Coffee intake was self-reported and grouped into rough consumption categories. People with severe mental illness often have uneven lifestyles and medication profiles, so residual confounding is substantial. The models adjust for some variables, but unmeasured factors such as sleep, socioeconomic circumstances, smoking patterns, and medication effects could readily produce the pattern seen. The shape of the association (“up to about two cups a day”) is typical of non-linear confounding in observational nutrition research, yet the article treats this as a plausible biological threshold.
Quotes such as “shows that moderate coffee consumption was associated with slower biological ageing” suggest a degree of mechanistic insight that the study cannot provide. Nothing in the design tests whether coffee causes any change in the underlying biology. No intervention was performed, and the cohorts were not designed to explore caffeine metabolism, brew type, or the many additives that accompany coffee drinking.
The framing around mental illness implies a specific benefit in this group, but the evidence only shows a statistical association in a subset of observational datasets. The article does not mention that biological age algorithms differ in what they measure, can disagree with one another, and often reflect current health status rather than ageing processes. It also omits that the confidence intervals around subgroup effects may be wide, especially when stratifying by diagnosis and consumption band.
Overall, the data are narrow: observational, self-reported exposure, proxy biomarkers, multiple potential confounders. The article overstates the finding and treats a modest association as evidence of a limit to coffee’s “helpfulness”, when the study cannot define such a threshold or establish causation.