The Permanent Prescription: Why Gene Therapy Is Rewriting the Rules of Drug Safety
John Praveen, Vice President, Pharmacovigilance & Medical Writing Delivery and PV Shared Services Offering Lead, Accenture
Gene therapy is transforming medicine by offering the potential for one-time, curative treatments for previously untreatable genetic diseases. However, the very characteristic that makes these therapies revolutionary, their long-lasting or permanent biological effect, also creates unprecedented challenges for pharmacovigilance. Traditional drug safety systems were designed for therapies administered repeatedly over short timeframes, whereas gene therapies may produce safety events years or even decades after treatment.
Gene therapy has moved from laboratory promise to clinical reality, offering something medicine has rarely been able to promise before: a single treatment capable of curing a disease rather than merely managing it. Yet this same quality of permanence is quietly undermining the safety systems the pharmaceutical industry has relied on for decades. Adverse events can emerge a decade or more after treatment, patient populations are often too small for conventional statistical methods, and the standard safety check of withdrawing a drug to see if symptoms resolve does not apply when there is only ever one dose. This article examines what gene therapy actually offers patients, why its safety profile differs so fundamentally from anything pharmacovigilance has previously handled, where established monitoring systems fall short, and what the field is doing to address the gap.
A New Kind of Medicine, A New Kind of Risk
For most of medical history, treating a genetic disease meant managing its symptoms for life. Gene therapy introduced a different proposition: that a single infusion could correct the underlying defect itself. That proposition is no longer theoretical. Children with spinal muscular atrophy, patients with inherited forms of blindness, and individuals living with sickle cell disease have all received gene therapies that have meaningfully altered, and in some cases transformed, the course of their illness.
But any treatment that changes how it works also changes how it can fail. Pharmacovigilance, the discipline responsible for tracking drug safety after approval, was developed for medicines taken repeatedly, where problems tend to surface within months and resolve once the drug is withdrawn. Gene therapy follows none of these conventions. It is, in effect, compelling an entire safety discipline to be rebuilt while it is already treating patients.
How We Got Here
Gene therapy works by introducing new or corrected genetic material into a patient's cells, typically delivered by a modified virus — most often an adeno-associated virus (AAV) or a lentivirus. Once inside the cell, this genetic material may integrate into the patient's own DNA or remain active independently, directing the cell to continue producing a missing or corrected protein indefinitely.
This is precisely what allows a single dose to be potentially curative. It is also what makes the resulting safety profile so unusual. A conventional drug is metabolized and cleared from the body within hours or days. A gene therapy vector may still be active in a patient's cells five, ten, or fifteen years later. The effect regulators wish to preserve, and the risk they must monitor are, in essence, the same biological process, simply unfolding over a far longer timeline.
Regulators have responded with reasonable speed. The FDA approved its first gene therapy product in August 2017, and within a year issued several draft guidance documents covering manufacturing controls, viral testing, and, critically, long-term follow-up of treated patients. These replaced earlier guidance from 2006, underscoring how rapidly the field had progressed from theory to approved product. Today, both the FDA and the European Medicines Agency (EMA) require sponsors to enroll gene therapy recipients in long-term follow-up studies lasting five to fifteen years after treatment, a commitment with no real precedent in conventional drug development.
The Promise: What Gene Therapy Actually Delivers
It is worth pausing to consider why this field commands such attention and investment, because the benefits are substantial.
A shift from management to cure: For conditions such as spinal muscular atrophy or certain inherited forms of blindness, gene therapy has moved outcomes from slowing decline to restoring or preserving function in ways no earlier treatment class could achieve. For some patients, a single infusion has replaced what would otherwise have been a lifetime of progressive disability.
Durability of effect: Because the treatment is designed to persist, many gene therapies eliminate the need for repeat dosing or lifelong daily medication required by older treatments for the same conditions. This has meaningful implications for patients' quality of life, and potentially for long-term healthcare costs, even though the upfront price of these therapies is often very high.
A genuine option for rare disease: Gene therapy is particularly well suited to conditions caused by a single faulty gene, many of which are rare and previously had no meaningful treatment at all. This has opened real therapeutic pathways for patients that the pharmaceutical industry had largely overlooked.
A foundation for future innovation. The manufacturing and delivery techniques developed for current gene therapies are already being adapted for gene editing approaches such as CRISPR, cell therapies, and more advanced combination treatments. Each new product advances the field's understanding and improves the next.
None of this is in dispute. The tension lies in the fact that the same quality driving these benefits — permanence — is precisely what makes the safety side of the equation so much harder to manage than anything pharmacovigilance has addressed before.
The Problem: Where Safety Monitoring Breaks Down
Adverse events that emerge late: Because vector DNA may integrate into the genome or simply remain active for years, risks such as abnormal cell growth, secondary malignancy, or delayed immune reactions may not appear until long after a product has been approved and is in wide use. A safety signal that takes a decade to surface is, by definition, invisible to a system designed around a three-to-five-year monitoring window.
The long-term follow-up burden: Follow-up periods of five to fifteen years are straightforward in principle but difficult to execute in practice. Patients relocate, change insurers, transition out of pediatric care, lose contact with the specialist centers that treated them, or simply grow weary of years of check-ins for a condition they feel is behind them. Every gap in that follow-up represents a gap in the field's ability to detect a genuine late-onset risk.
Populations too small for conventional statistics: Most gene therapies target rare or ultra-rare diseases, meaning the total number of treated patients worldwide can number only in the dozens or low hundreds. Conventional safety monitoring relies on comparing expected and observed rates of adverse events across considerably larger groups. In populations this small, a genuine safety signal can be statistically indistinguishable from background noise, and there is often no sufficiently large comparison group to resolve the ambiguity.
No mechanism for de-challenge: Standard causality assessment relies on a simple test: withdraw the treatment, observe whether the problem resolves, and if it recurs on re-exposure, causality is strongly supported. A one-time gene therapy removes this test entirely. When a symptom appears years later, there is no reliable way to determine whether it stems from the therapy itself, the natural course of the underlying disease, a pre-treatment regimen such as chemotherapy, or an unrelated cause.
Risks tied to the vector and its manufacture: Different delivery vectors carry distinct risk profiles — AAV vectors are associated with immune activation and liver stress, while lentiviral vectors carry a more direct cancer-related risk through genomic integration. Manufacturing itself is complex and can vary from batch to batch in purity and composition, introducing a layer of risk that conventional, chemically stable drugs do not present.
Fragmented data across a specialized network: Gene therapy is typically administered at a limited number of specialized centers, and patients often move between those centers, registries, and broader healthcare systems, sometimes across national borders, with little connectivity between the records each holds.
The Blind Spot: Why Established Tools Fall Short
It is worth being precise about why this mismatch occurs, because the issue is not that safety teams are working insufficiently hard — it is that several of the foundational assumptions behind their tools no longer hold. Standard signal-detection methods require a population large enough to support meaningful statistical comparison; in cohorts of a few dozen to a few hundred patients, that assumption breaks down. Standard causality assessment tools depend heavily on whether a problem resolves after treatment is withdrawn, a criterion unavailable for a permanent, and single-dose therapy. Standard monitoring windows are weighted toward the first few years after launch, precisely the period preceding when many gene therapy risks are expected to appear. And standard reporting depends on a patient or physician connecting a new health problem to a treatment administered years, sometimes a decade, earlier — a link that becomes progressively harder to make as time passes and care teams change. None of these are hypothetical shortcomings. They are structural, which is why conducting existing safety work more diligently will not, on its own, resolve them.
The Fix: Building a New Safety Model
The field is not starting from first principles. A combination of regulatory mandate and industry-led innovation is converging on a more workable approach.
Embed registries into the approval process: Disease- and product-specific patient registries are now frequently required as a condition of approval, and safety plans are expected to be submitted alongside the marketing application rather than added afterward. This builds long-term data collection into a product's lifecycle from the outset.
Connect real-world data sources: Linking hospital records, insurance claims, and registries through shared data formats allows safety signals to be identified across the entire global patient population, even when any single treatment center has observed only a handful of cases.
Use digital tools to sustain engagement over decades: Wearables, patient-facing applications, and telehealth check-ins can ease the burden of fifteen-year follow-up on patients and families, helping to reduce the attrition that otherwise weakens long-term safety data.
Monitor biological markers ahead of clinical symptoms: Rather than waiting for a reportable health event, sponsors can track early indicators, where the vector has integrated, whether it is being shed from the body, and the immune response it provokes — that may flag emerging risk before it becomes clinically apparent.
Tailor safety plans to each specific product: A generic monitoring template is poorly suited to a field where an AAV-related liver risk and a lentiviral cancer risk call for entirely different surveillance. Safety plans built around a therapy's actual biology allow limited monitoring resources to be deployed far more effectively.
Pool safety data across borders: For ultra-rare diseases, no single country's patient population is likely to be large enough to detect a rare signal in isolation. International partnerships among regulators, researchers, and patient advocacy groups are increasingly combining data to achieve the statistical power that no single program could reach alone.
Apply machine learning to sparse, complex data: Given the mismatch between small patient cohorts and the considerable volume of genetic, biological, and clinical data each patient generates, machine learning methods designed for exactly this kind of sparse-data problem represent a promising avenue for identifying patterns that conventional statistics would otherwise miss.
The Bottom Line
Gene therapy's greatest strength a durable, potentially curative effect from a single treatment — is also its greatest safety challenge. A field accustomed to monitoring short-acting drugs across large populations is now responsible for tracking permanent biological changes in some of medicine's smallest patient groups, over timeframes measured in decades rather than years. How effectively the field meets this challenge will shape not only the safety record of individual products, but public and regulatory confidence in gene therapy as it expands to treat more patients and more conditions in the years ahead.
John Praveen is Vice President, Pharmacovigilance & Medical Writing Delivery and PV Shared Services Offering Lead at Accenture, with over 18 years of experience in pharmacovigilance, medical writing, and product safety. He leads global teams delivering innovative safety solutions for leading biopharmaceutical organizations and has successfully built and scaled multiple capabilities across pharmacovigilance, medical writing, medical device vigilance, cosmetovigilance, and safety operations. Passionate about creating new capabilities from the ground up, John focuses on driving transformation through innovation, technology, and collaborative leadership. He is an accomplished author, speaker, and thought leader, with numerous publications and industry contributions focused on advancing pharmacovigilance practices and improving patient safety outcomes worldwide.
