Longitudinal study designs play an increasingly important role in studies of cancer treatment and prevention. Applications that concern our group include preclinical tumor growth delay experiments in mice, large multi- center cancer chemoprevention trials, and practice-based interventions to improve the provision of preventive services by primary care physicians. Although these studies cover a wide spectrum of research areas in cancer, they share a key statistical feature in that each involves measurement of outcomes for individuals or clusters on two or more occasions. We propose to develop, implement, and apply new statistical methodologies to improve the design and analysis of longitudinal studies in cancer treatment and prevention. Specifically, we plan to extend our previous research on tumor growth curves to develop simple and efficient statistical methods for modeling characteristics of nonlinear growth curves; develop methods to perform sample size calculations for longitudinal prevention studies based on GEE models; develop regression diagnostic methods for longitudinal models; and develop methods to analyze multiple outcomes in cancer treatment and prevention studies. To encourage the successful translation of these techniques into applied research settings, we will develop practical software implementations in a format familiar to investigators in cancer research. We will also integrate the techniques into the analyses of the studies conducted by our group.
