Here we have another one of these stories that should be waved obnoxiously in anyone’s face who lectures you about blindly “following the science.”
For decades—really as long as I can remember—researchers have been telling us that moderate consumption of alcohol was not just safe but in fact beneficial. This wonderful news was welcomed by those who “needed a drink” after a hard day, or self-medicated with a glass of wine (or good scotch) before bedtime, or who tended to have just a bit more than a moderate amount of alcohol now, then, or frequently, but who’s counting?
Along comes a report from the Canadian Institute for Substance Use Research at the University of Victoria in British Columbia, Canada, that appeared a week ago in the Journal of Studies on Alcohol and Drugs published by the Center of Alcohol & Substance Use Studies at Rutgers University in New Jersey. It announced, in essence, “Oopsie! All of us trained scientific researchers made just a teeny mistake in our previous studies on this topic, repeatedly, over and over, and for half a century or more!”
Researchers confessed that previous “science-based” recommendations promoting the life-extending benefits of wine (for example) had been based on flawed scientific research, as were all those studies showing that “moderate drinkers” live longer than teetotalers because that single beer or scotch or Merlot each day decreases the risk of heart disease, some cancers and other dire illnesses.
It turns out that this finding was hooey. The public should “be skeptical of claims that alcohol use in moderation may prolong life and reduce the risk of serious health conditions and premature death,” the study’s lead author, Tim Stockwell, a scientist at the Canadian Institute for Substance Use Research, says. “Drinking less or not at all is more consistent with improving health and well-being.”
Oh. And why were we getting the opposite advice for so long, pray tell? The studies that showed how moderate alcohol use was good for you compared moderate drinkers with non-drinking groups that included older adults who had quit drinking or had reduced their alcohol consumption because they had health conditions that were exacerbated by drinking. This made those who could keep on drinking look healthy by comparison.
Researchers in the new study identified 107 published studies that followed people over time and examined the relationship between drinking habits and longevity, and sure enough, it appeared that light-to-moderate drinkers— those who imbibed between one drink per week and two per day—had a 14% lower risk of dying during the study periods compared with the squares who didn’t drink at all. Then the researchers—Finally!—-realized that most of these studies had the same confounding flaw, which was failing to distinguish between subjects who had never had alcoholic beverages regularly and those who had stopped drinking.
Now, the best research available indicates that the less an individual drinks, the healthier he or she is likely to be. In a way, the study was really about how misleading the majority of the research has been and and how “the appearance of health benefits vanishes when studies use more rigorous methods.”
I know this is hindsight, but why did it take so long for our scientific geniuses to recognize a rather obvious and glaring mistake? Among non-drinkers are recovering alcoholics who have imbibed large amounts of alcohol before quitting (or trying to quit) by necessity. Nonetheless, the damage their addiction has caused may be permanent. Alcohol use has been part of human society for thousands of years. I have to assume that if scientists couldn’t figure out it wasn’t good for us, it was because they didn’t want to.
Bias made them stupid….
…and made some of the people who “trusted them”followed the science” sick or dead.
Ethics Alarms 
“I have to assume that if scientists couldn’t figure out it wasn’t good for us, it was because they didn’t want to.”
We have to assume that studies that have uprooted other bits of conventional wisdom over the years have the same motivation: the results are based on what the researchers want or don’t want to be true.
Like most papers dealing with health, I would take the results of the study with a grain of salt (and a margarita). The research focused on how the results of studies can be skewed by failing to offer reasonable controls on the data, but they themselves only used a small handful of controls. Granted, the problem with trying to implement all the needed controls is that they quickly become intractable, with so many variants that a huge number of participants are needed to have any statistically meaningful conclusions. Research into health is messy, and for that reason, I don’t have good confidence in any health studies.
The general conclusions from previous research has been that there is a J-shaped curve regarding health results and alcohol consumption. The worst health results are for heavy drinkers, followed by abstainer, followed by medium drinkers, with the best results for light drinkers. This study took 107 studies that met their listed criteria and examined how this J curve looks between good studies and poor studies. Predominantly they seem to have been interested in separating those who never (or very rarely drank) from those who drank for a time and then quit, especially when cohorts like that could be studied from much younger ages. Over all, the J-curve flattened at the loop, with the best health results statistically tied between abstainers and light drinkers, followed as expected by moderate and then by heavy drinkers.
Interestingly enough, their study tried to control for smoking, and the results were indeterminate. Because of a strong correlation between drinking and smoking, they wonder if the cause of the large J result in previous studies came from the possibility that drinking offset some of the ill-effects of smoking.
What really stood out to me was how, despite every effort to make a longitudinal comparison between the healthiest possible people who never drank, and those who were light drinkers, the worst they could say was that there was no statistical difference between health outcomes. Moderate drinking did drop a notch in the health rankings, but that really shouldn’t be any kind of surprise.
I think overall, the authors made reasonable assumptions in constructing the controls on their study. Certainly it seems quite reasonable to make a distinction between people who quit drinking for health reasons, and those who never drank at all. Trying to filter out the noise that comes from pre-existing health conditions, from smoking, and from socioeconomic status also seem fairly prudent. Overall, though, their conclusions only warrant “more study necessary.” There are a lot of factors at play in any health assessment, and while I agree that the controls they identified as necessary are good, they are also chosen predominantly to weigh heavily against alcohol consumption altogether.
A final note: these researchers are funded, as they fully disclose, by groups that are against alcohol consumption. Take that as you will. I would have to study the paper a lot more to see if that really biases the results, but my initial thoughts are that the bias is slight.
All of which supports my point, which is that “studies,” “research” and “Science” are so often and essily distorted by human conflicts, biases, agendas and imperfections that to hold science up as an absolute authority in all (or even most) cases is irrational.
Or dishonest. Or stupid and naive.
Unfortunately, it also has the effect of making me cringe every time I hear the words “data-driven” or “evidence-based.”
These words mean that someone has found a correlation via an unknown amount of data manipulation in order to justify doing what they already wanted to do but can now cloak it in The Science.
Science has been contaminated and I don’t know if there’s any fixing of it.
What would help is if scientists accepted that their discoveries, conclusions and edicts will be subjected to rigorous examination and inherent skepticism, and that the level of trust they can maintain with the public depends upon their overcoming their egos and arrogance and earning that trust.
I’m sure we’d all accept that Stephen Hawking is a genius; but even he is still zero for two; at least for another thirty years!
Now, the best research available indicates that the less an individual drinks, the healthier he or she is likely to be. In a way, the study was really about how misleading the majority of the research has been and and how “the appearance of health benefits vanishes when studies use more rigorous methods.”
Admittedly OT, but this is a perfect axe grinding opportunity. If one is talking about an individual, then that person’s gender is known. Otherwise, the subject is plural: the less people drink, the healthier they are likely to be.
Better: those who drink less are more likely healthy.
The he-or-she construct is almost invariably a sign of sloppy thinking.
Back to the subject at hand. My bet is that aside from heavy drinking — that which is mind altering — alcohol consumption, or lack thereof, has no measurable effect on life span.
I genuinely don’t understand your grammatical complaint. “An individual” is of unknown gender, so “he or she” is fair. I could default to the neutral, Mankind “he” but this is my small concession to complaints about gender weighted language. I think “the less people drink, the healthier they are likely to be” is sloppy, because the issue is individual health, not collective health.
Except the issue isn’t individual health, it is statistical health. The assertion being that if two otherwise identical groups differ only in their alcohol consumption, the group that consumes less will be healthier.
There is absolutely no way to know at the level of individuals, for moderate consumption, whether alcohol caused their specific health outcomes to be different than would have been the case absent alcohol.Because such studies are stochastic, invoking individuals doesn’t fit.
More broadly, unless one is speaking of an individual, in which case gender is known, then the he/she construct just doesn’t fit.
Again, Jeff: if the message is that YOU are better off and likely to be healthier if you don’t drink, and that is the message that health officials want to get through, then we are talking individuals, and “he or she” is not inappropriate. Otherwise it’s like group stereotyping: OK, voters as a group don’t know their asses from their elbows: that doesn’t mean anything to me. I don’t care. If, however, the message is, if YOU are a human being, all evidence points to you being better off if you don’t drink at all. And an unspecified individual is similarly warned. An individual is either a he or she—if you refer to that individual with a pronoun, that’s the correct choice.
And he or she is what you do when the gender ISN’T known.
Again, Jeff: if the message is that YOU are better off and likely to be healthier if you don’t drink, and that is the message that health officials want to get through, then we are talking individuals, and “he or she” is not inappropriate.
No, it isn’t appropriate, nor is it correct.
Here is their conclusion: Studies with life-time selection biases may create misleading positive health associations. These biases pervade the field of alcohol epidemiology and can confuse communications about health risks. Future research should investigate whether smoking status mediates, moderates, or confounds alcohol-mortality risk relationships.
Studies with life-time biases produced health associations that can’t be relied upon. That is the message. Maybe eliminating those biases will vitiate those associations. Maybe not.
Regardless, by definition, epidemiology, being inherently statistical, is meaningless at the individual level. Assume that it has been demonstrated to a fare-thee-well that, everything else completely accounted for, non-drinkers live, on average, two years longer than drinkers. How much longer will non-drinker Jack live than drinker Jeff?
There is no rational answer to that question.
What is the difference in average life span between 10,000 non-drinkers, and 10,000 drinkers? There is, in principle, a rational answer to that question, and “he or she” has no logical place in it.This is just one example of why “he or she” is nearly always conceptually wrong.
Whether it is right or not, and I agree with all of that, the fact is that if you link the study to an individual, that individual must be he or she, him or her. The alternative is to talk about group probabilities only. From the linked quotes from the researchers: “to be consistent with our messaging to the public — you may like alcohol and it may be enjoyable, but it doesn’t prolong your life.” You is singular, referring to me, or you, but an individual is at the receiving end of that statement one way or the other. That individual must be a he or she.
The “public” doesn’t have a “life,” so you can’t say “you” is intended to be plural in that statement.
Jeff,
I don’t know if this was intentional, but in your axe-grind, I think you point to a particular problem that plagues understanding the results of studies. Compare the two statements:
The less someone drinks, the healthier he is likely to be.
Those who drink less are more likely healthy.
Notice that the first statement strongly indicates causality, in that the healthier status comes from less drinking. The second statement is more correlative, merely noting that drinking less and a healthier status tend to correlate. Now, I think we could quibble about how the first statement could be taken in a correlative sense, too, but I think a general read would interpret it as “A implies B”.
Why is this important? Because with so many of these health-related studies, the noted correlations are often presented to the public as causal, and in the particular direction that fits the desired narrative. Maybe drinking less leads one to be healthier, but might it be the case that healthier people simply drink less? Maybe less healthy states cause an increase in alcohol consumption.
The study we’re examining here notes that the bias in previous studies came from counting people who quit drinking as abstainers. They note that most often people who quit alcohol do so for health reasons, which means that including them in the abstaining category adds a pool of unhealthy people into that group. But they don’t consider (at least from what I read) the possibility that unhealthy people might remove themselves from the abstaining category. The closest they come is considering the impact of smoking and the impact of socioeconomic status. They also do not seem to differentiate among those who quit alcohol whether the poor health was directly due to alcohol or not. Note what follows from that: in order to address what they perceived as bias in the question of whether or not alcohol leads to poor health, they essentially assumed that alcohol led to the poor health conditions in those who chose to quit drinking. Now, undoubtedly most of those who quit drinking probably did so because either doctors warned them that continuing to drink would cause poorer health, or because they noticed that drinking was having an adverse impact on their health, so the assumption the researchers made was probably reasonable. But I call this out to address just how much we can be blinded to the difference between causation and correlation.
Excellent comment.
It just occurred to me that a way to get around those complications is to compare two groups that are very similar in all respects except alcohol consumption: non-smoking Catholics, and Mormons.
If the average lifespan of the two groups is statistically distinguishable, then there is an a priori reason to suspect that alcohol is a contributing factor; otherwise, it likely isn’t.
So, just like all the data that shows that marijuana usage is generally harmful. People want to get high so badly and the social pressure to not ‘be a square’ made the data interpretation impossible.