Critique 257:  Does statin use modify potential beneficial effects of moderate alcohol consumption on cardiovascular risk?   –   21 September 2022

Anderson JL, Le VT, Bair TL, Muhlestein JB, Knowlton KU, Horne BD.   Is Alcohol Consumption Associated with a Lower Risk of Cardiovascular Events in Patients Treated with Statins? An Observational Real-World Experience.  J Clin Med 2022;11:4797.  https://doi.org/10.3390/jcm11164797

Authors’ Abstract

Introduction: Alcohol consumption has long been associated with cardiovascular (CV) benefit, but it also has adverse potential. Statins are currently widely used for CV prevention.  We evaluated whether alcohol use is associated with lower CV risk in patients on statins.

Methods: We searched Intermountain Medical Center cardiac catheterization laboratory medical records for patients with a prescription history of statin use or non-use and a self-report of alcohol use or non-use.  Alcohol and statin prescription data were available together with long-term (mean [SD], 4.4 [2.4] years) major adverse CV events (MACE, including death, myocardial infarction, stroke, and heart failure hospitalizations) in 1701 patients at primary and 3266 patients at secondary CV risk.

Results:  MACE rates were lower for primary prevention alcohol users than non-users not on statins (adjusted hazard ratio [adj-HR] 0.50 (95% CI 0.33, 0.78, p = 0.002), but not for those on statins (adj-HR 0.84, CI 0.54, 1.32, p = 0.45).  MACE rates for secondary prevention were not reduced by alcohol consumption either in statin non-users or users (adj HR 1.18, CI 0.85, 1.64, p = 0.33; adj HR 1.08, CI 0.87, 1.35, p = 0.45, respectively).

Conclusions:  These findings, together with other recent supportive studies, can help inform personal choices in alcohol consumption and professional society recommendations for CV prevention

Forum Comments

Many factors have been shown to affect the risk of cardiovascular disease, including genetic factors, diet, obesity, smoking, dyslipidemia, alcohol consumption, and the use of a variety of prescribed medications.  The present study evaluated the risk associated with alcohol consumption according to whether the subject was taking a statin or not taking a statin.  (It did not evaluate the alternative question of how adding a statin relates to cardiovascular risk according to whether or not the subject was a regular consumer of alcohol.) 

Does alcohol consumption modify the effects of statins on liver function?   Forum member Skovenborg provided an excellent summary on the topic of this paper: “There have been a number of papers relating the use of alcohol along with statins, primarily relating to hepatic function.  The use of statins is included in the list of possible harmful interactions mixing alcohol with medicines published by the National Institute on Alcohol Abuse and Alcoholism (2014) with liver damage as the possible harm.  A Chinese study of hepatic safety of statins used in very elderly patients (Guo et al) found alcohol consumption was associated with hepatic aminotransferase elevation: OR 2.737 (95% CI 1.373-5.454) but all patients returned to normal liver enzyme levels after statin therapy cessation.  In the Heart Protection Study (2003), no upper limit for alcohol intake was imposed provided that liver function tests were within acceptable ranges, and patients were thought likely to be compliant.  In that study, over 2000 (11%) participants reported baseline alcohol intake >21 units per week and there was no evidence that these people were at any greater risk of statin-associated excess of raised alanine transaminase.

“Further, in a clinical study of dyslipidemic patients treated with simvastatin 20 mg/day, the group on simvastatin that reported 30 g alcohol/day had higher levels of HDL-cholesterol than the group treated with simvastatin alone (Zdrenghea et al).

Are the participants in the present study typical of the population in western countries?  “Compared with most other populations in the Western world, the Intermountain population in this study had a very high percentage of nondrinkers: 65.9% of the group of 1701 patients at primary cardiovascular risk were nondrinkers, and 72.7% of the 3266 patients at secondary cardiovascular risk were nondrinkers.  Intermountain Healthcare is a nonprofit, integrated healthcare system that, at the time of this study, included 24 hospitals and 215 clinics in Utah, Idaho, and Nevada.  The especially high rate of nondrinkers could relate to religious restrictions.  Statewide, Mormons (the faith that teaches its members to abstain from alcohol) account for nearly 62% of Utah’s 3.1 million residents. While religious preference was not considered in this study, the external validity of this observational study might be compromised by that high percentage of abstainers in the study population.”  

Other problems with the analysis in this study:  Reviewer Hendriks echoed the concerns of others that the study should have focused on “moderate” alcohol consumption.  “Although the information on drinking level seemed to have been present for these patients, no data were provided on the numbers of patients in these groups and no analysis of the association with drinking category was shown. The category of  ≥ 2 drinks per week is an open-end group, which does not exclude the possibility that a relatively small number of alcohol drinkers in this population contained heavy drinkers or alcohol abusers.”

Forum member Ellison stated: “Showing zero coronary obstruction at catheterization was necessary to be included in the ‘primary’ group.  However, about 30% of the subjects in the ‘secondary’ group also showed no coronary obstruction, but were moved to the secondary group because of previous evidence of disease (myocardial infarction, a revascularization procedure, or obstruction of 70% or more of a coronary artery either at a previous or at the current catheterization). It is worrisome that the latter might mean that the secondary risk group was a very mixed group.  Given questions about the categories into which subjects were placed, it would have been enlightening to see some analyses with results shown with the relations between alcohol and statin for all subjects combined, but no data are presented on the effects of alcohol for the entire cohort.”  Reviewer Hendriks added: “The selection and categorization of the patient groups is unclear to me. There were patients undergoing cardiac catheterization and angiography and patients undergoing angiography only. Then patients received classification into primary or secondary prevention groups.  Criteria for categorization into primary and secondary prevention group are unclear and in both groups statins were prescribed.  If the research question concerns statin use and alcohol consumption why not combine primary and secondary prevention patients?”

Reviewer Ellison had another concern: potential under-reporting of alcohol consumption in a population with a large proportion of subjects whose religion requires abstinence.  It would be assumed that many Mormon subjects who may have consumed small amounts of alcohol on occasion would not report any alcohol use on a form going to a hospital where they were being treated.  Thus, there is a considerable risk of misclassification of exposure.  Even more problematic is an inadequate ascertainment of alcohol use, obtained on one occasion with no information on the pattern of drinking. 

“In addition, the authors appear to insinuate that lack of statistical significance denotes ‘no effect.’  However, for statin users the estimated hazard ratio was 0.84, a 16% lower risk than that of reported non-drinkers (not statistically significant but still compatible with a reduction in risk).”  Reviewer Hendriks agreed that “the absence of statistical significance in a small group does not mean the absence of an effect; it may imply that group size was insufficient.”  He stated further: “Drinking alcohol (in any quantity and in any pattern) either with or without statins rather improved than worsened the outcomes.  So why would the clinical implication have to be an even stronger message against the use of alcohol for cardiovascular prevention?  Especially when the authors state elsewhere: ‘Moreover, it cannot exclude small advantages (e.g., of the order of 10%) in MACE reduction by alcohol in the setting of statin use.’”

Reviewer Hendriks added: “Indeed this a complex paper with lots of uncertainties on the registration of the data and the statistical methodology applied in this small group of patients.  Correction for diabetes in this group is incorrect since diabetes is an intermediate variable, and correction for diet quality is lacking.  The part ‘mechanistic considerations’ does not really contain such considerations.  If any, they are not reflecting the state-of-the-art on mechanistic studies performed.  This may have been particularly interesting since statins lower LDL-cholesterol mainly whereas moderate alcohol consumption increases HDL-cholesterol.”

Other reviewers also expressed concern that diabetes was included, along with a huge list of other variables, as a potential confounder in the analyses.  The risk of diabetes has been shown consistently as being reduced from the moderate intake of alcohol, and is considered generally as an intermediate variable on a causal path from alcohol to cardiovascular disease.  In the present study, including diabetes as a confounder is probably an example of over-adjustment bias (Enrique et al).  Also, the inclusion of many other indices of cardiovascular disease, some known to relate to alcohol use, as potential confounders could be problematic. 

Forum member Djoussé wrote: “The main objective was to evaluate if alcohol intake has some benefits on CVD among statin users.  The authors stated: ‘We asked the question whether alcohol use would still be associated with a lower risk of adverse CV events in patients taking statins for primary or secondary prevention.’  Tables in the paper show data only on age and sex but nothing about major cardiovascular confounders to assess comparability of information among subjects who reported alcohol intake and those that did not among statin users.  Already here, we note that drinkers among statin users were more likely to be male and were younger than non-drinkers.  Confounding by comorbidity (hypertension, diabetes, etc.) or risk factors (mean LDL, physical activity, adiposity status, smoking, etc.) cannot be assessed given a lack of those data between drinkers and non-drinkers who were prescribed statin.

“Furthermore, the term ‘statin use’ is very broad and it is unclear what proportion of subjects were on high vs. low-intensity statin or degree of compliance with prescribed strength of statin. These are major determinants of residual risk for cardiovascular disease, even among patients prescribed statins, that should be carefully examined when addressing their study question.  Further, the number of subjects taking statin in primary prevention is very limited for a meaningful inference (164 drinkers and 313 non-drinkers).  Overall, while the primary study question is of interest, data presented are insufficient to answer adequately whether alcohol drinking confers additional benefits among patients treated with statins. The role of confounding and other biases in the current findings lead to uncertainty about their conclusions.”

Forum member Zhang commented on other concerns: “Besides problems with ascertaining alcohol consumption and actual consumption of prescribed statin, I have concerns about the study design and the interpretation of results of this paper.  Both alcohol consumption and statin use were prevalent exposures; thus, it always raises a concern on potential survival bias, especially when the study was conducted among older adults, when many people may have already developed the disease under study.  However, my major concern on this study is its design and results interpretation.  In general, most people start alcohol drinking in their early years, at 20-30 years of age; thus, alcohol consumption occurs before occurrence of MACE or its early pathological changes.  Subjects included in the current study (aged around 65 years) consisted of those who underwent cardiac catheterization and coronary angiography; thus, many of them may have already developed the early pathological changes in their cardiovascular system, especially subjects who were in the ‘secondary prevention’ group or who took statin.

“As shown in the following diagram, analysis of association between alcohol consumption and the risk of MACE among subjects who already have early pathological changes in their vascular system (such as those who took statin and had coronary artery stenosis) is to assess the direct effect of alcohol consumption on the risk of MACE, but not the total effect of alcohol consumption on the risk of MACE.  In other words, the study attempts to examine whether there are other mechanisms that could ‘cause’ MACE by alcohol consumption besides its ‘effect’ on the early pathological changes of MACE (which would be shown by a direct arrow from alcohol consumption to MACE in the figure below).  Analysis of restricting subjects among those who had early pathological changes in their vascular system is akin to adjusting for potential mediators.  As a result, the authors cannot obtain the total effect of alcohol on the risk of MACE, and direct effect of alcohol on the risk of MACE is susceptible to selection bias (i.e., collider bias) (Zhang et al).

           Total Effects of Alcohol Consumption (in Addition to the Direct Effects)  

         ​​

Alcohol consumption  Early pathological changes in CV system   MACE  

                                               Intermediary Factors

*In addition to a direct effect of alcohol on MACE, alcohol consumption has also been shown to relate to increases or decreases of a number of intermediary early pathological changes in the CV system, which themselves relate to the subsequent risk of MACE. In the present study some such changes were used incorrectly in the analysis as confounders.

Additional material from Forum member Di Gaetano, with input from his colleague Augusto Di Castelnuovo, agreed with concerns expressed by other Forum members.  They point out that “In view of the well-known J-shaped association of alcohol consumption with cardiovascular outcomes, the use of unique categories of drinkers is a severe weakness of the manuscript.  Also, there is no mention of how the authors dealt with former drinkers.  With the small available sample size and the observed event rate, the multivariable model used would not be valid (the rule of thumb is to include one variable for adjustment per 10 events).”

The law of diminishing returns:  Reviewer Skovenborg noted: “According to the authors ‘As preventive and treatment therapies evolve and advance, older therapies often showed diminished benefit, and indications for their use are more limited.’  Per se, the statement is correct, however it does not cover the effects of ‘the law of diminishing returns.’  The law of diminishing returns may be illustrated with data from the Heart Protection Study (Duerden et al).  In their analyses, aspirin, beta-blockers, ACE inhibitors and statins were each stated to lower the risk of future vascular events in high-risk patients by about 25%.  However, cumulative relative-risk reduction if all four drugs are used simultaneously is about 75%, not 100%.  The effect of the law of diminishing returns shows that each additional drug gives a smaller absolute reduction in risk, and the overall reduction achieved in relative and absolute risk by adding drugs to reduce risk of subsequent events in cardiovascular disease will be less than the sum of each drug’s effects when given separately. (Duerden et al).  

“The 2022 USPSTF systematic review of statin use for primary prevention of cardiovascular disease (Chou et al) found the following: Pooled relative risk reductions (95% CI); all-cause mortality 0.92 (0.87-0.98); CVD mortality 0.91 (0.81-1.02); myocardial infarction 0.67 (0.60-0.75); and stroke 0.78 (0.68-0.90).  The effect of the law of diminishing returns would diminish the protective effect of a moderate alcohol consumption in statin treated patients relative to patients with no statin treatment, exactly as the results of the Intermountain study found.  That the risk reduction in statin-treated patients was non-significant (HR 0.82 [0.36-1.90] is probably partly due to the moderate size of the study population.  Regarding the ’no effect’ statement results of moderate alcohol intake for patients at secondary cardiovascular risk, it is important to remember that these patients are treated not only with lipid lowering medication but also with  other medications.  Accordingly, the result of adding a moderate alcohol consumption effect to this group of patients would be expected to be small, or perhaps negligible.”

Do the results of this study answer the question posed by the authors?  Because of potentially serious problems in categorization of subjects and in the analyses, questions arise as to the conclusions of the authors.  Forum member Finkel wrote: “I consider that the source of data for this study is so limited that it is not possible to make any conclusions from these analyses that would apply to any other group of people.”  Further, the authors’ choice of very selected and not necessarily relevant references to support their inferences weakens their key conclusions. 

References from Forum Critique

Chou R, Cantor A, Dana T, et al.  Statin use for the primary prevention of cardiovascular disease in adults: updated evidence report and systematic review for the US Preventive Services Task Force. JAMA. Published online August 23, 2022. doi:10.1001/jama.2022.12138.)

Duerden M, Avery T, Payne R.  Polypharmacy and medicines optimisation. The King’s Fund 2013.  Information obtained in September, 2022 from the following source:  https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.734.5288&rep=rep1&type=pdf)

Enrique F, Schisterman EF, Cole SR, Platt RW.  Overadjustment bias and unnecessary adjustment in epidemiologic studies. Epidemiology 2009;20:488-495.

Guo M, Zhao J, Zhai Y, Zang P, et al.  A prospective study of hepatic safety of statins used in very elderly patients. BMC Geriatrics 2019;19:352.

Heart Protection Study Collaborative Group.  MRC/BHF Heart Protection Study of cholesterol-lowering with simvastatin in 5963 people with diabetes: a randomised placebo-controlled trial. Lancet 2003;361:2005–2016.

National Institute on Alcohol Abuse and Alcoholism.  Harmful Interactions – mixing alcohol with medicines. NIH Publication No 13-5329. Published 2003, revised 2014.

Zdrenghea D, Gligor E, Ossian V, Pop D.  The effect of simvastatin associated with ranitidine and alcohol upon serum lipids. Rom J Intern Med 2004;42:143-148.

Zhang YQ, Neogi T, Hunter D, Roemer F, Niu J.  What Effect Is Really Being Measured? An Alternative Explanation of Paradoxical Phenomena in Studies of Osteoarthritis Progression. Arthritis Care Res (Hoboken). 2014;66:658–661.

Forum Summary

The goal of this study was to determine if statin use (defined as the subject having been given a prescription for a statin, although actual consumption of statin was not ascertained) among a group of subjects who underwent cardiac catheterization modified the effects of alcohol consumption on subsequent cardiovascular events and total mortality.  Their primary outcome was the occurrence during a follow-up period averaging 4.4 years of a Major Cardiac Adverse Event (MACE = documented myocardial infarction, ischemic stroke, heart failure hospitalization, or all-cause mortality).  

From their adjusted analyses for “primary prevention” subjects MACE outcomes were 6.5% for alcohol users versus 14.2% for alcohol non-users (HR 0.50, CI 0.33, 0.78) for those not on a statin.  If the subjects were on a statin, the protection from alcohol was less (19.5% for alcohol users versus 22.7% for alcohol non-users), not statistically significant.  For subjects classified as ‘secondary prevention,’ alcohol users had only slightly lower risk of a MACE outcome, 18.2% versus 19.9% for no statin users and 19.9% versus 22.7% for statin users, not statistically significant.  (In the Authors’ abstract, the HRs reported as greater than 1.0 are apparently for non-users of alcohol versus users of alcohol, as the reported adjusted risk of MACE was lower for alcohol users in all categories.)

There were several problems identified in the Forum review regarding the study design, analysis, and conclusions of the authors.  Subjects were from the Intermountain Medical Center with patients primarily from Utah and several surrounding states.  The characteristics of the subjects are quite different from those in most other areas of the United States in that the large majority reported that they consumed zero alcohol (between 66% and 75% in the categories of subjects); this is probably related to most being of the Mormon faith where alcohol use is not acceptable.  (Religious affiliation of subjects is not reported in the paper.)  Further, it would be expected that many Mormons who may have had consumed some alcohol in the previous year would deny it on a form being collected by the Intermountain Center, where they are being treated; this would lead to under-reporting of the exposure being studied and misclassification.

Further, all subjects were placed into categories classifying them as at “primary cardiovascular risk” or “secondary cardiovascular risk”; also if they reported any alcohol use (regardless of the amount or the drinking pattern) they were classified as alcohol users; otherwise they were in the no alcohol category.  Subjects with zero coronary obstruction on angiography were classified as being at “primary risk” if they had not previously reported a myocardial infarction, coronary revascularization procedure, or an obstruction of a coronary artery of ≥ 70% identified on a prior angiogram (n=477).  Otherwise, regardless of current findings of coronary obstruction of zero or of any amount of obstruction greater than zero, subjects with a previous cardiovascular event (including ≥70% obstruction of a coronary artery on a previous catheterization) were classified as being at secondary risk (n=2,108).  Such complex and unusual categorizations could lead to misclassification of exposure and various types of bias.  No data that are based on all patients, regardless of classification into primary or secondary risk, are presented.

Unfortunately, in their main analyses there were no distinctions according to the reported amount of alcohol.  They do report “exploratory analyses” of reported amount: those consuming < 2 drinks/day seemed to have the most protection from alcohol – however, the numbers of subjects in the groups were inadequate for definitive results.  Further, there were no data on the “pattern of drinking” (regularly or sporadically, with or without food, etc.) for subjects reporting any alcohol use.  Having all subjects evaluated only within categories could lead to bias and limits the usefulness of the results. 

In their analyses, the authors adjusted the effects of alcohol use on MACE with an extremely long list of potential confounders.  Unfortunately, among these was the presence of diabetes, for which alcohol consumption has clearly been shown to reduce the risk of cardiovascular disease; diabetes is considered to be an important intermediary mechanism of the effect of alcohol.  Thus, it would not be appropriate as a confounder.  Also, others of the factors included as potential confounders were those that would be expected to be affected (either adversely or beneficially) by alcohol consumption, thus should not be used as potential confounders if one is seeking to determine the total effect of alcohol on cardiovascular disease.

Overall, a number of factors weaken the implications cited by the authors.  These include the database used for this study (a majority of subjects reporting zero alcohol at baseline, unusual for Western populations), problems in categorization of subjects into primary or secondary risk, inappropriate factors adjusted for as “confounders” (including some directly affected by alcohol use), failure to discuss the law of diminishing results, and suggesting “no effect” when results were not statistically significant.  Such problems make it difficult to support the authors’ conclusion that the prescription of a statin negates any effect of alcohol consumption on cardiovascular risk.   

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

Giovanni de Gaetano, MD, PhD, in conjunction with his colleague, Augusto Di Castelnuovo,  Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy

Luc Djoussé, MD, DSc, Dept. of Medicine, Division of Aging, Brigham & Women’s Hospital and Harvard Medical School, Boston, MA, USA

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

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

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

Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Principal, Stockley Health and Regulatory Solutions; Adjunct Senior Lecturer, The University of Adelaide, Adelaide, Australia

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

Yuqing Zhang, MD, DSc, Clinical Epidemiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.

With additional invited comments from Henk Hendriks, PhD, Senior Researcher, Centre for Earth, Environmental and Life Sciences, Zeist, Netherlands