Heart. The review inserted in the 2020 collection reports cardiovascular mortality reduced by about 19% in habitual consumers, with a maximum around three cups a day. Stroke mortality is reduced by 30% there.

Poole and his coauthors published in 2017 in the BMJ an umbrella review, that is, a review that does not reread the studies but the meta-analyses already done on them. This one compiles 201, all observational, covering 67 health outcomes, plus 17 meta-analyses of trials. It finds the same shape.

The risk of death from all causes is lowest at about three cups a day there, and three to four cups are associated with less coronary disease than abstinence. What matters here is not the figure, it is the shape of the curve. A relationship that rises and then falls again is harder to manufacture by a simple bias than one that always rises. It demonstrates nothing by itself.

Blood pressure, for its part, resists. The authors of the collection write that the results are contradictory: a single dose raises pressure in the non-consumer, little or not at all in the habitual drinker. Beyond these two points, the literature does not converge.

MDPI 2020, pp. 117-119, 209-210; Poole et al., Coffee consumption and health: umbrella review of meta-analyses of multiple health outcomes, BMJ 359 (2017)

Metabolism and liver. The meta-analyses cited give about 25% lower risk of type 2 diabetes for three to four cups a day compared with fewer than two, and 6% per additional cup, an effect more marked in women; no one has shown that coffee prevents diabetes, and nothing suggests that it aggravates it. The liver is the organ where the associations are strongest and most consistent in Poole's review: less cirrhosis, chronic liver disease and liver cancer among drinkers, and decaf takes part in it. Gout and kidney stones also figure among the outcomes associated with a lower risk. All these sentences describe associations.

MDPI 2020, p. 209; Poole et al., BMJ 2017; LC ch. 19

Brain. The collection reports a possible association between moderate consumption and less cognitive decline, a lower risk of Parkinson's and Alzheimer's, mechanism not demonstrated. On Parkinson's, the meta-analyses are more substantial: Qi and Li (2014) find a relative risk of 0.72 at about three cups a day after adjustment for tobacco, and Hong and coauthors (Nutrients, 2020) a risk ratio of 0.80 in regular consumers, with a dose effect mainly in men.

The methodological problem is more severe here than elsewhere: these diseases develop over decades, consumption is measured once by questionnaire, and the first damage changes habits well before diagnosis. The A2A receptor, target of the antagonists tested against Parkinson's, gives at least one biological lead.

MDPI 2020, pp. 42-45, 209; Qi & Li, Geriatr. Gerontol. Int. 14 (2014); Hong et al., Nutrients 12 (2020) 1860; Jacobson et al., Br. J. Pharmacol. 2022

Cancer. The collection's chapter on colorectal cancer relies on a meta-analysis of 26 prospective studies and 3.3 million subjects. It finds no overall association, neither with colorectal cancer nor with rectal cancer. On the colon alone, it finds a modest but significant protective effect, more marked in European men and Asian women.

A detail that general-audience articles omit nevertheless reverses the reading. The protective effect appears in decaf drinkers, not in those who drink caffeinated. If an effect exists, caffeine is therefore probably not its driver.

The reverse cannot be concluded either. Decaf drinkers are not a sample like the others: one often comes to it on medical advice, and the confounding factor described above takes over again. The authors' conclusion: the data are not enough to demonstrate a protective role.

The institutional verdict has moved in the same direction. The International Agency for Research on Cancer had classified coffee as “possibly carcinogenic”, group 2B, in 1991. It reclassified it in June 2016 in group 3, “unclassifiable”, after reviewing more than a thousand studies (Monograph 116).

Group 3 does not mean harmless. It means that the data do not allow a decision in either direction. It is not an acquittal of coffee, it is a dismissal of the case. And the same monograph places beverages drunk above 65 °C in group 2A, “probably carcinogenic” for the esophagus. It is the temperature that receives the verdict, not the drink.

MDPI 2020, pp. 269-274; IARC, press release 244 of 06/15/2016 and Monograph 116, Drinking coffee, mate, and very hot beverages (2018); LC ch. 19