Is Wine Good for You?

A pair of recent news items got me thinking about wine and health again.  The first was the newspaper article based on a BBC radio series, stating that we should consume less than one drink per day for good health.  On the flip side, we heard about the sad death of Serge Renaud, the man who gave the world the “French Paradox.”  These two stories exemplified the often-heard and irritatingly conflicting opinions that a reasonable amount of wine is either good for us or bad for us.  Well, which is it?  And why can’t we get a straight answer?  I’m going to try to answer these questions and clear away at least some of the confusion (but not all).  I won’t be going into great depth about the many scientific studies relating alcohol consumption to a whole host of human diseases and conditions.  There are many good reviews out there, including an excellent article by Rusty Gaffney in his Pinotfile newsletter, and the meta-study Alcohol and Cardiovascular Health, published in the Journal of the American College of Cardiology in 2007.  In addition, you can find a good review of the subject by Chris Kissack and Jamie Goode in

Now, why does the research community continue to come out with such conflicting and confusing results?  I see several reasons for this situation:

  1. The first issue is what statisticians call confounding, where there are factors at play that are not the focus of the research.  It is very difficult to take into account all influences on the relationship between alcohol and health.  Some possible confounding factors include smoking, eating patterns (including whether or not alcohol is taken with a meal), physical activity, age, gender, ethnicity, dietary supplements, rate of ingestion, pharmaceuticals, and the local environment.  An ideal statistical study is carried out with a population in which all confounding factors are controlled, where the study is “double blind” (i.e. neither the experimenter nor the participant knows who is getting the tested treatment and who is getting a placebo), and where it is randomized, i.e. who receives the treatment or the placebo is randomly selected.  Now, double blind is almost impossible with alcohol – we can usually tell whether or not a drink contains it!  And it is very difficult to control the confounding factors.  One partial solution is to do a meta-study, where researchers look at all the relevant studies and statistically synthesize them so that there is a larger, more significant population, and more confounding factors can be addressed.  One example is referenced above; another is the paper by Castelnuovo et al. from the journal JAMA Internal Medicine in 2006.
  2. A related problem is self-reporting.  Since a double blind study is difficult, most research relies on participants in a study reporting their own alcohol consumption.  If there were only random inaccuracies, then the problem would be minor.  However, what has been found consistently is something statisticians call a systematic error, an error that tends in one direction.  In this case, the problem is under-reporting.  It seems to be human nature to minimize the amount of alcohol consumed, both in terms of the number of drinks and the size of each drink.  In principal, this phenomenon should mean that the consumption limits for good health are actually higher than study results suggest.  In practice, drinkers who try to follow guidelines will still often indulge themselves with a bit more to drink than they are admitting to themselves, so for the most part, the bias probably cancels itself.  In addition, scientific studies rely on a standard drink, set in the US to be 17.7ml or 14g of pure alcohol.  This is equivalent to one 5 ounce glass (just under 150ml) of 12% wine.  However, an individual may pour more than 5 ounces and call it one drink, or the wine may contain more than 12% alcohol (very likely these days).
  3. There is also the problem of the typical study that focusses on only one outcome, or at best a small set of related outcomes.  A prime example of this phenomenon turned up in 2011 when it was found in a report from the Nurses’ Health Study  that even fairly small amounts of alcohol increased the chances of breast cancer.  I’ve placed the data on the following graph.  Notes:  (1) alcohol intake was averaged over 28 years; (2) the alcohol intake in the paper was grouped into ranges – I have plotted the average of the range, except for the >30g/day range, where I somewhat arbitrarily chose 40 to be representative.

    Dependence of breast cancer risk on average alcohol consumption, derived from the Nurses’ Health Study (JAMA 306, 1884 (2011)) over a 28 year period. The straight line is fitted to the data and yields a 12% increase in risk for each 10g/day consumption, similar to the value of 10% per g/day obtained from statistical analysis by the authors of the paper.

    These results created a mini panic with the press enjoining women to limit their alcohol intake to a fraction of a drink per day.  Some analysis of the numbers gives a different picture.  Breast cancer causes 13.7% of cancer deaths in women and cancer is responsible for 12.5% of all deaths in women.  Therefore breast cancer causes 1.7% of women’s deaths.  An increase of 33% in breast cancer mortality (for 2 drinks per day) works out to an overall mortality increase of a negligible 0.6%.   Yet no one took into account the benefits of alcohol consumption to other areas of health.  The best established positive effect of wine consumption is improvement in cardiovascular health.  Since heart disease is a much more frequent killer of women than breast cancer, the resulting improvement in longevity from better heart health easily offsets the small increased risk of breast cancer.  This brings me to a key concept:  the J-curve.

  4. If we plot relative risk of mortalilty (over some fixed period of time) against alcohol consumption, then the relationship usually turns out to look roughly like the letter J, i.e. a J-curve, where risk drops rapidly for low consumption and then turns around and rises as consumption increases.  An example is shown below, where I have summarized some of the results from the meta-study by Castelnuovo et al.

    Fitted relationship between alcohol consumption and risk of mortality. Risk is rapidly reduced for very low consumption and then rises, reaching the non-drinker (baseline) risk at ~2 drinks per day for women and ~3 drinks per day for men (a standard drink is considered to contain 14g of alcohol).

    Women react to alcohol at lower doses than men, due to lower average body weight and a slightly reduced ability to absorb alcohol.  This type of J-curve illustrates the collective effect of a host of conditions, at least those that affect longevity.  It does not say anything about mechanisms.  There is a lot of consensus now that it is the alcohol that is mainly responsible for health improvement, through the enhanced production of desirable HDL cholesterol at the expense of LDL cholesterol and through inhibition of blood clotting.  The possible additional benefits from wine, especially red wine, may arise from compounds called polyphenols, of which resveratrol is the most interesting.  There is, however, still a lot of contradictory evidence with regard to resveratrol.

  5. One possible contributor to the confusion over the benefits of alcohol consumption is the medical mantra “Do no harm.”  This fundamental dictum is highly worthy, but can cause problems when medical practitioners don’t look at the whole picture.  If an isolated negative effect is observed (e.g. with cancers, as in the breast cancer report discussed above), then the behaviour is considered undesirable.  Instead a holistic approach is required where benefits and drawbacks are considered together.  I should add that most researchers understand this point well, but front line health care workers and the press may not always be provided with the whole picture.
  6. There is one last point that I would like to emphasize since it is often neglected in discussions of alcohol and health.  How does the rate of consumption, and in particular the slower consumption rate with food, change the effects of alcohol?  It is fairly well established that binge drinking, that is consuming five or more drinks at a time and then skipping for a few days, is very unhealthy compared to the same amount of alcohol spread over several days.  That difference occurs because the liver can only metabolize alcohol at a fixed rate; the rate varies significantly from person to person but one half drink per hour is a good working number.  However, before alcohol is metabolized, it must first be absorbed, i.e. passed from the digestive tract to the blood.  It is alcohol that has been absorbed but not metabolized that causes problems.  At least one study has shown that a meal eaten with alcohol not only resulted in a 35% reduction in peak blood alcohol content, but it also took 36-50% less time to metabolize the alcohol.  Perhaps this result explains why wine sometimes appears to be more beneficial than other drinks.  Since it is much more often taken with food, wine drinkers may be in effect getting a lower dose.  More research in this area would be very welcome.

So what can we conclude from all this?  These are my thoughts:

  • Don’t take up drinking if you are currently an abstainer.  There is universal agreement on this point.
  • Don’t binge drink.
  • Do drink with food.
  • If you are looking for the maximum health benefit from alcohol, drink only small amounts, say half a standard drink or less.
  • If you wish to enjoy your good wine, and still be at least as healthy as non-drinkers, stick to 2-3 drinks per day (for women/men respectively).
  • Finally, it might be speculated that if your drinking consists almost entirely of a fine wine with dinner, then the limits may be more relaxed, but there is not yet enough direct evidence to support this enticing possibility.