Review of Evidence on Alcohol and Health (2025)
Chapter: 9 Future Directions
9
Future Directions
As the committee discussed its approach to the Statement of Task, it deliberated on and outlined its methodology, reviewed papers and obtained additional support for the systematic reviews of current literature. In the course of drafting and finalizing its findings and conclusions the committee identified additional methodological considerations as well as a consistent set of research issues that could strengthen the existing evidence on moderate alcohol consumption and health outcomes. This culminated in development of a list of specific research gaps for consideration for future studies looking at the questions in the Statement of Task. These future directions are discussed below.
METHODOLOGICAL CONSIDERATIONS
Exposure Measurement
A common challenge for studies examining the effects of alcohol on health is a lack of standard definitions of alcohol consumption levels and a lack of standardized limits for exposure categories. As discussed in Chapter 1, not all studies define exposure subgroups with reference to the U.S. Dietary Guidelines for Americans (DGA). Additionally, within the boundaries of moderate alcohol consumption there is a paucity of data on how variations in the volume, beverage type, frequency, and pattern of moderate alcohol consumption (i.e., low versus higher moderate intake) affect the associations of moderate alcohol consumption with health outcomes.
Standard Drink Sizes
In the United States, the Centers for Disease Control and Prevention (CDC) and the National Institute on Alcohol Abuse and Alcoholism (NIAAA) define a standard drink as containing 14 grams, 19 mL, or 0.6 ounce (oz) of ethanol (CDC, 2024; NIAAA, n.d.). Fourteen grams is the approximate ethanol content of 12 oz of beer, 5 oz of wine, and 1.5 oz of spirits. Some U.S.-based researchers use 14 grams as the definition of a “standard drink” (CDC, 2024; NIAAA, n.d.). Uniformity is further complicated by the fact that the definition of standard drink size varies by country and ranges from, for example, 8 grams in Korea to 10 grams in the United Kingdom to 8 grams in Sweden. When investigators use different definitions (e.g., 14 grams versus 10 grams), alcohol intake quantification must be adjusted accordingly to facilitate comparisons. The variation in the definition of a standard drink also complicates the categorization of moderate drinking that make evidence synthesis efforts more difficult.
Type of Alcoholic Beverage
Alcohol beverage type is typically divided into predominantly wine, predominantly beer, or predominantly spirits. Some individuals will consume only one beverage type while others consume multiple types of beverages and will thus be categorized into a mixed beverage group. If the health effects of alcohol (ethanol) are due solely to alcohol, comparable quantity, frequency, and pattern of intake should provide similar health effects across those beverage types; however, there are certainly opinions regarding differential benefits associated with specific types. This additional detail of exposure measurement could add important specificity to determining the health effects of moderate drinking.
Drinking Pattern
Drinking pattern refers to the number and timing of occasions where alcohol is consumed per week and may include further details, such as whether the alcohol is consumed with food. While research practice is less defined for this concept, a preferred approach for assessing the number of occasions where alcohol is consumed is to categorize consumption as frequent (e.g., ≥3 times per week) or infrequent (e.g., 1–2 times per week). In both cases, the amount of alcohol consumed must be within the limits of “low risk drinking,” that is, no more than two drinks per day and 14 drinks per week for men, and half the maximum for women. With a large enough sample, it is possible to examine the interrelations of average total intake and drinking pattern. Given the pharmacologic properties of alcohol, it would be unlikely that consumption of one drink each day for one week
(average 1 drink/day) has the same health effect as seven drinks on a single night (average 1 drink/day). To improve the specificity of evidence on the health effects, future metrics for research on alcohol consumption should include these intake patterns to better evaluate health effects.
Intake Reporting
An underlying issue for assessing any alcohol intake data is that the reported amounts and patterns of intake are derived from self-reported questionnaire-based data. In longitudinal studies, the reproducibility of self-report data appears to be good, but the validity of the data is uncertain (Ravelli and Schoeller, 2020). In populations with alcohol use disorder (AUD), self-reported alcohol consumption data are inconsistent with data based on alcohol biomarkers, and the former under-reports consumption by 5.5 percent to 56.0 percent (Nielsen et al., 2021). Some studies attempt to address this issue using collateral reports from family or friends, but these also are not reliable. Biochemical markers such as phosphatidylethanol and ethyl glucuronide have high reliability, but they may have a short duration in the body depending on the tissue sampled; their quantitation incurs a significant financial cost in population-level studies (Afshar et al., 2022). Thus, self-reported data are used to assess alcohol intake, considering the underlying assumption that alcohol intake is commonly under-reported by participants (Stockwell et al., 2016).
A further challenge regarding use of self-reported data is a difference between alcohol consumption levels obtained from self-reports and data based on alcohol purchase records for geographic locations. The latter are more objective as they are derived from taxation records. When converted to per capita alcohol consumption, some studies have found that self-report consistently underestimates alcohol purchase reports by as much as 60 to 80 percent (Stockwell et al., 2018; Subbaraman et al., 2020). The committee notes that there is no consensus on how to apply this discrepancy to sub-cohorts within a population. For example, if the response error affects all respondents similarly, self-reported alcohol intake levels are “underestimated” but retain their rank-order validity. However, if the response error affects occasional, moderate, and heavy alcohol consumer sub-cohorts differently, estimates of the association of alcohol intake with outcomes could over- or underestimate the true association depending on the sub-cohort. Additionally, evidence suggests that under-reporting is greatest among those having the highest intake levels (Bhattacharya et al., 2018). Generally, response error is a potential challenge for most observational research on diet and health (Ravelli and Schoeller, 2020).
Under-reporting of alcohol consumption will continue to be a challenge, and more accurate tools to quantify alcohol exposure are needed. Additionally, alcohol consumption should be assessed at multiple time
points across the entire lifespan (i.e., adolescence, young adult, middle age, older age) because the most vulnerable periods for alcohol’s impact on health outcomes are unknown. For example, are health risks (or benefits) incurred during younger or older drinking, or is alcohol’s impact cumulative? Examination of the full spectrum of alcohol consumption, from never-drinker to alcohol use disorder, would enable the modeling of outcome trajectories and cut points for positive or negative health outcomes.
Comparison Groups
A limitation of many studies assessing alcohol consumption is the continued practice of using nondrinkers, including former and never drinkers, as the reference cohort. As discussed in Chapter 1, this practice introduces substantial bias because nondrinkers are a heterogenous group comprised of individuals who never consumed alcohol due to personal preferences, those who never consumed alcohol or stopped because of health problems, and former heavy alcohol consumers including individuals with AUD. These last two groups may carry a burden of illness that is absent from a moderate drinking cohort and thus possibly bias outcomes more favorably toward the moderate drinkers.
Awareness of “abstainer bias” is growing, as per the number of studies in this report that were eligible and could be analyzed. Future studies must ensure that individuals who are true abstainers are not included with former users of alcohol in reference groups. This issue is especially critical for the analysis of moderate drinking. However, because abstainers also are a heterogeneous group, as noted above, a preferred approach may be to incorporate multiple (separate) comparison groups such as lifetime abstainers, former moderate drinkers, or current infrequent drinkers. A similar finding across these groups would suggest that the choice of comparison group did not influence the results, while differences would be important to note and further understand. The committee recognizes that there has been an increased use of ‘occasional drinkers’ (e.g., <1 drink/week or <1 drink/month) as a reference cohort, which may further complicate conclusions regarding health effects, as well as raise the issue of the potential value of creating comparison groups that would allow assessment of the magnitude of health benefits within the definition of moderate drinking.
Analysis Issues
Confounders, Mediators, and Effect Modifiers
There are many additional biologic and behavior factors including demographic factors (e.g., age, sex, genetic ancestry, socioeconomic status),
social determinants of health, lifestyle factors (e.g., diet, physical activity, tobacco, recreational drug use), and comorbid health conditions (e.g., obesity, blood pressure, diabetes, dyslipidemias) to consider in research on the impact of alcohol consumption on health where the factors may act as confounders, effect modifier and/or mediators.
Age
Age is a major determinant of health and, more specifically, it is a predictor of the types of illness that pose the greatest risk for adverse outcomes. For example, young adults are more likely to experience trauma, while older adults are more likely to experience myocardial infarction (MI). Age can also contribute to heterogeneity and moderating factors in assessing outcomes related to alcohol intake. Conducting an analysis of moderate drinking that focuses on younger adults would emphasize trauma and minimize MI and potentially lead to a different conclusion from an analysis focused on older adults. Age itself may also modify the health effects of alcohol. For example, alcohol interacts both directly and indirectly with certain medications that could interfere with the intended action of prescribed drugs and thus affect disease risk or severity. Moreover, both alcohol metabolic rate and lean mass decline with aging and contribute to higher blood alcohol concentration (BAC) per drink equivalent and reduced alcohol clearance rates in that population (Meier and Seitz, 2008).
Sex
A person’s sex (at birth) is a relevant determinant of health, and current research is limited to this binary so it is unknown how outcomes may differ for transgender individuals. Women and men differ in their alcohol pharmacokinetics. Although they metabolize alcohol at similar rates, women have lower rates of intestinal alcohol metabolism than men (5 percent in women versus 25 percent in men) and thus women absorb more alcohol into the bloodstream per drink than do men who consume an equivalent amount (Mumenthaler et al., 1999). Moreover, because alcohol is excluded from lipid compartments, the higher relative percent fat mass in women further concentrates the alcohol in their lean tissue mass (Mumenthaler et al., 1999). Thus, women experience higher BACs per drink than men, and this may increase their risk for adverse health effects, as seen in their greater risk for cirrhosis (Roerecke et al., 2019). Women and men also vary in their risk for health outcomes that may be modifiable by alcohol. For example, women are more likely than men to develop breast cancer. Lastly, perimenopause and menopausal status are important measures to include in future studies.
Genetic Ancestry
Genetic ancestry (e.g., European, East Asian) is also an important effect modifier for alcohol-related outcomes. For example, genetic variants in the enzymes that metabolize alcohol affect peak BAC and ethanol clearance, and thus modify the extent of an alcohol exposure. Another example—an enzyme variant that rapidly converts ethanol to acetaldehyde would reduce ethanol exposure but prolong acetaldehyde exposure (Edenberg, 2007). Additionally, the proteins that ethanol interacts with have genetic variants that further modify alcohol-related outcomes. For instance, variants that affect the expression or activity of proteins mediating neurotransmission can modify the risk for AUD (Gameiro-Ros et al., 2023; Zhou and Gelernter, 2024).
Additional Factors
Factors such as race and ethnicity can be confounded with socioeconomic status, health disparities, and educational differences and be further influenced by genetic variants that can modify outcomes. Race and ethnicity differences can exert additional influences on alcohol-related health outcomes. Studies must also consider socioeconomic status, as higher affluence correlates with both moderate alcohol use and factors that are protective for health, including educational attainment, better health care access, and nutritional adequacy; conversely, abstention is associated with risk factors for worsened health status including low income, reduced healthcare access, poor nutrition, and low educational attainment. Additionally, alcohol is often consumed in the context of diet, e.g., as part of the Mediterranean Diet. The committee encourages further research on how the dietary context influences the relation of moderate drinking to health.
Mediators
Analytic strategies initially should not include factors in the model that could act as mediators of a causal link between alcohol intake and outcome because this may mask or lessen the true effects of alcohol. Controlling for mediators, for example, is likely to obscure the potential “benefit” of moderate alcohol intake on an outcome. Thus, controlling for potential mediators should be used in the final analyses to assess the extent to which a measured variable explains any observed association between alcohol intake and a particular health outcome.
Causal Inference Study Designs
As noted earlier, most evidence regarding the health effects of moderate drinking is based on observational data from cohort studies. While there are
many smaller, short-term randomized controlled trials (RCT) designed to evaluate the effect of moderate drinking on intermediate outcomes—such as high density lipoprotein, low density lipoprotein, and apolipoprotein A-1 (Brien et al., 2011; Huang et al., 2017; Spaggiari et al., 2020); fibrinogen (Brien et al., 2011; Huang et al., 2017); interleukin-6 (Huang et al., 2017); and glucose control (Schrieks et al., 2015)—large RCTs evaluating actual clinical outcomes are lacking. Although large RCTs would be ideal, their design and implementation present major challenges. From a logistical standpoint, an RCT on the effect of moderate drinking, versus abstention, on risk of MI, for example, would require a large sample and long study duration. Moreover, it would require provision of alcohol to study participants, who, ideally, would be blinded to their randomly assigned group.
Ethics boards would likely be concerned about the potential health risks of assigning any abstainers to drinking for years, and there would be substantial challenges convincing abstainers to begin drinking. The alternative to asking current moderate drinkers to stop drinking for years also may be unrealistic and, over time, as the “abstention” group resumed drinking, might eventually result in a trial of moderate drinking versus moderate drinking and a spurious conclusion that moderate drinking had no impact on health. Focusing the trial on small changes in intake (e.g., asking daily drinkers to drink a little more or a little less) would tend to bias results toward null due to the small contrast between the randomly assigned groups. In short, such major challenges make it unlikely that there will be large RCTs on this important topic.
Mendelian randomization is a technique to study causal effects of modifiable exposures (e.g., moderate drinking) on health and other outcomes using genetic variants that are associated with exposures of interest (Burgess et al., 2019). However, moderate alcohol drinking is a complex and time-varying phenomenon, and currently identified gene(s) do not adequately capture individual differences in level of alcohol intake and/or drinking pattern. If the chosen genes cannot distinguish drinkers in the moderate range (e.g., 0.5 versus 1.5 drink per day), the results of Mendelian randomization studies on the health effects of moderate drinking will be biased toward the null.
RESEARCH GAPS BY TYPE OF HEALTH OUTCOME
For studies of all-cause mortality and moderate alcohol consumption, additional studies are needed to further elucidate the all-cause findings, especially because the direction of association may differ across outcomes. For example, many studies suggest that moderate drinking is associated with lower risk of myocardial infarction but higher risk of breast cancer. How do these disparate health findings affect overall (all-cause) mortality
in different socioeconomic groups? In the assessment of cardiovascular disease, studies of the association between moderate alcohol consumption and risk of major adverse cardiovascular events (MACE-3) (composite outcome) would be helpful. Moreover, there is a need for more studies that focus on the relationship of moderate drinking to other types of cardiovascular disease (e.g., heart failure, specific arrhythmias, and stroke).
With respect to weight changes, studies should include validated measures of adiposity, such as body composition measured via bioelectrical impedance analysis (BIA), with appropriate adjustment for factors such as hydration, or dual energy x-ray absorptiometry (DXA) instead of focusing on measures with established limitations like body weight, body mass index (BMI), and weight categories such as overweight and obesity defined by BMI. This is also relevant for studies of weight changes for women who are lactating because postpartum shifts in fluid balance similarly confound assessments that rely on body weight and BMI instead of body composition. In resource-poor settings or very large cohort studies where BIA and/or DXA) are unavailable, waist circumference and waist-to-hip ratio measures may still be better options than BMI. Further, self-reported measures of body size including weight and height are less desirable than relying on standardized, validated measures obtained by trained staff. Finally, studies assessing impacts of moderate alcohol consumption on weight-related outcomes should also assess dietary intake (e.g., energy-yielding nutrients, kcal/d), and measures of activity, sleep, and energy expenditure.
Several research gaps were identified for cancer of various sites. For breast cancer, further examination of moderate alcohol consumption by menopausal status is needed to determine if there are differences in those strata, particularly to determine risk associated with moderate alcohol consumption for premenopausal breast cancer and tumor type. Further examination of moderate alcohol consumption and colorectal cancer is warranted. The finding of a modest risk but with confidence intervals that include the null, merits additional consideration to determine if, with larger numbers of study participants and with greater power, a significant association would be identified. While current evidence is suggestive of a dose-response relationship, studies focusing on the dose-response relationship within the moderate consumption range are also needed with careful attention to abstainer bias.
Examination of moderate alcohol consumption with risk of cancer for the other sites identified as being associated with overall alcohol consumption is needed: oral, pharyngeal, laryngeal, esophageal, and liver. Examination of moderate alcohol consumption is needed with risk of other cancer sites such as gastric, pancreas, prostate, bladder, renal, and endometrium. Lastly, research is needed for moderate alcohol consumption with cancer
risk within strata of smoking status, especially for those cancer sites with smoking as a strong risk factor.
When considering neurocognition, diagnoses of dementia or Alzheimer’s disease must be made by medical professionals and follow established guidelines, such as Diagnostic and Statistical Manual of Mental Disorders (DSM)-5 or International Classification of Diseases (ICD)-10. Cognitive assessments should use standardized tests that are well-accepted to assess cognitive capacity (e.g., Montreal Cognitive Assessment [MoCA]; Mini-Mental State Examination [MMSE]) and should be performed, at a minimum, at two distinct ages to capture potential differences in cognitive performance and change in drinking patterns. Abstainers of alcohol to low drinking comparison groups are essential to account for test practice effects, also known as testing experience, that can endure over decades and under-estimate disease-related impairments. Additionally, focusing on one to a dozen variables as potential moderators of cognitive decline, impairment, or dementia may be inadequate to determine with confidence a direct correlation between current drinking amount by category and cognitive outcome. This includes a consideration of genetic influences that in themselves affect the risk for developing dementia-related disorders. Comorbidities are also common concomitants of drinking. For example, some people may use alcohol to self-medicate against certain psychiatric symptoms, notably depression, anxiety, obsessive-compulsiveness, traumatic stress, learned helplessness, and more. Other comorbidities include infections such as HIV or hepatitis C, nonalcohol illicit drug use, and misuse of tobacco and cannabis, which is legal in many U.S. states. Aging, sex, race and ethnicity, and socioeconomic status are also leading factors that have been shown to influence cognitive status (Delker et al., 2016; Sullivan et al., 2023).
With respect to alcohol consumption during lactation, studies are needed to evaluate the impact of acute and chronic maternal alcohol consumption on holistic milk composition and infant milk consumption. Milk composition varies within and among individuals, and factors affecting this variability (e.g., time of day, time within feed, time postpartum, maternal diet, physical activity, and body composition) should be accounted for, and controlled for if possible, using optimized and standardized collection methodology. For instance, complete breast expressions should be obtained when assessing milk composition, and infant milk consumption should be estimated using validated methods (e.g., test weighing or use of stable isotopes). Otherwise, research findings related to the potential impact of maternal alcohol consumption during lactation on milk composition or production are not useful. With respect to infant outcomes, future studies must control for confounding variables such as the influence of prenatal versus postnatal maternal alcohol use and the extent and duration of breastfeeding. A non-breastfeeding (formula feeding) cohort should be included as a reference
group. Finally, the committee concurs that the lack of research on alcohol consumption during lactation reflects the overall lack of alcohol research involving women (NASEM, 2024). Although logistic and experimental challenges certainly exist, the committee urges all studies that address the impact of alcohol consumption on human health to include postpartum women (both breastfeeding and non-breastfeeding) and their infants when possible.
REFERENCES
Afshar, M., K. Baker, J. Corral, E. Ross, E. Lowery, R. Gonzalez, E. L. Burnham, R. A. Callcut, L. Z. Kornblith, C. Hendrickson, E. J. Kovacs, and C. Joyce. 2022. Internal and external validation of an alcohol biomarker for screening in trauma. Annals of Surgery 276(6):e961–e968.
Bhattacharya, A., C. Angus, R. Pryce, J. Holmes, A. Brennan, and P. S. Meier. 2018. How dependent is the alcohol industry on heavy drinking in England? Addiction 113(12):2225–2232.
Brien, S. E., P. E. Ronksley, B. J. Turner, K. J. Mukamal, and W. A. Ghali. 2011. Effect of alcohol consumption on biological markers associated with risk of coronary heart disease: Systematic review and meta-analysis of interventional studies. BMJ 342:d636.
Burgess, S., G. Davey Smith, N. M. Davies, F. Dudbridge, D. Gill, M. M. Glymour, F. P. Hartwig, Z. Kutalik, M. V. Holmes, C. Minelli, J. V. Morrison, W. Pan, C. L. Relton, and E. Theodoratou. 2019. Guidelines for performing Mendelian randomization investigations: Update for summer 2023. Wellcome Open Research 4:186.
CDC (Centers for Disease Control and Prevention). 2024. About standard drink sizes. https://www.cdc.gov/alcohol/standard-drink-sizes/index.html (accessed September 23, 2024).
Delker, E., Q. Brown, and D. S. Hasin. 2016. Alcohol consumption in demographic subpopulations: An epidemiologic overview. Alcohol Research 38(1):7–15.
Edenberg, H. J. 2007. The genetics of alcohol metabolism: Role of alcohol dehydrogenase and aldehyde dehydrogenase variants. Alcohol Research Health 30(1):5–13.
Gameiro-Ros, I., D. Popova, I. Prytkova, Z. P. Pang, Y. Liu, D. Dick, K. K. Bucholz, A. Agrawal, B. Porjesz, A. M. Goate, X. Xuei, C. Kamarajan, C. Collaborators, J. A. Tischfield, H. J. Edenberg, P. A. Slesinger, and R. P. Hart. 2023. 5. Collaborative study on the genetics of alcoholism: Functional genomics. Genes, Brain and Behavior 22(5):e12855.
Huang, Y., Y. Li, S. Zheng, X. Yang, T. Wang, and J. Zeng. 2017. Moderate alcohol consumption and atherosclerosis: Meta-analysis of effects on lipids and inflammation. Wiener Klinische Wochenschrift 129(21-22):835–843.
Meier, P., and H. K. Seitz. 2008. Age, alcohol metabolism and liver disease. Current Opinion in Clinical Nutrition & Metabolic Care 11(1):21–26.
Mumenthaler, M. S., J. L. Taylor, R. O’Hara, and J. A. Yesavage. 1999. Gender differences in moderate drinking effects. Alcohol Research & Health 23(1):55–64.
NASEM (National Academies of Sciences, Engineering, and Medicine). 2025. A new vision for women’s health research: Transformative change at the National Institutes of Health. Washington, DC: The National Academies Press. https://doi.org/10.17226/28586 (accessed December 12, 2024).
NIAAA (National Institute on Alcohol Abuse and Alcoholism). n.d. What is a standard drink. https://www.niaaa.nih.gov/alcohols-effects-health/overview-alcohol-consumption/what-standard-drink (accessed September 19, 2024).
Nielsen, D. G., K. Andersen, A. Sogaard Nielsen, C. Juhl, and A. Mellentin. 2021. Consistency between self-reported alcohol consumption and biological markers among patients with alcohol use disorder - a systematic review. Neuroscience & BiobehavioralReviews 124:370–385.
Ravelli, M. N., and D. A. Schoeller. 2020. Traditional self-reported dietary instruments are prone to inaccuracies and new approaches are needed. Frontiers in Nutrition 7:90.
Roerecke, M., A. Vafaei, O. S. M. Hasan, B. R. Chrystoja, M. Cruz, R. Lee, M. G. Neuman, and J. Rehm. 2019. Alcohol consumption and risk of liver cirrhosis: A systematic review and meta-analysis. American Journal of Gastroenterology 114(10):1574–1586.
Schrieks, I. C., A. L. Heil, H. F. Hendriks, K. J. Mukamal, and J. W. Beulens. 2015. The effect of alcohol consumption on insulin sensitivity and glycemic status: A systematic review and meta-analysis of intervention studies. Diabetes Care 38(4):723–732.
Spaggiari, G., A. Cignarelli, A. Sansone, M. Baldi, and D. Santi. 2020. To beer or not to beer: A meta-analysis of the effects of beer consumption on cardiovascular health. PLoS One 15(6):e0233619.
Stockwell, T., J. Zhao, T. Greenfield, J. Li, M. Livingston, and Y. Meng. 2016. Estimating under- and over-reporting of drinking in national surveys of alcohol consumption: Identification of consistent biases across four English-speaking countries. Addiction 111(7):1203–1213.
Stockwell, T., J. Zhao, A. Sherk, J. Rehm, K. Shield, and T. Naimi. 2018. Underestimation of alcohol consumption in cohort studies and implications for alcohol’s contribution to the global burden of disease. Addiction 113(12):2245–2249.
Subbaraman, M. S., N. Mulia, W. C. Kerr, D. Patterson, K. J. Karriker-Jaffe, and T. K. Greenfield. 2020. Relationships between us state alcohol policies and alcohol outcomes: Differences by gender and race/ethnicity. Addiction 115(7):1285–1294.
Sullivan, E. V., S. Lannoy, A.-P. Le Berre, R. Fama, and A. Pfefferbaum. 2023. Alcohol drinking and alcohol use disorder across the ages: Dynamic effects on the brain and function. In APA handbook of neuropsychology: Neurobehavioral disorders and conditions: Accepted science and open questions, edited by T. Z. King, G. G. Brown, K. Y. Haaland, and B. Crosson. American Psychological Association. Pp. 569–607.
Zhou, H., and J. Gelernter. 2024. Human genetics and epigenetics of alcohol use disorder. Journal of Clinical Investigation 134(16).
