Safety and efficacy of allogeneic umbilical cord-derived mesenchymal stem cell infusion for frailty: a phase 2, single-centre, randomised, open-label controlled trial

Summary

Background

Frailty syndrome in older adults is a growing public health concern associated with increased vulnerability and adverse outcomes. Umbilical cord-derived mesenchymal stem cell (UC-MSC) therapy shows promise in improving physical function and quality of life. This study evaluated the safety and efficacy of intravenous UC-MSC infusion in older frail adults.

Methods

A phase 2 randomised controlled trial at Vinmec Times City International General Hospital (2021–2024) enrolled 147 frail adults aged 60–85 (Modified Fried Criteria ≥3). Participants were randomly assigned to receive two intravenous UC-MSC infusions (1.5 × 10⁶ cells/kg, three months apart) plus oral supplements, or supplements alone, with nine months of follow-up. Safety was assessed by adverse events (AEs) and serious adverse events (SAEs). Efficacy outcomes included Short Physical Performance Battery (SPPB), handgrip strength, physical activity (CHAMPS), fatigue (MFI), knee function (WOMAC), and quality of life (SF-36).

Findings

No UC-MSC therapy-related SAEs occurred. Twelve mild AEs (headache, dizziness, chest discomfort) resolved spontaneously. At nine months, the UC-MSC group showed higher SPPB scores than controls (least-squares mean difference 1.1 points, 95% CI: 0.6–1.6).

Interpretation

In this study, UC-MSC infusion was associated with no safety concerns and was associated with higher SPPB scores than in the control group. These findings provide the necessary rationale and foundation for larger, double-blind, multicentre phase 3 trials.

Funding

Vingroup Joint Stock Company, grant ISC.21.11.

Research in context.

Evidence before this study

Before undertaking this trial, we searched PubMed/MEDLINE, Scopus, Web of Science, Embase, ClinicalTrials.gov, and reference lists of relevant articles for studies on frailty and mesenchymal stromal/stem cell therapy. The search covered studies published from database inception to 31 December 2024, with no language restrictions. Search terms included combinations of “frailty”, “physical frailty”, “older adults”, “ageing frailty”, “mesenchymal stem cells”, “mesenchymal stromal cells”, “umbilical cord-derived mesenchymal stem cells”, “umbilical cord tissue-derived mesenchymal stem cells”, “bone marrow-derived mesenchymal stem cells”, “cell therapy”, and “clinical trial”. We included preclinical studies relevant to biological mechanisms and clinical studies reporting safety, physical function, quality of life, or inflammatory outcomes in older adults with frailty treated with mesenchymal stromal/stem cells. We excluded studies not focused on frailty, studies using non-MSC cell products, conference abstracts without sufficient outcome data, and duplicate reports. The available preclinical evidence suggested that MSCs could improve muscle regeneration, modulate inflammation, and attenuate age-related functional decline. The available clinical evidence was limited to a small number of early-phase studies with modest sample sizes and relatively short follow-up, which consistently supported safety but provided only limited and imprecise evidence for efficacy. Owing to heterogeneity in cell source, dose, trial design, and outcome measures, a pooled quantitative meta-analysis was not considered appropriate for this study context.

Added value of this study

This phase 2 randomised controlled trial extends the available clinical evidence by evaluating two intravenous infusions of allogeneic umbilical cord-derived mesenchymal stem cells in a substantially larger cohort of older adults with frailty and with nine months of follow-up. Compared with controls, UC-MSC therapy was not associated with serious safety concerns and was associated with clinically meaningful improvements in physical performance, including SPPB score, gait speed, handgrip strength, physical activity, fatigue, knee-related symptoms, and general health status. The study also adds translational value by showing favourable changes in inflammatory and senescence-related biomarkers, supporting a plausible biological basis for the observed functional benefits.

Implications of all the available evidence

Taken together, the available preclinical, early-phase clinical, and current trial evidence suggests that UC-MSC therapy is a promising candidate for a disease-modifying approach to frailty, particularly through immunomodulatory and anti-inflammatory mechanisms. These findings support progression to larger, double-blind, multicentre phase 3 trials to confirm efficacy, define optimal dose and treatment schedule, assess durability of benefit, and determine cost-effectiveness and applicability in broader, more comorbid older populations. For human health, this work is relevant because frailty is a major cause of disability, loss of independence, and healthcare burden in ageing populations, and effective biological therapies remain limited.

Introduction

Frailty syndrome is a multifactorial geriatric condition characterised by declines in physiological reserve, reduced muscle strength, and impaired resilience to stressors. This syndrome substantially increases the risk of adverse health outcomes, including falls, hospitalisation, and mortality.1, 2, 3 As global populations age, frailty has become a major public health concern, with prevalence estimates among adults aged 65 years and older ranging from 4.9% to 27.3%.^4,5 The Fried phenotype, defining frailty as the presence of at least three of five criteria: unintentional weight loss, self-reported exhaustion, muscle weakness (commonly measured by handgrip strength), slow walking speed, and low physical activity, is widely used.⁶ Other tools, such as the Edmonton Frail Scale (EFS; score ≥ 9)^7,8 and Canadian Frailty Scale (CFS; range 1–9),⁹ are also employed.

The increasing prevalence of frailty poses considerable challenges for healthcare systems worldwide, affecting medical and socioeconomic outcomes. Effective management requires multidisciplinary strategies, including physical activity promotion, nutritional support, treatment of underlying medical conditions, and targeted social interventions.10, 11, 12 However, existing therapies have limited efficacy, especially in advanced frailty.

In recent years, stem cell therapy has emerged as a promising approach in regenerative medicine for age-related degeneration, including frailty. Preclinical studies using animal models, such as young bone marrow injections, have prevented muscle atrophy in senescence-accelerated mice¹³ and umbilical cord-derived mesenchymal stem cells (UC-MSCs), improving muscle function and reducing inflammation, provide strong mechanistic evidence.¹⁴ Additionally, MSC therapies mitigate bone loss and promote osteogenesis, which may also benefit frail patients.¹⁵

Initial clinical trials using bone marrow-derived MSCs have demonstrated safety and potential physical improvements, including reduced tumour necrosis factor-alpha (TNF-α) levels and enhanced physical performance.¹⁶ More recently, trials with allogeneic UC-MSCs reported improved quality of life and inflammatory profiles.^17,18 However, these studies were limited by small sample sizes and short follow-up durations.

Our large, randomised controlled trial addresses these limitations by evaluating the safety and efficacy of allogeneic UC-MSC therapy in frailty with an extended nine-month follow-up and comprehensive clinical and biomarker assessments. This study represents a significant translational advance by providing robust data supporting UC-MSCs as a potential disease-modifying treatment for frailty, with implications for regenerative medicine and age-related degenerative conditions.

Methods

Trial design

This phase 2 randomised, open-label, assessor-blinded, parallel-group clinical trial was conducted at Vinmec Times City International General Hospital, Hanoi, Vietnam. Participants were recruited from September 2021 to December 2023, and final follow-up for all outcomes was completed in December 2024. This was a single-centre trial conducted in a GCP-compliant stem cell research Institute and a tertiary care private hospital, where UC-MSC infusions were administered by experienced physicians and research nurses trained in infusion and emergency management. The trial protocol was prospectively registered at ClinicalTrials.gov (NCT04919135).

The initial protocol published by Hoang DM et al. planned for 44 participants.¹⁹ Following approval by the Vietnamese Ministry of Health, the sample size was increased to 158 participants to accommodate two distinct study phases (phase 1 with ten participants and phase 2 with 148 participants). The complete updated protocol and statistical analysis plan (SAP) are provided in the Supplementary Appendix 1 and Supplementary Appendix 2.

An independent Data Safety Monitoring Board (DSMB) oversaw the trial. The DSMB conducted a scheduled formal safety review after completion of phase 1 safety run-in (n = 10, with nine months of follow-up). The pre-defined criterion for progression to phase 2 was the absence of any treatment-related Serious Adverse Events (SAEs) or Grade ≥3 Adverse Events (AEs) attributable to the cell infusion. As safety was deemed acceptable based on these criteria, the DSMB formally endorsed the continuation to the randomised phase 2 efficacy phase. No emergency SAE-triggered reviews were subsequently required during phase 2 (Supplementary Appendix 3).

Specifically, the first 10 participants were enrolled in an open-label phase 1 safety study to confirm the safety of the infusion protocol (baseline characteristics and safety outcomes are summarised in Supplementary Table S1). The subsequent 148 participants were enrolled in the phase 2 randomised efficacy trial, which is the subject of this report. No further major amendments were made after these adjustments. The change ensured adequate power for evaluating the safety and efficacy of mesenchymal stem cell therapy for frailty. No patients or members of the public were involved in the design, conduct, reporting, or dissemination of this trial.

Sample size

The sample size was calculated for the primary endpoint (SPPB total score at nine months) following DELTA2 guidance for superiority trials.²⁰ We targeted an absolute between-group difference of 1.1 point (substantial MCID, small MCID 0.5 points) in the SPPB change from baseline, informed by prior frailty trials (mean baseline SPPB of 10.5 ± 2.2¹⁷). Using a two-sample t-test (two-sided α = 0.05, 85% power, SD = 2.2), 65 per group were required. With 14% attrition anticipated, the total recruitment was 148 participants (74 per group). No interim analyses or formal stopping guidelines were planned for this trial.

Patients

The inclusion criteria for this study were as follows: (1) participants aged 60–85 years at the time of signing the informed consent form; (2) participants meeting at least three criteria of the Modified Fried frailty phenotype; (3) participants with CFS scores ranging from 3 to 6; and (4) participants who provided written informed consent. Age and sex were self-reported by participants at enrolment.

Exclusion criteria included: (1) Mini-Mental State Examination (MMSE)²¹ score ≤20; (2) current listing or expected listing for organ transplantation; (3) abnormal laboratory findings, including haemoglobin <8 g/dL, white blood cell count <3000/mm³, platelet count <80,000/mm³, alkaline phosphatase >3 times the upper limit of normal, total bilirubin >1.5 mg/dL; (4) presence of severe comorbidities that, in the investigators’ judgement, could compromise patient safety, compliance, or study completion, such as HIV infection, liver failure, renal failure, heart failure (NYHA class II-IV), myocardial infarction, chronic obstructive pulmonary disease (COPD) classified as GOLD stage C or D, ischaemic stroke with NIHSS >5, type 2 diabetes mellitus with HbA1c > 8.5%, or acute infection; (5) history of malignancy within the past five years; and (6) other medical conditions with an expected survival of less than one year.

Intervention

UC-MSC infusion group

Preparation and characterisation of UC-MSCs

UC-MSCs were sourced from a high-quality master biobank, isolated and screened from 30 umbilical cords of full-term, healthy newborns free of infectious diseases.²² The UC-MSCs line 25 (UC-25) were cultured in serum-free and animal component-free conditions, and passed all quality control tests, as described previously.²³ The cells were expanded at large scale up to passage 5 (P5) and cryopreserved in CryoStore CS10 reagent (Stem Cell Technology, Canada) in the vapour phase of liquid nitrogen at −196 °C in a Brooks system (Brooks Life Science, USA). Each manufacturing lot was required to meet the release criteria for UC-MSC products, including cell viability (≥85%), surface marker expression (≥95% of cells expressing CD73, CD90, and CD105, and ≤2% of cells expressing CD45, CD34, CD11b, CD19, and HLA-DR), sterility testing (negative for microorganisms, including bacteria, fungi, and mycoplasma), and confirmation of a normal karyotype (Supplementary Figs. S1–S3 and Table S2). Before infusion, UC-MSCs were thawed, washed, counted, and resuspended in 50 mL of Ringer’s lactate for intravenous (IV) administration at a dose of 1.5 × 10⁶ cells per kilogram of body weight. Quality control of the released UC-MSC product included the following criteria: cell viability of at least 70%. Endotoxin levels were ≤5 EU/kg body weight/hour for IV administration and absent of bacterial, fungal, and mycoplasma contamination (Supplementary Figs. S1 and S2).

Administration of UC-MSCs

Patients in the intervention group received intravenous infusions of UC-MSCs at a dose of 1.5 × 10⁶ cells/kg body weight (measured via Day 0 scale weight, within ±5%), resuspended in 50 mL of infusion solution. The pre-cryopreservation cell count, together with post-thaw cell viability and post-thaw cell number, ensured that the intended infusion dose was achieved. This dose (∼80 million cells for a typical 54 kg patient) was selected from phase 1 safety data (n = 10, mean weight 53.6 kg) and prior frailty MSC trials showing optimal efficacy/safety.¹⁷ Each infusion was administered over 30 min, with two infusions given at a three-month interval. Two infusions spaced three months apart enabled first-dose response assessment and recapitulated anti-inflammatory kinetics documented at 3–6 months in prior frailty MSC trials.^16,18

Before UC-MSC infusion, patients received oral Xarelto 10 mg (anticoagulation) and intravenous Dimedrol 20 mg (allergy prophylaxis), administered 45–60 min prior to treatment. These single pre-infusion doses were given to reduce the risk of infusion-related thrombosis in frail older adults, who commonly have endothelial dysfunction and an increased baseline risk of venous thromboembolism.^24,25 Clinical examination and routine laboratory tests (haemoglobin, coagulation profile, D-dimer) were used to monitor for bleeding, and no major bleeding events were observed.

In addition, all patients received oral supplementation with HighTamine, Total Calcium, Bioflex, and Zinc for 9 months.

Control group

Patients in the control group did not receive UC-MSC infusion. Instead, they were administered oral supplementation with High-Tamine, Total Calcium, Bioflex, and Zinc for 9 months.

Follow-up and monitoring

All patients were followed for 9 months postintervention, with follow-up visits scheduled at baseline, 1, 3, 6, and 9 months.

Outcome measures

Primary efficacy endpoint

The primary endpoints were safety, defined as incidence and severity of adverse events (AEs) and serious adverse events (SAEs) over nine months, and efficacy, defined as the SPPB total score at nine months post-first infusion (assessed at baseline, one month, three months, six months, and nine months).

Safety

Safety was assessed by monitoring AEs and SAEs throughout the study. While initial grading utilised the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 4.03²⁶ Per protocol, all events reported in this manuscript were re-evaluated and mapped to CTCAE version 5.0.²⁷ This update required no reclassification of event severity.

Efficacy

Efficacy was assessed using a comprehensive set of validated outcome measures, including the Short Physical Performance Battery (SPPB), handgrip strength, the Community Healthy Activities Model Program for Seniors (CHAMPS) questionnaire, frailty status (CFS), the Multidimensional Fatigue Inventory (MFI), the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), the 36-Item Short Form Health Survey (SF-36), left ventricular ejection fraction (LVEF), pulmonary function tests (FEV1/FVC), plasma cytokine, chemokine profiling, p16^INK4a mRNA expression in CD3⁺ T cells, mitochondrial function.

Mobility assessment

Lower extremity function was evaluated using the Short Physical Performance Battery (SPPB),²⁸ which comprises three components: balance tests, chair stand test, and gait speed. Each component is scored from 0 to 4, yielding a total score ranging from 0 to 12, with higher scores indicating better physical performance. A change of 0.5 points is considered a small meaningful change, and 1.0 point represents a substantial meaningful change in older adults.²⁹

For the gait speed assessment, participants were instructed to walk 4 m at their usual pace, and the time to complete the distance was recorded in seconds. A reduction in the time required to complete the walk reflects improved gait speed, indicating enhanced mobility and physical function. Improvements of 0.05 m/s and 0.10 m/s are defined as small and substantial clinically meaningful changes, respectively²⁹ (Supplementary Fig. S4).

Handgrip strength

Handgrip strength³⁰ was assessed using a calibrated hand dynamometer. Participants performed two maximal grip attempts with each hand, and the highest value obtained for each hand was used for analysis. Results were recorded in kilograms (kg). An increase in handgrip strength indicates an improvement in upper limb muscle function and overall physical performance.

Physical activity assessment

Physical activity was evaluated using the Community Healthy Activities Model Program for Seniors (CHAMPS)³¹ questionnaire. Participants reported the frequency and duration of 41 daily and recreational activities performed over the previous four weeks. Weekly energy expenditure for moderate-to-vigorous activities (MET ≥ 3.0) was calculated and expressed in kilocalories per week (kcal/week). Higher CHAMPS scores indicate greater levels of physical activity and better functional capacity.

The clinical frailty scale (CFS)

The Clinical Frailty Scale (CFS) is a validated 9-point scale ranging from 1 (Very Fit) to 9 (Terminally Ill), with higher scores indicating greater frailty severity. Frailty status was assessed at baseline and all follow-up visits to monitor longitudinal changes.³²

Fatigue assessment

Fatigue was assessed using the Multidimensional Fatigue Inventory (MFI),³³ a validated 20-item questionnaire that evaluates five dimensions: general fatigue, physical fatigue, mental fatigue, reduced motivation, and reduced activity. Each item is rated on a 5-point Likert scale, resulting in subscale and total scores ranging from 4 to 20 and 20 to 100, respectively. Higher scores indicate greater fatigue severity.

Knee pain and function

Knee pain, stiffness, and physical function were assessed using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC),³⁴ a validated 24-item questionnaire comprising three subscales: pain (5 items), stiffness (2 items), and physical function (17 items). Each item is scored on a 5-point Likert scale (0–4), with higher scores reflecting greater symptom severity or functional impairment. Subscale and total scores are calculated by summing the relevant items, with higher scores indicating worse outcomes.

Health-related quality of life

Health-related quality of life was assessed using the 36-Item Short Form Health Survey (SF-36).³⁵ For this study, we specifically analysed the “general health” domain. Scores for general health were transformed to a 0–100 scale, with higher scores indicating better perceived health status.

Left ventricular ejection fraction

Left ventricular ejection fraction (LVEF)³⁶ was assessed to evaluate cardiac systolic function using standard transthoracic echocardiography. LVEF was calculated via the Simpson’s biplane method, the current guideline-recommended technique, by measuring end-diastolic and end-systolic volumes. Values are expressed as percentages (%), with higher values (≥55%) indicating preserved systolic function and lower values (<50%) suggesting impaired cardiac performance.

Pulmonary function (FEV1/FVC)

Pulmonary function³⁷ was assessed using standard spirometry, measuring forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC). The FEV₁/FVC ratio was calculated to evaluate the presence and degree of airflow limitation. Values are expressed as percentages (%), with higher ratios indicating better pulmonary function and lower risk of obstructive airway disease.

Blood sample analysis

For each visit (baseline, 1, 3, 6, and 9 months), about 20 mL of peripheral blood from each participant was collected in EDTA tubes. The blood samples were centrifuged at 1000 g for 10 min at 4 °C, and plasma was collected and stored at −80 °C until analysis. Plasma concentrations of the following cytokines and chemokines were quantified using Human ProcartaPlex (Invitrogen) on a Luminex 200 system: interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-10 (IL-10), tumour necrosis factor alpha (TNF-α), and monocyte chemoattractant protein-1 (MCP-1). Measurements were performed according to the manufacturer's instructions.

The remaining blood cells were used to isolate PBMCs by Ficoll density gradient centrifugation. At the indicated time points, one aliquot of PBMCs was analysed by flow cytometry, and CD3⁺ cells were subsequently isolated from the remaining PBMCs using CD3 microbeads (Miltenyi Biotec) for p16^INK4a mRNA and mitochondrial function analyses.

To check the composition of immune cell populations in patients after UC-MSC infusion, PBMCs were stained with antibody panels including CD45, CD3, CD56, CD19, CD14, CD4, and CD8 (Miltenyi Biotec). Dead cells were excluded using 7-AAD (Miltenyi Biotec). Immune cell compositions were analysed using MACSQuant Analyser 10 (Miltenyi Biotec) and FlowJo software (BD Biosciences).

To check the expression of p16^INK4a mRNA in CD3+ T cells, the total RNA was extracted using TRIzol™ Reagent (Invitrogen), reverse-transcribed into cDNA using SuperScript™ IV VILO with ezDnase Master Mix (Thermo Fisher Scientific), and quantified by real-time PCR using Power SYBR™ Green PCR Master Mix (Applied Biosystems) on an ABI 7500 Real-time PCR system. Expression levels were normalised to the reference gene GAPDH. Primers for p16^INK4a were: forward: GAAGGTCCCTCAGACATCCCC, reverse: CCCTGTAGGACCTTCGGTGAC; primers for GAPDH were: forward: GGCATGGACTGTGGTCATGAG; reverse: TGCACCACCAACTGCTTAGC.

Mitochondrial function in CD3+ T cells was assessed using the Seahorse XF Mito Stress Test Kits (Agilent Technologies) on a Seahorse XFe96 Analyser, following the manufacturer’s instructions. Parameters measured included basal respiration, ATP-linked respiration, maximal respiration, and spare respiratory capacity.

Randomisation

Participants were randomly allocated in a 1:1 ratio to receive either intravenous UC-MSC therapy or a control using computer-generated random numbers with permuted block randomisation to ensure balanced group allocation throughout the recruitment period. The random allocation sequence was generated by an independent trial statistician not otherwise involved in patient recruitment or assessment. Sequentially numbered, opaque, sealed envelopes prepared by this statistician were used to conceal allocation until assignment. Investigators enrolled patients, and research nurses opened the envelopes and assigned the allocated intervention. No emergency unblinding occurred (no SAEs, DSMB authorised if needed).

Blinding

Because the stem cell product had a distinct appearance compared to the control supplements, double blinding was not feasible. Thus, the study was open-label regarding participants and treating clinicians. To minimise detection and performance bias, outcome assessment was performed in a blinded manner. All efficacy endpoints (SPPB, walk speed, handgrip strength) were evaluated independently by physicians and rehabilitation technicians who were strictly blinded to the treatment allocation and not involved in intervention delivery.

Data management

Data were captured using REDCap v12.1.0 hosted at Vinmec Research Institute servers. Double data entry (90% of records), automated range checks, and real-time SAE alerts to DSMB ensured data integrity. Database locked December 15, 2024 (Supplementary Appendix 2).

Statistics

Analysis populations were intention-to-treat (ITT: all 148 randomised participants), modified ITT (mITT: all 147 dosed participants), and safety (all dosed participants). The estimand was the treatment policy estimand: between-group difference in observed SPPB total score at nine months, irrespective of intercurrent events.

Baseline characteristics were summarised using descriptive statistics. Group differences in categorical variables were assessed with the chi-square test, while continuous variables were compared using independent two-sample t-tests when normality assumptions were satisfied and the Wilcoxon rank-sum test otherwise.

No missing data occurred (0% attrition): all 148 randomised participants completed 9-month follow-up, although one UC-MSC participant withdrew consent pre-treatment (73 treated, 74 control). Complete-case analysis was used for all outcomes.

Longitudinal analysis: Fixed effects comprised treatment group, visit (1, 3, 6, 9 months), treatment-by-visit interaction, baseline value, and baseline-by-visit interaction; random intercept per participant. Models used REML estimation with Kenward-Roger degrees of freedom and an unstructured covariance structure. Least squares mean calculated via margins (α = 0.05). No multiplicity adjustment applied (single primary endpoint).

The primary efficacy endpoint was the between-group difference in SPPB total score at 9 months after the first UC-MSC infusion. Safety outcomes were summarised by treatment group. Incidence rates of adverse events were calculated as the number of patients with at least one event divided by the total number of patients in the safety population, and are presented with 95% confidence intervals (CIs) calculated using continuity-corrected Wilson score intervals. All efficacy analyses used the intention-to-treat population, defined as all randomised participants, and the safety population included all participants who received at least one UC-MSC infusion or at least one dose of supplements. No participants were lost to follow-up; therefore, all analyses used complete case data, and no imputation for missing values was required. Exploratory Spearman correlations between biomarker changes (nine months-baseline) and ΔSPPB were calculated in the UC-MSC group.

Subgroup analyses by baseline CFS score (3, 4, or ≥5) were prespecified and are presented in the Supplementary material. Individual longitudinal trajectories for physiologic outcomes and biomarkers were visualised using spaghetti plots to assess the consistency of treatment effects across participants. All analyses were conducted using STATA, and statistical significance was defined as a two-sided p-value < 0.05.

Ethics

The study was conducted in accordance with the Declaration of Helsinki and ICH-GCP guidelines. The trial was approved by the Vinmec Institutional Review Board (No. 75/2021/QĐ-VMEC) and the National Ethics Committee (No. 4191/QĐ-BYT, No. 2909/QĐ-BYT). Prior to enrolment, all participants provided written informed consent. The consenting process included a clear and comprehensive explanation of the study’s objectives, procedures, potential risks, and benefits, and participants were encouraged to ask questions before signing. Participation did not incur any costs—covering umbilical cord collection and processing, cell culture and infusion, clinical examinations, hospital stays, laboratory tests, imaging, rehabilitation therapy, or follow-up visits.

Role of funders

This research was funded by Vingroup Joint Stock Company under grant number ISC.21.11. The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Baseline characteristics

After a preceding open-label phase 1 safety study (n = 10), a total of 148 separate patients were enrolled in the phase 2 trial and randomised to either two intravenous UC-MSC infusions plus supplements (n = 74) or supplements alone (n = 74). One patient in the UC-MSC group withdrew consent after randomisation but prior to the first infusion. The remaining 73 patients received two intravenous UC-MSC infusions spaced 3 months apart. All participants in both the treated UC-MSC cohort (n = 73) and the control group (n = 74) completed all scheduled follow-up visits at 1, 3, 6, and 9 months. No patients were lost to follow-up after treatment initiation. The trial ended as planned after the target sample size had been enrolled and all follow-up visits had been completed (Fig. 1).

Fig. 1.

Trial profile of the phase 2 efficacy study (CONSORT flow diagram). This flowchart depicts the enrolment and follow-up of the 148 participants in the phase 2 trial (data from the preceding phase 1 safety study, n = 10, are excluded). Of 166 patients assessed for eligibility, 148 met the inclusion criteria and were randomised (1:1) to receive either allogeneic umbilical cord-derived mesenchymal stem cell (UC-MSC) infusion plus standard supplements (n = 74) or standard supplements alone (control group, n = 74). In the UC-MSC group, one patient withdrew consent prior to the first infusion. The remaining 73 patients received two intravenous infusions spaced 3 months apart. All randomised participants in the control group (n = 74) and the treated UC-MSC cohort (n = 73) completed scheduled follow-up visits at 1, 3, 6, and 9 months. No patients were lost to follow-up after treatment initiation.

Baseline demographic and clinical variables were well balanced between groups (p > 0.05). The mean age was 68.2 ± 6.3 years in the UC-MSC group and 68.7 ± 6.1 years in the control group. Females comprised 71.2% and 64.9% of each group, respectively. There were no statistically significant differences between groups across functional, physical, or cognitive measures, including SPPB total score, 4-m walking test, handgrip strength, body weight, physical activity (CHAMPS), MFI total score, WOMAC total score, SF-36, MMSE score, FEV1/FVC ratio, and LVEF. The distribution of CFS grades was balanced between groups: the UC-MSC cohort comprised 58.9% grade 3, 32.9% grade 4, and 8.2% grade ≥5, while the control cohort comprised 56.8% grade 3, 33.8% grade 4, and 9.5% grade ≥5 (p > 0.05). Participants exhibited a substantial burden of age-related comorbidities, predominantly hypertension (ranging from 56.5% to 73.7% across groups). Other metabolic and musculoskeletal conditions, including diabetes and osteoarthritis, were also present but less frequent. The overall comorbidity profile did not differ statistically significantly between groups (p > 0.05) (Table 1).

Table 1.

Baseline demographic and clinical characteristics of the modified intention-to-treat population (n = 147).

Variable	UC-MSC group (n = 73)	Control group (n = 74)
Age, years	68.2 ± 6.3	68.7 ± 6.1
Height, cm	155.2 ± 7.6	154.7 ± 6.6
BMI, kg/m2	23.4 ± 2.9	23.0 ± 2.7
Female, n (%)	52 (71.2)	48 (64.9)
SPPB total score	7.6 ± 1.6	7.5 ± 1.5
4-m walking test, second	7.6 ± 2.3	7.5 ± 2.6
Handgrip strength, kg	14.9 ± 5.8	16.3 ± 5.9
Body weight, kg	56.5 ± 9.0	55.1 ± 7.6
CHAMPS, kcal/week	1527.6 ± 715.9	1662.6 ± 742.1
MFI total score	14.6 ± 1.3	14.5 ± 1.8
WOMAC total (median, IQR)	22.9 (6.3–41.7)	28.1 (14.6–45.8)
SF-36 general health	25 (20–30)	25 (20–30)
MMSE score	26.6 ± 1.8	26.1 ± 1.5
FEV1/FVC, %	77.9 ± 6.5	78.2 ± 7.7
LVEF, %	69.0 ± 7.2	70.1 ± 6.1
CFS grade 3, n (%)	43 (58.9)	42 (56.8)
CFS grade 4, n (%)	24 (32.9)	25 (33.8)
CFS grade ≥5, n (%)	6 (8.2)	7 (9.5)
Hypertension	26 (56.5)	28 (73.7)
Diabetes mellitus	7 (15.2)	2 (5.3)
Osteoarthritis	5 (10.9)	2 (5.3)
Osteoporosis	3 (6.5)	0 (0)
Others	5 (10.9)	6 (15.8)

Values are mean ± SD (standard deviation) or median (IQR, interquartile range), or n (%). Modified ITT population (n = 147). One UC-MSC participant withdrew consent after randomisation but before treatment (baseline data included). Study conducted in Hanoi, Vietnam (100% Kinh ethnicity). Abbreviations: CHAMPS, Community Healthy Activities Model Program for Seniors; SPPB, Short Physical Performance Battery; MFI, Multidimensional Fatigue Inventory; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index; SF-36, Short Form-36; MMSE, Mini-Mental State Examination; FEV1/FVC, ratio of forced expiratory volume in 1 s to forced vital capacity; LVEF, left ventricular ejection fraction; CFS, Clinical Frailty Scale; BMI, body mass index; cm, centimetre; kg, kilogram.

Baseline levels of inflammatory and senescence biomarkers were also balanced between the UC-MSC and control groups (Supplementary Table S3).

Cell quality assessment

A total of 20 UC-MSC batches were manufactured to support 146 infusions in 73 patients. All produced batches met the predefined release criteria (Supplementary Table S2; Supplementary Figs. S1–S3). Expression of MSC-positive markers (CD73, CD90, CD105) exceeded 95%, while expression of negative markers (CD34, CD45, CD19, CD11b, HLA-DR) was below 2%. All batches were free of bacterial, fungal, and mycoplasma contamination and exhibited normal karyotypes.

Post-thaw quality assessment demonstrated consistently high cell quality, with mean viabilities of 91.8 ± 2.2% and 90.5 ± 1.0% for the first and second infusions, respectively. No microbial contamination was detected, endotoxin levels remained below 5 EU/kg/hour, and all administered cell doses complied with the study protocol (Supplementary Table S4).

AE and SAEs

No SAEs were reported in either group throughout the study period. In the UC-MSC group, 19 patients (26.0%; 95% CI: 17.3–37.1) experienced at least one adverse event, compared with 6 patients (8.1%; 95% CI: 3.8–16.6) in the control group. Treatment-related AEs occurred in 10 patients (13.7%; 95% CI: 7.6–23.4), comprising 12 mild, transient events (headache n = 6, dizziness n = 3, chest discomfort n = 3), with eight events after the first infusion and four after the second. All resolved spontaneously without sequelae. Additionally, nine patients (12.3%; 95% CI: 6.6–21.8) in the UC-MSC group and six patients (8.1%; 95% CI: 3.8–16.6) in the control group experienced 14 and 7 non-treatment-related AEs, respectively, reflecting intercurrent conditions such as infections, musculoskeletal problems, and other age-related morbidities. All reported AEs were CTCAE v5.0 grade 1, and no grade 3–5 events occurred. No deaths or thromboembolic events were observed in either group (Table 2, Supplementary Table S5).

Table 2.

Summary of adverse and serious adverse events.

Event category	UC-MSC (n = 73)	Control (n = 74)	Difference in proportions, % (95% CI)
Patients with ≥1 AE, n (%)	19 (26.0) [17.3–37.1]	6 (8.1)[3.8–16.6]	17.9 [6.1–29.8]
Patients with treatment-related AEs, n (%)	10 (13.7) [7.6–23.4]	0 [0.0–5.0]	13.7 [4.6–22.7]
Headache (events)	6	0	–
Dizziness (events)	3	0	–
Chest discomfort (events)	3	0	–
Patients with non-treatment AEs, n (%)	9 (12.3) [6.6–21.8]	6 (8.1) [3.8–16.6]	4.2 [−5.5 to 14.0]
Non-treatment events	14	7
SAE, n (%)	0 [0.0–5.0]	0 [0.0–5.0]	0.0 [−4.9 to 5.0]
Death, n (%)	0 [0.0–5.0]	0 [0.0–5.0]	0.0 [−4.9 to 5.0]
Thromboembolic events, n (%)	0 [0.0–5.0]	0 [0.0–5.0]	0.0 [−4.9 to 5.0]

CI = Confidence Interval; CTCAE = Common Terminology Criteria for Adverse Events; SAE = Serious Adverse Event; UC-MSC = Umbilical Cord-derived Mesenchymal Stem Cell. Note: Data are n, n (%), or n (%) [95% CI], as indicated. Percentages and CIs calculated using safety population (UC-MSC n = 73; control n = 74) based on the number of patients with ≥1 event of each type; total event counts shown additionally. Difference in proportions (treatment − control) with 95% CI calculated using continuity-corrected Wilson score interval method. Treatment-related AEs represent anticipated adverse events of special interest (AESIs) monitored during the study, whereas non-treatment AEs reflect other study-emergent conditions. Headache (4 events after first infusion, 2 after second); dizziness (2 after first, 1 after second); chest discomfort (2 after first, 1 after second). All AEs were CTCAE grade 1; no grade ≥3 AEs, SAEs, deaths, or thromboembolic events occurred.

Efficacy

At baseline, demographic and clinical characteristics were comparable between groups. Over the 9-month follow-up, the UC-MSC group consistently demonstrated superior functional improvements compared to the control group across multiple domains, as shown in the individual patient trajectories for physical performance and patient-reported outcomes (Fig. 2, Supplementary Figs. S5 and S6).

Fig. 2.

MMRM-adjusted mean trajectories of functional outcomes. MMRM: least-squares means ± 95% CIs from mixed-effects model repeated measures (fixed effects: treatment, visit, treatment × visit interaction, baseline; random intercept; REML). Control (red lines), UC-MSC (blue lines); individual patient trajectories. Safety population (UC-MSC n = 73, control n = 74). SPPB total score, 4 m walk speed, handgrip strength, CHAMPS total, MFI, CFS, WOMAC, SF-36.

Mixed-effects model analysis confirmed these observations. For physical performance, the UC-MSC group showed statistically significantly greater improvement in SPPB total scores, with a least-squares mean difference (MD) of +1.1 points (95% CI: 0.6; 1.6) at three months, +1.1 points (95% CI: 0.6; 1.6) at six months, and +0.8 (0.3; 1.3) at nine months compared to controls. Notably, the lower bound of these confidence intervals (+0.6) exceeded the established minimal clinically important difference (MCID) of 0.5 points. Gait speed (4-m walk) also improved more in the treatment group, with the largest statistically significant difference observed at 9 months (MD = −0.8 s; 95% CI: −1.4; −0.2). Similarly, handgrip strength was consistently higher in the UC-MSC group, with differences ranging from +2.6 kg to +4.0 kg across all time points. Regarding patient-reported outcomes, participants receiving UC-MSC therapy reported substantially higher physical activity levels (CHAMPS: MD = +1402.0 kcal/week at 3 months; 95% CI: 950.3; 1853.8) and lower fatigue scores (MFI: MD = −4.0 points; 95% CI: −4.7; −3.2) compared to controls. CFS showed sustained statistically significant reductions in the intervention group from 3 months onwards. Although WOMAC scores showed variability, a supportive difference in favour of the UC-MSC group was observed at 9 months (MD = −6.4; 95% CI: −11.6; −1.2). Finally, general health perception (SF-36) was markedly better in the UC-MSC group, with a mean difference of +22.5 points (95% CI: 17.7; 27.2) at 3 months (Table 3).

Table 3.

Analysis of clinical primary and secondary efficacy outcomes over time between UC-MSC and control groups using a mixed-effects model, intention-to-treat population.

Outcome	1 month	p	3 months	p	6 months	p	9 months	p
SPPB total score (primary)	+0.6 (0.1; 1.1)	0.02	+1.1 (0.6; 1.6)	<0.001	+1.1 (0.6; 1.6)	<0.001	+0.8 (0.3; 1.3)	0.002
4-m walk speed, second	−0.5 (−1.1; 0.1)	0.105	−0.8 (−1.4; (−0.2))	0.008	−0.7 (−1.3; (−0.1))	0.021	−0.8 (−1.4; (−0.2))	0.012
Handgrip, kg	+2.9 (1.3; 4.6)	0.001	+4.0 (2.3; 5.7)	<0.001	+2.6 (0.9; 4.2)	0.003	+2.7 (1.1; 4.4)	0.001
CHAMPS, (kcal/week)	+1157.7 (705.9; 1609.4)	<0.001	+1402.0 (950.3; 1853.8)	<0.001	+706.3 (250.0; 1162.6)	0.002	+764.7 (308.5; 1221.0)	0.001
MFI total score	−2.3 (−3.0; (−1.6))	<0.001	−4.0 (−4.7; (−3.2))	<0.001	−2.9 (−3.6; (−2.2))	<0.001	−2.1 (−2.8; (−1.4))	<0.001
CFS score	−0.2 (−0.4; 0.01)	0.06	−0.4 (−0.6; (−0.2))	<0.001	−0.3 (−0.5; (−0.1))	0.006	−0.2 (−0.5; (−0.02))	0.032
WOMAC total	−3.4 (−8.6; 1.8)	0.203	−4.7 (−9.9; 0.5)	0.079	−5.5 (−10.7; (−0.3))	0.039	−6.4 (−11.6; (−1.2))	0.017
SF-36 general health	+13.0 (8.2; 17.8)	<0.001	+22.5 (17.7; 27.2)	<0.001	+12.4 (7.7; 17.2)	<0.001	+10.6 (5.9; 15.4)	<0.001

MD, mean difference (UC-MSC minus control); CI, confidence interval; SPPB, Short Physical Performance Battery; CHAMPS, Community Healthy Activities Model Program for Seniors; MFI, Multidimensional Fatigue Inventory; CFS, Clinical Frailty Scale; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index; SF-36, Short-Form 36. Notes: Data are presented as least-squares mean difference (95% confidence interval) between groups, derived from a mixed-effects model for repeated measures (MMRM) adjusted for baseline score as a covariate. Positive MD values indicate greater improvement in the UC-MSC group. For 4-m walk, negative MD denotes faster time in the UC-MSC group. Negative values for MFI, CFS, and WOMAC represent improvement. All differences favour the UC-MSC group.

Pulmonary function (FEV₁/FVC) and left ventricular ejection fraction (LVEF)

There were no statistically significant between-group differences in FEV₁/FVC ratio or left ventricular ejection fraction (LVEF) at any time point during follow-up (Supplementary Fig. S7).

Changes in biomarkers

Longitudinal visualisation of p16^INK4a, inflammatory cytokines, and mitochondrial function markers revealed distinct patterns of improvement at the individual level (Supplementary Figs. S8–S10).

Mixed-effects model analysis demonstrated that the UC-MSC group exhibited a statistically significant reduction in p16^INK4a mRNA expression compared to controls at 3 months, with a least-squares mean difference (MD) of −23% (95% CI: −42; −4). Regarding mitochondrial function, the most notable improvements were observed at 6 months, where maximal respiration and spare respiratory capacity were both higher in the UC-MSC group than in controls (p = 0.094 and p = 0.057, respectively), with mean differences of +0.2 nmol/min/μg protein (95% CI: −0.04; 0.5) and +0.2 nmol/min/μg protein (95% CI: 0.01; 0.4), respectively. The most pronounced anti-inflammatory effects were observed at 9 months. Most notably, TNF-α levels were statistically significantly reduced in the UC-MSC group compared to controls (MD = −8.3 pg/mL; 95% CI: −15.9; −0.7; p = 0.032), alongside statistically significant reductions in IL-6 (MD = −37.6 pg/mL; 95% CI: −62.8; −12.4; p = 0.004) and IL-10 (MD = −3.4 pg/mL; 95% CI: −5.7; −1.0; p = 0.005). No statistically significant between-group differences were observed for IL-8 or MCP-1 at any time point (Supplementary Table S6).

In addition, the composition of immune cell populations was assessed by flow cytometry at baseline and 3, 6, and 9 months post-treatment in both the UC-MSC and control groups. No statistically significant differences were detected between the two groups at any of the analysed time points (Supplementary Fig. S11 and Table S7).

Exploratory analyses within the UC-MSC group revealed weak inverse correlations between improvements in physical function (ΔSPPB) and changes in several biomarkers, including TNF-α (ρ = −0.19; p = 0.262), spare respiratory capacity (ρ = −0.17; p = 0.316), and maximal respiration (ρ = −0.18; p = 0.291) over 9 months, although these associations were not statistically significant (Supplementary Table S8).

Impact of CFS subgroup on clinical and biomarker outcomes

Subgroup analyses stratified by baseline frailty severity (CFS 3, 4, or ≥5) revealed differential treatment effects. The CFS 4 subgroup showed the most robust early benefits compared with the control group. At 3 months, the between-group difference in SPPB total score and handgrip strength was statistically significantly larger in the CFS 4 subgroup compared to others (MD vs. control: +1.5 points [95% CI: 0.7; 2.3] and +4.5 kg [95% CI: 1.1; 7.9], respectively). Regarding physical activity, the most notable treatment effect was observed in the CFS 5 subgroup at 6 months, where CHAMPS scores were statistically significantly higher in the UC-MSC group compared to controls (MD: +1870.9 kcal/week; 95% CI: 126.6; 3615.2). Similarly, the treatment effect on general health (SF-36) was most consistent in the CFS 4 subgroup, peaking at 3 months with a mean difference of +24.7 points (95% CI: 16.8; 32.6) compared to controls (Supplementary Fig. S12 and Table S9).

Biomarker responses also varied by baseline frailty status. Regarding cellular senescence, the CFS 3 subgroup showed a statistically significant treatment-associated reduction in p16^INK4a mRNA expression at 3 months compared to controls (MD: −0.3; 95% CI: −0.6; −0.0). The most pronounced anti-inflammatory effects relative to controls were observed in the CFS 5 subgroup at 1 month, with statistically significantly lower levels of IL-6 (MD: −127.8 pg/mL; 95% CI: −245.9; −9.7) and TNF-α (MD: −73.2 pg/mL; 95% CI: −139.9; −6.5) in the UC-MSC arm. In contrast, sustained suppression of IL-6 at 9 months compared to controls was most evident in the CFS 4 subgroup (MD: −46.9 pg/mL; 95% CI: −90.7; −3.1) (Supplementary Table S10).

Discussion

Our findings demonstrate that UC-MSC administration is both safe and effective in enhancing functional status among older adults. No SAEs occurred during or after cell infusion, and all observed AEs were mild and resolved spontaneously. These results align with previous research. In a randomised, double-blind, placebo-controlled trial by Tompkins et al. (2017),¹⁷ no SAEs were reported, and the incidence of mild AEs was similarly low. Similarly, a 2024 trial by Zhu et al.¹⁸ found comparably low rates of transient AEs, reinforcing the safety profile of MSC therapy.

Participants in this trial presented with established multimorbidity, predominantly hypertension and metabolic disorders, which drive frailty via endothelial dysfunction and chronic inflammaging. Despite this high-risk profile, UC-MSC therapy demonstrated efficacy, with statistically significant reductions in systemic IL-6 and TNF-α levels compared with controls. No exacerbation of blood pressure or cardiovascular events occurred, providing evidence supporting the safety of UC-MSC administration in geriatric populations.

Our results provide strong evidence that UC-MSC therapy yields statistically significant and sustained improvements across multiple domains of physical function. Total SPPB scores were statistically significantly higher in the UC-MSC group than in the control group starting at one month post-intervention, a difference that persisted through nine months. Gait speed was also statistically significantly faster in the UC-MSC group, with between-group differences observed from three months onwards.

The magnitude of these between-group differences exceeded established thresholds for clinical relevance. The mean difference in SPPB scores between groups was 1.1 points at 3 months (exceeding the 1.0-point threshold for substantial change), and the difference in gait speed was 0.16 m/s (surpassing the 0.10 m/s benchmark).²⁹ These findings confirm that the functional benefits of UC-MSC therapy are clinically meaningful.

Handgrip strength was consistently greater in the UC-MSC group than in the control group, reinforcing the positive effect on upper body strength. Similarly, scores for self-reported physical activity (CHAMPS) were statistically significantly higher in the intervention group, indicating enhanced engagement in daily living.

General health status (SF-36) scores were statistically significantly higher in the UC-MSC group, with the largest between-group difference observed at three months. Fatigue scores were consistently lower in the UC-MSC group than in controls, indicating durable alleviation of fatigue. In contrast, statistically significant between-group differences in knee function (WOMAC) were observed only at nine months. This delayed separation implies that regenerative processes in joint tissues may require a longer timescale to manifest clinically compared to systemic functional effects.

Despite the statistically significant between-group differences in physical function, no differences were noted in FEV₁/FVC or LVEF between the two groups. Subgroup analysis based on the CFS revealed distinct response patterns. Participants with mild-to-moderate frailty (CFS 3–4) demonstrated consistent functional benefits relative to controls, whereas those with advanced frailty (CFS 5) exhibited variable responses, primarily in joint function (WOMAC) and LVEF parameters. This heterogeneity suggests a potential “therapeutic window” where MSC intervention is most efficacious before physiological reserves are critically depleted. However, given the small sample size of the severe subgroup, these findings warrant cautious interpretation.

Our findings align with previous trials by Tompkins et al.¹⁷ and Zhu et al.,¹⁸ which reported superior physical performance and favourable immunological profiles following MSC infusion. Consistent with these studies, we observed robust treatment effects in mild-to-moderate frailty.

With respect to cell quality, the consistent manufacturing performance observed across all UC-MSC batches supports the robustness and reproducibility of the production process. The cell viability release criterion of ≥70% applied in this study aligns with established MSC manufacturing standards and multiple prior MSC clinical trials.^16,38, 39, 40, 41 Importantly, all UC-MSC products administered substantially exceeded this minimum threshold, with post-thaw viabilities of approximately 90%, indicating high cell quality at the time of infusion. Collectively, these findings suggest that the observed biological and clinical outcomes are unlikely to be confounded by cell viability variability and instead reflect treatment-related effects.

An interesting dichotomy exists between our findings and the recent study by Ruiz and colleagues,⁴² who reported improvements in the 6-min walk test (6 MW T) distance but not SPPB. In our trial, we used the 4-m walk test to assess mobility rather than the 6 MW T. Our baseline 4-m walk times were 7.6 ± 2.3 s in the UC-MSC group and 7.5 ± 2.6 s in the control group. The differing SPPB results between these two studies might stem from differences in cell source (UC-MSC vs. BM-MSC) or in dosing protocols.

In the present study, we analysed plasma levels of IL-6, TNF-α, IL-8, IL-10, MCP-1, and IL-1β, cytokines and chemokines that have been extensively investigated in relation to frailty.⁴³ We observed that the UC-MSC-treated group exhibited statistically significant reductions in plasma levels of IL-6, TNF-α, and IL-10 at both one-month and nine-month follow-up, as well as a statistically significant decrease in IL-1β at one month. In contrast, no statistically significant differences between groups were observed for IL-8 or MCP-1 levels. The reduction of IL-6, TNF-α, IL-10, and IL-1β suggests the anti-inflammatory and immunomodulatory effects of MSC therapy. Although IL-10 is classically anti-inflammatory, it can also reflect a compensatory regulatory response to heightened immune activation. Thus, reduced IL-10 in this context may indicate normalisation of immune homoeostasis rather than loss of anti-inflammatory control. Studies have shown that MSCs exert their immunomodulatory properties to decrease inflammatory markers.⁴⁴ However, our study did not demonstrate a statistically significant difference in IL-8 and MCP-1 levels between the UC-MSC-treated and control groups. This observation may be attributed to the context-dependent immunomodulatory mechanisms of UC-MSCs, which likely operate through specific signalling pathways that do not directly influence IL-8 and MCP-1 expression under the experimental conditions employed.

We observed that baseline TNF-α levels were notably high (median 4.1 pg/mL in UC-MSC vs. 3.32 pg/mL in control) (Supplementary Table S3). This likely reflects the substantial burden of age-related comorbidities in our study population, such as hypertension (reaching 73.7% in some groups) and metabolic disorders, which are known drivers of chronic inflammaging and elevated baseline inflammatory markers.

Our findings were supported by the study of Golpanian et al., 2017 and Tompkin et al., 2017 and Zhu et al. (2024), where they reported a statistically significant decrease in TNF-α in treated groups.^16,17 Additionally, a study by Zhu et al. (2024) found that no statistically significant changes were observed in IL-8 levels.¹⁸ The results of this study suggested that MSCs have immunomodulatory and anti-inflammatory effects related to ageing. Nonetheless, the molecular and cellular pathways through which mesenchymal stem cells mediate their immunomodulatory functions remain incompletely characterised, necessitating further in-depth mechanistic studies.

p16^INK4a mRNA expression has been highlighted in various studies as a biomarker of cellular senescence and biological ageing.^45,46 Increased p16^INK4a mRNA expression in peripheral blood T cells and adipose tissue has been associated with key components of the frailty syndrome, including reduced grip strength, slower gait speed, and impaired mobility.^47,48 Given that frailty reflects multisystem physiological decline, we selected peripheral blood T cells as a biologically relevant compartment in which senescence markers may capture organism-level ageing. p16^INK4a mRNA expression in T cells has been shown to increase robustly with age, associate with functional measures central to frailty, and respond to systemic stressors, supporting its role as a clinically accessible marker of systemic ageing relevant to frailty.^45,47, 48, 49 We compared the relative expression levels of p16INK4a mRNA in patients' T cells at baseline and at 3, 6, and 9 months post-intervention in both groups. The UC-MSC group showed a statistically significant reduction in p16INK4a mRNA expression compared with the control group at 3 months post-infusion. The biological mechanisms linking reductions in p16^INK4a to clinical improvement remain incompletely elucidated. However, decreased p16^INK4a is proposed to reflect a reduced burden of cellular senescence and attenuation of senescence-associated dysfunction.^50,51 In experimental models, selective elimination of p16^INK4a -positive senescent cells results in improvements in physical function and physiological resilience, supporting a potential mechanistic relationship between reduced p16^INK4a expression and functional recovery.^52,53 In this study, p16^INK4a expression was not sustained at later time points, which may reflect the fact that p16^INK4a is widely regarded as a marker of cumulative chronological ageing rather than a dynamic indicator of short-term biological change. Consequently, given the relatively short nine-month follow-up period and the limited sample size, the observed changes in p16^INK4a did not reach statistical significance.

Previous studies have shown that diminished mitochondrial function is associated with frailty.54, 55, 56 Interestingly, emerging evidence suggests that MSCs can transfer healthy mitochondria to injured cells, thereby restoring mitochondrial respiratory function, enhancing ATP production, and reducing oxidative stress.57, 58, 59, 60 In this study, we evaluated mitochondrial function in patients following MSC infusion. At six months post-infusion, statistically significant differences were observed in mitochondrial maximal respiration and spare respiratory capacity, two indicators of cellular metabolic health and energy production flexibility. Both parameters were notably higher in the UC-MSC group, suggesting that MSC therapy may enhance mitochondrial function. Studies have demonstrated that exercise-induced improvements in mitochondrial function are associated with enhanced muscle strength, faster gait speed, and improved physical performance in older adults.^61,62 In addition, improved mitochondrial function has been proposed to attenuate chronic inflammation and cellular senescence, biological processes implicated in the pathophysiology of frailty.⁶³

Patients received UC-MSC infusions at baseline and at three months. Two groups showed no statistically significant differences in mitochondrial function at three months, suggesting that MSC effects may depend on dose or exposure time. Although a difference was observed at six months, it did not persist at nine months, indicating short-lived therapeutic benefits that require further study for durability and clinical relevance.

Subgroup analyses demonstrated heterogeneity in biological responses across frailty severity, suggesting that baseline clinical status influences responsiveness to UC-MSC therapy. Reductions in p16^INK4a mRNA expression and modest improvements in mitochondrial respiration were most evident in the CFS 3 subgroup, indicating that senescence and bioenergetic-related pathways may be more responsive in individuals with mild-to-moderate frailty, where physiological reserve is relatively preserved. In contrast, these changes were attenuated or inconsistent in more advanced frailty (CFS 4–5), consistent with greater biological impairment and variability. Conversely, inflammatory markers showed the largest early reductions in the most frail subgroup (CFS 5), reflecting higher baseline inflammatory burden and potentially greater short-term immunomodulatory benefit. Together, these findings suggest that UC-MSC therapy may confer distinct biological and potentially clinical benefits depending on frailty stage, with early frailty showing greater responsiveness in senescence and mitochondrial pathways and advanced frailty exhibiting more prominent inflammatory modulation.

Although UC-MSC therapy produced early improvements in inflammatory markers, p16^INK4a mRNA expression, and mitochondrial function, several effects were not sustained at nine months. This pattern is consistent with the transient, paracrine mechanisms of MSC action and suggests that repeated or maintenance dosing may be required in a chronic condition such as frailty. While direct evidence in optimal maintenance dosing for frailty is limited, prior studies support repeated dosing to prolong biological and clinical effects.^16,17 In addition, combination strategies pairing MSC therapy with interventions targeting complementary pathways, such as exercise or metabolic support, may further enhance durability and translation into sustained clinical benefits.

A limitation of this study is the single-centre design, which limits generalisability, and the open-label design, which poses a risk of detection bias, particularly regarding patient-reported outcomes where participant expectations may influence subjective scores. The sample size, while adequate for the primary endpoint, limits the detection of rare adverse events. However, a placebo-controlled sham infusion was ethically precluded due to the risks associated with invasive intravenous access, thrombophlebitis, and potential systemic inflammatory responses in this frail population. We minimised this bias through strictly blinded outcome assessment for physical endpoints. Furthermore, the observed efficacy is substantiated by objective improvements in inflammatory and senescence biomarkers (IL-6, TNF-α, p16INK4a), which provide independent corroboration, immune to placebo effects, notwithstanding the weak linear correlations between biomarker changes and functional gains. It is important to note that all subgroup analyses and exploratory biomarker assessments are hypothesis-generating in nature and require confirmation in larger, adequately powered trials. Finally, while all patients were diagnosed with frailty based on the Fried phenotype criteria, we enrolled a large number of patients who were subsequently categorised as CFS grade 3 (well managed with treated comorbidities). This indicates that, on the Rockwood clinical spectrum, a substantial portion of our cohort was not strictly in advanced stages of frailty. Nevertheless, their positive functional response is highly encouraging and highlights the potential of UC-MSCs for early intervention and frailty prevention.

Conclusion

Intravenous infusion of allogeneic UC-MSCs showed no safety concerns and was well tolerated in older adults with frailty. Compared to standard care, UC-MSC therapy was associated with statistically significant and sustained changes in physical function, muscle strength, activity, fatigue, knee pain, and quality of life. The treatment correlated with favourable changes in inflammatory, senescence, and mitochondrial biomarkers. These findings provide the necessary rationale and foundation for larger, double-blind, multicentre phase 3 trials to confirm the definitive efficacy of UC-MSC infusion for frailty.

Contributors

LTN, KTN, LTMD, TTKH, QMN, and VTH conceptualised and designed the study. LTN, KTN, TTNN, VTH, NTHD, TTKP, HTN, and ATPN analysed the data. LTN, KTN, and LTMD accessed and verified the underlying data. LTN, KTN, LTMD, and VTH wrote the manuscript. All the authors critically revised the manuscript and approved the final version of the manuscript. LTN, KTN, and LTMD agreed to be accountable for all aspects of the work and ensured that questions regarding the accuracy or integrity of any part of the work were appropriately investigated and resolved.

Data sharing statement

The data supporting the findings of this study are available from the corresponding author (liem.nt@vinuni.edu.vn) upon reasonable request.

Declaration of interests

All authors have completed the ICMJE uniform disclosure form and declare no support from any organisation for this work, other than the funding information provided. They have no financial relationships or other conflicts that could have influenced this manuscript. No competing interests exist.

Appendix A. Supplementary data

Supplementary Fig. S1.

Supplementary Fig. S2.

Supplementary Fig. S3.

Supplementary Fig. S4.

Supplementary Fig. S5.

Supplementary Fig. S6.

Supplementary Fig. S7.

Supplementary Fig. S8.

Supplementary Fig. S9.

Supplementary Fig. S10.

Supplementary Fig. S11.

Supplementary Fig. S12.