All the study claimed was an „association“, nothing more.
> We used Cox proportional hazards models to estimate the associations between coffee/tea consumption and incident stroke and dementia, adjusting for sex, age, ethnicity, qualification, income, body mass index (BMI), physical activity, alcohol status, smoking status, diet pattern, consumption of sugar-sweetened beverages, high-density lipoprotein (HDL), low-density lipoprotein (LDL), history of cancer, history of diabetes, history of cardiovascular arterial disease (CAD), and hypertension.
It's extremely difficult to do this in practice because people lie, there are other factors they might have missed and so on.
Typically the easiest way to turn it around is to switch where the authors chose to place the word "risk" and see if the new statement (with some appropriate changes, and adding "predisposition" to account for future events) still makes sense.
E.g. Predisposition to dementia lowers lifetime "risk " of coffee consumption.
For example, while that would mean that simply increasing coffee consumption wouldn't directly lower your risk of dementia and stroke, we can improve our assessment of the risk of a certain patient developing these conditions by asking whether they consume coffee. Even though the coffee isn't the cause, since it correlates with the cause, we still get information by doing that.
> Covariates
> In the present study, the selection of covariates based on (1) demographic variables, including sex, age, ethnicity background, education level, and income; and (2) a priori knowledge of potential confounding factors associated with incident stroke and dementia [30,31]. Covariates were documented including sex, age, ethnicity (White, Asian or Asian British, Black or Black British, and Other ethnic group), qualification (college or university degree, Advanced [A] levels/Advanced Subsidiary [AS] levels or equivalent, Ordinary [O] levels/General Certificate of Secondary Education [GCSE] or equivalent, Certificate of Secondary Education [CSE] or equivalent, National Vocational Qualification [NVQ] or Higher National Diploma [HND] or Higher National Certificate [HNC] or equivalent, other professional qualifications, or none of the above), income (less than £18,000, 18,000 to 30,999, 31,000 to 51,999, 52,000 to 100,000, and greater than 100,000), BMI (<25, 25 to <30, 30 to <35, and ≥35 kg/m2), smoking status (never, former, and current), alcohol status (never, former, and current), physical activity (low, moderate, and high), consumption of sugar-sweetened beverages, history of diabetes, history of coronary artery disease, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and diet pattern (healthy and unhealthy, healthy diet was based on consumption of at least 4 of 7 dietary components: (1) fruits: ≥3 servings/day; (2) vegetables: ≥3 servings/day; (3) fish: ≥2 servings/week; (4) processed meats: ≤1 serving/week; (5) unprocessed red meats: ≤1.5 servings/week; (6) whole grains: ≥3 servings/day; (7) refined grains: ≤1.5 servings/day [32–35]) (S1 Table).
> Information on cardiovascular arterial disease (CAD) was derived from medical records (ICD-10 codes I20 to I25). Diabetes was ascertained on the basis of medical records (ICD-10 codes E10 to E14), glycated hemoglobin ≥6.5%, and the use of antidiabetic drugs. Hypertension was defined as systolic blood pressure (SBP) ≥140 mm Hg or diastolic blood pressure (DBP) ≥90 mm Hg, use of antihypertension agents, or medical records (ICD-10 codes I10 to I13 and I15). Cancer was identified through linkage to the National Health Service (NHS) Central Register (ICD-10 codes C00 to C97).
That thought crossed my mind, actually.
I have a genetic disorder. A lot of my food preferences turn out to be "medicinal" so to speak for that condition.
[1] https://www.hsph.harvard.edu/news/press-releases/genetic-var...
Discovered the gene thanks to consumer genetic testing, and the heart rate thing with a fitbit. I'm more careful these days.