Bias in Epidemiological Research, Coursework Example

This cohort study examines the potential association between exposure to pesticides and incidence of childhood leukemia. The main bias of this study is due to possibility of differential misclassification between “exposed” and “unexposed.”  That is, if the margin of error is +- 12, the “unexposed” range could vary up to 31 “units”; the “exposed” range could vary down to 8 “units.”  With this substantial overlap, the study will not potentially be measuring unexposed and exposed individuals leading to biased calculations.

The main potential source of bias in this question is recall bias.  That is, if an individual is knowledgeable of his/her HIV/AIDS disease status, he/she may provide an inaccurate number of sexual partners in response to the question. In particular, the number of sexual partners will likely be inflated for HIV/AIDS individuals as they are more to the putative link between the number of sexual partners and HIV/AIDS.  Individuals who tested negative for HIV/AIDS may underestimate the number of sexual partners.

This cohort study examines the potential association between exposure to PCB’s in the workplace to the incidence of cancer over a 20 year time period.  This study suffers from the potential of loss to follow-up bias.  That is, loss to follow-up bias occurs when individuals drop out of a study, and they may have different characteristics than those individuals that remain in the study.  This is particularly of concern in a study regarding cancer because some of those individuals may have developed cancer- or suffered from other maladies- that would be of use to the study.

The potential for bias in this clinical study is the bifurcation of study administration between the intervention and the placebo group- perhaps leading to a conflict of interest.  That is, a doctor working for the pharmaceutical company is calculating the outcomes- he/she may have incentive to a better to the intervention group.  In order to resolve this bias, the study should be administered by a neutral third-party.