One CRISPR Infusion Cut LDL 52.5%. Twelve Months Later It Has Not Come Back

52.5 percent lower, sustained for 365 days without a booster, second dose, or daily pill to remember, is what happened to LDL cholesterol in people who received a single 0.8 mg per kilogram infusion of CTX310 and were followed for a full year.

For anyone who has watched a statin prescription gather dust in a medicine cabinet, that durability matters more than the headline percentage because adherence, not potency, is the dominant failure mode in preventive cardiology. Cleveland Clinic reported August 28 that its Phase 1 first-in-human trial of CTX310, an investigational lipid nanoparticle-encapsulated CRISPR-Cas9 therapy developed by CRISPR Therapeutics AG in Zug, Switzerland, showed sustained lipid lowering at one year in 15 patients with refractory dyslipidemia who had failed maximally tolerated therapy. Results were presented as late-breaking science at the European Society of Cardiology Congress 2026 and simultaneously published as a letter in the New England Journal of Medicine titled Durability of CRISPR-Cas9 Gene Editing Targeting ANGPTL3 with CTX310.

CTX310 is deceptively simple in concept, even if execution is complex, because it carries Cas9 mRNA and a single guide RNA targeting ANGPTL3 inside a lipid nanoparticle designed for selective hepatic uptake, enters hepatocytes, cuts the ANGPTL3 locus, and lets error-prone repair create permanent loss of function that persists as hepatocytes turn over. ANGPTL3 is a hepatically secreted inhibitor of lipoprotein lipase and endothelial lipase, and losing it lowers both LDL and triglycerides through derepressed lipase activity that accelerates clearance of triglyceride-rich lipoproteins. People born with naturally occurring loss-of-function variants prove the point with decades of human data. Dewey et al. 2017 found 13 LoF variants in 58,335 DiscovEHR participants, about 1 in 250 people of European heritage carries a single copy, and carriers have roughly 39 percent lower odds of coronary artery disease with 70 percent lower plasma triglycerides and LDL when homozygous, with no apparent detrimental consequences to health.

Early data looked promising but brief, which is typical for first-in-human gene editing where safety is the primary endpoint and durability is unknown. Laffin et al.'s initial Phase 1 report in November 2025, published as N Engl J Med 2026;393:2119-2130, enrolled 15 adults across Australia, New Zealand, and the UK with a median age of 53, 87 percent men, median LDL 155 mg per deciliter, median triglycerides 192 mg per deciliter, 40 percent with established atherosclerotic cardiovascular disease, 40 percent with familial hypercholesterolemia, all with persistent dyslipidemia despite maximally tolerated lipid-lowering therapy. Dosing ranged from 0.1 to 0.8 mg per kilogram lean body weight after corticosteroid and antihistamine pre-treatment, and at 60 to 90 days the highest dose produced a 48.9 percent LDL reduction and 55.2 percent triglyceride reduction that immediately raised the durability question every reviewer asked.

Durability was the question that mattered because gene editing is irreversible, and if effects wane you cannot re-dose the same guide without immunogenicity risk against Cas9 or the LNP, while if they persist you have a genuine one-and-done that bypasses adherence entirely. Twelve-month follow-up answers that question, at least for year one, with data that will be scrutinized for years.

Highest dose cohort at 12 months showed LDL minus 52.5 percent from baseline and triglycerides minus 47.8 percent, with no dose-limiting toxicities, no new serious adverse events, and no elevations in liver-function measures beyond a transient aminotransferase bump previously reported in one participant shortly after treatment that resolved without intervention. Two serious adverse events previously reported, including one death months after the lowest dose and one preexisting condition, were deemed unrelated by investigators after adjudication. FDA-mandated 15-year long-term safety follow-up continues, as recommended for all gene-editing therapies, because rare oncogenic events and clonal expansions can appear years later as the liver regenerates.

Dose response is not linear

Table 1 shows the initial dose-response at 60 to 90 days drawn from ACC and NEJM summaries, and it reveals a threshold effect rather than a smooth dose-response curve that would be easy to optimize.

Notice the anomaly that will drive Phase 1b dose selection. The 0.7 mg per kilogram group achieved the deepest ANGPTL3 knockdown at minus 79.7 percent but the weakest LDL response at minus 21 percent among effective doses, while 0.6 and 0.8 mg per kilogram produced similar LDL effects despite different ANGPTL3 suppression levels. That heterogeneity likely reflects hepatic steatosis, inflammatory status, LNP uptake pathways, and germline variation in lipid metabolism, as Mondal and Misra noted March 2026 in their review of in vivo ANGPTL3 editing. In other words, fatty liver may edit worse, and patients with fatty liver are precisely those with metabolic syndrome who need this most, which creates a selection paradox for trial enrichment.

Original calculation: what does 52.5 percent LDL cut mean for heart attack risk

No outcomes data exist for CTX310 yet because Phase 1 measured surrogates only, including LDL cholesterol, triglycerides, non-HDL cholesterol, apolipoprotein B, apolipoprotein C-III, and remnant cholesterol, with no major adverse cardiovascular events adjudicated. We can project using established epidemiology with full transparency about assumptions and limitations, because that projection is more useful than waiting five years for outcomes while patients make decisions today.

Method uses Cholesterol Treatment Trialists' Collaboration meta-analysis of 26 statin trials published in Lancet 2010 and 2012, which found each 1 mmol per liter LDL reduction, equivalent to 38.7 mg per deciliter, yields about 22 percent reduction in major vascular events, with a log-linear relationship that appears consistent across mechanisms even though ANGPTL3 inhibition has not been tested in outcomes trials and may differ. Inputs from the CTX310 trial are median LDL 155 mg per deciliter with 52.5 percent reduction equaling 81.4 mg per deciliter absolute reduction, and dividing 81.4 by 38.7 equals 2.10 mmol per liter equivalent, then multiplying 2.10 times 22 percent equals 46.2 percent relative risk reduction from LDL lowering alone under this model.

Triglyceride contribution adds independent risk reduction that is less well quantified but clinically meaningful. Rosenson et al. NEJM 2024 zodasiran, an RNAi therapeutic targeting ANGPTL3 for mixed hyperlipidemia, and Copenhagen General Population Study data suggest triglyceride-rich lipoprotein reduction adds roughly 8 to 12 percent residual risk reduction after LDL adjustment for 40 to 50 percent triglyceride lowering. Taking the midpoint 10 percent and applying a conservative 50 percent independence discount to avoid double-counting overlapping pathways, combined projected major adverse cardiovascular event reduction equals about 50 to 52 percent relative risk, which aligns directionally with human genetics where Dewey 2017 reported carriers of ANGPTL3 loss-of-function have 34 to 39 percent lower coronary risk despite only 15 to 20 percent lower LDL lifelong because our pharmacologic knockdown is larger than natural heterozygosity. This is not a clinical claim. It is a model that Phase 2 and 3 must test with hard outcomes.

Adherence math changes everything

Here is a calculation nobody ran in ESC press coverage, yet it determines population impact more than molecule potency. Benner et al. JAMA 2002, cited in Mondal and Misra 2026, found nearly 50 percent of elderly patients discontinued statin therapy within one year of initiation, and Cleveland Clinic cardiologist Luke Laffin, first author of CTX310, noted in his ESC press conference that up to 50 percent of people prescribed statins or PCSK9 inhibitors are not taking them at one year, with adherence waning over time as refills lapse.

Statins in ideal conditions lower LDL about 50 percent with high-intensity dosing, but real-world 12-month population mean accounting for 50 percent discontinuation where discontinued equals 0 percent reduction is 0.5 times 50 percent plus 0.5 times 0 percent equals 25 percent mean reduction, while even allowing partial adherence where half of discontinuers still take intermittently achieving 30 percent reduction yields population mean of 0.5 times 50 plus 0.25 times 30 plus 0.25 times 0 equals 32.5 percent. CTX310 at 0.8 mg per kilogram achieved 52.5 percent in per-protocol analysis, but because it is a one-time infusion there is no discontinuation, so population mean equals per-protocol mean minus rare non-responders, and assuming 10 percent non-response gives 0.9 times 52.5 equals 47.3 percent.

Ratio of 47.3 divided by 25 equals 1.89 times more effective at population level at 12 months under full discontinuation model, or 47.3 divided by 32.5 equals 1.46 times under partial adherence model, which means despite similar molecule-level efficacy the one-time therapy is substantially more effective where it matters, which is in the community rather than the clinic. Cost break-even versus PCSK9 monoclonal antibodies at about 6,000 dollars per year US net price, totaling 180,000 dollars undiscounted over 30 years, occurs at about 16.7 years without discounting and about 22 years at 3 percent discount rate if CTX310 is priced at 100,000 dollars one-time, a plausible analyst estimate for in vivo LNP editing far simpler than ex vivo Casgevy at 2.2 million dollars. Versus generic statins at 48 dollars per year it never breaks even on drug cost alone, but if it prevents one major adverse cardiovascular event per 12 treated over 10 years, assuming 80,000 dollars per hospitalization plus rehabilitation, it breaks even on event costs alone even before quality-adjusted life years.

Editing efficiency inference

Assuming liver ANGPTL3 protein reduction correlates linearly with hepatocyte editing efficiency, which Mondal and Misra's Figure 1 conceptual model supports with LNP uptake into hepatocytes leading to nuclear Cas9 delivery and locus cleavage, roughly 75 percent hepatocyte editing yields roughly 50 percent LDL reduction in this cohort, and each 1 percent ANGPTL3 knockdown corresponds to about 0.67 percent LDL reduction calculated as 52.5 divided by 78.5 average high-dose ANGPTL3 knockdown, which is useful for dose optimization and patient stratification in Phase 1b where hepatic steatosis may impair LNP uptake.

Limitations

This analysis relies on publicly reported summary data and has not seen individual patient trajectories, Figure 1 error bars, Table S2 lipid biomarker distributions, or Supplementary Appendix adverse event details for the 12-month letter, which remain behind the NEJM paywall as of August 28, so we could not assess individual variation or outlier-driven means. We also did not have access to individual-level hepatic imaging or inflammatory biomarker data that would clarify heterogeneity.

Sample size is 15 total with highest dose 4 patients at initial report and fewer with complete 12-month data, so all results are preliminary and confidence intervals are wide. No clinical outcomes were measured, only surrogate lipids including LDL cholesterol, triglycerides, non-HDL cholesterol, apolipoprotein B, apolipoprotein C-III, and remnant cholesterol, which are validated surrogates but not outcomes.

Follow-up is 12 months, while FDA requires 15-year long-term safety monitoring for gene-editing therapies because hepatotoxicity profiles beyond transient aminotransferase elevation, immune activation to Cas9 or LNP, low-frequency off-target edits detectable only by ultra-deep sequencing, and potential pro-oncogenic consequences from clonal hepatocyte repopulation in steatotic or inflamed livers remain unknown. These risks are particularly salient because liver injury and regeneration can amplify rare mutational events over years, and current bulk sequencing has limited sensitivity for rare, cell-restricted edits.

Efficacy varied substantially, including poor LDL response despite excellent ANGPTL3 knockdown in the 0.7 mg per kilogram cohort, and the mechanism of that heterogeneity is unexplained, which raises concerns about treating patients with metabolic syndrome whose fatty livers may edit inefficiently. Population studied was refractory dyslipidemia, median age 53, 87 percent men, 40 percent ASCVD, 40 percent FH, not general primary prevention, so younger patients who would benefit most from lifetime editing were not studied and may have different risk-benefit profiles.

Irreversibility is inherent, with no reversal agent for ANGPTL3 editing, and if long-term harm emerges, even though natural LoF carriers appear healthy with no apparent detrimental consequences, acute knockout at age 53 in a fatty liver is not identical to lifelong LoF from conception with developmental compensation and alternative pathway upregulation.

CTT-based projection uses statin trial data, not ANGPTL3 editing, and may not generalize, because we assumed linear 22 percent per 38.7 mg per deciliter relationship that is conservative but unvalidated for this mechanism and could overestimate benefit if ANGPTL3 editing has different pleiotropic effects.

Strongest counterargument

The strongest case against one-time ANGPTL3 editing is not that editing is unsafe in a sensational sense, but that we already have a cheap, reversible, effective solution, and this irreversibly solves adherence by removing patient autonomy and pharmacologic flexibility. Statins cost 4 dollars per month, ezetimibe costs 10 dollars per month, and inclisiran, a twice-yearly siRNA against PCSK9, achieves about 50 percent LDL reduction with no DNA editing and reversible dosing that is FDA-approved today and solves adherence nearly as well as editing without genome alteration.

ANGPTL3 loss of function in humans appears benign, but those people had loss of function from conception with developmental compensation, whereas acutely knocking out ANGPTL3 at age 53 in a metabolically diseased liver is not the same biology and could unmask liabilities that lifelong carriers never experience. The trial's own heterogeneity, where 0.7 mg per kilogram achieved excellent ANGPTL3 knockdown but poor LDL response, hints that liver health confounds editing efficiency, so precisely the patients with metabolic syndrome who need this most may edit worst, creating a paradox where those with greatest need have least benefit.

The 15-year FDA follow-up requirement exists because AAV gene therapy taught us that rare oncogenic events and clonal expansions can appear years later, often in regenerating tissue, and a one-time infusion that permanently alters your liver genome to save 6,000 dollars per year in PCSK9 costs trades a known, manageable adherence problem for an unknown, unmanageable genomic liability that cannot be recalled. If adherence is the problem, fixing adherence through 90-day prescriptions, single-pill combinations, and behavioral nudges may be more ethical than editing the genome, and that argument deserves serious engagement rather than dismissal as technophobia, because it is answered only by larger trials, longer follow-up, ultra-deep sequencing for off-targets, and rigorous cost-effectiveness analysis that includes the price of permanence.

What you can do

If you have severe hypercholesterolemia or hypertriglyceridemia refractory to statins, ezetimibe, and PCSK9 inhibitors and you are in the United States, UK, Australia, or New Zealand, Phase 1b trials for CTX310 are ongoing in US and ex-US sites with a severe hypertriglyceridemia cohort update expected in the second half of 2026 per CRISPR Therapeutics. Search clinicaltrials.gov for CTX310 and discuss referral with your lipidologist, but do not seek this clinically because it is Phase 1 only with no outcomes data and requires corticosteroid pre-treatment.

If you are a clinician, do not offer this outside a trial because there is no outcomes data, no dose optimization, and FDA 15-year monitoring implies unknown long-term risk that you cannot counsel patients on. For patients struggling with daily pills today, inclisiran twice-yearly already solves adherence without genome editing and is available now through standard lipid clinics.

If you are at elevated ASCVD risk but struggle with adherence, talk to your doctor about adherence barriers, 90-day prescriptions, and single-pill combinations before considering experimental editing, because statins remain the most cost-effective intervention in preventive cardiology with decades of safety data. Statins remain foundational.

If you are a payer or health technology assessment body, model break-even at 100,000 dollars one-time price versus PCSK9 chronic therapy, demand major adverse cardiovascular event outcomes before coverage, and recognize this is not a statin replacement today even if press releases suggest otherwise. Signal to watch is Phase 1b severe hypertriglyceridemia cohort in H2 2026, because if triglyceride lowering exceeds 60 percent sustained, that unlocks pancreatitis prevention indication where no good chronic therapy exists and where one-time editing could be transformative even at high price.

The Bottom Line

A single IV infusion of lipid nanoparticle CRISPR-Cas9 targeting ANGPTL3 lowered LDL 52.5 percent and triglycerides 47.8 percent at 12 months in the highest dose group, with no treatment-related serious adverse events in 15 patients, representing the first demonstration of one-year durability for in vivo CRISPR editing in cardiometabolic disease. It is also 15 people, 12 months, surrogate endpoints, and no reversal agent, so small numbers can mislead, durability can still fade, and rare events have not had time to appear, which means enthusiasm must be tempered with appropriate scientific caution.

What changed August 28 is not that cholesterol editing works in principle, which mouse data showed in 2018, but that it lasts 12 months in humans, which is the inflection point that justifies larger trials and investment in delivery optimization. For a field that has been promising one-and-done since base editing papers in Circulation, lasting a full year in humans is the signal that this approach may finally move from rare monogenic disease toward broadly prevalent conditions where adherence dominates outcomes. Next step is seeing whether 12 months becomes 24, and whether 15 people becomes 1500 with outcomes.

Sources: Cleveland Clinic Newsroom 2026-08-28; NEJM Durability Letter DOI 10.1056/NEJMc2609825; ACC Phase 1 Summary; Mondal and Misra Signal Transduct Target Ther 2026; Dewey et al. NEJM 2017; Benner JAMA 2002; CRISPR Therapeutics Press Release

Related: CRISPR Cas12a2 Shreds Chromatin When It Sees p53 — Cancer Cells Die in Hours · Prime Editing Triple Breakthrough: 49 Percent Liver Editing In One Dose · Penn Engineered Nanoparticles That Edit Blood Stem Cells Without Leaving the Body