Weighing the Evidence: Bias in Pediatric Studies

Abstract

Bias is defined as any process that leads to the systematic deviation of study results or inferences from the truth.1 The term bias does not mean that the investigator’s conscious or unconscious views led to a particular subjective interpretation, as in conventional usage of the word. Bias in evidence-based medicine refers to deviations from the truth that generally result from limitations in study design or reporting rather than from personal prejudices. Some forms of bias are inevitable in research, but need to be recognized and accounted for by investigators and readers.Many distinct forms of bias have been recognized. In 1979, Sackett identified 57 types.2 In 2005, Choi and Pak noted 48 varieties of bias solely in studies using questionnaires.3 Dorak described 108 bias categories in 2008.4 I will describe a few of the more frequent types of bias below.Sackett has noted that bias may occur at any of the following stages of research:2The effects of bias may be positive when the observed measurement of effect is larger than the true measurement, or negative when the observed effect measure is smaller than the true one. In the case of an odds ratio or risk ratio, a positive bias would distort the observed statistic farther away from 1.0 than the true measurement, and a negative bias would misrepresent the observed ratio closer to 1.0 than the truth.Some important types of bias are:Many conditions, such as cancer and HIV infection, have variable survival. When measurement of possible risk factors is selectively done on existing survivors (prevalent) rather than newly diagnosed (incident) subjects, distorted results may occur, since the measure may differ related to survival.Individuals who participate or do not participate may differ systematically and create error in study inferences. This may occur in the design of selection criteria or by the choice of individuals to participate or respond to a survey.This is a special type of selection bias that may occur in hospital-based studies when admission rates of exposed and unexposed cases and controls differ.Cases and control subjects may have differences in accuracy of recalling past events. This may occur in studying the causes of congenital anomalies when the recall of past exposures by parents of affected children are compared with those of unaffected children.This may occur when different groups selectively suppress or reveal information, such as sensitive information on sexual behavior or substance use. Reporting bias may also occur in a study such as the one by Slapak, et al reported in this issue. Here, neither the physicians nor the subjects were blinded to their treatments, so their knowledge of which treatment they received may have selectively influenced their reporting of either benefits or adverse effects of the intervention.This is the tendency of authors and editors to publish studies with positive results and to reject those with negative results.When an extraneous factor is associated with both the exposure and outcome, but is not an intermediate step in the process, there may be distortion in the estimated effect of the exposure on the outcome. An example of this is in looking at the effect of prenatal cocaine exposure on childhood behavior and development, where sources of confounding may include tobacco exposure and other factors. If these extraneous factors are disproportionate between cocaine-exposed and unexposed groups, confounding may occur.How can we deal with bias? Investigators are usually aware of potential sources of bias when designing studies. Methods such as randomization and blinding in clinical trials and selection of matched control groups in observational studies are used to reduce bias and confounding effects. In study analysis, stratification and multivariate analysis are used to account for some confounding effects. Publication bias may be reduced by preregistration of clinical trials in a public registry. Readers should be aware of sources of bias and how they were managed when they interpret study results.