The guide

The 10 studies that shape what we know about cholesterol

By GyroscopeUpdated November 15, 202615 sources

Ten studies do most of the work behind what this site says about cholesterol. Three show that LDL causes heart disease, three show that lowering it by different routes helps, and four show the limits: a marker that wasn’t a cause, diet trials that disappointed, and the other risks that sit beside LDL. Each one comes with what it found, why it counts and where it falls short.

1. The PCSK9 families (Cohen and Hobbs, 2006)

   What it found. In ARIC, a 15-year US heart study, 2.6% of 3,363 Black participants carried a broken PCSK9 gene. Their LDL was 28% lower from birth and they had 88% less coronary heart disease. White participants with a milder variant had 15% lower LDL and 47% less heart disease.

   Why it counts. Genes are handed out at random, so this is close to a lifelong trial. A modest drop held for life protected far more than the same drop from a drug in middle age, which is where “lower for longer” comes from.

   Its limit: the carriers were a small group in one study, so the exact size of the protection is uncertain.

2. The Cholesterol Treatment Trialists (2010)

   What it found. Across 26 statin trials and about 170,000 people, each 39 mg/dL (1 mmol/L) off LDL cut heart attacks, strokes and artery procedures by about a fifth and deaths from any cause by 10%, over about five years.

   Why it counts. It’s the dose. The benefit follows the size of the LDL drop, not the drug or the starting number.

   Its limit: most people were in trials for about five years, so it can’t show a lifetime.

3. Different routes, the same benefit (Silverman and others, 2016)

   What it found. In 49 trials and 312,175 people, statins cut major events by 23% for each 39 mg/dL, and diet, ezetimibe, bile acid binders and even surgery that bypassed part of the gut cut them by 25%, with no real difference between the two.

   Why it counts. If LDL were just a side effect of whatever statins really do, other ways of lowering it wouldn’t help. They did, by the same amount.

   Its limit: CETP drugs did worse than their LDL drop predicted, a reminder that a drug can do other things.

4. It’s the particle count (Ference and others, 2019)

   What it found. In 654,783 people, genes that lowered triglycerides and genes that lowered LDL cut heart disease by the same amount per 10 mg/dL less ApoB. With ApoB counted, LDL and triglycerides added nothing.

   Why it counts. It explains why LDL works as a marker most of the time and fails some of the time. What causes plaque is the number of ApoB particles, and LDL cholesterol is a rough stand-in for it.

   Its limit: it’s genetic, and ApoB-targeted treatment goals haven’t been tested head to head with LDL goals.

5. Very low is safe and helps (FOURIER, 2017, and its extension)

   What it found. Evolocumab took LDL from 92 to 30 on top of a statin in 27,564 people with heart disease and cut the main events 15%. In the extension, people stayed on it up to 8.4 years without more side effects than placebo.

   Why it counts. It answered the worry that very low LDL might harm the brain or cause diabetes, and it showed a non-statin route working.

   Its limit: the main trial lasted about two years, and everyone already had heart disease.

6. Ezetimibe on top of a statin (IMPROVE-IT, 2015)

   What it found. After a heart attack, adding ezetimibe took LDL from 70 to 54 and cut events from 34.7% to 32.7% over seven years.

   Why it counts. Ezetimibe works in the gut, nothing like a statin, and it still helped in proportion to the LDL drop, below the old goals.

   Its limit: the benefit was small in absolute terms, 2 people in 100 over seven years.

7. HDL genes don’t protect (Voight and others, 2012)

   What it found. People born with a variant that raised HDL had the same heart attack risk as everyone else. A score of HDL-raising genes didn’t protect either. LDL-raising genes did raise risk, more than the population studies had suggested.

   Why it counts. It’s the clearest case of a number linked to risk that isn’t the cause, and it explained why HDL-raising drugs kept failing.

   Its limit: HDL particles do many things, and the study tested only the cholesterol inside them.

8. Saturated fat trials (Cochrane review, 2020)

   What it found. Across 15 trials and about 59,000 people, eating less saturated fat cut heart attacks, strokes and other heart events by 21%, more when cholesterol fell more. It found little or no effect on deaths.

   Why it counts. It’s the best test of the food advice itself, done by randomizing people instead of comparing what they chose to eat.

   Its limit: the trials varied a lot, many are old, and the evidence on deaths and on single outcomes is weak.

9. The recovered Minnesota trial (Ramsden and others, 2016)

   What it found. In 9,423 people in a nursing home and mental hospitals from 1968 to 1973, corn oil and margarine in place of animal fats lowered cholesterol 13.8% and didn’t lower deaths. Much of the data had never been published.

   Why it counts. It’s the skeptics’ strongest trial, and it’s fair. It shows that swapping in one kind of vegetable oil isn’t proven to save lives, and that how you lower cholesterol matters.

   Its limit: most people were on the diets for less than a year, in institutions, and it tested one oil pattern from the 1960s.

10. Risk is many things (INTERHEART, 2004)

    What it found. In heart attack patients and controls from 52 countries, nine risk factors accounted for about 90% of the risk: blood fats (ApoB against ApoA1), smoking, high blood pressure, diabetes, belly fat, stress, few fruits and vegetables, little activity and alcohol. Blood fats were the biggest single share, about half.

    Why it counts. It keeps LDL in its place. It’s the biggest single cause, and it’s one of several.

    Its limit: comparing people who had heart attacks with people who didn’t can’t prove cause on its own.

Just off the list

The 4S trial in 1994 first showed a statin saving lives, 8% deaths against 12%. The 2012 genetic analysis of 312,321 people put a number on lower for longer, about three times the protection per point. The 2022 Cholesterol Treatment Trialists’ review of 19 blinded trials found that most muscle aches blamed on statins happened just as often on placebo. And Lp(a)HORIZON, published in November 2026, found that cutting Lp(a) didn’t cut events, which keeps Lp(a) an open question.

Does LDL really matter? uses all of these to weigh the case for and against, and the story of cholesterol puts them in order.