*2.1. Data Source and Participants*

We used electronic data of Taiwan Biobank (TWB) participants recruited between 2008 and 2015. Participants provided blood samples for DNA extraction and completed questionnaires covering a wide range of medical, social, and lifestyle information. All participants provided informed consent. Genotyping was done using the Axiom™ Genome-Wide TWB 2.0 Array plate (Santa Clara, CA, USA). Data on CHD between 1998 and 2015 were obtained from the National Health Insurance Research Database (NHIRD). The TWB database was linked to the NHIRD using encrypted personal identification numbers. This study was approved by the Institutional Review Board of Chung Shan Medical University (CS2-16114).

In total, 9001 biobank participants were recruited. After excluding persons with incomplete questionnaires (*n* = 13) and genotype information (*n* = 19), 1116 coronary heart disease patients and 7853 controls were included in the study.

#### *2.2. Assessment of Variables*

Coronary heart disease was identified based on either two outpatient visits or one admission with reported International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 410–414. Participants were classified as regular coffee drinkers if they drank coffee at least three

days per week in the last 6 months. Details of the covariates and physical measures used in the text have been described in our recent publication [28].

#### *2.3. Selection of the Polymorphic Variant*

The rs17321515 variant in the *TRIB1* gene was selected based on the literature search. This variant was selected because of its previous associations with CHD and dyslipidemia, especially in Han Chinese populations [16,17]. We also searched Google Scholar and selected rs762551 variant in the CYP1A2 gene which has been associated with caffeine metabolism and increased risk of myocardial infarction. We followed a standard quality control procedure and excluded SNPs with (1) a low call rate (<95%), (2) *p*-value of <1.0 <sup>×</sup> 10−<sup>3</sup> for the Hardy–Weinberg equilibrium test, and (3) minor allele frequency of <0.05. Moreover, we removed one individual from the pair of related samples based on pairwise identity-by-descent (IBD).

#### *2.4. Statistical Analysis*

We used the statistical analysis system (SAS) software (version 9.4, SAS Institute, Cary, NC, USA) and PLINK (v1.09, http://pngu.mgh.harvard.edu/purcell/plink/) to perform analyses. Differences between groups were compared using the chi-square test. Associations of coffee and the rs17321515 variant with CHD were determined using logistic regression analysis. Adjusted variables included sex, age, educational level, smoking, alcohol intake, tea consumption, vegetarian diet, body mass index (BMI), diabetes, hypertension, hyperlipidemia, atrial fibrillation, and *CYP1A2* rs762551 variant. Odds ratios with their 95% confidence intervals were estimated.

### **3. Results**

The descriptive data of 1116 participants with CHD and 7863 control individuals are shown in Table 1. Significant differences existed between patients and controls for coffee drinking, sex, age, educational level, cigarette smoking, exercise, body mass index (BMI), diabetes, hypertension, hyperlipidemia, atrial fibrillation, and vegetarian diet (*p* < 0.05). However, there were no significant differences between patients and controls for the *TRIB1* rs17321515 and *CYP1A2* rs762551 genotypes, alcohol, and tea consumption. Differences in coffee consumption habits between men and women as well as between those in different age groups are shown in Table 2.


**Table 1.** Descriptive data of the study participants.


**Table 1.** *Cont.*

CHD: Coronary heart disease, BMI: Body mass index, TRIB1: tribbles pseudokinase 1; CYP1A2: cytochrome P450 1A2. GG, GA, and AA represent genotypes in the TRIB1 rs17321515 variant while AA, AC, and CC represent genotypes in the CYP1A2 rs762551 variant.


**Table 2.** Characteristics of study participants based on coffee consumption.

CHD: Coronary heart disease, BMI: Body mass index, *TRIB1: tribbles pseudokinase 1, CYP1A2: cytochrome P450 1A2*.

Coffee drinking was associated with a lower risk of CHD (OR, 0.84; 95% CI, 0.72–0.99), as shown in Table 3. Association with the *TRIB1* rs17321515 variant was not significant; the OR was 1.01, 95% CI = 0.87–1.18. However, for the *CYP1A2* rs762551 variant, the OR was 0.86 with a 95% CI of 0.74–0.99 for AC+CC, compared to the AA genotype. Corresponding ORs (95% CI) for CHD

were 1.53 (1.07–2.19) for ages 40–49 years, 3.92 (2.82–5.46) for ages 50–59 years, 6.46 (4.59–9.09) for ages 60–70 years, 1.23 (1.04–1.46) for overweight, 1.35 (1.11–1.63) for obesity, 1.19 (1.01–1.41) for diabetes, 3.40 (2.91–3.98) for hypertension, 2.25 (1.91–2.63) for hyperlipidemia, and 4.09 (2.14–7.82) for atrial fibrillation.


**Table 3.** Association of CHD with associated variables.

Ref: reference, CHD: Coronary heart disease, BMI: Body mass index, OR: odds ratio, CI: confidence interval, *TRIB1: tribbles pseudokinase 1, CYP1A2: cytochrome P450 1A2*.

There was a significant interaction (*p* = 0.0330) between *TRIB1* rs17321515 and coffee drinking on CHD risk (Table 4). After stratification by rs17321515 genotypes, coffee drinking remained significantly associated with a lower risk of CHD only among those with the GG genotype (OR, 0.62; 95% CI, 0.45–0.85). There was no interaction between the *CYP1A2* rs762551 variant and coffee consumption.



Ref: reference, CHD: Coronary heart disease, BMI: Body mass index, OR: odds ratio, CI: confidence interval, *TRIB1: tribbles pseudokinase 1, CYP1A2: cytochrome P450 1A2*.

#### **4. Discussion**

In the current study, we determined whether an interactive association exists between coffee intake and the *TRIB1* rs17321515 variant with the risk of CHD. Our findings offered unique evidence that coffee intake might have a protective effect on CHD. We also found that contrary to previous findings [17,29], rs17321515 was not associated with CHD. Importantly, we found evidence of an interaction between rs17321515 and coffee intake. After stratification by rs17321515 genotypes, we found that CHD risk was significantly lower among those with GG genotype who consumed coffee relative to their non-coffee-drinking counterparts. However, there was no association among those with the GA+AA genotype, indicating that the genotype may not have any effect on CHD. *TRIB1* rs17321515 has been associated with a decreased risk of CAD among Europeans, Malays, and Asian Indians [15,30,31]. However, their analyses were not performed based on coffee intake.

So far, several studies have investigated the independent effects of coffee intake and *TRIB1* rs17321515 on cardiovascular disease risk. Of the studies, those investigating coffee consumption and cardiovascular disease risk have shown conflicting results. Contrary to findings from case–control studies which suggested that coffee intake was detrimental to coronary arteries [32], umbrella reviews of observational and intervention studies have found it to be beneficial even in little amounts [33,34]. An increased risk of CHD previously reported among heavy coffee drinkers was attributed to smoking [35]. In light of this, we included smoking in our analysis.

Regarding the rs17321515 polymorphism, its AA+GA genotypes were previously associated with an increased risk of CHD among Han Chinese [36]. In a Singapore Malay Eye study of 3280 adults aged 40–79 years old, the odds ratio for CHD among carriers of this variant was 1.23 for each copy of the A allele [31]. Even though the rs17321515 variant has been assessed in Asian populations as noted above, attempts have not been made to replicate it in Taiwan. This was the motivation behind the selection of this variant for the current study.

As stated earlier, lifestyle changes and genetic factors play a substantial role in the development of cardiovascular diseases. Of note, the interactive associations of both factors with CHD have not been widely reported. When coffee intake and the *TRIB1* rs17321515 variant were included in our model with adjustments for smoking and other lifestyle variables, we found that the GG genotype was significantly protective against CHD disease in individuals who consumed coffee compared to those who did not. The underlying mechanisms of interaction between coffee drinking and *TRIB1* rs17321515 SNP on CHD are not completely understood. However, metabolites in coffee are believed to influence protective endogenous pathways by modulation of gene expression [37].

One of the main variables included in our model was the rs762551 variant in the *CYP1A2* gene. We chose this variant based on its previous association with caffeine metabolism and its role in modifying the association between caffeinated coffee and the risk of heart disease [27]. Contrary to expectation, we found that AC+CC, compared to the AA genotype was protective against CHD in both the adjusted (OR, 0.86; 95% CI (0.74–0.99) and the separate model (Supplementary Table S1). By performing stratified analyses, we found that associations of *CYP1A2* rs762551 genotypes with CHD were not significant (Supplementary Table S2). Besides, there was no interaction between the variant and coffee consumption. Given that our findings are based on a limited number of coffee consumers, further investigations would be needed to clarify these associations.

In this study, we also observed that coffee consumption habits between cases and controls differed significantly based on gender and different age groups. However, differences in consumption based on gender and age are yet to be adequately determined, particularly in Taiwan.

We believe that these results will help to enhance the knowledge on the role of coffee in the association between rs17321515 variant and CHD among Taiwanese adults. However, the current study is just a first step to examine this association, which remains a fundamental issue for future research.

This study was limited in several ways. First, about 70% of the population studied did not consume any coffee. Such a limited number of coffee drinkers may preclude the possibility of observing meaningful associations between coffee and CHD. Next, our questionnaire did not have information on the type of coffee, caffeine content (that is, caffeinated or decaffeinated), methods of preparation, and the daily amount of consumption. We understand that these attributes may have different effects on CHD. Therefore, we recommend further research in this area. Second, well-established risk factors such as smoking, exercise, education, male sex, diabetes, tea-drinking, and vegetarian diet were not associated with the risk of CHD in the current population. This is an indication that our study population might not be representative of typical CHD study populations. Third, there is a possibility of nondifferential misclassification bias as information on coffee intake was based on self-report Lastly, even though the TWB is representative of the general population, only individuals who are 30–70 years old were recruited in the project. Therefore, we could not analyze data of adults under 30 or over 70 years of age.

#### **5. Conclusions**

In conclusion, our findings highlight the interactive association of coffee drinking and *TRIB1* rs17321515 polymorphism on coronary heart disease in Taiwanese adults. Taken together, we found that the risk of CHD was significantly lower among those with GG genotype who consumed coffee compared to their non-coffee-drinking counterparts. These results have provided considerable knowledge on gene–nutrient interaction in relation to cardiovascular disease.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/12/5/1301/s1. Table S1: Association of CHD with rs762551 variant and associated factors, Table S2: Association of CHD with coffee drinking stratified by rs762551 genotypes.

**Author Contributions:** Conceptualization, Y.-T.L., D.M.T., L.W., O.N.N., S.-Y.H., C.-C.H., C.-C.L., H.-R.C., and Y.-P.L.; formal analysis, S.-Y.H., C.-C.L., and Y.-P.L.; methodology, Y.-T.L., D.M.T., L.W., O.N.N., S.-Y.H., C.-C.H., C.-C.L., H.-R.C., and Y.-P.L.; supervision, H.-R.C., and Y.-P.L.; writing–original draft, Y.-T.L., D.M.T., L.W., and O.N.N., writing–review and editing Y.-T.L., D.M.T., L.W., O.N.N., S.-Y.H., C.-C.H., C.-C.L., H.-R.C., and Y.-P.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** The Ministry of Science and Technology (MOST), Taiwan partly funded this work (MOST 107-2627-M-040-002, 108-2621-M-040-00, and 107-EPA-F-017-002).

**Acknowledgments:** Authors would like to thank the Ministry of Science and Technology for the financial support.

**Conflicts of Interest:** The authors declare no conflict of interest.
