Cancer Control Research
5R01CA040644-15
Breslow, Norman E.
STATISTICAL METHODS IN CANCER EPIDEMIOLOGY
AbstractDESCRIPTION (Adapted from Applicant's abstract): Epidemiology plays a major
role in the identification of carcinogenic agents and in the quantification
of dose time response relationships upon which regulation and preventive
strategies are based. epidemiology as a science depends critically upon
statistics. The goal of this project is the development of more efficient
statistical designs and methods of analysis for both analytic and
descriptive studies. There are three areas of emphasis. First, many
studies involve the estimation of a large number of related quantities:
multiple relative risks in case-control studies involving multiple diseases
and multiple risk groups; multiple cancer rates in small areas used for
construction of maps; and multiple individual responses to intervention in
longitudinal studies. A major goal is the development, evaluation and
implementation of hierarchical statistical models that allow for the
efficient estimation of such related quantities. Second, two-phase
case-control studies and other complex stratified designs are of great value
in limiting the collection of costly covariate data to those subjects who
are most informative regarding disease/risk factor associations. An
important example is the validation substudy conducted to alleviate the
effects of measurement error. Optimal methods for design and analysis of
data from such complex designs will be developed. Finally, epidemiologists
have proposed new study designs that involve comparison of the exposures of
diseased cases with those of internal or artificial controls. Examples are
the haplotype relative risk method in genetic epidemiology, the
case-specular design for study of electromagnetic fields of cancer, and the
case-crossover and case-time-control designs for studies of the effects of
intermittent exposures on event rates. Unfortunately, misleading inferences
occur when these methods are used in situations that do not meet the
underlying assumptions. A critical evaluation is planned of the logical
foundations of such case "pseudo-control" designs, with a goal of maximizing
the validity and efficiency of inferences based upon them.
The methods used to achieve these goals include mathematical and statistical
analysis, computer simulation and application to important datasets
collected by cancer epidemiologists and other public health scientists.
|