Age-standardised mortality and incidence rate of ischaemic heart disease
We obtained the annual age-standardised IHD mortality and incidence rates per 100,000 population for each country from 1990 to 2018 from the Global Burden of Disease Study (GBD) 2019 . The GBD is a comprehensive programme of global and regional burden studies conducted by the University of Washington’s Institute for Health Metrics and Evaluation as an international collaboration of more than 145 countries . The estimates of the GBD adhere to the Guidelines for Accurate and Transparent Health Estimates Reporting standards developed by the World Health Organisation and other organisations.
We obtained coffee intake data (cups/day/population) (1 cup = 8 oz) for each country in 1990, 1995, 2000, 2005, 2010, 2015 and 2018 from the Global Dietary Database (GDD) . The GDD is an ongoing collaborative effort to produce the most reliable estimates of worldwide dietary intake to inform research and policymaking on health and nutrition worldwide. Data from the GDD are employed to estimate individual food and nutrient intake worldwide by country, year, sex, etc. The estimates are used to determine the disease burden and evaluate and improve public health and nutrition policies, among other purposes. The GDD has identified and obtained data for 1634 eligible survey years of data from public and private sources. Of these, 1240 have been checked, standardised, and approved for inclusion in the GDD 2018 model. Information on handling cases where country-representative surveys are not available, as well as coding methods and the GDD 2018 prediction model and its validation, are described elsewhere .
Information on coffee supplies was obtained from the food balance sheet published by the United Nations Food and Agriculture Organisation (FAO). The data are available from the FAO Statistics Division database (FAOSTAT), which provides the public with annual data on more than 245 countries and territories . The data are calculated based on various statistics and sources from each country; the domestic supply of each food item is the amount of food produced and imported minus the amount of food exported. The food supply is calculated as the domestic supply minus food losses and food used for feed and seed. Currently, data are available from 1961 to 2018. Given that the methodology for estimating the data has changed since 2014, we used data from 1990 (when GBD data were made available) to 2013 (before the change in the estimating methodology) to determine the mean coffee supply (g/day/capita) by country. Further details can be found elsewhere .
Socioeconomic and lifestyle indicators
Multiple socioeconomic and lifestyle factors are associated with the incidence of IHD and mortality. To exclude the influence of these factors, we obtained covariates that might affect IHD. For the socioeconomic indicators, we obtained data from the World Bank database from 1990 to 2013 on the gross domestic product (GDP) per capita (US $1000/capita), ageing rate (percentage of the population aged 65 years and older) and total population by country .
We obtained lifestyle factors from the GBD covariate database from 1990 to 2019. It included energy intake (total kcal/day/population), mean age-standardised alcohol consumption (g/day/population), years of education, age-standardised current smoking rate (%), age-standardised physical activity (1000 metabolic equivalents-min/week), mean body mass index (BMI) for individuals older than 20 years (kg/m2), age-standardised mean systolic blood pressure (SBP) (mm Hg) and age-standardised mean low-density lipoprotein cholesterol level (LDL-C) (converting mmol to mg/dl by dividing by 0.02586) . We also obtained the alcohol supply (grams of ethanol/day/capita) and energy supply (kcal/day/capita) by country from FAOSTAT.
Covariates for lifestyle factors such as smoking, BMI and physical activity were taken from the country-specific frequencies and distributions. Since this study was conducted on a country-by-country basis, the age and sex of the individuals in the analysis could not be included. Forage, however, we used the ageing rate as a covariate. The sex distribution is almost the same in all countries, and when compared across countries, there is little bias due to differences in sex distribution. To account for various regional differences, such as cultural and climatic differences, we also used the “Super Regions” as covariates, which are the seven regions of the GBD’s country classification: Central Europe, Eastern Europe and Central Asia; Latin America and Caribbean; North Africa and Middle East; South Asia; Southeast Asia, East Asia and Oceania; Sub-Saharan Africa; and High-income.
The present analysis used a variety of statistical data from those years in which no variables were missing. We limited our analysis to countries with populations of 1 million or more. Countries with smaller populations often do not have their own statistical systems, year-to-year variations in statistical values can be large, and outliers can significantly impact the overall results. For our analysis, we included 147 countries for which all data were available.
We examined the distribution and change over time in IHD mortality and incidence rates, coffee intake, socioeconomic indicators and lifestyles. We tested the trend of the mean values of each variable in 1990 (the first year of the analysis), 2000 (the early middle year), 2010 (the late middle year) and 2018 (the last year) with a general linear model. We divided the countries based on the Super Regions, calculated the population-weighted mean values for coffee consumption, IHD mortality and incidence rates by region for each year, and plotted them using the locally estimated scatterplot smoothing method.
To examine the association between coffee intake and IHD mortality and incidence rates and the changes in the association by year, we conducted a linear mixed model analysis using each country’s IHD mortality and incidence rates over 28 years from 1990 to 2018 as the dependent variables. We used the independent variables of coffee intake, year and the interaction between coffee intake and year. No covariates were added to Model 1, GDP was a covariate in Model 2 and GDP, ageing rate, years of education, smoking rate, physical activity and energy supply were covariates in Model 3. All independent variables centred on the grand mean. The random effects in the mixed model were the intercept and slope of the year for each country. We also specified a composite symmetric structure for the covariance matrix for each country and year. The fitting of the model was performed by maximising the log-likelihood.
For the sensitivity analyses, we further adjusted for alcohol consumption, BMI, SBP, LDL-C and Super Regions to test the robustness of the results. For all of the above analyses over 23 years from 1990 to 2013, we changed the independent variable from the coffee intake data from the GDD to the coffee supply data obtained from the FAOSTAT, and the covariate was changed from the GBD’s energy intake and alcohol consumption to the FAOSTAT’s energy supply and alcohol supply.
We used R 4.0.5 for the analysis , and p-values < 0.05 were considered statistically significant. The generalised linear mixed-effects models were fitted using the “lme” function of the “nlme” package .
This study was carried out in compliance with Declaration of Helsinki. Only publicly available data was used in this study, and no personal information was handled.