Part I: Alcohol and breast cancer – a study in weak associations by Erik Skovenborg MD

Alcohol and breast cancer – a study in weak associations by Erik Skovenborg MD
Alcoholic beverages were declared “carcinogenic to humans” (Group 1) by the IARC Monographs Programme, first in 1988 and then again in 2007 and 2010. (IARC. Alcohol drinking. IARC Monogr Eval Carcinog Risks Hum 1988;44:1-378.) In the same year Harris et al. published a study on breast cancer and alcohol consumption. (Harris RE et al. Breast cancer and alcohol consumption – A study in weak associations. JAMA 1988;259:2867-71.) “Recent epidemiologic investigations have indicated that moderate alcohol consumption slightly increases the risk of breast cancer. However, only a weak association is suggested, with reported odds ratios generally being less than 2.0 for women who drink vs those who abstain.” The results from the Harris et al. case-control study of 1467 women with breast cancer and 10 178 sex- and age-matched hospital controls “did not provide compelling evidence that alcohol has a role in this malignancy”.
Relative risks below 3 are considered moderate or weak – in comparison smokers’ risk of lung cancer is 20 or more times higher than that of non-smokers. (Wynder EL. Workshop on guidelines to the epidemiology of weak associations. Prev Med 1987;16:139-141.) “In the workshop on the epidemiology of weak associations various problems were reviewed, including case-control selection, confounders, biases, subgroup analysis, criteria of judgment, and biologic plausibility. Clearly, when it comes to alcohol consumption and breast cancer, we are dealing with a weak association. Findings are not consistent from study to study. There is no definitive dose response. The association appears to be limited to certain subgroups, and there appears to be a lack of external consistency in that countries such as France and Italy, with the highest per capita intake of alcohol (about 16 and 12 L, respectively), do not have correspondingly high rates of breast cancer (about 16 and 15 deaths per 100000, approximate rank = 50th percentile).”
Alcoholic drinks and the risk of cancer
The Expert Panel investigating alcohol and breast cancer associations in the “Continuous Update Project Expert Report 2018” was less sceptical. (World Cancer Research Fund/American Institute for Cancer Research. Continuous Update Project Expert Report 2018. Diet, nutrition, physical activity and breast cancer. Available at dietandcancerreport.org) The CUP Panel concluded: Consumption of alcoholic drinks is a convincing cause of postmenopausal breast cancer and a probable cause of premenopausal breast cancer. The dose-response meta-analysis showed a significant 5% increased risk per 10 grams of ethanol per day: RR 1.05 (1.02-1.08). The dose-response meta-analysis for postmenopausal breast cancer showed a significant 9% increased risk per 10 grams of ethanol per day: RR 1.09 (1.07-1.12).
Cancer is caused by an interaction of risk factors, including internal (genetic susceptibility and hormonal factors) and external factors (lifestyle and environmental factors). Breast cancer is the most common cancer in women worldwide and the fifth most common cause of death from cancer in women, so the association of alcohol intake and breast cancer risk has to be taken seriously. The question is how to interpret the statement on page 59 in the “Alcoholic drinks and the risk of cancer” report (World Cancer Research Fund/American Institute for Cancer Research. Continuous Update Project Expert Report 2018. Alcoholic drinks and the risk of cancer. Available at dietandcancerreport.org): “Current evidence does not identify a generally “safe” threshold for consumption of alcoholic drinks for breast cancer. That is, there does not seem to be a threshold below which an effect on cancer risk is not observed.” However, there is some confusion as to the alcohol content of a drink in the texts. In the CUP Expert Report 2018 “Alcoholic drinks and risk of cancer” the alcohol content of a drink is presented with three different figures: 1) The intake of drinks was rescaled to grams of ethanol per day using 13 grams as the average content of one drink; 2) The dose-response analysis is reported as risk per 10 grams increase in alcohol consumed per day; and 3) The association of alcohol intake and kidney cancer is based on evidence for alcohol intakes up to 30 grams per day (about two drinks a day).
According to WHO no generally “safe” threshold for consumption of alcoholic drinks for breast cancer means “that any reduction in alcohol consumption will be beneficial for health through the reduction of cancer risk”. (World Cancer Report 2014). The WHO Report do not take account of the robust J-shaped relation between moderate alcohol consumption, CHD and all-cause mortality. (Skovenborg, E, Grønbæk M, Ellison RC. Benefits and hazards of alcohol − the J-shaped curve and public health. Drugs and Alcohol Today 2020:21(1)54-69.) In the large EPIC study (268 442 European women) light to moderate intake of alcohol was not related to incidence of alcohol-related cancer in women. With lifetime light alcohol users of ≤ 1 g/day as reference group instead of nondrinkers (“because these women consume alcohol at a level that is so low and infrequent that it presumably could not affect disease risk”) lifetime alcohol consumption of one drink per day in women was associated with lower risk of all-cause mortality: HR 0.92 (0.86-0.98). (Bergmann et al. The association of pattern of lifetime alcohol use and cause of death in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Int J Epidemiol 2013;42:1772-90.)
Causality, bias and confounders
With the one drink per day limit, that worked so favourably for the EPIC-women adhering to the limit, incorporated in many sensible drinking guidelines, an obvious question is whether WHO (and many national Cancer Societies) really have sufficient valid evidence for a message of total abstinence. “It is best not to drink alcohol,” American Cancer Society recommends in its latest cancer prevention guidelines. With the 18th-century empiricist David Hume’s observation that “proof is impossible in empirical science” in mind, how do epidemiologists separate out the causal from the noncausal explanations? (Rothman et al. Causation and Causal Inference in Epidemiology. Am J Public Health 2005;95:S144-S150.) “Scientific evidence can usually be viewed as a form of measurement. If an epidemiologic study sets out to assess the relation between exposure to alcoholic drinks and breast cancer risk, the results can and should be framed as a measure of causal effect, such as the ratio of the risk of breast cancer among drinkers to the risk among nondrinkers. Like any measurement, the measurement of a causal effect is subject to measurement error. In addition to statistical errors, the measurement error subsumes problems that relate to study design, including subject selection and retention, information acquisition, and uncontrolled confounding and other sources of bias. Nearly every study will have nearly every type of error. The real issue is to quantify the errors.” First, one has to discriminate between evidence for no causal relationship, and no evidence of a causal relationship. The former expresses an important piece of evidence that may have substantial consequences, whereas the latter simply expresses lack of knowledge. Because of the by far higher demands on quality and size of studies set out to dismiss the assumption of carcinogenicity, there is an inherent imbalance of classification concerning carcinogenicity and lack of carcinogenicity. (Kundi M. Causality and the Interpretation of Epidemiologic Evidence. Environ Health Perspect 2006;114:969-74.)
Cancer epidemiology is, to a large extent, the determination of small effects and weak associations and poses major challenges. (Boffetta P. Causation in the presence of weak associations. Critical Reviews in Food Science and Nutrition 2010;50:13-16.) So let us apply thorough criticism of the evidence on light drinking and risk of breast cancer. A commonly used set of criteria, proposed by Bradford Hill, was an expansion of a set of criteria offered previously in the landmark surgeon general’s report (1964) on smoking and health. Hill suggested that the following aspects of an association be considered in attempting to distinguish causal from noncausal associations: (1) strength, (2) consistency, (3) specificity, (4) temporality, (5) biological gradient, (6) plausibility, (7) coherence, (8) experimental evidence, and (9) analogy. These criteria should be considered only as aids to thought and practical tools; in fact, scientific causal assessment is a never-ending process centered around measurement of the exposure-disease relationship.
Strength. Strong associations are more likely to be causal than weak associations because, if they could be explained by some other factor, the effect of that factor would have to be even stronger than the observed association and therefore would have become evident. Weak associations, on the other hand, are more likely to be explained by undetected biases. The alcohol-breast cancer association is weak and several bias and effect modifiers have been identified.
Information bias: Underreporting of alcohol intake is a general and serious problem.
• It is generally agreed that alcohol consumption is substantially underestimated when estimations are based on self-reports. In most studies self-reported consumption amounts to only 40-60% compared to sales figures. (Høyer et al. The Svalbard study 1988-89: a unique setting for validation of self-reported alcohol consumption. Addiction 1995;90:539-44.)
• In a study of alcohol measures derived from the Cognitive Lifetime Drinking History among 2142 respondents, current alcohol consumption accounted for only about 10-25% of the variability in lifetime alcohol consumption. (Russell et al. Relations among alcohol consumption measures derived from the Cognitive Lifetime Drinking History. Drug Alcohol Rev 1998;17:377-87.) Most studies that have quantified the effect of alcohol of the risk for cancers of the female breast have captured only current drinking.
• The actual Spanish consumption was 9.5 l of pure alcohol/person-year in 2011, however the coverage of WHO availability, Tax Agency availability, self-reported purchases, and self-reported consumption was 99.5, 99.5, 66.3, and 28.0 %, respectively. To obtain conservative estimates of alcohol-attributable disease burden, self-reported consumption should be shifted upwards by more than 85 % of Tax Agency or WHO availability figures. The underreporting identified can probably also be found worldwide, thus much empirical work remains to be done to improve estimates of per capita alcohol consumption. (Sordo et al. Estimating average alcohol consumption in the population using multiple sources: the case of Spain. Population Health Metrics 2016;14:21.)
• In a cohort of 127,176 persons, participants reporting light–moderate drinking had increased cancer risk in this cohort: HR for risk of any cancer were 1.10 (1.04–1.17) at <1 drink per day and 1.15 (1.08–1.23) at 1–2 drinks per day. However, the increased risk of cancer was concentrated in the stratum suspected of underreporting. (Klatsky et al. Moderate alcohol intake and cancer: the role of underreporting. Cancer Causes Control 2014;25(6):693-9.)
Detection bias: The increased risk of incident breast cancer in moderate alcohol consumers may be overestimated by more intensive screening.
• In the Maryland Behavioral Risk Factor Survey of 946 women ages 50 years and older, women who drank alcoholic beverages had higher mammography rates than nondrinkers: OR = 1.37 (1.03–1.83). (Fredman et al. Cigarette Smoking, Alcohol Consumption, and Screening Mammography among Women Ages 50 and Older. Preventive Medicine 1999;28:407-17.)
• Women who drank alcoholic beverages had significantly higher mammography rates than women who abstained from alcoholic beverages among respondents to the 1990 National Health Interview Survey. (Rakowski et al. Integrating behavior and intention regarding mammography by respondents in the 1990 National Health Interview Survey of Health Promotion and Disease Prevention. Public Health Rep 1993;108:605-24).
• In six iterations (2002–2012) of the Behavioral Risk Factor Surveillance System, a telephone survey of US adults conducted by the Centers for Disease Control and Prevention, participants reported their alcohol use and recent screening for several cancers. Moderate alcohol consumers were 84.7% more likely to report mammography screening than non-consumers. Among binge consumers the weighted prevalence of screening was lower than in non-binge consumers (binge vs non-binge moderate consumers 80.5 vs 85.5%). (Mu et al. Alcohol consumption and rates of screening: Is cancer risk overestimated? Cancer causes Control 2016;27:281-89.)
Effect modification: Uniform adjustments for effect modifiers have not been made in the observational studies of alcohol intake and breast cancer risk.
• Folate intake: Prospective studies indicated a U-shaped relationship for the dietary intake of folate intake: Increased dietary folate intake reduced breast cancer risk for women with higher alcohol intake level, but not for those with lower alcohol intake. (Chen et al. Higher dietary folate intake reduces the breast cancer risk: a systematic review and meta-analysis. British Journal of Cancer 2014;110:2327-38.) A dose-response meta-analysis of observational studies found folate intake inversely correlated with the breast cancer risk when the highest and lowest categories were compared: OR = 0.85 (0.79–0.92), and the dose-response result showed that folate intake had a linear correlation with the breast cancer risk. A higher folate intake correlated with a lower breast cancer risk in premenopausal women: OR = 0.80 (0.66–0.97), but not significantly in postmenopausal women: OR = 0.94 (0.83–1.06). (Ren et al. Association of folate intake and plasma folate level with the risk of breast cancer: a dose-response meta-analysis of observational studies. Aging 2020;12(21):21355-375.)
• Diet: In the Norwegian Women and Cancer study a significantly higher risk of breast cancer was found in women with high consumption of alcohol and low consumption of fruit and vegetables and low consumption of fatty fish. (Engeset et al. Dietary patterns and risk of cancer of various sites in the Norwegian European Prospective Investigation into Cancer and Nutrition cohort: the Norwegian Women and Cancer study. European Journal of Cancer Prevention 2009;18:69-75.) A systematic review and meta-analysis found highest adherence to the MedDiet related to lower risk of cancer mortality in the general population and all-cause mortality among cancer survivors. The association of the MedDiet and cancer was only reflected by inverse associations between overall cancer risk and consumption of fruit, vegetables, whole grains and moderate intake of red wine. (Morze et al. An updated systematic review and meta‑analysis on adherence to Mediterranean diet and risk of cancer. Eur J Nutr 2020.) A Randomized Clinical Trial to evaluate the effect of 2 interventions with Mediterranean diet vs the advice to follow a low-fat diet on breast cancer incidence found multivariable-adjusted hazard ratios vs control: HR = 0.31 (0.13-0.77) for the Mediterranean diet with extra-virgin olive oil group and HR = 0.53 (0.23-1.26) for the Mediterranean diet with nuts group. (Toledo et al. Mediterranean Diet and Invasive Breast Cancer Risk Among Women at High Cardio-vascular Risk in the PREDIMED Trial. JAMA Intern Med 2015;175(11):1752-60.)
• Drinking pattern: After controlling for cumulative alcohol intake, binge drinking was associated with increased breast cancer risk in Danish nurses (Mørch et al. Alcohol drinking, consumption patterns and breast cancer among Danish nurses: a cohort study. European Journal of Public Health 2007;17(6):624-29), US nurses (Chen et al. Moderate Alcohol Consumption During Adult Life, Drinking Patterns, and Breast Cancer Risk. JAMA 2011;306(17):1884-90), Puerto Rico women with a sister with breast cancer (White et al. Lifetime Alcohol Intake, Binge Drinking Behaviors, and Breast Cancer Risk. Am J Epidemiol 2017;186(5):541-49) and Spanish university graduates (Sánchez-Bayona et al. Binge Drinking and Risk of Breast Cancer: Results from the SUN (‘Seguimiento Universidad de Navarra’) Project. Nutrients 2020;12:731). However, information on drinking patterns is almost non-existent in the evidence base for the CUP Expert Report 2018 “Alcoholic drinks and risk of cancer”.
• Physical activity: An association of physical activity and cancer mortality risk was found in an analysis of data from participants aged 30 years and over in Health Surveys for England and the Scottish Health Surveys. Among the physical activity (PA) ≤ 7.5 MET-hour/week group, there was a significant association between alcohol consumption and risk of alcohol-related cancer mortality. Exdrinkers: HR = 1.53 (1.11 – 2.12), drinkers at hazardous levels: HR = 1.47 (1.07 – 2.02) and drinkers at harmful levels: HR = 1.64 (1.07 – 2.52) had significantly higher mortality risks than never-drinkers. The increased risks were eliminated among the PA > 7.5 MET-hour/week individuals. (Feng Y et al. Does adequate physical activity attenuate the associations of alcohol and alcohol-related cancer mortality? A pooled study of 54 686 British adults. Int J Cancer 2020;147(10):2754-63.)

Erik Skovenborg is a Danish physician with a special interest in the health benefits of moderate alcohol consumption. His published work includes In Vino Sanitas, 1990; Lead in Wine throughout the Ages, 1994; Wine and Health – Myths and Facts, 2000. Member of the Social, Scientific and Medical Council of AIM (Alcohol in Moderation) from 1992 and co-founder of the Scandinavian Medical Alcohol Board (SMAB) in 1994. Chairman of the 1996 Copenhagen “Health and Alcohol Symposium” and the 1998 Stockholm “Women and Alcohol Symposium”. For many years Erik Skovenborg has lectured extensively on alcohol and health to medical professionals and the general public and he is currently researching the effects of a moderate consumption of beer, wine and spirits.
In our next edition, the second part of this paper will look at consistency, biological gradient, plausible mechanisms, population attributable fractions and public health issues.