Part of an occasional series on new health information.
Frequently, reports of new medical findings seem to suggest measures we can take to improve our health. Let’s consider an example. In 2001, the news included an item about a new report from Harvard that moderate coffee drinking could “cut the risk” of developing Parkinson’s disease nearly in half. A few months later, a physician-authored column about Parkinson’s disease in a nationally circulated Sunday newspaper supplement stated, “moderate caffeine intake appears to reduce the risk of Parkinson’s in both sexes—so drink coffee or tea.”
Should you take his advice?
The new information came from two large observational studies, the Nurses’ Health Study and the Health Professionals’ Follow-up Study. The actual report concluded that the “results support a possible protective effect of moderate doses of caffeine...” Not all studies of Parkinson’s and caffeine have supported their conclusion. In fact, the authors also discussed two other explanations for their results. Furthermore, some experts have suggested that either something about Parkinson’s while it is developing but still hidden, or something about the people who get it, leads to avoidance of coffee. If this were the case, decreased coffee consumption would not be a cause of Parkinson’s, but rather be caused by the disease or be a feature of the people who get it.
The Harvard study could not tell the difference among these various possibilities because observational studies cannot establish a cause and effect relationship. The strongest type of evidence comes from a randomized controlled trial, which is truly an experiment testing a specific hypothesis. Participants are randomized into groups, each receiving a different treatment, diet or other intervention. Randomization makes all the groups essentially identical in makeup, with each group’s unique intervention being the only difference. Thus, any difference in health outcome which occurs should be due only to that group’s unique intervention.
If the hypothesis were that coffee intake influences the development of Parkinson’s, then a trial might, for example, randomize participants into groups: one drinking no coffee at all, 1-2 cups daily, and so on. If those consuming more coffee had a lower incidence of Parkinson’s, that would be strong causal evidence that drinking coffee actually could affect the development of Parkinson’s.
This kind of study is difficult to do. Participants have to be willing to adhere to their assigned coffee regimen for a long time. The researchers have to decide how long the trial should last without knowing whether caffeine has an effect, and if it does, how long a person has to drink coffee or how long before the Parkinson’s would have developed. So while a randomized controlled trial provides the strongest evidence for cause and effect, it is relatively impractical for dietary or nutritional issues and infrequently done.
More practical is an observational study in which groups of people are studied backwards or forwards in time by gathering information about what they do and what their health outcomes are. The participants make their own choices about their lifestyle and healthcare.
In our Parkinson’s example, researchers divided participants into groups according to daily intake of coffee or caffeine, and examined the incidence of Parkinson’s in each group. They observed an “inverse association” between coffee or caffeine intake and the occurrence of Parkinson’s. However, associations revealed by observational studies have multiple interpretations and sources of error. Hence, associations do not prove a hypothesis, but may support it and suggest that further study is warranted.
The vast majority of information on the effects of nutritional factors is based on observational studies, such as the Parkinson’s report was based on. The results may be reported in terms of “cutting” or “reducing” your risk or “protection” from some unwanted health outcome. This wording implies a cause and effect relationship that may not really have been established at all.
Should you take the advice of the Sunday supplement columnist and drink coffee or tea? What about all of the other effects of caffeine? You can find recommendations to avoid caffeine for people with certain conditions.
Does it make sense to act on one piece of information, which doesn’t even prove an effect, from one study about one possible effect of caffeine? Unfortunately, all too often this is how the information arrives. To me, it makes more sense to consider all of the evidence and known effects of an intervention before adopting it.
Future columns will discuss how to get your information from sources that review an entire issue and have the expertise to evaluate the strength of the evidence. Armed with a broader base of information, you can better decide what makes sense for your circumstances.
Mark H. Zweig M.D., previously a staff physician at NIH, has more than 35 years of experience in medical research and 25 years in patient care, with particular interest in medical decision-making and preventative medicine.