Coffee and health: what we still don't know

Menée aux Etats-Unis à partir des données de 2 cohortes de 15 551 femmes et 7 397 hommes, cette étude évalue l'association entre la consommation de café et les concentrations de 14 biomarqueurs plasmatiques des voies métaboliques et inflammatoires

Résumé en anglais

Coffee is a major source of caffeine, one of the most widely consumed stimulants in the world. Coffee also contains many other substances, such as polyphenols, diterpenoids, and trace minerals, that have been linked to various health conditions. The consumption of either caffeinated or decaffeinated coffee has been associated with increases and decreases in the risk of a number of chronic diseases. In the present issue of the Journal, Hang et al. (1) used several biomarkers to explore potential mechanisms underlying the putative protective effects of coffee in 2 large, well-characterized cohorts: the Nurses’ Health Study and the Health Professionals Follow-Up Study. They found that coffee consumption was associated with favorable profiles of several biomarkers in key metabolic and inflammatory pathways, including some that are specifically linked to cardiovascular disease. Many of the associations observed were similar for caffeinated and decaffeinated coffee, suggesting that the effects are not due to caffeine. Moreover, it appears that anti-inflammatory pathways might be driving many of the health-related effects of coffee. What remains unknown, however, is what components of coffee confer such benefits. One approach that can be used to home in on the relevant bioactive(s) is to include genetic variants affecting the metabolism or response to those bioactives, as proposed over a decade ago (2). Indeed, another study in this issue of the Journal did just that. Using prospective data from the UK Biobank, Zhou and Hypponen (3) evaluated the association between habitual coffee consumption and cardiovascular disease risk, and incorporated genetic markers of caffeine metabolism to determine the contribution of caffeine to this association. The authors identified a U-shaped association between coffee intake and cardiovascular disease that was not modified by the genetic markers tested. These findings contrast those of Cornelis et al. (4), who first reported that variation in cytochrome P450 family 1 subfamily A member 2 (CYP1A2), which affects the rate of caffeine metabolism and modifies the association between coffee consumption and the risk of myocardial infarction. Those earlier findings have since been extended to hypertension (5), impaired fasting glucose (6), and blood pressure (7), where in all cases coffee or caffeine was associated with adverse outcomes only in individuals with the CYP1A2 genotype associated with slow metabolism of caffeine. As such, the lack of a modifying effect of genetic differences in caffeine metabolism observed by Zhou and Hypponen seems unexpected (3). A recent study that used the same UK Biobank population found that coffee is inversely associated with overall mortality regardless of genetic differences affecting caffeine metabolism, although the association was only marginally significant for cardiovascular mortality (8). The disparate findings between the 2 studies (3, 8), despite using data from the same UK Biobank cohort and incorporating similar genetic risk scores based on the same genome-wide association study (9), is puzzling. Importantly, both studies have significant limitations that preclude a definitive conclusion about the role of genetics in modifying the association between coffee and health. For instance, Loftfield et al. (8) included decaffeinated coffee in the analyses, thus attenuating any effects of genetic differences in caffeine metabolism on the outcomes measured. Although both studies examined the role of CYP1A2, they also used genetic risk scores that have yet to be shown to predict differences in the rate of caffeine metabolism (3, 8). They also did not consider all sources of caffeine, which results in misclassification of the exposure of interest. Zhou and Hypponen (3) state that their prospective analysis minimizes the possibility of reverse causation. However, such a design reflects long-term intake, and coffee has distinct acute effects on cardiovascular phenotypes, such as blood pressure, that may affect its contribution to cardiovascular outcomes (10). A ute effects are better captured by case-control (4), or intervention (7), study designs. The previous study, showing that CYP1A2 modifies the association between coffee and myocardial infarction, found the effect was greatest in younger cases (4), suggesting an important modifying effect of age and that a lack of effect in an older cohort could be due to a healthy survivor effect. The authors found that a gene-coffee-age interaction did not reach the threshold for significance (P ≥ 0.11, data not shown), but a stratified analysis by median age should have been shown considering previous findings (4). As such, the modifying effects of CYP1A2 cannot be ruled out. Future studies that aim to focus on caffeine by incorporating the CYP1A2 genotype into the study design, as Zhou and Hypponen have done, must also include all sources of this bioactive. Otherwise, any estimate of the modifying effect of genetics will be affected by the imprecise estimate of exposure to caffeine. Studies that seek to better understand the role of other bioactives in coffee would be greatly strengthened by incorporating genetic variants that affect their metabolism or are potential targets of their action (2).