The Partial Cure Problem: What Happens When CRISPR Works Only Partly?

When a therapy is not a switch

Much of the public conversation about CRISPR treats gene editing as a switch. The faulty sequence is found, the edit is made, and the patient is cured. That framing is useful for explaining the idea, but it is a poor guide to clinical reality. In trials and in any future NHS rollout, response will not be binary. Some patients will receive a meaningful benefit. Some will respond partly. Some will lose the effect over time. Others will not respond at all, for reasons that are biological, technical or simply not yet understood.

In 2026, Britain has a growing CRISPR clinical research base and a policy conversation that is increasingly focused on delivery, affordability and long-term follow-up. Yet there is still little public discussion of what happens when the edit does not take. That gap matters. A gene-editing programme that cannot describe, measure and support non-response is not ready for routine care.

Three ways an edit can fail to deliver

Non-response is not one phenomenon. It is a cluster of outcomes that need to be separated before they can be managed.

First, there is primary non-response. The edit may be made in the laboratory but not reach enough cells, or enough of the right cells, to produce a clinical effect. In ex vivo therapies, where cells are removed, edited and returned, success depends on editing efficiency, the fitness of the modified cells, and their ability to engraft and persist. A high editing rate in a dish is not the same as a durable clinical benefit in a patient. In in vivo therapies, where editing is delivered directly into the body, the barriers are different: delivery to the target tissue, pre-existing immunity to the delivery vehicle, and the body’s response to the editing components can all limit effect.

Second, there is partial response. Some diseases have thresholds. A patient may need a certain proportion of edited cells, or a certain level of corrected protein, before symptoms improve. Falling just below that threshold may produce laboratory evidence of editing without the clinical change that patients and clinicians care about. Partial response is especially difficult because it can look like success in a press release and like disappointment in a clinic.

Third, there is loss of effect. An edit can be present initially and then fade in clinical importance as edited cells decline, as the underlying disease evolves, or as the immune system responds to edited cells. Durability is not a footnote to gene editing; it is part of the therapeutic claim. A treatment that works for six months and then stops is not the same product as a one-off cure, and it should not be described as one.

Why Britain needs to measure non-response, not just response

UK trial design has become more sophisticated, but non-response can still be obscured by the way studies are reported. Composite endpoints can blend molecular, biochemical and clinical outcomes. A statistically significant change in a laboratory marker may not translate into a benefit that patients notice. Conversely, a patient who feels better may not fit a narrow molecular definition of response.

Britain needs trials that prespecify how non-response will be captured. That means reporting the proportion of participants with no meaningful edit, the proportion with partial response, and the proportion whose response wanes. It means separating assay failure from biological non-response. It means publishing null and ambiguous results, not only the clean wins. The Medicines and Healthcare products Regulatory Agency, the Health Research Authority and research funders all have a role in making that expectation explicit.

The NHS will also need a practical definition of response for commissioning and follow-up. The National Institute for Health and Care Excellence already wrestles with uncertainty in advanced therapies. For CRISPR, the question will not only be whether a therapy works on average, but whether the health system can identify the patients for whom it is not working and offer an appropriate next step.

Consent, care and the cost of silence

Non-response is an ethical issue as much as a scientific one. Consent materials for CRISPR trials should say plainly that the edit may not work, that the effect may be partial, and that the duration of benefit is uncertain. That is not a counsel of despair. It is a condition of honest recruitment. Patients who understand the possibility of non-response are better placed to make decisions and less likely to experience the result as a betrayal.

Care pathways must also reflect that possibility. A patient who has undergone a demanding procedure and received no benefit still needs follow-up. They may need psychological support, access to alternative treatments, and a clear explanation of what is and is not known. In a system as pressed as the NHS, these needs can be quietly deprioritised because they do not fit the success story. That would be a mistake. Non-responders are not a public-relations problem; they are patients.

A registry for the unedited

Britain already has strengths it could use. The NHS’s data infrastructure and the Genomic Medicine Service offer a foundation for long-term follow-up. What is missing is a dedicated approach to capturing non-response across CRISPR programmes. A UK-wide registry, or a coordinated set of trial follow-up arrangements, could record molecular and clinical outcomes over time, including patients who leave trials early or whose response fades.

Such a registry would need governance, consent and privacy safeguards, but it would pay for itself in knowledge. It would help researchers understand whether non-response is tied to disease stage, delivery route, immune status or manufacturing variability. It would give regulators and payers a clearer picture of durability. And it would give patients a way to contribute to the evidence even when their own treatment has not worked.

The trust dividend

Gene editing will not earn public trust by promising that every edit will take. It will earn trust by being candid when it does not. Britain’s CRISPR sector has an opportunity to build that candour into trial reporting, consent, NHS follow-up and public communication from the start. The alternative is a familiar cycle: initial hype, partial results presented as breakthroughs, and then quiet disappointment among patients who were told to expect more.

A mature gene-editing programme does not hide its non-responders. It counts them, learns from them and cares for them. That is the test Britain should set itself now, before the therapies become routine.