Writing in JAMA Internal Medicine, University of Michigan physicians argue that treatments designed to be permanent need a higher safety bar when the alternative is a pill someone can simply stop taking.

Gene therapy has arrived in the clinic. Treatments that replace, silence or edit genes are approved and in use, mostly for rare diseases such as spinal muscular atrophy and sickle cell disease, where the genetic cause is known, the disease is severe, and the alternatives are poor.

The next frontier is different. Research is moving toward common conditions: high cholesterol, heart disease, obesity, diabetes. In a commentary in JAMA Internal Medicine, two cardiologists at the University of Michigan Health Frankel Cardiovascular Center argue that the shift requires rethinking how such treatments are evaluated.

“As therapies move from rare diseases to common conditions, the safety bar must rise accordingly,” said Venkatesh L. Murthy, who wrote the piece with Eric D. Smith.

The arithmetic of rarity

The authors’ central point turns on numbers. More than 6,000 rare genetic disorders exist, and only a fraction have treatable targets. A therapy for one of them might be given to a few hundred or a few thousand people, where a serious side effect occurring in one patient per thousand would be detected and weighed against a devastating disease.

A treatment for high cholesterol could be given to millions. At that scale, the same rate produces thousands of injuries, and the condition being treated is one for which effective, inexpensive drugs already exist.

Murthy’s argument is that this changes the comparison entirely. Common conditions are harder to justify precisely because existing treatments work well and can be adjusted or stopped. Statins can be titrated, switched or discontinued if a patient develops a problem. A permanent edit to the genome offers no such option.

What ‘one-time’ means

Smith raises a complication in the way these therapies are described. Many genomic therapies are designed as one-time treatments, but their effects may not last indefinitely, and the delivery mechanism can make a second attempt impossible.

Gene therapies are typically delivered using modified viruses, which carry genetic material into cells. The immune system can respond to the viral vector, and that response can block a second dose. A patient whose treatment wears off may have no route to another.

Current therapies also target somatic cells, the body’s ordinary cells, rather than germline cells, so the changes are not meant to be inherited. That distinction matters for the ethical debate but not for the individual patient, who lives with whatever the treatment did or failed to do.

“Until those safeguards are proven reliable, reversibility should be viewed as an important clinical benefit,” Murthy said.

That reframing is the piece’s most useful contribution. The ability to stop a treatment is usually treated as a neutral feature or an inconvenience, something a one-shot therapy improves upon. The authors argue it should be counted as a benefit on the ledger.

Five questions

The commentary proposes a framework of five questions to ask before using a genomic therapy: how severe the disease is; whether reversible alternatives exist; whether the target tissue is suited to a one-time intervention; whether the risk of the delivery method is proportionate; and whether long-term monitoring is feasible.

The questions are deliberately simple, and the point is that they can be asked of any candidate therapy, early. For a severe disease with no alternatives, a one-time treatment in a suitable tissue, the answers line up. For a common condition with good existing drugs, several of them do not.

The framework echoes a logic familiar from other areas of medicine, where the acceptable risk of an intervention scales with the severity of what it treats and the quality of the alternatives. A treatment carrying a small chance of serious harm is straightforward to justify for an otherwise fatal illness and difficult to justify for a condition already managed by a cheap daily tablet.

Trials that are too short

The monitoring question may be the hardest in practice.

“Clinical trials are often too small and too short to identify rare complications,” Smith said.

A trial of a few hundred patients over two years cannot detect an adverse event affecting one patient in ten thousand, or one that emerges after a decade. For most drugs, post-marketing surveillance fills the gap, and a drug can be withdrawn. A therapy that has already altered a patient’s genome cannot be recalled from the people who received it.

The authors argue that long-term surveillance has to be built into how these therapies reach patients rather than added afterwards, and that therapies used by millions will require independent, standardized systems. Existing registries tend to be run by manufacturers, cover single products, and are not designed to follow patients for decades.

A million dollars a patient

There is also the question of money. Some currently available gene therapies list at over $1 million per patient, and several cost substantially more. At rare-disease volumes, health systems have absorbed those prices, with difficulty.

Murthy argues that such pricing would be hard to sustain for common conditions. The arithmetic is unforgiving: a million-dollar treatment for a condition affecting tens of millions of people is not a pricing problem to be negotiated but an impossibility, which means either the cost structure changes or the therapies remain confined to a small number of patients who can access them.

Cholesterol as the test case

Early studies are already testing whether a single gene-editing treatment can produce lasting reductions in LDL cholesterol, the form associated with cardiovascular risk. It is a logical first target: the biology is well understood, the effect is measurable in a blood test, and the liver is a tissue that gene therapies can reach.

Smith’s position is that researchers must first demonstrate long-term safety, durability, affordability and adequate safeguards. Those are the conditions, not objections to the attempt.

“The fact that we can develop a genomic therapy for a condition does not necessarily mean that we should,” Murthy said.

“The goal is not to slow progress,” Smith said.

The commentary is a perspective piece rather than a study, and it reports no new data. Its argument is that the questions worth settling about durable genomic interventions are best settled before the therapies arrive, not after millions of people have received them.

Smith’s disclosures note grants from Lexeo Therapeutics.

---

Notes

1. Smith, E. D., Murthy, V. L. “Clinical Framework for Durable Genomic Interventions.” JAMA Internal Medicine, 2026. DOI: 10.1001/jamainternmed.2026.4690.
2. Michigan Medicine – University of Michigan. “Gene editing is coming for common diseases. Are we ready?” EurekAlert!, 20 October 2026. https://www.eurekalert.org/news-releases/1148457.