Comments on Nelson et al, a new publication relating alcohol to cancer risk — 14 February 2013

Requested comments on paper by Nelson et al

by R. Curtis Ellison, MD, Professor of Medicine & Public Health,

Boston University School of Medicine

NOTE:  These comments do not necessarily reflect the views of the members of the International Scientific Forum on Alcohol Research

Nelson DE, et al.  Alcohol-attributable cancer deaths and years of potential life lost in the United States.  Am J Pub Health, released February 14, 2013

We always welcome new scientific publications that help elucidate factors related to the development of cancer.  The new paper by Nelson et al focuses on the role that alcohol consumption may play in the risk of a number of cancers. 

Background:  It has long been appreciated that there are a number of upper aero-digestive cancers, such as cancer of the mouth, throat, and esophagus, that occur much more frequently in heavy drinkers, especially alcoholics, than in abstainers.  Physiologic studies suggest that these are not diseases of light to moderate drinkers, as a certain amount of alcohol is required to produce these diseases.  Similarly, among people consuming enough alcohol to lead to liver cirrhosis, the risk of liver cancer is also markedly increased.

These “alcohol-related cancers” should be discussed separately from other more common cancers – especially colon cancer and female breast cancer – for which the risk may be only slightly increased by alcohol.  In the case of breast cancer, there may be a slight increase in risk among some women consuming an average of only one drink/day (in some studies, such an increase in risk occurs primarily among women who binge drink, have an inadequate intake of folate, and/or are also on hormone replacement therapy.)

For these types of cancer, it is especially important to consider the net health effects of alcohol consumption.  For example, it is estimated that if a woman at average risk of breast cancer (i.e., does not have such a cancer in a first-degree relative) decides to avoid drinking completely in hopes of reducing her risk of breast cancer, her risk of breast cancer would be expected to be slightly decreased, on average, by perhaps 5-10%; however, her risk would be increased of dying from much more common diseases such as heart attack,  stroke, or other conditions for which small amounts of alcohol have been shown to reduce risk.  And, importantly, her risk of dying of any cause (total mortality) would actually be increased by her avoiding light alcohol consumption.

The pattern of drinking has often been shown to be even more important than the average amount of alcohol consumed.  A stronger association with beneficial effects is seen with the regular (up to daily) intake of small amounts of an alcoholic beverage; drinking larger amounts on fewer days (including binge drinking) is almost always associated with adverse health effects.  For example, a man having up to 2 drinks each day would have an average of 14 drinks per week, generally considered to be within recommended guidelines.  However, a man consuming 7 drinks on each of only two days each week, despite the same weekly average consumption, would not be considered to be a moderate drinker.

Comments on the present paper:  There are a number of concerns about the analyses and conclusions of the authors of the present paper.  These include the following:

(1)  The authors have “corrected” the reported data on alcohol consumption to make up for presumed under-reporting, using a method not generally accepted by statisticians and other researchers.  This means that even many “light” drinkers are listed as reporting greater amounts of alcohol.

(2)  The authors do not clearly separate the effects of truly moderate drinking from heavier drinking in their conclusions.  They use up to 20 grams of alcohol per day as their lowest drinking category; this is higher than the 14 g/day that is the current definition of responsible drinking for women in the US Guidelines.  Further, as stated, if reported intakes are increased artificially, many more light drinkers would be bumped up into higher categories of drinking.  The result of this mis-categorization is that bona fide moderate drinking, which has been shown by others to have no association with most types of cancer, is improperly associated in this study with increased cancer. 

(3)  The authors’ implications that even regular, moderate drinking increases the risk of many cancers is not consistent with most previous research.  Further, by not having data on the pattern of drinking, the authors include binge drinkers in the same category as regular drinkers, further exaggerating the association of cancer with moderate drinking.  Others have clearly shown that there are large differences in effect between these two patterns.  (It is troubling also that in the paper, the estimated percentage of alcohol-attributable cancer risk among subjects reporting > 0 to 20 grams of alcohol per day is much higher than that of subjects reporting > 20 to 40 grams/day; while this partly relates to the large number of persons who drink only small amounts, such an association makes no sense biologically.  It is difficult to understand who the subjects are in the lowest drinking group, but it may include a large percentage of ex-alcoholics or heavier drinkers underreporting their intake.  However, this makes any conclusions in this paper regarding the risk of cancer among moderate drinkers highly suspect.)

(4)  The authors do not point out the demonstrated effects of alcohol on total mortality; regular, light-to-moderate drinkers live longer.  By focusing only on cancer risks, the authors fail to mention the effects on the risk of much more common conditions, such as coronary heart disease, stroke, dementia, other important health problems of ageing, and on total mortality.  The study of the health benefits and problems of drinking is a very mature field — authors generally discuss their observations in the context of total mortality or other major diseases that would be affected by their experimental design.  In nearly all cases, light drinking is shown to be beneficial; these studies are ignored here.

(5)  Overall, a criticism of this paper relates to the failure of the authors to put their results into perspective.  Statements such as “There is no safe threshold for alcohol and cancer risk” is more of a “scare” statement than a balanced discussion of their results.  Given that almost all prospective studies show that regular moderate drinkers have better health as they age and live longer than lifetime abstainers, even papers focused on the effects of alcohol on any particular disease should present a balanced view on its net effects on health and disease.

(6)  Finally, the authors of this paper have taken the results of their analyses (some of which are based on questionable assumptions) as “truth,” then expounding at length about the public health implications.  There should always be a certain amount of doubt when presenting the results of an individual study, as no one analysis can possibly reveal everything about an association.  (As stated by Voltaire: “Doubt is not a pleasant condition, but certainty is absurd!) 

As the authors acknowledge, observational epidemiologic data can never reveal the full “truth” about the causation of disease from exposures, and each new study’s results must be interpreted taking into consideration previous research.  However, the overall implications presented by the authors of this paper suggest that their goal may have been to support a presumed conclusion to discourage alcohol consumption, not to carefully interpret the available data to best advance the public health.

Key points of these comments

•  There are a number of assumptions taken by the authors in their analyses that raise  questions about their results. The authors present only “adjusted” data for reported alcohol intake (based on national sales, not on individual intake), making the relation of alcohol intake to the occurrence of cancer in individuals unclear.

•  There is poor differentiation between regular moderate drinking and periodic heavy drinking (binge drinking) or alcoholism, although there are marked differences in health effects between these groups.  Regular, moderate drinking is associated with net health benefits, whereas binge drinking and alcoholism have almost exclusively adverse effects (including increases in many types of cancer).

•  There has been a huge amount of previous research in this field, but the authors do not put their own results into perspective or discuss the overall health effects of alcohol consumption.  Previous data have clearly shown that regular moderate drinkers tend to have lower risk of cardiovascular disease, stroke, diabetes, and many other diseases, and have a lower overall risk of all-cause mortality.   

Submitted by

R. Curtis Ellison, MD, Professor of Medicine & Public Health; Director, Institute on Lifestyle & Health, Boston University School of Medicine

Co-Director, International Scientific Forum on Alcohol Research