Previously, we talked about one reason why studies of Alzheimer’s disease (and other aging diseases) can be hugely complicated: scientists have to rely on observational evidence and are only able to experiment on humans in narrow circumstances. This means that it’s sometimes difficult to tell the difference between cause and effect. In the case of Alzheimer’s, it meant it was difficult to tell the difference between “side effects” of a disease process and the actual cause, meaning scientists can spend a lot of time developing treatments for something that may not be causing the actual disease.
Another problem with clinical studies arises when we try to study the impact of so-called lifestyle factors (exercise activity level, diet) on aging and disease. Since we usually can’t control peoples’ daily lives, we have to use metrics to group people into categories…and those categories can mean something completely different than what you might expect.
The best example of this is in studies of alcohol consumption on aging health. Depending on the study, alcohol consumption either has a U-shaped effect (worse in no or heavy drinkers) or a linear effect (better with more alcohol) on a wide variety of health outcomes, including cognition, cardiovascular disease, and overall protection from death. This has led to a small number of physicians to recommend that mild alcohol consumption is beneficial.
But a new study done on alcohol consumption published by researchers out of University College London shows the problem with taking data on health and alcohol consumption at face value. In this case, the researchers took data from three eastern European populations where alcohol consumption is relatively high (Russia, Poland, and Czechoslovakia), grouped participants into non-, moderate-, and heavy-drinkers, and determined whether or not drinking alcohol offered protection from cognitive problems during aging at the end of the study. Surprisingly, they found that heavy drinking had no consistent effect on cognitive impairment, and that moderate drinkers had better cognitive performance than non-drinkers over the course of the study.
But a deeper analysis showed this wasn’t a simple case of alcohol protecting against the ravages of aging on the brain. When the researchers confined their analysis to just patients who had stopped drinking during the course of the study, they found that they had worse performance on cognitive tests than those that maintained a similar amount of alcohol consumption over the whole study. When questioned at follow-up, the researchers found that more than half of the people in this group had quit drinking due to health problems—suggesting that their poorer performance on thinking tasks was due to poorer overall health having an effect on the brain.
While other health studies attempt to control for this problem (say, by not including people who have signs of health impairment due to alcohol at the beginning of the study), there isn’t really an obvious way to get around this type of problem. Measuring the effects of alcohol consumption in only extremely healthy people isn’t likely to be useful for guessing how it will affect the population at-large, and eliminating ex-drinkers may bias for groups (such as Mormons or Seventh-day Adventists) who have generally healthier lifestyles.
From the perspective of epidemiologists who study health issues on a broad scale, this problem fits into the category of problems known as “selection bias”. It’s a well-known phenomenon that occurs in all kinds of research that involve categorizing data. But this term is just a way of describing the fact that ordering people into categories can select for other, unintended traits that can come alongside and can disrupt your analysis. In this case, the category “non-drinker” selects for the unintended trait “former heavy drinker”—a silent category of people who have worse cognitive performance due to long-term effects of alcohol. In a population with lower overall rates of drinking, clinicians wouldn’t be aware of the phenomenon, but the high alcohol consumption in these countries allowed clinicians to pick it up on a personal level and add it to their analysis.
How do we deal with this issue? The only real way to tell the difference between these two possibilities would be to actually intervene in peoples’ lives—to assign people with similar health and lifestyles to different drinking groups at the beginning of a study and follow the health effects. Clinicians refer to these types of studies as “interventional” studies, and the intervention (say, a controlled, precise dose of alcohol randomly given to participants) removes most of the problems of selection bias. However, since alcohol has some clear long-term negative effects, this is a fairly difficult study to get approved by a regulatory panel.
While we can’t do this easily in people, these sort of interventional experiments have been tried in flies and mice. The typical outcome is that small amounts of alcohol consumption protect against cognitive decline due to disease, and this may be due to stress-induced growth of new neurons in the hippocampus of our brains.
But free-range humans don’t live (or consume alcohol) in an intervention-driven clinical study either. We drink for various different reasons, and in different amounts and patterns, and the interactions of these with other aspects of our lifestyle (exercise, diet) can drastically affect the overall outcome of alcohol on our health.
So wait? Are you recommending moderate drinking or not? Well, we here at SAGE are not all MDs, we’re Ph.Ds, so we’re not licensed or insured to practice medicine or give medical advice. But this story gives you an idea of why it’s so difficult for clinicians to recommend any particular lifestyle modification that’s even mildly harmful—it’s unclear how it’ll interact with our health when the unknown factor of human behavior is included. Even when something appears to work in animals and clinical studies, something like the fact that humans change their alcohol consumption throughout their lives can dramatically alter how results from these types of studies are interpreted.