Aaron 2007.

Methods	Design: A randomised, double‐blind, placebo‐controlled, parallel group trial from October 2003 to January 2006. The trial included 27 Canadian medical centres; 20 centres were academic hospital‐based pulmonary clinics, 5 were community‐based pulmonary clinics, and 2 were community‐based primary care clinics.
Participants	Population: 293 adults with a clinical history of moderate or severe COPD as defined by ATS and GOLD guidelines. Baseline Characteristics: Mean age 68 years. COPD severity moderate to severe with mean FEV1 predicted of 39%. 58% men. Inclusion Criteria: At least 1 exacerbation of COPD that required treatment with systemic steroids or antibiotics within the 12 months before randomisation; age older than 35 years; a history of 10 pack‐years or more of cigarette smoking; documented chronic airflow obstruction, with an FEV1/FVC ratio < 0.70 and a post‐bronchodilator FEV1 < 65% of the predicted value. Exclusion Criteria: History of physician‐diagnosed asthma before 40 years of age; history of physician‐diagnosed chronic congestive heart failure with known persistent severe left ventricular dysfunction; those receiving oral prednisone; those with a known hypersensitivity or intolerance to tiotropium, salmeterol, or fluticasone‐salmeterol; history of severe glaucoma or severe urinary tract obstruction, previous lung transplantation or lung volume reduction surgery, or diffuse bilateral bronchiectasis; and those who were pregnant or were breastfeeding.
Interventions	1. Tiotropium + salmeterol + fluticasone: tiotropium (Spiriva, Handihaler (Boehringer Ingelheim Pharma, Ingelheim, Germany)), 18 g once daily, plus fluticasone‐salmeterol (Advair (GlaxoSmithKline)), 250/25 g/puff, 2 puffs twice daily 2. Tiotropium + salmeterol: tiotropium, 18 g once daily, plus salmeterol (Serevent (GlaxoSmithKline, Research Triangle Park, North Carolina)), 25 g/puff, 2 puffs twice daily
Outcomes	Primary: Proportion of patients with one or more exacerbation of COPD. Secondary: Mean number of COPD exacerbations per patient‐year; the total number of exacerbations that resulted in urgent visits to a healthcare provider or emergency department; the number of hospitalisations for COPD; the total number of hospitalisations for all causes; changes in health‐related quality of life, dyspnoea, lung function.
Notes	Co‐medication: All study patients were provided with inhaled albuterol and were instructed to use it when necessary to relieve symptoms. Any treatment with inhaled corticosteroids, long‐acting beta2‐agonists, and anticholinergics that the patient may have been using before entry was discontinued on entry into the study. Therapy with other respiratory medications, such as oxygen, antileukotrienes, and methylxanthines, was continued in all patient groups.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was done through central allocation of a randomisation schedule that was prepared from a computer‐generated random listing of the 3 treatment allocations, blocked in variable blocks of 9 or 12 and stratified by site.
Allocation concealment (selection bias)	Low risk	The metered‐dose inhalers containing placebo, salmeterol, and fluticasone‐salmeterol were identical in taste and appearance, and they were enclosed in identical tamper‐proof blinding devices. The medication canisters within the blinding devices were stripped of any identifying labelling.
Blinding (performance bias and detection bias) All outcomes	Low risk	Neither research staff nor patients were aware of the treatment assignment before or after randomisation.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The number of people who stopped drug therapy was high, with large variations between the groups (37/145 (26%) tiotropium + LABA/ICS and 64/148 (43%) tiotropium + LABA). However, the number of people who did not complete the trial was lower, although there was still some variations between the groups (15 (10%) tiotropium + LABA/ICS and 20 (14%) tiotropium + LABA). The issue of incomplete data was however addressed by sensitivity analyses of the data for the primary outcome (exacerbations). In the primary analysis it was assumed that all patients who were lost to follow‐up did not have an exacerbation (ITT analysis). In the sensitivity analyses it was assumed that patients who were lost to follow‐up either all had an exacerbation or had exacerbations at the same rate as those who continued in the study. Both analyses gave a similar result to the primary ITT analysis.
Selective reporting (reporting bias)	Low risk	Results for all listed primary and secondary outcomes were reported.