Coffee and Cholesterol: Does Unfiltered Coffee (Turkish, French Press) Raise LDL While Filtered Coffee Does Not?

Cafestol and kahweol are diterpenes — oily compounds that occur naturally in the fat of coffee beans. Cafestol is found in both arabica and robusta, whereas kahweol occurs almost only in arabica. They are lipophilic, so they pass into the cup together with micro-droplets of coffee oil — wherever the grounds are in direct contact with water and the drink is not strained through a dense barrier.

A paper filter acts as a mechanical trap for these oil droplets. Metal mesh filters, French press screens and moka pots, as well as no filter at all (Turkish coffee, 'cowboy' coffee poured over grounds), let far more diterpenes through. The difference is visible in measurements: in a 2025 Swedish study that measured the compounds by liquid chromatography–mass spectrometry, the median cafestol in home-brewed, paper-filtered coffee was 12 mg/L, while in boiled unfiltered coffee it was as high as 939 mg/L. Pouring the same boiled coffee through a fabric filter lowered the concentration to 28 mg/L.

One practical consequence is worth remembering: the type of filter, not the type of bean or the roast level, determines the effect on lipids. The same ground coffee brewed once through paper and once in a French press is two different drinks from a cafestol point of view.

Two things deserve attention. First, the LDL effect persisted throughout the six months — the body does not adapt to cafestol in a way that cancels its action. Second, the dose in this trial was very high (almost a litre of strong coffee a day), so the result does not scale linearly to one cup a day. It does show, however, that the direction of the effect is real and measurable.

These two studies give a picture that a later meta-analysis confirmed: cafestol behaves like a drug with a linear dose-effect relationship. The more unfiltered coffee, the higher the LDL. One cup a day is a completely different exposure from five.

A mean LDL rise of 5 mg/dL is small in a healthy person with a normal lipid profile, but it is not negligible when LDL is already elevated or in people who drink a lot of unfiltered coffee. Sensitivity also varied: people with hyperlipidemia reacted more strongly, which is a clinically useful clue.

The Tromsø data complement the interventional trials with what they could not test: what happens in real life with everyday habits. For espresso — a drink that the randomized trials of the 1990s barely covered — the result is a cautious signal: a modest but statistically significant association with total cholesterol at 3-5 cups a day.

The exact mechanism is not fully understood, but studies in humans have measured specific changes in lipid-transfer proteins. Dutch researchers who gave volunteers French press or filtered coffee for 24 weeks observed that cholesteryl ester transfer protein (CETP) activity rose by 12% after just 2 weeks and by 18% after 12 weeks, PLTP activity by 10%, and LCAT activity fell by 6-7%. Importantly, the rise in CETP clearly preceded the rise in LDL — suggesting it is one link in the process, although the authors do not claim it is the only one.

Reversibility is as important as the effect itself. In Urgert's trial and the other studies, the rise receded after switching to filtered coffee or stopping coffee. There is no permanent damage here — it is a metabolic effect that disappears with the exposure. In practice, this makes changing your brewing method a 'testable' intervention: you can get a lipid panel, switch to paper filters for 6-8 weeks and see the difference in yourself.

Two methodological caveats. First, the figures in the table are concentrations per litre of drink — an espresso serving is several times smaller than a cup of French press coffee, so the cafestol dose per serving can be comparable or lower than the concentration alone suggests. Second, pod coffee and moka pots were not measured separately in the studies we opened; the fact that they do not pass through paper is a hint that they may contain more diterpenes than filtered coffee, but that is an inference from the mechanism, not from measurements, and should be treated cautiously.

In the same trial that showed the LDL rise, ALT activity (alanine aminotransferase, a liver enzyme) in people drinking cafetière coffee rose by up to 80% relative to filtered coffee, and after 24 weeks was still 45% higher. In 8 of 22 participants ALT exceeded the upper limit of normal, and in three it was twice as high. The rise is asymptomatic and — as the authors stress — completely reversible after stopping unfiltered coffee.

What does this mean in practice? On one hand, the result does not mean liver damage in a clinical sense — this is a small sample of healthy volunteers and a large dose. On the other hand, if someone drinks a lot of brewed coffee and has unexplained, mildly elevated ALT, this is one of the rarely considered causes (the authors say so themselves). Before launching a broad work-up, it is worth keeping in mind. We discuss the general data on coffee and the liver — including the surprisingly modest interventional results — in a separate article.

The picture is consistent with the lipid mechanism: when cholesterol was omitted from the statistical model, cardiovascular risk in unfiltered-coffee groups rose, and after it was included the differences shrank — which fits the hypothesis that cholesterol is one of the mediating links (but that is a conclusion from a model, not proof). Note the scale, though: even unfiltered coffee in these data was not worse than no coffee for all-cause death in men (HR 0.96, the interval includes 1.0), and in women it was even associated with lower risk. The most worrying signal concerned older people drinking large amounts.

We cover caffeine dose and its effect on sleep and blood pressure in other articles in this series — here we focus only on lipids. An important distinction: cafestol is not caffeine. Decaffeinated unfiltered coffee still contains cafestol, and filtered caffeinated coffee does not.

This is one of the few cases in nutrition where we have a mechanism, randomized evidence, a meta-analysis and population data all pointing the same way — and at the same time a simple, cheap and reversible intervention: a paper filter. There is no reason to give up your favourite coffee; there is a reason to know how you brew it, especially if you drink a lot and your LDL is elevated. The best test is a lipid panel before and after the change.