Most decaf won’t raise your cholesterol. What decides it isn’t the decaf at all, it’s how you brew. Two oils in coffee, cafestol and kahweol, nudge LDL up by slowing the way the body clears cholesterol, and a paper filter strips them out while a cafetière lets them through. That holds for caffeinated and decaf alike. So filtered decaf is fine. It’s the unfiltered cup, decaf or not, that does the nudging: a cafetière, a moka pot, or a workplace bean-to-cup machine.
The verdict, brew by brew
Paper-filtered decaf almost certainly doesn’t raise your cholesterol. Cafetière, espresso, moka pot, Turkish or workplace bean-to-cup decaf might nudge it up a touch, but so will the caffeinated versions of those brews. Which decaffeination method you chose (Swiss Water, sugar cane, CO2, methylene chloride) does not meaningfully change the picture. The lever sits at the filter.
What the studies actually say (and don’t)
The decaf-and-cholesterol SERP is anchored in one 1991 study that’s still doing the rounds in 2026.
Superko et al., 1991 (Am J Clin Nutr, PMID 1877516), randomised 181 men to caffeinated coffee, decaffeinated coffee, or no coffee for eight weeks. The decaf group ended around 9 mg/dL higher on LDL than the other two. Apolipoprotein B also went up. The headline that travelled was “decaf raises cholesterol”.
The bit the SERP misses is everything that came after. Later commentary has often attributed the bump to the specific decaf used in that trial rather than to decaffeination as a process, but the paper itself does not break the coffees down by bean species. What the wider evidence base does show is that follow-up trials using different decafs failed to reproduce the LDL rise.
Three years later, Wahrburg et al., 1994 (Eur J Clin Nutr, PMID 8194502), switched 119 healthy students from caffeinated to two different decaffeinated coffees. Their finding, verbatim: “switch from regular to decaffeinated coffee had no cholesterol-elevating effects, irrespective of the type of coffee”.
Then the meta-analyses. Jee et al., 2001, found increases in serum lipids were greater in patients with hyperlipidemia and in trials of caffeinated or boiled coffee. Filtered coffee trials demonstrated very little increase in serum cholesterol. Cai et al., 2012, pooled the trial evidence and confirmed unfiltered coffee raised total cholesterol by 8.1 mg/dL and LDL by 5.4 mg/dL. Filtered coffee did almost nothing.
Read together, the evidence isn’t “decaf raises cholesterol”. It’s “high-diterpene coffee raises cholesterol, and filtration is what removes diterpenes”. A quietly different story, and the one the top of the SERP keeps not telling.
The actual mechanism: cafestol and kahweol
The lab work that pinned the mechanism down was the Dutch and German strand from the late 1990s. Cafestol and kahweol survive roasting essentially intact and survive decaffeination essentially intact. Caffeine is a different molecule, and decaffeination targets it specifically; the lipid layer comes through untouched.
Post et al., 1997 (Arterioscler Thromb Vasc Biol), showed cafestol downregulates cholesterol 7α-hydroxylase (CYP7A1), the enzyme that turns cholesterol into bile acids for excretion. Less bile acid synthesis means more cholesterol staying in circulation. Rustan et al., 1997 (same journal, same year), showed pre-incubation with cafestol cuts LDL receptor uptake by 15 to 20% and LDL degradation by 20 to 30%. Two mechanisms pulling in the same direction.
Dose-response confirms the effect in humans. Urgert et al., 1997 (Am J Clin Nutr, PMID 9022539), gave 10 healthy men around 60 mg per day of purified cafestol for four weeks. Total cholesterol rose by 0.79 mmol/L, about 31 mg/dL, roughly a 12% increase. These are real effects at exposures readers can plausibly reach from cafetière or workplace bean-to-cup decaf.
Cafestol and kahweol are not affected by decaffeination. They are affected by what your brewer does with them.
Why brew method matters more than decaf vs regular
Numbers from Orrje et al., 2025, in Nutrition, Metabolism and Cardiovascular Diseases (PMID 40089392), the most comprehensive recent quantification of cafestol across brew methods.
| Brew method | Cafestol (mg/L) | Notes |
|---|---|---|
| Paper-filtered drip or pour-over | ~12 (range 4 to 24) | Paper retains diterpenes |
| AeroPress with paper filter | ~12 | Same logic |
| Espresso (typical) | ~80 | Concentrated cup, small serving |
| Espresso (high end) | up to 2,447 | Sample-dependent outlier |
| Cafetière (French press) | ~90 | Metal mesh lets oils through |
| Workplace bean-to-cup machine | 176 (range 24 to 444) | Often higher than cafetière |
| Boiled or Turkish | up to 939 | Worst case for diterpene exposure |
A V60 of decaf delivers around 0.2 mg of cafestol per cup. A cafetière of decaf delivers around 7 mg. Same bean. Same caffeine status. Roughly 35 times the cafestol load.
That is the variable doing most of the work. The “decaf vs regular” framing the SERP uses is a sideshow. If you drink three cafetières of decaf a day, you’ll deliver more cafestol than someone drinking three V60s of full-caffeine coffee.
Paper filters trap diterpenes because the oils are lipid-soluble and bind to cellulose. Metal mesh doesn’t. AeroPress with paper behaves like a paper filter. AeroPress with a reusable metal disk doesn’t. Cold brew traditionally uses no paper, so it retains diterpenes too.
Does the decaffeination method itself matter?
In a word, no. Not for cholesterol.
The four mainstream methods of decaffeination are Swiss Water, sugar cane (ethyl acetate), CO2 and methylene chloride. All four target caffeine. None of them documents a meaningful drop in cafestol or kahweol content. Decaffeination is selective for the molecule that gives coffee its kick. The oily diterpene fraction comes along for the ride.
Swiss Water’s own consumer page is candid about this: the cafestol and kahweol oils “naturally occur in all coffee, regardless of caffeine content”, and the brand recommends paper-filter brewing as the actual mitigation. That is the correct answer. It cost nothing to surface in the top result of the SERP. Nobody did.
The one variable in the bean that does matter is species. Robusta has roughly twice the cafestol content of arabica. Modern specialty-grade decaf is overwhelmingly arabica, which keeps the diterpene load comparable to caffeinated arabica.
Roast level shifts diterpene content modestly, with darker roasts losing a touch. Not enough to change practical advice. The lever sits at the filter.
If you’re watching your cholesterol, drink this
The brewer is the first decision. Paper-filtered drip (V60, Chemex, Kalita Wave, Melitta), pour-over or AeroPress with a paper filter all pull the oil layer out before it reaches your cup. Anything else lets the oils through.
If you drink more than two or three cups a day and your LDL needs attention, the kit to avoid is cafetière, AeroPress with the metal disk, moka pot, Turkish, Scandinavian boiled and the office bean-to-cup. Espresso sits in the middle: the drink is concentrated but the serving is small. One or two shots a day is a very different exposure than four cafetières.
Bean choice is the easy bit. Any clean arabica decaf will do. The decaffeination process is a flavour choice. It isn’t a cholesterol choice.
Five from the directory that fit a paper-filter setup well:
- Origin Coffee, Atlas Decaf. Sugar cane EA, medium roast, apple and chocolate. Origin is one of the more respected UK specialty roasters and Atlas brews cleanly in a V60.
- Caribe Coffee, Swiss Water Decaf SHG. Honduran Swiss Water, walnut and toffee. Reliable filter coffee from a steady UK roaster.
- Decadent Decaf, Swiss Water Colombia. UK direct-to-consumer specialist. Classic Swiss Water profile, easy to repeat-order.
- Artisan Roast, Decaf Brazil Swiss Water. Almond, molasses, cocoa. The archetypal comfort-chocolate filter cup at £9.50.
- Apostle Coffee, The Needle’s Eye Organic Decaf. Sumatran Swiss Water. Butterscotch, marjoram, nutmeg. Breaks the chocolate gravity if you want something different.
All five paper-filter cleanly. None will deliver enough cafestol to move LDL at sensible volumes.
What we still don’t know
Individual response varies more than the trial work captures. CYP7A1, the enzyme cafestol downregulates, has genetic variants that the population-scale coffee-cholesterol work hasn’t mapped onto. Some people probably absorb and respond to diterpenes more than others. The dose-response curves are population averages.
Threshold effects in normal-cholesterol drinkers aren’t well-mapped either. The Urgert 1997 dose-response trial used around 60 mg per day of purified cafestol, which is workplace-bean-to-cup territory at several cups a day. Whether one cafetière of decaf moves LDL meaningfully in someone with already-good numbers is genuinely uncertain.
Real-world ranges are huge. The Orrje 2025 cafestol numbers for workplace bean-to-cup machines run from 24 to 444 mg/L, an eighteen-fold spread across machines, beans and grinds. “Avoid the office machine” is a useful rule with wide individual variance behind it.
Swiss Water versus sugar cane versus CO2 has never been compared head-to-head in a diterpene assay at scale. The verdict above (decaffeination method doesn’t change cafestol exposure) is best read as absence of evidence for an effect, not strong evidence of no effect.
The LDL question and the heart-disease-outcome question aren’t the same. Most large cohort studies find no excess coronary risk from decaf consumption, including in groups whose LDL rose slightly in trial work. A 5 mg/dL LDL bump from a cafetière habit may or may not translate to a clinical endpoint. The evidence currently says probably not, but probably is a real word doing real work in that sentence.
If you’re on statins, blood-pressure medication, or have familial hypercholesterolaemia, this is editorial guidance from a decaf specialist, not medical advice. Your GP or lipid clinic has the actual numbers that matter for you.
Where to go from here
The directory has every UK decaf we track, filterable by method and roaster. If you want to dig further on the method comparison, the full pieces on Swiss Water, sugar cane EA, CO2 and chemical methods go further on the chemistry than this article does.
If you’re cholesterol-conscious, the move is a V60 and any clean arabica decaf you fancy. The rest is detail.