Genetically engineered medicines, from gene therapies that rewrite DNA to lab-made proteins and engineered immune cells, carry a range of risks that go well beyond typical drug side effects. Some of these consequences are biological, rooted in how unpredictably living systems respond to genetic alteration. Others are structural, tied to cost, access, and the difficulty of monitoring patients for decades after a one-time treatment. The field has produced genuine breakthroughs for diseases that were once untreatable, but those breakthroughs have come with hard lessons about what can go wrong when you modify the machinery of life.
Unintended Edits to the Genome
The most fundamental risk of any gene-editing therapy is that the tool does not cut where it is supposed to. CRISPR-Cas9, the most widely used editing system, works by guiding a molecular scissor to a specific DNA sequence. But DNA is long and repetitive, and the guide can land at the wrong spot. These “off-target” edits are a major concern, and early research flagged off-target activity rates that could exceed 50% at sites other than the intended one, depending on the guide design and delivery method.1Molecular Therapy – Nucleic Acids. Off-target Effects in CRISPR/Cas9-mediated Genome Engineering Guide RNA design has improved considerably since then, but the problem has not disappeared. Unexpected alterations to the genome remain a central safety concern in every CRISPR application intended for humans.2PubMed Central. Off-target effects in CRISPR/Cas9 gene editing
What makes this worse is that off-target effects are not always small, clean nicks. Recent work has revealed that CRISPR can cause large structural changes to chromosomes, including translocations (where chunks of DNA swap between chromosomes) and deletions spanning millions of base pairs. These large-scale rearrangements are harder to detect with standard sequencing methods and raise serious questions about long-term safety, particularly in therapies that edit cells inside a living patient rather than in a dish.3Nature Communications. The hidden risks of CRISPR/Cas: structural variations and genome integrity A small insertion or deletion at the wrong site might be harmless, but a megabase-scale deletion could wipe out tumor-suppressor genes or disrupt essential developmental programs. The ability to detect these events reliably in clinical settings is still catching up to the ability to cause them.
Immune Reactions to Gene Therapy Vectors
Most gene therapies need a vehicle to carry their genetic payload into cells, and the most common vehicle is a modified virus called adeno-associated virus, or AAV. AAV vectors are popular because they are relatively safe compared to other viral delivery systems. But “relatively safe” is not the same as harmless. At high doses, AAV vectors can trigger severe immune reactions. Patients have died from liver, kidney, heart, or lung failure driven by the body’s innate and adaptive immune responses to the vector itself.4PubMed Central. Lethal immunotoxicity in high-dose systemic AAV therapy
One particularly underappreciated mechanism is complement activation, a part of the immune system that was initially thought to barely react to AAV. Clinical trials have since shown otherwise: complement-driven reactions can cause serious adverse events, particularly in patients receiving large vector doses.5Frontiers in Immunology. Immunogenicity and toxicity of AAV gene therapy Other dose-dependent complications include toxicity to the dorsal root ganglia (nerve clusters along the spine that relay sensory information) and severe liver damage. Immunosuppressive drugs given alongside therapy can help blunt T-cell responses, but they do not work reliably in all patients, especially those getting the highest doses.
The dose problem is a genuine bind. Many genetic diseases require the therapeutic gene to reach a large proportion of cells across an entire organ. That demands a high dose. But high doses are exactly what triggers the worst immune reactions. Researchers are exploring more efficient vector designs that could deliver the same benefit at lower doses, but for now, the trade-off between efficacy and immune toxicity remains one of the field’s central tensions.
Cancer Caused by the Treatment Itself
Before CRISPR existed, the main gene therapy approach was to use retroviral vectors that insert a corrective gene into a patient’s DNA at a semi-random location. In some of the earliest and most celebrated gene therapy trials, for a severe immune deficiency called SCID-X1, this approach worked. Children who had been living in sterile isolation gained functioning immune systems. But several of those children later developed leukemia. The retroviral vector had inserted itself near a gene called LMO2, a known proto-oncogene, and its enhancer elements switched LMO2 on.6The Journal of Clinical Investigation. Insertional mutagenesis combined with acquired somatic mutations causes leukemogenesis following gene therapy of SCID-X1 patients The leukemia was not caused by the vector insertion alone; other mutations accumulated alongside it. But the insertion was the initiating event, and this pattern of insertional activation of cancer-promoting genes became a recurring finding across early gene therapy trials using similar vectors.7PubMed Central. Stem cell gene therapy: the risks of insertional mutagenesis and approaches to minimize genotoxicity
Newer vector designs, particularly lentiviral vectors with self-inactivating features, have substantially reduced this risk. But they have not eliminated it. Any therapy that permanently integrates foreign DNA into the genome carries some baseline risk that the integration site will disrupt something it shouldn’t. This is one reason regulators now require long-term follow-up studies, sometimes spanning 15 years, for patients who receive integrating gene therapies.
The Dangers of CAR T-Cell Therapy
Chimeric antigen receptor T-cell therapy, or CAR T, is one of the most dramatic examples of genetically engineered medicine. A patient’s own immune cells are removed, genetically modified to recognize cancer cells, and infused back. When it works, it can produce complete remissions in blood cancers that had resisted every other treatment. When it goes wrong, it can be life-threatening.
The most common serious side effect is cytokine release syndrome, an inflammatory storm triggered when the engineered T cells activate en masse and flood the bloodstream with immune signaling molecules. In up to about a third of patients, this produces toxicities severe enough to require intensive care.8PubMed Central. Cytokine release syndrome and associated neurotoxicity in cancer immunotherapy Symptoms range from high fever and low blood pressure to multi-organ dysfunction. The inflammatory molecules interleukin-1 and interleukin-6 play central roles, and blocking them with targeted drugs has become a standard part of managing CAR T patients.9Nature Medicine. Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syndrome and neurotoxicity due to CAR T cells
The second major risk is neurotoxicity. Known formally as ICANS (immune effector cell-associated neurotoxicity syndrome), this can cause confusion, difficulty speaking, seizures, and in rare cases fatal brain swelling. Fatal brain edema in early trials led researchers to routinely exclude patients with existing brain involvement from clinical studies, which means that the people who might benefit most from the therapy are often the ones kept out of trials for safety reasons.10PubMed Central. CAR T-cell-associated neurotoxicity in central nervous system hematologic disease: Is it still a concern?
Even when patients survive these acute toxicities, their cancers can come back. A substantial number of CAR T patients relapse because the cancer cells lose or modify the surface marker the engineered T cells were designed to target, essentially becoming invisible to the therapy. In other cases, the modified T cells simply do not persist long enough in the body to maintain a durable response.11PubMed Central. Mechanisms of resistance to CAR T cell therapy
The Re-Dosing Problem
A less dramatic but practically devastating consequence of genetically engineered medicine is that many treatments cannot be given twice. This is especially true for AAV-based gene therapies. After a patient receives an AAV vector, their immune system generates neutralizing antibodies against the viral shell. Those antibodies persist, and if the same vector type is administered again, the immune system destroys it before it can deliver its payload. Patients who are already seropositive for AAV antibodies (from natural exposure to wild-type AAV in childhood) are typically excluded from clinical trials altogether.12Trends in Biotechnology. Evading and overcoming AAV neutralization in gene therapy
This matters enormously for diseases where the therapeutic effect fades over time. If a gene therapy for a progressive liver disease wears off after five or ten years, the patient may have no option for retreatment. Researchers are working on strategies to evade neutralizing antibodies, including engineered capsids that the immune system does not recognize and antibody-depleting protocols, but none of these has yet become routine in the clinic.13PubMed Central. Evading the immune response upon in vivo gene therapy with viral vectors For now, gene therapy is often a one-shot deal, which puts extraordinary pressure on getting the dose and the vector design right the first time.
Antibodies That Undermine Engineered Proteins
Gene therapy is only one branch of genetically engineered medicine. Recombinant proteins, including insulin, clotting factors, and enzyme replacement therapies, are produced using genetically modified cells and have been in clinical use for decades. They are generally safer than gene therapies, but they carry their own immunological risk: the body can produce anti-drug antibodies that neutralize the therapeutic protein, rendering it ineffective.
This is a well-documented problem in enzyme replacement therapy for rare metabolic diseases. In Fabry disease, for instance, treatment with recombinant enzyme can provoke infusion reactions and the formation of neutralizing antibodies, which then blunt the therapy’s effect and allow the disease to progress despite ongoing treatment.14PubMed Central. Mechanisms of Neutralizing Anti-drug Antibody Formation and Clinical Relevance on Therapeutic Efficacy of Enzyme Replacement Therapies in Fabry Disease The same pattern occurs across a range of recombinant biologics. In some cases, the antibodies do more than neutralize the drug; they cross-react with the patient’s own naturally produced version of the protein, creating a new deficiency that did not exist before treatment began.15Trends in Pharmacological Sciences. Mechanisms of unwanted immunogenicity to recombinant human protein therapeutics
Part of the immunogenicity risk comes from the manufacturing process itself. Recombinant proteins are made in living cells, often Chinese hamster ovary cells or bacteria, and trace amounts of the host cell’s own proteins inevitably end up in the final product. These impurities are present at very low levels, but they can act as immune-stimulating contaminants, priming the body to mount a broader immune response that includes attacks on the therapeutic protein.16PubMed. Evaluating Immunogenicity Risk Due to Host Cell Protein Impurities in Antibody-Based Biotherapeutics Purification technology has improved steadily, but the diversity of host cell proteins makes complete removal an ongoing challenge.17PubMed Central. Host cell protein impurities in therapeutic proteins: overview of advances in detection, nonconventional removal technologies and immunogenicity assessment
Pleiotropy and the Ripple Effects of Editing Genes
Even when a gene edit lands exactly where it is supposed to, the consequences may not be what anyone intended. Most genes do more than one thing. A variant that protects against one disease can increase the risk of another. This property, called pleiotropy, means that correcting a genetic defect can have unpredictable knock-on effects throughout the body.
The concern is especially acute for any future application that attempts to edit multiple genetic variants at once, as has been proposed for polygenic diseases like heart disease or diabetes. Large-effect genetic changes are, on average, predicted to be harmful, because evolution has already filtered out most variants that have big effects without trade-offs. Variants that lower the risk of one disease often increase the risk of another, and little is known about how these trade-offs play out during fetal development or across a full lifespan.18Nature. Heritable polygenic editing: the next frontier in genomic medicine? For single-gene disorders with clear, well-understood mutations, these pleiotropic concerns are more manageable. But as the ambitions of genetic medicine expand toward common diseases, pleiotropy becomes a much bigger wild card.
When Somatic Edits Could Reach Future Generations
Current gene therapies are designed to modify only somatic cells, the non-reproductive cells of the body. The assumption has been that these edits die with the patient and are never passed to children. That assumption is now being questioned. New technology platforms, including induced pluripotent stem cells (lab-reprogrammed adult cells that behave like embryonic stem cells), open a pathway by which edits made in ordinary body cells could, in theory, be used to create eggs or sperm through a process called in vitro gametogenesis. If that chain of events were ever carried out, edits originally intended to be non-heritable could become heritable.19PubMed Central. Do not overlook the possibility of genome-edited somatic cells ending up in the human germline
There is also speculation that gene editing performed on fetal cells during pregnancy could unintentionally reach the germline, because the human fetus undergoes extensive tissue remodeling and cells retain a high degree of plasticity. Neither of these scenarios is routine or imminent, but they illustrate how the boundaries between somatic and germline modification are less absolute than regulators and ethicists originally assumed. The governance frameworks built around the idea that somatic editing is inherently non-heritable may need updating.
The Cost and Access Problem
Many genetically engineered medicines carry price tags that are orders of magnitude higher than conventional drugs. Gene therapies approved in the United States have been priced at $1 million to $3.5 million per patient, reflecting both the complexity of manufacturing and the small patient populations involved. One simulation of the U.S. market estimated that annual spending on gene therapies could reach roughly $20 billion under conservative assumptions as more products enter the market.20PubMed Central. The estimated annual financial impact of gene therapy in the United States
High cost is consistently identified as the single biggest obstacle to patient access. A comprehensive review of the literature found that affordability concerns, especially around the upfront cost of therapy and how to structure reimbursement, appeared in the vast majority of published discussions on gene therapy access barriers.21BMJ Innovations. Patient access to gene therapy medicinal products: a comprehensive review Traditional insurance models are built for recurring costs spread over time, not for one-time cures that cost as much as a house. Outcomes-based payment models, where the manufacturer is paid only if the therapy works, have been proposed and piloted, but they add complexity and have not scaled.
The access problem is not purely financial. Gene therapies often need to be administered at specialized centers with cell-processing laboratories, intensive care units, and multidisciplinary teams. Patients in rural areas or in countries without these facilities may have no realistic path to treatment. For diseases concentrated in low-income regions (like sickle cell disease, which disproportionately affects sub-Saharan Africa), the mismatch between where the patients are and where the treatments are available is stark.
Environmental Release and Vector Shedding
When a patient receives a viral-vector gene therapy, the modified virus does not always stay inside the patient. Treated individuals can shed viral particles through saliva, urine, stool, or other bodily fluids. This “vector shedding” raises questions about unintended environmental release of genetically modified organisms. In most cases, the shed virus is replication-deficient and cannot infect other people in a meaningful way. But the regulatory frameworks for assessing this risk vary widely between countries. In Canada, for example, shedding intersects with environmental protection law, and the potential for release through medical waste disposal or from the patient directly has prompted calls for clearer regulatory guidelines.22Frontiers in Medicine. Recommendations for Regulating the Environmental Risk of Shedding for Gene Therapy and Oncolytic Viruses in Canada
The practical risk to bystanders from current gene therapies is almost certainly very low. But as gene therapies move toward more potent vectors and broader patient populations, the cumulative environmental load of shed material becomes a question that regulators will have to take more seriously.
Long-Term Monitoring in Uncharted Territory
One of the most challenging consequences of genetically engineered medicine is that many of its risks may not appear for years or decades. A gene therapy administered to a child today could cause chromosomal instability or oncogene activation that manifests as cancer in middle age. The pharmacovigilance systems designed for conventional drugs, which focus on adverse events within weeks or months, are not built for this kind of timeline. Regulators have mandated long-term follow-up periods for cell and gene therapies, but actually tracking patients for 15 years is logistically difficult, and the infrastructure for doing so varies enormously between countries.23PubMed Central. Pharmacovigilance in Cell and Gene Therapy: Evolving Challenges in Risk Management and Long-Term Follow-Up
Pediatric patients face a particular version of this challenge. Children have decades of life ahead of them, amplifying the window in which late-onset complications could appear. The preclinical models used to test gene therapies for childhood cancers often fail to replicate pediatric tumor biology accurately, which means the safety data from animal studies may be less reliable than it is for adult applications.24PubMed Central. Advances and challenges in gene therapy strategies for pediatric cancer: a comprehensive update Consent is another wrinkle: a five-year-old cannot meaningfully agree to a treatment whose full risks may not be known until they are forty.
Biosecurity and Dual-Use Risks
The tools used to create genetically engineered medicines are the same tools that could, in the wrong hands, be used to create biological weapons or harmful organisms. Synthetic biology enables the design of novel biological systems from scratch, and the same gene-synthesis platforms that produce therapeutic vectors could theoretically produce dangerous pathogens. The risks include unauthorized access, misuse, and intentional release of engineered biological material.25Synthetic and Systems Biotechnology. Regulation and management of the biosecurity for synthetic biology
Governance has struggled to keep pace. The challenges include gene modifications that produce organisms with unpredictable behavior, threats to species diversity, the potential for laboratory accidents, and the difficulty of controlling knowledge that is increasingly accessible to small groups outside traditional research institutions.26Journal of Biosafety and Biosecurity. Challenges and recent progress in the governance of biosecurity risks in the era of synthetic biology DNA synthesis companies now screen orders against databases of known pathogen sequences, but the screening systems are voluntary in many jurisdictions and imperfect even where they are required. As gene-editing tools become cheaper and easier to use, the barrier to misuse continues to drop.
Engineered Bacteria and the Gut Microbiome
An emerging branch of genetically engineered medicine involves modifying bacteria that live inside the human body. Researchers have demonstrated systems that reprogram a host’s own gut bacteria using horizontal gene transfer, essentially delivering new genetic functions directly into the microbiome.27Columbia University Department of Systems Biology. Personalized Gene Delivery to the Gut The potential applications are exciting: engineered gut bacteria could produce therapeutic molecules on demand, detect disease biomarkers, or correct metabolic deficiencies locally.
But modifying a microbial ecosystem that contains trillions of organisms and hundreds of species introduces risks that are qualitatively different from editing a patient’s own cells. Engineered genes could spread to unintended bacterial species through the same horizontal transfer mechanism that makes the technology work. The long-term ecological effects of introducing synthetic genetic functions into a complex community are largely unknown. And unlike a gene therapy that modifies a fixed number of the patient’s cells, engineered gut bacteria reproduce, evolve, and potentially shed into the environment through normal bodily functions, creating containment challenges that do not exist for other forms of genetic medicine.