06/05/2026
Caffeine metabolism is something that shows up daily in my practice. I have come across many people who are "slow metabolisers" and as suggested in this article, they demonstrate certain symptoms, including early insomnia, increased anxiety and sympathetic NS arousal. Personally, if I drink caffeine after 11am, I am still awake at 2am!
Every few years a meta-analysis comes out claiming coffee is either protective or harmful for cardiovascular disease, and the headlines treat the result as universal. It is not. Coffee's effect on your body depends on a single genetic variant in a single enzyme, and about half the population carries the slow version.
CYP1A2 is the liver enzyme responsible for metabolizing roughly 95% of the caffeine you ingest. A single nucleotide polymorphism called rs762551 (also written as -163C>A) changes how inducible the enzyme is. Individuals homozygous for the A allele (genotype AA) produce a highly inducible form that clears caffeine quickly, with a plasma half-life around 3 hours. Individuals carrying one or two C alleles (AC or CC) have reduced inducibility, with half-lives of 6 to 10 hours. In population data from people of European descent, roughly 45% are AA (fast), 44% are AC (intermediate/slow), and 11% are CC (slow). About 55% carry at least one slow allele.
The clinical consequences are not subtle. Cornelis and colleagues (JAMA, 2006) conducted a case-control study in Costa Rica with 2,014 cases of first nonfatal myocardial infarction and 2,014 matched controls. Among slow metabolizers, drinking 4 or more cups of coffee per day was associated with an odds ratio of 1.64 for nonfatal MI (95% CI, 1.14 to 2.34). Among fast metabolizers, the same intake produced an odds ratio of 0.99 (0.66 to 1.48), essentially no change. The gene-coffee interaction was statistically significant (p = 0.04). In participants younger than 59, the effect amplified. Slow metabolizers drinking 4 or more cups showed an OR of 2.33 (1.39 to 3.89). More than double the risk.
Palatini and colleagues (Journal of Hypertension, 2009) followed 553 young Italian adults screened for stage 1 hypertension over a median of 8.2 years. Among slow metabolizers, heavy coffee consumption tripled the hazard of developing physician-diagnosed hypertension (HR 3.00, 95% CI 1.53 to 5.90). Among fast metabolizers, heavy coffee was protective (HR 0.36, 0.14 to 0.89). The gene-coffee interaction on blood pressure was highly significant. Urinary epinephrine was elevated only in slow metabolizers who drank coffee. Same beverage, opposite cardiovascular signal.
Guest and colleagues (Medicine & Science in Sports & Exercise, 2018) took this into athletic performance. In a randomized, double-blind, placebo-controlled crossover trial of 101 competitive male athletes performing 10-km cycling time trials, the interaction pattern repeated. At 4 mg per kilogram of body weight (roughly 300 mg for a 75-kg athlete), AA genotype carriers improved cycling time by 6.8%. CC genotype carriers got 13.7% slower. AC heterozygotes showed no effect in either direction, which is worth noting. Cornelis and Palatini grouped AC and CC together as slow carriers. Guest distinguished them and found the ergolytic response specifically in CC homozygotes. The caffeine-gene interaction was significant at p less than 0.0001.
One honest caveat belongs here. The hypertension finding has not replicated cleanly across every population. A large Taiwan Biobank analysis of over 19,000 participants found coffee protective for AC and CC genotypes in that cohort. Population-specific differences in diet, smoking history, coffee preparation, and linkage with other haplotypes likely explain part of the variance. The MI and performance data have been more consistent. The overall principle, that CYP1A2 genotype substantially modifies the response to caffeine, is not seriously disputed in the pharmacogenomics literature. What gets disputed is the magnitude and direction of each specific outcome in each specific population.
This is the piece nutrition headlines routinely miss. When a new meta-analysis reports that coffee reduces cardiovascular mortality in a pooled cohort, the pooling averages across genotypes showing opposite effects. The published summary statistic is a population-weighted compromise between two genuinely different biological responses. Your personal response is not the average. Your personal response is AA or AC or CC.
You do not need a genetic test to start assessing your phenotype. Caffeine clearance shows up in everyday experience. If a 2pm coffee still affects your sleep at 11pm, your body is clearing it slowly. If you drink coffee at 6pm and sleep soundly by 10pm, you are clearing it quickly. A week of structured notes on dose, timing, and sleep quality will tell you more about your individual response than any one-time lab result. Genotyping for rs762551 is available through clinical pharmacogenomic panels for those who want the genetic answer, but the functional question is answerable today with a notepad.
The question is not whether coffee is good for you. The question is which half of the population you are in, and nobody has told you.
Cornelis et al., JAMA, 2006
Palatini et al., J Hypertens, 2009
Guest et al., Med Sci Sports Exerc, 2018
