Critique 229: A J-shaped curve for the relation of alcohol consumption to the risk of colorectal cancer — 10 July 2019

McNabb S, Harrison TA, Albanes, Berndt S, Brenner H, Caan BJ, et al.  Meta-analysis of 16 studies of the association of alcohol with colorectal cancer.   Int J Cancer. 2019 Apr 29. doi: 10.1002/ijc.32377. [Epub ahead of print]

Authors’ Abstract

Alcohol consumption is an established risk factor for colorectal cancer (CRC). However, while studies have consistently reported elevated risk of CRC among heavy drinkers, associations at moderate levels of alcohol consumption are less clear.

We conducted a combined analysis of 16 studies of CRC to examine the shape of the alcohol-CRC association, investigate potential effect modifiers of the association, and examine differential effects of alcohol consumption by cancer anatomic site and stage. We collected information on alcohol consumption for 14,276 CRC cases and 15,802 controls from 5 case-control and 11 nested case-control studies of CRC. We compared adjusted logistic regression models with linear and restricted cubic splines to select a model that best fit the association between alcohol consumption and CRC. Study-specific results were pooled using fixed-effects meta-analysis.

Compared to non-/occasional drinking (≤1 g/day), light/moderate drinking (up to 2 drinks/day) was associated with a decreased risk of CRC (odds ratio [OR]: 0.92, 95% confidence interval [CI]: 0.88-0.98, p = 0.005), heavy drinking (2-3 drinks/day) was not significantly associated with CRC risk (OR: 1.11, 95% CI: 0.99-1.24, p = 0.08) and very heavy drinking (more than 3 drinks/day) was associated with a significant increased risk (OR: 1.25, 95% CI: 1.11-1.40, p < 0.001). We observed no evidence of interactions with lifestyle risk factors or of differences by cancer site or stage.

These results provide further evidence that there is a J-shaped association between alcohol consumption and CRC risk. This overall pattern was not significantly modified by other CRC risk factors and there was no effect heterogeneity by tumor site or stage.

Forum Comments

It is unfortunate that some scientists and the press often report associations of certain diseases with “alcohol consumption” without noting (perhaps intentionally in some cases) what the relation is according to the amount consumed and, especially, the drinking pattern.  It is clear that periodic drinking, such as large amounts on the week-end, has very different effects on health than consuming the same average amount spread out over the week, especially if the beverage in consumed with food.

The present study of the relation of alcohol consumption to invasive CRC was conducted by leading investigators in the field, and is a meta-analysis of 16 studies with a number of strengths; these include the large number of subjects, the ability to use individual level data in their meta-analysis, and the use of appropriate and complete analytic techniques.  The key weaknesses, in the opinion of the Forum, is that having to use only the average alcohol intake they did not have the ability to judge potential effects according to drinking pattern (regular moderate versus binge drinking) and the lack of ability to judge beverage-specific results.

In general, the ability of physicians to predict the development of CRC is limited, although a positive family history, obesity, red meat and dietary intake, folate levels, physical activity, and a few other factors have some value.  The investigators have included all of these variables into their analyses as potential confounders or modifiers of effect.   Knowing that the association with alcohol may be non-linear, the investigators evaluated the association appropriately by regression using splines and higher level terms.

Overall, their sample, from the USA, two European countries, and Australia, consisted of 41% non-drinkers (included both ex-drinkers and never-drinkers), 47% light or moderate intake (1.1 to 28 g of alcohol per day, with the latter being the equivalent of two typical drinks), 6% reporting 28.1 to 42 g/day (2-3 typical drinks), and 6% reporting >42 g/day, about 3 drinks or more per day.  While their analyses suggested a greater protective effect of light or moderate drinking among women, and a lack of an increase in risk with the highest category of intake, there were relatively few females (< 2%, n= 275) in the highest drinking category.  However, there were adequate numbers for rather precise estimates of effect for the other sex-specific categories.

The key finding of their study was that there was a J-shaped curve in all of their spine analytic approaches: linear splines with varying nodes and cubic splines.  Thus, their results strongly support a J-shaped curve for the relation of alcohol intake to CRC: a slightly lower risk of cancer for light to moderate drinkers and an increased risk for subjects reporting an average of 3 or more drinks/day.  Further, they found that their results did not vary by age, obesity, smoking, or family history of CRC.  Overall, the effects of benefits/harms were somewhat more striking for cancers in the distal colon than in the proximal colon.

Specific comments from Forum members 

Reviewer Finkel noted: “Although this paper has on first reading no obvious impairing defects, my initial reflex was, ‘What does it tell us that’s new?’  After a little thought, however, it occurs to me that with breast cancer, colorectal cancer is one of the only two cancers that may, by credible evidence, be associated with increased risk by moderate drinking, and thus a highlighted exemplar of the J-shaped curve.”

Indeed, as noted by Forum member Ellison, this new meta-analysis clearly shows that the increase in colorectal cancer associated with alcohol relates to higher levels of consumption, not consumption averaging no more than 14 drinks/week.  This finding was present in the paper even though the pattern of consumption was not included in these analyses.  It would be expected that if “moderate” drinkers (by weekly averages only) who consumed alcohol only in binges, as on one or two days/week, were removed, a more precise estimate of protection from truly moderate drinking (such as < 2 drinks/day on a regular basis) may have been demonstrated.

Forum member Goldfinger stated: “More studies differentiating moderate from heavy drinking, providing a substrate to define responsible from excessive drinking, where consequences are significantly different, are critically important.  For those of us who follow the literature, commentaries, etc., this is a welcome addition to the literature, underscoring safety and clinical benefit of moderation.”

Dr. Goldfinger continued: “Missing from the present study is an analysis of wine consumers versus consumers of spirits/beer.  This would be most interesting, but certainly difficult data to collect.  And, the moderate drinker with periodic excess would be an interesting subset to look at in this scenario.  The moderate drinker with periodic excess likely reflects many drinkers.  Is the J-curve protected, shifted, or eliminated?”

Reviewer Skovenborg stated: “I agree with the initial comments by other members of the Forum on this study on CRC and alcohol intake.  The study has many of the advantages of the case-control study design and also some of the drawbacks.  One of the problems with case-control studies is information bias: the participants trying to remember their alcohol intake, where the CRC diagnosis may affect the memory – especially in patients that are aware of the association of CRC and alcohol.”  Other reviewers noted that the subjects considered in this analysis included both those from case-control and cohort studies, and the authors stated that there were no differences between the two groups.

Skovenborg continued: “Another issue is the problems with determination of alcohol exposure. The authors write: ‘When studies reported alcohol consumptions in terms of servings, we converted this assuming an alcohol content of 14 g/serving.’  However, the 14 g alcohol/serving is the US standard drink definition, while Finland (221 cases, 141 controls) has a standard drink of 12 g alcohol; Germany (2879 cases, 2325 controls) 12 g; Australia (780 cases, 699 controls) 10 g; Canada (319 cases, 686 controls) 13.5 g; and Sweden (566 cases, 1434 controls) 12 g alcohol/standard drink (IARD).

“The issue of drinking pattern (and also beverage preference) is very important and not addressed in the study.  Goldfinger presents an interesting question: ‘The moderate drinker with periodic excess likely reflects many drinkers:  Is the J-curve protected, shifted, of eliminated?’ The conclusion of a Danish observational study of the effects of occasional binge drinking suggests that occasional excess by regular moderate drinkers may not be a problem, as Skov-Ettrup et al reported:: ‘Among light-to-moderate alcohol drinkers, binge drinking was not associated with increased risk of IHD and all-cause mortality.’”

Reviewer De Gaetano stated: “A J-shaped curve has been observed almost invariably when correlating different doses of alcohol consumption and fatal/non fatal cardiovascular events or all-cause mortality.  It is a diffuse belief in the literature that the relationship between alcohol consumption and cancer is, in contrast, direct and linear, and there is no dose that may be associated with some protection (de Gaetano & Costanzo).  The present study offers strong support to the fact that a window of health benefit (of different breadth, according to different conditions) does exist even for a diffuse cancer type such as CRC.  Such a benefit is expressed by a J curve.”

Reviewer Van Velden noted: “It should be stressed that a healthy lifestyle may include moderate consumption of alcohol except for those individuals with a genetic abnormality that indicates they cannot metabolize ethanol.  All other confounders such as diet, activity profile, smoking, obesity, etc., should be taken into consideration to make meaningful conclusions.”

Potential mechanisms for a protective effect of moderate alcohol on risk of CRC:  Forum member Mattivi wrote: “The reported different effect on proximal vs distal colon raises the question of the mechanisms.  Actually, a number of active, beverage specific metabolites other than ethanol can reach the intestine.  In particular, our attention goes to the several microbial catabolites which can be formed via our host microbiota in the colon.  Several gamma-valerolactones and valeric acids are formed in high concentration in the colon via cleavage of procyanidins after intake of red wines.  Dozens of catabolites of other phenolics are detected (at lower concentration) in the colon after intake of white wines or beer.  All these are absent or at baseline levels in the case of intake of spirits.”  Mattivi concluded: “After this interesting meta-analysis it could be appropriate to further investigate the putative mechanisms.  I would not be surprised if compounds other than alcohol would play a role.”

References from Forum Critique

de Gaetano G, Costanzo S.  Alcohol and Health: Praise of the J Curves.  J Am Coll Cardiol 2017;70:923-925. doi: 10.1016/j.jacc.2017.07.710.

IARD (http://www.iard.org/resources/drinking-guidelines-general-population).

Skov-Ettrup LS, Eliasen M, Ekholm O, Grønbæk M, Tolstrup JS.  Binge drinking, drinking frequency, and risk of ischaemic heart disease: A population-based cohort study.  Scand J Public Health 2011;39:880.  DOI: 10.1177/1403494811425605.

Forum Summary

While heavy alcohol consumption is recognized as a contributor to the development of colorectal cancer (CRC), an increasing number of studies suggest that moderate alcohol consumption may not have an adverse effect on the risk of CRC, or even decrease the risk.

The present paper, a meta-analysis of 16 studies on the relation of alcohol consumption to invasive CRC, was conducted by leading investigators in the field.  Forum members consider this to be a well-done study, with a large number of subjects, the ability to use individual level data in their meta-analysis, and the use of appropriate and complete analytic techniques.

The key finding of their study was that there was a J-shaped curve evident in all of their analytic approaches; their results strongly support a J-shaped association for the relation of alcohol intake to CRC: a slightly lower risk of cancer for light to moderate drinkers and an increased risk for subjects reporting an average of 3 or more drinks/day.  Further, they found that their results did not vary by age, obesity, smoking, or family history of CRC.  Overall, the effects of benefits/harms were somewhat more striking for cancers in the distal colon than in the proximal colon.

Potential mechanisms for heavy alcohol consumption being related to an increase in risk were discussed by the authors.  As for moderate drinking decreasing the risk, potential mechanisms are discussed by a Forum member in this critique.  He describes active, beverage-specific metabolites other than ethanol that reach the intestine.  In particular, our attention goes to the several microbial catabolites which can be formed via our host microbiota in the colon.

The finding of a J-shaped curve between alcohol and CRC is similar to the relation seen for a number of other diseases, including coronary artery disease, ischemic stroke, dementia, and total mortality. While a linear relation between heavy drinking and certain upper aero-digestive tract cancers has been shown, for certain cancers there may be a J-shaped curve of effect, with reductions in risk for light-to-moderate drinkers.

Comments in this review by the International Scientific Forum on Alcohol Research were provided by the following members:

David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa

Susan J van Rensburg, PhD, Department of Pathology, Stellenbosch University, Tygerberg, South Africa

Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway

Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Adjunct Senior Lecturer at the University of Adelaide, South Australia

Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Fulvio Mattivi, MSc, CAFE – Center Agriculture Food Environment, University of Trento, via E. Mach 1,  San Michele all’Adige, Italy

Imke Janssen, PhD, Department of Preventive Medicine, Rush University Medical Centre, Chicago, IL, USA.

Tedd Goldfinger, DO, FACC, Desert Cardiology of Tucson Heart Center, University of Arizona School of Medicine, Tucson, AZ, USA

Harvey Finkel, MD, Hematology/Oncology, Retired (Formerly, Clinical Professor of Medicine, Boston University Medical Center, Boston, MA, USA)

R. Curtis Ellison, MD, Professor of Medicine, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA

Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy