Heavy alcohol consumption but not smoking predicts mortality in patients with acute coronary syndrome Authors: Allan Andersen, Steven R H Beach, Willem Philibert, James A Mills, Emily Milefchik, Emma Papworth, Kelsey Dawes, Joanna Moody, Gracie Weeks, Ellyse Froehlich, Kaitlyn deBlois, Jeffrey D Long, Ferhaan Ahmad, Robert Philibert Published: 2024 ABSTRACT ABSTRACTThe relationship of heavy alcohol consumption (HAC) and smoking to mortality in those with CHD, and mechanisms through which these effects are elicited are not clear. In order to improve our understanding, we examined the relationship of Alcohol T-Scores (ATS), an epigenetic biomarker of chronic HAC, and cg05575921 methylation, a biomarker of smoking intensity, with all-cause mortality and degree of coronary artery obstruction in a cohort of 217 subjects admitted for CHD-related acute coronary syndrome (ACS). We found that 65% of the subjects had ATS values indicative of chronic HAC. ATS values, but not cg05575921 values, were significantly associated (p < 0.02) with subsequent proband death (total of 28 deaths) with a Cox Proportional Hazards model showing a slightly larger effect of ATS levels than age on all-cause mortality survival (overall model, p < 0.003). Subjects in the highest decile of ATS scores had a 2.4-fold increase in the risk for mortality as compared to those in the lowest decile. In contrast, cg05575921 methylation (p < 0.003) but not ATS scores, were significantly inversely associated with degree of obstruction. Only 2 of the 217 subjects were referred for treatment for either smoking or drinking. We conclude that HAC is an underappreciated driver of CHD-related mortality, that those with ACS who smoke are much less likely to have significant obstruction upon cardiac imaging and that substance use treatment may be underutilized in those with CHD.KEYWORDS: DNA methylation, alcoholism, heavy alcohol use, digital PCR, epigenetics, coronary heart disease Introduction Both smoking and excessive alcohol consumption are well established as significant risks for coronary heart disease (CHD). It is generally accepted that any degree of smoking including passive smoking is unsafe with respect to risk for CHD [1]. However, the relationship for alcohol is more complex with the majority of epidemiological studies showing that the relationship between CHD and alcohol consumption follows a J-shaped curve with light drinkers being at the lowest risk and heavy drinkers being at the greatest risk [2,3]. Unfortunately, a more exacting understanding of this relationship to CHD for alcohol has eluded the field because of the difficulty of establishing reliable diagnoses of CHD in large epidemiologically sound populations and the reliance on often unreliable self-reports of alcohol use as the demarcator of alcohol consumption [4,5]. Progress in DNA methylation biomarkers may enable a more granular, quantitative understanding of the relationship of both smoking and drinking to the development of CHD. The work with cg05575921 methylation is perhaps better established. Since the discovery of this aryl hydrocarbon receptor repressor (AHRR) site in 2012, over 100 papers have demonstrated the relationship between demethylation at this CpG site with increasing smoking intensity and it is now generally accepted as a biomarker for smoking intensity [6–8]. In contrast, the progress with respect to alcohol consumption has been more recent. But, over the past 6 years, we have shown that a metric known as the Alcohol T Score (ATS), which is formed using the data from four methylation sensitive digital PCR (MSdPCR) assays, is capable of detecting heavy alcohol consumption (HAC), defined as drinking six or more drinks per day for eight consecutive weeks, with a Receiver Operating Characteristic area under the curve of 0.96 [9,10]. In several published studies, we have shown that the ATS outperforms carbohydrate deficient transferrin (CDT) assessments, which is considered to be the current gold standard for detecting excessive alcohol consumption [11,12], in predicting HAC, accelerated aging and alcohol related immune cell changes [13–15]. Furthermore, methylation at one of the four sites in the ATS, cg04987734, has been associated with all-cause mortality in at least three prior studies [16–18]. Previously, we have shown that these epigenetic measures are useful for understanding the impact of smoking and drinking in clinical settings, where self-reported alcohol consumption often can be highly unreliable [5,19,20]. For example, in a study of alcohol withdrawal syndrome, we found that objective assessments of alcohol consumption, including blood alcohol concentrations, methylation, and CDT assessments, were highly correlated with one another and predicted clinical outcomes. In contrast, self-reported alcohol consumption was not correlated with any of these biological markers and did not predict any clinical outcomes [15]. We believe that epigenetic tools such as the ATS can improve our ability to assess the effects of alcohol on CHD-related healthcare outcomes as well. Therefore, as the next step, we determined cg05575921 and ATS values in 217 subjects admitted for acute coronary syndrome (ACS), then assessed their relationship to survival and the degree of coronary artery obstruction. Methods The clinical and biological data were obtained from the Iowa CHD Repository Study. Each of the subjects in the current study were individuals over the age of 18 years who presented to the University of Iowa Hospital and Clinics (UIHC) and were hospitalized for the evaluation and treatment of ACS between 25 November 2019 and 18 March 2024 [21]. After informed consent was obtained from the subject or a legally authorized representative, each subject was interviewed, when possible, to confirm the CHD presentation history, and then underwent phlebotomy. After discharge, the medical records were abstracted by research assistants blinded to methylation status to confirm the clinical diagnosis of ACS and to obtain additional data for this study. Specifically, alcohol and tobacco consumption (i.e., drinking) information was retrieved through a search of the admission and discharge summaries, and daily clinical notes by a staff member (EP). All protocols and procedures used in this repository have been approved by the University of Iowa Institutional Review Board (IRB # 201910834). DNA was extracted from whole blood using our standard procedures [22]. DNA methylation assessments of cigarette and alcohol consumption were then conducted using MSdPCR following our previously described procedures [9,23]. In brief, quantification of cg05575921, a generally accepted biomarker of smoking intensity, was conducted using a MSdPCR assay (Behavioral Diagnostics, Coralville, IA) and droplet digital PCR machinery and reagents (Bio-Rad, Carlsbad, CA) according to manufacturer’s instructions. Similarly, ATS values for each subject were determined using the Alcohol Signature assay (Behavioral Diagnostics, Coralville, IA) and droplet digital PCR machinery and reagents (Bio-Rad, Carlsbad, CA) as previously described [9,10]. The Alcohol Signature is a battery of four assays that assess methylation status at the cg02583484, cg04987734, cg09935388, and cg04583842 loci, which are highly responsive to alcohol consumption and are not affected by smoking [10]. The ATS is a zero centered metric produced from these four assessments by calculating an unweighted sum of alcohol-induced changes in the T-score (TS) at the four sites [9]. The TS for each CpG site is produced by subtracting the previously established mean of methylation at the locus in abstinent controls, defined as those who report that they did not consume any alcohol or illegal substances over the past year and had negative tests for cannabinoids, from the observed methylation value for each subject at each locus, and then dividing the result by the standard deviation of the controls at that locus. Mortality was assessed by two methods with a cutoff date of 26 April 2024. First, the electronic health record (Epic, Verona, WI) was abstracted to determine whether the system had been notified of the subject’s death and if so, the date of the death. In addition, a research assistant blind to epigenetic status conducted a Google search for obituaries using the subject name and hometown as the initial keywords as described by Lund and associates [24]. The resulting obituaries were downloaded, and the identification of the deceased individual confirmed by the use of other personal identifying information, such as age and birth date, by the research assistant and another staff member. Subjects were designated as deceased if a matching obituary or Epic designation was documented. Duke Jeopardy Scores were determined by a medical student blind to methylation status using the method of Dash and colleagues using angiograms during the current or up to 6 months prior to hospital admission [25]. In the Dash et al., approach, coronary artery distribution is divided into six segments: the left anterior descending coronary artery (LAD), diagonal branches of the LAD, septal perforating branches, the circumflex coronary artery, obtuse marginal branches, and the posterior descending coronary artery (PDA). All segments distal to ≥70% stenosis are assigned 2 points with a total maximum score being 12. Initial analysis of data was conducted using the suite of general linear model analytic algorithms embedded in JMP Version 17 (SAS Institute, Cary, SC). Comparisons of normally and non-normally distributed continuous variables between groups were conducted using t-tests and Wilcoxon rank sum tests, respectively, where indicated. Comparison of categorical variables was conducted using chi-square testing. Comparisons between continuous variables were conducted using linear regression. Where indicated, correction for multiple comparisons was conducted using the method of Bonferroni [26]. Cox proportional hazards regression was used to assess ATS in predicting all-cause mortality, adjusting for age and sex [27]. Hazard ratios were estimated from spline-based baseline hazard models to compare those in the highest decile of ATS scores to those in the lowest decile [28]. Linear regression models were used to investigate the relationship of Duke Jeopardy Scores with ATS and cg05575921, with and without an adjustment for age and sex. Results Table 1 delineates the clinical and demographic characteristics of 217 subjects who participated in this study. Consistent with the demographics of the state of Iowa and the epidemiology of ACS, the majority of subjects in this study were White. Although males (68 ± 12 years) were older than females (65 ± 15 years), the difference was not statistically significant. Although CHD strikes both males and females relatively equally, over two thirds of the subjects were male (148 of 217). Table 1.Clinical and demographic characteristics of the subjects. MaleFemale 14869Age (years)68 ± 2165 ± 12Race White141 (95%)69 (100%) White Hispanic5 (3%)- Black2 (1%)-Type of ACS STEMI45 (30%)14 (20%) NSTEMI77 (52%)47 (68%) Other26 (18%)8 (12%)Deceased208Average survival (days)521 ± 521565 ± 552Self-reported alcohol use in chart Yes77 (52%)29 (42%) No71 (48%)40 (58%)Chart Diagnosis of AUD* Yes36 (24%)3 (4%) No112 (76%)65 (95%) Missing-1 (1%)Self-Report of smoking Yes42 (28%)14 (20%) No106 (72%)55 (80%)Angiograms with 6 months92 (62%)45 (65%)Duke Jeopardy Scores4.0 ± 2.94.2 ± 2.8Epigenetic Markers cg05575921**67.6 ± 21.6%77.6 ± 16% cg05575921 <80%70 (47%)22 (32%) >80%75 (51%)45 (65%) Missing3 (2%)2 (1%)Alcohol T Score5.2 ± 3.84.6 ± 3.2*AUD is an abbreviation of Alcohol Use Disorder.**p < 0.01, the ‘n’ for each category may differ due to missing data.n.b., not all percentages add up to 100% due to rounding. Clinical and demographic characteristics of the subjects. *AUD is an abbreviation of Alcohol Use Disorder. **p < 0.01, the ‘n’ for each category may differ due to missing data. n.b., not all percentages add up to 100% due to rounding. Review of the discharge summaries of the subjects showed that non-ST-elevation myocardial infarctions (NSTEMIs) were listed as discharge diagnoses for 57% of the subjects and 27% had ST-elevation myocardial infarctions (STEMIs) listed as discharge diagnoses. The remaining 16% had any of variety of non-infarct related discharge diagnoses listed such as non-specific electrocardiographic changes. Over the course of follow-up, at least 28 of the subjects died as documented by either an obituary identified through Google searches or the electronic medical record. Google searches identified obituaries for 25 of the 28 (89%) deceased subjects while the electronic health record reported that 23 of the 28 (82%) subjects were deceased. Males were not significantly more likely to report consuming alcohol (Χ2 = 1.89, p < 0.17) or smoking (Χ2 = 2.35, p < 0.31) than female subjects. However, the presence of a chart diagnosis of an alcohol use disorder (AUD) was considerably higher in male (24%) than in female (4%) subjects (Χ2 = 17.2, p < 0.0006). Table 1 provides the MSdPCR values for cg05575921 methylation. As assessed by this epigenetic measure, males have a significantly higher smoking exposure than females (p < 0.0001). However, using the established cutoff of 80% as indicative of smoking, both males (objective rate 48%, subjective rate 28%) and females (objective rate 33%, subjective rate 20%) significantly under-reported smoking. Consistent with our prior findings, cg05575921 was significantly correlated with ATS values (Adj. R2 = 0.195, p < 0.0001). Table 1 and Figure 1 show the distribution of the ATS in this cohort of subjects with ACS. In abstinent populations, the ATS is a unitless, zero-centered metric with a standard deviation of 2.2 [9]. However, in subjects admitted for ACS, the average ATS value was 5.0 ± 3.7 with a marked positive skew in the distribution. Using the lower threshold value of 3.5 that has been previously established to predict HAC [9], 140 of the 217 (65%) subjects in this study had current chronic HAC at enrollment. There was no significant difference in the intensity of alcohol consumption, as assessed by the ATS, between males (5.2 ± 3.8) and females (4.6 ± 3.2). Distribution of alcohol T scores (ATS, n = 217). The ATS is a unitless zero centered metric in abstinent individuals. Values of 3.5 or higher (black arrow) are predictive of chronic heavy alcohol consumption. In some studies, HAC consumption has been shown to be a risk factor for the development of CHD. However, whether its presence affects subsequent survival in those with preexisting CHD is unknown. To address this question, we first contrasted the ATS levels in those who died (n = 28) as compared to those who lived through the observation period (n = 199). ATS values (6.6 ± 3.1 vs 4.8 ± 3.6, p < 0.02) were significantly higher in those who died. Next, we conducted a standard semi-parametric survival analysis (Cox regression) of the cohort. The median follow-up period of the cohort was 234 days with an interquartile range of 122–838 days. A simple model of age at baseline and ATS values significantly predicted survival (see Table 2, p < 0.03 and p < 0.02, respectively). Interestingly, the addition of sex, a traditional risk factor for CHD, or other variables did not improve the model. The comparison of participants in the highest decile of ATS scores (cutoff = 9.3) to those in the lowest decile (cutoff = 1.1) yielded a hazard ratio estimate of 2.35 (95% CI: 1.16–4.76), or a 2.4-fold increase in mortality risk. In a sensitivity analysis that excluded participants over 80 years of age (N = 22), we found nearly identical results in the association of higher ATS values and increased mortality risk for the participants younger than 80 (N = 195; data not shown). Table 2.Relative contribution of variables to a cox proportional hazards model of risk for mortality.VariableParameter EstimateSEχ2p-valueAge0.04260.01845.39060.0202ATS0.10430.04395.63960.0176Sex (F vs. M)−0.12270.42510.08330.7729 Relative contribution of variables to a cox proportional hazards model of risk for mortality. An illustration of the combined effects of ATS and age is shown in Figure 2. The figure shows the cumulative hazard based on a spline-based baseline hazard that yields a smooth curve of estimates [29]. The figures are for males and show the cumulative hazard as a function of year in study among those at the 10th and 90th percentiles of ATS and age (cutoffs of 50 and 80 years, respectively). The lower right hand panel shows that the increase in risk is fastest for the combination of the older age (≥80) and larger ATS (≥9.3). In contrast, the risk increases at the slowest rate in the top-lefthand panel for the younger age (≤50) and the smaller ATS (≤1.1). At each age, heavier drinkers show a steeper slope of hazard over time than do those who are not heavy drinkers. Cumulative hazard (95% confidence interval) as a function of year on study, age, and ATS. Values for the latter two are set at the 10th and 90th percentiles, and the curves are for male participants. Conceivably, HAC could be causing this increased mortality in these subjects through a number of mechanisms. However, given that the subjects all have CHD, one obvious mechanism may be through increasing the degree of coronary artery obstruction. To help test that hypothesis, we determined Duke Jeopardy Scores, a measure of the degree of coronary artery obstruction and amount of myocardium at risk [30], for those subjects who had angiograms performed during or within 6 months of admission (see Table 1). There was no difference in the average jeopardy score between 92 males and 45 females who received angiograms. Furthermore, there was no significant relationship of ATS values to jeopardy scores before or after adjustment for age and sex (p < 0.52 and p < 0.30, respectively). However, there was a significant inverse relationship between increasing intensity of smoking (i.e., demethylation at cg05575921) and Duke Jeopardy Score (Figure 3; β = 1.80, p < 0.003). This relationship remains significant after adjustment for age, sex, and ATS (β = 1.44, p < 0.009). Therefore, even though the majority of subjects in this study were male, the findings also apply to females. The relationship between MSdPCR assessments of cg05575921 methylation and Duke Jeopardy scores (n = 133). Bivariate regression p-value is p < 0.003. Discussion In summary, the major conclusions of this study are: 1) that ATS scores significantly predict all-cause mortality in those acutely hospitalized for ACS with those at the highest quartile of alcohol consumption having a 2.4-fold greater risk for death than those at the lowest quartile, 2) that 65% of subjects admitted for ACS had ATS > 3.5 and 3) increasing smoking intensity is inversely associated with the degree of obstruction of the major coronary arteries. Limitations of the study include the lack of affirmation by other biomarkers of alcohol, medical conditions other than ACS were not included, the use of only one time point for methylation and clinical assessments, the lack of a matched control cohort for comparison and that the subjects were almost exclusively White and collected from a single site. In some respects, the overall findings should not be surprising since methylation status at some loci used in the ATS predict mortality in other populations that are also enriched for CHD [16–18,31]. Still, the current analyses do not show whether the observed excessive alcohol consumption caused excess mortality or was simply associated with causative factors. We speculate that both possibilities may be contributing to our findings. With respect to causality, alcohol is toxic to cells at relatively modest concentrations [28]. Unlike liver cells, cardiomyocytes cannot directly metabolize alcohol and its presence impedes the generation of ATP by the mitochondria [28]. Finally, the consumption of alcohol in the diet also displaces the consumption and/or colonic absorption of vital nutrients needed for cardiac function [32]. For these and other reasons, it is likely that HAC is directly acting on the heart to increase dysfunction and drive excess cardiac mortality. At the same time, excessive alcohol consumption is also a strong indicator of overall stress, which in turn could drive mortality directly or indirectly. Classically, excessive alcohol consumption is a dysfunctional behavior that occurs in response to a variety of psychological or physiological stressors. For example, chronic financial strain, which is associated with HAC, is associated with hypertension [33]. Similarly, as we have shown here and elsewhere, HAC is highly correlated with smoking, a well-known driver of CHD [14]. In addition, as we show elsewhere, ATS values are asssociated with increased systolic blood pressure, hemoglobin A1c and serum triglycerides, all of which are established clinical risk factors for CHD [25]. Finally, stress- related changes in sleep patterns similarly are associated with an increased likelihood of ACS [34]. Thus, there is a myriad of factors associated with HAC that can cause the occurrence of an acute exacerbation of CHD. Furthermore, excessive alcohol consumption is associated with non-compliance with medical therapy and risk for accidental death. Patients with serious diseases are routinely cautioned not to drink alcohol. Those who continue to consume alcohol are more likely than those who do not to be non-compliant with medical therapy [35,36]. Finally, excessive alcohol consumption is a well-established risk factor for accidental death, both in the home and in a motor vehicle [37]. The second major finding in this study is that there appears to be an alarmingly high rate of HAC in this hospitalized population, suggesting that HAC may be a risk factor for ACS and may be underdiagnosed. However, we do not know the distribution of ATS values in the catchment area served by our academic medical center. Still, in both published and unpublished studies, we have examined ATS values in at least 4000 individuals including those in alcohol treatment centers and those admitted for alcohol withdrawal monitoring/treatment [9,10,13–15,38–40]. The ATS values in this study are some of the highest values that we have ever observed, and we note that the ATS values even in non-abstinent, well-controlled subjects recruited from our academic medical center community are low. However, to truly understand the significance of this observation, it will be necessary to understand the distribution of ATS values in the general population and in medical populations. These studies are in progress elsewhere. One concern of some may be that medications used for the treatment of CHD could affect ATS values. This is not an unreasonable fear because our assessment method uses leukocyte DNA, and any pathophysiological process that affects the leukocytes, such as leukemia or treatment with chemotherapeutic agents, could result in anomalous ATS values. Still, in the training set of 143 heavy drinking inpatients used to develop the ATS algorithm, we observed no unusual effects of medications on DNA methylation [10]. Similarly, we observed no effects of medications in our recent study of 120 subjects admitted for alcohol withdrawal [15]. However, while re-assuring that does not rule out other potential factors that may be enriched in subjects with CHD. For example, clonal hematopoiesis is a recently characterized risk factor for CHD [41]. Furthermore, commercial testing using a different MSdPCR technology excludes those who have undergone bone marrow transplants [42]. However, in truth, there is a plethora of potential other explanations. For example, the poor nutritional status found in many alcoholics may generate part of the signal and may continue even in the absence of alcohol use if the poor nutrition continues. Other physiological factors such as chronic uremia or hypoxemia may similarly affect the signature. Therefore, just as we have noted the effect of statins on CDT performance, it very well may be that in the future, others may note that conclusions derived using DNA methylation testing may have to take into account other factors that can alter epigenetic signatures. Studies that use multiple objective methods to characterize alcohol consumption will be necessary to generate that improved understanding. The strong inverse relationship of smoking intensity to the degree of cardiac obstruction may be of interest to those examining the pathophysiology of Ischemia with No Obstructive Coronary Artery (INOCA) [43]. This subtype of CHD is particularly common in females with up to 65% of females undergoing diagnostic angiography having no significant obstruction detected [44]. A substantial portion of the vulnerability to INOCA presentations is thought to result from smoking-related inflammation of small vessels [43,45]. These data show the potential benefit of objective determinations of smoking intensity in studies of INOCA and emphasize prior smoking cessation should be emphasized in those who smoke and have INOCA [43]. The current findings should not be construed to minimize the effect of smoking on mortality. Smoking is a well-established driver of CHD whose influences of the development of CHD likely begin after the initiation of smoking and by determining cg05575921 methylation we may be able to better determine that initiation of pathogenesis. In prior studies, cg05575921 methylation has been used to quantify nascent smoking [46,47]. However, some have noted that demethylation of this locus occurs even in response to passive smoking [48,49]. Therefore, given the clear relationship between the intensity of smoking and drinking intensity shown in prior studies [10,13,50,51], the current findings instead emphasize the need to completely eliminate smoking at any stage of life. The finding that HAC increases mortality is already well established and association of HAC with hospitalization and/or death already has been noted in the epigenetic literature [16–18]. Still, the magnitude of the effect in this study is unusual. We believe that our stronger findings may be because the reliance on self-reporting used in prior studies may have obscured the association between alcohol use and CHD as it did for associations of HAC with accelerated aging [52]. In addition, the CDT, which is commonly used by the insurance industry and sometimes by clinicians to detect HAC, is confounded by the use of statins [53]. Fortunately, other tests, such as phosphatidylethanolamine, which is sensitive to heavy alcohol consumption over the past 10 days, are now available [54]. However, these tests require special sample preparation and laboratory equipment not found in most major medical centers. Hence, they are not often performed. The potential clinical implications of these findings are straightforward. We recognize that it may be difficult to gather comprehensive biopsychosocial histories in the acute cardiovascular setting. Given these and other constraints in the practice of clinical medicine, it simply may not be possible to use clinical histories to detect the presence of AUD in these individuals in similar situations. Instead, DNA methylation or perhaps other biomarker methods, such as the use of phosphatidylethanolamine, may be necessary to determine the need to address the presence of an AUD [55]. Indeed, if these findings are confirmed, it is imperative that steps be taken to detect and treat HAC in those presenting with ACS. Current treatment guidelines mandate that other better established drivers of CHD, such as smoking and hypercholesteroemia, which are clinically modifiable, be addressed. Although AUD is similarly treatable, it also must be detected as part of the medical assessment process. The current study, which uses clinically implementable MSdPCR techniques, suggests that the use of DNA methylation technologies may be one mechanism to achieve that detection and to help better understand the mechanisms through which the excessive mortality is moderated. Acknowledgments The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Funding Statement This work was supported by NIH [R21 AA029435]. Disclosure statement Dr. Philibert is the Chief Executive Officer of Behavioral Diagnostics. The use of cg05575921 to assess smoking status is covered by existing and pending patents including US Patents 8,637,652 and 9,273,358, while the use of DNA methylation to assess alcohol is covered by existing and pending patents including European Union Patent 3149206. On behalf of Drs. Philibert and Andersen, Behavioral Diagnostics and the University of Iowa have filed an intellectual property claim on the use of DNA methylation to predict AWS and related phenomena. Dr. Philibert is also a co-founder, board member, and Chief Medical Officer of Cardio Diagnostics Inc. The use of DNA methylation and gene-methylation interaction effects for the assessment, diagnosis, and monitoring of cardiovascular disease is covered by United States Patent 11414704, European Patent EP3472344B1, and other both granted and pending intellectual property claims elsewhere. The University of Iowa Research Foundation is entitled to royalties from use of this technology. Data availability statement The data used in this study are available at the Open Science Foundation website. https://osf.io/adnpm/?view_only=b1a2e2c22b1f4611b4c6eb2820ee95a5. References Associated Data Data Availability Statement The data used in this study are available at the Open Science Foundation website. https://osf.io/adnpm/?view_only=b1a2e2c22b1f4611b4c6eb2820ee95a5.