Freedman ND, Park Y, Abnet CC, Hollenbeck AR, Sinha R. Association of coffee drinking with total and cause-specific mortality. NEJM. 2012;366:1891-1904.
Longitudinal cohort study
The National Institutes of Health–American Association of Retired Persons (NIH–AARP) Diet and Health Study enrolled participants from 1995–1996 and has followed them for a number of health conditions since then. This study included a total of 402,260 participants (n=229,119 for men; n=173,141 for women). All participants resided in 1 of 6 states (California, Florida, Louisiana, New Jersey, North Carolina, or Pennsylvania) or 2 metropolitan areas (Atlanta or Detroit). Participants’ median age at enrollment was 60.3–63.5 years (reported age stratified by gender and coffee intake). For both genders, participants reporting consuming 1 cup of coffee/day had the oldest median age, while those reporting ≥6 cups/day were the youngest. Exclusion criteria included cancer, heart disease, and history of stroke. In addition, questionnaires were excluded if any of the following applied: questionnaires completed by a surrogate or representative (e.g., spouse, caregiver), no information was provided about tobacco use, reported energy (kcal) intake was very high or very low, or the participant died before the completed questionnaire was received by investigators.
All participants completed a 124-item food frequency questionnaire (FFQ) at baseline (1995–1996). A subset of participants (n=1,953) also completed 24-hour dietary recall questionnaires on 2 non-consecutive days. Years of follow-up were from baseline until date of death, or December 31, 2008, when mortality was assessed. Daily coffee intake was categorized as: none; <1 cup; 1 cup; 2–3 cups; 4–5 cups; or ≥6 cups. Nearly all participants (96.5%) completing the FFQs volunteered information about whether they consumed caffeinated or decaffeinated coffee. The majority (79%) reported drinking ground coffee; another 19% reported drinking instant coffee. One percent each reported drinking espresso or did not indicate what kind of coffee was consumed.
Primary Outcome Measures
Statistics were computed and reported separately, according to gender and level of reported coffee consumption. The primary outcome measures were total mortality and cause-specific mortality based on ICD-9 and ICD-10 codes listed on death certificates. Deaths due to cancer, heart disease, respiratory disease, stroke, injuries and accidents, diabetes, infections, and all other causes were each considered separately.
Participants were followed for 14 years, during which 33,731 men and 18,784 women died. Overall, coffee drinkers were more likely to smoke cigarettes; consume >3 alcoholic drinks/day; eat red meat; have a lower level of education; be less likely to exercise; eat fewer fruits, vegetables, and white meat. They were less likely to report having diabetes (more so with women).
In their primary analysis, investigators computed hazard ratios (95% confidence intervals) for death associated with coffee consumption (<1, 1, 2, 3, 4, 5, ≥6 cups daily) as compared to no coffee consumption. The level of significance was set at P<0.5. For their secondary analysis, investigators determined risk estimates for categories of consumption of caffeinated and decaffeinated coffee.
Approximately 2/3 of coffee drinkers reported drinking caffeinated coffee. A subgroup analysis found similar results for both caffeinated and decaffeinated coffee drinkers. Smoking status (current vs. never or former smokers) was shown to significantly influence both total and cause-specific mortality for both genders (men: P<0.001 for interaction; women: P=0.002 for interaction). Self-reported health at baseline (good or very good to excellent health vs. poor to fair health) was also shown to significantly influence total and cause-specific mortality (P<0.001 for interaction, both genders). Stronger associations were noted with better health.
After adjusting for potential confounders, drinking ≥6 cups of coffee daily was associated with a 10% reduction in total mortality for men [HR=0.90 (0.85–0.96); P-value for trend <0.001]. This analysis also revealed significant reductions in cause-specific mortality. Men who reported drinking ≥6 cups of coffee daily were 40% less likely to die from diabetes [HR=0.60 (0.39–0.94); P-value for trend <0.001]; 41% less likely to die from infections [HR=0.59 (0.37–0.95); P-value for trend=0.001]; and 28% less likely to die from all other causes [HR=0.72 (0.61–0.84); P-value for trend <0.001].
Among women who reported drinking ≥6 cups of coffee daily, the adjusted analysis showed an associated 15% reduction in total mortality [HR=0.85 (0.78–0.93); P-value for trend <0.001]. Drinking ≥6 cups of coffee daily was also associated with significant reductions in cause-specific mortality. Women were 28% less likely to die from heart disease [HR=0.72 (0.59–0.88); P-value for trend <0.001]; 23% less likely to die from respiratory disease [HR=0.77 (0.61–0.99); P-value for trend <0.001]; and 28% less likely to die from all other causes [HR=0.72 (0.57–0.90); P-value for trend <0.001].
Beyond being a beverage richly steeped in culture and tradition, coffee contains a complex mixture of more than 1,000 different biologically active compounds.1 Mounting scientific evidence points toward equally complex biological effects, some positive, others less so. Until recently, research has focused largely upon coffee’s methylxanthine content (primarily caffeine), with its acute effects on the cardiovascular system and metabolism. These results supported a perception of coffee drinking as potentially harmful.2
Data from large cohort studies such as this one by Freedman et al suggest that the effects of coffee may be beneficial.3–6 This study was published on the heels of another large (N=42,659) cohort study reporting on the association between coffee and chronic disease, the European Prospective Investigation into Cancer and Nutrition (EPIC)-Germany study.7 This study by Floegel et al collected similar baseline information to the present study, with some key differences. For example, EPIC-Germany participants were approximately 10 years younger than NIH-AARP Diet and Health participants.8,9 Also, although EPIC-Germany investigators collected caffeinated and decaffeinated coffee consumption data separately, very few (4.5%) reported drinking exclusively decaffeinated coffee. An additional 18.6% reporting drinking both, while fully 72.6% of EPIC-Germany participants reported drinking exclusively caffeinated coffee.10 This study found no association between coffee consumption and total chronic disease. Further analysis found no association between coffee consumption and combined cardiovascular disease, myocardial infarction, stroke, and total cancers. The investigators found a possible decrease in risk for type 2 diabetes mellitus.11
Multiple studies have demonstrated associations between coffee consumption and decreased diabetes,12,13 stroke,14–18 and death due to inflammatory diseases19,20 or injuries and accidents.21–23 Although investigations into the relations between coffee consumption and heart disease have been more mixed, a 2009 meta-analysis found similar results to Freedman et al and Floegel et al, both of whom found no association.24–26 Coffee drinking has also been shown to help reduce the risk of developing Parkinson’s disease.27
Coffee’s various phenolic compounds (eg, chlorogenic acid, ferulic acid, p-coumaric acid) possess antioxidant activities and may act in conjunction with magnesium, trigonelline, and quinides to help improve insulin sensitivity.28,29 Coffee’s diterpenes (cafestol and kahweol) have demonstrated anticarcinogenic properties.30 Although they may also increase the LDL lipid fraction, LDL drawn from coffee drinkers has demonstrated a higher resistance to oxidation, an effect believed due to the uptake of the antioxidant phenolic acids (eg, caffeic acid, ferulic acid, p-coumaric acid, all in conjugated forms).31 Thus, any small increase in LDL may be offset by an increased resistance to oxidation.
Coffee drinkers are also self-selected, and it is likely that key differences exist between those who drink coffee and those who do not.
As with most epidemiologic studies evaluating the effects of coffee on health, these studies relied upon assessment of coffee consumption at a single time point.32,33 The Nurses’ Health study showed that coffee intake is relatively stable over time.34 Coffee drinkers are also self-selected, and it is likely that key differences exist between those who drink coffee and those who do not. The degree to which these differences may influence studies evaluating associations between drinking coffee (at any level) and not drinking coffee at all remains to be fully explored.
Both Freedman et al and Floegel et al noted that coffee drinkers are also more likely to drink alcohol and use tobacco.35,36 As smoking and drinking each pose independent health risks, failing to control for these may lead to spurious findings. In fact, many earlier analyses did not control for the effects of smoking.37 Freedman et al’s study illustrated this well. An initial analysis controlled only for age and showed increased mortality associated with coffee drinking. After controlling for tobacco use and other variables, the same data showed the opposite—reductions in mortality that increased with increasing cups of coffee/day.38 One of the most striking results of Freedman et al’s study was the dose-response element. That is, the P-value for trend was positive, and significant, for many of the mortality states studied. It was most pronounced for women and for type 2 diabetes mellitus.39 This supports the theory that coffee’s antioxidant content underlies the observed health benefits.
For years, the public, along with much of the naturopathic community, has viewed coffee (especially caffeinated) drinking as less than healthy, if not downright unhealthy. It may be time to rethink this notion.40 The Western diet (or the Standard American Diet) is relatively low in antioxidant-rich foods such as fruits and vegetables. Coffee, tea, and chocolate represent 3 foods high in antioxidants that remain common. While essential that we encourage patients to eat more of antioxidant rich colored fruits and vegetables, it may not be advisable to suggest a blanket decrease in another (coffee), unless clear evidence points to harm for that individual patient.