Critique 222. A new analysis relating alcohol intake to coronary heart disease – 20 December 2018
Fan AZ, Ruan WJ, Chou SP. Re-examining the relationship between alcohol consumption and coronary heart disease with a new lens. Pre-publication. Preventive Medicine 2019;118:336=343.
Moderate alcohol consumption has been related to lower risk of coronary heart disease (CHD) in the literature. To examine whether alcohol drinking during the past12 months and heaviest drinking period were diﬀerentially associated with the risk of CHD, we designed a case-control study using a population-based health survey of U.S. adults conducted from 2012 to 2013. Respondents who reported to have doctor-ascertained CHD served as cases (n=1671), and those free of CHD and other alcohol-related health conditions served as controls (n=17,629) in logistic regressions. Sex-speciﬁc quartiles of average daily ethanol intake were ascertained and calculated for the past 12 months and during the period of heaviest lifetime drinking. We further split current drinkers into reducers and non-reducers (past 12 months relative to the heaviest drinking period) to examine CHD risk proﬁles in association with the 12-month drinking level.
Current-drinker reducers (AOR, 95% CI=1.57 [1.10–2.27] for men; AOR, 95% CI=1.33 [1.02–1.72] for women) and former drinkers (AOR, 95% CI=2.06 [1.43–2.97] for men; AOR, 95% CI=1.51 [1.19–1.92] for women) more often had CHD than lifetime abstainers. Male heavy drinkers during the heaviest drinking period (AOR, 95% CI=2.25 [1.52–3.32]) were more likely to manifest CHD than lifetime abstainers. In addition, individuals with diagnosed CHD were signiﬁcantly more likely to have reduced drinking in the past. A change in alcohol consumption over the life course among former and current drinkers may distort the true alcohol-CHD relationship.
Data from well-done cohort studies, with follow up observations over decades, have consistently shown that the risks of myocardial infarction and other aspects of cardiovascular disease are lower among light-to-moderate drinkers than among abstainers A few reports have raised questions about such an association because in many early studies the investigators were unable to separate lifetime abstainers from ex-drinkers, some of whom may have stopped drinking due to health problems associated with alcohol abuse or even due the development of cardiovascular disease itself (e.g., Shaper et al). However, many well-done cohort studies have been able to separate ex-drinkers from lifetime abstainers, and the latter continue to show significant and rather large decreases in risk of coronary diseases among moderate drinkers (e.g., Kunzmann et al; Mukamal et al).
Upon seeing the results of the present paper, as perceived by the authors, Forum members first began to evaluate the methodology to see if there were obvious reasons why the results of these analyses are very different from essentially all other scientific publications. It became clear very quickly that this was not only a very biased paper, but many of the assumptions made by the authors are misguided and not supported by scientific data.
Evaluating for CHD in a relatively young cohort: First of all, the study was carried out among relatively young subjects; at the time of the baseline interview, when the assessment of current alcohol intake was collected, men averaged 36.3 years of age and women, 39.9 years. Very few subjects have CHD at a young age, as the mean age of myocardial infarction in the United States is now in the 70s. Yet, the investigators excluded 1,671 subjects as having “CHD” at baseline.
Weakness of diagnostic criteria for CHD: Further, the diagnosis of the main outcome, coronary heart disease, was very weak, being based only on the subject reporting that he/she had been told by a doctor that “you had chest pain or angina, a heart attack or myocardial infarction,” a completely inadequate approach for diagnosing CHD.
Unusual exclusions of subjects from control group: It appears that the investigators excluded from their “control” group anyone who reported any risk factor for CHD, including “high blood pressure, high cholesterol, high triglycerides, diabetes, stroke, rapid heart beat, or tachycardia.” These include factors that are known to be affected, either adversely or favorably, by alcohol consumption Most confusing is that they state that they also excluded any subjects with conditions that “may increase the risk of CHD (e.g., pneumonia, influenza, tuberculosis, liver disease, neurologic conditions, osteoporosis, arthritis, STD) and can be viewed as comorbid conditions related to alcohol consumption.” This makes no sense to reviewers. All in all, there were 17,629 subjects (of the total group of 36,309) remaining for the control group.
Other analytic problems: Other problems appeared in this analysis. The categories for drinkers were “very light drinker,” “light drinker,” “moderate drinker,” or “heavy drinker,” but the authors do not give the specific limits they set for men or women. Instead, they used quartiles of the distribution of each group to compare effects between those in the upper and lower quartiles. Unfortunately, this does not permit comparisons with previous research or guidelines, which are generally presented as so-many typical drinks per day or per week.
The authors attempted to judge past drinking by asking subjects if they had ever consumed more alcohol than during the past year, and those saying yes were classified as “reducers.” Evidently, they did not determine for how many years subjects consumed more alcohol, and they had no way to verify the estimates of earlier drinking. (It is noted that those diagnosed as “reducers” were the drinkers reporting the most current alcohol intake.) Also, while the investigators state that they controlled for smoking, the text states that they divided subjects into groups of current smokers, former smokers, and never smokers. Apparently there was no adjustments for level of smoking or pack-years of smoking.
Forum member Zhang wrote: “It appears to me that the CHD cases are prevalent cases over the past 12 months and alcohol intake is also being assessed over the same past 12 months. So it is likely that reduced alcohol consumption may be due to patients have suffered from CHD. Further, former drinkers are older than abstainers in both men and women, making the comparison very challenging.” Others noted that no data are given for subsequent occurrence of CHD after the period of assessment of current drinking.
Several Forum investigators stated that after reading the methods used by the authors, they felt it unnecessary to look at the results, considering that with these marked methodologic problems any results would be meaningless.
Other specific comments from Forum members: Professor Keil agreed with other Forum members of the problems with the design and implementation of the study. “It is an impressive example of the hasty review process which so many so-called scientific journals apply. How can reviewers accept such a poorly done and highly biased paper? The review process of many journals is really in big disarray – we are obviously dealing with an inflation of submitted manuscripts and with journal editors who have great difficulties in soliciting help from qualified reviewers.” He added, “Why did the authors start an ill-designed and performed case-control study, when we have more than 100 good cohort studies clearly showing that light to moderate alcohol intake reduces the risk of CHD and the risk of ischemic stroke?”
Reviewer Keil listed “three previous well-carried out analyses [two of which (Maclure; Ronksley et al) can be considered as ‘landmark papers’] that make this paper superfluous. We should all fight the inflation of scientific publishing and the publishing of papers which are not only superfluous but utterly wrong. For example, reading and re-reading the deductive meta-analysis by Maclure is always a special experience. Our ‘changes in alcohol’ paper (Wellmann et al) has been included in the excellent meta-analysis by Ronksley et al.
Stated Forum member De Gaetano: “I really was confused by the different groups and sub-groups these authors studied, making it especially difficult to understand the paper. However, I read with pleasure the earlier comments of other reviewers; I agree with all of them.”
Forum member Skovenborg had a number of concerns about the paper: “I agree with the Forum review comments and would add some observations on the question of evidence as stated in the paper. For example, the authors state: ‘It is hard to believe that drinking one drink or more per day on average for women would confer any cardioprotection. In fact, epidemiologic studies usually provide opposite evidence.’ The first author of the present paper (Fan) quoted her own paper, stating ‘Drinking exceeding one drink per drinking day for women has been found to be related to higher prevalence of metabolic syndrome which is a clustering of cardiovascular risk factors (Fan et al., 2008a).’
“Instead, the authors of the present paper could have used a meta-analysis of observational studies on alcohol consumption and the prevalence of metabolic syndrome (Ala’a A et al), which showed that alcohol consumption of less than 40 g/day in men and 20 g/day in women significantly reduced the prevalence of metabolic syndrome. Or they might have quoted Baliunas et al, a paper based on data from 20 cohort studies, which concluded that a U-formed relationship was found for both sexes. In the latter analysis, compared with lifetime abstainers, the relative risk for type 2 diabetes among men was most protective when consuming 22 g/day alcohol (RR 0.82 [0.76-1.00]) and became deleterious at just over 60 g/day alcohol (RR 1.01 [0.71-1.44]). Among women, consumption of 24 g/day alcohol was most protective (RR 0.60 [0.52-0.69]) and became deleterious at about 50 g/day alcohol (RR 1.02 [0.83-1.26]).”
Skovenborg continued: “Fan et al stated in this paper that “The analyses using drinking intensity (drinks per drinking day) and frequency yielded similar patterns.” However, the results of a recent large Danish cohort study told another story, concluding: “Compared with current alcohol consumers consuming <1 day/week, consumption of alcohol on 3–4 days weekly was associated with significantly lower risk for diabetes in men (HR 0.73 [95% CI 0.59, 0.94]) and women (HR 0.68 [95% CI 0.53, 0.88]) after adjusting for confounders and average weekly alcohol amount (Holst et al). Even changes in alcohol consumption and subsequent risk of type 2 diabetes has been studied in a large prospective cohort study by Joosten et al, who stated: ‘A total of 1,905 cases of type 2 diabetes occurred during 428,497 person-years of follow-up. A 7.5 g/day (approximately half a glass) increase in alcohol consumption over 4 years was associated with lower diabetes risk among initial nondrinkers (multivariable hazard ratio [HR] 0.78; 95% CI: 0.60 –1.00) and drinkers initially consuming <15 g/day (HR 0.89; 95% CI: 0.83–0.96), but not among men initially drinking ≥15 g/day (HR 0.99; 95% CI: 0.95–1.02).’”
As expressed by many reviewers, it seems unusual to attempt to use cross-sectional data to try and determine how changes in alcohol intake relate to CHD. Skovenborg noted: “If Fan et al prefer cross-sectional data they might have quoted an analysis of data obtained by linking 13 waves of the National Health Interview Surveys (1997 to 2009) to the National Death Index records through December 31, 2011 (Xi et al). Those authors concluded: ‘Compared with lifetime abstainers, those who were light or moderate alcohol consumers were at a reduced risk of mortality for all causes (light—hazard ratio [HR]: 0.79; 95% confidence interval [CI]: 0.76 to 0.82; moderate—HR: 0.78; 95% CI: 0.74 to 0.82) and CVD (light—HR: 0.74; 95% CI: 0.69 to 0.80; moderate—HR: 0.71; 95% CI: 0.64 to 0.78), respectively. In contrast, there was a significantly increased risk of mortality for all causes (HR: 1.11; 95% CI: 1.04 to 1.19) and cancer (HR: 1.27; 95% CI: 1.13 to 1.42) in adults with heavy alcohol consumption. Binge drinking ≥1 d/week was also associated with an increased risk of mortality for all causes (HR: 1.13; 95% CI: 1.04 to 1.23) and cancer (HR: 1.22; 95% CI: 1.05 to 1.41).’”
Wine consumption as part of a healthy lifestyle: The present study does not discuss how alcohol intake must be considered in conjunction with other lifestyle habits and socio-economic factors, which have large effects on the risk of CHD. Further, the study did not comment on beverage-specific effects; the polyphenols in wine have frequently been shown to provide additional effects to those from alcohol. Forum member Van Velden noted: “It must be acknowledged that the etiology of CHD is multifactorial, that risk factors have a multiplicative effect, and that physicians have to deal with the whole person, not with isolated risk factors. The prevention of CHD is multifacited, and wine consumption, in particular, cannot be seen in isolation from other preventative strategies such as exercise, diet, not smoking and weight loss. Wine as part of a healthy, Mediterranean-like diet and lifestyle, could play a role in this regard. Our group indicated that wine can improve endothelial dysfunction, combat low-grade chronic inflammation of the endothelium, and prevent the premature onset of the metabolic syndrome (Van Velden et al). The health benefits of wine or alcohol are only part of a healthy, enjoyable lifestyle”.
Bias in the analyses and interpretation: To reviewer Ellison, the Discussion in this paper reflects bias of the authors against any potential healthy aspects of alcohol consumption. They indicate repeatedly that they think that any alcohol consumption must be bad for health. “The authors ‘explain away’ every finding that may not support their (probably pre-conceived) conclusions. For example, in discussing the effects of alcohol on risk factors, they quote two studies, their own previous study and another from NHANES, then a report that shows that blood pressure increases with alcohol and state: ‘There is no doubt that these alcohol-induced exacerbated risk proﬁles would accumulate with aging and result in higher tendency of clinical manifestation of CHD’; this statement is not consistent with most previous research. They quote one paper from the National Alcohol Survey in the United States stating that more than half of those who reported never having a drink of any alcoholic beverage in the 1992 survey reported drinking in previous surveys, stating that this indicates: ‘The contaminated reference group may result in an underestimation of alcohol-attributable morbidity and mortality. Thus, they may not represent the optimal reference group to evaluate the beneﬁcial eﬀects of any current drinking level. On the other hand, it is hard to believe that drinking one drink or more per day on average for women would confer any cardioprotection.’ A huge amount of basic science and epidemiologic data do not support such a conclusion.
“Finally, even when their own data indicate potential benefits from moderate alcohol intake, they seek to explain away their results. For example for their findings of protection against CHD for r females consuming more alcohol, they state: ‘In our study, the current drinker non-reducers initiated drinking at an older age and consumed alcohol for a shorter number of years at a lower level; thus, they may not have accumulated suﬃcient dosage exposure to manifest chronic clinical outcomes including CHD. If we assume that alcohol consumption may be beneﬁcial to some extent, the worsened CHD proﬁle among current drinker reducers may be attributed to the reduction in alcohol consumption.’ They conclude: ‘Therefore, it makes sense to attribute reducers’ excess CHD risk to accumulated harms caused by heavier drinking earlier in their lives.’
“Elsewhere they state: ‘. . . the increased risk of CHD among lower-quartile reducers was the underlying driving force to artiﬁcially produce a relatively lower risk of CHD among higher volume quartiles before splitting was done.’ Such reasoning is difficult to follow, and contrasts with most previous research on the topic. It is noted that many of the references given to support these statements are from previous publications by the authors, and do not reflect the massive amount of experimental and epidemiologic data on this topic. The authors end up stating that their conclusions are supported by the Mendelian randomization paper by Holmes et al, even though that paper has been repeatedly shown to be erroneous it its conclusions.
“The authors appear to believe that they are the first group to ever consider changes in alcohol intake over time, ignoring many well-done cohort studies that have repeated assessments of alcohol intake over many decades. Yet, they select for their analyses a study that basically does not have adequate data to be able to judge changes in alcohol.” Reviewer Finkel used few words to express his views: “I would reduce this paper to the trivia that it is, and ignore it.”
References from Forum Critique
Ala’a A et al. Alcohol consumption and the prevalence of metabolic syndrome: A meta-analysis of observational studies. Atherosclerosis 2009;204:624-635.
Baliunas DO et al. Alcohol as a risk factor for type 2 diabetes. Diabetes Care 2009;32:2123-2132.
Holmes MV, et al. Association between alcohol and cardiovascular disease: Mendelian randomisation analysis based on individual participant data. BMJ 2014;349-g4164.. doi: https://doi.org/10.1136/bmj.g4164.
Holst C et al. Alcohol drinking patterns and risk of diabetes: a cohort study of 70,551 men and women from the general Danish population. Diabetologia 2017;60:1941-1950.
Joosten MM et al. Changes in Alcohol Consumption and Subsequent Risk of Type 2 Diabetes in Men. Diabetes 2011;60:74-79.
Kunzmann AT, Coleman HG, Huang WY, Berndt SI. The association of lifetime alcohol use with mortality and cancer risk in older adults: A cohort study. PLoS Med 2018;19;15:e1002585.
Maclure M. Demonstration of Deductive Meta-Analysis: Ethanol Intake and Risk of Myocardial Infarction. Epidemiologic Reviews 1993;15:328-351.
Mukamal KJ, Chen CM, Rao SR, Breslow RA. Alcohol consumption and cardiovascular mortality among U.S. adults, 1987 to 2002. J Am Coll Cardiol 2010;55:1328–1335.
Ronksley PE, Brienp 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.
Shaper AG, Wannamethee G, Walker M. Alcohol and mortality in British men: explaining the U-shaped curve. Lancet. 1988;2(8623):1267-1273.
Van Velden D P et al. The bio-pshyco-social approach to health and disease. SADJ 2003;58;191.193-194.
Wellmann J, Heidrich J, Berger K, Döring A, Heuschmann PU, Keil U. Changes in alcohol intake and risk of coronary heart disease and all-cause mortality in the MONICA/KORA-Augsburg cohort 1987-97. Euro J Cardiovasc Prevent Rehabilitation 2004;11:48-55. DOI: 10.1097/01.hjr.0000118174.70522.20.
Xi B, et al. Relationship of Alcohol Consumption to All-Cause, Cardiovascular, and Cancer-Related Mortality in U.S. Adults. J Am Coll Cardiol 2017;70:913–922.
In an attempt to judge the effects of alcohol on the risk of coronary heart disease (CHD), the authors have carried out analyses using a sample of subjects from the National Epidemiologic Survey on Alcohol and Related Conditions III (NESARC-III). They based their exposure to current alcohol use on a single assessment at baseline of self-reported consumption during the past year, when the subjects were an average of 38.9 years of age. In addition they asked subjects if they had consumed larger amounts of alcohol when they were younger.
Their assessment of CHD was based on answers to questions at baseline as to having been told by a doctor that they had chest pain or CHD, apparently within the same 12-month period when they were asked about current alcohol consumption. Nothing in the paper indicates that the authors had data on the subsequent development of CHD during a follow-up period. When relating reported alcohol intake to CHD, they excluded 1,671 who reported “previous CHD,” although for this relatively young group it is questionable how many actually had CHD, as no data on validation of the diagnosis was done. For unclear reasons, for their comparison group they excluded any subjects with conditions that “may increase the risk of CHD (e.g., pneumonia, influenza, tuberculosis, liver disease, neurologic conditions, osteoporosis, arthritis, STD) and can be viewed as comorbid conditions related to alcohol consumption.”
Given that the exposure (alcohol) and outcome (CHD) were apparently from the same time period, there is reason to worry about reverse causation (i.e., subjects may have developed CHD and changed, either decreased or increased, their alcohol consumption during the same period).. As for using the self-reported amount of alcohol consumed at some time in the past as the exposure, there is a large chance of serious misclassification of exposure due to recall bias and bias according to current alcohol intake (as subjects currently reporting greater amounts of alcohol were more likely to report even higher intake in the past).
With only a single assessment of current alcohol intake, an inadequate assessment of long-term alcohol intake, a lack of data during a follow-up period for the main outcome, CHD, and no validation of the outcome, Forum members considered that this was an unusual cohort to use when attempting to judge the relation of changes in alcohol consumption to the development of CHD. The necessary elements for a useful analysis were not available to these investigators using this population sample.
In any case, results from many well-done long-term cohort studies with repeated assessments of alcohol and validated CHD outcomes during follow up have already provided key data on this topic: they almost uniformly demonstrate that light-to-moderate drinkers who do not binge drink have significantly lower risk of developing CHD. The present analysis does not provide data that would question such findings.
* * * *
Comments on this critique by the International Scientific Forum on Alcohol Research were provided by the following members:
Ulrich Keil, MD, PhD, Professor Emeritus, Institute of Epidemiology & Social Medicine, University of Muenster, Germany
Harvey Finkel, MD, Hematology/Oncology, Retired (Formerly, Clinical Professor of Medicine, Boston University Medical Center, Boston, MA, USA)
Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy
David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa
Yuqing Zhang, MD, DSc. Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.
R. Curtis Ellison, MD, Professor of Medicine, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA
Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark
Ramon Estruch, MD, PhD, Hospital Clinic, IDIBAPS, Associate Professor of Medicine, University of Barcelona, Spain
Pierre-Louis Teissedre, PhD, Faculty of Oenology–ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France
Fulvio Mattivi, MSc, CAFE – Center Agriculture Food Environment, University of Trento, via E. Mach 1, San Michele all’Adige, Italy
Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; (Formerly, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia), Australia