The Role of Allogeneic Stem Cell Transplantation in Adult Patients With Peripheral T-Cell Lymphoma: A Systematic Review and Meta-Analysis

Abstract

Objectives:

Peripheral T-cell lymphomas (PTCLs) are rare, aggressive non-Hodgkin lymphomas with limited treatment options and poor prognosis, especially upon relapse. Both autologous and allogeneic stem cell transplants have been used, although allogeneic transplantation is preferred for resistant cases. This systematic review and meta-analysis aim to clarify the impact of allogeneic transplantation on survival and treatment outcomes in PTCL patients.

Methods:

We systematically searched the electronic databases PubMed, Scopus, and Web of Science from each database’s inception to September 2024. Following the PRISMA guidelines, we included adult patients with PTCL undergoing allogeneic hematopoietic stem cell transplantation (Allo-HSCT). We used proportional meta-analysis by executing a random-effects model (DerSimonian-Laird), with data transformation via the Freeman-Tukey method. Heterogeneity was assessed through I² and prediction intervals. We evaluated the risk of bias with the Methodological Index for Non-Randomized Studies (MINORS) tool and assessed evidence quality using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. All analyses were performed using the R software package.

Results:

Our search retrieved 448 articles after first scanning and removing the duplicates. A total of 112 titles were eligible for full-text screening. Finally, 10 studies were included with a total of 1586 patients. The 1-year overall survival (OS) proportion of 63.4% (95% confidence interval [CI] = 56.6% to 70%) across 9 studies with 1555 observations and 1077 events. The 1-year progression-free survival (PFS) proportion of 56.1% (95% CI = 51.1% to 61.1%) across 7 studies, with 1224 observations and 711 events. The 1-year non-relapse mortality (NRM) proportion of 22.3% (95% CI = 15.4% to 29.9%) across 6 studies with 1379 observations and 243 events recorded.

Conclusions:

This systemic review supports the use of allogeneic hematopoietic stem cell transplantation for patients with PTCL; however, further research is needed to confirm its various benefits and limitations as a possible innovation in the field.

Introduction

Peripheral T-cell lymphomas (PTCLs) are relatively rare heterogeneous hematologic malignancies categorized under non-Hodgkin lymphomas that require interdisciplinary involvement. ¹ Studies focusing specifically on PTCL are emerging to improve our understanding of disease biology and the development of more effective therapies. ² Most PTCL subtypes are aggressive and chemotherapy-resistant, with poor prognosis. Multiple mechanisms, such as tumor heterogeneity, tumor microenvironment, and signaling pathways, contribute to PTCL resistance. ³ Relapse is common after the administration of most currently available agents, and there are few effective options for salvage therapy. In addition, only a small number of patients have each type of PTCL, which further complicates studying new treatments for these diseases in clinical trials. ² T-cell lymphomas have traditionally been treated much like B-cell lymphomas, with a combination chemotherapy regimen. Cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) was the most widely used. ² Both autologous stem cell transplantation (AutoSCT) and allogeneic stem cell transplantation (AlloSCT) have been used as consolidation in the first remission and at relapse. ⁴ It has been concluded that outcomes for AutoSCT are best in CR1, while for patients with resistant or relapsed disease, AlloSCT should be preferred over AutoSCT. To our knowledge, this study represents the first systematic review and meta-analysis (SR-MA) focusing on adults with PTCL without cutaneous T-cell lymphoma (CTCL). This study aims to provide evidence and further clarify the role of Allo-HSCT in improving outcomes for patients with PTCL, focusing on overall survival (OS), progression-free survival (PFS), non-relapse mortality (NRM), and complications associated with this treatment.

Methods

Our study employed a systematic review and meta-analysis methodology, adhering to the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions (PRISMA) ⁵ to ensure methodological rigor. The completed PRISMA checklist is provided as a Supplementary File 2. Given the nature of our review study, patient consent and ethical approval were deemed unnecessary. The study protocol was registered in PROSPERO (CRD42024597436) to enhance transparency and preclude duplication of efforts.

Search strategy

We conducted a comprehensive search on PubMed, Scopus, and Web of Science (WOS), covering articles published until September 2024. The search utilized the keywords “Allogeneic Hematopoietic Stem Cell Transplantation” and “Peripheral T-cell Lymphoma.” The search strategy for each database is outlined in Table 1 in Supplementary File 1.

Inclusion criteria

We utilized the population, intervention, comparator, outcomes, and study design (PICOS) selection criteria to determine the included studies. We included studies meeting the following PICOS criteria: (1) Population: adults aged 18 years and older diagnosed with peripheral non-cutaneous T-cell lymphoma (PTCL) of both genders; (2) Intervention: allogeneic hematopoietic stem cell transplantation (Allo-HSCT); (3) Comparison: no control group; (4) Outcomes: OS, PFS, NRM, and 1-year chronic graft vs host disease (GvHD) in at least on time point from 1 year to 5 years or providing Kaplan-Meier curves where data can be extracted; and (5) Study Design: retrospective or prospective cohort studies and clinical randomized controlled trials (RCTs).

Studies that included mixed populations of PTCL and CTCL were carefully reviewed to distinguish the 2 groups based on histopathological confirmation, clinical presentation, and available data. In cases where the study did not provide separate results for PTCL and CTCL patients, we excluded those studies from our analysis.

The articles retrieved through the systematic search were imported to an Excel sheet, and then the titles and abstracts of the search results were screened for relevance by 2 authors, where duplicates were determined and removed manually. After duplicates had been removed, potentially eligible studies were then retrieved for full-text screening. The final list of included trials was agreed upon by discussion between all authors. Disagreement among reviewers was resolved through consensus. The reference lists of the retrieved studies were manually screened for any additional eligible studies.

Exclusion criteria

We excluded studies that did not meet the specified criteria: (1) a sample size of fewer than 30 patients; (2) including more than 10% of the total sample size comprised of patients with CTCL; (3) involved pediatric populations; (4) not reporting separated results for PTCL patients treated with Allo-HSCT; and (5) animal studies, in vitro experiments, meeting and conference abstracts, secondary analyses, or subgroup analyses.

Data extraction

A standardized data extraction form using Google Spreadsheet was developed, then 2 authors independently extracted the data, and a third author resolved any discrepancies in data extraction by discussion. We also used WebPlotDigitizer (WebPlotDigitizer, 2023), a tool that allowed us to extract data from the Kaplan-Meier curves when needed. The form contained essential study characteristics, including study design, sample size, demographics such as median age and gender distribution, histological subtypes, and disease stage at diagnosis. Treatment details included conditioning regimens—classified as Myeloablative Conditioning (MAC) or Reduced Intensity Conditioning (RIC)—stem cell transplant type, stem cell source (bone marrow or peripheral blood), and donor type (matched related, unrelated, mismatched, or haploidentical). Outcomes of interest included the status at transplantation (eg, complete remission, partial remission, stable disease, progressive disease, or refractory), follow-up duration (median and range), and survival outcomes, such as OS, PFS, treatment-related mortality (TRM), and chronic graft vs host disease (GvHD).

Risk-of-bias assessment

The risk of bias was assessed in the included studies independently by 2 authors using the Methodological Index for Non-Randomized Studies (MINORS). ⁶ The global ideal score is 16 for non-comparative studies and 24 for comparative studies. The items are scored 0 (not reported), 1 (reported but inadequate), or 2 (reported and adequate). Disagreement among reviewers was resolved through consensus.

Data analysis

A proportional meta-analysis was conducted by pooling the overall proportions of OS, PFS, NRM, and chronic graft vs host disease (cGvHD) at different time points with 95% confidence intervals (CIs). A random-effects model (DerSimonian-Laird) was employed, and the variances of the raw proportions were stabilized with arcsine square root using the Freeman-Tukey method. High heterogeneity is expected in such studies, and depending solely on I² level can be misleading as it does not necessarily indicate true heterogeneity, so herein we adopted prediction intervals (PIs) in conjunction with I² levels as a measure of heterogeneity, but in general, I² value of ⩾ 50% with a significance level of P < .05 is considered as high heterogeneity. Subgroup analysis was not applicable due to the inevitable limitations of the study and the high level of diversity observed among them. We performed a leave-one-out test by excluding 1 study in each outcome to assess the effect of individual studies to eliminate heterogeneity. If there is a significant decrease in I² levels or PIs observed, we will report these scenarios along with the results. Publication bias was done using the Egger test in 1 outcome (4-year OS) because it necessitates including at least 10 studies to obtain accurate results. All meta-analyses were conducted using R version 4.4.0.

Assessment of the strength of the evidence

We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) ⁷ scale to assess the certainty of the effect estimate by classifying evidence into high quality, where further research is unlikely to alter effect estimates; moderate quality, indicating that additional studies may affect confidence; low quality, which highlights the need for more research that could significantly influence confidence; and very low quality, reflecting high uncertainty regarding the estimates across the following predefined domains: the number and design of studies, risk of bias, heterogeneity, indirectness, imprecision, and other considerations.

Results

Search result

Our search retrieved 448 articles after first scanning and removing the duplicates. A total of 112 titles were eligible for full-text screening. Finally, 10 studies were included with a total of 1586 patients, as presented in the PRISMA flow diagram (Figure 1). All included studies were retrospective cohort studies. The detailed baseline characteristics of patients in the included studies are demonstrated in Table 1.

Figure 1.

Study selection flow diagram.

Table 1.

Summary of the characteristics of the included studies.

Type of study		Massoud et al 8	Berning et al 9	Mamez et al 10	Castagna et al 11	Kyriakou et al 12	Modi et al 13	Savani et al 14	Hsu et al 15	Wulf et al 16	Rohlfing et al 17
		Retrospective	Retrospective	Retrospective	Retrospective	Retrospective	Retrospective	Retrospective	Retrospective	Retrospective	Retrospective
Number of patients		55	135	285	68	45	39	803	41	84	31
Age (median, range)	At transplantation	48 (18–67)	43.4 (18.3–67.7)	49.5	51 (21–70)	48 (23-68)	50	51 (18-65)	40	50 (17-74)	51 (21-65)
Gender (%)	Male	62	68	67	-	53	64	62	-	71	68
	Female	38	32	33	-	47	36	38	-	29	32
Follow-up in months (Median, range)		12	58 (42–71)	72.4	55 (12-116)	29 (6-76)	40	60 (3-145)	21.2	14.5 (1.5-114)	69.6 (16.8-147.6)
Histological subtype(s) (%)		AITL: 20%, ALCL: 15%, ALK pos.: 4%, ALK neg.: 4%, ALK unknown: 7%, ATLL: 11%, EATL: 9%, ENKTL: 2%, HSTL: 7%, MF: 4%, PTCL, NOS: 29%, T-PLL: 4%.	NK/TCL: 100%.	NOS: 39%, AITL: 29%, ALCL (ALK+): 7%, ALCL (ALK-): 7%, ALCL (unknown ALK status): < 1%, ATLL: 6%, NK/T-cell Lymphoma: 6%, HSTL: 4%, EATL: 1%, LGL: < 1%.	T-NOS: 37%, AITL: 26%, ALCL ALK−: 6%, ALCL ALK +: 12%, ALCL ALK unknown: 1%, SS/MF: 7%, T gamma-delta: 4%, NK nasal type: 3%, Others: 3%.	AITL: 100%.	PTCL, NOS: 41%, AITL: 21%, ALCL: 21%, HSTL: 5%, CTCL: 8%, NKCL: 5%.	PTCL, NOS: 45.4%, ALCL: 27.5%, AITL: 23.9%.	PTCL, NOS: 29.2%, ALCL: 7.3%, ALK (+): 2.4%, ALK (−): 2.4%, ALK unknown: 2.4%, AITL: 4.8%, Extranodal NK/T-cell Lymphoma: 19.5%, Other: 39.2%	PTCL-NOS: 35.7%, AITL: 20.2%, ALCL: 17.9%, NK/T: 4.8%, T-LBL: 6.0%, T-PLL: 7.1%, Other: 8.3%	AITL: 20%, ALCL: 15%, ALK pos.: 4%, ALK neg.: 4%, ALK unknown: 7%, ATLL: 11%, EATL: 9%, ENKTL: 2%, HSTL: 7%, MF: 4%, PTCL, NOS: 29%, T-PLL: 4%.
Karnofsky index (median, range)/percentage		80 (40–100)	80–100%: (63.0%), <80%: (4.4%), Unknown: (32.6%)	100% (35%), 90-80% (59%), less than 70% (6%)	-	80 or less: 20%	80 (50–100)	90% or less: 56.4%	-	-	-
Stage at Diagnosis (%)	I-Ii (Localized)	16	30	15	-	-	-	-	7	-	-
	III–IV (Advanced)	66	61	85	-	68	79	-	93	-	-
	Unknown/Other	18	8.6	-	-	32	21	-	-	-	31
Prior treatments		-	Anthracycline: 26.7%, Asparaginase-based: 50.4%, DEVIC or VIPD: 5.9%, Gemcitabine: 0.7%, Unknown: 5.9%	Previous AutoSCT: 23%, Tandem Auto/AlloSCT: 9%.	Previous AutoSCT: 28%	-	Previous AutoSCT: 31%	Previous AutoSCT: 30.5%	-	Previous AutoSCT: 23%	Previous AutoSCT: 39%
Status at transplantation (%)	CR	24	53	62	74	27	41	47	14.6	15	-
	PR	29	27	27	21	22	5	35	-	41	-
	SD	-	2	-	-	-	-	-	-	16	-
	PD	-	18	11	-	29	31	-	-	26	-
	unknown	-	1	-	5	-	-	-	-	-	45
	Refractory	47	-	-	-	22	23	15	4.9	-	55
Donor type (%)	MRD	36	36	45	29	2	54	43	51	26	55
	MUD	40	44	36	25	36	46	10	29	63	45
	MMRD	-	21	5	3	4	-	-	15	10	-
	Haploidentical	-	2	2	43	58	-	11	5	-	-
	Other/unknown	24	1	12	-	-	-	-	-	-	-
Stem cell source (%)	Bone marrow	11	9	17	15	9	5	12	0	-	-
	Peripheral blood	89	91	71	85	91	95	88	100	5	-
	Unknown	-	-	12	-	-	-	-	-	95	31
Conditioning regimen (%)	RIC	9	47	62	30	46.5	92	32	32	43	39
	MAC	90	53	38	40	53.5	8	68	68	57	58
	Other/unknown	-	-	-	-	-	-	-		-	3

AITL: Angioimmunoblastic T-cell Lymphoma; ALCL: Anaplastic Large-Cell Lymphoma; ALK: Anaplastic Lymphoma Kinase; ATLL: Adult T-cell Leukemia/Lymphoma; EATL: Enteropathy-Associated T-cell Lymphoma; ENKTL: Extranodal NK/T-cell Lymphoma; HSTL: Hepatosplenic T-cell Lymphoma; MF: Mycosis Fungoides; PTCL-NOS: Peripheral T-cell Lymphoma Not Otherwise Specified; T-PLL: T-cell Prolymphocytic Leukemia; NK/TCL: NK/T-cell Lymphoma; CTCL: Cutaneous T-cell Lymphoma; NKCL: NK-cell Lymphoma; LGL: Large Granular Lymphocytic Leukemia; SS/MF: Sézary Syndrome/Mycosis Fungoides; T gamma-delta: T gamma-delta Lymphoma; VIPD: Vincristine, Ifosfamide, Prednisone, Doxorubicin; DEVIC: Dexamethasone, Etoposide, Ifosfamide, Carboplatin; CR: Complete Remission; PR: Partial Remission; SD: Stable Disease; PD: Progressive Disease; RIC: Reduced Intensity Conditioning; MAC: Myeloablative Conditioning; AutoSCT: Autologous Stem Cell Transplant; AlloSCT: Allogeneic Stem Cell Transplant; MRD; matched related donor; MUD: matched unrelated donor; MMRD: Mismatched related donor.

Study characteristics

All the 10 included studies are retrospective cohorts,^{8 -17} with a total of 1586 patients, the median age varies from 40 to 51 years across most studies. Gender distribution varied slightly, with male predominance in nearly all cohorts. Median follow-up ranged widely, from 12 months to over 6 years. Histological subtypes were heterogeneous, with peripheral T-cell lymphoma not otherwise specified (PTCL-NOS), angioimmunoblastic T-cell lymphoma (AITL), and anaplastic large-cell lymphoma (ALCL) representing the most common diagnoses, although some studies focused exclusively on rarer subtypes (Berning et al ⁹ ) that included NK/T-cell lymphoma (NK/TCL): 100% and (Berning et al ¹² ) AITL: 100%. Moreover, most patients presented with advanced-stage disease (stage III-IV), underscoring the aggressive nature of PTCL, as represented in Table 1.

Risk-of-bias assessment

The overall distribution of the included studies indicated that all studies were below 70%, which suggests a high risk of bias among all of the included studies, as shown in Figure 2 and Table 2.

Figure 2.

Risk-of-bias scores based on MINORS criteria.

Table 2.

Risk-of-bias assessment using MINORS criteria.

Study	A clearly stated aim	Inclusion of consecutive patients	Prospective collection of data	Endpoints appropriate to the aim of the study	Unbiased assessment of the study endpoint	Follow-up period appropriate to the aim of the study	Loss to follow-up less than 5%	Prospective calculation of the study size	Total
Berning 2023	2	2	0	2	1	2	1	0	10/16
Castagna 2021	2	2	0	2	0	2	1	0	9/16
Hsu 2018	2	1	0	2	0	2	1	0	8/16
Kyriakou 2009	2	2	0	2	1	2	2	0	11/16
Mamez 2020	2	2	0	2	0	2	1	0	9/16
Massoud 2024	2	2	0	2	0	2	1	0	9/16
Modi 2019	1	2	0	2	0	2	2	0	9/16
Rohlfing 2018	2	2	0	1	0	2	1	0	8/16
Savani 2022	2	2	0	2	0	2	1	0	9/16
Wulf 2018	1	1	0	2	0	2	1	0	7/16

Outcomes

Overall survival

The random-effects model estimated a 1-year OS including 1555 observations and 1077 events across 9 studies.

The pooled proportion is 63.4% (95% CI = 56.6% to 70%). There was substantial heterogeneity across the studies, with an I² of 83% (P < .01) (Figure 1 in Supplementary File 1) A leave-one-out analysis revealed that removing Savani et al ¹⁴ resulted in a pooled estimate of 61.4% (95% CI = 0.5513 to 0.6758), with I² = 61.8% (P = .01) (Figure 2 in Supplementary File 1).

The 2-year OS proportion of 59.7% (95% CI = 53.2% to 66.1%) across 9 studies with 981 events and 1554 observations. There was substantial heterogeneity across the studies, with an I² of 78.1% (P < .01) (Figure 3 in Supplementary File 1). A leave-one-out test could not address the source of heterogeneity.

The random-effects model estimated over 8 studies with 751 events in 1270 patients, a 3-year (OS) proportion of 53.4% (95% CI = 44.3% to 62.5%). The heterogeneity across the studies was high, with an I² of 84.9% (P < .01) (Figure 4 in Supplementary File 1) A leave-one-out test could not address the source of heterogeneity.

The random-effects model estimated a 4-year OS proportion of 53.9% (95% CI = 47.6% to 60.2%) including 10 studies and 1585 patients with a total of 900 events. High heterogeneity was observed with I² of 76.6% (P < .01) (Figure 5 in Supplementary File 1). A leave-one-out test could not address the source of heterogeneity.

Finally, the 5-year OS proportion of 52.8% (95% CI = 43.7% to 61.8%), with an I² of 83.6% (P < .01), across 1299 patients in 7 studies with total events of 689 (Figure 6 in Supplementary File 1). A leave-one-out test could not address the source of heterogeneity. Table 3 summarizes OS results across different years.

Table 3.

Summary of overall survival results.

Outcome	Number of studies	Total events	Total patients	Proportion	95% CI	95% PI	Heterogeneity	Leave-one-out
1-year OS	9	1077	1555	63.46%	56.65% to 70.02%	44.67% to 80.39%.	I2 = 63.37% [56.24%; 70.23%]	Removing Savani (2022) I2 = 61.45% [55.13%, 67.58%], P = .01
2-year OS	9	981	1554	58.38%	51.19% to 65.40%	38.60% to 76.88%	I2 = 78.2% [58.7%; 88.4%]	-
3-year OS	8	1270	751	53.33%	43.27% to 63.26%	26.54% to 79.15%	I2 = 84.9% [72.1%; 91.8%]	-
4-year OS	10	900	1585	53.63%	45.86% to 61.31%	31.25% to 75.28%	I2 = 76.6% [56.8%; 87.3%]	-
5-year OS	7	689	1229	52.65%	41.69% to 63.49%	24.94% to 79.53%	I2 = 83.6% [67.8%; 91.7%]	-

Progression-free survival

The random-effects model estimated a 1-year PFS proportion of 56.1% (95% CI = 51.1% to 61.1%), with an I² of 40.2% (P = .03) including 711 events from 1224 patients across 7 studies (Figure 7 in Supplementary File 1). A leave-one-out test that removed Castagna et al ¹¹ showed a proportion of 0.5611 [0.5107; 0.6108], with an I² of 40.2% (P = .14) (Figure 8 in Supplementary File 1).

The random-effects model estimated a 2-year PFS proportion of 53.7% (95% CI = 46.9% to 60.3%) across seven studies, with 1224 observations and 643 events. Heterogeneity was substantial, with an I² of 69.5% (P < .01) (Figure 9 in Supplementary File 1). A leave-one-out sensitivity analysis, which excluded the Castagna study, resulted in a slightly lower OS proportion of 50.8% (95% CI = 45.3% to 56.4%), with a reduced I² of 49.5% (P = .08) (Figure 10 in Supplementary File 1).

The analysis included 7 studies, comprising a total of 1224 observations and 615 events. The random-effect model yielded a 3-year PFS proportion of 51.6% (95% CI = 44.5% to 58.6%). Heterogeneity among the studies was substantial, with an I² of 72.1% (P < .01) (Figure 11 in Supplementary File 1). A leave-one-out test could not address the source of heterogeneity.

The analysis encompassed 7 studies, with a total of 1224 observations and 636 events. The random-effect model yielded a 4-year PFS proportion of 52.3% (95% CI = 45.3% to 59.3%). There was significant heterogeneity among the studies, indicated by an I² of 71.7% (P < .01) (Figure 12 in Supplementary File 1). A leave-one-out test could not address the source of heterogeneity.

The analysis included six studies, comprising a total of 1183 observations and 551 events. The random-effect model yielded a 4-year PFS proportion of 48.7% (95% CI = 40.9% to 56.4%). Significant heterogeneity was observed among the studies, as indicated by an I² of 75.9% (P < .01) (Figure 13 in Supplementary File 1). The leave-one-out method demonstrated the random-effects model 0.4507 [0.4212; 0.4803] and reduced the heterogeneity to I² = 0.0% (P = .46) after removing Castanga (Figure 14 in Supplementary File 1). The PFS results are summarized in Table 4.

Table 4.

Summary of patients’ free-survival results.

Outcome	Number of studies	Total events	Total patients	Proportion	95% CI	95% PI	Heterogeneity	Leave-in-out Proportion (random-effect model), 95% CI
1-year PFS	7	711	1224	58.14%	51.75% to 64.41%	43.66% to 71.96%	I2 = 58.2% [3.3%; 81.9%]	Removing Castanga: 56.1% (51.1% to 61.1%), I2 = 40.2%
2-year PFS	7	643	1224	53.84%	45.92% to 61.65%	34.72% to 72.40%	I2 = 69.5% (32.9%; 86.1%)	Removing Castanga: 50.8% (45.3% to 56.4%), I2 = 49.5%
3-year PFS	7	615	1224	51.78%	43.47% to 60.04%	31.56% to 71.71%	I2 = 72.1% [39.7%; 87.1%]	-
4-year PFS	7	636	1224	52.46%	44.27% to 60.60%	32.54% to 72.00%	I2 = 71.7% [38.5%; 86.9%]	-
5-year PFS	6	551	1183	48.84%	39.72% to 58.00%	27.78% to 70.12%	I2 = 0.0% [0.0%; 79.2%]	Removing Castanga: 45.07% [42.12%; 48.03%], I2 = 0.0%

Non-relapse mortality

A total of 6 studies were included, encompassing 1379 observations and 243 events. The random-effects model estimated a 1-year NRM proportion of 22.3% (95% CI = 15.4% to 29.9%). The heterogeneity was significant, with an I² value of 86.0% (P < .01) (Figure 15 in Supplementary File 1), indicating substantial variability among the studies. A leave-one-out test could not address the source of heterogeneity. A leave-one-out test could not address the source of heterogeneity.

Six studies contributed a total of 1224 observations, with 263 events recorded. The random-effects model yielded an estimated 3-year NRM proportion of 26.2% (95% CI = 16.9% to 36.6%). The results indicated high heterogeneity among the studies, with an I² value of 89.7% (P < .01) (Figure 16 in Supplementary File 1), suggesting substantial variability in the NRM rates reported across the included studies. A leave-one-out test could not address the source of heterogeneity.

Four studies contributed data encompassing 950 observations and 250 events. The random-effects model calculated the 5-year NRM proportion to be 33.9% (95% CI = 19.5% to 49.9%). Notably, the results exhibited a high level of heterogeneity, as evidenced by an I² value of 89.3% (P < .01) (Figure 17 in Supplementary File 1), indicating substantial variation in NRM rates across the studies included in the analysis. A leave-one-out test could not address the source of heterogeneity. Table 5 summarizes the NRM results.

Table 5.

Summary of non-relapse mortality results.

Outcome	Number of studies	Total events	Total patients	Proportion	95% CI	95% PI	Heterogeneity	Leave-in-out proportion (random-effect model), 95% CI
1-year NRM	6	243	1379	22.41%	15.14% to 30.61%	6.93% to 43.20%	I2 = 86.0% [71.5%; 93.1%]	-
3-year NRM	3	263	1224	26.44%	15.64% to 38.85%	3.95% to 58.81%	I2 = 89.7% [80.2%; 94.6%]	-
5-year NRM	4	250	950	33.87%	20.88% to 48.17%	9.71% to 63.52%	I2 = 89.3% [75.4%; 95.4%]	-

Chronic graft vs host disease

Five studies contributed a total of 1040 observations, with 426 events recorded. The random-effects model estimated the proportion of events at 38.5% (95% CI = 31.1% to 46.2%) and (95% PI = 23.85% to 54.35%). The analysis revealed high heterogeneity among the studies, as indicated by an I² value of 68.4% (P = .013) (Figure 18 in Supplementary File 1). A leave-one-out test was conducted that showed removing (Berning et al) ⁹ decreased I² value to (43%) and an estimated proportion of 41% (95% CI = 34.8% to 48.2%).

Quality of the evidence

All studies were observational with a high risk of bias, inconsistency, and imprecision noted across the outcomes with high I² values and wide PIs. Indirectness was minimal. Overall, the importance was critical for most of the outcomes but with low certainty of evidence as illustrated in Table 2 in Supplementary File 1.

Publication bias

Publication bias could not be assessed in most of the outcomes except in one outcome statistical (4-year OS), using the Egger test that yielded asymmetry in the funnel plot (P = .3051) thus indicating insignificant publication bias.

Discussion

Our systematic review and meta-analysis studied the efficacy of allogeneic hematopoietic stem cell transplantation (Allo-HSCT) in adult patients with PTCL. We meta-analyzed 10 studies with a total of 1586 patients to evaluate the OS, PFS, TRM, and cGvHD at different time points. The pooled estimates indicated a 1-year OS rate of 63.5% (95% CI = 56.7% to 70.0%), a 1-year PFS proportion of 56.1% (95% CI = 51.1% to 61.1%), and a 1-year NRM proportion of 22.3% (95% CI = 15.4% to 29.9%). These findings may suggest the potential use of Allo-HSCT as a viable treatment option for patients with PTCL, particularly for those who are refractory to standard therapies.

In 2014, Jia Wei et al ¹⁸ published a similar meta-analysis encompassing 344 patients, including pediatrics, patients with T-cell lymphoma without excluding studies that included B-cell lymphomas. The pooled 3-year OS was 49.6% (95% CI = 41.7% to 57.5%), and the estimated pooled incidence of cGvHD was 29.9% (95% CI = 24.3% to 36.1%). Also, a recent meta-analysis ¹⁹ in 2021 included 888 patients without excluding pediatrics and CTCL showed a 2-year OS of 57%, a 3-year OS of 54%, and a 5-year OS of 51%, and PFS was 45%, 50%, and 45%; finally, a 3-year and 5-year NRM of 29%. Those results are relatively aligned with our results in terms of the outcomes except for cGvHD, possibly due to not reporting the time of the event.

This systematic review has several strengths. First, to our knowledge, this is the largest and the most comprehensive meta-analysis on this topic. In addition, it focuses mainly on adult patients with PTCL with a very minimal percentage of CTCL. Also, conducting a robust data analysis method by the employment of data transformation techniques. Furthermore, using the GRADE tool to assess the certainty of evidence.

Despite the strengths, this review is not without limitations, mostly inherited and inevitable: first, the relatively small number of retrospective single-arm cohort studies; also, heterogeneity and residual variation likely reflects differences in patient selection, transplant platforms, and institutional protocols. Factors such as sample size, subtypes, patients’ age, donor types, conditioning regimens, and disease status at transplant or prior treatments may contribute to the observed variability. Although we applied leave-one-out tests to mitigate the impact of individual studies, the high levels of heterogeneity (eg, I² > 80%) suggest that these clinical variables, which vary widely across studies, may play a significant role in the outcomes.

Further research and randomized clinical trials are needed to specifically address the role of Allo-HSCT in adult patients with PTCL, given the extreme shortage of literature. In addition, controlling confounders in patient demographics, such as age (especially in advanced age), conditioning regimens, post-transplant therapies, and other factors, may provide further understanding of the intervention and enhance patient treatment outcomes.

Conclusion

In conclusion, this systematic review provides consistent evidence supporting the usage of allogeneic hematopoietic stem cell transplantation in patients with PTCL, but despite the notable limitations and challenges, further research should address and control other confounders such as patient selection, transplant platforms, and treatment protocols.