score when the administration had been open-label compared to the ‘gold standard’ double-blind. A subjective end point such as quality of life is particularly susceptible to bias due to patients’ expectations towards the efficacy of an intervention. Therefore, the validity of studies using non-blinded designs to investigate such end points must be questioned.