Introduction

Type 2 diabetes mellitus (T2DM) is one of the leading global causes of morbidity and mortality. According to the International Diabetes Federation, 537 million adults have been clinically diagnosed with T2DM; this number is expected to increase to 643 million by 2030 [1].

Increased alcohol intake has been shown to be positively correlated with an increased incidence of T2DM [2]. Comparatively, moderate alcohol consumption has been shown to be associated with a reduced incidence of T2DM [3]. However, the aforementioned results are not consistent among other studies [4,5,6]. These discrepancies may reflect differences in analytical methodology used to examine the association, ethnic, and/or lifestyle differences among the study populations. In addition, some of the previous studies included participants who had preexisting diabetes whereas other studies included biased populations. The NAGALA (NAfld in the Gifu Area, Longitudinal Analysis) study was conducted at Murakami Memorial Hospital’s Medical Health Checkup Center (Gifu, Japan) from 2004 to 2015; key advantages of the NAGALA study are the large sample size and rigorous follow-up over an extended period of time.

The current study describes a secondary analysis of the NAGALA study to examine the relationship between alcohol consumption and incidence of T2DM.

Methods

Data source

Data was obtained from the publicly forum “DRYAD” (https://datadryad.org/) and from the following article: Ectopic fat obesity presents the greatest risk for incident T2DM: a population-based longitudinal study [7] (dataset: https://doi.org/10.5061/dryad.8q0p192). The database file contains the following variables: age, sex, body weight, body mass index (BMI), waist circumference, systolic blood pressure (SBP), diastolic blood pressure (DBP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), fasting plasma glucose (FPG), hemoglobin A1C (HbA1C), alcohol consumption, exercise, smoking status, fatty liver, year of follow-up and T2DM events during the follow-up. The authors of the original study forfeited all copyrights and related ownership of the data used in this secondary analysis. The original study was approved by the Murakami Memorial Hospital Ethics Committee. Written informed consent was obtained from all participants. The authors state that this study was conducted in accordance with the Declaration of Helsinki.

Study design and participants

The NAGALA study was conducted at Murakami Memorial Hospital’s Medical Health Checkup Center (Gifu, Japan). A total of 20,944 participants were recruited during a period from 2004 to 2015. The final analysis excluded participants with missing covariate data, known liver disease, excessively heavy alcohol intake (ethanol consumption over 60 g/day for men and 40 g/day for women), medication use, and diagnosis of T2DM or FPG ≥ 6.1 mmol/L. Ultimately, 15,464 participants were included in the original study [7].

Data collection and measurements

The following information was obtained at the initial screening: weight, waist circumference, blood pressure and a standard blood test panel that included alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), fasting plasma glucose (FPG), and hemoglobinA1c (HbA1c) levels in fasting blood samples. Medical history and lifestyle variables, such as smoking and drinking habits, were collected using a standardized questionnaire. Alcohol consumption estimation was based on the mean weekly alcohol intake, as assessed by asking the participants about the type and amount of alcohol consumed each week during the prior month, and classified into four categories according to the original study: heavy = > 280 g per week; moderate = 140–280 g per week; light = 40–140 g per week; and none or minimal = < 40 g per week [7, 8]. Smoking status was classified into three categories: never smoking, past smoking, and currently smoking [7]. Regular physical activity was defined as engaging in any sporting activity (exercise) more than once per week [9]. Abdominal ultrasound was performed by a trained technician, and images were reviewed by a gastroenterologist blinded to other personal data of the participants. The diagnosis of fatty liver disease was based on ultrasound images and classified into four criteria: hepatorenal echo contrast, liver brightness, deep attenuation, and vascular blurring [7]. An estimated 60% of the participants were examined once or twice per year [7]. T2DM was defined as FPG ≥ 7.0 mmol/L, HbA1c ≥ 6.5%, or self-reported. The study endpoint was the occurrence of T2DM during the follow-up.

Statistical analysis

Continuous variables were compared among groups with different alcohol consumption categories using one-way ANOVA, Kruskal-Wallis H test, and shown as mean ± standard deviation (SD) or median and quaternary ranges (25th-75th percentile) based on data distribution (normal vs. skewed). Categorical variables were compared using chi-square test, and shown as number and percentages. Cumulative incidence of T2DM was plotted using Kaplan–Meier curves and analyzed using a log-rank test. Multivariate Cox regression was used to evaluate the potential associations between alcohol consumption and T2DM risk. Results are shown as the hazard ratio (HR) and 95% confidence interval (95%CI). Adjustments were made when HR changed by 10% upon covariate addition. Smooth curve fitting was conducted for visual display of the relationship between alcohol consumption and new-onset T2DM.

All analyses were performed with the statistical software package R (http://www.R-project.org, The R Foundation) and Empower-Stats (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA). All T tests were two-sided, and P values < 0.05 were considered statistically significant.

Results

Baseline characteristics of the Study Population

The final analysis included a total of 15,464 participants (8430 men and 7034 women). The average age was 43.71 ± 8.90 years (Table 1).

Table 1 Baseline characteristics of participants classified according to alcohol consumption
Full size table

The percentages of the study cohorts with respect to different alcohol consumption categories were: 76.34% (11,805) for none or minimal, 11.37% (1758) for light, 8.80% (1360) for moderate and 3.50% (541) for heavy alcohol consumption. As the categorization of alcohol consumption was constrained by the original study, the heavy drink consumption in women could not be analysed, there were no women in the heavy consumption group. Overall, participants in the heavy consumption group were of older age comparatively, had higher BMI and waist circumference, higher blood pressure, and higher levels of ALT, AST, GGT, TG and FPG, but lower HbA1C. The rate of fatty liver disease of those currently smoking and the current smoker was higher in the heavy consumption group.

Kaplan–Meier analysis revealed higher cumulative risk of T2DM in heavy consumption group compared to all other groups (log rank test, P = 0.0002; Fig. 1).

Fig. 1
figure 1

Cumulative hazard curves for incidental T2DM by alcohol consumption

Full size image

Cox regression analysis

Univariate regression analysis revealed incidental T2DM was positively associated with age, BMI, waist circumference, ALT, AST, GGT, TG, FPG, HbA1C, SBP, DBP, current smokers, fatty liver disease and heavy alcohol consumption, and negatively with HDL-C (Table 2).

Table 2 Univariate analysis between variables and incident diabetes
Full size table

Multivariate regression analysis revealed incidental T2DM was independently associated with older age, higher BMI, SBP, ALT, GGT, HDL-C, TC, TG, FPG, HbA1C, fatty liver disease and smoking status, as well as alcohol consumption (Table 3). The adjusted HR relative to the none or minimal consumption group was 1.02 (0.71, 1.48) for light consumption, 1.06 (0.71, 1.57) for moderate consumption, and 2.06 (1.30, 3.24) for heavy consumption (P value = 0.024). The association between heavy consumption with incidental T2DM was significantly different in the subgroup analysis that only included men (HR: 2.06, 95%CI: 1.30–3.28; P value = 0.002), but not in women.

Table 3 Relationship between alcohol consumption and incident diabetes in different models
Full size table

Nonlinear relationship between Alcohol Consumption and T2DM

The nonlinear relationship between alcohol consumption and T2DM risk, after adjusting for confounding factors (e.g., age, BMI, SBP, ALT, GGT, HDL-C, TC, TG, FPG, HbA1C, fatty liver disease and smoking status) in men, is shown in Fig. 2. There were no saturation or threshold effects (P = 0.385).

Fig. 2
figure 2

The association between alcohol consumption and T2DM events in men, after adjustment for age, BMI, SBP, ALT, GGT, HDL-C, TC, TG, FPG, HbA1C, fatty liver disease and smoking status

Full size image

Discussion

In the present study, T2DM incidence was independently associated with heavy alcohol consumption. The adjusted HR relative to the none or minimal consumption group was 1.02 (0.76, 1.48) for light consumption, 1.06 (0.71, 1.57) for moderate consumption, and 2.06 (1.30, 3.24) for heavy consumption (P value = 0.024). Subgroup analysis confirmed the association between alcohol consumption and incidental T2DM in men but not in women. The subgroup analysis in men suggested a nonlinear relationship between alcohol intake and incidental T2DM; however, this relationship was not statistically significant.

Heavy alcohol consumption has been consistently associated with incident T2DM [10, 11]. For instance, Watanabe et al. [12] found that alcohol intake increased the probability of developing T2DM in Japanese men. In a study by Cullmann et al. [2], men with heavy alcohol consumption are more likely to have impaired glucose regulation. Moderate alcohol consumption has been associated with reduced risk of T2DM [13,14,15]. However, in some Asian groups, this protective effect on T2DM incidence was not found [16, 17]. In the current study, we did not find a reduced risk of incidental T2DM with moderate alcohol consumption. Similar discrepancies are reflected among ethnic backgrounds throughout associated literature.

There are several biological mechanisms that might explain the reported association between alcohol consumption and T2DM. Heavy alcohol drinking causes diabetes by increasing caloric intake and obesity, which disrupts carbohydrate and glucose metabolism. Heavy alcohol consumption has also been linked to chronic pancreatitis and liver injury, with decreased insulin secretion and insulin resistance [18, 19]. Previous research has associated disruption of β-cells to oxidative stress, mitochondrial dysfunction, and increased generation of reactive oxygen species in pancreatic islets [2, 15]. In vitro, ethanol metabolites 2,3-butanediol and 1,2-propanediol inhibit both basal and insulin-stimulated adipocyte metabolism [20]. Ethanol exposure in rats increases β-cell atrophy and decreases absolute pancreatic islet volume [21]. β-cells from ethanol-treated rats had extensive ultrastructural abnormalities, including deep invagination of the nuclear envelope, heterochromatin margination, and numerous empty granules or granules lacking distinct electrodense insulin crystals. Wang et al. [22] showed that excessive ethanol consumption disrupted pancreatic γ-aminobutyric acid (GABA) signaling in pancreatic islet β-cells and compromised glucose metabolism. The mechanisms underlying impaired pancreatic islet glucose metabolism and insulin secretion capacity are complex, and may involve resistance to fibroblast growth factor 21 (FGF21) [23]. A recent study linked increased T2DM by alcohol consumption to a HECTD4 genotype [24]. Specifically, HECTD4 mutation predisposes people to T2DM upon alcohol consumption.

Previous studies have also suggested a J- or U-shaped relationship between alcohol consumption and incidental diabetes [25, 26]. Subgroup analysis in the current study confirmed an independent association between alcohol consumption and incidental diabetes in men but not women. The subgroup analysis in men suggested a nonlinear relationship, albeit statistically non-significant, between alcohol consumption and incident T2DM. Several previous studies in the Western population showed that moderate alcohol consumption is associated with an increase in HDL-C, which in turn reduces the risk of incident T2DM [27, 28]. The current study failed to show reduced risk of incident T2DM with moderate alcohol consumption. The observed discrepancy may be partly attributed to different ethnic backgrounds; metabolic differences between Japanese and Western populations have been documented previously. For example, Western populations have higher levels of insulin resistance whereas Japanese populations are less likely to be obese yet have greater β-cell dysfunction [29]. Accordingly, the effect of heavy alcohol intake on incidental T2DM in Japanese men may be explained by its influence on β-cell function rather in insulin sensitivity.

A major strength of the current study is the large sample size and rigorous follow-up for an extended period of time in the original study. As a population-based retrospective cohort study, the data on alcohol consumption and incident diabetes, as well as a broad range of confounding factors, are relatively reliable.

The current study has several limitations. First, as a secondary analysis, we did not have access to all confounding factors. Second, the diagnosis of incidental T2DM was not based on an oral glucose tolerance test and thus may be underestimated. Third, alcohol consumption status was based on the information collected at baseline did not consider changes during the follow-up period. Fourth, according to the categorization of alcohol in women, the number of women with moderate alcohol consumptions was very low; heavy drinking was not reported by any women in this cohort.

In conclusion, the current study revealed an independent association between heavy alcohol consumption and increased risk of incidental T2DM. Subgroup analysis confirmed the association in men, but not in women.