What we believe determines what we want to study and how we study it. This is a somewhat dangerous statement in science, where being objective is an ideal. However, it’s true that the approach to designing and conducting a study may be inherently (and unintentionally) biased simply an investigator’s viewpoint and preconceived notions about the relationship in question. A study is also subject to bias by nature of its design, meaning how a study is implemented. Keep these points in mind as you participate in this discussion.
An epidemiologist has a close relative who recently died of brain cancer. She would like to evaluate the relationship between cellphone use and brain cancer using a hospital-based case-control study. The plan is to select cases of brain cancer from hospitalized patients, and to sample controls from individuals hospitalized for different reasons. The hospital is in a low-income area, and is well-known for its expertise in treating cancer, so cases may come from all over the U.S. to be treated. Controls, however, will likely be sampled from the population who lives in the area near the hospital.
First, let’s start with a question based on what we have learned so far in past modules:
Consider the viewpoint of the investigator. How might she conduct this study or not conduct this study differently from someone who did not have a close relative die of brain cancer?
Consider the selection of cases and controls for the study. Does the approach seem reasonable? Why or why not?
Now applying what you’ve learned from the readings , let’s go through the rest of the discussion together, one question at a time.
One of the tenets of scientific research is that it should remain open to inquiry. We may as a scientific community think we understand some exposure-disease relationship, but that theoretically remains true only until proven otherwise. You have learned that observational research is always potentially limited by many biases and other sources of error. You should read biomedical research articles with a healthy skepticism, using these newly developed skills which allow you to evaluate whether a particular study design is appropriate, whether adequate information was collected and assessed, and if the study findings are subject to potential bias. Recognizing the limits of research is as important as understanding its contributions in the context of existing knowledge.
Gary Taubes is a science writer who has long been a vocal critic of epidemiology. His views have caused the ire of many upstanding researchers, often because while he may miss some of the nuances of epidemiology, he also makes good points! Now that you have a background in study design, risk estimation, and bias, you are in a good position to determine for yourself if all of Mr. Taubes’ arguments hold sway. You may find it informative to read the earlier references to Taubes’ own work (“Do we really know what makes us healthy?” is a fascinating read), but this is not required for the discussion.
Read “Science, Psuedoscience, Nutritional Epidemiology, and Meat”. You may find it helpful to jot down some of your gut reactions to the article as you read – keep in mind your role as a healthy skeptic, and don’t be afraid to agree with him. Participate in the following discussion by addressing the following points about several of Taubes’ key arguments:
The magnitude of association for the cited nutritional epidemiology study by Willett et al. (and many other observational studies) is very small.
Given what you know about how bias can affect the measure of association and therefore what we infer from a study’s findings, do you agree with Taubes’ assertion that small risks essentially mean the absence of risk? Why or why not?
Taubes repeats his mantra that experimental studies are the only way to determine what factors are causal for a health endpoint.
Given what you know about study designs and bias, what might you say to refute his opinion? Provide clear and specific examples in your response.
