Moderate alcohol consumption and risk of depression: A longitudinal analysis in community-dwelling older adults

Mohebbi, M., Davoodian, N., Ganjali, S., Beilin, L.J., Berk, M., Forbes, M., McNeil, J.J., Nelson, M.R., Ryan, J., Wolfe, R., Woods, R.L., Lotfaliany, M. Nutrients. 2025 Aug 20;17(16):2688.
Abstract
Background/Objectives: Evidence suggests a J-shaped association between alcohol consumption and depression, but it remains unclear whether this reflects a true causal effect, reverse causation, or methodological bias. This uncertainty is particularly relevant in older adults, who are at increased risk for both depression and alcohol-related harms.Aim: This study aimed to examine the association between varying levels of alcohol consumption and depression risk in community-dwelling older adults.
Methods: We analysed 16,563 community-dwelling older adults (mean age 75.1 ± 4.6 years) from the ASPirin in Reducing Events in the Elderly (ASPREE) trial. Alcohol intake, reported at baseline and follow-up, was categorised as abstinent, occasional, moderate, or above guideline. Both intention-to-treat (classified by baseline alcohol consumption, regardless of later changes) and per-protocol (using annual time-updated alcohol consumption) analyses were performed. To address confounding, informative censoring, and selection bias, we applied marginal structural models with inverse probability weighting.
Results: In per-protocol analyses, abstainers (OR 1.17), occasional drinkers (OR 1.11), and above-guideline drinkers (OR 1.15) were significantly associated with a higher risk of depression compared with moderate drinkers, consistent with a J-shaped association. Sensitivity analyses excluding former drinkers and those with baseline depressive symptoms showed similar results. The association remained robust after adjusting for social isolation, social support, social interactions, physical activity, pain, sleep duration, sleep difficulties, and sleep medication use (n = 14,892; Australian sub-sample), and did not differ by sex.
Conclusions: Moderate alcohol consumption was associated with the lowest depression risk, confirming a J-shaped relationship after comprehensive confounder adjustment.
ISFAR Summary
Depression significantly adds to the global disease burden. Alcohol and depression are closely connected, and their relationship is bidirectional: alcohol can contribute to depression, while depression can also lead to alcohol use. This large epidemiological study reports, as previous studies have shown, a lower risk of depression among moderate alcohol consumers compared to abstainers and those who drink above guidelines.
The authors studied a well-characterised, apparently healthy, large group of older adults and employed various statistical techniques to both reduce confounding and assess a cause-and-effect relationship. However, Forum members believe that a cause-and-effect relationship has not been established since the aetiology of depression remains unknown. They note that depressive symptoms are evaluated rather than depression itself, effect sizes are small, and important confounders such as living arrangements and drinking patterns have not been adequately controlled for.
ISFAR Critique
Background
Depression significantly contributes to the global disease burden (Moussavi et al., 2007, Thapar et al., 2022), with a high incidence and frequent recurrence. The worldwide prevalence of depression is approximately 5.7% (Marwaha et al., 2023). Alcohol and depression are closely linked, and their relationship is bidirectional: alcohol can contribute to depression, while depression can also lead to alcohol use, creating a self-perpetuating cycle. This is because alcohol is a depressant that disrupts neurotransmitters such as serotonin and dopamine, which regulate mood. Heavy or long-term drinking can impair their normal function, resulting in low mood. There are public health concerns about the long-term effects of alcohol consumption on depression. Research has shown that alcohol misuse or heavy drinking is associated with depression, generally indicating an increased risk. Most studies have documented a higher risk of depression linked to heavy and binge drinking (An and Xiang, 2015, Paljärvi et al., 2009).
Few prospective studies have been conducted to assess this association among moderate drinkers, and most have been undertaken in younger adult populations, which have found a non-linear relationship. Compared with abstainers, light to moderate alcohol consumption may reduce the risk of depression (Liang et al., 2021), while heavier consumption increases risk, creating a J-shaped pattern (Kirchner et al., 2007, Gea et al., 2012, Gea et al., 2013, Li et al., 2020, Visontay et al., 2023). A meta-analysis by Li et al. (2020) reported J-shaped associations in both categorical and dose-response analyses where light-moderate drinking had a significantly decreased risk of depression, while heavy drinking did not show a significant association with depressive symptoms compared with non-drinkers (Li et al., 2020). Feng et al. (2024) also found that while alcohol consumption can alleviate major depression, increasing the frequency of alcohol consumption can aggravate it.
For example, these large observational studies have found that light or moderate alcohol consumption (often defined as up to 1 drink/day for women, 1 to 2 for men) is associated with a lower risk of developing depression compared to abstainers. Possible explanations are that moderate drinking may increase social interaction and reduce social isolation (protective against depression). Alternatively, small amounts of alcohol may transiently boost mood-related neurotransmitters like dopamine. However, this is association, not causation — people who drink moderately may also differ in lifestyle such as diet, exercise and social factors. Once consumption exceeds moderate levels, the risk of depression rises sharply. Heavy drinking causes neurochemical disruptions (serotonin, dopamine), sleep issues, and heightened stress responses with social, financial, and health consequences. Interestingly, some studies indicate that abstainers—especially lifelong non-drinkers—may have a marginally higher risk of depression than moderate drinkers. The reasons are debated; for example, some abstainers are “sick quitters” or people who stopped drinking due to health problems, including prior depression or alcohol misuse, or that cultural and social contexts really matter because in societies where moderate drinking is common, abstainers may have less social integration.
This paper is from the ASPirin in Reducing Events in the Elderly (ASPREE) study, which was a large, government-funded international clinical trial that investigated whether daily low-dose aspirin prevents age-related illnesses in healthy older adults (https://aspree.org/aus/). It is a long-term, multi-centre, bi-national study into how aspirin supports the health of older adults. It enrolled 19,114 initially healthy older adults (Australians aged 70+ years and U.S. minorities aged 65+ years) who were free of major illnesses such as cardiovascular disease and dementia, severe psychiatric disease and significant physical disability. The study excluded participants with baseline psychotropic medication use and then modelled medication initiation as time-varying confounders. The main finding of Mohebbi et al. (2025) is also a J-shaped association with the lowest depression odds in moderate drinkers; compared with moderate drinkers, abstainers had an OR of 1.17 (95% CI 1.08–1.26), occasional drinkers had an OR of 1.11 (1.03–1.19), and above-guideline drinkers had an OR of 1.15 (1.07–1.25). However, this specific subject selection limits generalisability to less healthy or institutionalised older adults.
Critique
This is a large, well-characterised sample of older adults (ASPREE) with repeated measures, providing strong statistical power and temporal insights. Modern causal-inference tools suitable for datasets with time-varying confounders were employed, including time-updated exposure, use of per-protocol marginal structural models (MSM), inverse probability weighting (IPW), and estimating equations (MSMs/IPW) (Breskin et al., 2019). These approaches likely enhanced their estimations, which could have been influenced by prior exposure, such as medication use. They markedly improve upon crude cross-sectional analyses.
The authors also attempt to address selection bias and censoring by using inverse probability of censoring weights (IPCW) and trimming extreme weights to avoid unstable estimates; they assess covariate balance post-weighting (Dong et al., 2021). In addition, the authors perform transparent sensitivity analyses, excluding former drinkers (to probe the “sick quitter” effect), excluding baseline depressive cases, and attempting alternative CES-D cutoffs; the results of these analyses were mainly robust. They also report E-values to quantify how strong an unmeasured confounder would need to be to explain away the association. As older adults are a group with a higher baseline risk of both depression and alcohol harms, the authors report subgroup checks such as sex and social factors, and analyses that adjust for many comorbidities and medications.
There are, however, important reasons to exercise caution when interpreting these findings as “moderate drinking protects against depression.” For instance, observational data remain vulnerable to unmeasured confounding. MSMs/IPTW enhance causal inference but depend on the assumption that there is no unmeasured confounding and that the treatment and censoring models are correctly specified. The reported E-values of approximately 1.5–1.6 indicate that an unmeasured confounder with only moderate strength could bring the OR down to null, not an implausibly large effect. In essence, MSMs lessen but do not eliminate the risk of confounding. Additionally, the study used moderate drinkers as the reference group (rather than never-drinkers), so the results are framed as “abstainers have higher odds than moderate drinkers.” While this choice is justifiable (the authors justify it based on prior ASPREE findings), it also raises the likelihood that the paper will emphasise the potential benefits of moderate drinking. It is crucial to compare moderate drinkers with never drinkers and to differentiate between former drinkers. Sensitivity analyses excluding former drinkers are conducted, but residual bias may still exist. Beverage type is not the main focus of the study, and although the authors mention wine/Mediterranean pattern analyses in previous ASPREE research, they do not thoroughly explore beverage type and context, such as consumption with meals versus alone, which could influence social integration effects.
The outcome is also a screening tool (CES-D-10) but not a clinical diagnosis, as it measures symptoms over the past week, resulting in a classification of “probable depression/clinically relevant symptoms,” rather than confirmed major depressive disorder. Consequently, misclassification (false positives/negatives) can bias the results. In addition, alcohol exposure is self-reported and subject to misclassification and underreporting, as older adults, especially those with cognitive decline, may underreport heavy drinking (Kamsvaag et al., 2021) and recall errors or social desirability bias can produce differential misclassification. If misclassification differs by emerging depression or by social factors, bias can result. Although the study tried to capture frequency, volume, and binge episodes, objective and sensitive biomarkers such as phosphatidylethanol, which is a direct biomarker of alcohol consumption indicating recent and heavy alcohol use for up to 28 days, and carbohydrate-deficient transferrin, which is an indirect, long-term biomarker for chronic heavy alcohol consumption, were not used (Aboutara et al., 2022).
The magnitude and clinical relevance of effect sizes are modest, with ORs close to 1 (for example, abstainer 1.17), so even if these are accurate, the absolute risk differences in a relatively healthy elderly sample may be minor. They should be weighed against known harms of alcohol, such as falls, cancer, and medication interactions. The authors do not (and should not) suggest that drinking is beneficial for mood. Although the study uses advanced methods that make this a better causal attempt than many observational studies, MSMs are only as good as the measured confounders and the correct specification of weight models. The e-values reported (approximately 1.5–1.8), which assess the strength of evidence for causality in observational studies, especially against unmeasured confounding, indicate moderately low robustness. This suggests that an unmeasured confounder with only moderate associations with both alcohol use and dementia risk—such as social support quality not fully captured, lifetime trauma, unmeasured socioeconomic factors, or genetic predispositions—could explain the observed association. Therefore, the evidence only points to a J-shaped relationship but does not definitively prove a protective causal effect of moderate alcohol consumption on depression.
Specific comments
Forum member Romano remarks that “the study by Mohebbi et al. (2025) investigates the relationship between moderate alcohol consumption and the risk of depression in community-dwelling older adults, using data from the ASPREE trial. The authors identified a J-shaped pattern, where moderate drinkers had a lower risk of depressive symptoms compared to abstainers, former drinkers, and those exceeding recommended alcohol limits. The sample is large and well-characterised, and the statistical methods employed—marginal structural models (MSM) with inverse probability weighting (IPW), sensitivity analyses, and E-values—are all appropriate for causal inference in longitudinal data with time-varying confounders. The authors also addressed selection bias and censoring through weighting and trimming of extreme weights. However, the results should be interpreted with caution. The reliance on self-reported alcohol consumption, the heterogeneity of abstainers (a mix of lifetime abstainers and former drinkers with potentially poorer baseline health), and the absence of a clinical depression diagnosis limit causal interpretation. Furthermore, effect sizes were modest (ORs close to 1), making clinical relevance limited and needing to be weighed against well-established alcohol-related risks.
Depression is a multifactorial disorder resulting from the interaction of biological factors (genetics, neurotransmitters, hormonal imbalances), psychological factors (early-life trauma, low self-esteem, negative thinking patterns), and social factors (isolation, chronic stress, loss, family or work-related problems). Chronic diseases, medications, and major life changes can also act as triggers. Light alcohol consumption may be associated with a slightly lower risk of depression in some studies, but an true protective effect has not been demonstrated. This association is more likely due to social and lifestyle factors. Abstainers, particularly former drinkers, may have poorer baseline health, which can bias comparisons and partly explain the apparent protection seen among moderate drinkers.
Thus, this study supports a J-shaped association between moderate alcohol consumption and depressive symptoms in older adults but does not establish a causal protective effect. The public health implications are limited, and alcohol should not be recommended as a preventive measure against depression. The study reflects the clinical complexity of depression and provides stronger evidence than many prior observational studies, although methodological limitations warrant cautious interpretation.
Forum member Ellison considers this is an excellent paper. “The authors spent considerable time seeking to evaluate mechanisms that either support or do not support the J-shaped association between alcohol consumption and depression found in their analyses. Such research can be crucial in determining the extent to which the observed associations may be causal. These authors were also diligent in seeking to determine how confounding factors may have affected their results. Perhaps surprisingly, they found that the effects of most of the generally accepted potentially confounding variables were minor.
The potential health benefits regarding depression shown in these analyses should stimulate future research into mechanisms by which moderate drinking can influence the risk of depression. While the vast majority of previous studies on alcohol and health have clearly shown strong inverse associations between moderate alcohol consumption and cardiovascular diseases, diabetes, and total mortality, increasing research is needed on alcohol and dementia, as well as on depression, as these two conditions contribute immensely to the burdens of health care for ageing populations around the world.
My main concern with this paper was the use of Mendelian randomisation results to negate a J-shaped curve based on observational data. We obviously will become better at evaluating self-reported information on the exposure to alcohol with improved assessment techniques. But the true health effects of alcohol relate more to the drinking pattern (that is, the type of beverage, consumed with or without food, regularly or only on a few days/week, associated or not with binge drinking, etc.), which is more important than just the total number of drinks consumed over a period of time. And while genetic factors are obviously important, they do not predict why some people drink to excess very well, nor do they account for the effects of social influences, the drinking patterns of the culture in which someone lives, or socioeconomic factors (Ellison et al., 2021). Emerging approaches for better estimating patterns of drinking should provide better data for which to judge the relation of alcohol intake to dementia.”
Forum member Skovenborg also states that “this is an excellent and well-done study. My only concern might be that they use the CES-D to diagnose depression. They themselves say that you can’t, but they use the word anyway. The CES-D has good sensitivity (around 80-90%), but very low specificity (around 40%). That is, the positive predictive value is woefully low. Exactly what this means for the study’s conclusions, I can’t quite figure out, but perhaps you should read “symptoms of depression and anxiety” where the authors write “depression”.
Forum member Harding considers that “the association of lower risk of depression in moderate drinkers compared to above guideline drinkers was observed to be small. In the Discussion section, the authors made a concerted effort to demonstrate that this is a causal relationship, which was always likely to be an uphill battle given that the causes of depression are not yet fully understood. I did find the last sentence of the first paragraph curious ‘our findings provide preliminary evidence to suggest that the observed relationship may reflect a possible causal link between alcohol consumption and the risk of depression in older adults, rather than being merely artifacts driven by methodological biases.’ Why one or the other? Couldn’t the association be a mere coincidence?
In support of this, the authors cite a paper by Adams et al. (2020) that reports an adverse immune response in chronic heavier drinkers (that is, those with Alcohol Use Disorder), but such drinkers are a long way from those in the cohort who were drinking above the alcohol drinking guidelines. Similarly, regarding the effect of alcohol on sleep, they refer to a paper by Kenney et al. (2013), which concerned 1,044 heavy-drinking college students. Such comparisons are not appropriate. They could not find any evidence that moderate drinking facilitated beneficial social interaction (and therefore less likely to lead to depression) and so concluded that such interactions are less important in later life, or that their database was not good enough to capture it. Similarly, for physical activity. As for sex differences, they said, “Although we observed no statistically significant sex differences, potential residual confounding means that this finding should be interpreted cautiously.” I don’t understand why. They found no difference.
For these reasons, I don’t think the sentence in the Conclusions section, ‘The findings suggest that the protective effects of moderate consumption may not be solely attributable to methodological bias’, can be justified.”
Forum member Waterhouse similarly muses that “the author’s pursuit of a causal link is fraught with challenges. One datapoint caught my eye-the study population was 54% female, but the moderate alcohol group was only 33% female. I am not sure what to make of this. Another was that 73% of moderate drinkers lived at home with another, while this was the case with only 64% of abstainers and occasional drinkers. In the Introduction section, they mention “moderate drinking often occurs in socially engaging contexts, with higher social integration and activity levels linked to lower depression risk”, but unfortunately, I did not see them pursue this issue in their data analysis. I say unfortunate because this may be one means by which moderate consumption is associated with lower levels of depression. However, the association may be linked to the living situation. In other words, is it possible that living in a better socially integrated home would lead to moderate consumption?”
References
Aboutara, N., Müller, A., Jungen, H., Szewczyk, A., van Rüth, V., Bertram, F., Püschel, K., Heinrich, F., Iwersen-Bergmann, S. (2022) Investigating the use of PEth, CDT and MCV to evaluate alcohol consumption in a cohort of homeless individuals- A comparison of different alcohol biomarkers. Forensic Sci Int., 331:111147. https://doi.org/10.1016/j.forsciint.2021.111147.
An, R., Xiang, X. (2015) Smoking, heavy drinking, and depression among U.S. middle-aged and older adults. Prev. Med., 81:295-302.
ASPREE XT. ASPirin in reducing events in the elderly-eXTension A study of the long term effects of aspirin and other factors on the health of older persons – protocol. https://aspree.org/aus/wp-content/uploads/sites/2/2021/05/ASPREE-XT-Protocol-v5-April-2021.pdf#:~:text=ASPREE%20(ASPirin%20in%20Reducing%20Events%20in%20the,in%202010%20and%20completed%20recruitment%20in%202014.
Breskin, A., Cole, S.R., Westreich, D. (2018) Exploring the Subtleties of Inverse Probability Weighting and Marginal Structural Models. Epidemiology, 29(3):352-355. https://doi.org/10.1097/EDE.0000000000000813.
Dong, G., Mao, L., Huang, B., Gamalo-Siebers, M., Wang, J., Yu, G., Hoaglin, D.C. (2020) The inverse-probability-of-censoring weighting (IPCW) adjusted win ratio statistic: an unbiased estimator in the presence of independent censoring. J Biopharm Stat., 30(5):882-899. https://doi.org/10.1080/10543406.2020.1757692.
Ellison, R.C., Grønbæk, M., Skovenborg, E. (2021) Using Mendelian randomization to evaluate the effects of alcohol consumption on the risk of coronary heart disease. Drugs & Alcohol Today, 21:84-95. https://doi.org/10.1108/DAT-09-2020-0061.
Feng, W., Zhang, B., Duan, P., Bi, Y.H., Jin, Z., Li, X., Zhao, X., Zuo, K. (2024) Risk of major depressive increases with increasing frequency of alcohol drinking: a bidirectional two-sample Mendelian randomization analysis. Front Public Health., 12:1372758. https://doi.org/10.3389/fpubh.2024.1372758.
Fredman Stein, K., Allen, J.L., Robinson, R., Smith, C., Sawyer, K., Taylor, G. (2022) Do interventions principally targeting excessive alcohol use in young people improve depression symptoms?: a systematic review and meta-analysis. BMC Psychiatry, 22(1):417. https://doi.org/10.1186/s12888-022-04006-x.
Gea, A., Beunza, J.J., Estruch, R., Sánchez-Villegas, A., Salas-Salvadó, J., Buil-Cosiales, P., Gómez-Gracia, E., Covas, M.I., Corella, D., Fiol, M., Arós, F., Lapetra, J., Lamuela-Raventós, R.M., Wärnberg, J., Pintó, X., Serra-Majem, L., Martínez-González, M.A.; PREDIMED GROUP. (2013) Alcohol intake, wine consumption and the development of depression: the PREDIMED study. BMC Medicine, 11:192. https://doi.org/ 10.1186/1741-7015-11-192.
Gea, A., Martinez-Gonzalez, M.A., Toledo, E., Sanchez-Villegas, A., Bes-Rastrollo, M., Nuñez-Cordoba, J.M., Sayon-Orea, C., Beunza, J.J. (2012) A longitudinal assessment of alcohol intake and incident depression: the SUN project. BMC Public Health, 12:954. doi.org/10.1186/1471-2458-12-954.
Hammerton, G., Lewis, G., Heron, J., Fernandes, G., Hickman, M., Lewis, G. (2023) The association of alcohol dependence and consumption during adolescence with depression in young adulthood, in England: a prospective cohort study. Lancet Psychiatry, 10(7):490-498. https://doi.org/10.1016/S2215-0366(23)00138-4. Epub 2023 Jun 1. Erratum in: Lancet Psychiatry. 2023 Jul;10(7):e18. https://doi.org/10.1016/S2215-0366(23)00198-0.
Kamsvaag, B., Bergh, S., Šaltytė Benth, J., Selbaek, G., Tevik, K., & Helvik, A. S. (2021). Alcohol consumption among older adults with symptoms of cognitive decline consulting specialist health care. Aging & Mental Health, 26(9):1756–1764. https://doi.org/10.1080/13607863.2021.1950618
Lasserre, A.M., Imtiaz, S., Roerecke, M., Heilig, M., Probst, C., Rehm, J. (2022) Socioeconomic status, alcohol use disorders, and depression: A population-based study. J Affect Disord., 301:331-336. https://doi.org/ 10.1016/j.jad.2021.12.132.
Li, J., Wang, H., Li, M., Shen, Q., Li, X., Zhang, Y., Peng, J., Rong, X., Peng, Y. (2020) Effect of alcohol use disorders and alcohol intake on the risk of subsequent depressive symptoms: A systematic review and meta-analysis of cohort studies. Addiction, 115:1224–1243.
Liang, L., Hua, R., Tang, S., Li, C., Xie, W. (2021) Low-to-moderate alcohol intake associated with lower risk of incidental depressive symptoms: A pooled analysis of three intercontinental cohort studies. J Affect Disord., 286:49-57. https://doi.org/10.1016/j.jad.2021.02.050.
Kirchner, J.E., Zubritsky, C., Cody, M., Coakley, E., Chen, H., Ware, J.H., Oslin, D.W., Sanchez, H.A., Durai, U.N.B., Miles, K.M., et al. (2007) Alcohol consumption among older adults in primary care. J. Gen. Intern. Med., 22: 92–97.
Marwaha, S., Palmer, E., Suppes, T., Cons, E., Young, A.H., Upthegrove, R. (2023) Novel and emerging treatments for major depression. The Lancet (London England), 401(10371):141–53.
Moussavi, S., Chatterji, S., Verdes, E., Tandon, A., Patel, V., Ustun, B. (2007) Depression, chronic diseases, and decrements in health: results from the World Health Surveys. The Lancet, 370(9590):851-8. https://doi.org/10.1016/S0140-6736(07)61415-9.
Paljärvi, T., Koskenvuo, M., Poikolainen, K., Kauhanen, J., Sillanmäki, L., Mäkelä, P. (2009) Binge drinking and depressive symptoms: a 5-year population-based cohort study. Addiction, 104(7):1168-78. https://doi.org/ 10.1111/j.1360-0443.2009.02577.x.
Qi, P., Huang, M., Zhu, H. (2024) Association between alcohol drinking frequency and depression among adults in the United States: a cross-sectional study. BMC Psychiatry, 24: 836. https://doi.org/10.1186/s12888-024-06296-9.
Thapar, A., Eyre, O., Patel, V., Brent, D. (2022) Depression in young people. The Lancet (London England), 400(10352):617–31.
Visontay, R., Mewton, L., Slade, T., Aris, I.M., Sunderland, M. (2023) Moderate alcohol consumption and depression: A marginal structural model approach promoting causal inference. Am. J. Psychiatry, 180: 209–217.
Comments on this critique by the International Scientific Forum on Alcohol Research were provided by the following members:
Henk Hendriks, PhD, Independent consultant and partner of the Nutrition Consultants Cooperative, Netherlands
Creina Stockley, PhD, MBA, Independent consultant and Adjunct Senior Lecturer in the School of Agriculture, Food and Wine at the University of Adelaide, Australia
Raquel Romano, PhD, Independent consultant and Professor of Applied Technology at the University of Aconcagua, Argentina
R. Curtis Ellison, MD, 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
Richard Harding, PhD, Formerly Head of Consumer Choice, Food Standards and Special Projects Division, Food Standards Agency, UK
Andrew L. Waterhouse, PhD, Department of Viticulture and Enology, University of California, Davis, US