A large prospective study evaluates the relationship between alcohol consumption and breast cancer. A total of 4000 middle-aged women are enrolled in the study through a random selection of residential addresses. Daily alcohol consumption and breast cancer incidence are assessed through the use of periodic questionnaires. On five-year follow-up, the investigators report the association between alcohol consumption and breast cancer as a relative risk of 1.32 (95% confidence interval = 0.90-1.85); 800 subjects were lost to follow-up by the end of the study, the majority of whom were moderate to heavy alcohol consumers. According to this information, which of the following biases is most likely to be present and may have affected the results?

Question

A large prospective study evaluates the relationship between alcohol consumption and breast cancer.  A total of 4000 middle-aged women are enrolled in the study through a random selection of residential addresses.  Daily alcohol consumption and breast cancer incidence are assessed through the use of periodic questionnaires.  On five-year follow-up, the investigators report the association between alcohol consumption and breast cancer as a relative risk of 1.32 (95% confidence interval = 0.90-1.85); 800 subjects were lost to follow-up by the end of the study, the majority of whom were moderate to heavy alcohol consumers.  According to this information, which of the following biases is most likely to be present and may have affected the results?

Quizplus · Accepted Answer

11ec160b_3702_9480_9249_937541a1037b_MD0001_00 Bias refers to a systematic error (due to nonrandom factors) that can result in an incorrect association between the exposure and outcome. In general, when bias is present, the results are consistently distorted in one direction. In prospective studies, if loss to follow-up occurs disproportionately between the exposed and unexposed groups, attrition bias can result if the lost subjects differ in their risk of developing the outcome compared to the remaining subjects. Attrition bias is a form of selection bias (a term that most often refers to systematic differences between groups in terms of treatment response or prognosis). Attrition bias does not occur when the losses happen randomly between the exposed and unexposed groups, as this simply leads to a smaller study population. In this example, a substantial number of subjects (800 / 4000 = 20%) were lost to follow-up, primarily those with higher levels of alcohol consumption. If these lost subjects had a higher incidence of breast cancer than those remaining in the study, then a selective loss of high-risk subjects would have occurred in the exposed group. As a result, the measure of the association between alcohol use and breast cancer would be underestimated, which might explain why no significant association was seen in this study (as the confidence interval of 0.90 to 1.85 crosses 1, which is the null value for relative risk). (Choice A) Lead-time bias occurs when a screening test diagnoses a disease earlier than it would have appeared by natural history alone, so that the time from diagnosis until death appears prolonged even though there might actually be no improvement in survival. (Choice B) Observer bias can occur in clinical trials when study participants or investigators are aware of individual treatment assignments. It can be prevented by performing a double-blind study in which neither participants nor investigators are aware of treatment assignments. (Choice C) Misclassification bias occurs when either the exposure or the outcome is not identified correctly. Random (or nondifferential) misclassification affects all groups to the same extent. For example, if a pediatric size sphygmomanometer cuff is used on all participants (treatment and control groups) as part of a study in adult patients, the blood pressure readings will be incorrect (due to the incorrect cuff size). However, the resulting misclassification will likely affect both groups to the same extent (because the same cuff was used). (Choice D) Recall bias results from the inaccurate recall of past exposure status. It is a potential problem for retrospective studies such as case-control studies, particularly when questionnaires are used to inquire about distant past exposure. However, in this prospective study, exposure status is determined through periodic questionnaires assessing daily alcohol consumption (without inquiring about distant past exposure). In such a scenario, recall bias is less likely to affect results. Educational objective: In prospective studies, disproportionate loss to follow-up between the exposed and unexposed groups creates the potential for attrition bias, which is a form of selection bias. As a result, investigators generally try to achieve high patient follow-up rates in prospective studies.

A Large Prospective Study Evaluates the Relationship Between Alcohol Consumption