Parameterization and estimation of the multivariate normal model in the presence of incomplete and censored data for the methacholine challenge.

摘要

In recent years, there has been a rise in the number of new asthma cases among children and adults within the United States. The Asthma Clinical Research Network was established to assess new and existing therapeutic approaches to asthma. Through clinical trials run by this network, large amounts of data are being collected. High quality clinical trials need to be complemented with new and improved statistical data analysis techniques.; The methacholine challenge is a commonly used assay within such trials. At each clinical visit, three correlated measures are obtained, and a study may incorporate multiple clinical visits. The statistical challenges of the resulting data are that missing data and censored observations are inherent within the methacholine challenge.; This work developed an observed-data likelihood multivariate normal model that performs better than existing data analysis techniques for the methacholine challenge. The bias of this method is compared with five other methods in the univariate setting of a single clinical visit, and the method is extended to the multivariate setting where it is used to analyze a complete clinical trial with repeated measures.