Does short-term placebo response predict the long-term observation? Meta-analysis on forced expiratory volume in 1 second from asthma trials

The objectives of this work were: (1) to characterize the placebo FEV1 response in asthma; (2) to identify the potential factors with strong influence on FEV1; (3) to determine the predictability of the early (week 2) placebo FEV1 response to the longer term (week 12) FEV1 response. Placebo FEV1 data of about 800 subjects from 11 randomized 12-week clinical trials in mild-to-moderate asthmatics were collected. Stepwise logistic regression methods using SAS were used to model the week 12 trough FEV1 change from baseline greater than a clinically relevant value (150 mL) and to select the predictive covariates. The study effect was assessed using hierarchical logistic regression models implemented in WinBUGS. The results indicated that the early (week 2) placebo response was significantly predictive of the FEV1 response at week 12. Age, baseline predicted FEV1, and season showed statistical significance in the model. The final model showed satisfactory predictability with the area under the receiver operating characteristic curve (ROC) of 80%. The late (week 12) FEV1 response with placebo was positively related to the early (week 2) FEV1 change. The use of the predictive modeling approach proposed in this article presents a valuable method to increase the efficiency of clinical trial design in asthma population.