Coffee drinking and Mortality

Paolo Boffetta-paolo.boffetta@mssm.edu

We plan to use intramural branch funds to support this project.

Coffee is one of the most widely and habitually consumed dietary items worldwide. Consequently, there is a growing scientific and public interest in how coffee consumption impacts public health. Studies have reported both positive and inverse associations of coffee with cardiovascular disease, but they have generally reported inverse associations with diabetes, Parkinson Disease, depression and certain cancers, including liver, colorectal, endometrial, melanoma, and nonmelanoma skin cancer. It is, however, the association of coffee drinking with mortality that is most relevant when considering public health impact and dietary guidance. In general, studies have observed inverse associations with all-cause mortality, and recently, the Scientific Report of the 2015 Dietary Advisory Committee concluded that moderate coffee consumption (3 to 5 cups per day) can be a part of a healthy diet. Yet a number of important questions that may impact dietary guidance remain unresolved.

First, there is concern that coffee drinking may increase mortality among younger individuals. In 2013, a prospective study reported that coffee drinking was positively associated with mortality among individuals under the age of 55. Since most cohorts have included older populations and mortality rates are lower among younger age groups, few studies have the ability to replicate this finding, and the association between coffee drinking and mortality among younger populations remains controversial. The proposed study, which would include cohorts with broad age ranges, would therefore provide an important contribution to the literature.

Second, existing studies have each had a limited number of cases among the highest consumption categories (e.g. ≥6 cups/day) precluding stable risk estimates at the extreme end of the distribution and in particular among never smokers who tend to drink less coffee. Coffee drinking and cigarette smoking are strongly positively correlated. Consequently, the impact of residual confounding is likely greater among the heaviest consumers. A sufficiently powered subgroup analysis of never smokers would clarify the impact of residual confounding by smoking and address the question of a nonlinear association between coffee drinking and mortality.

Third, racial and ethnic minorities are largely under-represented in studies of coffee drinking and mortality. Given that patterns of coffee consumption vary considerably by race/ethnicity, it is important to assess associations in these subgroups.

Fourth, only a few large cohort studies have considered causes of death beyond composite measures of cardiovascular disease and cancer. The NCI Cohort Consortium affords the opportunity to better ascertain the drivers of the observed association with all-cause mortality, including different cardiovascular and cancer mortality endpoints.

Finally, associations of caffeinated and decaffeinated coffee drinking with all-cause mortality have generally been similar, but the associations with cause-specific mortality are less clear due to limited sample sizes as decaffeinated coffee drinking is less common than caffeinated coffee drinking.
The NCI Cohort Consortium provides an opportunity to examine these issues in a large, diverse population with the advantage of a standard analytic approach across studies.

The overall goal of this study is conduct pooled analyses of coffee drinking and mortality in a large, diverse population using a standard analytic approach across studies in order to address unresolved questions related to age, heavy coffee consumption, race/ethnicity, cause-specific death, and decaffeinated coffee intake.

Our primary objectives are to:
1) Provide stable estimates of the mortality risks across the entire range of low, moderate, high and very high coffee intake.
2) Assess whether age modifies the association of coffee drinking and mortality
3) Evaluate associations separately among smokers and non-smokers
4) Evaluate mortality risks by subgroups of race/ethnicity and decaffeinated/caffeinated coffee drinkers.
5) Evaluate associations among participants who have and do not have comorbidities.

Our study will invite all studies in the NCI Cohort Consortium with >5 years of follow-up, ≥1000 deaths, and a baseline year of 1970 or later to participate. Studies must also have ascertained coffee intake, smoking status, and preexisting cancer and cardiovascular disease. We will use proportional-hazards models to estimate associations of coffee with total and cause-specific mortality and effect modification by age.

Retrospective assessment of coffee intake would require proxy respondents and is subject to recall bias. Furthermore, individual cohort studies have had only a modest number of deaths among several key subgroups including those <55 years old, heavy coffee drinkers, and ethnic/racial minorities. Information on preexisting disease and extended follow-up time address concerns about reverse causality.

1000

Total mortality and cause-specific mortality ascertained from death certificates or medical records and coded according to the International Classification of Diseases, Ninth Revision (ICD-9).

Our primary exposure of interest is coffee consumption. Ideally, coffee consumption would have been measured at baseline using a food frequency questionnaire that includes information on frequency of consumption and portion size. Coffee drinking would be defined as follows: none, <1 cup/day, 1 cup/day, 2–3 cups/day, 4–5 cups/day, ≥6 cups/day. Additionally, for studies with detailed information above 6 cups/day, we would evaluate higher categories of consumption. For secondary analyses, we also request any available data on decaffeinated and caffeinated coffee consumption and on former coffee drinking.

We propose using covariates that have already been harmonized for studies of mortality, BMI and physical activity including: smoking status, number of years since the person stopped smoking, alcohol intake, physical activity level, educational level, marital status, body mass index, age, gender, race/ethnicity, and co-morbidities (e.g. cancer other than nonmelanoma skin cancer and heart disease manifested as angina, arrhythmia or myocardial infarction). We also request data on diabetes status. We do not require additional dietary variables (e.g. fruit, vegetable, meat, and energy intake) as our previous studies of coffee and mortality suggest that adjustment for dietary variables does not meaningfully alter risk estimates. Thus, it is unlikely that the ability to adjust for these variables justifies the harmonization effort.

No