A general asymptotic theory for case-control designs was published, and methods to analyze cluster samples of controls are being adapted to analyze case-control studies of breast cancer and non-melanoma skin cancer. Methods to estimate absolute risk from hybrid cohort/case- control designs have been published and extended to allow for studying additional risk factors in subsamples of cases and controls. Recent developments for inference on attributable risk were reviewed. Computer programs to implement these new methods are in development. A paper appeared that showed how to adjust lung cancer risk attributable to radon for smoking, and another paper described how to estimate attributable risk even if some unexposed subjects are misclassified as exposed. Simulation studies demonstrated that measurement error and other factors limit the power of case-control studies to detect risk from exposure to radon in homes. A paper appeared that demonstrated that conventional adjustments for measurement error can be very misleading if error is correlated with the true exposure, as happens in studies of diet. Another publication demonstrated that linear relationships among macronutrients make it impossible to draw some desired types of inferences. New methods were developed to select a subset of questions to estimate nutrient intake. Methodological work on randomization-based inference for community intervention trials was developed and incorporated into analyses of the Community Intervention Trial to promote smoking cessation. Methods for monitoring Phase II cancer trials were also published. A description of appropriate graphical presentations for studying trends in cancer rates was published. Regression methods for overdispersed Poisson data were developed and applied to describe AIDS incidence data. Methods for age- and calendar-year-specific backcalculation were refined and applied to estimate HIV infection rates and prevalence in the United States.
