Abstract
Background: Coffee consumption has been associated with glucose metabolism and risk of type 2 diabetes.
Objective: We examined whether the genetic variation determining habitual coffee consumption affected glycemic changes in response to weight-loss dietary intervention.
Design: A genetic risk score (GRS) was calculated based on 8 habitual coffee consumption–associated single nucleotide polymorphisms. We used general linear models to test changes in glycemic traits in groups randomly assigned to high- and low-fat diets according to tertiles of the GRS.
Results: We observed significant interactions between the GRS and low compared with high dietary fat intake on 6-mo changes in fasting insulin and homeostasis model assessment of insulin resistance (HOMA-IR) (P-interaction = 0.023 and 0.022, respectively), adjusting for age, sex, race, physical activity, smoking, alcohol, seasonal variation, and baseline values of the respective outcomes. Participants with a higher GRS of habitual coffee consumption showed a greater reduction in fasting insulin and a marginally greater decrease in HOMA-IR in the low-fat diet intervention group.
Conclusions: Our data suggest that participants with genetically determined high coffee consumption may benefit more by eating a low-fat diet in improving fasting insulin and HOMA-IR in a short term. This trial was registered at clinicaltrials.gov as NCT00072995 and NCT03258203.
Keywords: genetic risk score, habitual coffee consumption, weight-loss diets, gene-diet interaction, glycemic traits
INTRODUCTION
Coffee is one of the most popular and widely consumed beverages in the world. In epidemiology studies, high coffee consumption has been consistently related to improved insulin sensitivity and glucose metabolism as well as a decreased risk of type 2 diabetes in a dose-response manner (1–3). Compelling evidence has shown that consumption of coffee is affected by genetic variation. Several genetic variants have been identified as being related to habitual coffee consumption (4–6). Furthermore, data from experimental studies suggest that dietary macronutrients, such as fat intake, may modulate the effects of coffee consumption on insulin resistance and glucose metabolism (7, 8). Whether macronutrient intakes modify the effects of the coffee consumption–associated genetic variants remains unclear.
A recent genome-wide transethnic meta-analysis identified a variety of loci implicated in habitual coffee consumption (9), including 5 loci located in or near genes potentially involved in the pharmacokinetics of caffeine [ATP binding cassette subfamily G member 2 (ABCG2), aryl hydrocarbon receptor (AHR), cytochrome p450 oxidoreductase (POR), cytochrome P450 family 1 subfamily A member 2 (CYP1A2)] and pharmacodynamics [brain derived neurotrophic factor (BDNF)]. Loci near BDNF potentially influence consumption behavior through regulating the acute behavioral and strengthening properties of caffeine, whereas others near AHR, CYP1A2, POR, and ABCG2 exert their effects indirectly through regulating the metabolism of caffeine and the physiologic levels of this stimulant (9). Glucokinase regulator (GCKR) is involved in glucose metabolism (10), but GCKR variation affects central pathways in response to coffee constituents through regulating the glucose-sensing process of the brain (11).
In this study, we used the 8 habitual coffee consumption–associated single nucleotide polymorphisms (SNPs) identified by the genome-wide meta-analysis (9) to calculate a genetic risk score (GRS) and analyzed whether the GRS was related to changes in glycemic traits in the POUNDS LOST (Preventing Overweight Using Novel Dietary Strategies) trial (registered at clinicaltrials.gov as NCT00072995 and NCT03258203). We particularly tested interactions between the GRS and diet interventions varying in macronutrient intake on glycemic changes. To the best of our knowledge, no study has assessed whether macronutrient intakes modified the relations of the coffee consumption–associated genetic variants with these metabolic outcomes. We hypothesized that the genetic variants of coffee consumption might affect changes of metabolism of insulin and glucose in the context of weight-loss interventions.
METHODS
Study population
The POUNDS LOST trial was a 2-y randomized multicenter clinical trial for weight loss conducted in 2004–2007 at the Harvard School of Public Health and Brigham and Women’s Hospital (Boston, MA) and the Pennington Biomedical Research Center of the Louisiana State University System (Baton Rouge, LA). The study design and sample collection have been described previously in detail (12). The trial participants were 811 overweight or obese individuals aged 30–70 y and with a BMI (in kg/m2) of 25–40. The major exclusion criteria were the presence of diabetes treated with oral medications or insulin, unstable cardiovascular disease, cancer, the use of medications that influence body weight, and inadequate motivation as estimated by an interview and questionnaire.
Participants were randomly distributed to 1 of 4 diets that composed a 2-by-2 factorial design: 1) a low-fat, average-protein diet (20% fat, 15% protein, and 65% carbohydrate), 2) a low-fat, high-protein diet (20% fat, 25% protein, and 55% carbohydrate), 3) a high-fat, average-protein diet (40% fat, 15% protein, and 45% carbohydrate), and 4) a high-fat, high-protein diet (40% fat, 25% protein, and 35% carbohydrate). After 2 y, 645 participants (80% of the total population) completed the trial. The participants’ flowchart is shown as Supplemental Figure 1. Foods with a low glycemic index were used in all diets. The study was approved by the human subjects committee at each institution and by a data and safety monitoring board appointed by the National Heart, Lung, and Blood Institute. All participants provided written informed consent.
The current study included 680 participants with genotype data available, and 594 actually completed the 2-y trial. Of the 680 participants who were genotyped successfully, 60.4% were women, 84.1% were white, and 15.9% were black by self-report. No important differences in the baseline characteristics between the participants with and without genotyping data were evidenced.
Measurements
After an overnight fast, body weight, height, and waist circumference were measured in the morning on 2 nonconsecutive days at baseline, 6 mo, and 2 y. The Baecke physical activity questionnaire, a valid and reliable 16-item self-report inventory used to determine an individual’s level of habitual physical activity (13, 14), was administered at baseline and at 12 and 24 mo. Self-reported physical activity was also tracked through the computer tracking system during each week of the study. Smoking was categorized as never smoked, former smoker, and current smoker. The completion date of demographic data was recorded and categorized into 4 seasons to indicate seasonal variation.
Fasting blood samples, 24-h urine samples, and measurement of the resting metabolic rate were obtained on 1 d at baseline, 6 mo, and 2 y. Analyses of glucose and urinary nitrogen were performed at the Clinical Laboratory at Pennington Biomedical Research Center of the Louisiana State University System. BMI was calculated as weight by height squared (kg/m2). Glucose and insulin were measured with the use of an immunoassay with chemiluminescent detection on the Immulite analyzer (Diagnostic Products Corp.). To be consistent with prior published studies in the POUNDS LOST trial, homeostasis model assessment models were used to estimate HOMA-IR (15, 16), which was calculated by the following approximation equation: HOMA-IR = [fasting insulin (μU/mL) × fasting glucose (mmol/L)]/22.5.
Dietary intake and adherence
Because the estimation of dietary intake and adherence required substantial manpower and financial resources, dietary intake was only estimated in a random sample of 50% of participants through a review of the 5-d diet record at baseline and a 24-h recall during a telephone interview on 3 nonconsecutive days at 6 mo and at 2 y. Biomarkers of nutrient intake were used to validate self-reported adherence to macronutrient targets as follows: urinary nitrogen excretion for protein and respiratory quotient for fat (12). The amount of alcohol consumed (grams per day) was obtained from diet recall. Owing to diminished adherence, <50% of the total participants were included in each group at 2 y.
Genotyping and GRS calculation
DNA was extracted from the buffy coat fraction of centrifuged blood with the use of the QIAamp Blood Kit (Qiagen). We decided candidate genes based on the criteria of satisfying genome-wide significance (log10 Bayes factor >5.64) from the newest and largest genome-wide transethnic (European ancestry and African-American ancestry) meta-analysis (9). Ten SNPs associated with habitual coffee consumption were selected and genotyped (9), including GCKR rs1260326, ABCG2 rs1481012, AHR rs4410790 and rs6968554, Max-like protein X interacting protein-like (MLXIPL) rs7800944, POR rs17685, BDNF rs6265, CYP1A member 1 (CYP1A1) rs2470893, CYP1A2 rs2472297, and EF-hand calcium binding domain 5 (EFCAB5) rs9902453 (9). The rs4410790 of AHR and rs2470893 of CYP1A1 were not included in the GRS owing to high linkage disequilibrium with rs6968554 and rs2472297, respectively. Therefore, 8 SNPs were included in the GRS score. These SNPs were genotyped successfully in 680 of 811 total participants with the use of the OpenArray SNP Genotyping System (BioTrove) (Supplemental Table 1).
To represent the overall genetic variation of habitual coffee consumption, we created the GRS by following the widely accepted methods to sum the genetic variations weighted by their effect sizes (β-coefficients) (17, 18). The GRS was calculated based on the 8 SNPs by summing risk allele’s numbers and weighted by their effect sizes (β-coefficients) derived from the genome-wide association meta-analysis (Supplemental Table 1). The following equation was applied: GRS = [β1 × SNP1 + β2 × SNP2 + … + β8 × SNP8] × [n/sum of the β-coefficients], where β is the β-coefficient of each SNP for higher habitual coffee consumption, and SNP1, SNP2, …, and SNP8 indicate the number of risk alleles (0, 1, or 2) for each SNP. We actually transformed the βs to indicate a higher coffee consumption with increasing allele, and the sum of the β-coefficients is 0.53 in the current analysis, meaning total allele effect sizes of 0.53 cups/d. The GRS ranged from 1.86 to 15.2 in the study participants, and a higher GRS means a higher likelihood to consume more coffee.
Statistical analysis
The primary outcomes were changes in glycemic traits, including fasting glucose, insulin, and HOMA-IR. Insulin and HOMA-IR were log-transformed to ensure normality. The Hardy-Weinberg equilibrium and comparison of categorical variables at baseline were assessed with the χ2 test. General linear models were used to test differences in continuous variables according to tertiles of the GRS at baseline, with adjustment for age, sex, and race. We used general linear models to test changes in glycemic traits in participants assigned to high- or low-fat diets according to tertiles of the GRS. Covariates adjustment included age, sex, race, physical activity, smoking, alcohol, seasonal variation, and the baseline value for the respective outcome in the model. Moreover, to analyze the potential interactions between GRS and diet interventions, an interaction product term of GRS-by-diet was included in the models. As a secondary analysis with time as a repeated variable, linear mixed models were used to test the GRS effect on the trajectory of changes in glycemic traits by including a GRS-by-time interaction term. Because our analysis is hypothesis driven and primarily focused on the GRS rather than on individual SNPs, we did not adjust for multiple testing. The level of significance for all tests was set to P < 0.05. At the significance level of 0.05, the study had 80% power to detect gene-diet interaction effect sizes of 0.10 U for changes in fasting insulin and 0.10 U for changes in HOMA-IR at 6 mo. Statistical analyses were performed with SAS 9.2 software (SAS Institute Inc.).
RESULTS
Characteristics of the study population according to GRS
Baseline characteristics of participants according to low, median, and high tertiles of GRS are reported in Table 1. The GRS was calculated on the basis of 8 SNPs associated with habitual coffee consumption (Supplemental Tables 1 and 2). No differences in baseline characteristics across the tertiles of GRS were observed, except for the ethnicity proportion (Table 1). Levels of glucose, fasting insulin, and HOMA-IR at baseline were not related to the GRS. Supplemental Figure 2 presents the GRS distribution of habitual coffee consumption. We did not observe significant genetic associations between the GRS and baseline levels of glycemic traits and changes of these measures over the dietary intervention.
TABLE 1.
Baseline characteristics of the study participants according to genetic risk score1
| Tertiles of genetic risk score to habitual coffee consumption |
||||
| T1 (n = 211) (lowest) | T2 (n = 220) | T3 (n = 249) (highest) | P | |
| Age, y | 51.34 ± 9.10 [51.82 (45.48, 57.76)] | 50.79 ± 9.01 [51.75 (44.86, 57.46)] | 52.16 ± 8.87 [52.91 (46.86, 58.09)] | 0.303 |
| Sex | ||||
| Female | 131 (62.1) | 133 (60.0) | 147 (59.0) | 0.356 |
| Male | 80 (38.9) | 87 (40.0) | 102 (41.0) | |
| Race/ethnicity | ||||
| White | 192 (91.0) | 170 (77.3) | 210 (84.3) | <0.001 |
| Black | 19 (9.0) | 50 (22.7) | 39 (15.7) | |
| Diet groups (% fat/%protein/% carbohydrate) | ||||
| Group 1 (20/15/65) | 54 (25.59) | 56 (25.45) | 61 (24.50) | 0.351 |
| Group 2 (20/25/55) | 40 (18.96) | 62 (28.18) | 68 (27.31) | |
| Group 3 (40/15/45) | 57 (27.01) | 42 (19.09) | 70 (28.11) | |
| Group 4 (40/25/35) | 60 (28.44) | 60 (27.27) | 50 (20.08) | |
| Weight, kg | 93.85 ± 16.19 [92.53 (80.10, 106.70)] | 94.28 ± 15.59 [92.45 (83.00, 104.50)] | 93.85 ± 14.11 [94.25 (83.45, 103.20)] | 0.863 |
| Alcohol, g/d | 5.30 ± 8.48 [1.22 (0.01, 7.52)] | 4.87 ± 7.93 [0.87 (0.01, 7.11)] | 5.95 ± 8.77 [1.39 (0.01, 8.71)] | 0.638 |
| Baseline activity factor | 1.58 ± 0.11 [1.59 (1.48, 1.48)] | 1.59 ± 0.10 [1.61 (1.51, 1.70)] | 1.59 ± 0.12 [1.63 (1.50, 1.70)] | 0.552 |
| Smoking | 0.113 | |||
| Current smoker | 14 (6.63) | 6 (2.73) | 8 (3.21) | |
| Former smoker | 81 (38.39) | 73 (33.18) | 96 (38.56) | |
| Never smoked | 116 (54.98) | 141 (64.09) | 145 (58.23) | |
| BMI, kg/m2 | 32.89 ± 3.92 [32.90 (29.90, 35.90)] | 32.85 ± 3.86 [32.90 (29.65, 36.25)] | 32.77 ± 3.74 [32.50 (29.70, 35.90)] | 0.718 |
| Waist circumference, cm | 103.96 ± 13.66 [105.10 (92.33, 114.68)] | 104.29 ± 12.77 [103.50, (93.61, 114.65)] | 104.62 ± 11.74 [103.98 (96.38, 112.75)] | 0.965 |
| Glucose, mmol/L | 5.11 ± 0.70 [5.00 (4.67, 5.33)] | 5.03 ± 0.57 [5.00 (4.67, 5.31)] | 5.18 ± 0.67 [5.06 (4.78, 5.50)] | 0.096 |
| Daily dietary intake | ||||
| Energy, kcal | 1976.39 ± 559.89 [1910.28 (1551.72, 2354.39)] | 1965.25 ± 555.33 [1924.53 (1618.70, 2269.53)] | 1929.73 ± 540.04 [1880.29 (1482.12, 2243.54)] | 0.840 |
| Carbohydrate, % | 45.06 ± 7.73 [45.73 (39.75, 51.24)] | 44.33 ± 7.23 [44.63 (40.00, 48.98)] | 44.45 ± 7.70 [44.89 (38.61, 49.98)] | 0.684 |
| Fat, % | 37.10 ± 6.00 [37.31 (32.76, 41.29)] | 36.71 ± 6.03 [37.34 (33.12, 41.24)] | 37.20 ± 5.82 [36.68 (33.42, 41.42)] | 0.817 |
| Protein, % | 17.92 ± 3.11 [17.51 (16.12, 19.57)] | 18.79 ± 3.74 [17.98 (16.62, 20.58)] | 17.87 ± 3.20 [17.17 (15.63, 20.48)] | 0.061 |
| Biomarkers of adherence | ||||
| Urinary nitrogen, g | 12.16 ± 4.66 [11.65 (8.90, 14.92)] | 12.16 ± 4.66 [11.69 (9.27, 14.35)] | 12.16 ± 3.89 [11.76 (9.31, 14.35)] | 0.753 |
| Respiratory quotient | 0.84 ± 0.04 [0.84 (0.82, 0.87)] | 0.83 ± 0.04 [0.84 (0.81, 0.86)] | 0.84 ± 0.04 [0.84 (0.81, 0.87)] | 0.314 |
| Weight loss at 6 mo, kg | −6.87 ± 5.61 [−6.55 (−9.55,−3.18)] | −6.70 ± 5.76 [−5.95 (−10.75, −2.45)] | −7.00 ± 5.95 [−6.33 (−10.83, −2.55)] | 0.968 |
| Weight loss at 2 y, kg | −3.73 ± 7.36 [−3.23 (−6.85, 0.65)] | −4.84 ± 7.66 [−4.13 (−7.79, −0.04)] | −4.18 ± 7.82 [−2.58 (−8.00, 0.35)] | 0.226 |
| Insulin, μU/mL | 11.42 ± 5.88 [10.50 (7.10, 14.30)] | 12.19 ± 7.54 [10.00 (7.00, 15.70)] | 13.07 ± 9.04 [11.30 (7.10, 16.50)] | 0.135 |
| HOMA-IR | 2.66 ± 1.58 [2.42 (1.56, 3.30)] | 2.78 ± 1.87 [2.23 (1.51, 3.64)] | 3.09 ± 2.28 [2.55 (1.53, 4.06)] | 0.101 |
Data are expressed as means ± SDs [medians (IQRs)] or as n (%). P values were calculated by the use of the χ2 test for categorical variables and F tests in general linear models for continuous variables after adjustment for age, sex, and race. T, tertile.
The dietary intake and adherence markers of the 50% participants are reported in Supplemental Table 3. No significant differences were found in mean values or changes of nutrient intakes and biomarkers of adherence at 6 mo or 2 y across the GRS tertiles after adjustment for age, sex, and race (Supplemental Table 3). The biomarkers of adherence confirmed that participants modified their intake of macronutrients in the direction of the goals, although the targets were not fully achieved.
GRS-diet interactions on changes in glycemic traits
At 6 mo, we observed significant interactions between the GRS and dietary fat intake (high fat compared with low fat) on changes in fasting insulin and HOMA-IR (P-interaction = 0.023 and 0.022, respectively), adjusting for age, sex, race, physical activity, smoking, alcohol, seasonal variation, and baseline values of the respective outcomes. Participants in the low-fat diet group with a higher GRS had a greater reduction in fasting insulin and a marginally greater decrease in HOMA-IR, whereas the GRS in the high-fat group was not significantly associated with changes in these glycemic traits (Figure 1, Supplemental Table 4). We also did a sensitivity analysis by removing BDNF rs6265 and GCKR rs1260326 (P-interaction = 0.034 for fasting insulin and 0.033 for HOMA-IR at 6 mo). At 2 y, the GRS did not show a significant interaction with dietary fat on any of the glycemic traits (P-interaction > 0.05). Similar interactions were observed when the analysis was restricted to the white participants (data not shown).
FIGURE 1.
The effects of high- and low-fat diets (x-axis) on changes in insulin (A) and HOMA-IR (B) (y-axis) over 6 mo were based on the genetic risk score. T1 = lowest; T3 = highest. Data are expressed as means ± SDs. General linear models were used to calculate P values with adjustment for age, sex, race, physical activity, smoking, alcohol, seasonal variation, and baseline values of the respective outcomes. Values are means ± SEs. T, tertile.
Trajectory of changes in glycemic traits
In a secondary analysis, we used linear mixed models to assess the genotype-by-time interactive effect over the 2-y trial in those assigned to the low- or high-fat diet. No significant genotype-by-time interactions on changes in glucose, fasting insulin, or HOMA-IR were observed.
DISCUSSION
In this 2-y randomized weight-loss intervention trial, we observed significant interactions between the genetic determinants of habitual coffee consumption, assessed by a GRS, and dietary fat intake in relation to changes in fasting insulin and HOMA-IR. Participants with a higher GRS of habitual coffee consumption showed greater improvement in these metabolic measures in the low-fat diet intervention group than in the high-fat diet group.
Compelling evidence from epidemiologic studies has consistently linked coffee consumption with improved insulin sensitivity and glucose metabolism and a decreased risk of type 2 diabetes (19–22). Our findings, which showed that genetically determined high habitual coffee consumption interacted with weight-loss dietary interventions on glycemic changes, lent further support to the relation between coffee consumption and glucose metabolism (9).
Notably, greater improvement in these metabolic measures was observed in participants with genetic determinants of high habitual coffee consumption in the low-fat diet intervention group. An experimental study showed that switching from a high-fat diet to a low-fat diet normalized glucose tolerance and improved insulin secretion and insulin sensitivity (23). Besides, a low-fat diet alone also resulted in distinct improvement in insulin sensitivity in humans (24–26). Inverse associations have also been reported between coffee consumption and fasting insulin and HOMA-IR (27–29). Both low fat and high coffee consumption have been related to the alleviation of systemic inflammation (30, 31) and encouraging fat oxidation (32), suggesting that interactions between genetic variants of coffee consumption and dietary fat intake might be through these shared pathways. Further studies are warranted to investigate the mechanisms underlying our findings.
That no significant interactions were observed at 2 y is not surprising, because most of the participants regained weight between 6 mo and 2 y in POUNDS LOST as a result of diminished adherence to the dietary interventions (12). We assumed that weight regain from 6 mo to 2 y might lead to changes of the glycemic measures that at least partly attenuated the gene-diet interactions on these measures. Our results suggest that the modification of dietary fat on the association between genetic variation of habitual coffee consumption and changes in glycemic traits might be more pronounced in a short-term intervention and that the lack of effect at 2 y may be the result of the diminished adherence of participants.
To the best of our knowledge, this is the first study to investigate interactions between the genetic determinants of habitual coffee consumption and dietary fat on changes in glycemic traits in a large and long-term randomized diet intervention trial. We also assessed the time-course effects on dynamic changes of glycemic traits by repeating measurements during the 2 y of intervention. We created the GRS by following the widely accepted methods to sum the genetic variations weighted by their effect sizes (17, 18). Individual genetic variants or SNPs explain a very small fraction of the overall variation, and studies have demonstrated that the combination of a higher number of included variants with weak to moderate effect sizes into a GRS can explain larger proportion of heritability (33, 34). In our study, the sum of β-coefficients is 0.53 from the GRS, which means total allele effect sizes of 0.53 cups/d, explaining more genetic variation of habitual coffee consumption than single genetic variants. Besides, we did a sensitivity analysis by removing BDNF rs6265 and GCKR rs1260326, for these 2 genes are involved in other metabolisms rather than specific to coffee consumption, and the statistically significant results did not materially change.
Several limitations of our study should be noted. The confirmed loci together explained ∼1.3% of the habitual coffee consumption (9), and the small proportion of the variance explained by the genetic variants is similar to the proportion reported by most genome-wide association studies on other traits. POUNDS LOST is a randomized intervention trial in which the intervention diets were prescribed (12). We therefore analyzed interactions with diet interventions rather than coffee intake. However, as indicated in this report, the relation between the reported genetic variants and coffee intake has been well established in genome-wide studies (9). The participants of our study were mostly white and overweight or obese, and therefore, further studies are warranted to investigate whether our findings are applicable in people with different ethnicities or with normal weight. Besides, coffee consumption may also be influenced by other loci; therefore, the use of the calculated GRS in the current study might not fully reflect information in genetic determinants of coffee intake (9).
In conclusion, our data indicate that weight-loss diets that vary in fat content may modify the effects of the genetic determinant of habitual coffee consumption on glycemic changes in response to weight-loss diet interventions in a short term. Participants with a higher GRS might benefit more in improvement in fasting insulin and HOMA-IR by consuming a low-fat diet. Our study provides supportive evidence for potential precision dietary interventions based on the genetic determinants of habitual coffee consumption. Our findings suggest that individuals with genetically determined high coffee consumption may improve insulin and glucose metabolism by adhering to a low-fat diet. However, our conclusions should be interpreted with caution, and the findings should be replicated in other populations.
Acknowledgments
The authors’ responsibilities were as follows—LH, TH, DD, TW, DS, YH, WM, YZ, and LQ: designed the research, acquired, analyzed, and interpreted the data, and drafted and critically revised the manuscript; LH and TH: analyzed the data; JGAB and FMS: were involved in the collection and analysis of data and the funding of the initial project; FMS, CMC, and LQ: conducted the research, administration, material support, and study supervision; LQ: had full access to all of the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis; and all authors: read and approved the final manuscript. None of the authors reported a conflict of interest related to the study.
Footnotes
Abbreviations used: ABCG2, ATP binding cassette subfamily G member 2; AHR, aryl hydrocarbon receptor; BDNF, brain derived neurotrophic factor; CYP1A, cytochrome P450 family 1 subfamily A; GCKR, glucokinase regulator; GRS, genetic risk score; POR, p450 oxidoreductase; POUNDS LOST, Preventing Overweight Using Novel Dietary Strategies; SNP, single nucleotide polymorphism.
REFERENCES
- 1.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;37:569–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Rustenbeck I, Lier-Glaubitz V, Willenborg M, Eggert F, Engelhardt U, Jorns A. Effect of chronic coffee consumption on weight gain and glycaemia in a mouse model of obesity and type 2 diabetes. Nutr Diabetes 2014;4:e123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Akash MS, Rehman K, Chen S. Effects of coffee on type 2 diabetes mellitus. Nutrition 2014;30:755–63. [DOI] [PubMed] [Google Scholar]
- 4.Cornelis MC, Monda KL, Yu K, Paynter N, Azzato EM, Bennett SN, Berndt SI, Boerwinkle E, Chanock S, Chatterjee N, et al. Genome-wide meta-analysis identifies regions on 7p21 (AHR) and 15q24 (CYP1A2) as determinants of habitual caffeine consumption. PLoS Genet 2011;7:e1002033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sulem P, Gudbjartsson DF, Geller F, Prokopenko I, Feenstra B, Aben KK, Franke B, den Heijer M, Kovacs P, Stumvoll M, et al. Sequence variants at CYP1A1-CYP1A2 and AHR associate with coffee consumption. Hum Mol Genet 2011;20:2071–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rodenburg EM, Eijgelsheim M, Geleijnse JM, Amin N, van Duijn CM, Hofman A, Uitterlinden AG, Stricker BH, Visser LE. CYP1A2 and coffee intake and the modifying effect of sex, age, and smoking. Am J Clin Nutr 2012;96:182–7. [DOI] [PubMed] [Google Scholar]
- 7.Jia H, Aw W, Egashira K, Takahashi S, Aoyama S, Saito K, Kishimoto Y, Kato H. Coffee intake mitigated inflammation and obesity-induced insulin resistance in skeletal muscle of high-fat diet-induced obese mice. Genes Nutr 2014;9:389. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Choi BK, Park SB, Lee DR, Lee HJ, Jin YY, Yang SH, Shu JW. Green coffee bean extract improves obesity by decreasing body fat in high-fat diet-induced obese mice. Asian Pac J Trop Med 2016;9:635–43. [DOI] [PubMed] [Google Scholar]
- 9.Coffee and Caffeine Genetics Consortium; Cornelis MC, Byrne EM, Esko T, Nalls MA, Ganna A, Paynter N, Monda KL, Amin N, Fischer K, Renstrom F, et al. ; International Parkinson’s Disease Genomics Consortium (IPDGC); North American Brain Expression Consortium (NABEC); UK Brain Expression Consortium (UKBEC) . Genome-wide meta-analysis identifies six novel loci associated with habitual coffee consumption. Mol Psychiatry 2015;20:647–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Raimondo A, Rees MG, Gloyn AL. Glucokinase regulatory protein: complexity at the crossroads of triglyceride and glucose metabolism. Curr Opin Lipidol 2015;26:88–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Alvarez E, Roncero I, Chowen JA, Vázquez P, Blázquez E. Evidence that glucokinase regulatory protein is expressed and interacts with glucokinase in rat brain. J Neurochem 2002;80:45–53. [DOI] [PubMed] [Google Scholar]
- 12.Sacks FM, Bray GA, Carey VJ, Smith SR, Ryan DH, Anton SD, McManus K, Champagne CM, Bishop LM, Laranjo N, et al. Comparison of weight-loss diets with different compositions of fat, protein, and carbohydrates. N Engl J Med 2009;360:859–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Baecke JA, Burema J, Frijters JE. A short questionnaire for the measurement of habitual physical activity in epidemiological studies. Am J Clin Nutr 1982;36:936–42. [DOI] [PubMed] [Google Scholar]
- 14.Westerterp KR. Assessment of physical activity level in relation to obesity: current evidence and research issues. Med Sci Sports Exerc 1999;31:S522–5. [DOI] [PubMed] [Google Scholar]
- 15.Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and fJ-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28:412–9. [DOI] [PubMed] [Google Scholar]
- 16.Levy JC, Matthews DR, Hermans MP. Correct homeostasis model assessment (HOMA) evaluation uses the computer program. Diabetes Care 1998;21:2191–2. [DOI] [PubMed] [Google Scholar]
- 17.Qi Q, Chu AY, Kang JH, Jensen MK, Curhan GC, Pasquale LR, Ridker PM, Hunter DJ, Willett WC, Rimm EB, et al. Sugar-sweetened beverages and genetic risk of obesity. N Engl J Med 2012;367:1387–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Qi Q, Chu AY, Kang JH, Huang J, Rose LM, Jensen MK, Liang L, Curhan GC, Pasquale LR, Wiggs JL, et al. Fried food consumption, genetic risk, and body mass index: gene-diet interaction analysis in three US cohort studies. BMJ 2014;348:g1610. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Shi X, Xue W, Liang S, Zhao J, Zhang X. Acute caffeine ingestion reduces insulin sensitivity in healthy subjects: a systematic review and meta-analysis. Nutr J 2016;15:103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lecoultre V, Carrel G, Egli L, Binnert C, Boss A, MacMillan EL, Kreis R, Boesch C, Darimont C, Tappy L. Coffee consumption attenuates short-term fructose-induced liver insulin resistance in healthy men. Am J Clin Nutr 2014;99:268–75. [DOI] [PubMed] [Google Scholar]
- 21.Lee AH, Tan L, Hiramatsu N, Ishisaka A, Alfonso H, Tanaka A, Uemura N, Fujiwara Y, Takechi R. Plasma concentrations of coffee polyphenols and plasma biomarkers of diabetes risk in healthy Japanese women. Nutr Diabetes 2016;6:e212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Santos RM, Lima DR. Coffee consumption, obesity and type 2 diabetes: a mini-review. Eur J Nutr 2016;55:1345–58. [DOI] [PubMed] [Google Scholar]
- 23.Agardh CD, Ahrén B. Switching from high-fat to low-fat diet normalizes glucose metabolism and improves glucose-stimulated insulin secretion and insulin sensitivity but not body weight in C57BL/6J mice. Pancreas 2012;41:253–7. [DOI] [PubMed] [Google Scholar]
- 24.Sanders TA. High- versus low-fat diets in human diseases. Curr Opin Clin Nutr Metab Care 2003;6:151–5. [DOI] [PubMed] [Google Scholar]
- 25.Rosenfalck AM, Almdal T, Viggers L, Madsbad S, Hilsted J. A low-fat diet improves peripheral insulin sensitivity in patients with type 1 diabetes. Diabet Med 2006;23:384–92. [DOI] [PubMed] [Google Scholar]
- 26.Straznicky NE, O’Callaghan CJ, Barrington VE, Louis WJ. Hypotensive effect of low-fat, high-carbohydrate diet can be independent of changes in plasma insulin concentrations. Hypertension 1999;34:580–5. [DOI] [PubMed] [Google Scholar]
- 27.van Dam RM, Dekker JM, Nijpels G, Stehouwer CD, Bouter LM, Heine RJ. Coffee consumption and incidence of impaired fasting glucose, impaired glucose tolerance, and type 2 diabetes: the Hoorn Study. Diabetologia 2004;47:2152–9. [DOI] [PubMed] [Google Scholar]
- 28.Pham NM, Nanri A, Kochi T, Kuwahara K, Tsuruoka H, Kurotani K, Akter S, Kabe I, Sato M, Hayabuchi H, et al. Coffee and green tea consumption is associated with insulin resistance in Japanese adults. Metabolism 2014;63:400–8. [DOI] [PubMed] [Google Scholar]
- 29.Agardh EE, Carlsson S, Ahlbom A, Efendic S, Grill V, Hammar N, Hilding A, Ostenson CG. Coffee consumption, type 2 diabetes and impaired glucose tolerance in Swedish men and women. J Intern Med 2004;255:645–52. [DOI] [PubMed] [Google Scholar]
- 30.da Cunha FM, Duma D, Assreuy J, Buzzi FC, Niero R, Campos MM, Calixto JB. Caffeic acid derivatives: in vitro and in vivo anti-inflammatory properties. Free Radic Res 2004;38:1241–53. [DOI] [PubMed] [Google Scholar]
- 31.Lopez-Garcia E, van Dam RM, Qi L, Hu FB. Coffee consumption and markers of inflammation and endothelial dysfunction in healthy and diabetic women. Am J Clin Nutr 2006;84:888–93. [DOI] [PubMed] [Google Scholar]
- 32.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:E122–33. [DOI] [PubMed] [Google Scholar]
- 33.McGovern DP, Kugathasan S, Cho JH. Genetics of inflammatory bowel diseases. Gastroenterology 2015;149:1163–76 e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Dudbridge F. Power and predictive accuracy of polygenic risk scores. PLoS Genet 2013;9:e1003348. [DOI] [PMC free article] [PubMed] [Google Scholar]