Predicting long term Placebo response based on short term observations:Meta-Analysis on FEV1 Response in Asthma Trials
Shuying Yang (1), Roberto Gomeni (2) and Misba Beerahee (1)
(1) Clinical Pharmacology Modeling and Simulation, GSK, UK; (2) Pharmacometrics, GSK, US
Objectives: To characterise the early (week 2) and late (week 12) FEV1 (forced expiratory volume in 1 second) response with placebo in mild-to-moderate asthma population and to investigate the potential influential covariates. The final goals of the meta-analysis were:
- To Identify criteria for population enrichment in novel trials in asthma population
- To provide the necessary model-based support for the implementation of alternative study design (i.e., sequential adaptive, seamless, etc.)
- To evaluate the expected outcomes of a Phase III program (including: benefit/risk and differentiating criteria), help in the definition of a preferred end-stage clinical development strategy including choice of the comparators, study population and experimental design.
Methods: Data from 11 large, well-defined historical 12-week clinical trials for the development of predictive models were utilised. Longitudinal placebo FEV1 data were collected in mild-to-moderate asthma patients. A total of over 1100 subjects' data were included in the analysis. This database includes baseline FEV1, demographics and disease history. All subjects were inhaled corticosteroid (ICS) naïve patients. The probability of the week 12 FEV1 change from baseline greater than a clinically relevant FEV1 response of 150mL (as an example) was modelled and predictive covariates on the FEV 1 response was selected using stepwise logistic regression using proc logistic in SAS (1).
Results: The modelling results indicated that the early (week 2) FEV1 response in the subjects treated with placebo was significantly predictive of the FEV1 change at the end of the 12-week trials. In addition, the baseline percent predicted FEV1 (adjusted by age, gender and height) and age were also important covariates in the model. The predictability of the model was satisfactory (receiver opererating characteristic curve (ROC) ~80%) for future potential application.
Conclusions: The late (week 12) FEV1 response with placebo was positively related to the early (week 2) FEV1 change. This finding has potential to be utilised as a rule to effectively select study population in asthma trials.
References:
[1]. SAS for windows 9.2 manual, SAS institute