Potential sources of bias in studies of drinking water exposures

Although some drinking water contaminates are rare or only present in specific geographic regions, exposure to other impurities, such as disinfection byproducts, is pervasive. As a result, even small effects of such exposures could have important public health implications. However, there are a number of potential sources of bias in studies of water exposures that if unaddressed, could create the appearance of small effects or mask true effects. Information bias due to misclassification of exposure is an area of particular concern because individual level exposure is difficult to assess. In addition, recall bias with respect to the outcome may occur in studies where the participants know that they have been exposed to contaminated water. While differential misclassification of the exposure or the outcome could introduce substantial bias, even non-differential misclassification is important when considering exposures that may have small effects. In addition, residual confounding is of concern in any study, but some water studies may be particularly subject to confounding depending on the study design. When the contaminant of interest is limited to a specific geographic area, a different region may be used to represent the unexposed population. This can result in confounding that cannot be removed analytically if there are risk factors for the outcome that are only present in one location. Therefore, a careful consideration of potential sources of bias is critical when designing studies of water exposures and when interpreting published results.