Coffee is the one item on the "is it healthy" list where the honest answer turns out to be mostly reassuring, and that is exactly why people distrust it. Headlines have swung between cancer risk and life extension for forty years, so any confident claim now arrives with a fair suspicion that something has been left out.
Something usually has. Most coffee articles quote a population average and stop, which hides the two variables that matter most to an individual drinker: how quickly you clear caffeine, and whether your coffee passed through a paper filter.
What does the population evidence actually say about coffee
The most complete summary is the 2017 umbrella review in The BMJ, which pooled 201 meta-analyses across 67 health outcomes. The largest risk reductions clustered at three to four cups a day compared with none, including all-cause mortality (relative risk 0.83, 95% CI 0.79 to 0.88) and cardiovascular disease (0.85). Coffee at usual intakes, the authors concluded, is more likely to benefit health than harm it (Poole et al., BMJ 2017).
It is still observational, and it carries a specific weakness. People who feel terrible on coffee quit, and people who become ill often cut back, so the non-drinker group mixes the caffeine-intolerant, the recently unwell and the never-interested.
UK Biobank shows the consequence. Across 347,077 people, compared with one to two cups a day, the odds of cardiovascular disease were 11% higher in non-drinkers, 7% higher in decaffeinated coffee drinkers and 22% higher above six cups (Zhou and Hyppönen, AJCN 2019). Non-drinkers on the wrong end of the curve is what you would expect if quitting tracks something other than coffee.
How much caffeine is in a cup and where does the ceiling sit
The cleanest European anchor is the European Food Safety Authority, whose 2015 opinion concluded that habitual intakes up to 400 mg of caffeine a day, and single doses up to 200 mg, raise no safety concern for healthy non-pregnant adults (EFSA, 2015).
Translating that into cups is where most people slip. A standard 8 oz brewed coffee holds roughly 95 mg of caffeine, while a 16 oz takeaway cup can reach around 330 mg on its own (Harvard Nutrition Source). Four home mugs and two large to-go cups get described the same way and are not the same exposure. And 400 mg is a ceiling below which no safety signal appears, not a target.
Does the way you brew your coffee change your cholesterol
Coffee beans contain two diterpenes, cafestol and kahweol, which raise LDL cholesterol. They are oil-soluble and they do not pass through paper. A Swedish analysis published in 2025 measured them across brewing methods and found a spread of roughly a hundredfold: paper-filtered home coffee held a median 12 mg/L of cafestol, workplace brewing machines 176 mg/L, and boiled coffee 939 mg/L, dropping to 28 mg/L once the same brew went through a fabric filter (Orrje et al., Nutr Metab Cardiovasc Dis 2025).
The effect shows up downstream. In the Tromsø Study, three to five espressos a day were associated with higher total cholesterol than none, by 0.16 mmol/L in men and 0.09 mmol/L in women, and six or more cups of boiled or plunger coffee with increases of 0.23 and 0.30 mmol/L (Svatun et al., Open Heart 2022). A randomised trial found one litre of French press coffee daily for two weeks also raised fasting homocysteine by 10% (Grubben et al., AJCN 2000).
None of that makes espresso dangerous. It makes brewing method a dose-dependent lever on LDL cholesterol and ApoB, and one of the few you can pull without changing anything else. The same logic runs through the eggs and cholesterol question: an average says nothing about which individuals respond.
What does caffeine do to cortisol blood pressure and sleep
Caffeine raises cortisol. In a placebo-controlled crossover trial, repeated doses lifted cortisol across the whole test day regardless of sex or stressor, and caffeine taken before mental stress pushed the stress response higher still (Lovallo et al., 2006). Daily drinkers develop partial tolerance, not complete tolerance, which is one reason a morning cortisol measurement needs context about what you drank and when.
Blood pressure follows a similar curve. The acute pressor effect lasts up to about three hours and is markedly blunted in habitual drinkers, and a 2023 systematic review in hypertensive adults found no support for a link between longer-term coffee drinking and elevated blood pressure or cardiovascular events in that group (AJCN 2023).
Sleep is where most real-world harm happens. In a double-blind trial, 400 mg of caffeine taken at bedtime, three hours before bed and six hours before bed all significantly disturbed sleep against placebo (Drake et al., J Clin Sleep Med 2013). Half-life averages four to five hours, with reported individual values from roughly 1.5 to more than 9 hours, so a slow clearer having a 4 pm espresso still carries a meaningful dose at midnight. The tiredness that follows gets blamed on the coffee, not the timing.
Where does coffee look protective rather than harmful
Liver markers are the clearest case. Coffee intake is inversely associated with ALT and GGT, and a meta-analysis of 16 studies found lower odds of advanced hepatic fibrosis in coffee drinkers (OR 0.73, 95% CI 0.58 to 0.92) and of cirrhosis at high intakes versus none (OR 0.53) (Liu et al., PLOS One 2015).
Glucose metabolism is stranger, because short-term and long-term data point opposite ways. A dose-response meta-analysis found progressively lower type 2 diabetes risk with higher coffee intake, holding for decaffeinated coffee too, which points at chlorogenic acids rather than caffeine (Ding et al., Diabetes Care 2014). Acute caffeine, meanwhile, measurably reduces insulin sensitivity in controlled trials (Shi et al., 2016). Both can be true, which makes HbA1c and fasting glucose worth watching rather than assuming.
Is it the coffee or what you put in it
A blended sweetened coffee drink carrying 40 to 60 g of sugar is a sugar-sweetened beverage with coffee in it, and it belongs to a different evidence base. The American Heart Association suggests an added sugar ceiling of about 25 g a day for women and 36 g for men, so one drink can clear the day's allowance (AHA). Caffeine in a formulated drink is a separate question again (energy drinks), and swapping sugar for a non-nutritive sweetener changes that question rather than closing it (sweeteners and erythritol).
Who needs a lower ceiling than 400 mg
Pregnancy is the clearest exception. EFSA puts the figure at 200 mg a day for pregnant and breastfeeding women, and caffeine is cleared more slowly during pregnancy. The BMJ umbrella review found harmful associations with coffee were largely nullified by adjusting for smoking everywhere except pregnancy, where higher intake remained associated with low birth weight, preterm birth and pregnancy loss.
Anyone with poorly controlled hypertension, an arrhythmia, an anxiety disorder or persistent insomnia has a lower practical ceiling than the population one, and some medicines slow caffeine clearance considerably. Those are conversations for a clinician.
Why the right amount of coffee is an individual number
CYP1A2 handles roughly 95% of caffeine clearance, which makes it the obvious candidate for why one person sleeps fine after an evening espresso and another does not. The most cited study on this, a case-control analysis of 2,014 Costa Rican heart attack cases, found four or more cups a day associated with higher risk only in carriers of the slow-metabolising variant (OR 1.64, 95% CI 1.14 to 2.34), with no increase in fast metabolisers (Cornelis et al., JAMA 2006).
That result has not held up cleanly. The far larger UK Biobank analysis found no interaction at all between CYP1A2 genotype, a broader caffeine metabolism genetic score and coffee intake with respect to cardiovascular risk. The same caution applies to ADORA2A variants, where associations with sleep complaints are real but modest and inconsistent (Erblang et al., Genes 2019). Anyone selling a caffeine gene report as a verdict is overselling a contested literature.
What is not contested is the size of the variation. Half-life differs several-fold between healthy adults, diterpene content about a hundredfold between brewing methods, and sleep, cortisol and lipid responses move independently of one another. Three to four cups is a reasonable opening hypothesis about a population and a weak claim about any single person inside it.
So the useful version of the question is not how much coffee is healthy, but what a particular habit has done to the markers it can plausibly move.
Five blood markers a coffee habit can plausibly move
| Marker |
What it tracks |
What coffee can do to it |
How fast it responds |
| LDL-C and ApoB |
Circulating cholesterol-carrying particles |
Unfiltered brews raise it, paper-filtered brews do not |
Two to four weeks after a brewing change |
| Homocysteine |
One-carbon metabolism and B vitamin status |
Rose about 10% on one litre of French press daily |
Within two weeks in trial conditions |
| Cortisol |
Stress axis output at the time of the draw |
Raises it acutely, with only partial habituation |
Same morning, so draw timing matters |
| ALT and GGT |
Liver cell stress and biliary enzyme activity |
Coffee intake tracks lower values, not higher |
Weeks to months, slow and habit-driven |
| HbA1c and glucose |
Average and current blood sugar control |
Additives move it far more than the coffee does |
HbA1c lags by roughly three months |
Read that as a list of things to check rather than things to fear. Four of the five rows move in a direction you can influence within one testing cycle.
How to see your own numbers
An Aniva membership costs EUR 199 per year and covers more than 100 biomarkers across 10 physiological systems, including lipids and ApoB, homocysteine, cortisol, liver enzymes, HbA1c and fasting glucose. The draw happens at a partner location in Germany or Finland and is performed by licensed physicians, and results land in the app in about a week with a personalised action plan, a biological age estimate and a personal health concierge chat. Membership is up to 100% reimbursable via German private health insurance (PKV), depending on your plan.
Retesting is part of the annual membership cycle, which suits this question well: change the brewing method or the cut-off time, keep the rest steady, then look at what moved. Getting started takes a few minutes.
Frequently asked questions
How many cups of coffee a day is healthy
Large pooled analyses put the lowest relative risk for all-cause and cardiovascular mortality at around three to four cups a day compared with none. EFSA separately concluded that habitual caffeine intakes up to 400 mg a day, roughly four to five standard cups, raise no safety concern for healthy non-pregnant adults. Both describe populations rather than individuals.
Is coffee bad for your cholesterol
It depends almost entirely on the filter. Cafestol and kahweol, the diterpenes that raise LDL cholesterol, are largely retained by paper, so filtered coffee has little effect while French press, boiled, Turkish and some espresso preparations raise cholesterol in a dose-dependent way. If your lipids are borderline and you drink several unfiltered cups a day, switching to paper-filtered coffee is a cheap experiment.
Does espresso raise cholesterol more than filter coffee
Per litre, espresso contains far more diterpenes than paper-filtered coffee, and levels vary widely between machines and beans. In the Tromsø Study, three to five espressos a day were associated with higher total cholesterol than none, more strongly in men. Serving size cuts the other way, since an espresso is a fraction of the volume of a mug of filter coffee.
How late can I drink coffee before it affects my sleep
A double-blind trial found that 400 mg of caffeine still measurably disturbed sleep when taken six hours before bedtime, which is where the common six-hour cut-off advice comes from. Caffeine's half-life averages four to five hours but ranges from under two to over nine hours between individuals, so a slow clearer needs a longer gap.
Can a genetic test tell me how much coffee to drink
Not reliably. CYP1A2 does most of the work of clearing caffeine, and one widely cited case-control study linked heavy coffee intake to higher heart attack risk only in slow metabolisers, but a much larger UK Biobank analysis found no such interaction. Treat caffeine gene reports as a hypothesis rather than an instruction.
How much caffeine is safe during pregnancy
EFSA sets 200 mg a day for pregnant and breastfeeding women, about half the general adult figure, because caffeine crosses the placenta and is cleared more slowly during pregnancy. Pregnancy is also the one context where the umbrella review found harmful associations survived adjustment for smoking. Anyone pregnant or planning to be should agree a personal limit with their own clinician.
This article is general information about coffee, caffeine and the blood markers they can affect. It is not medical advice and it does not diagnose or treat anything. Reference ranges are set by the analysing laboratory and differ between labs and methods, so a value flagged outside range in one report may sit inside range in another. Discuss your own results and history with a clinician.
