Abstract
Background: Relapsed/refractory (R/R) acute leukemia (AL) carries a poor prognosis, with allogeneic hematopoietic stem cell transplantation (allo-HSCT) the primary curative option. Total body irradiation (TBI)-based conditioning is a cornerstone, but optimizing intensity remains challenging. Thiotepa, a bifunctional alkylating agent, may enhance disease control when added to TBI, but data in R/R AL are lacking. Methods: This single-center retrospective study (2020–2025) included R/R AL and lymphoma patients receiving TBI-based conditioning with (TBI + TT) or without (TBI) thiotepa. Propensity score matching (1:2, caliper = 0.2) used 9 covariates. Primary endpoints were overall survival (OS) and disease-free survival (DFS). Secondary endpoints included relapse, non-relapse mortality (NRM), engraftment, GVHD, CMV infection, and mucositis. Kaplan-Meier, Fine-Gray competing risk, and inverse probability of treatment weighting (IPTW) analyses were employed. Results: After matching, 48 TBI + TT and 86 TBI patients were analyzed. OS (HR = 0.901, 95% CI 0.569–1.425, P = 0.655) and DFS (HR = 0.809, 95% CI 0.504–1.299, P = 0.380) were comparable. The 1-year relapse cumulative incidence was lower with TBI + TT (6.7% vs. 17.8%, Gray’s P = 0.035; multivariable sHR = 0.280, 95% CI 0.083–0.942, P = 0.040), though hypothesis-generating given only 3 relapse events. NRM was comparable (Gray’s P = 0.374). Mucositis was significantly higher with TBI + TT (81.2% vs. 64.0%, P = 0.049). IPTW confirmed comparable OS (P = 0.690) and DFS (P = 0.975). Conclusion: Adding thiotepa to TBI-based conditioning for R/R AL is associated with lower relapse incidence without compromising OS, DFS, engraftment, or GVHD, at the cost of increased mucosal toxicity. The relapse finding requires cautious interpretation and prospective validation.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s00277-026-07198-4.
Keywords: Relapsed/refractory acute leukemia, Total body irradiation, Thiotepa, Allogeneic hematopoietic stem cell transplantation, Conditioning regimen, Propensity score matching
Introduction
Relapsed or refractory (R/R) acute leukemia and lymphoma remain a major therapeutic challenge in hematologic oncology. For acute myeloid leukemia (AML), relapse after initial therapy is common, and mixed phenotype acute leukemia (MPAL) carries an even poorer prognosis with limited consensus on optimal treatment [1, 2]. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is the only potentially curative option for R/R disease across these subtypes [2, 3], leveraging the graft-versus-leukemia effect to eradicate residual malignant cells. However, disease status at transplantation is a dominant predictor of outcomes, and patients with active disease or measurable residual disease (MRD) face sharply diminished chances of sustained remission [4, 5], underscoring the need for more effective conditioning approaches that maximize disease control without prohibitive toxicity.
Total body irradiation (TBI)-based myeloablative conditioning has long been a cornerstone of preparative regimens, particularly in acute lymphoblastic leukemia (ALL). The FORUM trial demonstrated that fractionated TBI at 12 Gy with etoposide was superior to chemotherapy-only conditioning in pediatric high-risk ALL [6], and meta-analyses have confirmed a survival advantage for TBI-based regimens [7]. In haploidentical transplantation with post-transplant cyclophosphamide, TBI-based conditioning has been associated with lower non-relapse mortality and better leukemia-free survival [8]. However, TBI efficacy is dose-dependent and must be balanced against pulmonary toxicity [9], long-term pediatric sequelae including growth retardation and secondary malignancies [10–12], and the need for conformal techniques such as total marrow and lymphoid irradiation to spare organs at risk [13]. Thiotepa, a bifunctional alkylating agent with excellent tissue penetration and blood-brain barrier crossing, has emerged as a key component of TBI-free myeloablative conditioning. The thiotepa-busulfan-fludarabine (TBF) regimen is widely used in haploidentical transplantation [14, 15], with favorable outcomes reported in older adults with ALL [16] and R/R AML [17]. In autologous transplantation for lymphoma, thiotepa-containing regimens have demonstrated durable disease control across multiple histologic subtypes, including primary central nervous system lymphoma [18–21]. A promising strategy involves adding thiotepa directly to TBI-based conditioning rather than substituting for irradiation, as demonstrated in a pediatric cohort where this combination was associated with lower relapse rates without increased transplant-related mortality [22]. Intensified conditioning with agents such as decitabine has also shown encouraging results [14, 23]. Despite these advances, relapse remains the primary cause of treatment failure in R/R patients [8, 24, 25], and most existing comparisons have evaluated TBI and chemotherapy as mutually exclusive alternatives, primarily in patients transplanted in first complete remission [24, 25, 26, 27].
A critical unanswered question is whether adding thiotepa to a TBI-based platform yields superior outcomes compared with TBI-based conditioning alone in the R/R setting. The relative efficacy of TBI-based and thiotepa-based conditioning in the post-CAR T-cell era for relapsed lymphomas has not been systematically studied [28], and the ability of intensified conditioning to overcome MRD remains contested [23, 29, 30]. To address this evidence gap, the present study performed a single-center retrospective analysis with propensity score matching, comparing TBI-based conditioning with and without thiotepa in R/R acute leukemia and lymphoma patients, with overall survival and disease-free survival as primary endpoints.
Methods
Patients
This single-center retrospective study was conducted at the Department of Hematology, Aerospace Center Hospital, Beijing, China. Patients with R/R AL who underwent allo-HSCT between January 2020 and June 2025 were consecutively screened for eligibility. Inclusion criteria were as follows: (1) confirmed diagnosis of AML, ALL, mixed phenotype acute leukemia (MPAL), or myelodysplastic syndrome (MDS) with relapsed or refractory disease status; (2) age less than or equal to 60 years; and (3) receipt of TBI-based myeloablative conditioning with or without thiotepa. Exclusion criteria comprised: (1) severe cardiac, pulmonary, hepatic, or renal dysfunction; and (2) active, uncontrolled infection at the time of conditioning. The study was approved by the Institutional Review Board of Aerospace Center Hospital and conducted in accordance with the Declaration of Helsinki.
Conditioning regimens
Patients were categorized into two groups based on the conditioning regimen received. The TBI group received a TBI-based regimen consisting of TBI combined with either CLAG (clofarabine, cytarabine, and granulocyte colony-stimulating factor) plus cyclophosphamide (CY) and anti-thymocyte globulin-Fresenius (ATG-F), or FLAG (fludarabine, cytarabine, and granulocyte colony-stimulating factor) plus CY and ATG-F. The TBI + TT group received an identical backbone with the addition of thiotepa at a dose of 5 mg/kg/day administered intravenously for 2 consecutive days (total dose 10 mg/kg) on days − 7 and − 6, preceding the administration of TBI. TBI was delivered using a linear accelerator with a total dose of 8–12 Gy, fractionated over 3–4 days, with dose adjustments based on patient age, disease type, and organ function. GVHD prophylaxis was uniform across both groups and consisted of calcineurin inhibitors (cyclosporine A or tacrolimus) combined with short-course methotrexate. ATG-F was administered for ex vivo T-cell depletion in recipients of haploidentical or unrelated donor grafts.
Endpoint
The primary endpoints were overall survival (OS), defined as the time from allo-HSCT to death from any cause, and disease-free survival (DFS), defined as the time from allo-HSCT to relapse or death from any cause, whichever occurred first. Secondary endpoints included the cumulative incidence of relapse, defined as morphologic or molecular evidence of disease recurrence; non-relapse mortality (NRM), defined as death without preceding relapse; neutrophil and platelet engraftment; acute GVHD (aGVHD) graded by consensus criteria; chronic GVHD (cGVHD); CMV infection detected by polymerase chain reaction; and mucositis. Safety assessments additionally included mucositis-associated infections and mucositis-associated mortality.
Statistical analysis
Propensity score matching was performed using logistic regression to estimate the probability of receiving TBI + TT conditioning. Covariates included in the propensity score model were age, diagnosis, gender, disease status prior to transplant, central nervous system leukemia involvement, donor type, donor gender, minimal residual disease (MRD) status, and bone marrow blast percentage. A 1:2 nearest-neighbor matching algorithm was employed with a caliper width of 0.2 standard deviations of the logit of the propensity score. Covariate balance was assessed using standardized mean differences (SMD), with an SMD of less than 0.1 considered indicative of adequate balance.
Continuous variables were compared using the Student’s t-test or Mann-Whitney U test, as appropriate. Categorical variables were compared using the chi-square test or Fisher’s exact test. OS and DFS were estimated using the Kaplan-Meier method and compared using the log-rank test. The proportional hazards assumption was verified using Schoenfeld residuals. Cumulative incidence functions for relapse and NRM were estimated using the Fine-Gray competing risk model and compared using Gray’s test. Relapse and NRM were reported as 1-year and 3-year cumulative incidences, while CMV infection and aGVHD were reported as 100-day cumulative incidences.
Sensitivity analyses included subgroup analyses stratified by disease type (AML versus ALL) and age group (pediatric [age < 18 years] versus adult), as well as inverse probability of treatment weighting (IPTW) using the full unmatched cohort. Interaction tests were performed to evaluate effect modification by disease type and age group. A post hoc statistical power analysis was conducted to determine the range of hazard ratios detectable given the observed number of events. All statistical tests were two-sided, and a P-value of less than 0.05 was considered statistically significant. Statistical analyses were performed using R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria).
Results
Patient characteristics
A total of 415 patients were initially identified during the study period, comprising 337 patients in the TBI group and 78 patients in the TBI + TT group. After data cleaning, 400 patients were retained for analysis (337 TBI, 63 TBI + TT), with 15 patients in the TBI + TT group excluded due to missing overall survival time data. Following 1:2 propensity score matching, 134 patients were included in the matched cohort (48 TBI + TT, 86 TBI). All covariates achieved adequate balance with SMD values below the 0.1 threshold, with the exception of donor type (SMD = 0.135) and bone marrow blast percentage (SMD = 0.122), which slightly exceeded the threshold but remained within an acceptable range for comparative analysis. The baseline characteristics of the matched cohort are summarized in Table 1, and the propensity score matching quality assessment is presented in Fig. S1.
Table 1.
Baseline characteristics of patients in the propensity score-matched cohort
| Variable | Level | TBI (n = 86) | TBI + TT (n = 48) | SMD |
|---|---|---|---|---|
| n | 86 | 48 | ||
| Age at transplantation, years | Mean (SD) | 29.7 (13.6) | 29.8 (14.2) | 0.008 |
| Disease type | AML | 57 (66.3) | 33 (68.8) | 0.045 |
| ALL | 18 (20.9) | 9 (18.8) | ||
| MPAL | 2 (2.3) | 0 (0) | ||
| Lymphoma | 9 (10.5) | 6 (12.5) | ||
| Gender | Male | 50 (58.1) | 28 (58.3) | 0.003 |
| Female | 36 (41.9) | 20 (41.7) | ||
| Pre-transplant status | CR | 16 (18.6) | 10 (20.8) | 0.04 |
| NR | 70 (81.4) | 38 (79.2) | ||
| Blast percentage | ≤ 5 | 16 (18.6) | 10 (20.8) | 0.122 |
| 6–19 | 15 (17.4) | 7 (14.6) | ||
| 20–49 | 13 (15.1) | 7 (14.6) | ||
| 50–79 | 17 (19.8) | 13 (27.1) | ||
| ≥80 | 25 (29.1) | 11 (22.9) | ||
| MRD status | Negative | 10 (11.6) | 6 (12.5) | 0.019 |
| Positive | 76 (88.4) | 42 (87.5) | ||
| CNS leukemia | No | 67 (77.9) | 37 (77.1) | 0.014 |
| Yes | 19 (22.1) | 11 (22.9) | ||
| Donor type | HFD | 59 (68.6) | 37 (77.1) | 0.135 |
| MSD | 19 (22.1) | 8 (16.7) | ||
| URD | 6 (7) | 2 (4.2) | ||
| UCB | 2 (2.3) | 1 (2.1) | ||
| Donor gender | Male | 62 (72.1) | 37 (77.1) | 0.081 |
| Female | 24 (27.9) | 11 (22.9) |
Propensity score matching was performed using logistic regression with the following covariates: Age, disease type, gender, pre-transplant status, CNS leukemia, donor type, donor gender, MRD status, and blast percentage. Matching was performed using 1:2 nearest neighbor method with a caliper of 0.2 standard deviations of the logit of the propensity score. Cluster-robust standard errors (clustered by matched set) were used in subsequent analyses to account for the 1:2 matching design. Standardized mean differences (SMD) < 0.1 indicate adequate balance. Continuous variables are expressed as mean (standard deviation). Categorical variables are expressed as n (%)
Abbreviations: TBI Total Body Irradiation, TT Thiotepa, SMD Standardized Mean Difference, MRD Measurable Residual Disease, CNSL Central Nervous System Leukemia, HFD Haploidentical Family Donor, MSD Matched Sibling Donor, URD Unrelated Donor, CR Complete Remission, NR Non-Remission
Engraftment, GVHD and other complications
Neutrophil engraftment was achieved at a median of 13 (11.75-15) days in the TBI + TT group versus 14 (11–17) days in the TBI group (P = 0.208). Platelet engraftment occurred at a median of 15.5 (12.75-42) days in the TBI + TT group versus 16 (12-27.5) days in the TBI group (P = 0.450). The success rates of neutrophil engraftment were 48/48 (100) in the TBI + TT group and 83/86 (96.5) in the TBI group (P = 0.552), while platelet engraftment success rates were 41/48 (85.4) and 71/86 (82.6), respectively (P = 0.809). Neither engraftment parameter differed significantly between groups, indicating that the addition of thiotepa did not impair hematologic recovery. The cumulative incidence curves for engraftment are shown in Fig. S2.
With respect to transplant-related complications, the incidence of grades II-IV aGVHD was 16/48 (33.3%) in the TBI + TT group compared with 34/86 (39.5%) in the TBI group (P = 0.577). The incidence of cGVHD was 19/48 (39.6%) versus 36/86 (41.9%) (P = 0.856), and CMV infection occurred in 20/48 (41.7%) versus 42/86 (48.8%) (P = 0.473). Notably, mucositis was significantly more frequent in the TBI + TT group (39/48 (81.2%) vs. 55/86 (64%), P = 0.049). Mucositis-associated infections showed a trend toward higher incidence in the TBI + TT group (31/48 (64.6%) vs. 42/86 (48.8%), P = 0.103), though this did not reach statistical significance. Importantly, mucositis-associated mortality did not differ between groups (9/48 (18.8%) vs. 17/86 (19.8%), P = 1.000) (Figs. 1 and 2). These results are detailed in Table 2 and visualized in Fig. 3.
Fig. 1.
Kaplan-Meier curves for overall survival (A) and disease-free survival (B) in the propensity score-matched cohort. The TBI + TT group is shown in red and the TBI group is shown in blue. The number of patients at risk at each time point is displayed below the x-axis. P-values were derived from two-sided log-rank tests. HR, hazard ratio; CI, confidence interval; TBI, total body irradiation; TT, thiotepa; OS, overall survival; DFS, disease-free survival
Fig. 2.
Cumulative incidence of relapse (A) and non-relapse mortality (NRM) (B) in the propensity score-matched cohort, estimated using the Fine-Gray competing risk model. The TBI + TT group is shown in red and the TBI group is shown in blue. P-values were derived from Gray’s test. sHR, subdistribution hazard ratio; CI, confidence interval; NRM, non-relapse mortality; TBI, total body irradiation; TT, thiotepa
Table 2.
Engraftment and complications in the propensity score-matched cohort
| Category | Variable | TBI + TT (n = 48) | TBI (n = 86) | P value |
|---|---|---|---|---|
| Engraftment | Patients evaluable for WBC engraftment | 48 | 86 | |
| Patients evaluable for PLT engraftment | 48 | 86 | ||
| WBC engraftment time, days (median, IQR) | 13 (11.75-15) | 14 (11–17) | 0.2078 | |
| PLT engraftment time, days (median, IQR) | 15.5 (12.75-42) | 16 (12-27.5) | 0.4495 | |
| WBC engraftment success, n/N (%) | 48/48 (100) | 83/86 (96.5) | 0.5525 | |
| PLT engraftment success, n/N (%) | 41/48 (85.4) | 71/86 (82.6) | 0.8093 | |
| GVHD | aGVHD, n/N (%) | 16/48 (33.3%) | 34/86 (39.5%) | 0.577 |
| cGVHD, n/N (%) | 19/48 (39.6%) | 36/86 (41.9%) | 0.8558 | |
| aGrade 0 | 32 | 52 | ||
| aGrade 1 | 2 | 13 | ||
| aGrade 2 | 9 | 15 | ||
| aGrade 3 | 3 | 2 | ||
| aGrade 4 | 2 | 4 | ||
| Infection | CMV infection, n/N (%) | 20/48 (41.7%) | 42/86 (48.8%) | 0.4727 |
| Mucositis | Mucositis, n/N (%) | 39/48 (81.2%) | 55/86 (64%) | 0.0485 |
| Mucositis-related infection, n/N (%) | 31/48 (64.6%) | 42/86 (48.8%) | 0.1035 | |
| Mucositis-related death, n/N (%) | 9/48 (18.8%) | 17/86 (19.8%) | 1 |
All outcomes were assessed in the propensity score-matched cohort (1:2 matching). Continuous variables are expressed as median (interquartile range) and compared using the Mann-Whitney U test. Categorical variables are expressed as n/N (%) and compared using Fisher’s exact test
Abbreviations: TBI Total Body Irradiation, TT Thiotepa, WBC White Blood Cell, PLT Platelet, GVHD Graft-Versus-Host Disease, aGVHD acute GVHD, cGVHD chronic GVHD, CMV Cytomegalovirus, IQR Interquartile Range
Fig. 3.
Short-term complications following allogeneic hematopoietic stem cell transplantation in the propensity score-matched cohort. (A) Cytomegalovirus (CMV) infection cumulative incidence at 100 days. (B) Acute graft-versus-host disease (aGVHD) cumulative incidence at 100 days. (C) Mucositis incidence and mucositis-associated infection incidence. (D) Mucositis-associated mortality and cGVHD incidence. The TBI + TT group is shown in red and the TBI group is shown in blue. P-values were derived from Gray’s test or Fisher’s exact test, as appropriate. TBI, total body irradiation; TT, thiotepa; CMV, cytomegalovirus; aGVHD, acute graft-versus-host disease; cGVHD, chronic graft-versus-host disease
Competing risk analysis
In the competing risk analysis, the 1-year cumulative incidence of relapse was 6.7% in the TBI + TT group compared with 17.8% in the TBI group, and the 3-year cumulative incidence was 6.7% versus 21.9%, respectively (Gray’s P = 0.035). The univariable subdistribution hazard ratio (sHR) for relapse was 0.279 (95% CI 0.083–0.932, P = 0.038), and the multivariable sHR was 0.28 (95% CI 0.083–0.942, P = 0.040). A total of 3 relapse events occurred in the TBI + TT group and 20 in the TBI group.
For NRM, the 1-year cumulative incidence was 14% in the TBI + TT group versus 21.1% in the TBI group, and the 3-year cumulative incidence was 22.8% versus 28%, respectively (Gray’s P = 0.374). The univariable sHR for NRM was 0.676 (95% CI 0.321–1.425, P = 0.300). These findings are presented in Table 3; Fig. 2.
Table 3.
Competing risk analysis of relapse and non-relapse mortality
| Outcome | Time Point | TBI + TT | TBI | Gray’s P | Univariate sHR (95% CI) | Uni P | Multivariable sHR (95% CI)* | Multi P |
|---|---|---|---|---|---|---|---|---|
| Relapse | 1-year | 6.7% | 17.8% | 0.0346 | 0.279 (0.083–0.932) | 0.038 | 0.28 (0.083–0.942) | 0.04 |
| 3-year | 6.7% | 21.9% | 0.0346 | 0.279 (0.083–0.932) | 0.038 | 0.28 (0.083–0.942) | 0.04 | |
| NRM | 1-year | 14% | 21.1% | 0.3742 | 0.676 (0.321–1.425) | 0.300 | 0.694 (0.332–1.451) | 0.33 |
| 3-year | 22.8% | 28% | 0.3742 | 0.676 (0.321–1.425) | 0.300 | 0.694 (0.332–1.451) | 0.33 | |
| CMV infection | 100-day | 41.7% | 49.2% | 0.7788 | 0.927 (0.533–1.611) | 0.790 | 0.947 (0.543–1.652) | 0.85 |
| aGVHD | 100-day | 33.6% | 39.5% | 0.4806 | 0.809 (0.449–1.456) | 0.480 | 0.809 (0.444–1.473) | 0.49 |
Cumulative incidence was estimated using the cumulative incidence function (CIF) approach. Gray’s test was used for group comparisons. Univariate sHR: Fine-Gray model with treatment as sole covariate. *Multivariable sHR: Fine-Gray model adjusted for treatment, age, and disease type (AML, ALL). For relapse and NRM, competing events were NRM and relapse, respectively. For CMV infection and aGVHD, death was the competing event
Abbreviations: sHR Subdistribution Hazard Ratio, NRM Non-Relapse Mortality, CMV Cytomegalovirus, aGVHD acute Graft-Versus-Host Disease, CI Confidence Interval
Survival analysis
The 1-year OS rate was 51.6% in the TBI + TT group versus 50.8% in the TBI group, and the 3-year OS rate was 35.9% versus 34.3%, respectively (log-rank P = 0.660). The hazard ratio for OS was 0.901 (95% CI 0.569–1.425, P = 0.655), and the proportional hazards assumption was satisfied (Schoenfeld P = 0.589).
For DFS, the 1-year rate was 51.7% in the TBI + TT group versus 42.9% in the TBI group, and the 3-year DFS rate was 34.9% versus 30.7%, respectively (log-rank P = 0.378). The hazard ratio for DFS was 0.809 (95% CI 0.504–1.299, P = 0.380), with a satisfied proportional hazards assumption (Schoenfeld P = 0.590). It should be noted that at 5 years, the number of patients at risk was small (TBI + TT: 1, TBI: 11), and these estimates should be interpreted with caution. The survival curves are displayed in Fig. 1, and the detailed survival data are presented in Table 4.
Table 4.
Survival analysis results
| Outcome | Metric | TBI + TT (n = 48) | TBI (n = 86) | Log-rank P |
|---|---|---|---|---|
| OS | 1-year survival | 51.6% (38.5–69.3) | 50.8% (41.1–62.8) | 0.6597 |
| 3-year survival | 35.9% (23.1–55.7) | 34.3% (25.2–46.8) | 0.6597 | |
| 5-year survival | 35.9% (23.1–55.7) | 30.3% (21.2–43.2) | 0.6597 | |
| Median survival (months) | 14.6 (9.7-NA) | 12.2 (9.4–22.7) | 0.6597 | |
| HR (95% CI), P | 0.901 (0.569–1.425), P = 0.6547 | Reference | 0.6597 | |
| DFS | 1-year survival | 51.7% (38.5–69.5) | 42.9% (33.5–55) | 0.3779 |
| 3-year survival | 34.9% (22-55.3) | 30.7% (22.1–42.8) | 0.3779 | |
| 5-year survival | 34.9% (22-55.3) | 27.5% (19.1–39.6) | 0.3779 | |
| Median survival (months) | 14 (9-NA) | 9 (5–16) | 0.3779 | |
| HR (95% CI), P | 0.809 (0.504–1.299), P = 0.3802 | Reference | 0.3779 |
Survival probabilities were estimated using the Kaplan-Meier method and compared using the log-rank test. Hazard ratios (HR) with 95% confidence intervals (CI) were estimated using Cox proportional hazards regression with cluster-robust standard errors (clustered by matched set) to account for the 1:2 propensity score matching design. Proportional hazards assumption was tested using Schoenfeld residuals (cox.zph): OS P = 0.5888, DFS P = 0.5895. Survival rates are presented as percentage (95% CI). 5-year survival rates should be interpreted with caution given small at-risk numbers
Abbreviations: TBI Total Body Irradiation, TT Thiotepa, OS Overall Survival, DFS Disease-Free Survival, HR Hazard Ratio, CI Confidence Interval
Sensitivity analyses
Subgroup analyses were performed to assess the robustness of the primary findings across disease types and age groups. In the AML subgroup (n = 90), the OS HR was 0.913 (95% CI 0.516–1.616, P = 0.755) and the DFS HR was 0.906 (95% CI 0.5-1.641, P = 0.744). In the ALL subgroup (n = 27), the OS HR was 0.757 (95% CI 0.3-1.913, P = 0.556) and the DFS HR was 0.618 (95% CI 0.252–1.514, P = 0.292).
In the pediatric subgroup (n = 31), the OS HR was 1.001 (95% CI 0.394–2.542, P = 0.998) and the DFS HR was 0.962 (95% CI 0.369–2.506, P = 0.936). In the adult subgroup (n = 103), the OS HR was 0.869 (95% CI 0.521–1.448, P = 0.590) and the DFS HR was 0.764 (95% CI 0.454–1.285, P = 0.309).
IPTW analysis using the full unmatched cohort (n = 400) yielded an OS HR of 1.083 (95% CI 0.731–1.605, P = 0.690) and a DFS HR of 1.006 (95% CI 0.679–1.492, P = 0.975), consistent with the primary PSM analysis. Interaction tests for disease type (OS P = 0.938, DFS P = 0.499) and age group (OS P = 0.744, DFS P = 0.590) were not statistically significant, indicating no evidence of effect modification by these factors.
A post hoc power analysis indicated that with 81 OS events, the study was powered to detect hazard ratios of 0.733 or less, or 1.364 or greater, at 80% power. For DFS, with 86 events, the detectable hazard ratio range was 0.739 to 1.353. These results are summarized in Table 5, and the subgroup forest plot is provided in Fig. S3.
Table 5.
Sensitivity analyses including subgroup, IPTW, and post hoc power analyses
| Analysis | n | OS HR (95% CI) | OS P | DFS HR (95% CI) | DFS P |
|---|---|---|---|---|---|
| PSM cohort (overall) | 134 | 0.901 (0.569–1.425) | 0.6547 | 0.809 (0.504–1.299) | 0.3802 |
| Subgroup: AML only | 90 | 0.913 (0.516–1.616) | 0.7547 | 0.906 (0.5-1.641) | 0.7441 |
| Subgroup: ALL only | 27 | 0.757 (0.3-1.913) | 0.5564 | 0.618 (0.252–1.514) | 0.2924 |
| Subgroup: Pediatric (< 18 year) | 31 | 1.001 (0.394–2.542) | 0.9983 | 0.962 (0.369–2.506) | 0.9364 |
| Subgroup: Adult ( ≥18 year) | 103 | 0.869 (0.521–1.448) | 0.5896 | 0.764 (0.454–1.285) | 0.3094 |
| IPTW analysis (full cohort) | 400 | 1.083 (0.731–1.605) | 0.6903 | 1.006 (0.679–1.492) | 0.9749 |
| Interaction: disease type (AML vs. non-AML) | — | — | 0.9375 | — | 0.4989 |
| Interaction: age group (pediatric vs. adult) | — | — | 0.7438 | — | 0.5902 |
Hazard ratios (HR) compare TBI + TT vs. TBI group. PSM subgroup analyses used cluster-robust standard errors (clustered by matched set). IPTW analysis used robust sandwich variance estimation (robust=TRUE) on the full cohort. IPTW weights: median = 1.06, range = 1–10. Effective sample size (ESS): treated = 40.2, control = 300.6, total = 209.4 out of 400. Interaction P values test whether treatment effect differs across subgroups, extracted from the interaction term coefficient (not anova)
Abbreviations: PSM Propensity Score Matching, IPTW Inverse Probability of Treatment Weighting, HR Hazard Ratio, CI Confidence Interval, ESS Effective Sample Size
Discussion
This propensity score-matched analysis using a 1:2 matching design with cluster-robust standard errors provides real-world evidence that augmenting TBI-based conditioning with thiotepa for patients with R/R acute leukemia and lymphoma is associated with comparable OS (HR = 0.901, 95% CI 0.569–1.425, P = 0.655) and DFS (HR = 0.809, 95% CI 0.504–1.299, P = 0.380). Notably, the cumulative incidence of relapse was significantly lower in the TBI + TT group (1-year: 6.7% vs. 17.8%, Gray’s P = 0.035), and this association persisted in both univariate (sHR = 0.279, 95% CI 0.083–0.932, P = 0.038) and multivariable Fine-Gray models (sHR = 0.280, 95% CI 0.083–0.942, P = 0.040). The proportional hazards assumption was satisfied for both OS (cox.zph P = 0.589) and DFS (cox.zph P = 0.590). Cluster-robust standard errors were employed to account for within-matched-set correlation [31].
The significant reduction in relapse represents the first evidence that adding thiotepa to TBI-based conditioning may improve disease control in R/R acute leukemia and lymphoma. The persistence of significance after adjustment for age and disease type, with virtually unchanged point estimates (0.279 vs. 0.280), indicates robustness to confounding by these factors. Thiotepa exerts its antileukemic effect through DNA cross-linking and is cell-cycle independent, potentially complementing TBI’s mechanism of action. However, this finding must be interpreted with considerable caution given the extremely limited number of relapse events in the TBI + TT group (n = 3). The estimate is inherently fragile, and the confidence interval is wide. Accordingly, this finding should be considered hypothesis-generating rather than confirmatory. The dissociation between GVHD and relapse suggests that the GVL effect may be preserved without increased alloreactivity [32].
The significantly increased mucositis in the TBI + TT group (81.2% vs. 64.0%, P = 0.049) represents the primary safety concern, reflecting thiotepa’s non-selective cytotoxicity on rapidly dividing mucosal epithelial cells. However, mucositis-related infections (64.6% vs. 48.8%, P = 0.104) and mucositis-related deaths (18.8% vs. 19.8%, P = 1.000) did not differ significantly, suggesting that with adequate supportive care, this toxicity is manageable and does not translate into excess mortality [33]. The comparable incidence of aGVHD (33.3% vs. 39.5%, P = 0.577), cGVHD (39.6% vs. 41.9%, P = 0.856), and CMV reactivation (41.7% vs. 48.8%, P = 0.473) indicates that thiotepa intensification does not exacerbate alloreactivity. NRM was not significantly different between groups (1-year: 14.0% vs. 21.1%, 3-year: 22.8% vs. 28.0%, Gray’s P = 0.374; multivariable sHR = 0.694, 95% CI 0.332–1.451, P = 0.330). The fact that increased mucositis did not translate into higher NRM further supports the feasibility of this regimen.
A notable finding is the direction reversal between PSM and IPTW analyses. In the PSM cohort, both OS (HR = 0.901) and DFS (HR = 0.809) point estimates favored TBI + TT, whereas IPTW on the full cohort of 400 patients yielded opposite directions (OS HR = 1.083, DFS HR = 1.006), both non-significant. PSM analyzes only the 134 matched patients with overlapping propensity score distributions, while IPTW includes all 400 patients, including those with extreme propensity scores [34]. Residual confounding from unmeasured variables may differentially affect the two methods [35], and neither analysis had adequate power to detect a clinically meaningful difference [36]. Subgroup analyses by disease type and age group showed no significant interactions (OS P = 0.938 for disease type, P = 0.744 for age; DFS P = 0.499 and P = 0.590, respectively), supporting generalizability across the studied population. The divergent point estimates between pediatric (OS HR = 1.001) and adult (OS HR = 0.869) subgroups may reflect biological differences in treatment response [37], warranting validation in larger cohorts.
This study has several limitations. First, the retrospective, single-center design inherently carries selection bias, and despite 1:2 PSM, two covariates remained slightly above the 0.1 balance threshold [38]. Second, 15 TBI + TT patients could not be matched; these had higher propensity scores and more adverse features, meaning the matched cohort may not fully represent the spectrum of TBI + TT recipients. Third, the combination of AML, ALL, lymphoma, and MPAL across adult and pediatric populations introduces substantial heterogeneity. Fourth, the sample size is limited: post-hoc power calculations indicated the study could only detect hazard ratios of < = 0.733 or > = 1.37 for OS at 80% power, meaning it was underpowered for the observed effect sizes. Fifth, the relapse finding is based on only 3 events in the TBI + TT group, making it inherently fragile. Sixth, the events-per-variable ratio in the multivariable Fine-Gray model was 5.75, below the recommended minimum of 10. Finally, detailed mucositis severity grading and patient-reported quality-of-life data were not available.
Conclusion
In conclusion, the addition of thiotepa to TBI-based myeloablative conditioning for allo-HSCT in patients with relapsed or refractory acute leukemia yielded comparable overall survival and disease-free survival relative to TBI alone, accompanied by an observational trend toward reduced relapse incidence and increased mucosal toxicity. Engraftment, graft-versus-host disease, and cytomegalovirus infection were not adversely affected. These findings support the further exploration of the TBI+thiotepa regimen in carefully selected patients, but prospective multicenter studies are warranted to validate these observations and establish definitive efficacy and safety.
Supplementary Information
Below is the link to the electronic supplementary material.
Author contributions
Minglu Li and Jingbo Wang conceived of the presented idea. Minglu Li developed the theory and performed the computational analyses. Minglu Li drafted the initial manuscript. Shuqin Zhang, Hongxia Wen, Wei Wei, Yixin Yang and Xiaoyong Man were responsible for data collection, patient follow-up, and verification of clinical information. All authors discussed the results, revised the manuscript critically for important intellectual content, and approved the final version for submission.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Declarations
Permission to reproduce material
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Aerospace central Hospital.
Patient consent
Informed consent was obtained from all individual participants included in the study.
Competing interests
The authors declare no competing interests.
Conflict of interest
The authors have fully disclosed their conflicts of interest in the manuscript and during the submission process. This manuscript has not been published and is not under consideration for publication elsewhere.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Hennawi M, Pakasticali N, Tashkandi H, Hussaini M (2022) Genomic landscape of mixed-phenotype acute leukemia. Int J Mol Sci 23(19):11259. 10.3390/ijms231911259 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Paviglianiti A, Ngoya M, de la Peffault R, Schroeder T, Remenyi P, Morsink L et al (2026) Outcomes of mixed phenotype acute leukemia undergoing allogeneic stem cell transplantation: a retrospective study of the ALWP of EBMT. BMT. 10.1038/s41409-026-02920-2 [DOI] [PubMed]
- 3.Huang F, Yang K, Du J, Xuan L, Liu Q (2025) Current scenario of allogeneic hematopoietic stem cell transplantation in adult patients with relapsed/refractory acute myeloid leukemia: a narrative review. Ann Hematol 104(10):4913–4930. 10.1007/s00277-025-06693-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Aldoss I, Yang D, Malki MMA, Mei M, Mokhtari S, Artz A et al (2021) Allogeneic hematopoietic cell transplantation for relapsed and refractory philadelphia negative B cell ALL in the era of novel salvage therapies. Transpl Cell Ther 27(3):255.e1-255.e9 10.1016/j.jtct.2020.12.020 [DOI] [PubMed] [Google Scholar]
- 5.Short NJ, Rafei H, Daver N, Hwang H, Ning J, Jorgensen JL et al (2020) Prognostic impact of complete remission with MRD negativity in patients with relapsed or refractory AML. Blood Adv 4(24):6117–6126. 10.1182/bloodadvances.2020002811 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Peters C, Dalle JH, Locatelli F, Poetschger U, Sedlacek P, Buechner J et al (2021) Total body irradiation or chemotherapy conditioning in childhood ALL: A multinational, randomized, noninferiority phase III study. J Clin Oncol 39(4):295–307. 10.1200/JCO.20.02529 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Khimani F, Dutta M, Faramand R, Nishihori T, Perez AP, Dean E et al (2021) Impact of total body irradiation-based myeloablative conditioning regimens in patients with acute lymphoblastic leukemia undergoing allogeneic hematopoietic stem cell transplantation: systematic review and meta-analysis. Transpl Cell Ther 27(7):620.e1-620.e9 10.1016/j.jtct.2021.03.026 [DOI] [PubMed] [Google Scholar]
- 8.Dholaria B, Labopin M, Angelucci E, Tischer J, Arat M, Ciceri F et al (2021) Improved outcomes of haploidentical hematopoietic cell transplantation with total body irradiation-based myeloablative conditioning in acute lymphoblastic leukemia. Transpl Cell Ther 27(2):171.e1-171.e8 10.1016/j.jtct.2020.10.008 [DOI] [PubMed] [Google Scholar]
- 9.Vogel J, Hui S, Hua CH, Dusenbery K, Rassiah P, Kalapurakal J et al (2021) Pulmonary toxicity after total body irradiation - critical review of the literature and recommendations for toxicity reporting. Front Oncol 11:708906. 10.3389/fonc.2021.708906 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Hoeben BAH, Pazos M, Seravalli E, Bosman ME, Losert C, Albert MH et al (2022) ESTRO ACROP and SIOPE recommendations for myeloablative Total Body Irradiation in children. Radiother Oncol 173:119–133. 10.1016/j.radonc.2022.05.027 [DOI] [PubMed] [Google Scholar]
- 11.Handgretinger R, Lang P (2022) Could (should) we abandon total body irradiation for conditioning in children with leukemia. Blood Rev 56:100966. 10.1016/j.blre.2022.100966 [DOI] [PubMed] [Google Scholar]
- 12.Gruen A, Exner S, Kuhl JS, von Stackelberg A, Budach V, Stromberger C et al (2022) Total body irradiation as part of conditioning regimens in childhood leukemia-long-term outcome, toxicity, and secondary malignancies. Strahlenther Onkol 198(1):33–38. 10.1007/s00066-021-01810-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Naik Y, Kulkarni U, Lionel S, Selvarajan S, Devasia AJ, Korula A et al (2025) Total marrow and lymphoid irradiation (TMLI) is associated with good early outcomes in patients undergoing matched sibling donor and haplo-identical transplants for acute lymphoblastic leukemia. Transpl Cell Ther 31(1):22e. 10.1016/j.jtct.2024.10.007 [DOI] [PubMed] [Google Scholar]
- 14.Hu Z, Feng Z, Liu S, He H, Dong Y, Fan Z et al (2024) Intensified conditioning containing decitabine versus standard myeloablative conditioning for adult patients with KMT2A-rearranged leukemia: a multicenter retrospective study. BMC Med 22(1):605. 10.1186/s12916-024-03830-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Saraceni F, Labopin M, Raiola AM, Blaise D, Remenyi P, Sora F et al (2023) Thiotepa-busulfan-fludarabine compared to treosulfan-based conditioning for haploidentical transplant with posttransplant cyclophosphamide in patients with acute myeloid leukemia in remission: a study from the acute leukemia working party of the EBMT. Hemasphere 7(10):e952. 10.1097/HS9.0000000000000952 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Banet A, Bazarbachi A, Labopin M, Stocker N, Dulery R, Malard F et al (2023) Thiotepa, busulfan and fludarabine conditioning-regimen is a promising approach for older adult patients with acute lymphoblastic leukemia treated with allogeneic stem cell transplantation. Bone Marrow Transpl 58(1):61–67. 10.1038/s41409-022-01841-0 [DOI] [PubMed] [Google Scholar]
- 17.Luo C, Wu G, Huang X, Ding Y, Huang Y, Song Q et al (2023) Myeloablative conditioning regimens in adult patients with acute myeloid leukemia undergoing allogeneic hematopoietic stem cell transplantation in complete remission: a systematic review and network meta-analysis. Bone Marrow Transpl 58(2):175–185. 10.1038/s41409-022-01865-6 [DOI] [PubMed] [Google Scholar]
- 18.Cohen YI, Lebel E, Zimran E, Shaulov A, Stepensky P, Grisariu S et al (2023) Long-term results with thiotepa-containing conditioning regimens for autologous stem cell transplantation. Transpl Cell Ther 29(8):505.e1-505.e8 10.1016/j.jtct.2023.04.022 [DOI] [PubMed] [Google Scholar]
- 19.Kim DH, Park T, Hong J, Shin DY, Kim I, Yoon SS et al (2025) Autologous stem cell transplantation with thiotepa, busulfan, and cyclophosphamide conditioning in patients with central nervous system lymphoma: a phase II study. Ann Hematol 104(7):3843–3854. 10.1007/s00277-025-06405-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Puckrin R, Chua N, Shafey M, Stewart DA (2022) Improving the outcomes of secondary CNS lymphoma with high-dose thiotepa, busulfan, melphalan, rituximab conditioning and autotransplant. Leuk Lymphoma 63(10):2444–2452. 10.1080/10428194.2022.2068005 [DOI] [PubMed] [Google Scholar]
- 21.Scordo M, Wang TP, Ahn KW, Chen Y, Ahmed S, Awan FT et al (2021) Outcomes associated with thiotepa-based conditioning in patients with primary central nervous system lymphoma after autologous hematopoietic cell transplant. JAMA Oncol 7(7):993–1003. 10.1001/jamaoncol.2021.1074 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Chinnabhandar V, Tran S, Sutton R, Shaw PJ, Mechinaud F, Cole C et al (2020) Addition of thiotepa to total body irradiation and cyclophosphamide conditioning for allogeneic hematopoietic stem cell transplantation in pediatric acute lymphoblastic leukemia. Biol Blood Marrow Transpl 26(11):2068–2074. 10.1016/j.bbmt.2020.07.028 [DOI] [PubMed] [Google Scholar]
- 23.Willasch AM, Peters C, Sedlacek P, Dalle JH, Kitra-Roussou V, Yesilipek A et al (2020) Myeloablative conditioning for allo-HSCT in pediatric ALL: FTBI or chemotherapy?-A multicenter EBMT-PDWP study. Bone Marrow Transpl 55(8):1540–1551. 10.1038/s41409-020-0854-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zhang H, Fan Z, Huang F, Han L, Xu Y, Xu N et al (2023) Busulfan plus cyclophosphamide versus total body irradiation plus cyclophosphamide for adults acute B lymphoblastic leukemia: an open-label, multicenter, phase III trial. J Clin Oncol 41(2):343–353. 10.1200/JCO.22.00767 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Spyridonidis A, Labopin M, Savani B, Giebel S, Bug G, Schonland S et al (2023) Reduced 8-gray compared to standard 12-gray total body irradiation for allogeneic transplantation in first remission acute lymphoblastic leukemia: A study of the acute leukemia working party of the EBMT. Hemasphere 7(1):e812. 10.1097/HS9.0000000000000812 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Bonifazi F, Pavoni C, Peccatori J, Giglio F, Arpinati M, Busca A et al (2022) Myeloablative conditioning with thiotepa-busulfan-fludarabine does not improve the outcome of patients transplanted with active leukemia: final results of the GITMO prospective trial GANDALF-01. Bone Marrow Transpl 57(6):949–958. 10.1038/s41409-022-01626-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Swoboda R, Labopin M, Giebel S, Angelucci E, Arat M, Aljurf M et al (2022) Total body irradiation plus fludarabine versus thiotepa, busulfan plus fludarabine as a myeloablative conditioning for adults with acute lymphoblastic leukemia treated with haploidentical hematopoietic cell transplantation. A study by the Acute Leukemia Working Party of the EBMT. Bone Marrow Transpl 57(3):399–406. 10.1038/s41409-021-01550-0 [DOI] [PubMed] [Google Scholar]
- 28.Hilal T, Mountjoy LJ (2023) Allogeneic hematopoietic stem cell transplant for diffuse large B-cell lymphoma: evolving role in the era of CAR T-cell therapy. Curr Oncol Rep 25(6):599–607. 10.1007/s11912-023-01403-7 [DOI] [PubMed] [Google Scholar]
- 29.Harada K, Morita-Fujita M, Fukuda T, Ozawa Y, Doki N, Toyosaki M et al (2022) Overcoming minimal residual disease using intensified conditioning with medium-dose etoposide, cyclophosphamide and total body irradiation in allogeneic stem cell transplantation for Philadelphia chromosome-positive acute lymphoblastic leukemia in adults. Cytotherapy 24(9):954–961. 10.1016/j.jcyt.2022.03.004 [DOI] [PubMed] [Google Scholar]
- 30.Paras G, Morsink LM, Othus M, Milano F, Sandmaier BM, Zarling LC et al (2022) Conditioning intensity and peritransplant flow cytometric MRD dynamics in adult AML. Blood 139(11):1694–1706. 10.1182/blood.2021014804 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Imbens GW (2015) Matching methods in practice: three examples. J Hum Resour 50(2):373–419. 10.3368/jhr.50.2.373 [DOI] [Google Scholar]
- 32.Choi J, Ritchey J, Prior JL, Holt M, Shannon WD, Deych E et al (2010) In vivo administration of hypomethylating agents mitigate graft-versus-host disease without sacrificing graft-versus-leukemia. Blood 116(1):129–139. 10.1182/blood-2009-12-257253 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Curra M, Pellicioli ACA, Filho NAK, Ochs G, Matte UDS, Sant’Ana Filho M et al (2015) Photobiomodulation reduces oral mucositis by modulating NF-kB. J Biomed Opt 20(12):125008. 10.1117/1.jbo.20.12.125008 [DOI] [PubMed] [Google Scholar]
- 34.Sturmer T, Rothman KJ, Avorn J, Glynn RJ (2010) Treatment effects in the presence of unmeasured confounding: dealing with observations in the tails of the propensity score distribution–A simulation study. Am J Epidemiol 172(7):843–854. 10.1093/aje/kwq198 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Chatton A, Rohrer JM (2024) The causal cookbook: recipes for propensity scores, G-computation, and doubly robust standardization. Adv Methods Pract Psychol Sci 7(1). 10.1177/25152459241236149 [DOI]
- 36.Dudbridge F (2013) Power and predictive accuracy of polygenic risk scores. PLoS Genet 9(3):e1003348. 10.1371/journal.pgen.1003348 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Gruen A, Ebell W, Wlodarczyk W, Neumann O, Kuehl JS, Stromberger C et al (2013) Total body irradiation (TBI) using helical tomotherapy in children and young adults undergoing stem cell transplantation. Radiat Oncol 8(1):92. 10.1186/1748-717x-8-92 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Austin PC (2009) Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 28(25):3083–3107. 10.1002/sim.3697 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.