Regenerative medicine approaches in autism spectrum disorder: biological rationale, level of clinical evidence and a framework for practice

Introduction

ASD is a common, aetiologically heterogeneous neurodevelopmental condition. No approved medication targets its core features, and the evidence-based field of intervention remains limited to early, intensive behavioural and developmental education. This therapeutic gap has raised expectations of new biologically based approaches.

Those expectations feed two opposing positions. The first holds that no approach should be entertained until efficacy has been definitively demonstrated. The second assumes that any biologically plausible approach may be applied broadly. From opposite directions, both share the same methodological error: they substitute biological rationale for clinical evidence.

In paediatric neurology the cost of that error is high. The developing brain has a high capacity for neuroplasticity in early life, and developmental opportunities missed in that window cannot always be recovered. But time pressure is a reason to make the decision process more structured, not to lower the evidential threshold.

Method

PubMed and ClinicalTrials.gov were searched for stem cell, cord blood, mesenchymal stromal cell and exosome-based interventions in ASD. Completed and published clinical trials, systematic reviews and registered protocols were assessed together; trials recorded as completed in the registry but not yet published were noted separately. Retracted publications were not used as evidence and are discussed only as a methodological example.

Biological rationale

Peripheral immune dysregulation, chronic low-grade inflammation and microglial activation have been reported in a proportion of individuals with ASD. The immunomodulatory, anti-inflammatory and neurotrophic properties of MSCs align conceptually with this pathophysiological framework.

Experimental models support that alignment. In the valproic acid-induced mouse model, human cord blood-derived MSCs have been shown to improve social behaviour deficits, support cortical dendritic development, increase IGF-1R/Akt phosphorylation and the expression of genes associated with synaptic plasticity, and suppress pro-inflammatory cytokines.

For exosomes and other extracellular vesicles, the rationale rests on the assumption that the same paracrine signals can be delivered without transplanting cells. The assumption is biologically coherent, but it has not been tested clinically in ASD.

The methodological boundary here is clear: a biological effect demonstrated in cell culture and animal models does not amount to clinical benefit in humans. The shared weakness of work in this field is the absence of a defined therapeutic target and the uncertainty surrounding the mechanism of action of the intervention being applied.

Current state of clinical evidence

The most informative study in the field was conducted by Dawson et al. (2020). One hundred and eighty children aged two to seven were randomised to autologous cord blood (n = 56), allogeneic cord blood (n = 63) or placebo (n = 61) and assessed at six months. The infusion was safe and well tolerated. However, no significant improvement was found on the primary endpoint of social communication or on the secondary endpoints. In the subgroup without intellectual disability, changes were observed in the communication domain and in exploratory measures such as eye tracking and EEG alpha–beta power; the clinician-rated overall improvement rate did not reach statistical significance. This subgroup finding is hypothesis-generating, not evidence.

In the crossover randomised trial by Chez et al. (2018) no serious adverse events were reported and a trend was seen in socialisation, but statistical significance was not reached on any endpoint.

The Phase II randomised trial of a cord tissue MSC product in 137 participants was completed in 2023 and did not meet its primary endpoint; to date this result has been reported only in the trial registry.

For exosomes the picture is clearer still: no controlled clinical trial in ASD has been completed and published. Registered studies include a 60-participant Phase I/II trial of intranasal cord-derived MSC exosomes (NCT07243561) and a 100-participant study combining photobiomodulation, platelet-rich plasma and exosomes (NCT06600529). MuseCell-based applications in ASD are represented at the level of observational registries rather than randomised trials.

Exosome and MuseCell applications in ASD therefore sit one step behind cord blood and MSC applications: the first group has randomised trials that have reported, and those results are negative; the second has no randomised trial yet capable of reporting.

Three methodological traps in reading the evidence

First, the mismatch between the concluding sentence and the confidence interval. A meta-analysis of 11 studies and 461 patients concludes that stem cell therapy significantly improves ASD scales. Yet the reported mean difference on the Aberrant Behavior Checklist is −11.97 with a 95% confidence interval of −91.45 to 67.52. An interval of that width carries no information about the direction of effect. The source of this uncertainty is that a substantial proportion of the included studies were uncontrolled and that the outcome measures differed from one another. The existence of a meta-analysis does not by itself raise the level of evidence; the quality of the included studies is what determines it.

Second, retracted publications remaining in circulation. A frequently cited 20-participant study of cord tissue MSCs in ASD was subsequently retracted. Retracted papers can continue to circulate for a long time in the reviews and promotional texts they entered before retraction. Checking the current status of a source is a precondition for citing it.

Third, results that remain only in the registry. Trials with negative results are less likely to be published. That the 137-participant Phase II trial did not meet its primary endpoint is, to date, recorded only in the trial registry. Assessing this field requires searching trial registries alongside the published literature.

Methods and products are not equivalent to one another

Cord blood mononuclear cells, cord blood-derived MSCs, cord tissue-derived MSCs, bone marrow-derived cells, adipose tissue-derived cells, MuseCells and exosomes are distinct products in terms of cellular content, mechanism of action, manufacturing process and immune profile. Grouping them under a single heading is both scientifically incorrect and makes it harder to inform families properly.

Progress in the field is argued to require product standardisation, source-specific randomised trials, patient stratification and optimisation of dose, route and timing together. It has been proposed that the active component may be the purified MSC population rather than the heterogeneous mononuclear cell fraction; this has not been confirmed in any clinical study and remains a hypothesis.

Route of administration is an independent variable in its own right. Intravenous, intrathecal and intranasal administration do not share the same biodistribution or risk profile; safety data obtained with one cannot be transferred to another.

Safety

The most consistent finding in the published studies is short-term safety. No serious adverse events have been reported in the randomised trials, and the adverse effects observed are generally mild and short-lived.

This picture should be read with caution for two reasons. First, sample sizes are not large enough to detect rare events. Second, short-term safety does not mean long-term safety. In the Phase I cord tissue MSC trial, class I anti-HLA antibodies developed in 5 of 12 children. These antibodies remained clinically silent; but the significance of alloimmunisation in a child who may later require an organ or cell transplant cannot be assessed within the duration of a trial alone.

It should also be said that safety cannot be defined by a list of adverse events alone. Misplaced expectations, a shift in the direction of a family's time and attention, and the delay of education programmes of demonstrated effectiveness are measurable losses for the child. In developmental paediatrics, what cannot be recovered is often not the intervention that was given but what was not given during that period.

A framework for assessment in clinical practice

What makes an intervention defensible is not the name it carries but the process by which it is applied. When an approach still under investigation is raised, it is advisable that the following seven questions can be answered in advance:

1. 1Indication. Does the rationale rest on this child's clinical and biological findings, or on a general expectation about the method?
2. 2Endpoint. Has a defined, measurable endpoint specific to this child been set before starting?
3. 3Assessment and stopping rule. Have the assessment time points and the criterion for stopping the intervention been written down in advance?
4. 4Control of confounders. If education intensity, medication or school arrangements also changed in the same period, will the observed change be separable?
5. 5Safety monitoring. Are dose, route, session and adverse event records being kept? Is alloimmunisation monitoring planned for allogeneic administration?
6. 6Informed consent. Has it been explained to the family, in terms they can follow, that the method has not yet entered routine practice, what the current level of evidence is, what the known uncertainties are and what the alternatives are? Are ethics committee and regulatory conditions met?
7. 7Long-term follow-up. Are the duration of follow-up and the person responsible for it defined?

If these questions cannot be answered in advance, the problem lies not in the choice of method but in the way it is being applied. If they can be answered, the method may be considered within a defined and auditable framework even if it has not yet entered routine practice.

The place of education and rehabilitation

Supporting the biological environment and teaching a child a new skill are different processes. Speech, communication, motor function, social interaction and daily living skills are acquired through learning; no biological intervention takes the place of that learning process.

In uncontrolled studies where cell administration was given together with intensive education, the source of the observed change cannot be determined. This is the fundamental design problem of the field: the contribution of a second intervention added on top of an effective one can only be measured under an appropriate control condition.

At the centre of clinical practice should be: comprehensive diagnostic and aetiological assessment; child-specific, measurable developmental goals; intensive, high-quality and individualised education; speech and communication therapy, occupational therapy and, where needed, physiotherapy; management of co-occurring medical problems; family participation and regular developmental follow-up. The possible role of regenerative medicine approaches should be defined as supporting the gains these interventions produce, not replacing them.

Conclusion

The biological rationale for stem cell and extracellular vesicle-based approaches in ASD is coherent and supported in experimental models. In contrast, the randomised controlled trials completed to date have not demonstrated efficacy on the primary endpoint for core ASD features. Short-term safety is consistently reported; long-term safety has not been adequately studied.

This picture requires two conclusions together. Rejecting the field outright is inconsistent with the fact that many treatments now routine were once under investigation. Moving to large-scale application, on the other hand, is not supported by the current evidence.

Progress depends not on new treatment centres but on better-designed studies: patient stratification, a defined mechanism of action, biomarker-based endpoints, product standardisation and source-specific randomised controlled designs.

In paediatric neurology time is limited and cannot be recovered. But the way to protect it is not to lower the threshold of evidence; it is to avoid delaying diagnosis, to start education of demonstrated effectiveness without delay, to follow development closely and to assess new approaches within a defined framework.

References

1. 1.Lv YT, Zhang Y, Liu M, et al. Transplantation of human cord blood mononuclear cells and umbilical cord-derived mesenchymal stem cells in autism. J Transl Med. 2013;11:196. doi:10.1186/1479-5876-11-196
2. 2.Chez M, Lepage C, Parise C, Dang-Chu A, Hankins A, Carroll M. Safety and observations from a placebo-controlled, crossover study to assess use of autologous umbilical cord blood stem cells to improve symptoms in children with autism. Stem Cells Transl Med. 2018;7(4):333–341. doi:10.1002/sctm.17-0042
3. 3.Price J. Cell therapy approaches to autism: a review of clinical trial data. Mol Autism. 2020;11:37. doi:10.1186/s13229-020-00348-z
4. 4.Dawson G, Sun JM, Baker J, et al. A phase II randomized clinical trial of the safety and efficacy of intravenous umbilical cord blood infusion for treatment of children with autism spectrum disorder. J Pediatr. 2020;222:164–173.e5. doi:10.1016/j.jpeds.2020.03.011
5. 5.Sun JM, Dawson G, Franz L, et al. Infusion of human umbilical cord tissue mesenchymal stromal cells in children with autism spectrum disorder. Stem Cells Transl Med. 2020;9(10):1137–1146. doi:10.1002/sctm.19-0434
6. 6.Nguyen Thanh L, Nguyen HP, Ngo MD, et al. Outcomes of bone marrow mononuclear cell transplantation combined with interventional education for autism spectrum disorder. Stem Cells Transl Med. 2021;10(1):14–26. doi:10.1002/sctm.20-0102
7. 7.Villarreal-Martínez L, González-Martínez G, Sáenz-Flores M, et al. Stem cell therapy in the treatment of patients with autism spectrum disorder: a systematic review and meta-analysis. Stem Cell Rev Rep. 2022;18(1):160–174. doi:10.1007/s12015-021-10257-0
8. 8.Hong CE, Lyu SY. Cord-derived stem cells for pediatric autism spectrum disorder: biological rationale and source-specific trial design. Int J Mol Sci. 2026;27(19):8634. doi:10.3390/ijms27198634
9. 9.Tian J, Deng H, Hu Z, et al. Human umbilical cord blood mesenchymal stem cells ameliorate autism-like behaviors in a valproic acid-induced mouse model via the IGF-1/Akt signaling pathway. Brain Behav. 2026;16(4):e71335. doi:10.1002/brb3.71335
10. 10.ClinicalTrials.gov. NCT04089579 · NCT07243561 · NCT06600529.