Abstract:
Case-control studies, that investigate combinations of circumstances causing many diseases is largely in practice in many real world situations. Matching in case-control studies adjust for the effects of confounding variables and increase the efficiency of the study. Several authors have discussed the application of theories to 1:1 and 1:many matched case-control studies. In general, many:many matched case-control studies are useful as they eliminate the problem of disposing of available data due to loss of cases/controls. Design and analysis of many:many matched case-control studies is a topic of interest among epidemiologists. This paper primarily focuses on designing a many:many matched case-control study using the data of Ille-et-Vilaine case-control study in France and illustrating the way in which to analyze such a study.
The data base for the study was the Ille-et-Vilaine study. Chi-square tests confirmed that age is a risk factor for esophageal cancer and is associated with exposures tobacco and alcohol. Thus, age was chosen as the matching variable for the analysis as it is a confounder. Nine strata were formed using age. All the cases in each stratum were chosen for the analysis and a random sample of controls from each stratum was selected for the analysis. This approach resulted in a many:many case-control study.
A conditional logistic model was fitted to the data to find out the factors effecting esophageal cancer. It was found out that wine, cider, beer and tobacco affect the risk of getting esophageal cancer. Every increase in one gram of wine would increase the risk of getting esophageal cancer by 1.035 times. The risk of getting esophageal cancer is 1.024 times higher for every increase in one gram of beer. Every increase in one gram of cider would increase the risk of getting esophageal cancer by 1.028 times and the risk of getting esophageal cancer is 1.029 times higher when one gram of tobacco is increased.
This study illustrates the method of designing and analyzing many-to-many matched case-control studies. Both univariate analysis and model fitting is explained by way of the example used. In addition diagnostics for checking the goodness of fit of the model are proposed.
