Studies have been selected according to the criteria outlined below taking up of the PICOS method. Acronym stands for: Population, Intervention which is in our review defined as the exposure, Comparators, Outcome and Study design.

Four electronic bibliographic databases were used for the systematic search: PubMed (Medline), Science direct, Embase and Europe PMC. To ensure literature saturation, we searched the reference lists of included studies and relevant journals identified during the search.

The key words used for the bibliographic search were “Alcohol - Mendelian randomization – Cancer - Cardiovascular disease - Liver disease - Neurological disease-”. They were defined by two examiners (NM and BN). The search was carried out taking into account two methods: a search with “MeSH” terms and a search with “free text” terms in order to correctly translate the concepts and synonyms into English. A separate search was carried out for each type of disease. The keywords were combined with the Boolean operators “AND, OR” to form the search equations represented in the Supplementary Table S1 that we introduced in the “PubMed”, “Science Direct”, “Embase” and “Europe PMC”.

The articles were entered in the RAYYAN QCRI software (47), we then proceeded to count and remove the duplicates. Titles and abstracts were independently reviewed by two reviewers (NM and BN) respecting the inclusion and exclusion criteria. Full-text studies were also independently reviewed by the same two reviewers. Disagreements were resolved through discussion with the participation of a third reviewer (AHJ).

Using Microsoft Excel®, a standard data collection table was used to extract data from each study. Data was extracted by one investigator (BN) and verified by another investigator (NM). Disagreements were resolved through discussion and the participation of a third investigator (AHJ). When additional information was needed, the study authors were contacted.

The following items were taken from each included study: Reference, Name of the first author, date of publication, Sample size, Ancestry, Sex, Age, Instrumental variable associated to alcohol, Exposure dataset, outcome, Outcome dataset and results.

The quality of the studies was independently assessed by three investigators (NM, BN and CD) by using first the Q-Genie tool “Quality of Genetic Association Studies” which contains 11 items rated on a Likert scale. Seven points covered the following topics: rational for the study, selection and definition of the outcome of interest, selection and comparability of comparison groups, technical classification of exposure, non-technical classification of exposure, other sources of bias, sample size and power, a priori planning of analyzes, statistical methods and control of confounding factors (48). Second, since the Q-genie tool is not quality assessment tool for MR studies and there was no formal predefined protocol for systematic review of MR to date, we therefore rated other key elements of the MR design including the assumptions that are crucial to the validity of MR as showed in Figure 1: -

Validation of IV1 – Genetic variants are strongly associated with the exposure: Relevance assumption

The validation of the first assumption in MR studies is evaluated by the selection of strongly associated SNPs with the exposure of interest (P-values of the variant-exposure association <5 × 10⁻⁸). Moreover, to test the robustness of the instrument, the F-statistic of the variant-exposure association is usually provided. Values >10 suggest an adequate instrument with a low risk of bias due to a weak instrument (49). The F-statistic depends on the proportion of the variance of the exposure explained by the genetic variants (r²) (50), the sample size and the number of IVs included in the instrument. -

Validation of IV2 – Genetic variants are not related to confounders of the exposure-outcome association: Independence assumption

Taking into account the random allocation of genetic variants at conception, the IV2 of no association between the IVs and confounders is often fulfilled because of the random allocation of alleles to gametes. In a one-sample MR, in which the variant-exposure and variant-outcome association are computed using individual data from the same sample, the association between the genetic variant and the observed confounders should be tested. However, in a two-sample MR, the variant-exposure and variant-outcome association come from two independent samples, most often using published summary statistics of large genome wide association studies (GWAS), which does not allow to test this hypothesis. -

Validation of IV3 – Genetic variant does not affect the outcome except through the exposure: Exclusion restriction assumption

Horizontal pleiotropy (i.e., a genetic variant affects other traits which influence the outcome independently of the exposure) may lead to biased MR results because the effect of the variant on the outcome is not exclusively due to the exposure (51). Through assessment of horizontal pleiotropy, the third assumption can be partially verified (52). In addition to the main methods used to estimate the causal MR effect—the two-stage least-squares (2SLS) and the inverse-variance weighted (IVW) methods in a one and two-sample design, respectively (53)—several MR methods have been developed to detect and correct for the bias due to pleiotropy, such as the weighted median (54), weighted mode (55), MR-PRESSO (56) and MR-Egger (57) These methods, originally created for a two-sample MR design, can also be used on large one-sample MR, except for the MR-Egger method which is not recommended for one-sample MR unless the correlation between the variant-exposure and variant-outcome estimates due to confounding can be kept low, or the variability in instrument strength is very high (58). -

Assessment of non-linearity

In order to clarify the causal relationships, it is valuable to identify and characterize non-linear effects when they are present. Non-linear association may result in opposite effects depending on the level of exposure. Such opposing effects have been observed particularly in many observational studies examining the relationship between alcohol consumption and cardiovascular events (59).

These assumptions were rated as “Good” if the assumption was assessed using above-mentioned approaches, “moderate” if the assumption was only described, and “poor” if the assumption was not checked or described. This evaluation of the validity of each assumption using this protocol was assessed independently by three reviewers; any inconsistencies were resolved by discussion. The quality assessment data are presented in Supplementary Tables S2–S9.

Bibliometric analyzes are represented as a flow diagram that describes the complete process of searching and selecting articles. The summary of the results is presented in tables grouping together the data detailed in supplementary information (Supplementary Tables S10–S13). We synthesized narrative genetic evidence on causal associations between alcohol consumption and cancers, cardiovascular, liver and neurological diseases. The main findings have been classified by outcome.

A flowchart summarizing the study selection process has been produced in accordance with PRISMA guidelines (Figure 2). We identified 3,066 articles from electronic database searches. After removing the duplicates (336 articles), we excluded 2,368 after reading the title and among them, 190 were excluded after reading the abstracts and 110 were excluded among the full text articles. We reviewed the full texts of 172 articles, 57 of which met our inclusion criteria. We identified 13 additional studies via backward reference searches of the 57 included studies, such that a total of 70 studies were eligible for inclusion in our systematic review. The reasons of exclusion are presented in the flow chart.

The main characteristics of the included studies are presented in the tables available in Supplementary Materials. These characteristics have focused on a wide range of outcomes: Cancer outcomes (26 studies), cardiovascular outcomes (24 studies), liver outcomes (6 studies) and neurological outcomes (17 studies). While the majority of the studies were based on populations of European ancestry (47 studies), 12 were Asian and 7 were mixed-ancestry. The majority of studies involved middle-aged and/or older participants.

As for IVs, the studies used either a single genetic variant (12 studies), a combination of genetic variants (51 studies), or a combination of both (8 studies). The variants rs1229984 and rs671, respectively of the genes encoding ADH1B and ALDH2, are the genetic variants of alcohol consumption most mentioned in these studies. The rs1229984 variant of ADHB was used as an IV of alcohol consumption in 12 studies, as was the rs671 variant of ALDH2 which was used in 7 studies.

Using the Q-Genie tool, most of the included studies were rated as good overall quality, while 12 studies were rated as moderate quality. Detailed tables of the quality assessment are available in supplementary materials (Supplementary Tables S2–S5).

The MR methodological quality of the included studies was assessed using our protocol, and the results are presented in Supplementary Tables S6–S9.

All results are synthesized in Supplementary Tables S2–S13. All confidence intervals were estimated at 95% (95% CI) and p-value were two-sided in the included studies.

Our review on the relationship between genetically predicted alcohol consumption and various cancers provides important insights, yet reveals a complex and heterogeneous set of findings.

The association of genetically-predicted alcohol consumption with cancer was assessed for all cancers in one study (60), for breast and ovarian cancers in five additional studies (62–66) and for endometrial cancer in one study (67).

Regarding breast cancer, the five MR analyzes reported no significant association except with genetically predicted problematic alcohol use (60, 62, 63, 65, 66). These results were in contradiction with the findings of observational studies showing that any drinking, starting from light to moderate alcohol consumption is associated with a dose-dependent increase in breast cancer risk (62, 132, 133). However, it is crucial to note that the MR study may have lacked sufficient statistical power to identify such a moderately sized association, as demonstrated in other studies, where there is an 8% to 12% increase in risk for each 10 g/day increment in alcohol consumption, as some authors have suggested (68). As for ovarian cancer, MR findings were consistent with those of observational studies suggesting no association between alcohol consumption and ovarian cancer (134, 135).

For endometrial cancer, MR analysis indicates a strong causal relationship between genetically predicted alcohol consumption and EC risk. Specifically, highlighting the protective effect of genetically predicted alcohol. Subgroup analyses revealed that genetically predicted alcohol consumption had a protective effect only in the endometrioid subtype of endometrial cancer (EEC), while no protective effect was observed in non-endometrioid endometrial cancer (NEC). The study suggests that the protective effect of alcohol consumption on EC is mediated by other factors, specifically a decrease in human chorionic gonadotropin (HCG), which is associated with endometrial proliferation and malignant tumors (136), and insulin-like growth factor 1 (IGF1), which is involved in the occurrence and development of EC (137).

This relationship has been controversial in observational studies. A prospective cohort study involving 41,574 participants found that daily alcohol consumption of ≥2 drinks/day increased postmenopausal EC risk [RR = 2.01 (1.3–3.11)] compared to no alcohol consumption (138). Conversely, another prospective Nurses’ Health Study by Je et al. (139) in a population of 68 067 female participants showed a decreased risk of endometrial cancer for light alcohol consumption. Women with an alcohol intake of less than 5 g per day had a 22% lower risk of endometrial cancer. When comparing categories of alcohol intake and risk of endometrial cancer, the evidence remains inconsistent. A meta-analysis by Bagnardi et al. (20), based on results from 21 observational studies, found no significant association between endometrial cancer risk and moderate (≤50 g per day) or light (≤12.5 g per day) alcohol intake compared to no alcohol intake.

The relationship between alcohol consumption and prostate cancer risk remains inconclusive in observational studies (140). While some studies suggest a potential increased risk among heavy drinkers (141), others find no significant association, particularly for low-to-moderate consumption (142). In contrast to these mixed findings, the two MR studies reported no significant association between genetically-predicted alcohol consumption and prostate cancer. However, alcohol consumption was associated with increased prostate cancer mortality in men with low-grade prostate cancer (60, 68).

Oral/oropharyngeal cancers is the site-specific cancer clearly associated with genetically-predicted alcohol consumption in our review (69). Importantly, this association remained significant even after adjusting for smoking, demonstrating alcohol's independent effect which contrasts observational studies suggesting alcohol interacts synergistically with smoking to elevate the risk of head and neck cancer (143). Observational studies have identified alcohol as a significant risk factor for head and neck cancer (144, 145). Genetic factors, such as slow ADH1B and ALDH2 enzyme activity, further increase susceptibility to alcohol-induced HNC (146). Biochemical studies further elucidate this relationship, showing that alcohol's toxicity involves its metabolic products, oral microbiota, and oxidative stress, leading to genetic and epigenetic alterations (147), suggesting carcinogenic effects of alcohol are mediated not only through systemic exposure but also through local factors (148).

The results regarding the relationship between alcohol consumption and lung cancer were contradictory. One (72) of four MR studies (60, 71–73) examining the effects on lung cancer suggested that genetically predicted habitual alcohol consumption with meals (also described as light to moderate) was a protective factor against lung cancer and squamous cell cancer lung. This alcohol consumption during meals was unrelated to lung adenocarcinoma or small cell lung cancer in the MR analysis. Similarly, another MR study suggested a decreased risk in both men and women although it was statistically insignificant in women. These results confirm observational studies supporting the idea that light to moderate alcohol consumption may prevent lung cancer, most prominently squamous cell carcinoma, and increase survival after the diagnosis (18, 149). In contrast, In Larsson et al. (60), genetically predicted alcohol drinking was significantly positively associated with lung cancer in the ILCCO but not in the UK-Biobank study. In any case, the findings reported by this study call into question a protective effect of alcohol consumption on the onset of lung cancer while another study reported no association (73). However, the association between alcohol consumption and lung cancer is still under debate for never-smokers (150).

Regarding esophageal cancer, results from the six studies were contradictory. While four studies reported a positive association between alcohol intake and esophageal risk (60, 71, 74, 76), another study reported no association with genetically predicted alcohol frequency (75). Similarly, Zhang et al. (77) reported no association with esophageal squamous cell carcinoma, however, they did find a significantly increased risk associated with the variant rs1229984 of ADH1B. Despite these discrepancies, this set of six MR studies supports previous observational studies. A dose-response meta-analysis indicated that alcohol intake significantly increases the incidence of esophageal cancer, especially for ESCC (151). Additionally, another recent dose-response meta-analysis reported that even light alcohol consumption is significantly associated with higher risks of esophageal cancer, extending beyond heavy alcohol consumption (152).

As for gastric cancer, the set of five studies overall provides little support for a causal relationship between genetically predicted alcohol consumption and the risk of gastric cancer. Four out of the five studies reported no significant association. Yuan et al. (74) did report a positive, though non-significant, association even after adjusting for smoking. Interestingly, observational data showed a similar direction in a recent meta-analysis of eighty-one epidemiological studies demonstrated that increased daily intake of alcohol was correlated with a heightened incidence of gastric cancer (153).

Seven studies assessed the association between genetically predicted alcohol intake and colorectal cancer. Of these, five studies reported an increased risk, with one study maintaining a positive association even after adjusting for smoking (74). While one study reported no association with current, never and previous alcohol intake (80), another found a negative, though non-significant, association (76). Overall, these results align with observational studies suggesting that alcohol consumption is a risk factor for colorectal cancer, particularly in Asian populations. A case-control study conducted in an Asian population revealed that being a current or former drinker was positively associated with the risk of CRC compared to controls [OR = 5.4 (1.1–27.8), p = 0.043] (154). Moreover, an updated meta-analysis of prospective cohort studies by Zhou et al. (79) demonstrated that moderate and heavy drinking were statistically significantly associated with an increased risk of colorectal cancer.

The association between genetically predicted alcohol consumption and the risk of liver cancers was analyzed in six MR studies. Consistent with previous observational studies, the MR studies demonstrated that genetically predicted alcohol consumption increases the risk of hepatocellular carcinoma. Notably, one of these studies reported a non-linear relationship, finding a positive association between pure alcohol intake (g/day) and the risk of HCC, with a statistically significant linear trend (p < 0.0001). However, these studies showed positive but non-statistically significant associations with liver cancer overall. A similar positive relationship between alcohol consumption and HCC risk in Chinese populations was reported in a meta-analysis of 18 case-control studies, which included 3,812 HCC cases and 10,927 controls. The meta-analysis found that ever drinkers had a significantly higher risk of HCC compared to never drinkers (OR = 1.56; 95% CI, 1.16–2.09) (155). Another meta-analysis of 11 case-control studies indicated a positive association between alcohol consumption and liver cancer risk, with a higher intake increasing the risk compared to lower intake (156). The underlying mechanisms explaining this association in previous studies are multifaceted. Acetaldehyde, the first metabolic product of alcohol, can induce oxidative stress and DNA damage, promoting carcinogenesis. Furthermore, chronic and heavy alcohol consumption can lead to alcoholic cirrhosis, a known precursor to HCC. Moreover, alcohol acts as a solvent, enhancing the penetration of other carcinogens into liver cells, while also impairing hepatic detoxification and immunity (156, 157).

As for pancreatic cancer, all three studies from our review did not show any significant causal association with genetically predicted alcohol drinking. These findings contrast with previous observational studies, which reported an increased risk of pancreatic cancer only with heavy drinking (158, 159) likely mediated by inflammatory pathways related to chronic pancreatitis (160).

Also, our review found no relationship between alcohol consumption and bladder cancer risk. The link between alcohol consumption and bladder cancer risk has been inconsistent in observational studies, with varying results likely due to the type of alcohol consumed (e.g., beer, wine, or spirits) rather than the amount of alcohol consumed (161, 162). There was a linear dose-response relation in those who consume alcohol from liquor or spirits (163).

Conversely, a statistically significant increased risk of cutaneous melanoma was observed, aligning with observational studies that reported an increased risk even with moderate alcohol intake. Ethanol has been shown to inhibit the developmental process of melanocytes, directly contributing to the progression of melanoma, it also enhances its metastatic ability (164–166).

Contrary to observational studies suggesting an association between moderate alcohol consumption and a lower risk of thyroid and biliary tract cancers (167, 168), our systematic review did not confirm any significant association.

Finally, genetically predicted alcohol consumption was not associated to overall cancer in the only study having all cancers as outcomes (60). This lack of association is likely to be reliable to some extent as the study showed good methodological qualities.

Overall, MR studies suggest that genetically predicted alcohol consumption is associated with an increased risk for certain cancers, particularly oral/oropharyngeal, lung, head and neck, cutaneous melanoma, colorectal, and liver cancers particularly hepatocellular carcinoma. For breast cancer, a statistically significant increased risk was only observed in association with genetically predicted problematic alcohol use. for esophageal cancer, a statistically significant association was reported in the two studies performed on Asian ancestry population. Particularly, an increased risk for esophageal squamous cell carcinoma (ESCC) in association with the ADH1B rs1229984 variant. In contrast, no significant association was found for ovarian, prostate, lung, esophageal, gastric, pancreatic, bladder, thyroid, kidney, non-Hodgkin lymphoma, biliary tract cancers. Endometrial cancer showed a unique protective effect particularly in the endometrioid subtype.

These findings highlight the complex and varied relationship between alcohol consumption and cancer risk. The inconsistencies in the results can be attributed to differences in population ancestry, sample size, and genetic variants across the studies. To obtain more definitive conclusions, it is essential to conduct studies with more generalized populations. Additionally, it is important to note that we were unable to assess possible U- or J-shaped relationships between alcohol consumption and cancer risk in this review. Future research should aim to address these gaps and explore the potential non-linear associations between alcohol consumption and cancer.

The relationship between alcohol drinking and CVD is complex, controversial and still highly debated (169). There have been a growing number of studies outlining both harmful and potentially protective effects of alcohol consumption on the risk of various cardiovascular outcomes but there is less clarity about estimates of risk with low levels of consumption (131). Observational studies have consistently reported that compared with non-drinkers, light to moderate drinking exhibits a reduced cardiovascular risk (170, 171). However, relying on the participants’ self-report, imply the possibility of reverse causality and confounding; for example, Schutte et al. demonstrated that using never drinkers as reference entails a bias leading to underestimating the cardiovascular risk. When considering the drinking reference bias and excluding ischemic heart disease and wine beverage from the analysis, there is no overall cardiovascular protection from alcohol and instead there is an association with increased cardiovascular risk even when consuming 112 g or less per week (172).

Although two meta-analyses of observational studies reported an association of light or moderate alcohol consumption with a reduced risk of heart failure (26, 173), two large two-sample MR analyzes that validated all three key assumptions and reported a good quality according to the Q-genie tool) found no association between genetically predicted alcohol intake and the risk of heart failure (40, 105). As these two MR studies did not distinguish between levels of alcohol consumption, it is difficult to confront their findings with those deriving from the observational studies.

Five MR studies examined alcohol intake and hypertension (100–102, 174, 175). All these MR studies showed a significant association of alcohol consumption with an increased risk of hypertension, with the exception of a Chinese study where no association was found in women, a result that may derive, as suggested by the authors, from the fact that Chinese culture does not encourage women to drink (101). As these MR studies were not designed to reveal nonlinear associations, no conclusion could be reached regarding the light-moderate alcohol use possible causal effect on hypertension. These findings are consistent with two recent meta-analyses of cohort studies showing an elevated risk of hypertension compared to abstainers for any amount of alcohol consumption in men, and above 20–24 g per day in women (176, 177). In addition, it has been shown that a reduction in alcohol consumption was associated with a decrease of blood pressure in a dose-dependent manner with an apparent threshold effect at 24 g of alcohol per day. Overall, the convergent findings of MR and observational studies findings supports the hypothesis of causal effect of alcohol consumption on blood pressure and hypertension risk, particularly in men.

Four MR studies investigated the association of genetically predicted alcohol consumption with CHD (including coronary artery disease and myocardial infarction) (40, 91, 95, 104). These studies mostly showed non-significant associations between CHD and genetically predicted alcohol consumption when adjusted to smoking (40, 104). One study (91) showed that a genetic variant associated with non-drinking and lower alcohol consumption had a reduced risk of CHD, an association that disappeared when the analysis was restricted to non-drinkers.

Two MR studies in our review investigated the association of genetically predicted alcohol consumption with AF. The two-sample MR study conducted by Jiang et al. (107) reported no significant association between genetically predicted alcohol consumption and the risk of AF. As this MR study did not distinguish between levels of alcohol consumption, it is difficult to confront its negative findings with those deriving from the observational studies, showing an association between heavy alcohol consumption and an increased risk of AF (178). While Larsson et al. suggested a positive significant association between genetically predicted alcohol consumption and AF [OR_IVW, 1.17, (1.00–1.37); p = 0.05], the association was no longer significant after adjustment for smoking initiation (40). All four MR studies that investigated stroke reported a positive association with genetically predicted alcohol consumption (40, 88, 91, 95). One study was particularly methodologically sound, and showed that the association resisted to the adjustment to smoking initiation (40). Another study (91) showed that a variant associated with non-drinking and lower alcohol consumption was associated with an ischemic stroke (but not with the combined subtypes of stroke), an association that disappeared when the analysis was restricted to non-drinkers. Finally, in another study performed in a Chinese population, the MR analysis showed a log-linear rather than a U-shaped association with stroke, advocating against a possible protective effect of light or moderate alcohol consumption on stroke (95).

Finally, MR studies clearly confirmed the causal effect of alcohol on the risk of increased blood pressure, hypertension, and stroke. Beyond this, as most MR studies were unable distinguish between levels of alcohol consumption, the negative findings were difficult to confront with observational studies claiming a protective effect of light or moderate alcohol consumption. There is however an exception deriving from the study from Millwood and colleagues (95) in Asian descents, where the MR analysis showed a log-linear rather than the U-shaped association found in the conventional epidemiology analysis with stroke (in men only). Thus, demonstrating that MR methodology can call into question the results from classical observational methodology when applied to the same sample.

Several MR studies have challenged the previously suggested protective effect of moderate alcohol consumption on cardiovascular disease.

Traditional observational studies had shown a potential association between moderate alcohol consumption and a reduced risk of CVD, particularly CHD. However, these studies were subject to confounding factors, such as lifestyle, diet, socioeconomic status, and reverse causality, which may have biased the results.

Several MR studies have found that the protective effects of alcohol on heart health may be overstated or nonexistent. The study of Holmes et al. (91) used genetic variants related to alcohol metabolism (particularly the ADH1B gene) to examine the causal relationship between alcohol consumption and cardiovascular health. The study found that individuals with genetic variants that lead to lower alcohol consumption had a reduced risk of coronary heart disease. This suggests that lower alcohol intake, rather than moderate consumption, is beneficial for heart health. The study did not support a protective effect of moderate drinking on CVD risk. In the MR study of Millwood et al. (95), alcohol consumption was associated with an increased risk of stroke and other cardiovascular events, even at relatively low levels. There was no protective effect observed, and the study concluded that alcohol consumption increases the risk of CVD. Larsson et al. (40) found no evidence of a protective effect of moderate alcohol consumption on coronary artery disease. Instead, the study pointed to harmful effects of alcohol on heart health.

These studies suggest that the supposed cardiovascular benefits of moderate alcohol consumption may have been overstated in earlier observational research, largely due to confounding variables. Mendelian randomization, being less prone to confounding, provides stronger evidence that alcohol consumption likely does not confer a protective effect on CVD risk and may in fact be harmful, even in moderate amounts.

Only a few MR studies have been published on liver outcomes in relation to genetically predicted alcohol intake. Our review highlighted a strong association between genetically predicted alcohol consumption and alcoholic liver disease (ALD), with a markedly increased risk consistently reported across studies. Furthermore, significant positive associations were found between alcohol consumption and cirrhosis, indicating that higher alcohol intake significantly elevates the risk of developing these severe liver conditions.

Regarding MASLD, the three studies showed mixed results. Sookoian et al. (111) found that even moderate alcohol consumption could be harmful in patients at high risk for progressing to end-stage liver disease, with lower alcohol consumption correlating with less severe histological features. Despite the small sample size and the modest level of alcohol consumption due to MASLD inclusion criteria, carriers of the A-allele exhibited lower degrees of histological steatosis, lobular inflammation, and MASLD-Activity Score compared to non-carriers. While, Yuan et al. (112) reported a significant inverse association between genetically predicted alcohol consumption and the risk of MASLD. Nonetheless, the findings require further validation due to moderate-to-high heterogeneity and potential pleiotropy in the genetic instruments. Finally, another reported no significant association with the risk of MASLD. These results contradict those of previous observational studies indicating that, compared to nondrinkers, people who drink moderate levels of alcohol (≤30 g/day) not only have a lower prevalence of MASLD, but also less severe disease from a histological standpoint (179).

Two MR studies investigated the impact of alcohol consumption on liver function biomarkers. While non-significant positive associations were observed with alanine aminotransferase (ALT) and gamma-glutamyl transferase (GGT) in one study, significant positive associations were found with alkaline phosphatase (ALP). These findings indicate potential liver function impairment linked to alcohol consumption. Another study also reported significant associations between alcohol consumption and the incidence of liver disease, despite non-significant results for ALT and GGT levels. These results are consistent with observational studies indicating that even moderate drinking can elevate γ-GT, ALT, and ferritin levels compared to abstainers (180).

In summary, while the available MR studies provide some insights into the link between alcohol consumption and liver disease, the results are limited and varied. It is crucial to note that no studies tested for non-linearity in these associations, highlighting a significant gap in the current understanding that needs to be addressed in future research.

The association of alcohol consumption with cognitive outcomes is controversial. Epidemiological studies have repeatedly reported better brain health in moderate drinkers compared with abstainers. Observational studies suggest that heavy alcohol consumption leads to deterioration of cognitive and executive function and is related to an increased risk of dementia (37) and other neurodegenerative diseases (181),while moderate alcohol intake may have a protective effect, leading to a J-shaped or U-shaped relationship (182–184). Overall, concerns about confounding and inconsistencies between epidemiological studies make it difficult to define what level of intake may have beneficial effects for cognition.

We reviewed MR studies with cognitive performance as an outcome and found that the methodological quality was generally low, with relatively small sample sizes. Although some studies observed an association between alcohol consumption and better cognitive performance, the five MR analyses did not confirm a causal relationship between genetically predicted alcohol consumption and cognitive performance. One study employed an additional approach, conducting a gene × environment interaction analysis (118). This study found that alcohol consumption interacted with the genotype score reflecting ADH activity to significantly predict cognitive ability around age 70. The authors interpreted this finding to mean that individuals with high alcohol processing efficiency experienced modest improvements in intelligence with increased alcohol consumption later in life. In contrast, individuals with lower alcohol processing efficiency experienced a decline in cognitive ability with greater alcohol consumption throughout their life. However, it should be noted that this study had a number of limitations, including a limited sample size, and the latter finding is not the result of an MR analysis, falling somewhat outside the scope of our review. In any case, these MR studies do not support the causal effect of alcohol consumption suggested by observational findings that moderate levels of alcohol consumption are associated with better cognitive performance (183, 184).

Four MR studies examined the association between genetically predicted alcohol consumption and AD. In all studies, the associations were generally non-significant, albeit mostly in the negative direction. However, a significant association was reported between average beer plus cider consumption and a higher risk of AD (119). It should be noted, however, that although these studies were of good methodological quality, they seemed to use an overlapping set of subjects from the IGAP. Here again, we find no support for a possible protective effect of alcohol consumption on the risk of developing AD.

Five studies examined the association between genetically predicted alcohol consumption and PD. Two studies reported a protective effect of genetically predicted alcohol use on the risk of developing PD. In one study, the protective effect was observed only with alcohol intake frequency (not alcohol amount) (119). Importantly, the other study demonstrated that the protective effect of alcohol intake was unlikely to be due to confounding by smoking behavior or the presence of survival bias (126). Two studies reported no association of genetically predicted alcohol use with the risk of developing PD, while one study reported a positive association of genetically predicted alcohol use with the risk of developing PD (124). It should be noted that this latter study had notable methodological shortcomings. Despite some discrepancies, this set of five MR studies provides some support for a possible protective effect of alcohol use on the risk of developing PD, as suggested by the findings of a meta-analysis of case-control studies (185).

An MR analysis of the association between genetically predicted alcohol consumption and ALS was performed in two studies. One of these studies reported a positive association, which remained significant after adjusting for smoking (127). This latter finding suggests a causal detrimental effect of alcohol consumption on the risk of developing ALS, contrasting with the findings of a meta-analysis of five observational studies that showed a potentially neuroprotective effect of alcohol (186). One study reported a positive association of genetically predicted higher alcohol intake frequency with a higher risk of MS, contrasting with the report of a dose-dependent inverse association of observed alcohol consumption with MS (187). In one MR study (128), no significant association was found between genetically predicted alcohol consumption and the age of onset of Huntington's Disease, a finding that does not help to clarify the conclusions of a few observational studies, which reported contrasting results (128).

The association between genetically predicted alcohol consumption and the risk of epilepsy was analyzed in two MR studies. Both studies reported a positive association, reaching statistical significance in only one (129). These results suggest a causal detrimental effect of alcohol on the risk of developing epilepsy, providing clarity in an area with previously inconsistent associations. While mild to moderate alcohol consumption may decrease the risk of seizures, alcohol consumption is generally thought to lower the seizure threshold, and alcohol withdrawal is a recognized cause of seizures (188).

In summary, the findings from MR studies on alcohol consumption and neurological outcomes provide mixed results. While some studies suggest potential protective effects of moderate alcohol consumption on certain neurodegenerative diseases, others highlight detrimental associations, particularly with higher intake frequencies. The methodological limitations and inconsistencies observed in these studies underscore the need for more robust and large-scale research to clarify these associations. Overall, these MR studies do not provide strong evidence to support the protective effects of alcohol consumption suggested by some observational studies, and they highlight the complex and multifaceted nature of alcohol's impact on brain health.

This systematic review has some limitations. First, the lack of an existing formal quality assessment tool to assess the quality of MR studies may limit the strength of our findings (189). To minimize this limitation, we computed our validation protocol for MR analyses based on the STROBE-MR (Strengthening The Reporting of Observational studies in Epidemiology using Mendelian Randomization) check list (190), which is inspired by the original STROBE checklist and developed to assist researchers in reporting their MR studies clearly and transparently.

Mendelian randomization studies also display some methodological limits and, as other studies, may lack power. Some genotypes used in MR studies may not be strong instruments due to limited variation in alcohol intake that could be explained by theses genotypes in a specific population. For instance, genotypes of ADH (fast alcohol metabolizers) and ALDH (lower acetaldehyde clearance rate) are associated with alcohol intake because higher acetaldehyde levels lead to unpleasant symptoms, but peak levels of acetaldehyde are also involved in the development of alcohol-related cancers (191, 192) and possibly CVD (193). Therefore, the level of alcohol drinking cannot be studied independently of acetaldehyde. Moreover, the available studies included in our systematic review are often heterogeneous, including variations in genetic IVs used, adjustments for confounding exposures, the study populations, and the outcome measurements. However, we included a large number of studies, with many two-sample MR design. These patterns of MR confer a higher statistical power and use a large number of independent IVs that come from the biggest GWAS of alcohol to date (from the GSCAN consortium (61) which increases the proportion of the variance of alcohol consumption explained by the IVs. The majority of the MR studies found consistent results, which provides an additional level of evidence for the role of alcohol consumption in the occurrence of these diseases.

Unfortunately, genetic variants are often unable to clearly distinguish between abstainers, light drinkers, and moderate drinkers, making it difficult to distinguish between U or J-shaped vs. linear relationships between exposure and outcome. As one of the controversies regarding the impact of alcohol consumption concerns the possible protective effect of light or moderate alcohol consumption on health outcomes, particularly cardiovascular outcomes, our systematic review does not allow us to draw clear conclusions regarding the causal effect of moderate alcohol consumption on these outcomes.

To address these limitations, further research is needed to develop formal quality assessment tools for MR studies and to conduct more MR studies on diverse populations, including those with non-European ancestries.