Journal List > Nutr Res Pract > v.17(4) > 1516083806

TOOLS
Cho, Okekunle, Yie, Youn, Kang, Jin, Sung, and Lee: Association of coffee consumption with type 2 diabetes and glycemic traits: a Mendelian randomization study
Original Research

Nutrition Research and Practice 2023; 17(4): 789-802.

Published online: 10 March 2023

DOI: https://doi.org/10.4162/nrp.2023.17.4.789

Association of coffee consumption with type 2 diabetes and glycemic traits: a Mendelian randomization study

Hyun Jeong Cho1, 2, *, Akinkunmi Paul Okekunle1, 3, *, Ga-Eun Yie1, Jiyoung Youn1, Moonil Kang4, 5, Taiyue Jin6, Joohon Sung5, Jung Eun Lee1, 3

1Department of Food and Nutrition, College of Human Ecology, Seoul National University, Seoul 08826, Korea.

2K-BIO KIURI Center, Seoul National University, Seoul 08826, Korea.

3Research Institute of Human Ecology, Seoul National University, Seoul 08826, Korea.

4Department of Medicine (Biomedical Genetics), Boston University School of Medicine, Boston, MA 02118, USA.

5Complex Disease & Genome Epidemiology Branch, Department of Public Health, Graduate School of Public Health, Seoul National University, Seoul 08826, Korea.

6Division of Cancer Prevention, National Cancer Control Institute, National Cancer Center, Goyang 10408, Korea.

Corresponding Author: Jung Eun Lee. Department of Food and Nutrition, College of Human Ecology, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Korea. Tel. +82-2-880-6834, Fax. +82-2-884-0305, jungelee@snu.ac.kr

Equal contributor:

*Hyun Jeong Cho and Akinkunmi Paul Okekunle contributed equally and are joint first authors of this work.

Received 9 August 2022       Revised 23 December 2022       Accepted 20 January 2023

©2023 The Korean Nutrition Society and the Korean Society of Community Nutrition

The Korean Nutrition Society and the Korean Society of Community Nutrition

https://creativecommons.org/licenses/by-nc/4.0/
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

BACKGROUND/OBJECTIVES

Habitual coffee consumption was inversely associated with type 2 diabetes (T2D) and hyperglycemia in observational studies, but the causality of the association remains uncertain. This study tested a causal association of genetically predicted coffee consumption with T2D using the Mendelian randomization (MR) method.

SUBJECTS/METHODS

We used five single-nucleotide polymorphisms (SNPs) as instrumental variables (IVs) associated with habitual coffee consumption in a previous genome-wide association study among Koreans. We analyzed the associations between IVs and T2D, fasting blood glucose (FBG), 2h-postprandial glucose (2h-PG), and glycated haemoglobin (HbA1C) levels. The MR results were further evaluated by standard sensitivity tests for possible pleiotropism.

RESULTS

MR analysis revealed that increased genetically predicted coffee consumption was associated with a reduced prevalence of T2D; ORs per one-unit increment of log-transformed cup per day of coffee consumption ranged from 0.75 (0.62–0.90) for the weighted mode-based method to 0.79 (0.62–0.99) for Wald ratio estimator. We also used the inverse-variance-weighted method, weighted median-based method, MR-Egger method, and MR-PRESSO method. Similarly, genetically predicted coffee consumption was inversely associated with FBG and 2h-PG levels but not with HbA1c. Sensitivity measures gave similar results without evidence of pleiotropy.

CONCLUSIONS

A genetic predisposition to habitual coffee consumption was inversely associated with T2D prevalence and lower levels of FBG and 2h-PG profiles. Our study warrants further exploration.

Keywords: Coffee, diabetes mellitus, type 2, Mendelian randomization analysis, Koreans

INTRODUCTION

Type 2 diabetes (T2D) is a leading cause of mortality and morbidity worldwide [1234], with an estimated prevalence of 9.3% [1], 67.9 million disability-adjusted life-years [2] and increase in treatment cost [3] worldwide. Evidence for the physiological significance of habitual coffee consumption on T2D manifestations keeps evolving, with some cohort studies [5678] and randomized trials [910] suggesting an inverse association, but others [1112] found a null association. The limitations of these trials included small sample sizes and a short follow-up period. Also, observational studies may not be exonerated from biases that may confound the observed associations, thereby limiting the causal significance of habitual coffee consumption in T2D manifestation.
Mendelian randomization (MR) is a unique method that potentially overcomes these limitations and determines the causal significance of exposure to a disease event, taking advantage of the random assortment of alleles at conception [1314]. With the MR, the exposure-disease association essentially overpowers inherent limitations in traditional epidemiological reports as exposure phenotypes are genetically determined. Even though genetic variants account for a finite exposure variation, multiple independent genetic alleles from genome-wide association studies (GWAS) can be applied to determine causal associations in disease epidemiology [14].
The GWAS and MR studies have been suggested to help understand the significance of coffee consumption in the epidemiology of T2D [151617]. Such information may support nutrition recommendations and public health interventions to manage and prevent T2D. The extensive trans-Atlantic GWAS across multi-ethnic ancestry has identified some genetic alleles associated with T2D [18], but whether genetically predicted habitual coffee consumption (using MR methods) is associated with T2D, particularly among Asian populations, is yet to be clearly understood.
Therefore, we used genetic variants for habitual consumption as a genetic instrument variable to evaluate the association of coffee consumption with T2D prevalence and glycemic traits using the MR analysis.

SUBJECTS AND METHODS

Study population

Participants for this study were drawn from the Korea Association REsource (KARE) study [19], a sub-cohort of the Korean Genome and Epidemiology Study (KoGES) in Korea [20]. The KARE study was designed to distinguish genetic variants linked with several diseases among Asians. The study is a population-based cohort of > 10,000 Koreans that assessed > one million genetic variants [1921]. In this study, our primary outcome was T2D, and the secondary outcomes were glycemic traits; fasting blood glucose (FBG), 2h-postprandial glucose (2h-PG) and haemoglobin A1c (HbA1c) profiles. In the first step, we previously conducted the GWAS and identified genetic variants related to coffee consumption [22]. Second, we used an instrument variable from the GWAS-identified genetic variants to examine the associations of genetically predicted coffee consumption with T2D and glycemic traits. The flow chart of the study population is summarized in Fig. 1. The Institutional Review Board (IRB) of Seoul National University Korea (IRB No. E2104/001-010) approved this study.
Fig. 1

Flow diagram of the study population.

KARE, Korean Association REsource; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, haemoglobin A1c; MR, Mendelian randomization; PRESSO, Pleitropy RESidual and Outlier methods; SNP, single nucleotide polymorphism; CVD, cardiovascular disease.
nrp-17-789-g001

Assessment of coffee consumption and other variables

A validated semi-quantitative 103-item food frequency questionnaire was used to assess coffee consumption in this study [172123]. Participants reported the amount and frequency of coffee consumption in the last twelve months preceding the study. Responses to portion sizes of foods and drink items were a half of, equal to, and 2 times a standard serving size of coffee. Also, the frequency of consumption ranged from rarely to three or more times daily. Coffee consumption was transformed into cups/day. Details of how demographic and lifestyle covariates were obtained and characterized have been reported elsewhere [20]. In brief, information on age, sex, smoking status, alcohol status, and frequency and amount of alcohol drinking was collected using structured questionnaires. We calculated total ethanol consumption (g/day) through the product of frequency and amount. Body mass index (BMI) in kg/m2 was estimated from body weight (kg), and height (m), measured by trained personnel during the physical examination.

Definition of outcomes

HbA1c tests and serum levels of FBG and 2h-PG (from an oral glucose tolerance test) were assessed using the Hexokinase method (ADVIA 1650; Bayer, Inc., Tarrytown, NY, USA). Using the American Diabetes Association criteria [24], T2D was defined as one of the following conditions; FBG ≥ 126 mg/dL or HbA1c ≥ 6.5%, history of diabetes diagnosed by a physician or current use of blood glucose-lowering medications, or 2h-PG ≥ 200 mg/dL from the oral glucose tolerance test. Furthermore, glycemic traits in this study were presented in continuous models as secondary outcomes in the MR; FBG (mg/dL), 2h-PG (mg/dL), and HbA1c (%) profiles.

Genotyping

DNA extraction, genotyping, and imputation were conducted by the Korea National Research Institute of Health, Centers for Disease Control and Prevention and the Ministry of Health and Welfare, Korea. Affymetrix Genome-Wide Human single nucleotide polymorphism (SNP) Array 5.0 (Affymetrix, Santa Clara, CA, USA) was used for genotyping DNA samples extracted from the peripheral blood of participants. The Bayesian robust linear model (Mahalanobis distance genotyping algorithm) was used for genotype calling [25], and imputation for non-typed or missing genotypes was achieved using IMPUTE v246 with 1,000 genomes data [26]. Overall, 500,568 SNPs were identified in the genotype calling, but 352,228 SNPs remained after control filtering [27]. Minutiae of these methods and quality control have been reported elsewhere [25262728].

Statistical analysis

We previously identified five SNPs related to habitual coffee consumption in a GWAS using linear regression adjusting for age in years, sex and alcohol consumption in g/day in a continuous model. Log-transformed cups per day of coffee consumption was applied as a continuous outcome in the multivariable-adjusted model, in which SNPs were included as independent variables. When the GWAS was conducted for caffeine consumption in a continuous box-cox transformed linear model, SNPs identified for caffeine consumption were the same as those identified for coffee consumption. A statistical threshold of P < 1 × 10−5 was considered significant for the selected SNPs. Details of the Manhattan plots, quantile-quantile plots, the inflation factor, regional association plots of the SNPs that reached the statistical threshold, linkage equilibrium of the significant genetic variants and how the genetic risk score (GRS) was calculated in the GWAS have been detailed elsewhere [22].
The GRS was calculated by multiplying each weighted beta (β) coefficient of 5 SNPs associated with habitual coffee consumption by the number of corresponding minor alleles (0, 1, 2); the GRS = 5 × (β1 × SNP1 + β2 × SNP2 + β3 × SNP3 + β4 × SNP4 + β5 × SNP5)/(β1 + β2 + β3 + β4 + β5) [29]. The presence of minor allele was associated with increased habitual coffee consumption in our GWAS study. We tested the strength of the instrumental variable for the prediction of coffee consumption using F statistics, and the F values range was 24.2–33.8 for FBG, 2h-PG and HbA1c profiles.
To estimate the β-coefficient and SE of SNPs with log-transformed cups/day for coffee consumption, we used linear regression models adjusted for age (yr), sex, and alcohol consumption (g/day). Eighteen SNPs on chromosome 12q24 were identified in the multivariable-adjusted models (P < 1 × 10−5). In selecting SNPs for the GRS, pairwise correlations were applied to eliminate all imputed SNPs having a high correlation (r2 > 0.8) with genotyped SNPs. Also, SNPs were selected among strongly-correlated genotyped SNPs. Three genotyped SNPs (rs2074356, rs11066015, and rs12229654) and two imputed (rs11066015 and rs79105258) with minor allele frequencies range of 0.143 to 0.172 were identified in this population. For our primary outcome –T2D (no or yes), logistic regression (adjusting for age in years continuous, sex and alcohol consumption in g/day continuous) was used to calculate the β-coefficient and SE of the identified SNPs. For the glycemic traits, linear regression (adjusting for age in years continuous, sex and alcohol consumption in g/day continuous) was used to estimate the β-coefficients and SEs of SNPs with the raw concentrations of FBG (mg/dL), 2h-PG (mg/dL) and HbA1c (%).
MR inverse-variance-weighted (IVW) method was used to assess causal relations between genetically predicted coffee consumption and T2D or glycemic traits [30]. β-coefficients and SEs from the data where SNP and coffee consumption were treated as an independent variable and dependent variable, respectively, were applied to compute the Wald estimates and the SEs of the Wald estimates using the delta method [31], assuming a fixed effect model in the individual-level data [32]. Forest plots were used to visualize the associations of genetically predicted coffee consumption with T2D and glycemic traits at individual and pooled levels. In addition, MR-weighted median [30] and mode [33] methods were applied, and the degree of heterogeneity of the associations among 5 genetic variants was assessed to evaluate the robustness of the association. MR-Egger intercept and MR-Egger regression were applied to assess and adjust for directional and unbalanced pleiotropy of identified SNP in the MR analysis [34]. Because MR techniques are susceptible to pleiotropic effects, we used the MR-Pleitropy RESidual and Outlier methods (PRESSO) to assess and correct for horizontal pleiotropic outliers [35]. In the MR-PRESSO global test, we tested whether there was horizontal pleiotropy by alcohol consumption [35]. Scatter plots of the four MR methods (IVW, weighted median, weighted mode, and MR-Egger methods) were used to visualize and compare the methods. For glycemic traits, we also examined the associations between genetically predicted coffee consumption and glycemic traits using the two-stage least square technique. Using the GRS as an instrument variable, MR analyses were performed to examine the associations of genetically predicted coffee consumption with T2D and glycemic traits; Wald estimates for T2D and glycemic traits, and 2-stage least squares estimates for glycemic traits only. All statistical analysis was carried out using SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA) and R software (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Genetic variants related to coffee consumption

We previously identified 5 independent genomic loci associated with a unit increment of log-transformed coffee consumption (cups/day); rs11065828 in CUX2, rs11066015 in ACAD10, rs12229654 in MYL2, rs2074356 in HECTD4, and rs79105258 in CUX2. The strongest loci was rs2074356 within the HECTD4 gene (P = 6.62 × 10−8), and the weakest loci was rs11066015 in ACAD10 gene (P = 7.79 × 10−6) [22]. The GRS calculated from 5 SNPs in an additive model presented a significant association with T2D, FBG, and 2h-PG (Table 1). However, the GRS was not associated with HbA1c levels.
Table 1

Characteristics of SNPs associated with coffee consumption from GWAS in the KARE data (P < 1 × 10−8) and regression coefficients (SE) of SNPs with type 2 diabetes and glycemic traits

nrp-17-789-i001
Locus SNP Major allele Minor allele1) T2D2) FBG3) 2h-PG4) HbA1c5)
beta (SE) P-value beta (SE) P-value beta (SE) P-value beta (SE) P-value
CUX2 rs11065828 C A −0.07 (0.06) 0.24 −0.62 (0.18) 0.001 −1.73 (0.66) 0.009 −0.001 (0.007) 0.85
ACAD10 rs11066015 G A −0.12 (0.06) 0.04 −0.63 (0.18) 0.001 −2.45 (0.66) 0.0002 0.005 (0.007) 0.52
MYL2 rs12229654 T G −0.16 (0.06) 0.01 −0.60 (0.20) 0.002 −2.42 (0.71) 0.0006 0.01 (0.008) 0.11
HECTD4 rs2074356 G A −0.18 (0.06) 0.01 −0.73 (0.19) 0.0002 −2.70 (0.70) 0.0001 0.01 (0.008) 0.19
CUX2 rs79105258 C A −0.09 (0.06) 0.16 −0.57 (0.19) 0.003 −2.02 (0.68) 0.003 0.01 (0.008) 0.43
Genetic scores (per one-unit increment of genetic score) −0.03 (0.01) 0.03 −0.15 (0.04) 0.0003 −0.54 (0.15) 0.0003 0.002 (0.002) 0.35
All models were adjusted for age (yr, continuous), sex (men, women), and alcohol consumption (g/day, continuous).
SNP, single nucleotide polymorphism; GWAS, genome-wide association studies; KARE, Korea Association REsource; T2D, type 2 diabetes; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, haemoglobin A1c.
1)Genotypes coded as 0, 1 or 2, counting the number of the minor allele.
2)Association of SNPs (additive) or the genetic risk scores (per one-unit increment of genetic risk score) with T2D in the logistic regression model.
3)Association of SNPs (additive) or the genetic risk scores (per one-unit increment of genetic risk score) with fasting blood glucose – FBG (per an increment of 1 mg/dL) in the linear regression model.
4)Association of SNPs (additive) or the genetic risk scores (per one-unit increment of genetic risk score) with 2h-postprandial glucose (per an increment of 1 mg/dL) in the linear regression model.
5)Association of SNPs (additive) or the genetic risk scores (per one-unit increment of genetic risk score) with HbA1c (per an increment of 1%) in the linear regression model.
We examined the distribution of the GRS and each SNP according to the potential confounding factors and found no association with age (years in continuous; Spearman’s r = 0.02 and P = 0.17 for the GRS, Spearman’s r = 0.01 and P = 0.46 for rs79105258, Spearman’s r = 0.02 and P = 0.09 for rs2074356, Spearman’s r = 0.02 and P = 0.07 for rs12229654, Spearman’s r = 0.01 and P = 0.31 for rs11066015, and Spearman’s r = 0.01 and P = 0.23 for rs11065828), sex (t-test P = 0.49 for the GRS, P = 0.88 for rs79105258, P = 0.45 for rs2074356, P = 0.26 for rs12229654, P = 0.11 for rs11066015, and P = 0.72 for rs11065828), BMI (kg/m2 in continuous; Spearman’s r = −0.01 and P = 0.22 for the GRS, Spearman’s r = −0.02 and P = 0.07 for rs79105258, Spearman’s r = −0.02 and P = 0.08 for rs2074356, Spearman’s r = −0.02 and P = 0.04 for rs12229654, Spearman’s r = −0.01 and P = 0.22 for rs11066015, and Spearman’s r = −0.005 and P = 0.68 for rs11065828), or smoking status (never, past, current smokers; analysis of variance [ANOVA]; P = 0.39 for the GRS, P = 0.59 for rs79105258, P = 0.21 for rs2074356, P = 0.10 for rs12229654, P = 0.23 for rs11066015, and P = 0.41 for rs11065828). However, the GRS and SNP were associated with alcohol drinking (never, past, and current drinkers; ANOVA P < 0.001 for the GRS and each SNP). We examined the distribution of each SNP according to the potential confounding factors and found a similar distribution with the genetic score except for the distribution of rs12229654 according to BMI (Spearman’s r = −0.02 and P = 0.04).

Genetically predicted coffee consumption and T2D

Mendelian randomization suggests that genetically predicted coffee consumption was associated with a lower T2D prevalence: odds ratio (OR), 0.79; 95% confidence interval (CI), 0.71–0.88 using the MR-IVW method (Table 2, Fig. 2). An analogous risk estimate was observed using the MR-weighted-median (OR, 0.78; 95% CI, 0.68–0.89) and MR-weighted-mode (OR, 0.75; 95% CI, 0.62–0.90) methods. The intercept (β = 0.35, SE = 0.43) of MR-Egger regression analysis suggested no indication of directional horizontal pleiotropy (P for intercept = 0.32). Furthermore, the scatter plots of the MR methods in the observational analysis suggested no virtual evidence of heterogeneity across variants or potential pleiotropy regarding T2D (Fig. 3A).
Table 2

Genetically predicted coffee consumption with type 2 diabetes

nrp-17-789-i002
MR estimates OR (95% CI)1) P-value
Wald ratio estimator (GRS)2) 0.79 (0.62–0.99) 0.04
Inverse-variance-weighted method3) 0.79 (0.71–0.88) < 0.001
Weighted median-based method3) 0.78 (0.68–0.89) 0.03
Weighted mode-based method3) 0.75 (0.62–0.90) 0.03
MR-Egger method3)4) 0.39 (0.08–2.03) 0.35
MR, Mendelian randomization; OR, odds ratio; CI, confidence interval; GRS, genetic risk score; SNP, single nucleotide polymorphism.
1)For one-unit increment of log-transformed cup per day of coffee consumption.
2)Estimated genetic risk score was included as an instrument variable.
3)Estimates for each 5 SNP are combined to yield an overall estimate.
4)Beta (SE) was 0.35 (0.43) and P for intercept = 0.32 in the MR-Egger regression.
Fig. 2

The Mendelian randomization estimate of coffee consumption with (A) T2D, (B) FBS, (C) 2h-PG, and (D) HbA1c using a fixed-effects model in the inverse-variance-weighted method.

T2D, type 2 diabetes; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, hemoglobin A1c; SNP, single nucleotide polymorphism; OR, odds ratio; CI, confidence interval.
nrp-17-789-g002
Fig. 3

Scatter plot of the MR methods. The x-axis represents the genetic association with coffee consumption; the y-axis represents the genetic association with type 2 diabetes and glucose traits. Each line represents a different MR method. (A) T2D, (B) FBS, (C) 2h-PG, and (D) HbA1c.

MR, Mendelian randomization; T2D, type 2 diabetes; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, hemoglobin A1c; IVW, inverse-variance-weighted.
nrp-17-789-g003

Genetically predicted coffee consumption and glycemic traits

One-unit increment of log-transformed cups per day of genetically predicted coffee consumption was related to a decrease in FBG and 2h-PG levels in all methods except the MR-Egger method (Table 3, Fig. 2). For example, when the MR-IVW method was used, there was a 1.24 mg/dL decrease in FBG (β = −1.24, SE = 0.19, P < 0.001) and a 4.43 mg/dL diminution in 2h-PG (β = −4.43, SE = 1.40, P < 0.001). The intercept of MR-Egger regression analysis for FBG and 2h-PG suggested no indication of directional horizontal pleiotropy (P for intercept > 0.80). Genetically predicted coffee consumption was unrelated to HbA1c (β = 0.01, SE = 0.007, P = 0.07), and comparable estimates were observed for MR-weighted median and mode methods. Likewise, scatter plots of genetically predicted log-transformed coffee consumption (cups/day) with glycemic traits revealed similar findings; FBG (Fig. 3B), 2h-PG (Fig. 3C), and HbA1c (Fig. 3D). We observed similar results after examining the association using the MR-PRESSO global method (Table 4). MR-PRESSO global method revealed no evidence of directional horizontal pleiotropy (P for global test in the MR-PRESSO global method > 0.8).
Table 3

Genetically predicted coffee consumption with FBG, 2h-PG, and HbA1c

nrp-17-789-i003
MR estimates FBG (mg/dL) 2h-PG (mg/dL) HbA1c (%)
beta (SE) P-value beta (SE) P-value beta (SE) P-value
2-Stage least squares estimator1) −1.44 (0.20) < 0.001 −4.17 (0.62) < 0.001 0.02 (0.006) 0.006
2-Stage least squares estimator (GRS)1) −1.44 (0.40) < 0.001 −4.19 (1.27) < 0.001 0.002 (0.01) 0.16
Wald ratio estimator (GRS)1) −1.24 (0.39) 0.001 −4.44 (1.40) 0.001 0.01 (0.01) 0.36
Inverse-variance-weighted method2) −1.24 (0.19) < 0.001 −4.43 (0.69) < 0.001 0.01 (0.007) 0.07
Weighted median-based method2) −1.28 (0.22) 0.005 −4.68 (0.81) 0.005 0.01 (0.008) 0.19
Weighted mode-based method2) −1.30 (0.25) 0.006 −4.79 (0.94) 0.007 0.01 (0.01) 0.25
MR-Egger method2)3) −1.06 (2.55) 0.70 −6.95 (9.18) 0.50 0.11 (0.10) 0.37
Coffee consumption was per 1 cup/day increment of log-transformed.
MR, Mendelian randomization; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, haemoglobin A1c; GRS, genetic risk score; SNP, single nucleotide polymorphism.
1)Estimated genetic risk score was included as an instrument variable.
2)Estimates for each 5 SNP are combined to yield an overall estimate.
3)Beta (SE) and P-values for the intercept in the MR-Egger regression were as follows; −0.09 (1.29) and P = 0.95 for fasting blood glucose; 1.27 (4.65) and P = 0.80 for 2h-PG; and −0.05 (0.05) and P = 0.42 for HbA1c.
Table 4

Genetically predicted coffee consumption with type 2 diabetes and glycemic traits: MR-PRESSO methods

nrp-17-789-i004
MR-PRESSO global method
OR (95% CI) P-value beta (SE) P-value P for global test
Coffee and T2D 0.79 (0.73–0.84) 0.002 0.84
Coffee and FBS −1.24 (0.04) < 0.001 > 0.99
Coffee and 2h-PG −4.46 (0.27) < 0.001 0.95
Coffee and HbA1c 0.01 (0.004) 0.05 0.79
MR-PRESSO, Mendelian Randomization-Pleitropy RESidual and Outlier methods; OR, odds ratio; CI, confidence interval; T2D, type 2 diabetes; FBG, fasting blood glucose; 2h-PG, 2h-postprandial glucose; HbA1c, hemoglobin A1c.

DISCUSSION

We applied multiple MR methods with correction for pleiotropic effects and found that higher coffee consumption was causally associated with a lower T2D prevalence in this study. Similarly, higher coffee consumption was inversely associated with FBG and 2h-PG profiles in a dose-response manner. This study applied MR techniques to present evidence of the significance of coffee consumption in glucose metabolism among Asian populations.
Several prospective studies [223637], reviews [538394041], and randomized trials [1112] have studied the coffee consumption-T2D relationship without genetic evidence to support causal relationships. Earlier, MR studies on this subject have been reported in the Copenhagen General Population and City Heart Studies [42], Social Science Genetic Association Consortium [43], and UK biobank [164445] among European descents, but none has been reported among Asian descents. These studies presented genetic evidence to support null association [42434445], except for Wang et al. [16]. In that study, genetically predicted ground coffee consumption only (but not total or decaffeinated) was associated with a decreased risk of T2D. Similarly, a review of MR studies of genetically predicted coffee consumption in cardiometabolic outcomes concluded that higher coffee consumption was unrelated to T2D [46]. However, most of these studies were exclusively conducted among European descent, and genetic variants of coffee consumption in this population did not appear to differ geographically [47]. We hypothesize that these differences are likely related to the degree of exposure to coffee consumption. At least none of these studies has reported a genetically-causal association of deleterious risk for T2D. The absence of evidence does not necessarily imply the lack of evidence. In tandem with our observation, genetically proxied coffee consumption was associated with a low risk of gallstone diseases [48] and small vessel ischemic stroke [15] in European populations.
The pathophysiology of the inverse coffee consumption-T2D association can be explained in the following ways. First, coffee is a repository of phenolic compounds rich in antioxidant activity [4950], protective against inflammation [5152] and neurodegenerative diseases, including T2D, in some observational studies [5354], intervention trials [55], and reviews [505657]. Second, suppressed fat accumulation due to the inhibitive potential of quinic acids from phenols on the nuclear activity of sterol regulatory element-binding protein 1c, acetyl-CoA carboxylase activity, and cellular malonyl-CoA levels were reported in C57BL/6J mice on a high-fat diet supplemented with coffee-derived phenols [58].
Our findings should be contextualized in light of some empirical understanding. First, coffee consumption transformed into cups/day was used as an exposure without considering data on the types of coffee consumed and the additives used. Even though we did not distinguish the types of coffee, it is unlikely to bias our findings because several prospective studies have reported the inverse association between coffee consumption and T2D independent of coffee types [3841596061]. However, genetically predicted evidence in this regard is yet to be documented. Second, this study was exclusively conducted among Koreans. The vitality of multi-ancestral GWAS (including diverse Asian populations) cannot be underestimated in the reliability of MR analysis for robust associations between genetic variants and phenotypes.
Our study had some limitations and strengths. First, selecting unstable SNPs as instrumental variables cannot be ruled out. In order to overcome this, we used SNPs associated with coffee consumption in our study, and the multiple MR analysis method revealed consistent results. Second, MR methods are susceptible to pleiotropic effects, which can be either horizontal (pathways) [13] or vertical (involving multiple factors). Our study surmounted this limitation by conducting MR analysis using multiple complementary approaches. We used pleiotropy-detecting MR methods such as the MR-PRESSO to assess and correct for horizontal pleiotropy, and there was no evidence of direct horizontal pleiotropy in this study. Third, the applicability of our findings across diverse populations is quite limited as this study was exclusively among Koreans. Coffee consumption data was presented at a single time point, and the precise composition of different coffee types was unaccounted for in this study. Also, alcohol consumption correlated with genetic variants in this population and the exclusivity of coffee consumption in the reported association should be interpreted cautiously. However, correction for horizontal pleiotropy showed consistent results. We observed a statistically insignificant association between genetically predicted coffee consumption and HbA1c profiles. A rational explanation for this phenomenon is currently elusive, but further longitudinal studies and well-articulated intervention trials are necessary to clarify these findings.
In conclusion, our study revealed that genetically predicted coffee consumption was associated with a lower prevalence of T2D among Koreans. Future studies should be designed to discern the mechanistic pathophysiology of this association.

ACKNOWLEDGMENTS

This study was conducted with bioresources from the National Biobank of Korea, the Center for Disease Control and Prevention, Republic of Korea (KBN-2018-044).

Notes

Funding: APO was supported by the Brain Pool Program through the National Research Foundation of Korea, funded by the Ministry of Science and ICT (2020H1D3A1A04081265). HJC was supported by Korea Initiative for fostering University of Research and Innovation Program of the National Research Foundation (NRF) funded by the Korean government (MSIT) (No. NRF-2020M3H1A1073304). The funders had no role in the study design, data collection and analysis, decision to publish, or manuscript preparation.

Conflict of Interest: The authors declare no potential conflicts of interests.

Author Contributions:

  • Conceptualization: Lee JE.

  • Formal analysis: Cho HJ, Yie GE, Youn J, Lee JE.

  • Funding acquisition: Cho HJ, Okekunle AP, Lee JE.

  • Methodology: Cho HJ, Okekunle AP, Yie GE, Youn J, Kang M, Jin T, Sung J, Lee JE.

  • Supervision: Lee JE.

  • Writing - original draft: Okekunle AP, Lee JE.

  • Writing - review & editing: Cho HJ, Okekunle AP, Yie GE, Youn J, Kang M, Jin T, Sung J, Lee JE.

References

1. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N, Colagiuri S, Guariguata L, Motala AA, Ogurtsova K, et al. Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. 2019; 157:107843. PMID: 31518657.
crossref
2. Lin X, Xu Y, Pan X, Xu J, Ding Y, Sun X, Song X, Ren Y, Shan PF. Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025. Sci Rep. 2020; 10:14790. PMID: 32901098.
crossref
3. Bommer C, Sagalova V, Heesemann E, Manne-Goehler J, Atun R, Bärnighausen T, Davies J, Vollmer S. Global economic burden of diabetes in adults: projections from 2015 to 2030. Diabetes Care. 2018; 41:963–970. PMID: 29475843.
crossref
4. Cousin E, Duncan BB, Stein C, Ong KL, Vos T, Abbafati C, Abbasi-Kangevari M, Abdelmasseh M, Abdoli A, Abd-Rabu R, et al. Diabetes mortality and trends before 25 years of age: an analysis of the Global Burden of Disease Study 2019. Lancet Diabetes Endocrinol. 2022; 10:177–192. PMID: 35143780.
5. Jiang X, Zhang D, Jiang W. Coffee and caffeine intake and incidence of type 2 diabetes mellitus: a meta-analysis of prospective studies. Eur J Nutr. 2014; 53:25–38. PMID: 24150256.
crossref
6. Bhupathiraju SN, Pan A, Manson JE, Willett WC, van Dam RM, Hu FB. Changes in coffee intake and subsequent risk of type 2 diabetes: three large cohorts of US men and women. Diabetologia. 2014; 57:1346–1354. PMID: 24771089.
crossref
7. Kim Y, Je Y, Giovannucci E. Coffee consumption and all-cause and cause-specific mortality: a meta-analysis by potential modifiers. Eur J Epidemiol. 2019; 34:731–752. PMID: 31055709.
crossref
8. Cho HJ, Yoo JY, Kim AN, Moon S, Choi J, Kim I, Ko KP, Lee JE, Park SK. Association of coffee drinking with all-cause and cause-specific mortality in over 190,000 individuals: data from two prospective studies. Int J Food Sci Nutr. 2022; 73:513–521. PMID: 34779701.
crossref
9. Alperet DJ, Rebello SA, Khoo EY, Tay Z, Seah SS, Tai BC, Emady-Azar S, Chou CJ, Darimont C, van Dam RM. A randomized placebo-controlled trial of the effect of coffee consumption on insulin sensitivity: Design and baseline characteristics of the Coffee for METabolic Health (COMETH) study. Contemp Clin Trials Commun. 2016; 4:105–117. PMID: 29736473.
crossref
10. Ohnaka K, Ikeda M, Maki T, Okada T, Shimazoe T, Adachi M, Nomura M, Takayanagi R, Kono S. Effects of 16-week consumption of caffeinated and decaffeinated instant coffee on glucose metabolism in a randomized controlled trial. J Nutr Metab. 2012; 2012:207426. PMID: 23193459.
crossref
11. Alperet DJ, Rebello SA, Khoo EY, Tay Z, Seah SS, Tai BC, Tai ES, Emady-Azar S, Chou CJ, Darimont C, et al. The effect of coffee consumption on insulin sensitivity and other biological risk factors for type 2 diabetes: a randomized placebo-controlled trial. Am J Clin Nutr. 2019; 111:448–458.
crossref
12. Kempf K, Herder C, Erlund I, Kolb H, Martin S, Carstensen M, Koenig W, Sundvall J, Bidel S, Kuha S, et al. Effects of coffee consumption on subclinical inflammation and other risk factors for type 2 diabetes: a clinical trial. Am J Clin Nutr. 2010; 91:950–957. PMID: 20181814.
crossref
13. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008; 27:1133–1163. PMID: 17886233.
crossref
14. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013; 37:658–665. PMID: 24114802.
crossref
15. Qian Y, Ye D, Huang H, Wu DJ, Zhuang Y, Jiang X, Mao Y. Coffee consumption and risk of stroke: a Mendelian randomization study. Ann Neurol. 2020; 87:525–532. PMID: 32034791.
crossref
16. Wang X, Jia J, Huang T. Coffee types and type 2 diabetes mellitus: large-scale cross-phenotype association study and Mendelian randomization analysis. Front Endocrinol (Lausanne). 2022; 13:818831. PMID: 35222278.
crossref
17. Kim AN, Cho HJ, Youn J, Jin T, Kang M, Sung J, Lee JE. Coffee consumption, genetic polymorphisms, and the risk of type 2 diabetes mellitus: a pooled analysis of four prospective cohort studies. Int J Environ Res Public Health. 2020; 17:5379. PMID: 32722593.
crossref
18. DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) Consortium. Asian Genetic Epidemiology Network Type 2 Diabetes (AGEN-T2D) Consortium. South Asian Type 2 Diabetes (SAT2D) Consortium. Mexican American Type 2 Diabetes (MAT2D) Consortium. Type 2 Diabetes Genetic Exploration by Nex-generation sequencing in muylti-Ethnic Samples (T2D-GENES) Consortium. Mahajan A, Go MJ, Zhang W, Below JE, Gaulton KJ, et al. Genome-wide trans-ancestry meta-analysis provides insight into the genetic architecture of type 2 diabetes susceptibility. Nat Genet. 2014; 46:234–244. PMID: 24509480.
19. Hong KW, Kim HL, Oh B. Genome-wide association studies of the Korea Association REsource (KARE) Consortium. Genomics Inform. 2010; 8:101–102.
crossref
20. Kim Y, Han BG. KoGES group. Cohort profile: the Korean Genome and Epidemiology Study (KoGES) Consortium. Int J Epidemiol. 2017; 46:e20. PMID: 27085081.
crossref
21. Hong CB, Kim YJ, Moon S, Shin YA, Cho YS, Lee JY. KAREBrowser: SNP database of Korea Association REsource Project. BMB Rep. 2012; 45:47–50. PMID: 22281013.
crossref
22. Jin T, Youn J, Kim AN, Kang M, Kim K, Sung J, Lee JE. Interactions of habitual coffee consumption by genetic polymorphisms with the risk of prediabetes and type 2 diabetes combined. Nutrients. 2020; 12:2228. PMID: 32722627.
crossref
23. Younjhin A, Lee JE, Paik HY, Lee HK, Inho J. Development of a semi-quantitative food frequency questionnaire based on dietary data from the Korea National Health and Nutrition Examination Survey. Nutr Sci. 2003; 6:173–184.
24. American Diabetes Association Professional Practice Committee. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes-2022. Diabetes Care. 2021; 45:S17–S38.
25. Cho YS, Go MJ, Kim YJ, Heo JY, Oh JH, Ban HJ, Yoon D, Lee MH, Kim DJ, Park M, et al. A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing eight quantitative traits. Nat Genet. 2009; 41:527–534. PMID: 19396169.
crossref
26. Kim YJ, Go MJ, Hu C, Hong CB, Kim YK, Lee JY, Hwang JY, Oh JH, Kim DJ, Kim NH, et al. Large-scale genome-wide association studies in east Asians identify new genetic loci influencing metabolic traits. Nat Genet. 2011; 43:990–995. PMID: 21909109.
crossref
27. Rabbee N, Speed TP. A genotype calling algorithm for affymetrix SNP arrays. Bioinformatics. 2006; 22:7–12. PMID: 16267090.
crossref
28. Kim YK, Moon S, Hwang MY, Kim DJ, Oh JH, Kim YJ, Han BG, Lee JY, Kim BJ. Gene-based copy number variation study reveals a microdeletion at 12q24 that influences height in the Korean population. Genomics. 2013; 101:134–138. PMID: 23147675.
crossref
29. Wang T, Huang T, Heianza Y, Sun D, Zheng Y, Ma W, Jensen MK, Kang JH, Wiggs JL, Pasquale LR, et al. Genetic susceptibility, change in physical activity, and long-term weight gain. Diabetes. 2017; 66:2704–2712. PMID: 28701334.
crossref
30. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol. 2016; 40:304–314. PMID: 27061298.
crossref
31. Palmer TM, Sterne JA, Harbord RM, Lawlor DA, Sheehan NA, Meng S, Granell R, Smith GD, Didelez V. Instrumental variable estimation of causal risk ratios and causal odds ratios in Mendelian randomization analyses. Am J Epidemiol. 2011; 173:1392–1403. PMID: 21555716.
crossref
32. Burgess S, Dudbridge F, Thompson SG. Combining information on multiple instrumental variables in Mendelian randomization: comparison of allele score and summarized data methods. Stat Med. 2016; 35:1880–1906. PMID: 26661904.
crossref
33. Burgess S, Foley CN, Allara E, Staley JR, Howson JMM. A robust and efficient method for Mendelian randomization with hundreds of genetic variants. Nat Commun. 2020; 11:376. PMID: 31953392.
crossref
34. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015; 44:512–525. PMID: 26050253.
crossref
35. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018; 50:693–698. PMID: 29686387.
crossref
36. Lee JK, Kim K, Ahn Y, Yang M, Lee JE. Habitual coffee intake, genetic polymorphisms, and type 2 diabetes. Eur J Endocrinol. 2015; 172:595–601. PMID: 25755232.
crossref
37. Shin S, Lee JE, Loftfield E, Shu XO, Abe SK, Rahman MS, Saito E, Islam MR, Tsugane S, Sawada N, et al. Coffee and tea consumption and mortality from all causes, cardiovascular disease and cancer: a pooled analysis of prospective studies from the Asia Cohort Consortium. Int J Epidemiol. 2022; 51:626–640. PMID: 34468722.
crossref
38. Ding M, Bhupathiraju SN, Chen M, van Dam RM, Hu FB. Caffeinated and decaffeinated coffee consumption and risk of type 2 diabetes: a systematic review and a dose-response meta-analysis. Diabetes Care. 2014; 37:569–586. PMID: 24459154.
crossref
39. O’Keefe JH, Bhatti SK, Patil HR, DiNicolantonio JJ, Lucan SC, Lavie CJ. Effects of habitual coffee consumption on cardiometabolic disease, cardiovascular health, and all-cause mortality. J Am Coll Cardiol. 2013; 62:1043–1051. PMID: 23871889.
crossref
40. Huxley R, Lee CM, Barzi F, Timmermeister L, Czernichow S, Perkovic V, Grobbee DE, Batty D, Woodward M. Coffee, decaffeinated coffee, and tea consumption in relation to incident type 2 diabetes mellitus: a systematic review with meta-analysis. Arch Intern Med. 2009; 169:2053–2063. PMID: 20008687.
crossref
41. Carlström M, Larsson SC. Coffee consumption and reduced risk of developing type 2 diabetes: a systematic review with meta-analysis. Nutr Rev. 2018; 76:395–417. PMID: 29590460.
crossref
42. Nordestgaard AT, Thomsen M, Nordestgaard BG. Coffee intake and risk of obesity, metabolic syndrome and type 2 diabetes: a Mendelian randomization study. Int J Epidemiol. 2015; 44:551–565. PMID: 26002927.
crossref
43. Kwok MK, Leung GM, Schooling CM. Habitual coffee consumption and risk of type 2 diabetes, ischemic heart disease, depression and Alzheimer’s disease: a Mendelian randomization study. Sci Rep. 2016; 6:36500. PMID: 27845333.
crossref
44. Nicolopoulos K, Mulugeta A, Zhou A, Hyppönen E. Association between habitual coffee consumption and multiple disease outcomes: a Mendelian randomisation phenome-wide association study in the UK Biobank. Clin Nutr. 2020; 39:3467–3476. PMID: 32284183.
crossref
45. Zhou A, Hyppönen E. Long-term coffee consumption, caffeine metabolism genetics, and risk of cardiovascular disease: a prospective analysis of up to 347,077 individuals and 8368 cases. Am J Clin Nutr. 2019; 109:509–516. PMID: 30838377.
crossref
46. Nordestgaard AT. Causal relationship from coffee consumption to diseases and mortality: a review of observational and Mendelian randomization studies including cardiometabolic diseases, cancer, gallstones and other diseases. Eur J Nutr. 2022; 61:573–587. PMID: 34319429.
crossref
47. Coffee and Caffeine Genetics Consortium. Cornelis MC, Byrne EM, Esko T, Nalls MA, Ganna A, Paynter N, Monda KL, Amin N, Fischer K, et al. Genome-wide meta-analysis identifies six novel loci associated with habitual coffee consumption. Mol Psychiatry. 2015; 20:647–656. PMID: 25288136.
48. Yuan S, Gill D, Giovannucci EL, Larsson SC. Obesity, type 2 diabetes, lifestyle factors, and risk of gallstone disease: a Mendelian randomization investigation. Clin Gastroenterol Hepatol. 2022; 20:e529–e537. PMID: 33418132.
crossref
49. Kusumah J, Gonzalez de Mejia E. Coffee constituents with antiadipogenic and antidiabetic potentials: a narrative review. Food Chem Toxicol. 2022; 161:112821. PMID: 35032569.
crossref
50. Cao H, Ou J, Chen L, Zhang Y, Szkudelski T, Delmas D, Daglia M, Xiao J. Dietary polyphenols and type 2 diabetes: Human Study and Clinical Trial. Crit Rev Food Sci Nutr. 2019; 59:3371–3379. PMID: 29993262.
crossref
51. Boon EA, Croft KD, Shinde S, Hodgson JM, Ward NC. The acute effect of coffee on endothelial function and glucose metabolism following a glucose load in healthy human volunteers. Food Funct. 2017; 8:3366–3373. PMID: 28858362.
crossref
52. Hang D, Zeleznik OA, He X, Guasch-Ferre M, Jiang X, Li J, Liang L, Eliassen AH, Clish CB, Chan AT, et al. Metabolomic signatures of long-term coffee consumption and risk of type 2 diabetes in women. Diabetes Care. 2020; 43:2588–2596. PMID: 32788283.
crossref
53. Kosmalski M, Pękala-Wojciechowska A, Sut A, Pietras T, Luzak B. Dietary intake of polyphenols or polyunsaturated fatty acids and its relationship with metabolic and inflammatory state in patients with type 2 diabetes mellitus. Nutrients. 2022; 14:1083. PMID: 35268058.
crossref
54. Micek A, Godos J, Cernigliaro A, Cincione RI, Buscemi S, Libra M, Galvano F, Grosso G. Polyphenol-rich and alcoholic beverages and metabolic status in adults living in Sicily, Southern Italy. Foods. 2021; 10:383. PMID: 33572478.
crossref
55. Grabež M, Škrbić R, Stojiljković MP, Vučić V, Rudić Grujić V, Jakovljević V, Djuric DM, Suručić R, Šavikin K, Bigović D, et al. A prospective, randomized, double-blind, placebo-controlled trial of polyphenols on the outcomes of inflammatory factors and oxidative stress in patients with type 2 diabetes mellitus. Rev Cardiovasc Med. 2022; 23:57. PMID: 35229548.
crossref
56. Rudrapal M, Khairnar SJ, Khan J, Dukhyil AB, Ansari MA, Alomary MN, Alshabrmi FM, Palai S, Deb PK, Devi R. Dietary polyphenols and their role in oxidative stress-induced human diseases: insights into protective effects, antioxidant potentials and mechanism(s) of action. Front Pharmacol. 2022; 13:806470. PMID: 35237163.
crossref
57. Fernandes I, Oliveira J, Pinho A, Carvalho E. The role of nutraceutical containing polyphenols in diabetes prevention. Metabolites. 2022; 12:184. PMID: 35208257.
crossref
58. Murase T, Misawa K, Minegishi Y, Aoki M, Ominami H, Suzuki Y, Shibuya Y, Hase T. Coffee polyphenols suppress diet-induced body fat accumulation by downregulating SREBP-1c and related molecules in C57BL/6J mice. Am J Physiol Endocrinol Metab. 2011; 300:E122–E133. PMID: 20943752.
crossref
59. Park SY, Freedman ND, Haiman CA, Le Marchand L, Wilkens LR, Setiawan VW. Association of coffee consumption with total and cause-specific mortality among nonwhite populations. Ann Intern Med. 2017; 167:228–235. PMID: 28693036.
crossref
60. Freedman ND, Park Y, Abnet CC, Hollenbeck AR, Sinha R. Association of coffee drinking with total and cause-specific mortality. N Engl J Med. 2012; 366:1891–1904. PMID: 22591295.
crossref
61. Loftfield E, Freedman ND, Graubard BI, Guertin KA, Black A, Huang WY, Shebl FM, Mayne ST, Sinha R. Association of coffee consumption with overall and cause-specific mortality in a large US prospective cohort study. Am J Epidemiol. 2015; 182:1010–1022. PMID: 26614599.
crossref
TOOLS
Similar articles