Application of Case-Cohort methods to data on pulp and paper mill workers in British Columbia
There are two common methods for comparing disease incidence rates (such as cancer) in two populations (such as pulp and paper workers vs non-pulp and paper workers). In cohort studies, the two groups are followed over time and the incidence rates are directly compared. These types of studies can be inefficient for low incidence diseases when very large sample sizes are needed. Case-control methods take each incidence of disease and match it to a control. Then contributions from variables such as exposure to chemicals to the disease incidence rate can be determined. While more efficient than cohort studies, direct incidence rates cannot be computed.
Gregory's thesis used a newly proposed method, the case-cohort study, that combines features of both types of studies. Because it uses two cohorts, it uses more information than the case-control study but also gains efficiency from the matching of cases with controls.
While this method has been extensively theoretically developed in the literature, it has only be applied to simple problems or simulations. Gregory used this new method to reanalyze a long running study conducted by the British Columbia Cancer Agency. While the new methodology did not give dramatically different results, it did yield improved precision in estimates (implying that it will be easier to detect excess disease rates). Some potential dangers in the uncritical use of this method were also identified.
This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Jacqueline Gregory (jgregory (at) stat.sfu.ca) or her supervisor Randy Sitter (sitter (at) stat.sfu.ca).