CAR-T therapy or stem cell transplant: choosing the path for pediatric leukemia

The more clinically relevant question is whether CAR-T therapy should be used as a definitive treatment or as a bridge to transplant.

One National Cancer Institute trial led by Dr. Nirali N. Shah reported a striking difference: patients who received CD19-directed CAR-T therapy followed by allogeneic stem cell transplantation had a median overall survival of 70.2 months, compared with 10.5 months for patients treated with CAR-T therapy alone. That is not a minor statistical fluctuation. It is the kind of efficacy signal that changes the treatment pathway.

But it does not mean every child should automatically proceed to transplant after CAR-T infusion. The result came from a defined protocol and patient population, not from a universal law of pediatric oncology. Disease burden, response depth, measurable residual disease, prior therapies, donor availability, CAR-T construct, toxicity, and the child’s condition all shape the decision.

In my experience running clinical cohorts, the danger is not only undertreatment. It is also mistaking an impressive early remission rate for durable disease control.

CAR-T can remove leukemia quickly. The harder clinical question is whether it can keep it away without a second curative intervention.

The shift from salvage monotherapy to combined treatment

Relapsed or refractory B-ALL remains one of the most difficult problems in pediatric cancer care. Historical five-year survival for this group is approximately 30%, despite advances in chemotherapy, antibody-based treatment, and transplantation. Standard salvage regimens may induce another remission, but remission is not the same as durable leukemia-free survival.

CAR-T cell therapy changed the immediate treatment landscape by offering a mechanism that is fundamentally different from conventional cytotoxic chemotherapy. T cells are collected, genetically modified to recognize the CD19 antigen on B-lineage leukemia cells, expanded, and then returned to the patient. Once activated, these cells can produce a rapid antileukemic response.

The clinical attraction is obvious:

• CAR-T therapy can induce complete remission in heavily pretreated patients.
• It can work when conventional salvage chemotherapy has failed.
• It does not require an immediately available stem cell donor.
• Its activity is directed through an engineered immune-recognition mechanism rather than nonspecific chemotherapy exposure.

The limitations are equally important. CAR-T therapy is not a single standardized product. Different constructs behave differently, and the patient population receiving treatment is not uniform. Some children enter a deep remission and remain free of leukemia. Others relapse after the CAR-T cells lose activity, become undetectable, or fail to control leukemia with reduced or absent CD19 expression.

Allogeneic HSCT addresses the problem through a different biological strategy. After conditioning therapy, the patient receives donor hematopoietic stem cells. The treatment is intensive, but the donor immune system can provide a graft-versus-leukemia effect: immune cells from the donor may recognize and eliminate residual malignant cells.

This creates a practical division between the two approaches:

The most defensible interpretation of current evidence is therefore not that CAR-T replaces HSCT. It is that CAR-T may make HSCT possible for children whose disease could not otherwise be brought under sufficient control.

Why the 70.2-month versus 10.5-month result matters

The survival difference reported in the NCI trial is clinically meaningful because it compares two distinct post-CAR-T pathways. A median overall survival of 70.2 months after CAR-T followed by transplant substantially exceeds the 10.5-month median reported for CAR-T alone.

That finding supports the idea of consolidation. CAR-T therapy can act as a rapid disease-control tool, while HSCT provides a second immunological mechanism intended to suppress residual leukemia over the long term.

However, the number must be interpreted with discipline.

A median is not a guarantee for an individual child. It describes the point at which half of the cohort has experienced the measured outcome and half has not. It does not show that every patient who received HSCT survived for nearly six years, nor does it prove that transplant alone caused the entire difference.

The treatment groups may differ in clinically important ways. Children who proceed to HSCT may have:

• Achieved a sufficiently deep remission to tolerate transplantation.
• Maintained adequate organ function after CAR-T therapy.
• Had access to a suitable donor and a center capable of performing the procedure.
• Avoided severe or prolonged CAR-T-related toxicity.
• Been selected for transplant because their physicians judged relapse risk to be high.

Those factors can create selection effects. A child who deteriorates after CAR-T cannot simply be transferred into the transplant group. This is why a strong cohort result is persuasive but not identical to randomized evidence.

The July 2025 systematic review and meta-analysis adds weight to the consolidation strategy. It included 12 cohort studies covering 380 pediatric and young adult patients aged 25 years or younger. Across those studies, allogeneic HSCT after CAR-T was associated with lower relapse rates and improved leukemia-free survival compared with CAR-T therapy alone.

The wording matters: associated with. This was a meta-analysis of cohort studies, not a randomized trial. The result is still clinically important, particularly because the direction of benefit was consistent across multiple datasets. But statistical significance does not erase confounding, differences between protocols, or variation in follow-up.

A credible treatment decision has to hold both facts at once:

1. The combined pathway shows a stronger long-term signal than CAR-T alone.

2. The available evidence does not define one transplant rule for every child, every CAR-T construct, or every relapse pattern.

That is less satisfying than a universal algorithm. It is also more accurate.

CAR-T construct design changes the treatment logic

The phrase CAR-T therapy can hide a major technical distinction: the intracellular co-stimulatory domain used in the CAR construct.

CD28 and 4-1BB are not interchangeable design details. They influence the activation profile, expansion, persistence, and clinical behavior of the engineered T cells.

CAR-T constructs containing a CD28 co-stimulatory domain generally demonstrate shorter persistence than those using a 4-1BB domain. That shorter persistence can be a disadvantage if the treatment is expected to provide prolonged immune surveillance on its own. But it can also make CD28-based CAR-T therapy particularly useful as a bridge to allogeneic HSCT.

The logic is straightforward. If the clinical plan already includes transplant, the CAR-T product does not necessarily need to persist indefinitely. Its immediate job may be to reduce leukemia burden and create a remission window in which transplantation can be performed.

A 4-1BB-based construct, by contrast, is often discussed in the context of longer persistence. That may support durable control in some patients, but the available evidence does not establish that long-term persistence eliminates the need for consolidative HSCT across all pediatric B-ALL subgroups.

This is precisely where broad claims about CAR-T efficacy in children become unreliable. A result from one construct cannot automatically be transferred to another. Nor can outcomes from a patient treated in a controlled trial be assumed to apply to a child with a different disease burden, prior transplant history, antigen expression pattern, or level of measurable residual disease.

When reviewing a protocol, I look for the details that marketing summaries tend to flatten:

• Which CD19 CAR-T construct was used?
• Was the co-stimulatory domain CD28 or 4-1BB?
• How was remission defined?
• Was measurable residual disease assessed with a sensitive method?
• How many patients proceeded to HSCT, and how were they selected?
• What was the follow-up duration?
• Were relapses caused by antigen loss, inadequate persistence, or resistant leukemia?
• Were severe adverse events reported by grade and timing, or reduced to a general safety statement?

Without these details, the label of CAR-T therapy tells us less than it appears to.

Managing two different toxicity spectra

CAR-T therapy and HSCT are not safer and more dangerous versions of the same intervention. They carry different biological risks, often on different timelines.

The principal acute toxicities of CAR-T therapy are cytokine release syndrome, or CRS, and immune effector cell-associated neurotoxicity syndrome, known as ICANS.

CRS results from immune activation and inflammatory signaling after CAR-T cells engage their target. Clinical severity can vary, but the syndrome may require intensive monitoring and intervention. ICANS can involve neurological symptoms and may occur with or without severe CRS. These adverse events are not theoretical complications to be placed in a footnote. They are central to the treatment pathway and influence whether a child is fit to proceed to subsequent therapy.

HSCT has its own substantial toxicity profile. Conditioning therapy produces profound physiological stress, and donor immune cells can attack recipient tissues, causing acute or chronic graft-versus-host disease. The consequences may be immediate or persist long after hospital discharge. Chronic GvHD can affect multiple organ systems and complicate long-term recovery even when leukemia remains in remission.

A comparison of treatments that lists only remission rates is therefore clinically incomplete. Efficacy endpoints and safety endpoints must be reviewed together.

The correct question is not which treatment has fewer adverse events in the abstract. It is which risk profile is acceptable for a particular child given the probability of relapse without consolidation and the probability of serious transplant complications.

That requires more than a list of side effects. It requires individualized risk modeling, clear reporting of grade 3 and 4 adverse events, and an honest discussion about uncertainty.

A remission endpoint tells you what happened at one time point. A survivorship plan tells you whether the treatment strategy actually worked.

When CAR-T is a bridge rather than the destination

The strongest argument for combining CAR-T with HSCT is that the two treatments solve different parts of the clinical problem.

CAR-T can be highly effective at rapidly reducing active leukemia. That matters when relapse is progressing and standard salvage treatment has limited value. HSCT can then provide a longer-term immune platform, particularly for patients whose relapse risk remains substantial despite achieving remission.

This is why a pediatric leukemia treatment pathway often has to be planned before CAR-T infusion, not improvised afterward. The team may need to address:

1. Disease response after CAR-T. Complete remission is necessary for many transplant strategies, but the depth of that remission matters. Residual disease can influence relapse risk and transplant timing.

2. The child’s recovery from CAR-T toxicities. Severe CRS or ICANS, prolonged cytopenias, infections, and organ dysfunction can delay or prevent HSCT.

3. Donor and transplant logistics. A donor search, conditioning plan, and transplant-center evaluation cannot be treated as administrative details after remission has already been achieved.

4. The CAR-T product and persistence profile. A construct designed for shorter persistence may fit naturally into a bridging strategy, while a product intended to provide longer immune surveillance raises a different question about whether immediate HSCT is necessary.

5. The risk of relapse without consolidation. The decision is not CAR-T versus HSCT in isolation. It is the estimated risk of relapse after CAR-T alone versus the expected benefit and burden of HSCT.

There is no universal consensus on the optimal interval between CAR-T infusion and subsequent allogeneic transplantation. That timing remains dependent on the protocol, the patient’s recovery, disease status, and center-specific practice.

The same uncertainty applies to patients treated with longer-persistence CAR-T constructs. It remains unresolved whether such products can reliably eliminate the need for HSCT in every pediatric B-ALL subgroup. Some children may achieve durable control without transplant. Others may face a high probability of relapse if consolidation is omitted.

That distinction is not a weakness of the field. It is the current state of the evidence. The mistake would be to convert an unresolved question into a treatment promise.

What long-term recovery actually includes

Long-term recovery after pediatric CAR-T therapy is sometimes framed as a question of whether leukemia returns. That endpoint is essential, but it is not sufficient.

Families and clinicians also have to track immune recovery, infectious complications, blood count recovery, late effects of previous therapies, and the cumulative burden of treatment. If HSCT follows CAR-T, the surveillance framework becomes even broader because chronic GvHD and immune reconstitution can shape daily life long after the leukemia is no longer detectable.

For clinical trials, this means follow-up cannot stop at the first complete remission assessment. A study with a high early response rate but limited long-term observation may not answer the question families actually need answered: how many children remain alive and leukemia-free, and at what cost?

The most useful studies separate several efficacy endpoints rather than presenting one headline number:

• Overall survival.
• Leukemia-free survival.
• Duration of CAR-T persistence.
• Cumulative incidence of relapse.
• Non-relapse mortality.
• Incidence and severity of CRS and ICANS.
• Incidence and severity of acute and chronic GvHD after HSCT.
• Quality of life and functional recovery.

This level of detail is especially important when comparing immunotherapy versus transplant for childhood cancer. The treatments operate through different mechanisms, have different failure modes, and produce different burdens for the child and family. A single response percentage cannot summarize that comparison.

The practical verdict for pediatric treatment pathways

For relapsed or refractory pediatric B-ALL, the current evidence supports a combined strategy in many high-risk situations: use CD19 CAR-T therapy to achieve rapid remission, then consider allogeneic HSCT as consolidation when the child’s disease status and clinical condition make the procedure appropriate.

The NCI trial’s 70.2-month versus 10.5-month median survival result and the subsequent meta-analysis of 380 pediatric and young adult patients both argue against treating CAR-T alone as the default endpoint for every patient. The evidence points toward better long-term leukemia control when transplant follows CAR-T in selected children.

But the word selected is doing real clinical work. These findings do not prove that HSCT is required after every CAR-T product or for every patient who reaches remission. They also do not show that CAR-T is merely a temporary treatment with no independent value. Its ability to induce remission in heavily pretreated disease is clinically significant.

The sober conclusion is narrower and more useful: CAR-T therapy and HSCT should usually be evaluated as complementary tools, not competing brands of cure. CAR-T may open the door to remission. HSCT may provide the longer-term immune pressure needed to keep leukemia from returning. Whether that second step is justified depends on relapse risk, construct design, response depth, toxicities, donor feasibility, and the quality of evidence behind the proposed protocol.

For families and clinicians navigating pediatric leukemia treatment, the most important question is not which therapy sounds newer. It is which sequence has the strongest efficacy endpoints, the clearest safety data, and the most credible plan for durable disease control.