Associations of Coffee Consumption With Liver Injury Markers
Associations of Coffee Consumption With Liver Injury Markers
The Insulin Resistance Atherosclerosis Study (IRAS) is an epidemiological investigation of the relationship of insulin resistance (IR) with cardiovascular risk factors in a large multi-ethnic population. The design of IRAS has been published in detail. Briefly, data collection was initiated in October 1992 with participants recruited from four geographic areas in the United States (San Antonio TX, San Luis Valley CO, Oakland and Los Angeles CA). The recruitment target was to achieve equal participation across categories of sex, ethnicity (non-Hispanic white, Hispanic, African American), age and glucose tolerance status (normal, impaired glucose tolerance, diabetes). In total, the baseline IRAS population included 1625 participants, from which written informed consent was obtained. This study was conducted in accordance with the Helsinki Declaration and the institutional review boards at Kaiser Permanente Division of Research (Oakland, CA), University of California (Los Angeles, CA), University of Texas Health Science Center at San Antonio (San Antonio, TX), the University of Colorado (Denver, CO), and the Bowman Gray School of Medicine (Winston-Salem, NC) approved the project.
For the current cross-sectional study, the population was restricted to individuals without T2DM, as defined by the 1999 World Health Organization criteria. Further exclusions were made on the basis of missing liver enzyme data or extremes of energy intake, resulting in a final study population of n = 1005. Fetuin-A data were not available for all individuals however participant characteristics did not differ substantially between those for which fetuin-A data was available (n = 650) and those for which it was not (n = 355) (data not shown).
Participant assessments were conducted across two 4 h visits approximately 1 week apart and participants were asked to fast for 12 h prior to each visit. Diabetes status was assessed on the first visit using a 75 g oral glucose tolerance test (OGTT). Blood was collected for fasting and 2-h glucose samples. IR was assessed at the second visit using the insulin-modified frequently sampled intravenous glucose tolerance test (FSIGTT), and insulin sensitivity (SI), was calculated using mathematical modeling methods (MINMOD version 3.0, 1994, Los Angeles, CA), as has been described in previous publications. Socio-demographic information was collected by participant report and questionnaires assessed energy expenditure, alcohol consumption and smoking behaviours.
Dietary intake over the past year was assessed using a validated interviewer-administered 114-item food frequency questionnaire (FFQ) modified for the IRAS from the National Cancer Institute's Health Habits and History Questionnaire to include a number of regional and ethnic foods. This measure was validated in a sub-sample of the population through the administration of eight 24-h recalls administered on randomly selected days over a 1 year period. Results demonstrated desirable validity and confirmed the appropriateness of this instrument for use in this population. Although the validation did not specifically address coffee, it is established that coffee drinking behaviour is accurately captured in FFQs (American women r = 0.8, men r = 0.9, respectively, vs dietary records).
Caffeinated and decaffeinated coffee consumption were assessed individually and participants reported on both the frequency and quantity of consumption. Frequency was reported using a nine category scale with responses ranging from "never or less than 1 per month" to "six or more times per day". Individuals indicated whether their portion sizes were "small, medium or large compared with other men or women about your age". Frequency and portion size data were integrated in order to provide a single value reflective of intake by weighting the intake frequency by a factor of 0.5, 1 or 1.5 for the reported portion size of small, medium or large, respectively. Thus, one serving of an item reflects one participant-identified medium serving. Intakes of caffeinated and decaffeinated coffee were summed for a measure of total coffee. Categories of vegetable, fruit and whole grain intakes were calculated according to groupings established in previous work in IRAS. Nutrient analysis of FFQ data was conducted using the HHHQ-DIETSYS analysis software (version 3.0, 1993; National Cancer Institute, Bethesda, MD, USA).
Fasting blood samples were used to determine markers of liver injury. Liver enzymes ALT and AST were measured using standard methods at the central IRAS laboratory with a Paramax PLA instrument (Baxter). Fetuin-A was determined using a sandwich-format immunoassay developed by Tethys (Tethys Bioscience, Emeryville, CA) with antibodies from R&D systems (Minneapolis, USA). The assay had a detection limit of 0.5 ng/ml and an inter-assay coefficient of variation of 11 %. NAFLD liver fat score was calculated according to the equation developed and validated by Kotronen et al. which includes fasting serum insulin, ALT, AST and presence of the metabolic syndrome (International Diabetes Federation harmonized definition).
Statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC). Participant characteristics were tested for differences across categories of total coffee intake. Chi-square tests were used to test differences in frequencies of categorical variables. For continuous variables, differences in means were tested using ANOVA unless the variable was highly skewed. In this case, the Kruskal-Wallis test for non-parametric variables was used.
Linear regression analyses were conducted to assess the independent relationships of caffeinated and decaffeinated coffee consumption, modeled as continuous variables, with markers of liver injury. The distributions of outcome variables were evaluated and natural log transformations were used for all outcome variables as they resulted in more normal distributions. Due to the presence of negative values, a constant of 6.4 was added to all NAFLD liver fat score data before the transformation. The independent associations of caffeinated and decaffeinated coffee consumption with outcome variables were analyzed using staged multivariate models. A base model adjusted for age, sex, and ethnicity. A second model additionally adjusted for energy intake and expenditure, BMI, smoking, alcohol consumption and education. Finally, a third model additionally adjusted for a number of dietary factors including whole grain consumption, vegetable and fruit intake, percent of energy intake from saturated and polyunsaturated fat as well as intake of the alternate type of coffee consumed (caffeinated or decaffeinated), and sugar-sweetened beverages (regular soda and lemonade/sweetened mineral water). To test for potential modifiers of the relationship between coffee consumption and markers of liver injury, interaction terms were tested for age, sex, ethnicity, BMI and S I, and were judged to be statistically significant if the p value was <0.05.
A sensitivity analysis was performed in which multivariate linear regression analyses were conducted with total coffee consumption as the exposure variable. Because of the well-known association between high alcohol consumption and elevations in liver transaminases, sensitivity analyses were performed in which regression models were repeated when individuals in the two highest categories of alcohol consumption were excluded (≥1 alcoholic drink/day). In addition, a fourth mechanistic regression model which additionally adjusted for insulin sensitivity was added in order to assess the contribution of S I to the observed association of coffee and the outcome variables.
Methods
Study Population and Design
The Insulin Resistance Atherosclerosis Study (IRAS) is an epidemiological investigation of the relationship of insulin resistance (IR) with cardiovascular risk factors in a large multi-ethnic population. The design of IRAS has been published in detail. Briefly, data collection was initiated in October 1992 with participants recruited from four geographic areas in the United States (San Antonio TX, San Luis Valley CO, Oakland and Los Angeles CA). The recruitment target was to achieve equal participation across categories of sex, ethnicity (non-Hispanic white, Hispanic, African American), age and glucose tolerance status (normal, impaired glucose tolerance, diabetes). In total, the baseline IRAS population included 1625 participants, from which written informed consent was obtained. This study was conducted in accordance with the Helsinki Declaration and the institutional review boards at Kaiser Permanente Division of Research (Oakland, CA), University of California (Los Angeles, CA), University of Texas Health Science Center at San Antonio (San Antonio, TX), the University of Colorado (Denver, CO), and the Bowman Gray School of Medicine (Winston-Salem, NC) approved the project.
For the current cross-sectional study, the population was restricted to individuals without T2DM, as defined by the 1999 World Health Organization criteria. Further exclusions were made on the basis of missing liver enzyme data or extremes of energy intake, resulting in a final study population of n = 1005. Fetuin-A data were not available for all individuals however participant characteristics did not differ substantially between those for which fetuin-A data was available (n = 650) and those for which it was not (n = 355) (data not shown).
Data Collection
Participant assessments were conducted across two 4 h visits approximately 1 week apart and participants were asked to fast for 12 h prior to each visit. Diabetes status was assessed on the first visit using a 75 g oral glucose tolerance test (OGTT). Blood was collected for fasting and 2-h glucose samples. IR was assessed at the second visit using the insulin-modified frequently sampled intravenous glucose tolerance test (FSIGTT), and insulin sensitivity (SI), was calculated using mathematical modeling methods (MINMOD version 3.0, 1994, Los Angeles, CA), as has been described in previous publications. Socio-demographic information was collected by participant report and questionnaires assessed energy expenditure, alcohol consumption and smoking behaviours.
Exposure Variables
Dietary intake over the past year was assessed using a validated interviewer-administered 114-item food frequency questionnaire (FFQ) modified for the IRAS from the National Cancer Institute's Health Habits and History Questionnaire to include a number of regional and ethnic foods. This measure was validated in a sub-sample of the population through the administration of eight 24-h recalls administered on randomly selected days over a 1 year period. Results demonstrated desirable validity and confirmed the appropriateness of this instrument for use in this population. Although the validation did not specifically address coffee, it is established that coffee drinking behaviour is accurately captured in FFQs (American women r = 0.8, men r = 0.9, respectively, vs dietary records).
Caffeinated and decaffeinated coffee consumption were assessed individually and participants reported on both the frequency and quantity of consumption. Frequency was reported using a nine category scale with responses ranging from "never or less than 1 per month" to "six or more times per day". Individuals indicated whether their portion sizes were "small, medium or large compared with other men or women about your age". Frequency and portion size data were integrated in order to provide a single value reflective of intake by weighting the intake frequency by a factor of 0.5, 1 or 1.5 for the reported portion size of small, medium or large, respectively. Thus, one serving of an item reflects one participant-identified medium serving. Intakes of caffeinated and decaffeinated coffee were summed for a measure of total coffee. Categories of vegetable, fruit and whole grain intakes were calculated according to groupings established in previous work in IRAS. Nutrient analysis of FFQ data was conducted using the HHHQ-DIETSYS analysis software (version 3.0, 1993; National Cancer Institute, Bethesda, MD, USA).
Outcome Variables
Fasting blood samples were used to determine markers of liver injury. Liver enzymes ALT and AST were measured using standard methods at the central IRAS laboratory with a Paramax PLA instrument (Baxter). Fetuin-A was determined using a sandwich-format immunoassay developed by Tethys (Tethys Bioscience, Emeryville, CA) with antibodies from R&D systems (Minneapolis, USA). The assay had a detection limit of 0.5 ng/ml and an inter-assay coefficient of variation of 11 %. NAFLD liver fat score was calculated according to the equation developed and validated by Kotronen et al. which includes fasting serum insulin, ALT, AST and presence of the metabolic syndrome (International Diabetes Federation harmonized definition).
Statistical Analysis
Statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC). Participant characteristics were tested for differences across categories of total coffee intake. Chi-square tests were used to test differences in frequencies of categorical variables. For continuous variables, differences in means were tested using ANOVA unless the variable was highly skewed. In this case, the Kruskal-Wallis test for non-parametric variables was used.
Linear regression analyses were conducted to assess the independent relationships of caffeinated and decaffeinated coffee consumption, modeled as continuous variables, with markers of liver injury. The distributions of outcome variables were evaluated and natural log transformations were used for all outcome variables as they resulted in more normal distributions. Due to the presence of negative values, a constant of 6.4 was added to all NAFLD liver fat score data before the transformation. The independent associations of caffeinated and decaffeinated coffee consumption with outcome variables were analyzed using staged multivariate models. A base model adjusted for age, sex, and ethnicity. A second model additionally adjusted for energy intake and expenditure, BMI, smoking, alcohol consumption and education. Finally, a third model additionally adjusted for a number of dietary factors including whole grain consumption, vegetable and fruit intake, percent of energy intake from saturated and polyunsaturated fat as well as intake of the alternate type of coffee consumed (caffeinated or decaffeinated), and sugar-sweetened beverages (regular soda and lemonade/sweetened mineral water). To test for potential modifiers of the relationship between coffee consumption and markers of liver injury, interaction terms were tested for age, sex, ethnicity, BMI and S I, and were judged to be statistically significant if the p value was <0.05.
A sensitivity analysis was performed in which multivariate linear regression analyses were conducted with total coffee consumption as the exposure variable. Because of the well-known association between high alcohol consumption and elevations in liver transaminases, sensitivity analyses were performed in which regression models were repeated when individuals in the two highest categories of alcohol consumption were excluded (≥1 alcoholic drink/day). In addition, a fourth mechanistic regression model which additionally adjusted for insulin sensitivity was added in order to assess the contribution of S I to the observed association of coffee and the outcome variables.
Source...