Cord Blood Treatment for Children With Cerebral Palsy: Individual Participant Data Meta-Analysis Open Access

Cerebral palsy (CP) describes a group of lifelong, neurological conditions caused by nonprogressive injury or maldevelopment in the developing brain that affects movement and posture.¹ Usual care includes rehabilitation, orthopedics, and technology, with generally modest effect sizes from large investments in time and effort.² Stem cells and cellular therapies offer potential as a novel approach for the treatment of brain injury underlying neurological conditions like CP, with community backing.³ Originally focused on cell replacement as the key therapeutic mechanism, the field has now shifted to exploring cell therapies that do not require engraftment but rather function via anti-inflammatory and immune-modulating actions and help improve the local environment via paracrine signaling to promote endogenous repair after injury.⁴ Several types of cells are currently under investigation as treatments for CP, including umbilical cord blood (UCB) cells, bone marrow cells, mesenchymal stem/stromal cells, and neural stem cells.⁵ Of these, UCB is the most thoroughly studied cell treatment for CP, having been tested in clinical trials for almost 20 years. UCB contains a variety of cell types, including hemopoietic stem and progenitor cells, mesenchymal stem/stromal cells, endothelial progenitor cells, monocyte-derived cells, and T-regulatory cells.⁴ In small- and large-animal models of perinatal brain injury/CP, UCB treatment promotes brain repair, including reducing brain infarct size, astrogliosis, microglial activation, and neuroinflammation and increasing the number of neurons and oligodendrocytes and improving motor function.⁶ These mechanisms are consistent with clinical findings of brain imaging and outcomes after UCB treatment for CP,^7,8 with the working hypothesis that UCB works primarily via paracrine signaling when administered systemically for the treatment of the brain injury underlying a CP diagnosis. This is supported by a growing body of research suggesting that persistent brain and systemic inflammation contributes as a comorbidity of CP⁹ and is a target for UCB therapy.

A 2021 review indicated that there have been more than 300 participants treated in single-arm clinical trials of UCB cells (both autologous and allogeneic), with another 300 treated in controlled studies.⁵ Collectively, this research indicates that UCB treatment is safe and more effective than rehabilitation alone in improving gross motor function, measured using the gold standard, Gross Motor Function Measure (GMFM).¹⁰ Despite the accumulating research data, considerable heterogeneity has been noted across research design, participants (eg, age; type, topography, and severity of CP), and treatment protocols (eg, dose, UCB source, and use of premedication/immunosuppressants). This heterogeneity has limited the ability to interpret overall treatment effect of UCB on gross motor function, determine dosing thresholds for efficacy, and identify best responders.

The existing research data hold valuable information that could be used to better understand the clinical effect of UCB treatment for CP from brain injury, enable optimization of UCB treatment (eg, by ensuring sufficient dose), and design future trials with precision. To achieve this, we sought to conduct a systematic review and individual participant data meta-analysis (IPDMA) with the following aims: (1) determine the effect of UCB treatment on gross motor function measured using the GMFM-66 at 1, 3, 6, and 12 months after infusion in individuals with CP; (2) determine the effect of UCB cell dose on clinical improvement; (3) explore covariates, including baseline age, severity, and etiology of CP, to elucidate responder subgroups; and (4) report the safety of UCB treatment compared with controls.

This study is a systematic review and meta-analysis of individual participant data (IPD), which we have reported according to the Preferred Reporting Items for a Systematic Review and Meta-Analysis of Individual Participant Data (PRISMA-IPD) (Appendix S1).¹¹ This study was prospectively registered on PROSPERO (CRD42021259527). In the primary data sources, all participants had provided informed consent following relevant ethical approvals. Before release of nonidentifiable IPD, the IPDMA protocol and analysis plan were approved by the data contributor’s ethics committees as requested/required.

Searches of the CENTRAL, MEDLINE, and EMBASE databases via Ovid, ClinicalTrials.gov, and hand-searching were conducted on May 9, 2024. The full search strategy can be found in Appendix S2.

Eligible studies included clinical research studies (1) with or without a control group or comparator; (2) administering autologous or allogeneic UCB via any route (with or without concomitant treatments, eg, erythropoietin [EPO]); (3) including individuals with CP of any age; (4) collecting participant GMFM-66 total scores (or had item-level scoring data from the GMFM-88 to calculate GMFM-66) at baseline and at 1-, 3-, 6-, and/or 12-month follow-up time points; and (5) published in English.

Deduplicated results were imported into Covidence Systematic Review Software. Titles and/or abstracts of retrieved studies were screened before full texts were independently assessed for eligibility (by MFE and MCBP). There were no screening disagreements requiring a third author. Corresponding authors for all eligible studies were contacted via email regarding contribution of IPD. Nonresponders, after multiple attempts, were excluded from the IPDMA.

Deidentified IPD were uploaded by each data contributor into a secured SharePoint (Microsoft) folder. The list of variables requested and how these were coded for analysis is presented (Appendix S3). Once data were received, they were cleaned and recoded for uniformity across datasets and entered into a custom-built Research Electronic Data Capture form (REDCap; hosted at the University of Sydney) (by MFE, MCBP, ARG, or RKBW), with verification by a second author. The complete dataset was exported into R (version 4.3.1; R Foundation for Statistical Computing) for analysis. Suspected discrepancies or inaccuracies in source data were resolved via consultation with data contributor(s) as necessary (eg, rescoring GMFM-88 into GMFM-66).

The main analysis was conducted including all participants who had received UCB or placebo (excluding participants who received UCB and EPO and those who received EPO alone), including instances in which participants were enrolled in more than 1 study, but excluding any participant who received more than 1 dose of UCB in the included studies. A 1-stage IPDMA was conducted to estimate the effect of UCB treatment on GMFM-66 by fitting linear mixed-effects models, which included random-effects terms for intervention group (to account for between-study heterogeneity), adjusted for baseline GMFM-66, and included a random intercept term to account for repeat participants. Between-study heterogeneity was quantified by the estimated variance of the random-effects term for intervention group (τ²). Sources of between-study heterogeneity were investigated by including interaction terms between the intervention group and potential sources of heterogeneity in the mixed-effects model and examining the effect on the value of τ². Differences in the treatment effect associated with individual characteristics were investigated by adding an interaction term between intervention group and the individual characteristic variables of interest to the mixed-effects model. A secondary analysis investigated the overall effect of UCB treatment on GMFM-66 when including participants who received UCB and concomitant EPO treatment. Aggregate outcome data for studies were not extracted or analyzed because a 1-stage IPDMA was conducted.

Analysis of the effect of UCB cell dose on GMFM-66 included all participants who had received UCB or placebo, excluding any participant who had received EPO, those who did not receive UCB intravenously, and those who had received more than 1 dose of UCB due to enrollment in multiple studies. We imputed missing pre-thaw cell dose count values from post-thaw data using a multilevel multiple-imputation model,¹² and these imputed datasets were used for all subsequent cell dose analyses. Mixed-effects models, adjusted for baseline GMFM-66 with random-effects terms for UCB cell dose (to account for between-study heterogeneity), were used to test for any significant relationship between change in GMFM-66 and UCB cell dose. Nonlinear UCB cell dose trends were tested. Differences in the effect of UCB cell dose on GMFM-66 associated with individual participant characteristics were investigated by including interaction terms between UCB cell dose and individual characteristic variables in these mixed-effects models.

All GMFM-66 effect sizes for overall effect, effects associated with individual characteristics, and cell dose effects are reported relative to controls.

Adverse event (AE) and serious adverse event (SAE) data were extracted from all included studies. Incomplete or unclear data were verified through consultation with the study team(s) (eg, SAE relatedness). The total number of AEs, SAEs, and relatedness to treatment allocation was reported. AE/SAE designation and relatedness were determined by individual study teams, and no information was recoded.

Two study authors (split between MFE, ARG, and RKBW) independently assessed randomized controlled trials for risk of bias using the Cochrane Risk of Bias Tool 2.¹³ Studies with no comparator group were independently rated using the “Quality Assessment Tool for Before-After (Pre-Post) Studies With No Control Group” (MFE, MCBP, and RKBW).¹⁴ It was not possible to assess the risk of bias for unpublished studies. Risk-of-bias assessments were used to conduct sensitivity analyses investigating study quality as a potential source of between-study heterogeneity or a moderator of treatment effect size. In these sensitivity analyses, studies were grouped as low concern (randomized controlled trials with low risk of bias or single-arm studies of good quality), moderate concern (some concerns about risk of bias or fair study quality), and high concern (high risk of bias or poor study quality).

To explore potential publication bias, funnel plots were generated using R to visualize the between-group differences (UCB alone vs placebo and UCB or UCB and EPO vs placebo) for the randomized controlled trials, and the within-group pre-post change for single-arm studies, for each follow-up time point. As the number of individual studies included in each funnel plot was small (<10), both visual inspection and Egger’s tests for funnel plot asymmetry¹⁵ were conducted to determine whether publication bias was present.¹⁶ 

Screening of 98 published studies identified via databases and hand-searching resulted in 12 included published studies.^8,17–27 In addition, 3 unpublished studies were identified from ClinicalTrials.gov,^28–30 of which 1 was deemed eligible and included.²⁸ The PRISMA-IPD flow diagram and a summary of the 13 included studies are presented (Appendix S4 and Table S1). IPD were successfully obtained from 11 studies: 7 controlled trials and 4 single-arm, contributing a total of 498 participant data records for 447 unique participants. Of these, n = 170 (38%) received only UCB, n = 53 (12%) received UCB and EPO, n = 53 (12%) received only EPO, and n = 171 (38%) were controls; however, the n = 53 participants who received only EPO were not included in any analyses. No response was received for 2 reports,^20,21 resulting in exclusion. IPD summaries for unique participants as well as all included observations (with or without EPO-treated participants) are shown in Tables S2 and S3. Participant characteristics for the main analysis are shown in the Table.

Of participants included in the main analysis (Table S3), 60% were boys, and most had spastic CP (90%) and bilateral presentation (85%). All 5 Gross Motor Function Classification System (GMFCS) levels were represented, with 52% classified as moderate to severe (GMFCS level IV-V). Mean baseline age was 54.6 months (SD, 39.0), ranging from 8 months to 18.9 years. Just over half of participants (52%) were term-born or moderate-to-late preterm. Various presumed underlying etiologies/causal pathways to CP were reported (Table), including stroke/intraventricular hemorrhage (21%), hypoxic ischemic encephalopathy (13%), and “other” (45%) (Table S3). Notably, the “other” group comprised multiple well-defined causal pathways and risk factors for CP, such as presumed periventricular leukomalacia, surgery, meningitis, traumatic brain injury, and kernicterus, that appeared less frequently in the dataset.

From all included observations, 84% of infused UCB was sourced allogeneically, with 86% of these from an unrelated donor. Of participants receiving allogeneic UCB, 26% had no mismatches across the assessed human leukocyte antigen (HLA) markers. One study (n = 72, 18.4% of all included observations) administered cells intrathecally,²⁷ with these participants receiving 5 × 10⁶ mononuclear cells per kilogram of body weight. All other studies used the intravenous route, although 2 participants received cells intra-arterially.¹⁹ The mean pre-thaw cell dose for intravenous or intra-arterial UCB was 56.1 (SD, 39.2; range, 9.7–10.3) × 10⁶ total nucleated cells (TNCs)/kg. When using multiple imputation to account for missing pre-thaw data, the mean cell dose was 56.8 (SD, 40.0; range, 9.7–210.3) × 10⁶ TNCs/kg. Time points for GMFM-66 outcome data collection varied. The most-used follow-up time point was 6 months, with 72% of participants assessed.

Among participants included in the main analysis (all included observations), there was no difference between the treated and the control participants in terms of clinical or demographics characteristics including sex; age at baseline; type, topography, or severity of CP; gestational age (GA); or etiological factors (Table S4). There was no significant difference (P = .30) between the mean GMFM-66 score at baseline for UCB-treated participants (mean, 42.2; SD, 15.8; range, 8.1–85.0) and controls (mean, 40.3; SD, 17.1; range, 4.1–96.0). Potential interactions between variables were investigated and are reported in Tables S5 and S6. There was a significant association between CP severity and GA (P < .001), with participants with more severe CP more likely to have been born very or extremely preterm, and a significant association between CP severity and age at baseline (P = .03); participants with more severe CP were older at baseline. An association between cell dose (irrespective of UCB source) and causal pathway was also found (P = .003), with higher cell doses for those with hypoxic ischemic encephalopathy.

The 2 studies not included in the IPDMA were single-arm, comprising a total of 26 participants.^20,21 Both treated children with CP (range, 1.7–7.6 years) with autologous UCB administered intravenously. Follow-up time points varied, and variable effects on the GMFM were reported (Table S1).

Overall, 4 of the 7 (57%) randomized controlled trials assessed were rated as low risk, 1 (14%) as some concerns, and 2 (29%) as high risk of bias (Table S7). Quality assessment of the 5 single-arm studies resulted in 3 studies being rated as “good” and 2 as “fair,” often due to lack of detail in the report (Table S8).

The moderating effect of study quality on treatment effect size was investigated, demonstrating no significant difference in the effect of UCB treatment at the 1-, 3-, or 6-month follow-up time points across study quality levels (Table S9). A marginally significant difference in the effect size across study quality was detected at 12-month follow-up (P = .065); however, controlling for administration route, age at baseline, or cell dose reduced the association, suggesting that the observed marginal difference may instead be attributed to other study-specific factors (Table S9).

Visual inspection of funnel plots revealed no indication of publication bias at any time point for any of the comparisons analyzed, and Egger’s tests found no significant funnel plot asymmetry. Representative funnel plots for the 6- and 12-month time points are shown for the randomized controlled trials (between-group differences) and single-arm studies (pre-post analysis) (Figure S3), and P values for Egger’s test for funnel plot asymmetry are included (Table S10).

For the overall treatment effect of UCB, mean GMFM-66 scores were significantly higher in UCB-treated participants compared with controls at 6 and 12 months after intervention (6 months: 1.36 points [95% CI, 0.41–2.32; P = .005]; 12 months: 1.42 points [95% CI, 0.31–2.52; P = .012]; Figure 1). At these follow-up time points, there was negligible between-study heterogeneity (τ² = 0) in the effect observed for UCB. In contrast, there was no significant effect of UCB treatment at the 1- (P = .97) and 3-month (P = .26) time points. There was also substantial heterogeneity in the effect size (1-month τ² = 7.17; 3-month τ² = 1.07). Identified sources of this heterogeneity were route of UCB administration, baseline age, and CP severity. Similarly, the effect of receiving UCB with or without concomitant EPO therapy (as a combined treatment group compared with controls) was statistically significant at 6- and 12-month follow-up (6 months: 1.18 [95% CI, 0.32–2.03; P = .007]; 12 months: 1.23 [95% CI, 0.25–2.20; P = .014]) but not at 1 month (P = .95) or 3 months (P = .19) after intervention. To check whether inclusion of IPD from single-arm studies, with inherently higher potential for bias, influenced the overall treatment effect of UCB, we conducted a sensitivity analysis excluding these data. Similar results were observed with statistically significant findings at 6- and 12-month follow-up, but not at 1 or 3 months, with comparable mean effect sizes (Table S11). In addition, to check whether inclusion of data from the unpublished study (which could introduce substantial bias) influenced the study results, we conducted a sensitivity analysis excluding these data. Again, similar results were observed, with only minor changes to the mean effect sizes (Table S12).

Examination of the effect of cell dose on clinical improvement showed that mean GMFM-66 effect size vs controls increased with increasing TNC count for intravenous or intra-arterial UCB (Figure 2), and this response was statistically significant at 3- and 12-month follow-up (P < .001 and P = .047, respectively). Notably, across the 3-, 6- and 12-month time points, the mean effect size was close to or below 0 for cell doses less than ≈50 × 10⁶ TNCs/kg.

The mean difference in GMFM-66 between treatment groups differed significantly according to CP severity (GMFCS I-III vs GMFCS IV-V), with a greater effect of UCB in milder presentations at the 3-month (P_interaction = .003), 6-month (P_interaction = .047), and 12-month (P_interaction< .001) follow-up time points (Figure 3).

Given the possible confounding between baseline age and GMFCS level (P = .03, Table S5), we investigated the effect of UCB for both severity groups after adjusting for baseline age. At the 6- and 12-month follow-up time points, there were slightly reduced effects for the lower-severity group after adjustment, but the differences between the lower- and higher-severity groups remained statistically significant (6-month P_interaction = .031, 12-month P_interaction =.003) (Table S13). We likewise investigated the impact of adjusting for GA, given the association found between GA and GMFCS (P = .002), but adjusted effects were not significantly different from the unadjusted effects at any follow-up time point (Table S13).

Investigation of the effect of cell dose on GMFM-66 effect size across severity groups found a statistically significant difference between the GMFCS I-III vs GMFCS IV-V groups at the 3-month (P_interaction = .047) and 12-month follow-up time points (P_interaction = .028) (Figure S1). Specifically, there was a significantly increased GMFM-66 score when higher UCB cell doses were given in the GMFCS I-III group, whereas there was no significant effect of UCB treatment detected at any cell dose level in the GMFCS IV-V group.

When examining the effect of age at baseline, the mean GMFM-66 effect size differed significantly based on participant baseline age at the 3-month follow-up time point (P = .033) but not at 6 (P = .20) or 12 months (P = .11) (Figure S2). However, after adjusting for GMFCS, the change in effect of UCB treatment due to baseline age was no longer statistically significant at 3 months (P = .10) (Figure 4). Conversely, after adjustment, the change in effect of UCB treatment across baseline age became statistically significant at both the 6- (P = .035) and 12-month (P = .024) time points (Figure 4). In all cases, it was evident that the effect of UCB on GMFM-66 decreased with increasing age at baseline; ie, younger participants had increased response to UCB treatment.

When considering the impact of etiological factors on response to UCB treatment, there was a statistically significant effect of GA on GMFM-66 at 12-month follow-up (Table S14). Specifically, participants born term or moderate-to-late preterm (later GA) exhibited a greater effect of UCB treatment compared with those born extremely or very preterm (earlier GA) (P_interaction = .025). However, after adjusting for GMFCS, there was no longer a significant difference between the groups, suggesting the variability seen in the unadjusted analysis for the different GA groups is likely due to an association between prematurity and CP severity. No statistically significant between-group differences were observed at the 3- and 6-month follow-up time points, with or without adjustment (Table S14). Moreover, we found no significant difference in the effect of UCB treatment between participants based on the etiology information that could be extracted from the provided IPD (stroke/intraventricular hemorrhage or hypoxic ischemic encephalopathy vs other) at 3, 6, or 12 months in the unadjusted analysis, or after adjusting for GA, or cell dose (Table S15), nor based on CP type (Table S16).

Upon investigation of UCB characteristics, there appeared to be a significant difference in GMFM-66 effect size when comparing studies that administered UCB from an unrelated allogeneic source compared with autologous cells at both 6- and 12-month follow-up (Table S17), favoring unrelated allogeneic. However, investigation of the relationship between cell dose and UCB source revealed that those treated with unrelated allogeneic UCB received, on average, more than double the cell dose compared with autologous UCB (unrelated allogeneic: mean, 71.3 × 10⁶ TNCs/kg [SD, 43.3]; autologous: mean, 31.1 × 10⁶ TNCs/kg [SD, 15.8]; P < .001; Table S5). After adjusting for cell dose, the difference between the effect of unrelated allogeneic and autologous UCB was no longer statistically significant (Table S17). Notably, however, these results should be interpreted with caution given the small number of trials in which autologous cells were used.

For studies that used allogeneic UCB (including both related and unrelated donor), degree of HLA match was also interrogated. At 3, 6, and 12 months, there were no significant differences detected between those with no HLA mismatch compared with any mismatch (Table S18). Although there was an observably larger positive effect among those with no HLA mismatch (despite these participants receiving a significantly lower cell dose, P = .005, Table S5), wide CIs limit the interpretability of these results. After adjusting for dose, the difference in the effect increased but remained statistically insignificant, with wide CIs (Table S18).

Across all studies included in the systematic review, there was a total of 924 AEs and 87 SAEs reported (Table S19). Of the SAEs, 52 were experienced by participants who received UCB (n = 324, 16.0%) and 35 by participants who did not (n = 272, 12.9%). Thus, the rate of SAEs was similar between UCB- and non–UCB-treated participants. Only 1 SAE was definitely related to the UCB intervention, which was a treatable but serious adverse reaction to cryoprotectant (unilateral periorbital erythema and angioedema) that occurred at the time of infusion. Of note, there were at least 5 additional infusion reactions reported that were related to UCB treatment (eg, hives, fever, cough, nausea); however, these were not coded as SAEs due to their lesser severity.

The results of this IPDMA suggest overall benefit of UCB for improving gross motor function in individuals with CP greater than that for control participants, in both the presence and absence of concomitant EPO. UCB treatment showed most benefit 6–12 months after treatment, indicating that short-term clinical or trial follow-ups may not yield clinically or statistically significant findings. Given the hypothesized mechanism of action of UCB for treating the underlying brain injury in CP—namely, reduced neuroinflammation and stimulation of endogenous tissue repair via paracrine effects, leading to increased brain connectivity—the time taken to see clinical effect is justified. Improvements in brain connectivity are not immediate, and subsequent clinical improvements likely rely on the principles of neuroplasticity, with neural pathways strengthened by physical therapy over weeks to months. This is supported by analysis showing that UCB treatment increased normalized whole-brain connectivity in children with CP 1 year after treatment.^7,8 Importantly, the dose-response analysis indicates that treating with higher doses improves the efficacy of UCB without safety concerns, particularly in participants with mild CP. The pediatric and adult neurology fields have already shifted toward dosing at higher cell numbers given the robust safety data.³¹ This dose finding suggests that even higher dosing (eg, >150 × 10⁶ TNCs/kg) should be studied to elucidate maximal effect for improving gross motor function.

Our results also indicate that UCB treatment is safe, with similar rates of serious safety events reported across UCB-treated and control groups. Although rare, infusional toxicities can occur due to the traces of cryoprotectant in UCB units, even after unit washing before administration.³² These data emphasize the importance of infusion monitoring around the time of treatment, particularly for acute allergic reactions. Appropriate infusion setting, with access to emergency care where required, remains an integral consideration to uphold participant safety. We also recommend ongoing monitoring and reporting of safety data to identify rare AEs. This becomes especially relevant should the field move to very high dosing, which may require the use of multiple/pooled-donor UCB units and, thus, repeated donor exposure. In this instance, monitoring and capture of long-term safety and donor-specific anti-HLA antibodies may prove essential.

This IPDMA additionally identified participant characteristics of responders to UCB treatment. Namely, participants who were younger at baseline (before age ≈5 years) with milder CP (GMFCS I-III) had the most improved gross motor function after UCB treatment. In contrast, no effect of etiology was detected; however, this finding should be interpreted with caution because etiology reporting across studies was highly variable (ie, the type and quality of data provided), IPD included heterogeneous etiologies/causal pathways, there was a high proportion of missing data (although this was from a single study that exclusively enrolled children with white matter lesions on brain imaging), and untangling discrete etiological risk factors within an individual’s often-complex causal pathway to CP is a known challenge.³³ There does not exist a standard/consensus classification and reporting system for etiology in CP. It is hypothesized, although still largely unknown, that children with genetic and cryptogenic etiologies of CP may not benefit from UCB treatment. Indeed, in the included studies, children with known genetic disorders/syndromes, and even sometimes suspected genetic syndromes, were excluded. However, without detailed genomic analysis of potentially eligible participants, it is possible that some children with genetic contributions to their CP were inadvertently enrolled, and this may have altered the calculated effect size of UCB treatment. Future research should focus on collecting and reporting better etiology data so that potential benefits of treatments such as UCB can be better understood.

It is unsurprising that children with milder CP and of younger age showed improved response to UCB because this responsivity trend is seen in other CP interventions targeting gross motor function.³⁴ Regarding response in children with milder CP, this may be due in part to the ability for children with milder CP to participate in effective gross motor rehabilitation, leveraging the increased neuroplasticity and enhanced brain connectivity after UCB treatment. Moreover, children with milder CP reach 90% of their motor potential before age 5 years,³⁵ and early intervention harnessing neuroplasticity is accepted best practice.³⁶ Consistent with this, the UCB field has already shifted toward treating younger participants with less-severe CP. Indeed, the youngest participant in this analysis was just 8 months old at baseline; however, it is now established that an accurate diagnosis of CP can be before age 6 months (corrected for prematurity),³⁷ and the GMFM-66 is valid for use in CP from as early as age 5 months.³⁸ The natural history GMFM trajectories of children with CP, based on GMFCS level, are also well established.

Understanding why UCB treatment responsivity in CP is related to participant age is an area of research interest, particularly how age after initial injury relates to inflammatory status and immune changes over time. Notably, altered inflammatory profiles have been described in school-aged children with neurodevelopmental impairment who were born preterm,³⁹ and inflammation with altered immune function demonstrated to persist in children and even adults with CP,⁹ compared with controls, with more pronounced effects in younger children. Because UCB is postulated to provide therapeutic benefit via its paracrine effects and immune modulation,⁴ these findings support that UCB may be targeting elevated systemic inflammation in younger patients to provide the greatest benefit. Indeed, treating earlier may increase the feasibility of higher dosing per body weight and could allow for repeated UCB treatment during the therapeutic window (within the first few years of life). However, repeated UCB treatment for CP, especially when using UCB from multiple donors, requires more investigation in high-quality research studies to confirm safety and efficacy, given the theoretical risk of mounting an immune response with repeated exposure.

Encouragingly, this analysis suggests that donor source of UCB (autologous vs unrelated allogeneic) and HLA mismatch do not significantly affect safety or efficacy. Although the amount of available data limits the certainty of these findings, this is notable given the feasibility limitations of using autologous UCB, because most children with CP do not bank their UCB at birth. Moreover, the ability to use HLA-mismatched units enables greater choice of existing banked units with sufficient cell numbers. Given the declining use of UCB for treating hematological conditions,⁴⁰ these findings position banked UCB as a feasible, readily available treatment for improving gross motor function in CP and proposes a new indication for the existing stocks of high-quality UCB units in cord blood banks around the world.

It is important to consider the clinical significance of the reported findings, not just the statistical significance. Studies have calculated a minimum clinically important difference (MCID) for the GMFM-66 in comparable populations of children with CP after rehabilitation and surgical intervention.^41,42 When comparing the results of this study with previously published MCIDs, the GMFM-66 effect sizes after UCB treatment indicate medium-to-large MCIDs, particularly for the responder subgroups (eg, younger and milder), and 6–12 months after infusion. This suggests that the observed effect sizes are clinically relevant. Importantly, given the relatively low burden on participants to receive this intervention (eg, once-off infusion of ≈20 minutes), UCB treatment may be a preferred intervention choice by patients and families compared with time-intensive rehabilitation or highly invasive surgical procedures, with a similar financial cost.

Clinically meaningful improvement in gross motor function is a justified end point in research; however, UCB may convey other benefits for children with a brain injury that results in CP, warranting investigation. Indeed, a recent scoping review of outcomes reported in cell therapy clinical studies for CP, including UCB, identified a large range of outcome domains/categories.⁴³ These included outcomes measuring changes in fine motor, cognition, behavior, quality of life, and communication, with preclinical evidence to support some of these outcomes.^6,44 Due to high variability in the outcomes collected across studies, and resultant limited availability of data for other measures, it was not possible to analyze the effect of UCB on nonmotor outcomes in this IPDMA. More research is needed to better understand the potential effects of UCB on nonmotor outcomes that are important to individuals with CP and their families. This may be particularly relevant for older individuals and those with more severe CP receiving UCB treatment, who may not be expected to see significant improvements in gross motor function but may still experience changes in other domains.

We acknowledge several limitations of this IPDMA, including the participant and intervention heterogeneity across many variables and time points, leading to reduced power in subanalyses after adjustment. This IPDMA could also not control or evaluate the type of concomitant rehabilitation received by participants. Given that CP rehabilitation varies substantially across countries, with known widespread uptake of ineffective treatments, we could not tease apart the individual contribution of UCB compared with rehabilitation or ascertain whether rehabilitation produced an augmentative effect. Nevertheless, in the future, UCB is unlikely to be administered as a stand-alone treatment, as UCB paired with an intensive rehabilitation after infusion may harness greater neuroplasticity. In addition, the brain injury underlying CP may occur due to multiple causal pathways, which may respond differently to UCB, but the variability in coding of presumed etiological pathways in the original data sources precluded meaningful subgroup analysis. Moreover, the strength of the evidence, including the number of analyzed participants, and study quality may limit the interpretation of the results given the risk of bias in some included studies, as well as the inclusion of data from single-arm studies (including data from an unpublished study), which have a higher risk of bias especially from maturational effects. To maintain the benefits of a larger sample size but minimize confounding maturational effects, we adjusted single-group postinfusion GMFM scores using baseline scores, before data pooling with those treated in controlled studies. Nevertheless, we were able to confirm that there was no change in the overall effect of UCB treatment after excluding data from single-arm studies, confirm no change after excluding data specifically from the unpublished study, and verify that study quality did not significantly influence overall treatment effect in this analysis. Regardless, the findings of this IPDMA should be considered in light of the strength of the evidence, with acknowledgment that this would, no doubt, be improved via a well-designed, pivotal, phase 3 trial to determine the safety and effectiveness of UCB treatment for CP.

This IPDMA has helped to overcome the heterogeneity limitations of published clinical trial data to indicate safety and clinically meaningful efficacy of UCB for improving gross motor function in individuals with CP. Importantly, this analysis suggests best responders—namely, younger participants with milder CP—indicating who should be treated with UCB in future studies as well as in real-world implementation. The results will inform the design of pivotal clinical trials aimed at achieving regulatory approval. This will be achieved by enhancing feasibility through a reduced sample size and lower costs and by optimizing effect sizes through the targeting of known best responders.

We acknowledge Dr Anahita Majmaa, Dr Safdar Masoomi, and Mrs Elaheh Afzal for their assistance with data collection. We thank Professor Rod Hunt for assistance with etiology interpretation and coding and analysis advice.

AE

adverse event

CP

cerebral palsy

EPO

erythropoietin

GA

gestational age

GMFCS

Gross Motor Function Classification System

GMFM

Gross Motor Function Measure

HLA

human leukocyte antigen

IPD

individual participant data

IPDMA

individual participant data meta-analysis

MCID

minimum clinically important difference

PRISMA-IPD

Preferred Reporting Items for a Systematic Review and Meta-Analysis of Individual Participant Data

REDCap

Research Electronic Data Capture

SAE

serious adverse event

TNC

total nucleated cell

UCB

umbilical cord blood