Does Verve 102 Fix the Cholesterol Damage From Steroids?

In August 2026 the New England Journal of Medicine published something that sounds like science fiction. Thirty-five people were given a single intravenous infusion that permanently rewrote one letter of their DNA. At the highest dose their LDL cholesterol fell 62% and stayed down.

No daily tablet. No fortnightly injection. One infusion, one edited gene, done.

If you have ever watched your lipid panel fall apart six weeks into a cycle, you can see where this is going. So let me answer it up front, because the answer is more interesting than either the hype or the obvious cynical take.

What the Heart-2 trial actually showed

The drug is VERVE-102, made by Verve Therapeutics, which Eli Lilly acquired outright in 2025. The trial is Heart-2, a Phase 1 open-label single-ascending-dose study (Vafai et al., 2026).

The mechanism is genuinely elegant. An mRNA encoding an adenine base editor, plus a guide RNA aimed at the PCSK9 gene, wrapped in a lipid nanoparticle decorated with GalNAc so it homes to liver cells. Once inside, it changes a single A to a G, creating a premature stop codon. The gene is switched off. Because liver cells turn over slowly and DNA edits do not undo themselves, that is permanent.

PCSK9 is a protein that destroys LDL receptors on your liver cells. Fewer receptors means less LDL pulled out of your blood. Disable PCSK9 and the receptors survive longer, so your liver clears LDL faster. It is the same target as Repatha, approached from a completely different direction.

The numbers

Thirty-five participants across six dose cohorts, from 0.3 to 1.0 mg/kg.

At the top dose that 62% meant LDL-C falling roughly 78 mg/dL, from about 129 down to about 51.

Worth being precise about sourcing here. The peer-reviewed paper reports the two extremes, 51% to 88% for PCSK9 and 9% to 62% for LDL-C. The four middle numbers come from Lilly and Verve's own disclosures of the same interim dataset rather than from the published abstract. They are consistent with the paper's endpoints, but they are company figures.

The wobble nobody mentions

Look at that table again. The 0.7 mg/kg cohort came in at 33%, below both 0.45 and 0.6. The dose-response is not a clean line.

The paper does not comment on this, and no independent commentary I could find addresses it either. My read, and I want to flag clearly that this is my inference rather than something the investigators argue: the cohorts are tiny. The protocol specified three to nine participants per dose, and one interim disclosure noted as few as two people dosed at 0.7 mg/kg. Even at the top dose, individual responses ranged from 45% to 79%. With numbers that small, a mean landing below its neighbours is exactly what chance produces.

This is not a scandal. It is what Phase 1 data looks like before anyone has characterised the curve properly, and it is a good reason to hold the headline 62% loosely.

Safety looked clean. No dose-limiting toxicities, no treatment-related serious adverse events, no deaths, no withdrawals. Mild to moderate infusion reactions in 7 of 35. Transient ALT elevations that resolved. One serious adverse event, aspiration pneumonitis in someone with pre-existing reflux, judged unrelated.

Why this is biologically plausible in the first place

The strongest argument for deliberately breaking a gene is that nature already ran the experiment.

In 2006, Cohen and colleagues looked at people carrying naturally occurring nonsense mutations in PCSK9, meaning they had lived their whole lives with the gene partly or wholly disabled (Cohen et al., 2006). Among Black participants in the ARIC cohort, 2.6% carried such mutations. They had 28% lower LDL and an 88% reduction in coronary heart disease risk. A separate variant in 3.2% of white participants gave 15% lower LDL and a 47% risk reduction.

These were not people on a drug for a few years. These were people followed for decades who had been born this way, and they were fine. That is the bet VERVE-102 is making deliberately.

How a permanent edit differs from Repatha

Readers keep framing this as "a stronger Repatha." It is not. It acts on a different biological layer entirely.

Statins and ezetimibe work upstream of the receptor. The antibodies mop up the protein. Inclisiran stops the protein being made. VERVE-102 deletes the instruction to make it.

Everything above VERVE-102 in that table washes out if you stop. VERVE-102 does not have a stop.

Why it would not touch your HDL

Here is the mechanism argument, and it is solid.

Anabolic steroids do not damage your lipids through PCSK9. They do it by upregulating hepatic triglyceride lipase, the enzyme that catabolises HDL particles. The cleanest demonstration is a 1987 time-course study on oral stanozolol where the researchers measured both across the first week (Applebaum-Bowden et al., 1987).

Hepatic lipase activity rose 62% by day one, 161% by day two, 230% by day three. HDL had not moved at all for the first two days. Then it fell 14% by day three and 39% by day seven, with HDL2, the larger cardioprotective subfraction, down 71%.

The enzyme spikes first. The HDL crash follows. That lag is the whole argument for causation rather than coincidence, and it points at a pathway PCSK9 has nothing to do with.

Orals versus injectables, in one trial

Thompson and colleagues ran 11 weight lifters through both 6mg/day oral stanozolol and 200mg/week injectable testosterone enanthate, six weeks each, in a crossover design so every man acted as his own control (Thompson et al., 1989):

• Stanozolol: HDL down 33%, HDL2 down 71%, ApoA-I down 40%, LDL up 29%
• Testosterone enanthate: HDL down only 9%, confined to HDL3, and LDL down 16%

Same study, same duration, opposite LDL directions. This is why the blanket claim "steroids raise your cholesterol" is too crude to be useful.

It is not purely an oral phenomenon though. Supraphysiological testosterone at 600mg/week raised hepatic lipase more than 60% over three weeks and shifted LDL toward smaller, denser particles (Herbst et al., 2003). The effect is androgen-dose-dependent, and Bhasin's dose-response work found HDL declining linearly with testosterone dose (Bhasin et al., 2001). Orals are worse, not uniquely guilty.

Real-world stacks are worse than either. Two bodybuilders followed weekly through a cycle hit HDL troughs 69% and 72% below baseline, with LDL peaking 144% and 156% above, and ApoA-I down 84% and 91% (Lajarin et al., 1996). That is n=2, so treat it as illustration rather than evidence, but it matches what shows up in practice.

Interestingly, nandrolone decanoate alone at 200mg/week in a placebo-controlled arm did not significantly move most lipid parameters, only Lp(a) (Hartgens et al., 2004). Not every injectable behaves like testosterone.

The complication that makes this argument weaker than it sounds

I could stop there and you would leave thinking your crashed HDL is a ticking clock that gene editing cannot fix. That would be tidier than the truth.

Voight and colleagues ran a Mendelian randomisation study with two independent analyses (Voight et al., 2012). Observationally, higher HDL tracked with lower risk, odds ratio around 0.62, exactly what everyone expects. But genetically raised HDL showed no protection in either analysis: an odds ratio of 0.99 across 20,913 heart attack cases using a single HDL-raising variant, and 0.93 across up to 12,482 cases using a 14-variant score. Both are statistically indistinguishable from no effect. The parallel LDL analysis held up fine.

In other words, the best available human genetic evidence says HDL-C is largely a passenger, not a driver. This is why CETP inhibitors and niacin raised HDL beautifully and failed to save anyone in outcome trials.

So "a PCSK9 edit will not fix your HDL" is true, and it matters less than it sounds. Low HDL on cycle is best read as a marker that hepatic and endocrine machinery is being disrupted, rather than the damage itself. The damage is more likely happening through raised ApoB particle count, elevated blood pressure and left ventricular hypertrophy.

Which raises an obvious question: if ApoB is the thing that matters, and a PCSK9 edit lowers ApoB, does the objection collapse?

The trial did not measure the thing you care about

No, and here is why.

Heart-2 reported PCSK9 and LDL-C. It did not report HDL-C. It did not report ApoB. It did not report Lp(a). It did not report triglycerides.

Not "reported no change." Not measured, or at least not published in the interim results or any coverage of them.

So the claim "it would fix your ApoB" is an extrapolation from mechanism and from the PCSK9 antibody literature, not something this trial demonstrated. For an audience that should be tracking ApoB over LDL-C anyway, because anabolic use shifts particles toward the small dense variety that make LDL-C misleading, that gap is not academic. The trial measured the number that means less, in people unlike you, and did not measure the number that means more.

What if you are 28 with LDL of 4.1 mmol/L on cycle?

This is the scenario people actually have, so let us walk it rather than talk in the abstract.

Compare yourself to a Heart-2 participant. Mean age 52. Fifty-seven percent had heterozygous familial hypercholesterolaemia, a genetic condition, and another 26% had that plus premature coronary artery disease. Seventy-one percent were already on high-intensity statins, 43% on ezetimibe as well. Despite all of that their LDL-C was still averaging 129 mg/dL, which is why they qualified.

That is a person whose liver has been mishandling cholesterol since birth and who has run out of drug options.

You are a 28 year old whose LDL climbed because you introduced a compound eight weeks ago. Those are not the same problem wearing different clothes. One is a permanent genetic defect. The other is a drug effect with a known off switch.

And the off switch works. In 56 people followed for up to a year after stopping injectable AAS, the cholesterol profile normalised within six months, even while gonadotropin suppression persisted longer in some (Gårevik et al., 2011). Recovery is dose and duration dependent, and one case series found incomplete normalisation at 13 weeks, so it is not instant. But it happens on its own.

An irreversible fix for a reversible problem

This is the actual argument, and it does not depend on the HDL mechanism at all.

The first version had real safety signals. VERVE-101 targeted the same gene by the same approach. Enrollment was halted in April 2024 after a participant developed Grade 3 ALT elevation and Grade 3 thrombocytopenia within days of infusion. Both resolved, and Verve's investigation pointed at the lipid nanoparticle rather than the editing payload, since the delivery vehicle alone produced the same effect in animals and even a catalytically dead guide RNA did it. They changed the delivery chemistry and moved to VERVE-102, whose safety so far looks clean. Both things are true. Neither cancels the other.

Somatic does not mean reversible. The GalNAc targeting means hepatocytes, not germ cells, so this is not heritable and your children are unaffected. People conflate that with "safe" or "undoable." It is neither. It just is not inherited.

Lifelong differs from mid-life. The natural-carrier reassurance is real and strong. But those carriers had partial loss of function present from conception, so their entire development happened at that setting. VERVE-102 proposes an acquired, near-complete knockout, up to 88% protein reduction, in a liver that spent three decades with normal PCSK9. Whether those are equivalent has not been studied.

There is one unresolved signal worth knowing. Randomised trials of PCSK9 antibodies have not shown increased new-onset diabetes (Sabatine et al., 2017). But Mendelian randomisation studies, which model lifelong PCSK9 reduction, have fairly consistently suggested a diabetes association. A permanent edit resembles the genetic scenario more than the drug scenario. Unresolved is not the same as dangerous, but it is the kind of thing you want settled before signing up at 28.

Off-target editing has not been assessed clinically. Verve's preclinical work reported no meaningful off-target editing across roughly 6,000 candidate sites in cultured human liver cells. That is reassuring and it is not a clinical readout. The interim results report no off-target sequencing in dosed humans.

The statin adherence claim, corrected

You will see this everywhere alongside this story: "about half of statin users quit within a year, so a one-time treatment would be transformative."

The real numbers are close but not that. Pooling 82 studies and more than 3 million statin users aged 65 and over, 48.2% were nonadherent and 23.9% discontinued entirely within the first year (Ofori-Asenso et al., 2018).

So roughly half do not take them properly, and roughly a quarter stop altogether. "Half quit" merges those two into a punchier number that is not what the data says. The adherence problem is real and it does make a one-and-done therapy appealing. It just is not a coin flip.

What to do instead

Nothing here is available, so this is the practical half.

Track ApoB, not just LDL-C. Anabolic use shifts LDL toward small dense particles, which makes LDL-C an unreliable proxy for how many atherogenic particles you actually have. Guideline bodies already treat ApoB as equal or preferable. No outcome trial has validated this in AAS users specifically, so it is extrapolation, but it is well-founded extrapolation.

Know which compounds you are choosing. The 33% versus 9% HDL split between oral stanozolol and injectable testosterone is a choice you make at the start of a cycle, not a fate. Superdrol and other 17-alpha-alkylated orals are believed to behave like stanozolol here, though their published literature is thinner and mostly case-level, so I would not quote precise numbers for them.

Use the reversible tools. Statins, ezetimibe and, where genuinely warranted and prescribed, PCSK9 antibodies all wash out. Our ranked cholesterol supplements piece covers what actually moves a marker and what does not, and how to protect your cholesterol on steroids covers the on-cycle protocol properly.

Do not ignore blood pressure. If HDL is largely a marker, blood pressure and left ventricular hypertrophy are doing more of the real damage. Lowering blood pressure on cycle is the higher-yield intervention.

If you are curious about your genetics, our DNA and drug metabolism piece covers APOE and SLCO1B1, which are relevant to LDL clearance and statin tolerance respectively, and which you can actually act on today.

Key takeaways

• VERVE-102 permanently disables PCSK9 in liver cells with a single infusion. In 35 people, LDL-C fell 9% to 62% by dose, and PCSK9 51% to 88%.
• The dose-response is not clean. The 0.7 mg/kg cohort came in below two lower doses, which is what tiny Phase 1 cohorts do.
• Durability of at least a year is real but concentrated in the lower-dose cohorts. The 62% responders have the shortest follow-up.
• Anabolic steroids crash HDL via hepatic lipase, a pathway PCSK9 editing does not touch. Hepatic lipase rises 230% by day three on stanozolol, before HDL moves.
• But HDL-C is probably not causal. Genetically raised HDL confers no protection against heart attack, so the HDL objection matters less than it first appears.
• The trial did not measure HDL, ApoB, Lp(a) or triglycerides at all, so any claim about what it would do to those is extrapolation.
• Trial participants averaged 52 with inherited high cholesterol on maximal statins. Your cycle-driven dyslipidaemia resolves within about six months of stopping.
• The predecessor, VERVE-101, was halted after Grade 3 liver and platelet abnormalities. The delivery vehicle was changed and VERVE-102 looks clean so far. Both facts belong in the picture.
• It is not approved, not purchasable, and realistically years from availability. Phase 2 was slated to begin at the end of 2026.

---

References

1. Vafai, S. B., Täubel, J., Ashdown, T., Patel, R. S., Cegla, J., Soran, H., Gaudet, D., Brunham, L. R., Newby, D. E., Nicholls, S. J., Tardif, J.-C., Humphries, S. E., Khera, A. V., & Kathiresan, S. (2026). In vivo base editing of PCSK9 with VERVE-102 for hypercholesterolemia. New England Journal of Medicine, 395(7), 648-659. PubMed

2. Cohen, J. C., Boerwinkle, E., Mosley, T. H., & Hobbs, H. H. (2006). Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. New England Journal of Medicine, 354(12), 1264-1272. PubMed

3. Applebaum-Bowden, D., Haffner, S. M., & Hazzard, W. R. (1987). The dyslipoproteinemia of anabolic steroid therapy: Increase in hepatic triglyceride lipase precedes the decrease in high density lipoprotein2 cholesterol. Metabolism, 36(10), 949-952. PubMed

4. Thompson, P. D., Cullinane, E. M., Sady, S. P., Chenevert, C., Saritelli, A. L., Sady, M. A., & Herbert, P. N. (1989). Contrasting effects of testosterone and stanozolol on serum lipoprotein levels. JAMA, 261(8), 1165-1168. PubMed

5. Herbst, K. L., Amory, J. K., Brunzell, J. D., Chansky, H. A., & Bremner, W. J. (2003). Testosterone administration to men increases hepatic lipase activity and decreases HDL and LDL size in 3 wk. American Journal of Physiology-Endocrinology and Metabolism, 284(6), E1112-E1118. PubMed

6. Bhasin, S., Woodhouse, L., Casaburi, R., Singh, A. B., Bhasin, D., Berman, N., Chen, X., Yarasheski, K. E., Magliano, L., Dzekov, C., Dzekov, J., Bross, R., Phillips, J., Sinha-Hikim, I., Shen, R., & Storer, T. W. (2001). Testosterone dose-response relationships in healthy young men. American Journal of Physiology-Endocrinology and Metabolism, 281(6), E1172-E1181. PubMed

7. Lajarin, F., Zaragoza, R., Tovar, I., & Martinez-Hernandez, P. (1996). Evolution of serum lipids in two male bodybuilders using anabolic steroids. Clinical Chemistry, 42(6), 970-972. PubMed

8. Hartgens, F., Rietjens, G., Keizer, H. A., Kuipers, H., & Wolffenbuttel, B. H. (2004). Effects of androgenic-anabolic steroids on apolipoproteins and lipoprotein (a). British Journal of Sports Medicine, 38(3), 253-259. PubMed

9. Gårevik, N., Strahm, E., Garle, M., Lundmark, J., Ståhle, L., Ekström, L., & Rane, A. (2011). Long term perturbation of endocrine parameters and cholesterol metabolism after discontinued abuse of anabolic androgenic steroids. Journal of Steroid Biochemistry and Molecular Biology, 127(3-5), 295-300. PubMed

10. Voight, B. F., Peloso, G. M., Orho-Melander, M., Frikke-Schmidt, R., Barbalic, M., Jensen, M. K., et al. (2012). Plasma HDL cholesterol and risk of myocardial infarction: A mendelian randomisation study. The Lancet, 380(9841), 572-580. PubMed

11. Ofori-Asenso, R., Jakhu, A., Zomer, E., Curtis, A. J., Korhonen, M. J., Nelson, M., Gambhir, M., Tonkin, A., Liew, D., & Zoungas, S. (2018). Adherence and persistence among statin users aged 65 years and over: A systematic review and meta-analysis. Journals of Gerontology: Series A, 73(6), 813-819. PubMed

12. Sabatine, M. S., Giugliano, R. P., Keech, A. C., Honarpour, N., Wiviott, S. D., Murphy, S. A., et al. (2017). Evolocumab and clinical outcomes in patients with cardiovascular disease. New England Journal of Medicine, 376(18), 1713-1722. PubMed

13. Ray, K. K., Wright, R. S., Kallend, D., Koenig, W., Leiter, L. A., Raal, F. J., et al. (2020). Two phase 3 trials of inclisiran in patients with elevated LDL cholesterol. New England Journal of Medicine, 382(16), 1507-1519. PubMed