VERVE-102: Can Lilly's new drug end heart attacks?
A new Eli Lilly drug lowered PCSK9 protein by up to 88 percent in 35 patients and cut LDL by 62 percent on average at the top dose. If the results hold up, it might be the closest thing we have to a heart attack vaccine for high-risk patients.
Presented at the European Atherosclerosis Society, May 25 2026. Published simultaneously in The New England Journal of Medicine. Trial: NCT06164730.
01The problem
Roughly 805,000 Americans have a heart attack every year. That is one almost every 40 seconds, and the mechanism is almost always the same.
- Cholesterol is necessary. Cell membranes are built from it, the brain runs on it, and hormones like testosterone, estrogen, and cortisol are made from it. The trouble is if there's too much of it in the blood for too long, it can cause serious damage.
- Cholesterol needs to be transferred but is not blood soluble, so the body generates LDL (low-density lipoprotein) to carry it around.
- The more LDL particles accumulate in the bloodstream the more the chances it can slip through the inner lining of your coronary arteries and lodge in the artery wall.
- Once inside, the LDL gets chemically damaged, and your immune system reads the damage as a threat and sends white blood cells in.
- The white blood cells eat the cholesterol, swell up, die, and leave their contents behind.
- Over years and decades, the deposit grows into a plaque: cholesterol, dead cells, scar tissue, and eventually calcium, packed into the wall of an artery you cannot see.
- Eventually the plaque's thin protective cap rips. A clot forms in seconds, seals the artery, and the heart muscle downstream stops getting oxygen. This is how you get a heart attack.
By lowering the LDL in the blood, you slow every step in the chain. Less raw material going into the wall, fewer plaques forming, fewer ruptures, fewer heart attacks. That relationship has held across every class of LDL-lowering drug ever tested, going back to the original statin trials in the 1990s. LDL is one of the most thoroughly validated causal risk factors in any disease.
Modern LDL drugs do not attack cholesterol directly. They go after PCSK9, a protein the liver makes that controls how much LDL stays in the blood. The liver clears LDL by grabbing it with receptors on its surface. PCSK9 binds to those receptors and drags them off to be destroyed, which means fewer receptors, less LDL pulled out of the blood, and higher levels in circulation. Block PCSK9 and the receptors stay put. The liver clears more LDL, and blood levels fall.
Statins come at it from a different angle. They throttle the liver's own cholesterol production, which forces the liver to compensate by pulling more LDL out of the blood. The endpoint is the same, more LDL receptors at work, but the path is indirect. The PCSK9 drugs go straight at the regulator.
The current solutions are cheap and effective, but they share an issue: you have to stay on the medication for the rest of your life.
| Therapy | Dosing | Pros | Limitations |
|---|---|---|---|
| Statins | Daily pill | Cheap and effective. | Discontinuation can run up to ~70% in the first year in real-world settings. Side effects for some (muscle aches, small uptick in diabetes risk). Many forget or do not feel an immediate benefit. |
| PCSK9 antibodies (Repatha, Praluent) | Injection every 2–4 weeks, indefinitely | More effective LDL lowering for some patients. | Requires frequent ongoing injections. Adherence can still be difficult. |
| Leqvio | Injection every 6 months | Less frequent dosing than PCSK9 antibodies. | Still requires staying on injections long term. |
This is why Lilly's new gene-editing drug, VERVE-102, is such a breakthrough. The drug has only been administered once, and patients have seen lower LDL levels last for 18 months and counting.
02The breakthrough
At the top dose, a single intravenous infusion of VERVE-102 dropped circulating PCSK9 protein by 88 percent and LDL cholesterol by 62 percent on average. The effect held for up to 18 months in the patients followed longest. That 62 percent figure is roughly what a monthly PCSK9 antibody injection achieves. The difference is that the antibody is a monthly injection, and this was one infusion.
Heart-2 is the Phase 1b, open-label, single-ascending-dose study that produced those numbers. Thirty-five adults across six dose cohorts, all of them already on the maximum standard-of-care therapy, all of them with HeFH or premature coronary disease, received one roughly four-hour intravenous infusion at doses from 0.30 to 1.00 mg/kg.
Hover any row for the cohort's safety notes. Source: Heart-2 interim analysis, NEJM May 25 2026.
03How it works
The liver makes a protein called PCSK9 whose job, in effect, is to take LDL receptors off the surface of liver cells. Fewer LDL receptors means less LDL gets pulled out of the blood. So if you turn PCSK9 down, you turn LDL clearance up, and circulating cholesterol drops. This is why people born with naturally low PCSK9 have low LDL for life and rarely have heart attacks.
VERVE-102 turns PCSK9 off at the gene. It does this with base editing, which is the gentler cousin of CRISPR: instead of cutting DNA and letting the cell stitch it back together, a base editor converts one letter of DNA into another. The drug itself is a packet of instructions wrapped in a lipid nanoparticle. Inside the nanoparticle is messenger RNA for the editor protein, and a guide RNA that tells the editor where to go.
The infusion delivers the nanoparticle to the liver. The liver cells read the mRNA, make the editor protein for a few days, and the guide RNA walks the editor to a single adenine in the PCSK9 gene. The editor chemically converts that A into a G. The single-letter swap creates a premature stop codon, which means the cell can no longer build a full PCSK9 protein. The gene is broken. The mRNA and the editor protein degrade within days, but the DNA edit is permanent, and when the liver cell eventually divides, the edit is inherited by the daughter cell.
A single adenine (A) inside the PCSK9 gene is chemically converted to guanine (G). The one-letter swap creates a premature stop codon, silencing the gene. DNA is never cut — distinct from traditional CRISPR/Cas9 double-strand breaks, and a lower-risk way to edit a hepatocyte you intend to leave functional for life.
First-generation LNPs rely on a protein called ApoE to enter liver cells through the LDL receptor, which is fine until your target patient is somebody whose LDL receptors are precisely the things that do not work. VERVE-102's LNP keeps that LDLR pathway and adds a second one: a GalNAc molecule that also binds ASGPR, a different receptor that sits on every hepatocyte regardless of LDL-receptor status. The clinically important point is the backup: ASGPR uptake works even when a patient's LDLRs do not. In animal studies at six months, the edit shows up almost entirely in liver tissue, at trace levels in adrenal glands and spleen, and effectively nowhere else.
PCSK9 editing measured across 26 tissues in 4 primates six months after a 3 mg/kg infusion. The drug is designed to be picked up almost exclusively by the liver — and it is.
Liver, adrenals, and spleen percentages are reported explicitly in the appendix text. Remaining tissues marked * are estimated from Figure S2's bars — all measured below 1%. Source: Figure S2, NEJM Heart-2 Supplementary Appendix.
04The catch
This is, after all, a Phase 1 trial. This trial so far only involves thirty-five patients. And it has only run for eighteen months. While the signs are extremely promising, it does not yet conclusively prove that the drug prevents heart attacks. PCSK9 lowering is a very well-validated proxy for cardiovascular risk reduction, but proxies are not outcomes.
The clinical trial for the predecessor compound, VERVE-101, paused enrollment in April 2024 after serious issues were found in liver enzymes and platelet production. Verve (pre-Lilly acquisition) attributed the problem to the lipid nanoparticle rather than to the editing mechanism, and they switched to the GalNAc-decorated LNP that VERVE-102 now uses. Heart-2's clean safety read so far has validated that pivot. For a drug with this long of a time horizon, though, it is still early days.
The original VERVE-101 was paused when one patient simultaneously developed Grade 3 elevations in liver enzymes (ALT) and a Grade 3 crash in platelets within four days of dosing. The redesigned LNP behind VERVE-102 was supposed to fix that. In the clinical trial for VERVE-102, across all six doses through Day 60, no cohort breached the clinically meaningful 3× ULN threshold — mild transient ALT elevations were noted in the NEJM abstract, but nothing dose-limiting — meaning that the new solution seems to have worked, at least so far.
Liver enzyme. Spikes if liver cells are stressed or damaged. 3× ULN is the standard clinical signal for drug-induced liver injury; VERVE-101's halting event went well past it.
A second liver enzyme. Moves with ALT when there's hepatic stress.
Waste product the liver clears. Rising bilirubin signals the liver isn't keeping up.
Cells that clot blood. The Grade 3 platelet crash in the same VERVE-101 patient who showed the ALT spike was the second signal that prompted the LNP redesign.
Source: Figures S4 and S5, NEJM Heart-2 Supplementary Appendix. Mean values across each dose cohort, approximated from published figures. The editorial point is the shape: no cohort breached the 3× ULN threshold.
The one patient who ended up in the ER
A 40-year-old man, 0.45 mg/kg cohort
He had premature coronary artery disease and was on the maximum dose of two oral cholesterol drugs (atorvastatin 80 mg, ezetimibe 10 mg). He also had a history of acid reflux and a sliding hiatal hernia — a stomach that occasionally slips up through the diaphragm.
He received 42 mg of total RNA on Day 1, with only a mild headache during the infusion. On Day 12 he developed an unrelated Grade 2 respiratory tract infection. On Day 16 he woke up choking. Acutely short of breath, pale, sweaty, blue around the lips. At the emergency room a chest X-ray showed a cloudy patch in his right lung; a CT scan ruled out a pulmonary embolism.
He was treated for aspiration pneumonitis — the lung injury that follows when stomach contents come up at night and are inhaled — with a bronchodilator, steroids, antibiotics (including amoxicillin), oxygen, IV fluids, prophylactic anticoagulant, and stomach-acid suppression. He improved rapidly overnight and went home the next day on oral antibiotics. The acute event resolved quickly; residual Grade 1 symptoms fully cleared by Day 40.
The investigator graded the event as Grade 3 and judged it not related to VERVE-102 — consistent with the patient's pre-existing reflux and hernia. It is the only serious adverse event described in detail in the trial's supplementary appendix.
Everyone else
Every adverse event the investigators judged related to VERVE-102, across all 35 patients. The most common is "infusion-related reaction" (fever, chills, or mild discomfort during the four-hour drip). Almost everything else is a single patient — only fatigue shows up twice.
Each column is dose in mg/kg with cohort size. MedDRA v26.1; events with onset after VERVE-102 infusion. Source: Table S2, NEJM Heart-2 Supplementary Appendix.
The catch
The permanence of this drug is also a substantial issue. The edit is one-and-done by design. Once a patient is dosed, their PCSK9 gene is broken in their liver cells for life. The same property that makes a single infusion enough is the same property that means if something goes wrong, there is no easy undo.
The risk is not cholesterol starvation. The drug does not stop the body from making cholesterol or shipping LDL to cells. PCSK9 is the protein that normally tells the liver to destroy its own LDL receptors. Break PCSK9, and those receptors stay on the surface, and the liver pulls more LDL out of the blood. Production continues. Delivery continues. The clearance rate just goes up. Repatha and Praluent, the PCSK9 antibodies that have been on the market for years, have driven LDL into single digits without obvious trouble, and people born with naturally broken PCSK9 look healthy. The biology is well-trodden.
The real long-tail risk is somewhat subtler. It is that PCSK9 might do something else in the body we have not yet figured out, and a patient who has it edited out at fifty might quietly differ from one who was born without it. If that kind of safety problem shows up in five years, it is going to show up in the patients whose PCSK9 gene is already broken. Phase 2 and Phase 3 will start to tell us whether that risk is theoretical or real.