Kim et al. recently published the results of a prospective cohort study of the association between self-reported coffee consumption at baseline and incident cardiac tachyarrhythmias using a sample of 386,253 patients from the UK Biobank database.1 Contrary to the expectations of conventional medical wisdom, they report that patients who drank coffee had fewer tachyarrhythmias than those who did not in a dose-dependent manner. Each daily cup of coffee was associated with 3% fewer tachyarrhythmias (HR, 0.97; 95% CI, 0.96–0.98) over a 4.5 ± 3.1 year follow-up in a multivariate analysis adjusting for 19 other potential confounders. The specific tachyarrhythmias with statistically significant associations were atrial fibrillation and/or atrial flutter and supraventricular tachycardia. Of course, even with multivariate analysis, for many reasons, an observational study cannot conclusively prove that coffee consumption protects against tachyarrhythmias. In this case, reasons include self-report bias, time-dependent changes in coffee consumption, residual/unmeasured confounding, and reverse-causation. On the other hand, a randomized clinical trial of habitual coffee consumption is unlikely to ever be performed. Accordingly, data of this sort are the best that we have to guide recommendations for our patients.
Hill proposed nine characteristics of an observational trial that lend credence to ascribing causation to an observed association between an environmental factor and subsequent disease (sometimes referred to as the “Bradford Hill Criteria”).2 Relative to the study of Kim et al., causation would translate into protection and the most applicable of these characteristics would be biological plausibility, consistency, and biological gradient (ie, dose-response relationship). With respect to biological plausibility, the major, although not the only, pharmacologically active constituent of coffee is the methylxanthine caffeine. The direct effects of caffeine on arrhythmogenesis are complex with both potentially arrhythmic effects (including sympathetic activation) and potentially antiarrhythmic effects (including adenosine receptor blockade and antioxidant properties). When one also considers the indirect effects of caffeine and the possibility of effects of other active ingredients in coffee, it quickly becomes evident that prediction of the effects of habitual coffee ingestion on the propensity to tachyarrhythmias, in the general population or in a specific individual, is not possible. Accordingly, any association between coffee consumption and tachyarrhythmias is biologically plausible. With respect to consistency of results, other observational studies have reported that coffee consumption is associated with an increase in, a decrease in, or no change in the incidence of tachyarrhythmias. Nevertheless, a meta-analysis3 of six prospective studies involving 228,465 subjects published before 2014 concluded that habitual caffeine exposure was associated with a trend to reduction in incident atrial fibrillation (RR 0.90; 95% CI 0.81–1.01; p = 0.07) that narrowly missed statistical significance. These confidence intervals include the point estimate of the association between coffee intake and the risk of atrial fibrillation and/ or atrial flutter in the study of Kim et al. (HR 0.97; 95% CI 0.96–0.98, p < 0.001). Indeed, if the results of Kim et al. were included in a formal meta-analysis, the apparent association with a lower incidence of atrial fibrillation would likely be statistically significant. Accordingly, the results of the study of Kim et al. are consistent with previous studies. The presence of a biological gradient, or dose response, strengthens the inference of causality in an observational trial because, at least up to a point, the magnitude of an outcome associated with exposure to a truly causative environmental factor increases with the magnitude of the exposure. As indicated above, Kim et al. report that the inverse association between habitual coffee intake and incident tachyarrhythmias shows a dose-response relationship with each daily cup of coffee consumption being associated with 3% fewer tachyarrhythmias (HR, 0.97; 95% CI, 0.96–0.98). This observation is also consistent with those of the previous meta-analysis,3 which noted a decrease in incident atrial fibrillation of 6% (RR, 0.94; 95% CI, 0.90–0.99) for each 300 mg of habitual daily caffeine intake (equivalent to approximately 2–3 cups of coffee).
The study of Kim et al. adds a newer wrinkle to previous studies of this sort—Mendelian randomization analysis. Such an analysis is based on the concept that, if randomization to habitual coffee intake is not possible, then the next best thing would be to use another random process that determines coffee intake or another random process that determines caffeine exposure secondary to variations in caffeine metabolism. Genome-wide association studies have identified single-nucleotide polymorphisms (SNPs) associated with coffee intake and caffeine metabolism. Individuals genetically predisposed to higher levels of coffee intake or with slow metabolism of caffeine would be expected to have higher caffeine levels. Given that these SNPs are inherited randomly (Mendel’s Law of Segregation), provided that the SNPs are not associated with other confounders, nature provides a randomized trial of caffeine exposure that provides evidence of causation if the presence of the SNPs predicting higher caffeine exposure are associated with protection from cardiac tachyarrhythmias. Unfortunately for the hypothesis, the Mendelian randomization analysis reported by Kim et al. provided no evidence of causation. With respect to Hill’s characteristic of consistency, it is noteworthy that a previous report4 of a Mendelian randomization analysis of the association between coffee consumption and the risk of atrial fibrillation in a different subject population also found no evidence to support causation.
The patient population of Kim et al. included 1346 patients with known tachyarrhythmias at baseline – 0.3% of the overall sample. The analyses in this group of patients were comparable to the analyses in the overall group with each daily cup of coffee consumption being associated with 7% fewer tachyarrhythmias (HR, 0.93; 95% CI, 0.90-0.96).
How might we integrate these observations into clinical practice? Taken in conjunction with pre-existing literature, data of the report of Kim et al. do not convincingly support the inference that habitual coffee intake is causally related to a reduced probability of incident cardiac tachyarrhythmias. Thus, it would not be advisable for our patients to drink coffee for this purpose alone. Nevertheless, the probability that habitual coffee intake is causally related to an increased probability of cardiac tachyarrhythmias in the overall population is small. Accordingly, it seems unnecessary for our patients to avoid drinking coffee because of this concern. These conclusions also appear to be applicable to patients with known cardiac tachyarrhythmias, albeit with less certainty. In patients with known tachyarrhythmias, it would be appropriate to move away from the current practice of recommending avoidance of coffee or caffeine intake for all patients to reserving this recommendation for those patients who have recognized an association between their intake and tachyarrhythmia recurrences.