Regenerative Medicine Approaches for Children With Cerebral Palsy: A Systematic Review of Clinical Safety and Effectiveness

Major Points and Findings:

This systematic review from Jazan University in Saudi Arabia, with co-authors from other Saudi and Sudanese institutions, gathers 18 clinical studies of cell-based and related regenerative treatments in children with cerebral palsy, covering 1,087 children. It includes randomized trials as well as open-label, retrospective and uncontrolled studies.

The authors did not combine the results statistically. They judged the studies too different in design, cell product, dose, outcome measures and follow-up for a meta-analysis to be meaningful, so the findings are described study by study. Their reading is that umbilical cord-derived products, particularly cord blood with erythropoietin, gave the clearest motor gains, bone marrow-derived cells gave more modest and variable results, and side effects were mostly mild and short-lived. They conclude that the treatment should still be regarded as experimental and used within well-designed clinical trials. Because nothing was pooled, the review gives no overall effect size, and parts of the text contain counting and labelling errors that are described below.

Aim:

To evaluate the effectiveness and safety of regenerative medicine therapies in children with cerebral palsy, with attention to study design, outcome measures and the quality of the evidence.

Methods:

The protocol was registered on PROSPERO (CRD420251083077) and the review follows the PRISMA checklist. PubMed, Web of Science, the Cochrane Library and ScienceDirect were searched in July 2025 for English-language studies published from January 2010 to January 2025.

Eligible studies were randomized controlled trials, non-randomized interventional studies, prospective or retrospective cohort studies and case series of at least 3 children with any type of cerebral palsy. Interventions could be umbilical cord-derived products, mesenchymal stem cells (MSCs) from any source, bone marrow-derived cell therapies, or biological adjuncts given with them, such as erythropoietin (EPO) or granulocyte colony-stimulating factor (G-CSF). Studies had to report motor function (for example the Gross Motor Function Measure, GMFM), cognitive or developmental outcomes, or adverse events. Case reports, animal and laboratory work, reviews, conference abstracts and studies that were not open access were excluded.

Three reviewers screened the records and two extracted the data. Randomized trials were rated with the Cochrane RoB 2 tool and non-randomized studies with ROBINS-I. The results were synthesized narratively by intervention class, study design and outcome domain (motor, cognitive and safety).

Results:

The search found 410 records. After 176 duplicates and 173 records flagged by automation tools were removed, 61 were screened, 39 were assessed in full and 18 were included. Of the 21 excluded at the full-text stage, 12 were removed because they were not open access.

The 18 studies were 10 randomized controlled trials, three open-label trials and five non-randomized studies. They came from China, Korea, the United States, Iran, Poland and Vietnam, and ranged from 15 to 107 children each. Ages at enrolment were reported as 6 months to 17 years, most children had spastic or mixed cerebral palsy, and severity was often not reported in detail.

By number of children, 73.5% (799) received umbilical cord-derived products, 20.3% (221) bone marrow-derived products, 1.8% (20) a mixed umbilical cord or bone marrow product and 4.3% (47) G-CSF. By study, 13 used umbilical cord-derived products (including Wharton’s jelly MSCs), four used autologous bone marrow-derived cells (including neural stem cell-like cells made from bone marrow MSCs) and one used G-CSF. The products studied included:

• Allogeneic or autologous umbilical cord blood given intravenously, alone or with EPO (several Korean and US trials)
• Umbilical cord or cord blood MSCs given intravenously, in four infusions in two Chinese trials, one of which used a fixed dose of 5 × 10⁷ cells per infusion
• Cord blood mononuclear cells (single intrathecal dose of 5 × 10⁶ per kg, Iran) and allogeneic cord blood given intrathecally in 4 to 8 injections of 2 to 3 × 10⁷ cells (retrospective Chinese study)
• Wharton’s jelly MSCs, given by lumbar puncture in 4 to 8 doses (China) or as allogeneic infusions (retrospective Polish study)
• Bone marrow MSCs or mononuclear cells given intrathecally (China, Iran, Vietnam), and in one US crossover trial autologous bone marrow or cord blood mononuclear cells at 6 × 10⁶ cells per kg
• Intravenous G-CSF at 10 μg/kg for 5 days followed by reinfusion of mobilised peripheral blood cells (Korea)

Intravenous delivery was used in 10 studies and intrathecal delivery in six, with a few using combined or other routes. Seven studies gave a single dose and 11 gave repeated doses at intervals from weekly to monthly. Reported doses ranged from about 10⁶ to 10⁸ cells per kg. Most protocols were combined with rehabilitation, whose content and intensity were inconsistently reported.

Motor and other outcomes: The GMFM was the most consistently reported outcome. The authors describe moderate improvements with umbilical cord-derived products, particularly cord blood alone or with EPO, often exceeding the minimal clinically important difference in several trials. Umbilical cord-derived MSCs produced clinically significant but uneven gains, and bone marrow-derived products more modest ones. Several randomized trials reported statistically significant GMFM gains over controls (P<0.05), but effect sizes and confidence intervals were inconsistently reported and the review gives no figures of its own. Several placebo-controlled trials suggested that higher cell doses and combinations such as cord blood with EPO did better than single treatments or placebo. In the G-CSF crossover trial, the original investigators concluded that the improvement may have come from G-CSF alone without a contribution from the reinfused cells.

Safety: Safety data were available for nearly all studies. The common side effects were transient fever, headache, gastrointestinal upset, mild respiratory infections, back pain, and transient discomfort after intrathecal injection. The authors note that intrathecal protocols reported more procedure-related events (headache, back pain, irritability) than intravenous ones, where mainly mild infusion reactions were seen. Serious events were few. They included seizure exacerbation in isolated cases (one child in the Polish study needed a change of medication for worsened epilepsy) and one death, in a Korean cord blood trial, attributed to a cause unrelated to the treatment. No treatment-related cancers, transplant reactions or ectopic tissue formation were reported during the available follow-up, which was mostly short.

Risk of bias: The 10 randomized trials were rated low to moderate risk across the main domains, with allocation concealment and blinding sometimes unclear. The open-label and observational studies carried a higher risk of confounding and performance bias. Per-study ratings are given only in a figure and the supplementary tables.

Conclusions:

The authors conclude that regenerative approaches show promising functional improvement and reassuring short-term safety, particularly for umbilical cord-derived therapies. Variable study quality and limited long-term data prevent recommendations on the best cell type, dose or route. They state that the therapies should be considered experimental and limited to well-designed clinical trials, that the findings do not support routine clinical use, and that they should be seen as complementary to rehabilitation rather than a replacement for it.

The authors list these limitations: exclusion of studies that were not open access, no meta-analysis and therefore no pooled effect sizes, substantial heterogeneity between studies, no formal assessment of publication bias, small samples, frequent non-randomized designs, incomplete blinding even in the better trials, and selective outcome reporting with variable attribution of adverse events.

The following limitations are apparent from the full text:

• Excluding studies that were not open access removed 12 of the 21 full-text exclusions, more than any other reason, so the selection depends partly on publishing model rather than study quality.
• The narrative synthesis gives no numbers for the motor gains, so statements such as “moderate improvements” and “exceeding” the clinically important difference cannot be checked from the review itself.
• The open-label and uncontrolled studies are described alongside the randomized trials, and the authors themselves note that rehabilitation, caregiver involvement and regression to the mean may have inflated the uncontrolled results.
• There are internal inconsistencies. The abstract gives 73% as the share of interventions, while the results give 73.5% as the share of patients. One sentence states that 53% of participants in eight studies received MSCs of “area under the curve (AUC)” origin, which appears to be an error, and most of the cited studies used cord blood rather than MSCs. The text also describes 22 additional records found through trial registers, while the flow diagram shows 22 records excluded at screening. The 2017 North Carolina trial of autologous cord blood (Sun and colleagues) is labelled as allogeneic cord blood in the table, and the stated publication range (2012 to 2025) and age range (from 6 months) do not match the table, where the earliest studies are from 2013 and one study enrolled children from one month of age.
• The Korean open-label extension study is included as a separate study, although its children were drawn from an included randomized trial.
• The included studies were published up to 2025 and the search ended in January 2025, so several recent trials and meta-analyses are not covered.

The review received no funding, and the authors declare no conflicts of interest. One included randomized trial (Gu and colleagues, 2020) used umbilical cord MSCs supplied by Shenzhen Beike Biotechnology.

Background Information:

Cerebral palsy results from non-progressive damage to the developing brain and is the leading cause of physical disability in children. Current treatment focuses on spasticity, skeletal correction and rehabilitation, which cannot reverse the original injury.

The authors suggest that most protocols act through overlapping mechanisms: modulation of inflammation, neuroprotection, and support for axonal growth, synaptic plasticity and the brain’s own repair, rather than true replacement of lost cells. They see this as consistent with the functional gains reported without clear evidence of structural regeneration in most trials. In their view the route of administration mainly affects where the cells go and how much reaches the nervous system, rather than how they work, although they note that direct evidence for this in cerebral palsy is limited.