A Dose-Dependent Association between Alcohol Consumption and Incidence of Proteinuria and Low Glomerular Filtration Rate: A Systematic Review and Meta-Analysis of Cohort Studies

Previous cohort studies have reported conflicting associations between alcohol consumption and chronic kidney disease, characterized by proteinuria and low glomerular filtration rate (GFR). This systematic review, which included 14,634,940 participants from 11 cohort studies, assessed a dose-dependent association of alcohol consumption and incidence of proteinuria and low estimated GFR (eGFR) of <60 mL/min/1.73 m2. Compared with non-drinkers, the incidence of proteinuria was lower in drinkers with alcohol consumption of ≤12.0 g/day (relative risk 0.87 [95% confidence interval 0.83, 0.92]), but higher in drinkers with alcohol consumption of 36.1–60.0 g/day (1.09 [1.03, 1.15]), suggesting a J-shaped association between alcohol consumption and the incidence of proteinuria. Incidence of low eGFR was lower in drinkers with alcohol consumption of ≤12.0 and 12.1–36.0 than in non-drinkers (≤12.0, 12.1–36.0, and 36.1–60.0 g/day: 0.93 [0.90, 0.95], 0.82 [0.78, 0.86], and 0.89 [0.77, 1.03], respectively), suggesting that drinkers were at lower risk of low eGFR. In conclusion, compared with non-drinkers, mild drinkers were at lower risk of proteinuria and low eGFR, whereas heavy drinkers had a higher risk of proteinuria but a lower risk of low eGFR. The clinical impact of high alcohol consumption should be assessed in well-designed studies.

Alcohol consumption, a major global risk factor of attributable disability-adjusted life-years [2], is a potential modifiable lifestyle factor for CKD. Some recent systematic reviews have reported conflicting results of an association between alcohol consumption

Materials and Methods
The protocol for this systematic review and meta-analysis, registered in PROSPERO (CRD42023388228), adheres to the meta-analysis of observational study in epidemiology (MOOSE) reporting guidelines [22].

Literature Search and Selection Criteria
PubMed and Web of Science were searched between January 2000 and December 2022, to identify relevant cohort studies that investigated an association between alcohol consumption and the incidence of proteinuria or low eGFR of <60 mL/min/1.72 m 2 . The search strategy used is described in Appendix A in detail. Briefly, search terms included "proteinuria" or "glomerular filtration rate" with "alcohol," followed by terms to exclude non-cohort studies. In addition, we searched the reference lists of included publications and relevant reviews. The search was limited to publications available in the English language.
To assess a dose-dependent association between alcohol consumption, especially moderate to high consumption of >12 g/day, and incidence of the outcomes, we included studies in this review if a study (i) was a prospective or retrospective cohort study, (ii) measured the baseline alcohol consumption stratified by at least 3 categories, including non-/rare drinkers as a reference group, and current drinkers with at least 2 levels of alcohol consumption as exposure groups, (iii) had the highest alcohol consumption category with the lower boundary of >12 g/day, (iv) measured incidence of proteinuria or low eGFR of <60 mL/min/1.73 m 2 during the follow-up period, and (v) were published in the English language.
All records retrieved from the literature search were assessed by two reviewers (RY and QL) independently for inclusion, using a web app, Rayyan (Rayyan Systems, Inc. Cambridge, MA, USA. Available online: https://rayyan.ai) [23]. Full texts of potentially eligible studies were then reviewed to determine their final eligibility. Any disagreements between two reviewers were resolved through consensus.

Data Extraction and Quality Assessment
Two reviewers (RY and QL) independently extracted the following information from each study: lead author, study name, publication year, study location, numbers of participants and cases with outcomes, follow-up duration, age and eGFR of participants, male proportion, prevalence of diabetes and hypertension, alcohol consumption category (g/day or drinks/day), and multivariable-adjusted hazard ratios of each alcohol consumption cat-egory, and covariates used in statistical analyses. If more than one multivariable-adjusted model was reported in a study, the one with the largest number of adjusted variables was extracted.
We used the Newcastle-Ottawa quality assessment scale (NOS) to assess the methodological quality of 12 publications of 11 studies [24]. The NOS includes 8 items of 3 domains: (i) selection, including representativeness of the exposed cohort (0 or 1 score), selection of non-exposed (0 or 1), and ascertainment of exposed (0 or 1); (ii) comparability of cohorts on the basis of the design or analysis (0, 1, or 2); and (iii) outcome, including assessment of outcome (0 or 1), follow-up length long enough for outcome to occur (0 or 1), and adequacy of follow-up of cohorts (0 or 1). A study is considered of good quality if the total score is at least 7/9 [25]. Two reviewers (RY and QL) independently conducted quality assessment of the included studies. Disagreements were resolved by discussion between the two reviewers.

Statistical Analysis
If the mean or median values of baseline age, body mass index, and eGFR of all participants were not reported, we used the following equations to calculate the mean value of all participants. If the mean value of each alcohol consumption category was reported, where Mean all and Mean category are the mean value of all participants and participants with each alcohol consumption category, respectively, and N category is the number of participants in each alcohol consumption category. If the median value with the interquartile range of each alcohol consumption category was reported, we estimated the mean value of each alcohol consumption category using the following equation before calculating Mean all : Mean category = (Q1 category + Q3 category + Median categry ) ÷ 3 where Q1 category and Q3 category are its first and third quartiles, respectively, and Median category is its median value [26].
To assess a dose-dependent association between alcohol consumption (g/day) and the outcomes, we assigned the midpoint of the lower and upper boundaries of each alcohol consumption category as its representative value (Table 1). If the upper boundary was open-ended, 1.2 times its lower boundary was assigned as a representative value of the category [27]. A representative value of occasional drinkers was set at half a representative value of the adjacent category. If a unit of alcohol consumption was drinks/day (or drinks/week), it was converted to g/day (or g/week), according to a standard serving size of an alcoholic beverage in the study country, which is equivalent to 20 g of alcohol in Japan, 12 or 14 [28] g in the US and Canada, 8 g in the UK and Ireland, and 10 g in Australia [29]. Non-reference categories were divided into four groups of ≤12.0, 12.1-36.0, 36.1-60.0, and >60.0 g/day, based on the representative value of each alcohol consumption category. We pooled the relative risk (RR) estimates and their 95% confidence intervals of alcohol consumption for ≤12.0, 12.1-36.0, 36.1-60.0, and ≥60.0 g/day (vs. non-, rare, or never drinkers) with inverse weighting using the DerSimonian-Laird random-effects model to allow for between-study heterogeneity. If former drinkers were categorized separately from non-drinkers in a study, they were excluded from the present meta-analyses. Hazard ratios and odds ratios were considered as surrogate measures of RRs. Statistical heterogeneity among the studies was measured using the I 2 statistic. I 2 ≥ 50% suggested substantial heterogeneity. Publication bias was assessed by a visual inspection of funnel plot and by Egger's statistical tests [30]. We considered a p-value <0.05 to be evidence of small study effects.
To examine a potential nonlinear association between alcohol consumption and the outcomes, we performed a 2-stage random-effects dose-response meta-analysis [31] with the use of restricted cubic splines with 4 knots at fixed percentiles (5%, 35%, 65%, and 95%) of the distribution, including cohort studies which reported the number of participants with incidence of outcomes. The first stage of the meta-analysis estimated the dose-response association between alcohol consumption and the log RRs in each included study. The study specific estimates were then combined in the second stage of this meta-analysis.
All analyses were conducted using Stata version 17.0 (StataCorp LLC, College Station, TX, USA) and R version 4.2.1 (The R Foundation for Statistical Computing, https: //www.r-project.org). A p-value < 0.05 was considered statistically significant, unless otherwise specified.
The characteristics of the 12 publications from 11 studies are described in Tables 1 and 2. Of the 11 studies, 6 (7 publications) were conducted in Japan, 2 were from US, and 3 were from Italy, the Netherlands, and Korea ( Table 1). The highest alcohol consumption category in each study had a lower boundary of alcohol consumption of 14, 20, 30, 40, 46, 48, 60, and 69.1 g/day, to which we assigned 1.2 times the lower boundary values [27], namely, 16.8, 24.0, 36.0, 48.0, 55.2, 57.6, 72.0, and 82.9 g/day as its representative value. The incidence of proteinuria was defined as a dipstick urinary protein level of ≥1+ or urinary albumin excretion of >30 mg/day that was assessed in 7 studies, and the incidence of low eGFR was defined as <60 mL/min/1.73 m 2 or < 55 mL/min that was assessed in nine studies. All studies used serum creatinine-based equations to calculate eGFR. The study sample size ranged from 1539 [34] to 14,190,878 [19] ( Table 2). The prevalence of diabetes and hypertension was 0.0-13.9% and 0.0-64.9%, respectively and the mean (or median) follow-up period was ≤5, 6-10, and >10 years in 5, 2, and 3 studies, respectively. The study quality was good (NOS ≥ 7) for 5 (45.5%) studies (Table S1). In 6 publications from 5 studies, eGFR was not included as a covariate in the multivariable-adjusted model. ≥200 mg/dL, and/or use of anti-diabetic drugs; and hypertension defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, and/or use of anti-hypertensive drugs. ¶ Diabetes defined as fasting plasma glucose of ≥126 mg/dL, random plasma glucose ≥200 mg/dL, use of anti-diabetic drugs, and/or self-reported diabetes. ** Median value. † † A mean value of all men or women in a study was calculated after the mean value of each alcohol consumption group was estimated using the equation: mean = (Q1 + Q3 + median) ÷ 3, where Q1 and Q3 are the first and third quartiles, respectively [26].
An association between alcohol consumption and incidence of proteinuria stratified by alcohol consumption of 0.1-12.0, 12.1-36.0, 36.1-60.0, and >60.0 g/day is shown in Figure 1. The pooled result of 7 studies [19][20][21]35,37,38,40] that included 14,503,097 participants showed that drinkers were at a significantly lower risk of incidence of proteinuria than nondrinkers (overall RR 0.95 [95% confidence interval 0.93, 0.98]). However, the association was highly dependent on the alcohol consumption levels. Participants with low alcohol consumption of 0.1-12.0 g/day had a significantly lower risk of proteinuria (0.87 [0.83, 0.92]) than non-drinkers, while those with high alcohol consumption of 36.1-60.0 g/day had a significantly higher risk of proteinuria (1.09 [1.03, 1.15]). RR of alcohol consumption of >60.0 g/day (1.19 [0.93, 1.52]) was higher than that of 36.1-60.0 g/day, although a small number of studies with small sample sizes led to an underpowered analysis. A two-stage random-effects dose-response meta-analysis with use of a restricted cubic spline model, which included 14,410,068 participants from 5 studies [19][20][21]37,38], confirmed a J-shaped association between alcohol consumption and incidence of proteinuria (Figure 2a). Funnel plots suggested a potentially biased estimate of pooled RRs in drinkers with alcohol consumption of ≤12.0 g/day and 12.1-36.0 g/day (p = 0.041 and <0.001, respectively) ( Figure S2a,b and Table S2), but not in those consuming 36.1-60.0 g/day (p = 0.141) ( Figure S2c and Table S2). Because of substantial heterogeneity in 3 subgroups of alcohol consumption (I 2 = 94.60%, 89.99%, 57.06%, and 56.59% of ≤12.0, 12.1-36.0, 36.1-60.0, and >60.0 g/day, respectively) ( Figure 1 and Table S2), subgroup analyses stratified by sex; the median values of study size, body mass index, eGFR, prevalence of diabetes and hypertension, and follow-up duration; a NOS ≥7; and Asian and Western countries, were employed (Figures S3-S5 and Table S2). In the alcohol consumption subgroup of 36.1-60.0 g/day, age partly contributed to the high heterogeneity (I 2 = 39.16% and 0.00% in subgroup analyses of age < median and ≥ median, respectively) ( Figure S5c).   An association between alcohol consumption and the incidence of low eGFR was different from that between alcohol consumption and the incidence of proteinuria. A metaanalysis of nine studies, which included 14,375,672 participants [19,21,[32][33][34]36,37,39,40], drinkers with alcohol consumption of 0.1-12.0 and 12.1-36.0 g/day were at significantly lower risk for low eGFR than non-drinkers (0.93 [0.90, 0.95] and 0.82 [0.78, 0.86], respectively) (Figure 3). Those with alcohol consumption of 36.1-60.0 g/day were likely to have a lower risk of low eGFR (0.89 [0.77, 1.03]), although not at a statistically significant level. A two-stage random-effects dose-response meta-analysis with use of a restricted cubic spline model, which included 14,245,146 participants from five studies [19,21,36,37,39], showed that a negative linear association between alcohol consumption and incidence of low eGFR was blunted in the range of alcohol consumption of >24 g/day (Figure 2b). Funnel plots suggested publication bias in subgroups of alcohol consumption of 12.1-36.0 and 36.1-60.0 g/day (p = 0.048 and 0.020, respectively) ( Figure S2e,f and Table S3), although not in a subgroup of ≤12.0 g/day (p = 0.167) ( Figure S2d and Table S3). Substantial heterogeneity was observed in alcohol consumption of ≤12.0, 12.1-36.0, and 36.1-60.0 g/day (I 2 = 66.71%, 89.89%, and 84.35%, respectively) ( Figure 3 and Table S3). Subgroup analyses ( Figures S6-S8 and Table S3) suggested that sex (I 2 = 0.61% and 17.87% in subgroup analyses of men and women, respectively) ( Figure S6a) and body mass index (I 2 = 0.00% and 0.00% in subgroup analyses of body mass index < median and ≥ median, respectively) ( Figure S6d) possibly contributed to this heterogeneity in alcohol consumption of ≤12.0 g/day. Interestingly, four Western cohort studies showed a significantly stronger renoprotective effect of alcohol consumption of ≤12.0 g/day and 12.1-36.0 g/day than five Asian cohort studies (Table S3 and Figures S6j and S7j). However, only a single Western cohort study has assessed an association between alcohol consumption of 36.1-60.0 g/day and An association between alcohol consumption and the incidence of low eGFR was different from that between alcohol consumption and the incidence of proteinuria. A metaanalysis of nine studies, which included 14,375,672 participants [19,21,[32][33][34]36,37,39,40] A two-stage random-effects dose-response meta-analysis with use of a restricted cubic spline model, which included 14,245,146 participants from five studies [19,21,36,37,39], showed that a negative linear association between alcohol consumption and incidence of low eGFR was blunted in the range of alcohol consumption of >24 g/day (Figure 2b).  Table S3), although not in a subgroup of ≤12.0 g/day (p = 0.167) ( Figure S2d and Table S3). Substantial heterogeneity was observed in alcohol consumption of ≤12.0, 12.1-36.0, and 36.1-60.0 g/day (I 2 = 66.71%, 89.89%, and 84.35%, respectively) ( Figure 3 and Table S3). Subgroup analyses ( Figures S6-S8 and Table S3) suggested that sex (I 2 = 0.61% and 17.87% in subgroup analyses of men and women, respectively) ( Figure S6a) and body mass index (I 2 = 0.00% and 0.00% in subgroup analyses of body mass index < median and ≥ median, respectively) ( Figure S6d) possibly contributed to this heterogeneity in alcohol consumption of ≤12.0 g/day. Interestingly, four Western cohort studies showed a significantly stronger renoprotective effect of alcohol consumption of ≤12.0 g/day and 12.1-36.0 g/day than five Asian cohort studies (Table S3 and Figures S6j and S7j). However, only a single Western cohort study has assessed an association between alcohol consumption of 36.1-60.0 g/day and the incidence of low eGFR (Table S3).

Discussion
This systematic review, which included 12 publications from 11 cohort studies, showed that drinkers with low alcohol consumption were at lower risk of proteinuria and low eGFR. Inversely, high alcohol consumption was significantly associated with the incidence of proteinuria (Figure 2a), but not with the incidence of low eGFR (Figure 2b). Several advantages of the present systematic review were, first, assessment of high alcohol consumption of >36.0 g/day, second, separate analyses of a dose-dependent association of alcohol consumption between proteinuria and low eGFR, and third, a selection of publications with clinically relevant well-defined outcomes of proteinuria (dipstick urinary protein of ≥1+ or urinary albumin excretion of >30 mg/day) and low eGFR of <60 mL/min/1.73 m 2 (or <55 mL/min). This systematic review also disclosed a potential deleterious effect of high alcohol consumption on proteinuria, similar to the findings of the association between high alcohol consumption and cardiometabolic diseases, including hypertension [64], diabetes [65], stroke [66], and heart failure [66].
A previous American systematic review pooled RRs of the highest alcohol consumption category (vs. non-drinkers) in 6 cross-sectional, one case-control, and nine cohort studies, which included a total of 212,918 participants, and showed that the highest alcohol consumption was associated with lower prevalence/incidence of CKD with a wide variety of definitions, including ESKD, eGFR of <60 mL/min/1.73 m 2 , eGFR decline of >3 mL/min/1.73 m 2 , and proteinuria of ≥1+ [67]. This systematic review also reported no significant association between the highest alcohol consumption category and the incidence of proteinuria in four cohort studies with a total of 140,686 participants. However, it was difficult to draw any conclusion on the clinical impact of alcohol consumption on CKD because of the inclusion of many cross-sectional studies and a small number of cohort studies which assessed the association between alcohol consumption and incidence of proteinuria. Two Chinese systematic reviews, which included 268,723 participants from 15 cohort studies [17] and 514,148 participants from 25 prospective cohort studies [18], reported that high alcohol consumption of 26-60 g/day [17] and >24 g/day [18] was associated with a lower incidence of CKD, which was defined variously in each study including ESKD, eGFR <60 mL/min/1.73 m 2 , and eGFR decline of >3 mL/min/1.73 m 2 /year. The present study rigorously defined the outcomes of low eGFR and proteinuria and clarified different dose-dependent associations between alcohol consumption and these outcomes ( Figure 2). Inclusion of evidence from large cohort studies [19,20] enabled us to reveal a deleterious effect of high alcohol consumption on proteinuria.
Past drinkers were included as a reference category in nine of 11 cohort studies in this meta-analysis, possibly leading to a biased estimate of the association between alcohol drinking and incidence of proteinuria and low eGFR. Because former drinkers might be inspired to quit drinking due to health concerns, they might be at increased risk of proteinuria and low eGFR, known as the sick-quitter effect. An Australian cohort study clarified the sick-quitter effect among 97,852 drinkers aged ≥45 years [68]. During the median observational period of 5.3 years, 9438 (9.5%) drinkers quit drinking. Among a wide variety of 28 health conditions, including cancers, cardiovascular disease, endocrine conditions, genitourinary conditions, conditions affecting mobility, mental health conditions, and other conditions, 20 health conditions were significantly associated with quitting drinking. The most common health condition in those quitting drinking was heart disease (12.3%), a critical risk factor for incidence of CKD [69] and ESKD [70]. Given that the sick quitters at risk of incidence of CKD were categorized into a reference group, a beneficial effect of alcohol consumption might be overestimated, and its adverse effects might be underestimated. In the present meta-analysis, cohort studies with older age were more likely to show an antiproteinuric effect at ≤36.0 g/day of alcohol consumption and no detrimental effect at 36.1-60.0 g/day of alcohol consumption than those with younger age (Figures S3c, S4c and S5c and Table S2). Sick quitters might contribute to this significant age-dependent association between alcohol consumption and incidence of proteinuria. Evaluation of the association between alcohol consumption and incidences of proteinuria and low eGFR should be assessed more deliberately, considering the sick-quitter effect.
This study had several limitations. First, because seven of 11 (63.6%) cohort studies were reported from Asian countries, including six Japanese cohorts and one Korean cohort, and no Western study has assessed the clinical impact of alcohol consumption of >36.0 g/day on proteinuria and low eGFR (Tables S2 and S3 and Figures S5j and S8j), the results of the present meta-analysis were chiefly based on the genetic background of the Asian population. Alcohol metabolism is greatly dependent on genetic polymorphisms of major alcohol-metabolizing enzymes, alcohol dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH). Asian people are characterized by their unique allele frequencies of major gene polymorphisms of ADH and ALDH, including ADH1B*2 (rs1229984) and ALDH2*2 (rs679). Frequencies of A and C alleles of ADH1B*2 and ALDH2*2 are 74-77% and 11-28%, respectively, in Asian population, whereas their frequencies are almost 0% in other populations [71]. These alleles promote acetaldehyde production and suppress acetaldehyde metabolism, leading to the high acetaldehyde level [71]. These Asian genetic characteristics might contribute to a smaller renoprotective impact of alcohol consumption of ≤12.0 and 12.1-36.0 g/day on the incidence of low eGFR in Asian countries than in Western countries in this study (Table S3 and Figures S6j and S7j). The findings of this meta-analysis, especially the association between alcohol consumption of >36.0 g/day and the incidence of proteinuria and low eGFR, should be verified in non-Asian populations. Second, different definitions of the highest alcohol consumption category among 12 studies might lead to a biased estimate of clinical impacts of alcohol consumption on proteinuria and low eGFR. A large retrospective cohort study, which included 88,647 men and 88,925 women in Japan, carefully showed that the alcohol consumption category affected a dose-dependent association between alcohol consumption and incidence of dipstick proteinuria of ≥1+ [20]. If the alcohol consumption was categorized into four categories of rare drinkers, occasional drinkers, and daily drinkers with ≤19 and ≥20 g/day, alcohol consumption was associated with the incidence of proteinuria in a U-shape fashion in women ( Large cohort studies with deliberately categorized alcohol consumption are essential to assess a clinical impact of high alcohol consumption on CKD. Third, binge drinking defined generally as ≥5 and ≥4 standard drinks/occasion in men and women, respectively [72], was not assessed in this meta-analysis. Although binge drinking is a risk factor for cardiovascular diseases [73], little information has been available about an association between binge drinking and incidence of CKD. A prospective cohort study, which included 1883 Korean patients with CKD, reported that patients with occasional binge drinking were at a higher risk of incidence of a 50% increase in eGFR and/or ESKD. Clinical impact of binge drinking on CKD should be clarified in future studies. Fourth, GFR decline might be underestimated in participants with high alcohol consumption in this meta-analysis. Because high alcohol consumption is associated with low muscle mass [74,75] and serum creatinine level is heavily dependent on muscle mass [76], serum creatinine-based eGFR is likely to increase during the observational period among the participants with high alcohol consumption. Thus, a deleterious effect of high alcohol consumption on GFR might be blunted in the cohort studies included in this meta-analysis, in which GFR was estimated using serum creatinine-based equation. The association between high alcohol consumption and GFR trajectory should be assessed using an eGFR equation based on serum cystatin C level, which is independent of muscle mass [77].

Conclusions
This systematic review, which included 14,634,940 participants from 11 cohort studies, clarified that low alcohol consumption of ≤12 g/day was associated with lower incidence of proteinuria and low eGFR than non-drinkers. However, people with high alcohol consumption of ≥36 g/day were at a higher risk of proteinuria, whereas they were at a lower risk of low eGFR. Clinical impact of high alcohol consumption on the incidence of proteinuria and low eGFR have been assessed chiefly in the Asian population and scarcely in the non-Asian population with genetically different characteristics of alcohol metabolism. The association between high alcohol consumption and CKD should be assessed deliberately in well-designed cohort studies, including a wide variety of ethnic groups.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu15071592/s1, S1: Newcastle-Ottawa scale (NOS) of 12 publications from 11 cohort studies; Table S2: Alcohol consumption and incidence of proteinuria stratified by major study characteristics; Table S3: Alcohol consumption and incidence of eGFR of <60 mL/min/1.  Figure S3: Forest plots of relative risks to estimate an association between alcohol consumption of <12 g/day and incidence of proteinuria stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), and Newcastle-Ottawa scale (i); Figure S4: Forest plots of relative risks to estimate an association between alcohol consumption of 12.1-36.0 g/day and incidence of proteinuria stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), Newcastle-Ottawa scale (i), and Asian and Western countries (j); Figure S5: Forest plots of relative risks to estimate an association between alcohol consumption of 36.0-60.0 g/day and incidence of proteinuria stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), Newcastle-Ottawa scale (i), and Asian and Western countries (j); Figure S6: Forest plots of relative risks to estimate an association between alcohol consumption of <12 g/day and incidence of eGFR <60 mL/min/1.73 m 2 stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), Newcastle-Ottawa scale (i), and Asian and Western countries (j); Figure S7: Forest plots of relative risks to estimate an association between alcohol consumption of 12.1-36.0 g/day and incidence of eGFR <60 mL/min/1.73 m 2 stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), Newcastle-Ottawa scale (i), and Asian and Western countries (j); Figure S8: Forest plots of relative risks to estimate an association between alcohol consumption of 36.0-60.0 g/day and incidence of eGFR <60 mL/min/1.73 m 2 stratified by sex (a), study size (b), age (c), body mass index (d), eGFR (e), prevalence of diabetes (f), prevalence of hypertension (g), follow-up length (h), Newcastle-Ottawa scale (i), and Asian and Western countries (j).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the study design; collection, analyses, or interpretation of data; writing of the manuscript; or decision to publish the results.