The FVC values are only reported for adults because children often failed to complete the FVC maneuver correctly. Spirometry was performed (MicroLoop II; Micro Medical; Rochester, UK). The postbronchodilator FEV1 was measured at randomization, after 6 and 12 weeks, and then quarterly 30 min after inhaling terbutaline at 0.5 mg from dry powder inhaler, or 1 mg via pressurized metered-dose inhaler (Breathaire; Novartis Pharmaceuticals Corporation; East Hanover, NJ). The prebronchodilator FEV1 was measured at randomization and yearly thereafter. Here
Predicted normal values of FEV1 and FVC were calculated based on gender, age, and height at each visit. For male patients < 16.0 years old and for female patients < 15.0 years old, the prediction formulas of Quanjer at al were used, and race correction factors 1.00, 0.91, 0.87, and 0.88 were applied for white, Oriental, black, and other race, respectively. For male patients > 18.0 years old and female patients > 17.0 years old, predicted values were calculated from the official statement of the European Respiratory Society, and the race correction factors 1.00, 0.90, 0.87, and 0.85 were applied for white, Oriental, black, and other race, respectively. In the age range of 16.0 to 18.0 years for male and 15.0 to 17.0 years for female patients, predicted normal values were computed by linear (in age) interpolation between the two formulae.
For the purpose of analysis, the cohort was divided into subgroups determined by age, gender, and smoking history. Children were aged < 11 years old (28% of patients), adolescents were aged 11 to 17 years (17% of patients), and adults were aged > 18 years (55% of patients) at randomization. Cigarette smoking history was recorded as current (11.6%), previous (8.9%), nonsmoker (50.5%), or passive smoker (29.0%).
Longitudinal analysis of change from baseline values in the spirometric variables was made using mixed models, the main model including the factors baseline, treatment, time (since baseline), time squared, treatment X time, treatment X time squared, region, and region X time. A covariance structure of the type compound symmetry was assumed. To study differences in response between demographic subgroups, extensions to the main model were used.