The focus of this project is development and refinement of statistical procedures for the design and analysis of cancer screening and related studies. Problems under investigation include derivation and comparison of data analysis methods, assessment of case-control studies for screening evaluation, development of models of cancer screening, approaches to the analysis of categorical data, and geographic analysis and smoothing of cancer mortality rates. To assess the case-control design for screening evaluation, the MISCAN microsimulation model is being used to provide population data with and without screening. Case-control studies are then done in the screened populations and the results compared with the true effect to assess bias in the case-control approach. Criteria were developed for comparability of the restricted case subgroups used in the Limited Analysis of a cancer screening trial. Data from diagnostic testing and screening can often be analyzed using techniques for missing categorical data. A matrix model for incomplete multinomial data, the composite linear model, was developed to provide a unified approach for maximum likelihood inference for such data. A new study design for assessing screening was examined in which controls receive a screen at the end of the screening period, and only cases diagnosed to that time are followed and analyzed. Methods have been developed for estimating the benefit of screening unaffected by lead time bias and the average lead time, by examining the differences in case survival measured both from time of entry and time of diagnosis between screened and control groups. Approaches were defined for data monitoring of cancer screening trials. Further, exploratory analysis of cancer mortality rates for cancers of the lung, prostate, and skin was undertaken. Linear smoothing of standardized prostate cancer mortality rates revealed interesting features that otherwise might have been obscured.
