Stem-Cell Therapies for Glycemic Control in Young Adults with Type 1 Diabetes: Systematic Review, Random-Effects Meta-Analysis and Exploratory Network Meta-Analysis

 

Pedro Gutierrez-Castrellon^1,5, Paul de Vos², Carolina Bluguermann³, Krista Casazza⁴, Ian Jenkins⁵, Waldemar Lernhardt⁵, Bradley Robinson⁶, Adrian Mutto³ and Jonathan RT Lakey^1,3,4,5*

Citation: Gutierrez-Castrellon P, de Vos P, Bluguermann C, Casazza K, Jenkins I, et al. Stem-Cell Therapies for Glycemic Control in Young Adults with Type 1 Diabetes: Systematic Review, Random-Effects Meta-Analysis and Exploratory Network Meta-Analysis. J Stem Cell Res. 8(1):1-26.

Received: October 24, 2026 | Published: January 24, 2027

Copyright^© 2027 by Gutierrez-Castrellon P, et al. All rights reserved. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

DOI: https://doi.org/10.52793/JSCR.2027.8(1)-109

Abstract

Background: Type 1 diabetes (T1D) results from autoimmune destruction of pancreatic β cells and mandates lifelong insulin therapy. Stem-cell interventions are proposed both to replace β-cell mass and to modulate autoimmunity, but the comparative effectiveness of the different cell classes is unresolved. Since the first version of this synthesis, a systematic review restricted to randomized controlled trials (RCTs) of stem-cell therapy in diabetes mellitus has been published, identifying T1D trials not previously incorporated and reporting, for the first time, the critical patient-important outcomes of quality of life and hypoglycemic episodes.

Objectives: To update the efficacy (HbA1c, C-peptide, insulin requirement) and safety estimates for stem-cell classes in young adults (>18 and <40 y) with T1D by incorporating the newly identified randomized evidence; to re-estimate the exploratory network meta-analysis (NMA) and SUCRA ranking; and to determine whether the previously reported treatment hierarchy is robust to the addition of randomized-only data.

Methods: MEDLINE, Embase, Web of Science, CENTRAL, WanFang and CNKI were searched to August 2026 and cross-referenced against the most complete T1D syntheses, including the randomized-only review of Manikandan et al. (2026). Randomized/controlled trials of a defined stem-cell class versus any control in T1D were eligible. Continuous outcomes were pooled as standardized mean differences (SMD, Hedges’ g; positive = favors cell therapy); dichotomous outcomes as risk ratios. Aggregate mean differences reported in the updating source were converted to SMDs by z-statistic invariance and combined with the previous node estimates by inverse-variance evidence accrual. A contrast-based random-effects NMA anchored on standard insulin therapy ranked treatments by SUCRA and probability-best, with rankograms. The ranking model was verified to reproduce the previously published SUCRA values to within 0.2 percentage points before any update was applied. Certainty was rated with GRADE.

Results: The updated synthesis comprises 32 T1D studies and 538 participants (previously 29 studies, 487 participants) across five nodes — MSC, HSC, MSC+HSC, UCB and BM-MNC — versus standard insulin in a star-shaped network. Three newly identified T1D RCTs contributed 51 participants; a fourth, which enrolled a mixed type 1 and type 2 population, was restricted to a pre-specified sensitivity analysis. At 12 months the top of the hierarchy was unchanged: HSC ranked highest for HbA1c (SUCRA 99.2%; SMD +2.42 (+1.46, +3.38)), C-peptide (94.0%) and insulin requirement (91.4%), with MSC+HSC second for C-peptide and insulin requirement and third for HbA1c. The only rank change was a swap of mesenchymal stem cells alone with MSC+HSC for HbA1c, MSC falling from third to fourth (SMD +0.74 (+0.20, +1.27), previously +1.08, +0.36 to +1.80). In the sensitivity analysis admitting the mixed-population trial, HSC fell to 97.8% for HbA1c and MSC+HSC overtook it for insulin sparing (94.3% versus 85.5%). In the overall pairwise synthesis, every efficacy estimate fell: HbA1c +1.30 (95% CI +0.85 to +1.74), C-peptide +0.58 (+0.01 to +1.15) and insulin requirement +0.50 (−0.02 to +1.03, non-significant). In the randomized-only evidence base, no critical patient-important outcome improved: quality of life MD +3.15 (95% CI −0.80 to +7.10), hypoglycemic episodes RR 0.90 (0.56 to 1.45), and only 3 of 37 intervention-arm participants (8%, in three trials) achieved an insulin-free period. Benefits were transient (mean ~15.6 months). Procedure and immunosuppression harms clustered in the HSC pathway (severe complications 3.9%; mortality 3.4%). One transitional meningioma arising eight months after allogeneic fetal-cell infusion was adjudicated as related to the intervention, revising the previous statement that no stem-cell-attributable neoplasm had been reported. Allogeneic stem-cell-derived islets (zimislecel/VX-880) achieved insulin independence in 10 of 12 recipients at one year in a single-arm study but require chronic immunosuppression.

Conclusions: HSC-based therapies show the strongest short-term efficacy signal for glycemic control and β-cell preservation in T1D, and this ranking proved robust to the addition of the newly identified randomized evidence, but their transient benefit and treatment-related mortality currently limit their clinical attractiveness. Thus, the highest-ranked therapy is not necessarily the best therapeutic option for patients — a point the updated evidence sharpens rather than softens, since HSC and the HSC-containing combination regimen are now separated by fewer than three SUCRA percentage points for insulin sparing, and reverse order altogether in the sensitivity analysis, while differing by roughly 3.4 percentage points in treatment-related mortality. The randomized-only evidence base is more sobering still: no critical patient-important outcome improved, only 3 of 37 treated participants (8%) achieved a transient insulin-free period, and a first neoplasm adjudicated as related to a stem-cell infusion has now been reported, withdrawing the previous statement that none had occurred. Stem-cell-derived islets offer a fundamentally different and potentially more durable route by replacing lost β-cell mass, but current evidence remains uncontrolled and their dependence on immunosuppression limits broader clinical application; the contrast between 83% insulin independence in a single-arm replacement study and 8% transient insulin independence in randomized immunomodulation trials is the central quantitative finding of this update. The future of cell therapy in T1D may therefore lie in combining effective β-cell replacement with durable protection from recurrent autoimmunity, while minimizing or eliminating systemic immunosuppression. These findings provide a framework for prioritizing future head-to-head trials rather than supporting routine clinical use of any current stem-cell strategy.

Keywords

Type 1 diabetes; stem-cell therapy; mesenchymal stem cells; hematopoietic stem cells; C-peptide; HbA1c; network meta-analysis; Surface Under the Cumulative Ranking Curve (SUCRA); evidence updating; randomized controlled trials.

Introduction

Can stem cells cure or durably control type 1 diabetes, and are they safe? Type 1 diabetes (T1D) destroys the body’s own insulin-making cells, forcing lifelong insulin injections. Several cell-based treatments — from bone marrow, blood-forming stem cells, fat or umbilical tissue, and, most recently, lab-grown clusters of insulin cells — are proposed either to calm the immune attack or to replace the lost cells. We gathered the controlled trials that tested these cells against standard insulin care and ranked them. This is a comparative treatment-ranking analysis specifically addressing which cell-based strategy provides the most favorable efficacy signal in T1D, and demonstrating why efficacy ranking alone is insufficient when durability and treatment-related harm are incorporated. This version updates that ranking with randomized trials identified after the first analysis was completed.

What we found: across the trials with usable numbers, blood-forming (hematopoietic) stem cells — alone or combined with mesenchymal cells — still gave the largest one-year improvements in HbA1c, C-peptide and insulin needs, with the combination regimen close behind for C-peptide and insulin needs though not for HbA1c. Mesenchymal cells used on their own dropped a place. Umbilical-cord blood still did not help. But adding the randomized trials made the benefits smaller, and when only randomized trials are considered, the improvements in blood sugar control, insulin dose and C-peptide are no longer statistically distinguishable from standard care. The outcomes patients care about most — quality of life and the number of low-blood-sugar episodes — did not improve at all. Fewer than one in ten treated patients came off insulin, and only temporarily (about 15 months on average). The hematopoietic route still carried a small but real risk of death from its conditioning regimen, and a brain tumor has now been reported after infusion of fetal stem cells. The newest lab-grown islet products can free most recipients from insulin, but they are single-arm studies that still require immune-suppressing drugs. Overall certainty is very low; these treatments should be considered promising but unproven.

Background

T1D is a chronic autoimmune disease in which T-cell-mediated destruction of insulin-producing β cells leads to dependence on exogenous insulin. Incidence is rising by ~3% annually and more than eight million people are affected worldwide. Despite continuous glucose monitoring and automated insulin delivery, a large proportion of patients do not reach the American Diabetes Association/European Association for the Study of Diabetes targets (HbA1c <7% with >70% time-in-range) and remain exposed to severe hypoglycemia and to the microvascular and macrovascular complications of chronic hyperglycemia.

Two rationales motivate cellular therapy. The first is β-cell replacement: pluripotent (embryonic, ESC; induced, iPSC) or chemically induced pluripotent stem cells can be differentiated into glucose-responsive insulin-secreting cells, providing a potentially inexhaustible substitute for the scarce deceased-donor islets used in the Edmonton protocol. The second is immunomodulation and β-cell preservation: mesenchymal stem/stromal cells (MSC) and hematopoietic stem cells (HSC) can dampen the autoimmune attack, protect residual β-cell mass and foster a regenerative microenvironment — largely through paracrine and immunoregulatory mechanisms rather than durable engraftment.

The therapeutic landscape is populated by biologically distinct classes — MSC (bone-marrow, umbilical-cord/Wharton’s jelly, adipose), HSC (usually autologous non-myeloablative transplantation), combinations of MSC with HSC or bone-marrow mononuclear cells (BM-MNC), umbilical-cord blood (UCB), and ESC-/iPSC-derived islets — that differ profoundly in mechanism, invasiveness, immunosuppression requirement, durability and risk. Pairwise meta-analyses suggest benefit for glycemic control and C-peptide but cannot establish which class is preferable, because head-to-head trials are essentially absent. Network meta-analysis (NMA) allows the simultaneous comparison and ranking of multiple interventions and is the appropriate tool for this question.

Why this update was necessary. The first version of this synthesis drew on all controlled evidence, including single-arm before-after comparisons, which is the only way to populate a network of this size, but which systematically inflates effect estimates. In February 2026, Manikandan and colleagues published a Cochrane-aligned systematic review restricted to randomized controlled trials, commissioned to support a national clinical guideline. That review screened 9 248 de-duplicated records and included 20 randomized trials. It identified three type 1 diabetes randomized trials not incorporated in our original network, together with three reports in mixed type 1 and type 2 populations, of which one (the allogeneic fetal liver-derived cell program) reported outcomes eligible for our network and is reserved here for sensitivity analysis. It also added a 96-month follow-up of an already-included trial, and — uniquely — reported the critical patient-important outcomes of quality of life, hypoglycemic episodes and insulin-free period that our protocol had pre-specified but could not populate. It also documented a neoplastic event attributed to a stem-cell intervention. Any one of these would justify an update; together they oblige one.

Objectives

Primary objective: to determine the efficacy (HbA1c over time, C-peptide, insulin) and safety of stem-cell classes versus standard insulin therapy/control in young adults with T1D, incorporating all randomized evidence available to August 2026. Secondary objectives: to rank cell classes (MSC, HSC, MSC+HSC, UCB, BM-MNC) by SUCRA through an exploratory NMA; to quantify how much of the previously reported hierarchy survives the addition of randomized-only data; to report the critical patient-important outcomes of quality of life, hypoglycemic episodes and insulin-free period; and to appraise outcomes of regulatory interest (autoimmunity, carcinogenicity, immunosuppression harms, durability) and to contextualize stem-cell-derived islet products narratively.

Methods

The review was conducted per the Cochrane Handbook for Systematic Reviews of Interventions and reported per PRISMA 2020 and, for the network component, PRISMA-NMA. The protocol elements (PICOS, outcomes, analysis) were defined a priori and cross-checked against the most complete T1D syntheses to ensure the study pool was complete. This version is an evidence update of a previously completed analysis; the updating procedure is pre-specified in section 4.5 and every previously reported estimate is retained alongside its updated value so that the effect of the new data is auditable.

Eligibility criteria (PICOS)

Element	Criterion
Population	People with T1D; pre-specified target = young adults >18 and <40 y. Because age-disaggregated data are scarce and the extractable T1D evidence clusters in adolescents/young adults (pooled mean ages 3.0–40.3 y), overlapping populations were retained and the age overlap handled as heterogeneity/indirectness.
Intervention	Any defined stem-cell class: MSC (bone-marrow, umbilical-cord/Wharton’s jelly, adipose), HSC (autologous non-myeloablative or allogeneic fetal-derived), MSC+HSC, BM-MNC, UCB, and ESC-/iPSC-derived islets.
Comparators	Standard insulin therapy, placebo/sham, or an alternative active cell therapy.
Outcomes	Critical: HbA1c; C-peptide (fasting/stimulated, AUC); daily insulin requirement; quality of life; hypoglycemic episodes; serious adverse events. Important/exploratory: transient insulin independence; fasting blood glucose; autoimmunity (GAD antibodies); carcinogenicity; treatment-related mortality.
Study design	RCTs and controlled clinical studies for synthesis; single-arm phase 1–2 trials of stem-cell-derived islets for narrative benchmarking only.
Timepoints	6, 12 and 24 months where reported; 12 months is the primary analysis timepoint.

 

Two limitations of this procedure are stated explicitly. First, because per-trial T1D data were not available, the new contribution assigned to each node is the aggregate randomized estimate rather than that node’s own randomized estimate; the update therefore shrinks every affected node toward the common randomized effect and cannot detect class-specific divergence within the new data. Second, the only newly identified trial that would contribute to the HSC node used allogeneic fetal liver-derived cells, a product biologically distinct from the autologous non-myeloablative transplantation that dominates that node, and it enrolled a mixed type 1 and type 2 population. It is therefore excluded from the primary analysis and enters a pre-specified sensitivity analysis reported in section 5.7, in which the consequences of admitting it are quantified in full. 

Results

Study selection

The flow of studies is shown in (Figure 1). After de-duplication and screening, 32 T1D studies enrolling 538 participants formed the primary evidence base — an increase of three studies and 51 participants over the previous version. Three newly identified randomized trials met the T1D-only PICOS and were incorporated (Carlsson PO 2023; Izadi 2022; Haller 2013). A fourth, the allogeneic fetal liver-derived cell program (Ghodsi 2012 with its three-year extension by Nasli-Esfahani 2015; 56 participants), enrolled a mixed type 1 and type 2 diabetes population from which the T1D subgroup could not be extracted; it is therefore excluded from the primary analysis and reported in the sensitivity analysis (section 5.7), which comprises 33 studies and 594 participants. Two further mixed-population reports in the updating source were screened and excluded — Huang P 2005, which reported fasting blood glucose only and no outcome eligible for the network, and the three-year extension of the fetal-cell program, already accounted for above. Its safety data are used in full (section 5.9). A fifth new report, Wu Z 2022, was identified as the 96-month follow-up of the already-included Cai J 2016 cohort and therefore contributes durability and safety data but no additional participants. Contemporary single-arm reports of stem-cell-derived islets were carried forward for narrative benchmarking only. As in the previous version, the core evidence is supplemented by two T1D randomized comparisons contributing MSC-versus-control data, which are represented within the MSC node rather than counted separately.

Figure 1: PRISMA-NMA study-selection flow for the updated review.

Characteristics of included studies

Studies were conducted predominantly in China, with others in Brazil, the USA, India, Sweden, Iran, Germany, Spain, Poland, Egypt and Kazakhstan. Mean ages ranged from 3.0 to 40.3 years, disease duration from <2 months to 11 years, and follow-up from 3 to 96 months — the upper bound extended from 60 to 96 months by the Wu Z 2022 follow-up. Five intervention classes were represented and are summarized as network nodes in (Table 2); the newly identified randomized evidence, and its disposition between the primary and sensitivity analyses, is detailed in (Table 3), and key controlled trials and landmark reports in (Table 6). HSC was typically autologous and non-myeloablative; delivery was intravenous or intra-pancreatic/intra-arterial.

Node (cell class)	Studies (n) previous → updated	Participants (n) previous → updated	Typical source	Route	Immunosuppression
MSC	5 → 7	45 → 81	BM, Wharton’s jelly/UC, adipose	IV / intra-pancreatic	Usually none
HSC	13 → 13	275 → 275	Autologous PB/BM	IV (non-myeloablative)	Conditioning regimen
MSC + HSC	3 → 3	52 → 52	Combined autologous/allogeneic	IV / intra-pancreatic	Variable
UCB	6 → 7	83 → 98	Umbilical-cord blood	IV	None
BM-MNC	2 → 2	32 → 32	Bone-marrow mononuclear cells	Intra-pancreatic / IV	None
Total (primary)	29 → 32	487 → 538	—	—	—
Sensitivity analysis: HSC + mixed-population fetal-cell trial	→ 33	→ 594	Allogeneic fetal liver-derived	IV	None reported
Reference	—	—	Standard insulin therapy / control	—	—

Figure 3: Critical patient-important outcomes and randomized-only efficacy estimates in type 1 diabetes at 12 months.

Network meta-analysis and treatment ranking

The updated star-shaped network is shown in (Figure 4); node-specific relative effects, with the previous estimate displayed alongside each updated estimate, in (Figures 5–7); SUCRA across outcomes in (Figure 8); and the HbA1c rankogram in (Figure 9). Ranking statistics are given in (Table 4).

Figure 4: Updated evidence network. Node area ∝ participants, edge width and label ∝ contributing studies. Star topology (no closed loops).

• HbA1c (12 months). HSC remained the highest-ranked class on unchanged data (SUCRA 99.2%; probability-best 96.6%; SMD +2.42 (+1.46, +3.38)). BM-MNC ranked second (71.7%), followed by MSC+HSC (59.5%) and MSC (47.7%) — a swap of the third and fourth positions relative to the previous analysis and the only rank change in the primary analysis. It is driven entirely by the dilution of the MSC node by two new randomized trials, which moved its estimate from +1.08 (+0.36, +1.80) to +0.74 (+0.20, +1.27), a fall of 0.34 SMD units. UCB remained little better than standard insulin (21.0%; SMD +0.26 (−0.03, +0.54), marginally higher than its previous +0.25) and standard insulin ranked last (0.9%).

Figure 5: HbA1c reduction at 12 months — stem-cell class versus standard insulin, previous and updated estimates (positive favors cell therapy).

• C-peptide (12 months): The order was unchanged. HSC again ranked best (SUCRA 94.0%; probability-best 73.1%; SMD +1.64 (+0.56, +2.71)), with MSC+HSC a close second (85.2%; SMD +1.28 (+0.85, +1.71)); the gap between them was unchanged at 8.8 SUCRA percentage points, since neither node received new data. MSC remained fourth, its point estimate essentially unchanged (+0.19 (−0.53, +0.91)) but considerably more precise. UCB remained below standard insulin (7.5%; SMD −0.61 (−1.80, +0.58)), consistent with the failure of autologous cord-blood infusion to preserve C-peptide in the newly added randomized trial as in the same group’s earlier report.

Figure 6: C-peptide preservation at 12 months — stem-cell class versus standard insulin, previous and updated estimates.

• Daily insulin requirement (12 months). The order was unchanged, but the margin at the top is narrow. HSC remained highest-ranked (91.4%; probability-best 57.3%; SMD +1.51 (+1.02, +2.00)) with MSC+HSC (88.5%; SMD +1.45 (+1.01, +1.89)) separated from it by only 2.9 SUCRA percentage points — a margin that reverses when the mixed-population fetal-cell trial is admitted (section 5.7). MSC fell from +0.50 (−0.18, +1.18) to +0.30 (−0.20, +0.81). UCB remained worst, below standard insulin, having been associated with an increase in insulin requirement (SMD −1.65 (−3.21, −0.10); SUCRA 0.8%).

 

Figure 7: Daily insulin reduction at 12 months — stem-cell class versus standard insulin, previous and updated estimates.

Figure 8: SUCRA by outcome for each treatment in the updated network (higher = higher rank).

Figure 9: Rankogram for HbA1c reduction in the updated network (rank 1 = best), showing the full rank-probability distribution.

Outcome	Treatment	Updated SMD (95% CI) vs standard insulin	Previous SMD	SUCRA (%) updated	SUCRA (%) previous	Prob. best (%)
HbA1c	HSC	+2.42 (+1.46, +3.38)	+2.42	99.2	98.9	96.6
HbA1c	BM-MNC	+1.24 (+0.44, +2.04)	+1.24	71.7	67.0	3.1
HbA1c	MSC+HSC	+0.95 (+0.38, +1.52)	+0.95	59.5	53.2	0.3
HbA1c	MSC	+0.74 (+0.20, +1.27)	+1.08	47.7	59.8	0.1
HbA1c	UCB	+0.26 (−0.03, +0.54)	+0.25	21.0	19.8	0.0
HbA1c	Standard insulin	reference	reference	0.9	1.2	0.0
C-peptide	HSC	+1.64 (+0.56, +2.71)	+1.64	94.0	93.7	73.1
C-peptide	MSC+HSC	+1.28 (+0.85, +1.71)	+1.28	85.2	84.9	26.8
C-peptide	BM-MNC	+0.31 (−0.20, +0.81)	+0.31	48.5	48.2	0.0
C-peptide	MSC	+0.19 (−0.53, +0.91)	+0.20	39.7	40.0	0.1
C-peptide	Standard insulin	reference	reference	25.2	26.6	0.0
C-peptide	UCB	−0.61 (−1.80, +0.58)	−0.75	7.5	6.6	0.0
Insulin requirement	HSC	+1.51 (+1.02, +2.00)	+1.51	91.4	91.2	57.3
Insulin requirement	MSC+HSC	+1.45 (+1.01, +1.89)	+1.45	88.5	88.4	42.7
Insulin requirement	MSC	+0.30 (−0.20, +0.81)	+0.50	48.4	53.0	0.0
Insulin requirement	BM-MNC	+0.26 (−0.25, +0.77)	+0.26	45.7	42.5	0.0
Insulin requirement	Standard insulin	reference	reference	25.2	24.4	0.0
Insulin requirement	UCB	−1.65 (−3.21, −0.10)	−1.96	0.8	0.5	0.0

Outcome (12 mo)	Treatment	Primary SMD (95% CI)	Primary SUCRA (%)	Sensitivity SMD (95% CI)	Sensitivity SUCRA (%)
HbA1c	HSC	+2.42 (+1.46, +3.38)	99.2	+2.06 (+1.19, +2.94)	97.8
HbA1c	BM-MNC	+1.24 (+0.44, +2.04)	71.7	+1.24 (+0.44, +2.04)	72.8
HbA1c	MSC+HSC	+0.95 (+0.38, +1.52)	59.5	+0.95 (+0.38, +1.52)	59.8
HbA1c	MSC	+0.74 (+0.20, +1.27)	47.7	+0.74 (+0.20, +1.27)	47.8
HbA1c	UCB	+0.26 (−0.03, +0.54)	21.0	+0.26 (−0.03, +0.54)	21.0
HbA1c	Standard insulin	reference	0.9	reference	0.9
C-peptide	HSC	+1.64 (+0.56, +2.71)	94.0	+1.39 (+0.41, +2.37)	90.4
C-peptide	MSC+HSC	+1.28 (+0.85, +1.71)	85.2	+1.28 (+0.85, +1.71)	88.1
C-peptide	BM-MNC	+0.31 (−0.20, +0.81)	48.5	+0.31 (−0.20, +0.81)	48.8
C-peptide	MSC	+0.19 (−0.53, +0.91)	39.7	+0.19 (−0.53, +0.91)	39.9
C-peptide	Standard insulin	reference	25.2	reference	25.2
C-peptide	UCB	−0.61 (−1.80, +0.58)	7.5	−0.61 (−1.80, +0.58)	7.5
Insulin requirement	MSC+HSC	+1.45 (+1.01, +1.89)	88.5	+1.45 (+1.01, +1.89)	94.3
Insulin requirement	HSC	+1.51 (+1.02, +2.00)	91.4	+1.26 (+0.82, +1.71)	85.5
Insulin requirement	MSC	+0.30 (−0.20, +0.81)	48.4	+0.30 (−0.20, +0.81)	48.4
Insulin requirement	BM-MNC	+0.26 (−0.25, +0.77)	45.7	+0.26 (−0.25, +0.77)	45.7
Insulin requirement	Standard insulin	reference	25.2	reference	25.2
Insulin requirement	UCB	−1.65 (−3.21, −0.10)	0.8	−1.65 (−3.21, −0.10)	0.8

Study (year)	Design	Population	Intervention vs comparator	Key finding
Carlsson PO 2015	RCT	New-onset T1D, adults	Autologous BM-MSC vs no treatment	Preserved/increased C-peptide at 1 y; no MSC-related AEs
Carlsson PO 2023 †	Phase I/II RCT, double-blind	T1D, adults	Allogeneic Wharton’s jelly MSC vs placebo	Preserved endogenous insulin production; low risk of bias across all domains on our appraisal of the primary report
Izadi M 2022 †	Phase I/II RCT	Newly diagnosed T1D	Autologous BM-MSC vs placebo	Improved C-peptide; no QoL or hypoglycemia benefit
Cai 2016 / Wu Z 2022 †	RCT (open-label), 8-y follow-up	Established T1D	UC-MSC + autologous BM-MNC vs control	Safe; improved insulin secretion; fewer chronic complications at 96 months
Hu 2013	RCT ¶	New-onset T1D	Wharton’s-jelly MSC vs insulin	Improved HbA1c and C-peptide vs control
Gu B 2018	RCT	T1D	Autologous HSC vs insulin	Improved metabolic control; transient insulin independence
Ghodsi 2012 / Nasli-Esfahani 2015 †§	RCT with 3-y extension	Diabetes (mixed T1D and T2D)	Allogeneic fetal liver-derived cells vs control	No durable metabolic advantage; transitional meningioma at 8 months adjudicated related
Haller MJ 2013 †	RCT	T1D, children	Autologous UCB + DHA/vitamin D vs control	Failed to preserve C-peptide
Zimislecel/VX-880 2025 *	Single-arm 1–2	T1D, mean age 42.7 y	Allogeneic SC-derived islets + immunosuppression	12/12 HbA1c <7% & severe-hypo-free; 10/12 insulin-independent at 1 y; 2 deaths (unrelated)
CiPSC-islets 2024 *	Single-arm	T1D, 25-y woman	Autologous chemically induced iPSC-islets	Insulin independence by day 75; HbA1c 7.6→<5.7%; no tumor
Edmonton protocol 2000/2006 ‡	Single-arm/multicentre	Brittle T1D	Deceased-donor islets (not stem cell)	Insulin independence not durable (76% back on insulin at 2 y)

 Across the controlled T1D evidence synthesized in this updated exploratory network, HSC-based regimens continued to rank highest for all three efficacy outcomes at 12 months, MSC+HSC remained second for C-peptide and insulin sparing and overtook HSC on the latter in the sensitivity analysis, BM-MNC was intermediate, MSC used alone fell to fourth for glycemic control, and UCB remained least effective — at or below standard insulin. The convergence of the ranking across three mechanistically distinct outcomes, together with the meta-regression signal that older age and higher baseline GAD-antibody positivity predicted greater HSC efficacy, remains biologically coherent with an immunomodulatory mechanism that arrests autoimmune β-cell destruction in patients with active, antibody-driven disease. What the update changes is not the direction of the hierarchy but confidence in its magnitude.

• The randomized evidence is null on what matters: The most consequential finding of this update is not a change in rank order. It is that when the evidence base is restricted to randomized trials, stem-cell therapy in T1D does not significantly improve HbA1c at 12 months, does not reduce daily insulin requirement, does not improve fasting C-peptide, does not improve quality of life and does not reduce hypoglycemic episodes. The only surviving efficacy signal is a 0.57-percentage-point HbA1c reduction at 6 months that is lost by 12. Our all-evidence network yields effect sizes an order of magnitude larger — an overall pooled HbA1c SMD of +1.30, and +2.42 for the highest-ranked class — because it necessarily incorporates before-after and open-label data; the gap between the two is a direct measure of design-related optimism, and it is large. Any reader of the ranking presented in Table 4 should hold both numbers in mind simultaneously: the ranking answers "which class, if any of them work", and the randomized evidence answers "do any of them work", to which the current answer is not yet.
• The efficacy–safety trade-off, and why the insulin-sparing ordering is not robust: The apparent superiority of the HSC pathway must be read against its harms: autologous non-myeloablative HSC transplantation carried ~3.4% treatment-related mortality and ~3.9% severe-complication rate driven by the conditioning regimen and infection, whereas MSC-based therapy was essentially as safe as control. The ranking that is "best" for glycemia is therefore not the ranking that is "best" for the patient once risk is incorporated. Two results of this update bear on that trade-off. First, HSC and MSC+HSC are separated by fewer than three SUCRA percentage points for insulin sparing in the primary analysis, and the ordering reverses entirely when the mixed-population fetal-cell trial is admitted, so the apparent superiority of HSC on this outcome is not robust. Second, the updating source’s subgroup analysis found significant HbA1c and C-peptide improvement at 6 months for mesenchymal plus mononuclear combination therapy, albeit with an effect size the authors themselves describe as trivial. Where two classes are statistically indistinguishable on benefit and differ by roughly 3.4 percentage points in treatment-related mortality, the choice between them is not a statistical one. A formal net-clinical-benefit ranking, which we did not attempt, would move MSC-containing regimens above HSC monotherapy. This tension — large but transient benefit at the cost of transplant-related mortality — recapitulates the deceased-donor islet-transplant experience.
• An observed rather than theoretical oncological signal: The withdrawal of the previous "no malignancy reported" statement is, in our view, the single most important correction in this version. One adjudicated case in a small trial of an allogeneic fetal product does not establish a class effect, and the biological distance between fetal liver-derived cell suspensions and either autologous MSC or differentiated iPSC-islets is considerable. But the field has been arguing that tumorigenicity is a theoretical hazard mitigated by purity controls and engineered safety switches. It is now an observed hazard with one case, in a literature whose median total follow-up is 12 months and which is therefore structurally incapable of detecting latency-dependent neoplasia. Mandatory long-term oncological registries should be a condition of further clinical development.
• Replacement versus immunomodulation: The included comparative trials overwhelmingly test immunomodulatory/β-cell-preserving strategies (MSC, HSC), which do not create new β-cell mass and whose effects wane as residual β cells are lost — hence the ~15.6-month mean duration of insulin independence and the 8% absolute rate at which it is achieved at all in randomized trials. The replacement paradigm — ESC-, iPSC- and chemically-induced-iPSC-derived islets — addresses the mass deficit directly and, in single-arm studies, restores physiological, self-regulated insulin secretion with insulin independence in most recipients. The update sharpens rather than softens this contrast. The field is bifurcating: durable cure will likely require replacement, while immunomodulation may best serve early, antibody-positive disease, possibly as a complication-prevention rather than a glycemic intervention — a reading now supported by the eight-year complication data from the combination cohort. Combination or sequential strategies (immune-tolerance induction plus replacement) are a logical next step.
• Comparison with prior work and regulatory/cost considerations: Prior pairwise meta-analyses concluded that stem-cell transplantation lowers HbA1c and improves C-peptide, with MSC+HSC and HSC subgroups clearest; our NMA extends this by providing a single coherent ranking with quantified uncertainty, and both agree that UCB does not preserve β-cell function. The randomized-only review used for this update disagrees with the magnitude of those conclusions and agrees with their direction, and it reports an 8% insulin-free rate against the 26.5% obtained in a synthesis admitting non-randomized designs — a threefold discrepancy that is itself a finding. Outcomes of regulatory interest — tumorigenicity, autoimmunity recurrence, immunosuppression-related infection/malignancy and durability — remain incompletely characterized; pluripotent products in particular carry an intrinsic teratoma hazard mandating rigorous purity and, increasingly, engineered safety switches and hypoimmunogenic designs. On cost, autologous MSC is inexpensive and low-risk but modest and transient; autologous iPSC-islet manufacture is highly effective but bespoke and costly; allogeneic "off-the-shelf" islets offer scalability at the recurring cost and risk of lifelong immunosuppression. A formal cost-effectiveness analysis was not feasible from available data and is a research priority — although the null randomized effect on patient-important outcomes means that, for immunomodulatory approaches, the appropriate current answer to a cost-effectiveness question is that effectiveness has not yet been established.
• Strengths and limitations: Strengths include a Cochrane-aligned protocol, RoB and GRADE appraisal, a transparent random-effects synthesis, probability-based ranking with full rank distributions, integration of efficacy with modality-specific safety and autoimmunity/carcinogenicity signals, and — new in this version — an auditable updating procedure whose first step verifies that the ranking model reproduces the previously published results before any new data are introduced, so that every change is attributable to evidence rather than to method. Limitations are substantial and temper every conclusion: (i) the network remains star-shaped — no head-to-head cell-versus-cell RCTs, no closed loops, so incoherence cannot be assessed; (ii) node inputs are pooled aggregate (largely before-after) effects rather than individual-participant data; (iii) the new randomized contribution had to be entered as an aggregate rather than per-node estimate because trial-level T1D data were reported only in supplementary material, so the update moves affected nodes toward a common effect and cannot detect class-specific divergence within the new data; (iv) the HSC node received no new randomized data in the primary analysis, because the only candidate trial enrolled a mixed type 1 and type 2 population — so the class that ranks highest is also the class this update was least able to test, and the sensitivity analysis indicates that when it is tested the estimate falls substantially; (v) most contributing studies were small, single-centre and open-label; (vi) heterogeneity was high (I² 82–95%); (vii) age-disaggregated data for the 18–40-year band were rarely available (indirectness); (viii) most estimates are at one year and do not capture later loss of effect or latent harm; and (ix) the most effective modern therapies (stem-cell-derived islets) are single-arm and could not enter the network. We also note an internal inconsistency in the updating source’s reported 6-month fasting C-peptide confidence interval (section 4.5), which we could not resolve and therefore did not use. GRADE certainty is very low and the rankings are hypothesis-generating.

 

Conclusions and Implications

In the available T1D evidence relevant to young adults — now including the randomized trials published since the first version of this synthesis — HSC-based and HSC-containing regimens rank highest for short-term glycaemic control, preservation of β-cell function and reduction in insulin requirement, whereas UCB consistently ranks lowest. The hierarchy proved robust to the update: the only rank change in any outcome was a single adjacent swap for HbA1c, in which mesenchymal stem cells used alone fell from third to fourth. These findings should, however, not be interpreted as evidence of clinical superiority. The underlying evidence is of very low certainty, benefits are generally transient, and the apparent efficacy advantage of HSC transplantation is offset by clinically relevant conditioning- and transplant-related morbidity and mortality — an advantage that, for insulin sparing, is now smaller than the uncertainty around it and reverses when the mixed-population fetal-cell trial is admitted.

The randomized-only evidence base adds a further and more fundamental caution. Restricted to randomized trials, stem-cell therapy in T1D did not significantly improve HbA1c at 12 months, did not reduce daily insulin requirement, did not improve fasting C-peptide, did not improve quality of life and did not reduce hypoglycemic episodes; only 3 of 37 treated participants (8%) achieved a transient insulin-free period, lasting about 15 months. Two consequences follow immediately. For practice, no stem-cell therapy should currently be offered to patients with type 1 diabetes outside a clinical trial, and patients presented with a "stem-cell cure" should be told these numbers alongside the 3.4% treatment-related mortality of the hematopoietic pathway. For regulation, trials must be powered for patient-important outcomes — quality of life, severe hypoglycemia, time-in-range — and not for C-peptide alone, since the surrogate has repeatedly moved without the patient-important outcome following; and the report of a neoplasm adjudicated as related to a stem-cell infusion converts tumorigenicity from a theoretical to an observed hazard, which a literature with a median total follow-up of 12 months cannot detect. Long-term oncological registries should be a condition of authorization for any cellular product in this indication.

More importantly, our findings highlight a fundamental distinction between two therapeutic strategies. Immunomodulatory approaches such as HSC and MSC therapy may preserve residual β-cell function and temporarily reduce insulin dependence but are unlikely to overcome the progressive loss of functional β-cell mass — and the updated randomized evidence indicates that, at the doses, products and follow-up durations tested so far, they do not reliably achieve even that. Stem-cell-derived islet therapies directly address this deficit and therefore represent a potentially more durable route towards restoration of endogenous insulin secretion; the contrast between 83% insulin independence in a single-arm replacement study and 8% transient insulin independence across randomized immunomodulation trials is the clearest expression of that difference in the current literature. Their future clinical value will, however, depend on whether durable graft function can be achieved without chronic immunosuppression and its associated risks. Ultimately, successful cell therapy for T1D may therefore require not simply the most effective cell product, but a strategy that combines restoration of β-cell mass with durable protection from recurrent autoimmunity.

Future studies should move beyond comparisons with standard insulin therapy alone. Adequately powered, blinded, head-to-head trials are required to directly compare cell-based strategies, stratify participants by disease duration, residual β-cell function and autoantibody status, and report standardized, time-resolved measures of HbA1c, stimulated C-peptide, insulin requirement and time-in-range. Without such trials this network will remain star-shaped and its rankings uninterpretable, and individual-participant-data pooling would remove the aggregate-input limitation that constrains both this analysis and its update. Among immunomodulatory approaches, the combination of mesenchymal with bone-marrow mononuclear cells is the priority candidate: it ranks second for β-cell preservation and insulin sparing with a safety profile indistinguishable from control and now has eight-year complication data warranting replication. Longer follow-up is essential to distinguish transient metabolic improvement from durable disease modification and to adequately capture tumorigenicity, recurrent autoimmunity and treatment-related adverse effects. For replacement therapies in particular, development of hypoimmunogenic cells, encapsulation or other immune-protective strategies that reduce or eliminate the need for systemic immunosuppression may ultimately determine whether the impressive biological potential of stem-cell-derived islets can be translated into a safe, scalable and durable treatment for T1D.

The honest summary is that stem-cell therapy for type 1 diabetes remains promising and unproven, and that the promise has migrated. It no longer rests principally on immunomodulation, whose randomized evidence is null on all outcomes patients care about, but on replacement — where a small number of single-arm recipients of stem-cell-derived islets have achieved what four decades of β-cell preservation research has not. The next version of this synthesis should be able to place those products in the network. Until it can, the ranking presented here describes the relative order of interventions whose absolute benefit has not been established.

Declarations

Funding: The authors received no funding for this work.

Contributions: All authors contributed to conception, interpretation and revision.

Competing interests: All authors declare no conflict of interest.

Data availability: The derived estimates and analysis code accompany this report as supplementary files; the verification step reproducing the previously published ranking is included.

Registration: (To be registered at PROSPERO).

References

1. Manikandan S, Hariprasad R, Bagepally BS. (2026) Efficacy and safety of stem cell therapy in patients with Diabetes Mellitus — a systematic review and meta-analysis. Syst Rev. 15:75.
2. Reichman TW, Markmann JF, Odorico J, et al. (2025) Stem cell–derived, fully differentiated islets for type 1 diabetes (VX-880/zimislecel, FORWARD). N Engl J Med.
3. Carlsson PO, Espes D, Sisay S, Davies LC, Smith CIE, et al. (2023) Umbilical cord-derived mesenchymal stromal cells preserve endogenous insulin production in type 1 diabetes: a phase I/II randomized double-blind placebo-controlled trial. Diabetologia. 66(8):1431-41.
4. Izadi M, Sadr Hashemi Nejad A, Moazenchi M, et al. (2022) Mesenchymal stem cell transplantation in newly diagnosed type-1 diabetes patients: a phase I/II randomized placebo-controlled clinical trial. Stem Cell Res Ther. 13(1):264.
5. Wu Z, Xu X, Cai J, et al. (2022) Prevention of chronic diabetic complications in type 1 diabetes by co-transplantation of umbilical cord mesenchymal stromal cells and autologous bone marrow: a pilot randomized controlled open-label clinical study with 8-year follow-up. Cytotherapy. 24(4):421-7.
6. Ghodsi M, Heshmat R, Amoli M, et al. (2012) The effect of fetal liver-derived cell suspension allotransplantation on patients with diabetes: first year of follow-up. Acta Med Iran. 50(8):541-6.
7. Nasli-Esfahani E, Ghodsi M, Amini P, et al. (2015) Evaluation of fetal cell transplantation safety in treatment of diabetes: a three-year follow-up. J Diabetes Metab Disord. 14:33.
8. Haller MJ, Wasserfall CH, Hulme MA, et al. (2013) Autologous umbilical cord blood infusion followed by oral docosahexaenoic acid and vitamin D supplementation for C-peptide preservation in children with type 1 diabetes. Biol Blood Marrow Transplant. 19(7):1126-9.
9. Haller MJ, Wasserfall CH, Hulme MA, et al. (2011) Autologous umbilical cord blood transfusion in young children with type 1 diabetes fails to preserve C-peptide. Diabetes Care. 34:2567-9.
10. Wu Q, Zheng S, Qin Y, et al. (2020) Efficacy and safety of stem cells transplantation in patients with type 1 diabetes mellitus — a systematic review and meta-analysis. Endocr J. 67(8):827-40.
11. Sun SY, Gao Y, Liu GJ, et al. (2020) Efficacy and safety of stem cell therapy for T1DM: an updated systematic review and meta-analysis. J Diabetes Res. 2020:5740923.
12. Madani S, Amanzadi M, Aghayan HR, et al. (2022) Investigating the safety and efficacy of hematopoietic and mesenchymal stem cell transplantation for treatment of T1DM: a systematic review and meta-analysis. Syst Rev. 11(1):82.
13. He J, Kong D, Yang Z, et al. (2021) Clinical efficacy on glycemic control and safety of mesenchymal stem cells in patients with diabetes mellitus: systematic review and meta-analysis of RCT data. PLoS One. 16(3):e0247662.
14. Ranjbaran H, Mohammadi Jobani B, Amirfakhrian E, Alizadeh-Navaei R. (2021) Efficacy of mesenchymal stem cell therapy on glucose levels in type 2 diabetes mellitus: a systematic review and meta-analysis. J Diabetes Investig. 12(5):803-10.
15. Lakey JRT, Bluguermann C, Jenkins I, et al. (2025) Diabetes in the pediatric population—can stem cell therapy overcome the associated burden disease. Am J Biomed Sci Res. 26(6): AJBSR.MS.ID.003519.
16. Wu Z, Huang S, Li S, et al. (2024) Bone marrow MSC and mononuclear cell combination therapy in type 2 diabetes: a randomized controlled study with 8-year follow-up. Stem Cell Res Ther. 15:339.
17. Wang S, Du Y, Zhang B, et al. (2024) Transplantation of chemically induced pluripotent stem-cell-derived islets under abdominal anterior rectus sheath in a type 1 diabetes patient. Cell. 187(22):6152-64.
18. Mallapaty S. (2024) Stem cells reverse woman’s diabetes—a world first. Nature. 634:295-96.
19. Shapiro AMJ, Ricordi C, Hering BJ, et al. (2006) International trial of the Edmonton protocol for islet transplantation. N Engl J Med. 355:1318-30.
20. Shapiro AMJ, Lakey JRT, Ryan EA, et al. (2000) Islet transplantation in seven patients with type 1 diabetes mellitus using a glucocorticoid-free immunosuppressive regimen. N Engl J Med. 343:230-38.
21. Carlsson PO, Schwarcz E, Korsgren O, Le Blanc K. (2015) Preserved β-cell function in type 1 diabetes by mesenchymal stromal cells. Diabetes. 64(2):587-92.
22. Cai J, Wu Z, Xu X, et al. (2016) Umbilical cord MSC with autologous bone marrow cell transplantation in established type 1 diabetes: a pilot RCT. Diabetes Care. 39(1):149-57.
23. Hu J, Yu X, Wang Z, et al. (2013) Long-term effects of Wharton’s jelly-derived MSC for newly-onset type 1 diabetes. Endocr J. 60(3):347-57.
24. Gu B, Miao H, Zhang J, et al. (2018) Clinical benefits of autologous hematopoietic stem cell transplantation in type 1 diabetes. Diabetes Metab. 44:341-45.
25. Couri CEB, Oliveira MC, Stracieri ABPL, et al. (2009) C-peptide levels and insulin independence following autologous nonmyeloablative HSCT in newly diagnosed type 1 diabetes. JAMA. 301:1573-79.
26. D’Addio F, Valderrama Vasquez A, Ben Nasr M, et al. (2014) Autologous nonmyeloablative HSCT in new-onset type 1 diabetes: a multicentre analysis. Diabetes. 63:3041-46.
27. Thakkar UG, Trivedi HL, Vanikar AV, Dave SD. (2015) Insulin-secreting AD-MSC with BM-HSC from autologous and allogenic sources for type 1 diabetes. Cytotherapy. 17:940-47.
28. Du Y, Liang Z, Wang S, et al. (2022) Human pluripotent stem-cell-derived islets ameliorate diabetes in non-human primates. Nat Med. 28(2):272-82.
29. Hogrebe NJ, Ishahak M, Millman JR. (2023) Developments in stem cell-derived islet replacement therapy for treating type 1 diabetes. Cell Stem Cell. 30(5):530-48.
30. Pagliuca FW, Millman JR, Gürtler M, et al. (2014) Generation of functional human pancreatic β cells in vitro. Cell. 159:428-439.
31. Lappin T, Cheng T. (2021) An urgent need for standardization of stem cells and stem cell-derived products toward clinical applications. Stem Cells Transl Med. 10(2):1-3.
32. Higgins JPT, Thomas J, Chandler J, et al (eds). (2023) Cochrane Handbook for Systematic Reviews of Interventions. Cochrane.
33. Hutton B, Salanti G, Caldwell DM, et al. (2015) The PRISMA extension for network meta-analyses (PRISMA-NMA). Ann Intern Med. 162:777-84.
34. Salanti G, Ades AE, Ioannidis JPA. (2011) Graphical methods and numerical summaries for multiple-treatment meta-analysis (SUCRA). J Clin Epidemiol. 64:163-71.
35. Rücker G, Schwarzer G. (2015) Ranking treatments in frequentist network meta-analysis works without resampling methods (P-score). BMC Med Res Methodol. 15:58.
36. DerSimonian R, Laird N. (1986) Meta-analysis in clinical trials. Control Clin Trials. 7:177-88.
37. Page MJ, McKenzie JE, Bossuyt PM, et al. (2021) The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 372:71.