Ezetimibe costs a few pounds a month, has a side-effect profile close to placebo, and lowers LDL by about a fifth. It was written off in 2008 on the strength of an imaging test, and the story of how it came back is a better lesson in reading evidence than the drug itself is.
1. What it does, and why the mechanism matters
Your body handles cholesterol in two places, and the drugs split neatly along that line.
| Statins | Ezetimibe | |
|---|---|---|
| Where it acts | The liver | The small intestine |
| What it blocks | Production — HMG-CoA reductase, the rate-limiting enzyme in making cholesterol | Absorption — the NPC1L1 transporter in the gut lining |
| What it stops | New cholesterol being manufactured | Cholesterol being taken up from the gut — both from food and from bile, which is the larger share |
| Typical LDL reduction | 30–55% depending on drug and dose | 15–20% on its own; roughly a further 20% when added to a statin |
The two pathways compensate for each other, which is exactly why combining them works. Block production and the gut absorbs more; block absorption and the liver makes more. Doubling a statin dose gains only about 6% more LDL reduction — the well-known "rule of six" — because the body pushes back along the other pathway. Adding a drug that closes the second door gains far more than the third doubling of the first. That is the whole argument for combination therapy over dose escalation, and it is a mechanistic argument that the outcome data later supported.
A small historical oddity worth knowing: ezetimibe was approved in 2002 on its ability to lower LDL, and the protein it acts on was not identified until 2004–05. The drug was in pharmacies for two years before anyone could say what it bound to.
2. How it was written off — ENHANCE, 2008
ENHANCE gave people with familial hypercholesterolaemia either a high-dose statin or the same statin plus ezetimibe, and measured the thickness of the carotid artery wall on ultrasound. The combination lowered LDL substantially more. The artery walls did not differ.
The reaction was severe and the drug's reputation did not recover for years. But the trial had measured a surrogate, not an outcome — and a particularly unsuitable one:
- The participants had familial hypercholesterolaemia and had mostly been on statins for years already. Their arteries were both thickened and already treated, so there was little room left for a difference to appear in two years of imaging.
- Wall thickness is not plaque, and it is not events. It is a measurement that tracks risk across populations without necessarily moving when risk changes in an individual.
- The trial was not designed to detect events and could not have done so at that size and duration.
The lesson generalises well beyond this drug, and it is the same one the rest of this site keeps running into: a treatment that fails on a surrogate has not been shown to fail. It has been shown to fail on that surrogate. See how to read a trial — and note that the same logic cuts the other way in the inflammation drugs, where a treatment that succeeded on a biomarker went on to fail on events.
3. What rehabilitated it
Human genetics — the natural experiment
Some people are born with mutations that partially disable the NPC1L1 transporter — a lifelong, low-dose version of taking ezetimibe. Compared with everyone else they have LDL lower by about 12 mg/dL (roughly 0.3 mmol/L), which is a small difference. And they have around 53% less coronary heart disease.
That is a much larger effect than the LDL gap alone would predict from drug trials, and the reason is duration: this is a small difference maintained from birth rather than from age fifty-five. It is the clearest illustration on this site of why cumulative LDL exposure matters more than the number on any one blood test — the same argument made in the essentials. Strong for the causal direction; Mendelian randomisation cannot tell you what a drug started in middle age will achieve.
IMPROVE-IT — the outcome trial
18,144 people after an acute coronary syndrome, randomised to a statin alone or the same statin plus ezetimibe, followed a median of six years.
| Statin alone | Statin + ezetimibe | |
|---|---|---|
| Average LDL achieved | 1.8 mmol/L (69.5 mg/dL) | 1.4 mmol/L (53.7 mg/dL) |
| Cardiovascular events at 7 years | 34.7% | 32.7% |
| Absolute difference | 2.0 percentage points — hazard ratio 0.936, p = 0.016. About 50 people treated for 7 years per event avoided | |
| Death from any cause | 15.4% | 15.3% — no difference |
| Cardiovascular death | 6.9% | 6.8% — no difference |
| Side effects | No difference in muscle, gallbladder or liver problems, or cancer | |
How to hold those two facts together
- It worked, and it worked modestly. The reduction in heart attacks and strokes is real, statistically solid, and consistent with what the LDL difference predicted. Descriptions of IMPROVE-IT as having "definitively proven" ezetimibe overstate a 2-point difference.
- There was no mortality benefit. That is not a hidden failure — the trial was not powered for it, and few lipid trials of this size and duration show one — but anybody quoting this trial should say it.
- The value of the trial is what it established about LDL, not only about the drug. Lowering LDL further, by a mechanism that has nothing to do with statins, produced the benefit the LDL difference predicted. That is strong evidence that the LDL reduction is doing the work, rather than some statin-specific property.
- It also matters that ezetimibe is boring. A drug with placebo-level side effects and a 2-point benefit is a perfectly reasonable thing to take; a drug with real side effects and a 2-point benefit would not be.
4. How low should LDL actually go?
The guidelines moved to a target of 1.4 mmol/L (55 mg/dL) for very high risk — the tiers are on the calculator page — largely on the strength of drug trials in which reaching a low LDL was a consequence of the treatment rather than the goal being tested. Until 2026, no trial had compared the targets themselves.
Ez-PAVE did. 3,048 people with established cardiovascular disease in South Korea, randomised to aim for below 1.4 mmol/L or below 1.8 mmol/L, with clinicians free to use whatever combination reached it.
| At 3 years | Target < 1.8 | Target < 1.4 |
|---|---|---|
| Primary composite | 9.7% | 6.6% — hazard ratio 0.67, NNT 32 |
| Of which, revascularisation | 7.5% | 4.8% |
| Harder composite — CV death, heart attack or stroke | 3.6% | 2.3% |
| LDL actually achieved | 1.7 mmol/L | 1.45 mmol/L |
Moderate — a genuinely useful trial with three limits that should be read alongside it:
- It was open-label, because clinicians had to know the target they were aiming for. Outcomes were adjudicated blind, which helps.
- The benefit was driven mainly by revascularisation — a decision made by a clinician who knew which arm the patient was in. That is the softest component of the composite and the one most susceptible to open-label design. The harder composite moved in the same direction, which is reassuring, on smaller numbers.
- It was conducted entirely in South Korea, and newer agents such as inclisiran and bempedoic acid were not available there during the trial.
The dissent is worth knowing about. Some commentators argue Ez-PAVE does not validate the 1.4 target, precisely because of the revascularisation point. That is a reasonable position, and the two targets differ by less than most people assume — the achieved LDLs were 1.45 against 1.7 mmol/L, which is a narrower gap than the target labels suggest. What Ez-PAVE does establish is that aiming lower, and using combination therapy to get there, is not harmful — safety was similar in both arms.
It is corrected here rather than silently amended because the error had a practical consequence: a clinician reading the old wording would have concluded that the whole class was out of their hands, which is one of the reasons this treatment is under-prescribed. In Greater Manchester, by way of scale, roughly 1,900 people were receiving inclisiran against an estimated 3,700 eligible.
5. Where it sits in the order of things
| Step | Why here |
|---|---|
| 1. The highest statin dose you tolerate | Largest effect, most evidence, cheapest. Everything else is judged against it |
| 2. Add ezetimibe | Before escalating anything else. A different mechanism, so the effects add; a further ~20% LDL reduction; outcome evidence; placebo-level tolerability; a few pounds a month. In the UK it is a routine NHS prescription |
| 3. Bempedoic acid, if statins are not tolerated at all | Acts upstream of statins and bypasses muscle tissue, so it avoids the statin muscle problem. Has its own outcome trial in statin-intolerant patients |
| 4. A PCSK9 inhibitor or inclisiran | A 50–60% further reduction, and the only options that meaningfully lower Lp(a). In the UK all of them are injections, and access is restricted by criteria rather than by clinical judgement alone. The two routes are not the same, and this is widely misunderstood. The monoclonal antibodies — alirocumab (TA393) and evolocumab (TA394) — are specialist-initiated only. Inclisiran (TA733) is not: a GP can start it in someone with established cardiovascular disease whose LDL stays at or above 2.6 mmol/L on maximally tolerated statins. The PCSK9 chapter has the full eligibility table, the outcome evidence and the safety questions |
Ezetimibe is at step 2 for a specific reason, and it is not that it is the most powerful — it plainly is not. It is that the ratio of benefit to cost, risk and inconvenience is better than anything else on the list. Skipping it to chase a bigger percentage is a common pattern and rarely the right call.
One caution that applies specifically to people who have had trouble with statins. A minority of people who react to statins also react to ezetimibe — see the clinical depth on statin intolerance. Start it on its own, at a normal dose, and give it a fair trial before concluding anything.
6. The primary prevention gap, and why it exists
That gap is not a scientific one. It is a commercial one, and it is worth stating plainly because it affects how the evidence base looks:
- Ezetimibe is off patent and costs very little. A primary prevention outcome trial would need tens of thousands of participants followed for many years, at a cost of hundreds of millions.
- Nobody would ever recover that money, because the drug is generic. There is no commercial actor with a reason to fund it, and public funders rarely finance trials of this size for an existing cheap drug.
- So the trial will almost certainly never be done, and the absence of evidence will persist indefinitely rather than being resolved.
What follows from that, stated plainly. The mechanistic and genetic case for ezetimibe in primary prevention is strong — the NPC1L1 genetics are entirely a primary prevention finding. But "we are confident it works and nobody will pay to prove it" is a different claim from "it has been shown to work", and this page will not blur the two. In practice, the decision in primary prevention rests on how high the person's risk is, what their LDL is doing on a statin alone, and whether there is imaging — which is exactly where a calcium score earns its place.
7. What exists elsewhere — the first oral PCSK9 inhibitor
This page carries it anyway, and the reason is worth stating. The single largest practical objection to PCSK9 inhibition has never been the biology — it is that the drugs are injections, which deters patients, complicates delivery and keeps a very effective class confined to people who have already got as far as a lipid clinic. An oral option removes that objection entirely, and for someone who has declined injections it is the most consequential development in this area in a decade.
What to do with that, if you are in the UK
- Nothing yet, and be wary of anyone selling it. No UK licence means no legitimate NHS or private route today.
- But it is worth knowing about if injections are your objection, because the answer may be to wait rather than to conclude that this class is closed to you — and it is a reasonable thing to raise with a lipid specialist.
- And the discipline this site applies to inclisiran applies exactly as hard here. Enlicitide is approved on LDL reduction. Its cardiovascular outcome trial — CORALreef Outcomes — has not reported. That is the same position as inclisiran — a drug licensed on the strength of the surrogate rather than evidence for it — and a second example in the same class does not make the surrogate stronger, it makes the pattern clearer.
- Also newly approved in the US: lerodalcibep, a monthly fusion protein, likewise without UK availability. And NICE is currently appraising obicetrapib, an oral CETP inhibitor with a different mechanism again — appraisal ID6519, publication expected 14 October 2026.
Why this site reports drugs you cannot get. A UK reference that only described UK availability would leave a clinician unable to answer the question a well-read patient actually asks, and would leave the patient assuming nothing exists. The rule applied here is that a genuine advance gets reported with its status attached — what it is, where it is licensed, what has and has not been shown, and what if anything to do now. Compare the sunscreen chapter, where the same device runs the other way: filters the UK has had for decades that the US only recently approved.
8. The far end of the ladder — editing the gene instead of taking a drug
Section 3 made an argument that has an obvious next move. People born with a partially disabled NPC1L1 transporter have lower LDL for life and much less coronary disease. The same is true, more strongly, of people born with loss-of-function variants in PCSK9. If a genetic accident produces that protection, could you produce it deliberately?
The thing people worry about, which is not what is happening
This edits liver cells, not eggs or sperm
- It is somatic gene editing. The change is made in the cells of one organ in one adult. It is not heritable, and it does not affect that person's children.
- Germline editing — altering an embryo so the change passes down generations — is a different thing entirely, is not what any of these therapies do, and is prohibited in clinical use in the UK and most jurisdictions.
- The two get conflated constantly, and the conflation is the main reason this technology sounds more alarming than it is. The relevant comparison is not to designer babies; it is to a drug that does not wear off.
Why permanence is the whole question, in both directions
| The case for | The case for caution |
|---|---|
| Adherence is the unsolved problem in lipid lowering, and it is enormous — roughly half of people prescribed a lipid-lowering drug have stopped within a year. A treatment that cannot be forgotten removes the single largest cause of real-world failure Duration is what the genetics say matters. The NPC1L1 and PCSK9 carriers get their large effect from a modest difference maintained from birth. A one-off treatment in middle age is closer to that model than a tablet started at sixty and stopped at sixty-one The LDL surrogate is unusually well justified here. This site is normally sceptical of biomarker endpoints — but lowering LDL has cardiovascular outcome evidence across four pharmacologically distinct mechanisms: statins (HMG-CoA reductase), ezetimibe (NPC1L1), PCSK9 monoclonal antibodies, and bempedoic acid (ATP-citrate lyase, CLEAR Outcomes). Mendelian randomisation points the same way across the corresponding genes. It is about as validated as a surrogate gets. Note what that list excludes, because the discipline is the point: inclisiran is not on it. It lowers LDL substantially through the same PCSK9 pathway, but its outcome trials — ORION-4 and VICTORION-2 Prevent — have not reported. It is a drug licensed on the strength of the surrogate, not evidence for it |
It cannot be stopped. Every drug on this page can be discontinued if something emerges. This cannot, and an off-target edit or a late signal would be permanent These are Phase 1 numbers in 35 people. There is no cardiovascular outcome data at all, and there will not be for years The predecessor had problems. VERVE-101 ran into liver enzyme and platelet signals; the delivery particle was redesigned for VERVE-102, and the newer version has so far avoided them — about a fifth had mild infusion reactions. That is reassuring rather than settled Very long-term safety cannot be studied in advance of the long term. That is not a criticism of the programme; it is a structural feature of permanent interventions |
Emerging — and this is what Emerging is for. The mechanism is sound, the early human data is real and was published in a serious journal, and nothing here has been shown to prevent a single heart attack. It is being tested where the balance is most favourable: familial hypercholesterolaemia and premature coronary disease, in people facing decades of very high LDL, rather than as prevention in the generally well.
What to do with this information today: nothing. It is not available, it is not licensed, and it will not be for years. It is on this page because it is the logical end of the argument the rest of the chapter makes — that cumulative LDL exposure is what matters and that the mechanism of lowering it does not — and because when it does arrive it will be discussed as though it were settled. It is worth knowing now which parts are established (that lowering LDL by almost any route reduces events) and which are not (that this particular route is safe enough to do permanently).