Critique 091: Light drinking may relate to increase in risk for certain cancers — 18 September 2012

Bagnardi V, Rota M, Botteri E, Tramacere I, Islami F, Fedirko V, Scotti L, Jenab M, Turati F, Pasquali E, Pelucchi C, Bellocco R, Negri E, Corrao G, Rehm J, Boffetta P, La Vecchia C.  Light alcohol drinking and cancer: a meta-analysis.   Pre-publication: Annals of Oncology 2012; doi:10.1093/annonc/mds337

Authors’ Abstract

Background:  There is convincing evidence that alcohol consumption increases the risk of cancer of the colorectum, breast, larynx, liver, esophagus, oral cavity and pharynx.  Most of the data derive from studies that focused on the effect of moderate/high alcohol intakes, while little is known about light alcohol drinking (up to 1 drink/day).

Patients and methods:  We evaluated the association between light drinking and cancer of the colorectum, breast, larynx, liver, esophagus, oral cavity and pharynx, through a meta-analytic approach.  We searched epidemiological studies using PubMed, ISI Web of Science and EMBASE, published before December 2010.

Results:  We included 222 articles comprising ∼92 000 light drinkers and 60 000 non-drinkers with cancer.  Light drinking was associated with the risk of oropharyngeal cancer [relative risk, RR = 1.17; 95% confidence interval (CI) 1.06–1.29], esophageal squamous cell carcinoma (SCC) (RR = 1.30; 95% CI 1.09–1.56) and female breast cancer (RR = 1.05; 95% CI 1.02–1.08).  We estimated that ∼5000 deaths from oropharyngeal cancer, 24 000 from esophageal SCC and 5000 from breast cancer were attributable to light drinking in 2004 worldwide.  No association was found for colorectum, liver and larynx tumors.

Conclusions: Light drinking increases the risk of cancer of oral cavity and pharynx, esophagus and female breast.

Forum Comments

The majority of observational studies have shown that alcohol intake, especially heavy drinking, increases a number of upper-aero-digestive tract (UADT) and other cancers, and even moderate drinking is associated with a slight increase in the risk of breast cancer.  This paper compares the effects between light drinkers (an average reported intake of up to 1 typical drink/day) versus “non-drinkers” in terms of relative risks for a number of types of cancer.

Specific concerns about the analyses in this paper:  While the specific analyses appear to be appropriately carried out, there were a number of concerns by Forum reviewers about the data used for such analyses in this paper.  These include the following:

1.  Misclassification of alcohol intake could have affected the results.  The reference group of “non-drinkers” includes both ex-drinkers and never drinkers, which could bias the results in either direction.  If the “non-drinkers” included lots of ex-heavy drinkers, and alcohol increases cancer risk, the differences in cancer risk between light drinkers and “non-drinkers” would be decreased, and the risk associated with light drinking an underestimate of the true effect.  (The risk would presumably have been greater if only never drinkers were in the reference group.)  On the other hand, if the self-reported “light drinkers” included a large number of heavier drinkers who underreported their intake, the increases in the risk associated with light drinking would presumably be less than that calculated in the present study.

2.  There is no consideration in these analyses of the duration of alcohol consumption or of previous drinking for “non-drinkers” or “light drinkers.”  However, it is known that, unlike the mainly short-term effects of alcohol on cardiovascular disease, the relation of alcohol intake to cancer is usually found to reflect long-term exposure.

3.  As acknowledged by the authors, they were unable to adjust for the pattern of drinking or for under-reporting of alcohol consumption (such as heavier drinkers reporting only light drinking).  This could also strongly affect the estimates of the effect attributed to “light drinking.”

4.  While the authors reported differences between results of cohort studies and case-control studies, they apparently did not adjust for type of study in their main analyses.  The differences between these two study designs are particularly large for cancers of the oral cavity and pharynx; the RR is 1.22 (95% CI 1.11, 1.35) for case-control studies and 1.01 (CI 0.70, 1.45) for cohort studies.  The estimated world-wide numbers of cancer cases attributable to light drinking are based on their overall data, and are apparently not adjusted for study design. 

5.  The authors report large differences for certain cancers according to geographical areas.  For example, the calculated effects of light drinking (versus non-drinking) on esophageal cancer gave risk ratios (RR) of 1.05 for Europe, 1.21 for North America, but 1.49 for Asia.  However, these are apparently not adjusted for by the authors in calculating their world-wide attributable risk estimates. 

6.  The striking differences between “low drinking” and “any-dose alcohol” in this paper show that the increases in risk associated with light drinking were many times less that the risk for “any dose” alcohol (indicating that heavier drinking is what is primarily associated with these cancers).  For example, for men the attributable risk of cancers of the oral cavity/pharynx reported in this paper is 1.5 for low alcohol intake and 33.5 for any dose alcohol.  These striking differences are not discussed by the authors.

Differences in results from studies adjusted for important confounders (especially smoking) and those not adjusting for such factors:  Only 46% of the studies included in this meta-analysis adjusted their estimates of effect of alcohol for potentially confounding lifestyle factors of subjects.  In sensitivity analyses based on adjustments for confounding, estimates of the effect of light drinking on oral and pharyngeal cancer were higher for those based on unadjusted studies than for those based on individual studies that adjusted for potential confounders.  For example, the RR estimates for these cancers were 1.21 (95% 1.06, 1.39) for unadjusted studies but 1.11 (CI 1.00, 1.24) for estimates based on adjusted studies.  Much previous data suggest that for UADT cancers, estimates of alcohol’s effects that do not adjust for smoking are essentially meaningless.

The authors acknowledge that the risk of cancer “may be modulated by genetic factors, such as variants in genes of alcohol metabolism, folate and methionine metabolism, and DNA repair.”  However, such data were not available to permit their adjusting for such factors (which limits the applicability of their results when considering the risk of individuals).

Comments on the overall conclusions of the authors:  The net health effects of light drinking are conspicuously absent from this paper, which does not state anywhere that despite an increase in certain cancers reported in this paper, light drinkers overall have considerably lower rates of much more common cardiovascular diseases; further, light drinkers live longer than non-drinkers.  As stated by a Forum reviewer:  “The authors acknowledge the numerous limitations of their study, which suggests that they should have been more cautious in their conclusions.  Each time I read that the small increase in certain cancers (such as breast cancer) associated with light drinking is a matter of public health, I often see that there is no calculation of potential beneficial effects of light drinking in reducing cardiovascular diseases, which are many-times more common than the specific cancers studied in this paper.”

Stated another reviewer, “Clearly the issue is that the authors are focusing on the few cancers that are increased by alcohol and ignoring all of the improved health outcomes of light drinking.  So, while their data are likely to be ‘true,’ their approach ignores, apparently deliberately, the many diseases that are mitigated and reduced by light alcohol consumption, as well as the reduction in total mortality.  It is unfortunate that this approach is often accepted by journals in instances showing problems, but not when showing benefits, associated with light drinking.”

Forum reviewer Skovenborg provided comments that echoed those of many reviewers.  He stated, “The most worrying problems with the meta-analysis are the following:

(a.)  No ability to identify differences in drinking patterns; the effect of local acetaldehyde production might be very different from a glass of wine with food each evening compared with effects of binge drinking on a Saturday night.

(b.)  Information bias, especially underreporting of alcohol intake, as acknowledged by the authors in the discussion.  Underreporting by heavy drinkers could place many such subjects in the ‘light-drinking’ group and weaken the purported increase in risk of certain cancers from small amounts of alcohol.

(c.)  A large degree of heterogeneity between studies, which implies that the conclusions of such analyses should be cautious.  Such heterogeneity could tend to increase the apparent risk for certain cancers, but could also lead to no increase in risk being seen for other types of cancer.

(d.)  A question of external validity; most of the studies were case-control studies from North America rather than prospective studies that generally have less bias in their exposure data (alcohol intake). 

(e.)  The authors offer a possible biological pathway, local acetaldehyde production, for a causal relation between alcohol consumption and upper digestive tract cancers.  They also mention increases in estrogen, androgen, and plasma insulin-like growth factors to explain the slight increase in risk of breast cancer.  However, the association of light drinking with such factors is often found to be negligible.”

Skovenborg concludes: “The many caveats are not reflected in the unambiguous conclusion of the authors, and you may be sure that in the near future some public health authorities are going to make full use of the conclusions in their warning campaigns against even light alcohol consumption – without mentioning any caveats.”

Forum Summary

A meta-analysis by a distinguished group of scientists was designed to determine if “light drinking” (defined as an average of up to 1 drink per day) was associated with the risk of certain cancers that have been shown in previous studies to be associated with the risk of cancer.  The authors concluded that while the risk of these cancers was only slightly increased from such drinking, there were detectable increases in cancers of the oral cavity and pharynx, esophagus and female breast.  They report no increase in the risk of cancers of the colorectum, liver, and larynx to be associated with such drinking.

Forum reviewers were concerned about a number of aspects of this study.  While the statistical methodology was correct and done appropriately, the fact that the investigators (1) included both ex-drinkers and never drinkers in the reference group; (2) could not separate the effects of regular light drinking from binge drinking; (3) had no data on the duration of alcohol consumption at different levels; (4) did not adjust their analyses according to geographic region or type of study (both of which had large estimated effects on cancer risk); and (5) did not adjust their estimates of effect by other lifestyle habits, including smoking.  All of these factors tend to weaken the implications of their results.

Forum reviewers were also concerned that despite the acknowledged limitations of their data, the authors present conclusions indicating that even light drinking increases the risk of certain cancers without commenting on the net health effects.  They present only the effects on cancer (which was the topic of the meta-analysis) but do not comment on the overall or net health effects of light drinking: a marked reduction in the risk of much more common diseases, especially cardiovascular diseases, and a longer lifespan.  Further, the lack of data on genetic patterns, folate intake, and other lifestyle factors makes it difficult to apply their findings to individual subjects.  The Forum considers that while their analyses may be helpful in understanding associations between alcohol and cancer, the many limitations of this study indicate that it can provide only incomplete information on light alcohol consumption to be used as a basis for making recommendations to the public. 

                                                                      *                      *                      *

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

Giovanni de Gaetano, MD, PhD, Research Laboratories, Catholic University, Campobasso, Italy

Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA

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

Andrew L. Waterhouse, PhD, Marvin Sands Professor, Department of Viticulture and Enology, University of California, Davis; Davis, CA, USA

Fulvio Ursini, MD, Dept. of Biological Chemistry, University of Padova, Padova, Italy

Pierre-Louis Teissedre, PhD, Faculty of Oenology – ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France

Creina Stockley, PhD, MBA, Clinical Pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia

Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA

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

Francesco Orlandi, MD, Dept. of Gastroenterology, Università degli Studi di Ancona. Italy

Erik Skovenborg, MD, Scandinavian Medical Alcohol Board, Practitioner, Aarhus, Denmark

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