Cell and gene therapy has produced some of the most remarkable clinical results in the history of medicine. One-time treatments for previously untreatable diseases. Functional cures for conditions that once meant a lifetime of management. The science is extraordinary. So is the attrition rate.

Most CGT programs in development will not make it to approval. That is not pessimism — it is a baseline reality that the field has largely acknowledged but not fully reckoned with. Understanding why programs fail, and at what stage, is not an academic exercise. It is one of the most important strategic inputs available to anyone building or funding a cell and gene therapy program today.

The Data on Disruption

A review published in Nature Medicine — drawing on analysis from McKinsey and Company — compared the clinical development disruption rates for cell and gene therapies against traditional monoclonal antibody programs. The findings are clarifying.

CGT projects experience disruptions in clinical development approximately 43 percent of the time, compared to roughly 29 percent for monoclonal antibody programs. That gap of approximately 14 percentage points is not trivial. It reflects the fundamental difference in biological complexity between a small molecule or protein therapeutic and a living, patient-derived or patient-administered cell or gene construct.

More revealing than the headline disruption rate is the breakdown by cause. For CGT programs, efficacy failures account for 17 percent of disruptions — the single largest category. CMC-related issues account for 12 percent. Safety issues account for 9 percent. The remaining disruptions fall under other clinical and operational causes.

Each of these categories tells a different story about where the field’s vulnerabilities lie.

Table 1. CGT clinical development disruption categories, root causes, and MKA program design implications. Disruption frequency data sourced from McKinsey & Company analysis published in Nature Medicine. The ‘Program Design Implication’ column reflects MKA’s analytical framework for translating disruption data into development strategy.

Efficacy: The Hardest Problem to See Coming

The fact that efficacy is the leading cause of CGT clinical disruptions is both unsurprising and under-discussed. Unlike small molecule drugs, where dose-response relationships are relatively predictable and pharmacokinetics follow conventional ADME models, cell and gene therapies operate through mechanisms that remain poorly characterized even as the products advance into clinical trials.

Tumor heterogeneity is a foundational challenge. Solid tumors are not monolithic targets — they are dynamic, evolving collections of cells with different antigen profiles, different resistance mechanisms, and different relationships to the surrounding tissue. A therapy that ablates one population of tumor cells may leave behind another that proliferates. Identifying which antigen or clone to target, and proving that the targeting mechanism will be durable across a heterogeneous tumor population, requires more than early efficacy data can typically provide.

The tumor microenvironment compounds this. Solid tumors exist within a complex matrix of healthy and diseased tissue that actively suppresses immune activity — including the activity of engineered T cells. For cell therapies in particular, the question is not just whether the therapy reaches the tumor but whether it can function once it gets there.

Animal models, the conventional tool for building preclinical efficacy confidence, offer limited guidance in CGT. Because cell and gene therapies operate through human-specific immunological mechanisms, humanized rodent or non-human primate models cannot fully replicate the cascading downstream effects that occur in actual patients. Programs that look strong in preclinical data routinely encounter different realities at first-in-human.

CMC: Where Science Meets the Floor of Manufacturing Reality

Chemistry, manufacturing, and controls failures are the second largest cause of CGT program disruption, and in some ways the most actionable. Unlike efficacy, which often requires clinical data to reveal itself, CMC problems can frequently be anticipated — and in well-designed programs, largely prevented.

The CMC challenge in CGT begins with the starting material. Unlike conventional biologics manufactured from stable cell lines, many cell therapies start with a single patient’s leukapheresis product — a variable biological input with no guaranteed quality profile. For autologous therapies, the donor is the patient, which means the therapy begins at the moment of collection, under conditions that the manufacturer cannot fully control. Heavily pre-treated oncology patients may have compromised immune cells. Collection timing relative to prior chemotherapy affects cell health. These are not manufacturing problems in the conventional sense; they are upstream biological constraints that propagate downstream into product quality.

For gene therapies, the analogous challenge lies in vector production. The viral vector is not merely a delivery mechanism — it is the drug substance. Its titer, purity, identity, potency, and stability define the product. And unlike biological starting materials for conventional drugs, vector manufacturing involves living production systems — cell lines, transfection processes, purification steps — each of which introduces variability that must be characterized, controlled, and demonstrated to regulators before the product can advance.

CMC failures in CGT often emerge not because developers lack scientific knowledge of these challenges but because the rigor required to translate that knowledge into a controlled, documented, and reproducible manufacturing process is underestimated. The regulatory expectations for CMC in an IND, and even more so in a BLA, are substantial. The FDA’s 2020 CMC guidance for gene therapy INDs makes explicit the level of characterization required for drug substance, drug product, vector, starting materials, and in-process controls at each stage of development. Programs that treat CMC as a downstream problem — something to be solved once the science is validated — consistently discover that the two cannot be decoupled.

Safety: The Known Risks and the Uncertainty Beyond Them

Safety accounts for 9 percent of CGT program disruptions — a smaller share than efficacy or CMC, but one that carries disproportionate consequences when it materializes. CGT safety failures are not merely program-ending events; they are field-defining ones. High-profile adverse events have repeatedly shaped regulatory posture toward entire categories of therapy.

The safety risks specific to CGT are well-characterized in the scientific literature and in FDA guidance, but knowing them does not make them easy to manage. Immunogenicity — the ability of a foreign construct to provoke an immune response — is a fundamental concern for both cell and gene therapies. For gene therapies using viral vectors, pre-existing neutralizing antibodies can prevent the vector from reaching its target entirely. For cell therapies, immune rejection of allogeneic constructs is a core barrier to the promise of off-the-shelf approaches.

Cytokine release syndrome is a known risk for CAR-T therapies and, in severe cases, can be life-threatening. Graft-versus-host disease is an analogous risk for allogeneic cell therapies. Off-target binding — where an engineered receptor interacts with healthy tissue bearing the same antigen as the tumor target — is a mechanism-specific risk that requires rigorous preclinical characterization and ongoing clinical monitoring.

The structural challenge with CGT safety is not a lack of guidance — the FDA has issued detailed guidance on preclinical assessment, early clinical trial design, immunogenicity risk assessment, and post-administration monitoring. The challenge is that the guidance describes what to measure, while the biological systems being measured are still generating novel data with every program that enters the clinic. CGT is a field where the known risks are managed against a background of still-unknown ones.

What This Means Strategically

The disruption data carries a clear implication for how CGT programs should be designed and resourced. Efficacy failures are the leading cause of clinical disruption, which means the preclinical bar for mechanistic confidence needs to be higher than many programs have historically set it. The decision to advance from preclinical to IND is a scientific judgment call, but it is also a resource allocation decision — and programs that enter the clinic with unresolved efficacy hypotheses are statistically the most likely to fail.

CMC failures are the second leading cause, and they are the most preventable. The organizations that have navigated CGT development most successfully are those that treat manufacturing development as a co-equal workstream to clinical strategy, not a supporting function that catches up later. Building analytical capability early — establishing CQAs, developing potency assays, characterizing starting material variability — compresses the timeline to a controlled, regulatory-ready process.

Safety disruptions are less frequent but higher stakes. The programs that have survived adverse events and continued to advance are those with the strongest monitoring infrastructure and the most transparent regulatory relationships. Immunogenicity risk assessment, built into the IND, is not paperwork — it is the foundation of a safety strategy.

The CGT field has produced enough approved products to demonstrate that the challenges are solvable. The question for each new program is whether the development strategy is built with a clear understanding of where programs characteristically break down, and what it takes to build through those failure modes rather than around them.

MKA Insights works with CGT developers to build analytical, regulatory, and commercial strategies that reflect the actual risk landscape of the field. If you are evaluating a program or designing a development plan, we bring the operational experience that the disruption data demands.