The expiratory flow retardation created this website by the distal end produces positive back pressure on the airway. The expiratory pressure induced by resistance of the conical-PEP is flow dependent; the greater the expiratory flow the greater the back pressure (Mitchell 2007, Weng 1984). It produces a positive mouth pressure of 4.2–10.9 cmH2O at expiratory flows of 0.06− 0.41 L/s at rest and 4–20 cmH2O at flow rates of 0.09–0.51 L/s during exercise. This pressure range has been reported to be optimal for retarding airway collapse in patients with chronic obstructive pulmonary disease (O’Donnell et al 1988, Petrof

et al 1990, Plant et al 2000). Exercise was terminated when one of the following symptoms occurred: breathlessness ≥ 5/10 on the modified Borg scale, leg discomfort, or any other unpleasant symptoms such as dizziness. The control intervention was normal breathing during exercise. Lung function was measured as inspiratory capacity and slow vital capacity in litres according to ATS/ERS taskforce guidelines (Miller et al 2005) with a portable automated spirometera. The volume sensor was calibrated before each test. The duration

of exercise and the reasons for exercise termination were collected. Breathlessness was measured using the modified Borg scale (0 to 10) where 0 is no breathlessness and leg discomfort was measured using a 0–10 visual analogue scale PLX4032 where 0 is no discomfort. Cardiorespiratory function was also measured. SpO2 was measured by finger pulse oximeter and end tidal pressure of carbon dioxide (PETCO2) was measured in a side-stream of crotamiton expired air with a capnometerb. Electrocardiogram, expiratory mouth pressure and expiratory flow rate were continuously recorded on a PC with an A/D converterc. The flow and pressure sensors were calibrated before each data collection. Tidal volume, respiratory rate, inspiratory time, expiratory time and ratio (I:E ratio) were determined from the flow signal. Minute ventilation was calculated for the last minute of exercise. A pilot study

of two elderly participants without lung disease showed a between-intervention difference of 150 ml (SD 130) for inspiratory capacity. Therefore, we needed 11 participants to have a 90% power to detect between intervention difference of 150 mL at p = 0.05. Student’s paired t tests showed no evidence of either period effects or intervention-period interaction of the primary outcome and, therefore, the data for the two tests in each intervention were averaged to provide a single value for each participant. Statistical significance was considered at p < 0.05, therefore mean between-intervention differences (95% CI) are presented. Forty-three patients with moderate-severe stages of chronic obstructive pulmonary disease were screened and 17 (40%) agreed to participate in the study. Of these, 4 (24%) withdrew prior to randomisation for reasons that were unrelated to the procedures of the study.