To the Editor We read with interest the study by Loftfield and colleagues1 in which they found an inverse association between coffee consumption and mortality, regardless of genetic caffeine metabolism score.
We noticed that conventional Cox models were used to estimate the hazard ratios (95% CIs) of both all-cause and cause-specific mortality between groups.1 However, we wonder if this indiscriminate use might not have performed well in the hazard ratio estimation of the latter, because the true absolute risks of cause-specific mortality were overestimated owing to the presence of competing risks. A similar issue has been highlighted in previous literature.2,3