Critique 180: The association of alcohol intake with total mortality risk among women — 9 February 2016

Midlöv P, Calling S, Memon AA, Sundquist J, Sundquist K, Johansson S-E.  Women’s health in the Lund area (WHILA) – Alcohol consumption and all-cause mortality among women – a 17 year follow-up study.  Pre-publication: BMC Public Health 2016;16:22.  DOI 10.1186/s12889-016-2700-2

 Authors’ Abstract

Background:  Alcohol consumption contributes to many negative health consequences and is a risk factor for death. Some previous studies however suggest a J-shaped relationship between the level of alcohol consumption and all-cause mortality. These findings have in part been suggested to be due to confounders. The aim of our study was to analyze the relationship between self-reported alcohol intake and all-cause mortality in women, adjusted for sociodemographic, lifestyle factors and diseases such as diabetes and previous ischemic heart disease.

Methods: All women aged 50–59 years (born between 1935 and 1945) that lived in any of the five municipalities in southern Sweden were invited to participate in a health survey. From December 1995 to February 2000 a total of 6916 women (out of 10,766, the total population of women in 1995) underwent a physical examination and answered a questionnaire. We followed the women from the day of screening until death, or if no event occurred until May 31st 2015. Mortality was ascertained through the national cause-of-death register.

Results: In this study a total of 6353 women were included. Alcohol consumption showed a J-formed relationship with mortality, when adjusted for education, marital status, smoking, BMI, physical fitness, diabetes and ischemic heart disease before screening. Non consumption of alcohol was associated with increased mortality as well as higher levels of consumption, from 12 grams per day and upwards.

Conclusions: There was a clear J-shaped relation between the amount of alcohol consumption and all-cause mortality even after controlling for sociodemography, lifestyle factors and diseases such as diabetes and previous ischemic heart disease. The observed protective effect of light drinking (1–12 grams/day) could thus not be attributed to any of these known confounders.

Forum Comments

The large majority of prospective cohort studies have shown a “J-shaped curve” for the relation between moderate alcohol consumption and the risk of total mortality.  The degree of adjustment for potential confounders, especially previous alcohol intake and other lifestyle factors, varies among studies. This has led some scientists to raise the question as to whether or not the inverse association between alcohol and mortality is due just to confounding rather than to alcohol (e.g., Goulden).

The present study was designed to test the degree to which common potential confounders explain the “J-shaped” relation.  It is based on a large population-based cohort with good follow up.  There was a reasonable percentage of subjects who were non-drinkers, whose mortality was compared with women reporting two levels of drinking [light (up to 12 g/day, about 1 typical drink) and heavier (≥12 g/day)].  Baseline associations with reported alcohol intake clearly show the expected effects on mortality risk of education (inverse), partnership status (lower if married), fitness (better if fit), smoking (bad), and diabetes (lower with moderate drinking), which would support the outcome results.

Both unadjusted and fully adjusted analyses in this study support a beneficial effect of light drinking on mortality.  With spline analyses, the risk of mortality of drinkers up to about 2 typical drinks/day was lower than it was for non-drinkers.  The authors conclude that the observed protective effect of light drinking (1–12 grams/day), even after controlling for sociodemography, lifestyle factors and diseases such as diabetes and previous ischemic heart disease, could not be attributed to any of these known confounders.   Thus, while this study does not show anything new, it does support the J-shaped curve and suggests that the usual possible confounding factors do not explain the curve.

Specific comments by Forum members:  Reviewer Keil stated: “The study seems to be well done.  The results are not new at all, as the J shaped (or U-shaped) curve for alcohol consumption and all-cause mortality is familiar to us.  What is not familiar to me is that the nadir of the J-shaped curve does not reach below 1.0.  Obviously the model is set up in a way that the consumption level 0.1-11.9 gram alcohol per day is taken as the reference, showing that consumption of lower and higher amounts both carry an increased risk.  The study is based on a single measurement of alcohol consumption; changes of alcohol consumption over time are not assessed.  This means that the association between light alcohol consumption and total mortality (a protective effect) might have been diluted.

“The data presented clearly show that abstainers or women consuming more than 12 grams of alcohol per day have an increased risk compared to the 0.1-11.9 alcohol intake group, but the authors do not make a statement on how much total mortality is decreased among light drinkers.  In the meta-analysis by Ronksley et al, you can draw the conclusion and make the statement that light to moderate alcohol consumption reduces the risk of all-cause mortality by 13% (RR=0.87).”   Reviewer Barrett-Connor declared: “Again, a clear J-shaped curve!  The results of this study are very similar to what we’ve already seen.  I agree with the authors’ conclusions.”

Reviewer de Gaetano wrote: “A special interest of this study is the observed protective effect of moderate alcohol in women over 50 (at the enrollment).  This confirms that women over 50 benefit from moderate alcohol consumption, despite the fact that the risk of breast cancer might be slightly increased by even moderate alcohol drinking.”   Forum member Van Velden added: “Nothing was said about an increased risk for breast cancer in woman consuming alcohol, and no data on folate intake were presented.”  [While many studies have suggested that dietary folic acid reduces the risk of breast cancer associated with alcohol consumption, some recent studies do not support such protection from folate (Jung et al).]

Forum member Estruch stated: “The results of this study confirm the ‘J-shaped’ relationship between alcohol intake and all-cause mortality adjusted to age.  Unfortunately, the study protocol did not allow to differentiate between patterns of drinking, since the authors only evaluated weekly consumption of alcohol.  Further, although the authors asked about the consumption of beer, wine, and spirits separately, they put all the beverages in the same box (‘alcohol’) and did not differentiate between them.   Although the conclusions of this study are not new, they support the results of most previous studies on this topic, and confirm that ‘moderate alcohol consumption’ is protective for all-cause mortality.  However, we need more studies on the mechanisms of such effect.”  Reviewer Lanzmann-Petithory also regretted that beverage-specific results were not given.

Forum member Mattivi commented: “This seems to be a well conducted study leading to another confirmative result relating moderate drinking to better mortality.  It is positive here that data were available which allowed the authors to treat the self-reported weekly alcohol consumption at baseline as a continuous variable in the model (while most confounders where dichotomized).  Factors limiting the strength of the study are that a single self-reported value of alcohol consumption was available (which may have changed considerably over 17 years .  .  . as well as the marital status).  Also, the pattern of consumption and diet were apparently not considered.  It can be noted that high educational level, a factor frequently associated with higher income and better diet, was least common among abstainers.”

Reviewer Thelle commented: “I agree with the comments of other reviewers, but two additional issues may be mentioned. Only 64.2% of subjects contacted agreed to participate; the authors should have access to mortality data of the non-participants.  It would be interesting to see whether they differ from the participants to assess a possible selection bias.

“The authors claim (indirectly) that the study has sufficient external validity to allow public health implications.  More important is that the follow-up period in this study starts immediately after base-line. This may induce an element of reversed causality (as described by Gulsvik et al).  The first couple of years after base-line should have been excluded in order to assess whether this would influence the results.”

Forum member Skovenborg had several other concerns about the study: “The authors report that 26% of the women were non-drinkers, but they were not able to separate lifelong abstainers from former drinkers, which could lead to potential confounding by ‘sick quitters.’  Further, in order to achieve statistical significance the authors had to reduce the alcohol drinking categories to three, which turns the 3rd group (≥12 g alcohol/day) into an inhomogeneous group with a large range of weekly alcohol consumption: 109 – 1036 grams.”

Skovenborg continued: “In the questionnaire, only the weekly consumption volume was asked for; thus it was impossible to differentiate the drinking pattern during the week.  Among lifestyle factors, diet was not controlled for.  Also, in the discussion the authors claim that ‘the amount of alcohol intake, where the protective effects exceed the adverse effects, is unclear.’  However, they demonstrate that when alcohol is treated as a continuous variable, alcohol seems to be protective up to an intake of about 20 grams/day, which is in accordance with most published data.”

Reviewer Goldfinger added; “I agree with comments provided, but I am somewhat disturbed by the allegation of the authors that a higher mortality risk is noted beyond a threshold of 12 gm of alcohol per day.  This would not be consistent with other data or critical observation of the French or Italian Mediterranean lifestyle studies.  I suspect underreporting of alcohol consumption among the alcohol consumer group, and skewing of the results to a lower than likely threshold for benefit/risk.”

Reviewer Ellison agreed that under-reporting could increase the level of alcohol intake at which the risk of mortality exceeded that of non-drinkers, and move the “threshold level” for adverse effects higher.  However, he also noted: “The paper presents associations using spline analyses; in these, the fully adjusted data indicate that the point at which the reduced risk of mortality for light-to-moderate drinking reaches a level as high as non-drinkers is at a reported intake of between 20 and 25 grams/day; this is the equivalent of about two typical drinks.  Further, there were apparently too few heavy drinkers to allow the authors to use several categories for higher intake of alcohol (e.g., 12-24, ≥ 24 grams/day, etc.), and their confidence limits for the top category of intake were very wide.  Thus, these data should not be used to judge the specific level at which the risk of mortality among drinkers exceeds the risk of those who are not drinking alcohol.”

References from Forum review

Goulden R.  Moderate Alcohol Consumption Is Not Associated with Reduced All-cause Mortality.  Am J Med 2015; pre-publication.  (A critique of this paper by the Forum is available at

Gulsvik AK, Thelle DS, Mowé M, Wyller TB.  Increased mortality in the slim elderly: a 42 years follow-up study in a general population. Eur J Epidemiol 2009;24:683-690.

Jung S, Wang M, Anderson K, et al.  Alcohol consumption and breast cancer risk by estrogen receptor status: in a pooled analysis of 20 studies.  Int J Epidemiol 2016. doi: 10.1093/ije/dyv156  pii: dyv156 (pre-publication).

Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA.  Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis.  BMJ 2011;342:d671. doi: 10.1136/bmj.d671.

Forum Summary

The follow up of more than 6,000 women in a population-based cohort in an area of southern Sweden was used to estimate how baseline levels of alcohol consumption, at age 50-59 years, related to total mortality risk over the subsequent 17 years.  At baseline, 26% of women reported no alcohol consumption while 13% reported an average of 12 or more grams of alcohol per day.  Thus, the large majority (61%) were light drinkers, reporting the equivalent of no more than one drink per day.  The authors report that, even when adjusting for education, marital status, smoking, BMI, physical fitness, diabetes and ischemic heart disease before screening, mortality risk during follow up was significantly higher among non-drinkers and heavier drinkers than among women reporting the equivalent no more than one typical drink per day.  They state that their analyses thus support a “J-shaped” association between alcohol and total mortality risk, and that “The observed protective effect of light drinking (1–12 grams/day) could thus not be attributed to any of these known confounders.”

Forum members agreed this was a well-done study, but noted that the authors were unable to adjust for the pattern of drinking (regular versus binge), previous drinking among abstainers, or potential changes in drinking during follow up.  Also, beverage-specific results are not presented.  Further, the data presented in this study are not adequate to judge the specific level of alcohol consumption at which total mortality risk for drinkers reaches or exceeds the risk for non-drinkers; in other words, the “threshold” for adverse effects of alcohol on mortality cannot be determined very well.  Nevertheless, the results of these analyses are very consistent with most long-term follow-up studies and support a “J-shaped” association between alcohol and the risk of total mortality.

*                    *                    *                    *                      *                      *                     *                      *                     *                    *

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

Elizabeth Barrett-Connor, MD, Distinguished Professor, Division of Epidemiology, Department of Family Medicine and Public Health and Department of Medicine, University of California, San Diego, La Jolla, CA USA

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

R. Curtis Ellison, MD, Professor of Medicine & Public Health, Boston University School of Medicine, Boston, MA, USA

Ramon Estruch, MD, PhD.  Associate Professor of Medicine, University of Barcelona, Spain

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

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

Ulrich Keil, MD, PhD, Professor Emeritus, Institute of Epidemiology & Social Medicine, University of Muenster, Germany

Dominique Lanzmann-Petithory,MD, PhD, Nutrition/Cardiology, Praticien Hospitalier Hôpital Emile Roux, Paris, France

Fulvio Mattivi, MSc, Head of the Department of Food Quality and Nutrition, Research and Innovation Centre, Fondazione Edmund Mach, in San Michele all’Adige, Italy

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

Dag S. Thelle, MD, PhD, Senior Professor of Cardiovascular Epidemiology and Prevention, University of Gothenburg, Sweden; Senior Professor of Quantitative Medicine at the University of Oslo, Norway

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