Some articles approve fantastic properties of coffee drinking.. others dissaprove intake and associate it with dementia & other ailments. I am a 76 yr old Cardiologist in general practice, and I consume 2 -3 cups of this brain energizer during the morning and sometimes on discussions with colleagues in the same coffee sharing. We are al around the same age. we are not pursuing to leave our most enjoyful drink in tne mornings, articles or no article... Respectfully. .
In reference to your February 16 Sensible Medicine post, “Coffee is great but it does not prevent dementia,” I would like to offer a brief semantic and epistemological reflection.
First, a confession of origin: I come from a latitude where some claim the world’s most aromatic coffee is grown. I will nonetheless attempt intellectual sobriety.
Semantically, in epidemiology, few exposures “prevent” disease in an absolute sense. They modify probabilities. Thus, stating that coffee does not prevent dementia is technically correct if prevention implies total risk abolition. But the relevant question is whether moderate consumption reduces baseline risk—and by how much.
The study by Zhang et al. (2021), using UK Biobank data, reported non-linear associations between moderate coffee (and tea) intake and lower risk of stroke and dementia, with lowest risk around 2–3 cups daily. The authors claimed association, not causation. Their methodological caution was clear.
It is also true that long-term randomized controlled trials (RCTs) showing that coffee prevents incident dementia do not exist. But neither do such RCTs for walking every morning. Chronic disease epidemiology rarely allows decades-long molecularly isolated interventions with clinical endpoints.
Here lies the deeper issue.
Much research in cardiovascular disease and dementia has been framed within linear pathogenic models—lipid accumulation, metabolic overflow, amyloid deposition. RCTs inherit these ontologies. Randomization reduces allocation bias, but it does not correct the conceptual framing of disease as a single-pathway excess requiring molecular subtraction.
Chronic disease, however, is systemic, adaptive, and contextual.
Umbrella reviews by Poole et al. (2017) and Grosso et al. (2017) showed consistent inverse associations between moderate coffee consumption and all-cause mortality, cardiovascular disease, type 2 diabetes, liver disease, Parkinson’s disease, and depression—greatest around 3–4 cups daily. More recent syntheses (e.g., Gill et al., 2024; Ungvari et al., 2024) continue to report favorable associations with cardiometabolic health and plausible mechanisms involving insulin sensitivity, endothelial function, and inflammation—while acknowledging predominantly observational evidence.
Consistency and biological plausibility appear present; magnitude of independent causality remains debated.
Perhaps, however, the debate is misframed.
From a salutogenic perspective, the morning cup is not merely caffeine and polyphenols. It is a ritual embedded in circadian timing, social interaction, mood regulation, and behavioral organization. Dementia—particularly vascular and Alzheimer-type—shares substrates with cardiometabolic deterioration, chronic inflammation, endothelial dysfunction, and circadian disruption. In that broader system, coffee may act less as an isolated neuroprotective agent and more as a behavioral marker within a coherent metabolic ecology.
It is not the cup alone. It is the system in which the cup is embedded.
Rather than asking whether coffee prevents dementia, perhaps the better question is: within what biological and social metabolism do we place that daily cup? There—in the organization of the system—is where prevention may truly reside.
References
Echeverry-Raad, J. (2024). A falsehood that has been repeated many times becomes true: The origin of the diabesity pandemic, the most lethal of the 21st century? Journal of Diabetes, Metabolic Disorders & Control, 11(1), 39–50. https://doi.org/10.15406/jdmdc.2024.11.00276
Gill, D., et al. (2024). An umbrella review of meta-analyses to understand the effect of coffee consumption and the relationship between stroke, cardiovascular heart disease, and dementia. [Publication details pending].
Grosso, G., Godos, J., Galvano, F., & Giovannucci, E. L. (2017). Coffee, caffeine, and health outcomes: An umbrella review. Annual Review of Nutrition, 37, 131–156. https://doi.org/10.1146/annurev-nutr-071816-064941
Poole, R., Kennedy, O. J., Roderick, P., Fallowfield, J. A., Hayes, P. C., & Parkes, J. (2017). Coffee consumption and health: Umbrella review of meta-analyses of multiple health outcomes. BMJ, 359, j5024. https://doi.org/10.1136/bmj.j5024
Ungvari, Z., et al. (2024). Coffee consumption and cardiometabolic health. [Publication details pending].
Zhang, Y., Yang, H., Li, S., Li, W. D., & Wang, Y. (2021). Consumption of coffee and tea and risk of developing stroke, dementia, and poststroke dementia: A cohort study in the UK Biobank. PLoS Medicine, 18(11), e1003830. https://doi.org/10.1371/journal.pmed.1003830
Mandrola nails the core problem here: an observational association is being marketed as a preventive intervention, and the confidence of the coverage is wildly out of proportion to what the design can support. 
A few things I’d underline as a physician-scientist:
1. Your Bayesian prior should be low. Dementia is a heterogeneous syndrome (vascular, neurodegenerative, inflammatory, toxic, genetic), so the idea that one exposure like coffee would cleanly “prevent dementia” should trigger skepticism before you even look at a p-value. 
2. Adjustment is not randomization. You can only adjust for what’s captured, and “coffee drinker” is a cluster label for countless health, socioeconomic, sleep, work-pattern, and behavioral differences that don’t fit neatly in a spreadsheet. 
3. Effect-size + subgroup results are a tell. When a simple habit shows an ~18% lower dementia risk overall and looks even “better” in age strata, that’s often the signature of residual confounding/selection effects rather than biology doing something magical. 4. Analytic flexibility is the quiet elephant. With enough plausible modeling choices, you can often “find” a publishable association, especially in nutrition/behavior epidemiology, without anyone committing fraud. 
Enjoy coffee if you enjoy coffee. But if we’re serious about dementia prevention, we should keep the spotlight on interventions with stronger causal footing (BP control, physical activity, sleep, hearing, metabolic risk, smoking, education/cognitive engagement), not another headline that confuses correlation with protection.
All good with doubting observational data. If only people would do it consistently (ahemmm...vaccines...). But the claimed effect here with coffee is entirely plausible. This is what happens when you go to medical school and don't understand pleiotropy of natural medicines and experience supplements as an alternate dimension. No offense, but it's true.
A reflexive negative reaction to observational studies is itself a cognitive bias.
Observational and epidemiological studies are perfectly good puzzle pieces. The fixation on blinded controlled studies is not helpful. Those are great too, and in some ways better. But they can never cover the same size population or duration. All forms of evidence are imperfect.
In this particular case, the protective effects of caffeine have been seen in multiple observational studies over the years, enough that it should not be ignored. If there were no mechanism explaining why this effect would work, it would be a little shaky (though still valid-though-imperfect evidence). But there is an excellent, well supported, mechanism.
Studies demonstrate that, in mice caffeine and Eicosanoyl-5-hydroxytryptamide (EHT), which is found in coffee beans work together to improve activity of the PP2A catalyst, which protects the brain against abnormal protein accumulation. Studies in mice showed that the combination of EHT and caffeine reduced neuroinflammation and protected against oxidative stress, and manifested in actual effects in behavior.
"But that's just mice, not people."
No. When you have an observational study, a very plausible mechanism, and demonstrated efficacy in mice, that adds up to solid evidence. Is it enough to approve a drug? Maybe not. But is it enough to make it reasonable for people to take it seriously? Yes.
"Good enough for drug approval" is not the same thing as "good enough to incorporate into your personal practices."
This will inevitably be used frequently on an individual level (not saying JAMA or authors are suggesting this) to justify unhealthy high calorie coffee consumption with abundantly added sugar and creamer.
Unfortunately again, this article missed the point. He states that drinking coffee DOES NOT PREVENT DEMENTIA. He might have an argument that research about drinking coffee does not absolutely PROVE any benefits, but clearly, the research does NOT DISPROVE that coffee might provide a benefit to reduce dementia. In fact, the research indicates that coffee may be one of many factors involved in brain health, along with sunshine, avoiding sugar, avoiding statins, natural foods and drinks, exercise, and social interaction. Please stop being biased.
This sort of analysis can be applied to all epidemiologic and observational studies of the impact of "lifestyle factors" on health and disease. Unfortunately, the popularity of AI and the further construction of enormous data centers will probably assure that we will get a lot more of it. Take away number 1 defines the purpose and utility of Sensible Medicine and related websites. I suspect that numbers 2&3 were listed with a little tongue-in-cheek. Encouraging researchers "to resist the urge to perform these studies especially with taxpayer money from the NIH" is unlikely to accomplish much. Abolition of the NIH and all other agencies of the government that fund research would likely be much more effective. At least the promoters that are seeking "positive" results for their products would be more readily identified rather than just the academic institutions and researchers that are often bought and paid for---often in devious ways that may not be apparent in their declarations of conflicts of interest.
Amen! Thank you John! How did we get this far given the increased reliance on evidence based medicine. Do you see the day when AI can do an analysis like you did, so that even a practicing physician could understand the faults of a study? I have to rely on you..so thank you and keep at it.
What about the DECAF trial that randomized 200 patients after cardioversion for AF that showed less recurrence of AF in patients who consumed 1 cup of coffee a day, compared to patients who abstained? Is that study just as flawed as this one? Should DECAF be used to give patients with recurrent paroxsymal AF permission to have one cup a day?
I enjoy my morning coffee, black, and sometimes a little later in the work day. However, I am not convinced it has any impact on… I’m sorry, I can’t recall what was this Substack article was about.
To be fair, the first author is a PhD student at Harvard. As a trainee myself, I know that you learn a lot about how to do science by doing a study like this. Why should the NIH fund this? Because training the next generation of scientists is important. Everyone starts somewhere.
Now, should it have been published in JAMA and picked up and over-hyped by the media everywhere? That is a separate issue...
Some articles approve fantastic properties of coffee drinking.. others dissaprove intake and associate it with dementia & other ailments. I am a 76 yr old Cardiologist in general practice, and I consume 2 -3 cups of this brain energizer during the morning and sometimes on discussions with colleagues in the same coffee sharing. We are al around the same age. we are not pursuing to leave our most enjoyful drink in tne mornings, articles or no article... Respectfully. .
Dr. Mandrola,
In reference to your February 16 Sensible Medicine post, “Coffee is great but it does not prevent dementia,” I would like to offer a brief semantic and epistemological reflection.
First, a confession of origin: I come from a latitude where some claim the world’s most aromatic coffee is grown. I will nonetheless attempt intellectual sobriety.
Semantically, in epidemiology, few exposures “prevent” disease in an absolute sense. They modify probabilities. Thus, stating that coffee does not prevent dementia is technically correct if prevention implies total risk abolition. But the relevant question is whether moderate consumption reduces baseline risk—and by how much.
The study by Zhang et al. (2021), using UK Biobank data, reported non-linear associations between moderate coffee (and tea) intake and lower risk of stroke and dementia, with lowest risk around 2–3 cups daily. The authors claimed association, not causation. Their methodological caution was clear.
It is also true that long-term randomized controlled trials (RCTs) showing that coffee prevents incident dementia do not exist. But neither do such RCTs for walking every morning. Chronic disease epidemiology rarely allows decades-long molecularly isolated interventions with clinical endpoints.
Here lies the deeper issue.
Much research in cardiovascular disease and dementia has been framed within linear pathogenic models—lipid accumulation, metabolic overflow, amyloid deposition. RCTs inherit these ontologies. Randomization reduces allocation bias, but it does not correct the conceptual framing of disease as a single-pathway excess requiring molecular subtraction.
Chronic disease, however, is systemic, adaptive, and contextual.
Umbrella reviews by Poole et al. (2017) and Grosso et al. (2017) showed consistent inverse associations between moderate coffee consumption and all-cause mortality, cardiovascular disease, type 2 diabetes, liver disease, Parkinson’s disease, and depression—greatest around 3–4 cups daily. More recent syntheses (e.g., Gill et al., 2024; Ungvari et al., 2024) continue to report favorable associations with cardiometabolic health and plausible mechanisms involving insulin sensitivity, endothelial function, and inflammation—while acknowledging predominantly observational evidence.
Consistency and biological plausibility appear present; magnitude of independent causality remains debated.
Perhaps, however, the debate is misframed.
From a salutogenic perspective, the morning cup is not merely caffeine and polyphenols. It is a ritual embedded in circadian timing, social interaction, mood regulation, and behavioral organization. Dementia—particularly vascular and Alzheimer-type—shares substrates with cardiometabolic deterioration, chronic inflammation, endothelial dysfunction, and circadian disruption. In that broader system, coffee may act less as an isolated neuroprotective agent and more as a behavioral marker within a coherent metabolic ecology.
It is not the cup alone. It is the system in which the cup is embedded.
Rather than asking whether coffee prevents dementia, perhaps the better question is: within what biological and social metabolism do we place that daily cup? There—in the organization of the system—is where prevention may truly reside.
References
Echeverry-Raad, J. (2024). A falsehood that has been repeated many times becomes true: The origin of the diabesity pandemic, the most lethal of the 21st century? Journal of Diabetes, Metabolic Disorders & Control, 11(1), 39–50. https://doi.org/10.15406/jdmdc.2024.11.00276
Gill, D., et al. (2024). An umbrella review of meta-analyses to understand the effect of coffee consumption and the relationship between stroke, cardiovascular heart disease, and dementia. [Publication details pending].
Grosso, G., Godos, J., Galvano, F., & Giovannucci, E. L. (2017). Coffee, caffeine, and health outcomes: An umbrella review. Annual Review of Nutrition, 37, 131–156. https://doi.org/10.1146/annurev-nutr-071816-064941
Poole, R., Kennedy, O. J., Roderick, P., Fallowfield, J. A., Hayes, P. C., & Parkes, J. (2017). Coffee consumption and health: Umbrella review of meta-analyses of multiple health outcomes. BMJ, 359, j5024. https://doi.org/10.1136/bmj.j5024
Ungvari, Z., et al. (2024). Coffee consumption and cardiometabolic health. [Publication details pending].
Zhang, Y., Yang, H., Li, S., Li, W. D., & Wang, Y. (2021). Consumption of coffee and tea and risk of developing stroke, dementia, and poststroke dementia: A cohort study in the UK Biobank. PLoS Medicine, 18(11), e1003830. https://doi.org/10.1371/journal.pmed.1003830
Mandrola nails the core problem here: an observational association is being marketed as a preventive intervention, and the confidence of the coverage is wildly out of proportion to what the design can support. 
A few things I’d underline as a physician-scientist:
1. Your Bayesian prior should be low. Dementia is a heterogeneous syndrome (vascular, neurodegenerative, inflammatory, toxic, genetic), so the idea that one exposure like coffee would cleanly “prevent dementia” should trigger skepticism before you even look at a p-value. 
2. Adjustment is not randomization. You can only adjust for what’s captured, and “coffee drinker” is a cluster label for countless health, socioeconomic, sleep, work-pattern, and behavioral differences that don’t fit neatly in a spreadsheet. 
3. Effect-size + subgroup results are a tell. When a simple habit shows an ~18% lower dementia risk overall and looks even “better” in age strata, that’s often the signature of residual confounding/selection effects rather than biology doing something magical. 4. Analytic flexibility is the quiet elephant. With enough plausible modeling choices, you can often “find” a publishable association, especially in nutrition/behavior epidemiology, without anyone committing fraud. 
Enjoy coffee if you enjoy coffee. But if we’re serious about dementia prevention, we should keep the spotlight on interventions with stronger causal footing (BP control, physical activity, sleep, hearing, metabolic risk, smoking, education/cognitive engagement), not another headline that confuses correlation with protection.
All good with doubting observational data. If only people would do it consistently (ahemmm...vaccines...). But the claimed effect here with coffee is entirely plausible. This is what happens when you go to medical school and don't understand pleiotropy of natural medicines and experience supplements as an alternate dimension. No offense, but it's true.
A reflexive negative reaction to observational studies is itself a cognitive bias.
Observational and epidemiological studies are perfectly good puzzle pieces. The fixation on blinded controlled studies is not helpful. Those are great too, and in some ways better. But they can never cover the same size population or duration. All forms of evidence are imperfect.
In this particular case, the protective effects of caffeine have been seen in multiple observational studies over the years, enough that it should not be ignored. If there were no mechanism explaining why this effect would work, it would be a little shaky (though still valid-though-imperfect evidence). But there is an excellent, well supported, mechanism.
Studies demonstrate that, in mice caffeine and Eicosanoyl-5-hydroxytryptamide (EHT), which is found in coffee beans work together to improve activity of the PP2A catalyst, which protects the brain against abnormal protein accumulation. Studies in mice showed that the combination of EHT and caffeine reduced neuroinflammation and protected against oxidative stress, and manifested in actual effects in behavior.
"But that's just mice, not people."
No. When you have an observational study, a very plausible mechanism, and demonstrated efficacy in mice, that adds up to solid evidence. Is it enough to approve a drug? Maybe not. But is it enough to make it reasonable for people to take it seriously? Yes.
"Good enough for drug approval" is not the same thing as "good enough to incorporate into your personal practices."
But just to be safe....I'll still keep drinking my coffee in the morning 🤣-
This will inevitably be used frequently on an individual level (not saying JAMA or authors are suggesting this) to justify unhealthy high calorie coffee consumption with abundantly added sugar and creamer.
Yes-posts such as these are exactly why I’m a subscriber! Thanks Dr. Mandrola.
i enjoyed ready this commentary with my morning cup of joe. felt smarter in 2 ways
A study funded by the Coffee and Tea Marketing association? Come on guys do some research as to who paid for this fake science.
Unfortunately again, this article missed the point. He states that drinking coffee DOES NOT PREVENT DEMENTIA. He might have an argument that research about drinking coffee does not absolutely PROVE any benefits, but clearly, the research does NOT DISPROVE that coffee might provide a benefit to reduce dementia. In fact, the research indicates that coffee may be one of many factors involved in brain health, along with sunshine, avoiding sugar, avoiding statins, natural foods and drinks, exercise, and social interaction. Please stop being biased.
This sort of analysis can be applied to all epidemiologic and observational studies of the impact of "lifestyle factors" on health and disease. Unfortunately, the popularity of AI and the further construction of enormous data centers will probably assure that we will get a lot more of it. Take away number 1 defines the purpose and utility of Sensible Medicine and related websites. I suspect that numbers 2&3 were listed with a little tongue-in-cheek. Encouraging researchers "to resist the urge to perform these studies especially with taxpayer money from the NIH" is unlikely to accomplish much. Abolition of the NIH and all other agencies of the government that fund research would likely be much more effective. At least the promoters that are seeking "positive" results for their products would be more readily identified rather than just the academic institutions and researchers that are often bought and paid for---often in devious ways that may not be apparent in their declarations of conflicts of interest.
Amen! Thank you John! How did we get this far given the increased reliance on evidence based medicine. Do you see the day when AI can do an analysis like you did, so that even a practicing physician could understand the faults of a study? I have to rely on you..so thank you and keep at it.
What about the DECAF trial that randomized 200 patients after cardioversion for AF that showed less recurrence of AF in patients who consumed 1 cup of coffee a day, compared to patients who abstained? Is that study just as flawed as this one? Should DECAF be used to give patients with recurrent paroxsymal AF permission to have one cup a day?
I enjoy my morning coffee, black, and sometimes a little later in the work day. However, I am not convinced it has any impact on… I’m sorry, I can’t recall what was this Substack article was about.
To be fair, the first author is a PhD student at Harvard. As a trainee myself, I know that you learn a lot about how to do science by doing a study like this. Why should the NIH fund this? Because training the next generation of scientists is important. Everyone starts somewhere.
Now, should it have been published in JAMA and picked up and over-hyped by the media everywhere? That is a separate issue...