In cancer research, the statistical procedures that are developed are usually applied to ad hoc situations. For example, most statistical procedures in clinical trial evaluations focus on a particular treatment regimen that is effective against a specific type of cancer. On the other hand, reliability growth model methods can be used to analyze a sequence of clinical trials occurring over a period of years. Such an analysis can assess the growth in the success in treating a given cancer and predict success probabilities.The specific purposes of this research project are: (1) to convert the data obtained from previous clinical trials on specified types of cancer or the data on both specificity and sensitivity in cancer screening tests to "successes" and "failures" required for analysis by reliability growth methodology, (2) to estimate the requisite model parameters and test the fit of the model to the data for the parametric class of models, (3) to assess the applicability of the nonparametric and Bayes procedures where the parametric techniques are not suitable, and (4) to extend existing models to account for such factors as clinical trials lasting for different lengths of time.The principal research areas are: further development of parametric reliability growth models, extensions and modifications of the nonparametric models with an emphasis on Bayes and empirical Bayes techniques, and application and evaluation of existing and modified reliability growth models to assess clinical trial methodology and cancer screening procedures.Methods developed in the research project will provide invaluable tools for assessing and, in some instances, predicting patterns of response in several areas of cancer research, that is, assessing changes in response rates in cancer clinical trials and assessing changes in sensitivity and specificity in cancer screening procedures. The evaluation of different screening procedures will be invaluable in setting policy on the various screening procedures.National Cancer Institute
