Coffee and liver cancer: a lower-risk association with very-low-certainty evidence
A dose-response meta-analysis separated its striking estimate from the limitations of observational evidence.

Illustrative food image, not a photograph of the study intervention.
One liver-cancer endpoint, two observational designs
The review examined coffee consumption and hepatocellular carcinoma, not all cancers combined. It included 18 cohorts with more than 2.27 million participants and 2,905 cases, plus eight case-control studies containing 1,825 cases and 4,652 controls. Investigators used dose-response meta-analysis and evaluated evidence quality with the GRADE framework. No randomized trials supported the association. The large participant count is important context, but it should not be mistaken for a large randomized coffee experiment. The evidence combined different observational approaches to a specific cancer-incidence question.
Kennedy OJ, Roderick P, Buchanan R, Fallowfield JA, et al.. BMJ open, 2017. PMID 28490552 · DOI
The pooled estimate changed with study design
An extra two cups of coffee daily was associated with a pooled relative risk of 0.65 for hepatocellular carcinoma, with a 95% confidence interval from 0.59 to 0.72. The inverse association was weaker in cohort studies, at 0.71, than in case-control studies, at 0.53. The authors generally rated the cohorts as higher quality. That difference belongs alongside the headline estimate. It discourages treating the strongest pooled association as a settled causal effect and shows why the design of contributing studies matters as much as their combined sample size.
Kennedy OJ, Roderick P, Buchanan R, Fallowfield JA, et al.. BMJ open, 2017. PMID 28490552 · DOI
Decaffeinated coffee had thinner evidence
The caffeinated-coffee estimate drew on two cohort studies, while the decaffeinated estimate drew on three. For an extra two cups daily, the respective relative risks were 0.73 and 0.86. The decaffeinated confidence interval extended to 1.00, making its result less decisive than a claim of proven protection would suggest. These limited subgroup data do not establish that caffeine is the responsible ingredient or that removing caffeine preserves an identical effect. They support describing separate observational estimates with their own uncertainty rather than presenting every coffee preparation as an equivalent preventive treatment.
Kennedy OJ, Roderick P, Buchanan R, Fallowfield JA, et al.. BMJ open, 2017. PMID 28490552 · DOI
The certainty rating is part of the finding
The authors rated the evidence very low in quality because of the absence of randomized trials, possible publication bias, and the lack of an accepted coffee definition. Those limitations must remain visible even when the pooled estimate looks substantial. The study does not prescribe coffee for cancer prevention or establish benefit for a particular brewing method. The linked breakfast is merely a meal that could accompany a beverage; it contains no tested coffee intervention. Its inclusion carries no liver-cancer claim. The evidence concerns observed coffee consumption and hepatocellular carcinoma, not a protective breakfast recipe.
Kennedy OJ, Roderick P, Buchanan R, Fallowfield JA, et al.. BMJ open, 2017. PMID 28490552 · DOI