Nearly everything in coffee research is an association pulled out of observational data, hedged with caveats about confounding. This article is about the exception. The effect of unfiltered coffee on blood cholesterol has been demonstrated in randomised trials, the responsible molecule has been isolated and named, the mechanism is understood at the level of liver receptors, and the intervention that removes it costs a few pesos and takes no extra time. It is the most genuinely actionable finding in the field, and it is also the one fewest coffee drinkers have heard of.
The molecule
Coffee beans contain oil, and in that oil are two closely related diterpene compounds: cafestol and kahweol. The work identifying cafestol as the responsible agent was reviewed by Urgert and Katan in the Annual Review of Nutrition[1], which laid out the case that this specific compound, rather than caffeine or coffee generally, is what raises serum cholesterol in people drinking unfiltered coffee.
The mechanism is not vague. Cafestol interferes with the liver's regulation of cholesterol handling, reducing the rate at which LDL particles are cleared from the bloodstream. Less clearance means more LDL circulating. This is a pharmacological effect of a specific molecule on a specific pathway, which is why it produced consistent results in controlled feeding trials while so much other coffee research produced arguments.
Why the filter is the entire story
Cafestol and kahweol are carried in coffee oil and in the fine suspended particles that come through with it. A paper filter traps both almost completely. Anything that does not use paper lets them through.
In practical terms, that sorts brewing methods into two groups. On the low-diterpene side: drip machines and pour-over cones using paper filters, and pod machines that filter through paper. On the high side: Scandinavian boiled coffee, which carries by far the most, followed by French press, percolators, Turkish and Greek coffee, and to a smaller degree espresso and moka pot, which produce a small volume of concentrated, unfiltered coffee. Metal mesh filters and cloth filters do not remove the diterpenes, which is the part people using a metal-filter pour-over or a reusable mesh basket usually miss.
That is worth naming plainly because the specialty coffee world has drifted toward metal filters and immersion brewing for flavour reasons, since the oils carry body and texture that paper strips out. That is a genuine flavour gain, and it comes with this specific trade-off. Our home brewing method guide covers the taste side of the same decision.
How large is the effect?
Cai and colleagues pooled 12 randomised controlled trials in a 2012 meta-analysis[2] of coffee's effect on serum lipids and confirmed a significant effect on total cholesterol, LDL cholesterol and triglycerides, with the effect concentrated among trials using unfiltered preparations and larger at higher doses. Because these were randomised trials rather than surveys, the usual confounding objections do not apply: participants were assigned their coffee, and their blood was measured before and after.
The magnitude reported across this literature is on the order of several milligrams per decilitre of additional LDL for people drinking substantial quantities of unfiltered coffee daily, rather than a dramatic reversal of anyone's lipid panel. For someone drinking one cup of French press a day, this is a small effect that will be invisible next to diet and genetics. For someone drinking five or six cups of boiled or press coffee daily, it is the difference between two meaningfully different cholesterol readings, and it compounds over decades.
It is worth noting what kind of study produced these numbers, because it is unusual for this field. These were controlled feeding trials: participants were assigned a coffee preparation, drank it for a defined period, and had their blood lipids measured before and after. That design removes the confounding that plagues the observational coffee literature entirely. Whatever else is uncertain about coffee and health, the fact that unfiltered coffee raises LDL cholesterol is not in serious dispute.
Why espresso sits in an awkward middle
Espresso deserves more attention than it usually gets here, because the reasoning cuts both ways. On concentration, espresso is unfiltered and its diterpene content per millilitre is high, comparable to other unfiltered methods. On dose, a single shot is around 25 to 30 millilitres, so the absolute quantity delivered is small.
The arithmetic favours espresso drinkers, but only up to a point. Two shots a day is a modest exposure. Six shots a day, which is not unusual for someone who works in or lives around cafes, starts to approach the exposure from a couple of mugs of French press. Milk does not change this: a latte contains the same shot as an espresso, so the diterpene load is identical and only the volume of the drink differs. If you are counting, count shots rather than cups.
The moka pot, common in Filipino households, behaves similarly: unfiltered, concentrated, small serving. Cold brew depends entirely on how it is filtered, and paper-filtered cold brew removes the diterpenes just as hot paper filtering does.
The long-term data
The population-scale test of whether this matters came from Norway, a country that historically drank a great deal of unfiltered boiled coffee and then shifted substantially toward filtered, which gave researchers a rare natural comparison. Tverdal and colleagues followed 508,747 men and women aged 20 to 79 for an average of 20 years[3], recording 46,341 deaths from any cause, 12,621 from cardiovascular disease, 6,202 from ischaemic heart disease and 2,894 from stroke.
Taking non-drinkers as the reference, hazard ratios for death from any cause among men were 0.85 for filtered brew, 0.84 for people drinking both, and 0.96 for unfiltered brew. Among women the figures were 0.85, 0.79 and 0.91. For cardiovascular death specifically, filtered coffee was associated with hazard ratios of 0.88 in men and 0.80 in women, while unfiltered sat at 0.97 and 0.83.
The pattern is consistent: filtered coffee was associated with lower mortality than drinking no coffee at all, and unfiltered coffee was associated with higher mortality than filtered. Stratifying by age sharpened it. Among men aged 60 and over, the hazard ratio for cardiovascular death comparing unfiltered brew with no coffee was 1.19. The researchers also noted that the hazard ratios rose when total cholesterol was removed from the statistical model, which is exactly what you would expect if cholesterol is the pathway through which the effect operates. That is a coherence check, and the data passed it.
Reading this proportionately
Two mistakes are easy here, in opposite directions. The first is to conclude that French press coffee is dangerous. It is not. The Norwegian data show unfiltered coffee drinkers doing roughly as well as non-drinkers, not worse than them, in most strata. The effect is a lost benefit more than an inflicted harm, and it becomes clinically interesting mainly at high daily volumes, at older ages, and in people who already have elevated LDL or established cardiovascular risk.
The second mistake is to dismiss it as another nutrition scare. This is a randomised-trial-supported, mechanistically explained, dose-dependent effect of an identified molecule, which puts it in a completely different evidential category from most of what gets written about coffee. van Dam, Hu and Willett[4] treat it as one of the few coffee findings solid enough to inform actual recommendations, and note that people with elevated cholesterol are the group for whom brewing method genuinely matters.
What else the filter takes out
One more thing is removed by paper besides the diterpenes, and it is worth knowing because it explains the flavour trade-off. Paper traps the coffee oils that carry much of the drink's body and mouthfeel, along with the finest suspended particles. That is why filter coffee tastes clean and slightly thin next to a French press, and why immersion brewing has the reputation it does among people who like a heavier cup.
So the choice is not between good coffee and healthy coffee in any absolute sense. It is a specific trade of texture against a specific, measured lipid effect, and how that trade lands depends entirely on your own cholesterol and how much you drink. Someone with normal lipids drinking one press pot a day is making a decision with essentially no consequences. Someone with elevated LDL drinking a litre of it is making a decision with measurable ones.
The people for whom this genuinely matters are a narrow group, and if you are in it you probably already know, because someone has shown you a lipid panel and pointed at the LDL line. For everyone else this is a piece of trivia with a small practical tail.
Even then, the fix is a box of paper filters rather than giving up coffee, which is a considerably easier conversation than most cardiovascular advice.
It is also worth saying that decaffeination does not touch this at all. Cafestol and kahweol are unrelated to caffeine, so decaf French press carries the same diterpene load as caffeinated French press. Switching to decaf for cardiovascular reasons while keeping an unfiltered brewing method addresses the wrong variable entirely.
Reading a lipid panel with this in mind
If you want to know whether this applies to you, the practical test is straightforward and costs nothing beyond a blood test you may be having anyway. Keep your brewing method constant for the eight to twelve weeks before a lipid panel, then change it and repeat. Diterpene effects on LDL appear within weeks in the controlled trials, so a genuine change should be visible on that timescale.
That is not a formal experiment and a single pair of readings is noisy. But for the specific group this matters most to, people with elevated LDL who drink several cups of unfiltered coffee daily, it is a reasonable way to find out whether the population effect shows up in your own numbers rather than assuming it does or does not.
What to actually do about it
If you have high LDL cholesterol, or your doctor has flagged cardiovascular risk, and you drink several cups a day, switching to a paper-filtered method is a free, no-downside change that addresses the one coffee variable with clear evidence behind it. A V60 or Kalita cone with paper filters, or an ordinary drip machine, both qualify. If you like the body that paper strips out, a compromise many people land on is a paper filter placed inside a metal-mesh dripper.
This is genuinely one of the cheapest interventions in the whole of nutrition. A box of paper filters costs very little, the change takes no additional time, and the evidence behind it comes from randomised trials rather than from surveys.
Espresso is a middle case worth understanding. It is unfiltered and its diterpene concentration is high, but the serving volume is small, so total daily exposure from two shots is far below that of several mugs of press coffee. A latte is mostly milk, and the coffee fraction is one or two espresso shots. Someone drinking two lattes a day is not in the same exposure category as someone drinking a litre of French press.
The practical rule that falls out of all this is simple enough to remember: if the coffee passed through paper, the diterpenes did not.
And for everyone else, with normal cholesterol and a cup or two a day, this is genuinely a footnote. Drink what tastes good. If you want the flavour comparison rather than the lipid one, our brewing guide covers what each method actually does to the cup, and Franken Coffee Bar + Roastery in San Roque is a reasonable place in Marikina to taste filtered and unfiltered preparations side by side. As always, this is a summary of research and not medical advice.
The reason this article exists at all is that it is the rare coffee finding where the evidence is strong enough, and the fix cheap enough, to be worth acting on rather than merely knowing.
Everything else in this section is worth knowing; this one is worth doing something about.