Clinical research in smoking cessation has become an increasingly sophisticated enterprise. Although many of the design issues in smoking cessation research are similar to those encountered in other clinical areas, the types of outcomes measured and the structure of the outcome data have special features that set them apart. Standard methods of modeling and analysis that statisticians have developed over the years are therefore not well suited to handling data from smoking cessation trials. We propose to develop a range of statistical models and estimation procedures to handle the special features of smoking cessation outcomes. The availability of appropriate methods for this setting will further clinical research by permitting investigators to mine their data for new insights and design and analyze trials with maximum efficiency. This application describes a proposal by a team of experienced statisticians to address a range of critical methodologic problems arising in clinical trials of smoking cessation treatments.
The first aim will investigate the problem of estimating the rates of smoking cure, together with the effects of treatments on cure and relapse rates, from data on cigarette consumption when subjects may experience alternating periods of abstinence and relapse.
The second aim will consider models for daily cigarette consumption data in which the distribution of cigarette counts may exhibit excess zeros, a consequence of periods of abstinence;an integrated model based on smoking intensity rates will link models for abstinence and consumption given non-abstinence.
The third aim will address the issue of heaping of cigarette counts - i.e., the tendency of subjects to report cigarette consumption rounded to the nearest multiple of five, ten or twenty cigarettes. The methods will be applied to data from a number of smoking cessation trials conducted by the University of Pennsylvania's NIH-funded Transdisciplinary Tobacco Use Research Center and the University of Pittsburgh, in which the three investigators have been active collaborators. It is anticipated, however, that the resulting methods will be applicable more broadly to clinical research on outcomes that experience alternating periods of relapse and remission.
Showing the most recent 10 out of 11 publications
