Confounding factors in observational and intervention studies

Abstract

In studying the association of Helicobacter pylori with any particular illness, it is essential to consider the effects of confounding. The methods of dealing with confounding, a situation in which there is a true association of a risk factor and a disease, and the association is not causal, include matching during the design and carrying out of the study, or stratification and adjustment during the analysis. Adjustment during analysis of data collected in observational studies can reduce the effects of confounding. However, multivariate models are seriously limited in their ability to control for confounding, especially when covariates are imprecisely measured.