Meta-Analysis Systematic and Design Errors: Alcohol Use and Disease Outcomes

2009-2011
Principal Investigator: Kaye Fillmore, PhD
Supported by a grant from the National Institute on Alcohol Abuse and Alcoholism

The focus of this 24 month project is to test hypotheses regarding postulated associations between alcohol consumption and the risk of disease outcomes that have emanated from three types of epidemiological research designs: cohort studies, population case-control studies and hospital case-control studies. The general aim is to test hypotheses regarding possible systematic errors in epidemiological studies. Metaanalyses will be used to determine whether different methods and designs in these studies can unduly bias results towards: (i) creating apparent protection from alcohol use against some diseases; (ii) masking potentially statistically significant associations; (iii) contributing to Œmixed‚ outcomes that defy a clear resolution; and (iv) influencing the size and significance for established alcohol-disease associations. All known epidemiological studies, if applicable, that have evaluated alcohol use and, respectively, ischemic stroke, dementia/Alzheimer‚s, hemorrhagic stroke, and colorectal cancer will be compared.

The following hypotheses will be tested: i) Systematic misclassification error, whereby Œabstainer‚ groups erroneously include people who had reduced or stopped drinking due to increasing age or ill-health, biases towards finding that Œlight/moderate‚ use of alcohol Œprotects‚ against certain disease; ii) Studies employing prospective designs are more prone to systematic misclassification error and hence more likely to show Jshape curves (i.e. protective effects), whereas case-control studies are less likely to suffer from systematic misclassification error and hence more likely to show positive linear associations. iii) Controlling for design characteristics likely to increase systematic misclassification error across a range of different studies will bring study outcomes into closer agreement (e.g. control for proximity of the measure of alcohol consumption to disease incidence, exclusion of subjects with pre-existing illness in comparison groups). iv) Systematic misclassification error will reduce the size and significance of positive linear associations between alcohol consumption and disease.

The project will advance the understanding of alcohol‚s contribution to a range of chronic diseases and improve estimates of alcohol-attributable mortality and morbidity. It should clarify the evidentiary basis for advice from clinical practice and contribute understanding to national and international strategies to reduce the harm from alcoholic beverages.