Data were retracted from the Cooperative Weichteilsarkom Studiengruppe (CWS) European soft tissue sarcoma registry (SoTiSaR) after approval by the principal investigators’ institutional ethical commission. For 50 RMA patients (stage III–IV), information concerning patient history and therapeutic interventions were used for further evaluation. All patients were diagnosed with RMA after the year 2000. Database update and last data collection was June 2020. RMA was confirmed by a local and expert pathologist, with genetic confirmation of PAX3/PAX7-FOXO1 fusions in 35 of 38 cases; 3 cases were fusion-negative. Molecular diagnosis was not done in 12 cases.

Engraftment was defined as absolute neutrophil count (ANC) ≥0.5 × 10⁹/L. Engraftment failure: ANC <0.5 × 10⁹/L by day +28 after allo-HSCT. When patients succumbed within ≤100 days after allo-HSCT, status of chronic GvHD could not be assessed. Further information on chronic GvHD was not available, due to limited recording in the past. Transplant-related mortality (TRM) was defined as any transplant-related death occurring after allo-HSCT without evidence of underlying disease or disease progression. Death of disease (DOD) was defined as any death related to underlying disease progression or relapse. Response was evaluated as follows: progressive disease (PD) was defined as tumor volume progression ≥50%, partial remission (PR) was defined as tumor volume reduction ≥50%, very good partial response (VGPR) was defined as tumor volume reduction ≥90% or persistence of unclear residuals upon imaging, and complete response (CR) was defined as complete disappearance of all visible disease. Measurement of tumor volume was done according to CWS guidance (1.6.1) using the formula V = π/6 × length × width × thickness. HLA-mismatched was defined as ≤9/10 differing HLA-I and -II alleles. In case of the comparison within HLA-mismatched versus HLA-matched setting, survival data were calculated from the time of 1st allo-HSCT until an event (i.e., relapse—EFS or TRM) or DOD for OS. For matched-pair analysis, survival was calculated from first therapeutic intervention until an event occurred (relapse, TRM, or DOD). Glucksberg criteria were used to grade GvHD.

The study group consisted of a total of 50 transplanted patients, with 21 female patients (0.42) and 29 male patients (0.58). Median age at diagnosis was 14 years (range 0–24 years) with a median age at allo-HSCT of 14.0 years (range 0–24 years). Comparable to previously published analyses (18), patients were divided into two groups: HLA-mismatched (n = 35) and HLA-matched (n = 15). In the HLA-mismatched group, graft source was sibling in 2 patients, maternal in 17 patients, paternal in 15 patients, and both maternal and paternal in one patient. For HLA-matched patients, graft source was sibling in 14 (sister in 10/14 patients and brother in 4/14 patients). Patients qualified for allo-HSCT due to stage IV metastatic disease and/or age > 10 years and/or bone/bone marrow metastases at diagnosis or relapsed disease. Oberlin scores were used to assess further risk stratification (6). After conditioning therapy, 27/50 (0.54) patients were transplanted in CR or VGPR and 23/50 (0.46) patients were transplanted in PR or PD. Forty out of 50 (0.80) patients received primary allo-HSCT due to their very high-risk disease and/or lack of response to induction therapy. Allo-HSCT as relapse therapy was implemented in 10/50 patients (0.20). As a planned high-dose induction, 5/50 (0.10) received myeloablative chemotherapy and autologous HSCT (HLA-mismatched group = 4, HLA-matched group = 1) prior to allo-HSCT. Sex, age at diagnosis, age at allo-HSCT, primary site (favorable/unfavorable), bone or bone marrow involvement, stage, Oberlin score, eligibility for allo-HSCT, and disease status at allo-HSCT did not differ significantly in between groups A and B ( Table 1 ). Four patients underwent multiple allo-HSCT, either due to relapse after 1st allo-HSCT (n = 2) or engraftment failure of the 1st allograft (n = 2). Mainly, transplants consisted of peripheral blood-derived stem cells (49 of 55 applied allografts, fraction 0.89), and 5/55 allografts (0.09, 1/55 source unknown) were derived from bone marrow. Treatment decisions were in concordance with approvals of institutional review boards and in accordance with the Helsinki Conference Declaration. Informed consent was signed by all patients or their legal guardians before initiation of therapy.

Reduced-intensity conditioning (RIC) regimens (n = 37) were mostly based on fludarabine (n = 34) or clofarabine (n = 3), fludarabine monotherapy (n = 1), or in combination with either of the following: busulfan (BU, n = 3), cyclophosphamide (CYC, n = 7), etoposide (ETO, n = 3), melphalan (MEL, n = 31), thiotepa (TT, n = 34), or treosulfan (n = 1). Also, in most cases, either antithymocyte globuline (ATG, n = 11) or muromonab (OKT3, n = 18) was added. Total body irradiation (TBI) was performed in a minority of cases (n = 5). As for myeloablative condition regimens (MAC, n = 11), BU/CYC/ETO (n = 3), BU/TT/CYC (n = 1), BU/MEL/CYC (n = 1), BU/CYC/ATG (n = 1), CYC/TT/ATG (n = 1), CYC/treosulfan/ETO (n = 1), high-dose-ETO/ifosfamide (n = 1), and treosulfan/MEL (n = 1) were applied. TBI was given in 2 cases. Conditioning regimens were unknown in 2 cases.

Graft manipulation for HLA-mismatched grafts mainly consisted of CD3/CD19-depletion (n = 21, with TCRalpha/beta/CD19-depletion instead of CD3/CD19-depletion in 2 cases). CD3/CD19 depletions were combined with CD34 selection in 4 cases. CD34 selection was done alone in 3 cases, or in combination with T-cell depletion (n = 1). T-cell depletion alone was performed in 3 cases. Graft manipulation was unknown in 6 cases. Only one HLA-mismatched graft was not manipulated. HLA-matched grafts were not manipulated (n = 5), selected for CD34⁺ cells (n = 3), or CD3/CD19-depleted (n = 2). Status was unknown in 5 cases.

GvHD prophylaxis in the HLA-mismatched settings mainly contained mycophenolate mofetil (MMF, n = 27) alone or in combination with either corticosteroid, tacrolimus, post-transplant CYC, or ATG. Also, cyclosporine A (CSA) and methotrexate (MTX) were applied in/as combination. GvHD prophylaxis in the HLA-matched settings mainly composed of CSA (n = 11) alone or in combination with either MMF, tacrolimus, corticosteroid, MTX, sirolimus, or OKT3.

HLA-mismatched patients (n = 35) were treated according to protocols as follows: CWS96 (n = 2), CWS2000P (n = 9), CWS-IV-2002 (n = 6), SoTiSaR (n = 17), and one non-CWS regimen. HLA-matched patients (n = 15) were treated according to the following protocols: CWS96 (n = 6), CWS2000P (n = 5), SoTiSaR (n = 2), MMT-98 (n = 1), and one non-CWS regimen. Transplanted patients utilized for matched-pair analysis (n = 29) received systemic therapies according to CWS96 (n = 2), CWS2000P (n = 9), CWS-IV-2002 (n = 3), and SoTiSaR (n = 15). Non-transplanted patients (n = 87), who served as controls in the matched-pair analysis, were treated according to protocols as follows: CWS96 (n = 11), CWS2000P (n = 29), CWS-IV-2002 (n = 9), and SoTiSaR (n = 38).

Within the CWS database, three matching patients per one transplanted patient were identified for the analysis. For 29 transplanted patients, who received primary allo-HSCT for their high-risk/progressive disease, matching non-transplanted pairs could be identified. Here, 22 transplanted patients received HLA-mismatched grafts and 7 patients received HLA-matched allografts. Matching criteria for further statistical analyses consisted of sex, age at diagnosis, favorable/unfavorable tumor site, and metastatic sites (locations). Primary systemic therapies for transplanted patients were either CEVAIE-based (n = 22) or VAIA-based (n = 7). CEVAIE is an intensive chemotherapy consisting of alternating courses of ifosfamide, vincristine, actinomycin-D (I³VA), carboplatin, epirubicin, vincristine (CVE) and ifosfamide, vincristine, and etoposide (I³VE, CWS guidance v1.6.1). VAIA is another chemotherapy regimen, currently used for the very high-risk group or metastatic stage in past protocols, implementing alternating cycles and combinations of vincristine, adriamycin, ifosfamide, and actinomycin-D. Primary systemic therapies for non-transplanted controls mainly included CEVAIE (n = 55) with O-TI/E (43/55) or were VAIA-based (n = 28) with O-TI/E (21/28), unknown in 4. The specific matching criteria and patients’ characteristics are summarized in Table 4 .

End point values were recorded upon the last patient follow-up. Last database update was done in June 2020. Statistical analyses were performed using R studio 1.3.1093 (Public-benefit Corp), R for macOS X Cocoa GUI 4.1.0 (The R foundation for Statistical Computing) and Prism for macOS 9.1.2 (GraphPad Software). Fisher’s exact, Chi-square test, or Chi-square test for trend was applied to compare categorical variables in between study and control groups (for baseline characteristics). In multivariate analysis subdistribution hazard ratios (HR), standard errors, and confidence intervals (CI) are presented when appropriate. Survival probabilities were calculated using Kaplan–Meier curves and compared with a log-rank (Mantel-Cox) test. Cumulative incidence function and multivariate analyses were applied according to Scrucca et al. (19, 20). Statistical significance was attributed to p-values < 0.05.

Primary engraftment was stated in 45/50 (0.90) of cases; the rest experienced secondary engraftment. Median time to engraftment was 12 days (HLA-mismatched group, short mismatched: 13, HLA-matched group, short matched: 12). Incidence of acute GvHD in all analyzed patients was 0.85 (mismatched: 0.87 and matched: 0.80). No acute GvHD was reported in 15% of all cases (mismatched: 0.13 and matched: 0.20). Combined incidence with grade I–II acute GvHD was 0.66 (mismatched: 0.73 and matched: 0.50) indicating absent, or little to moderate toxicity in most of the cases. Incidence of extended chronic GvHD mounted up to 0.11 for all patients (mismatched: 0.09 and matched: 0.16). Status was unknown or not assessed in 10/50 cases for acute GvHD and 6/34 cases for chronic GvHD. Incidence of both acute and chronic GvHD did not differ significantly in between groups. More detailed information is provided in Table 2 .

Next, time intervals from the date of 1st allo-HSCT until occurrence of relapse, DOD, or TRM were compared. At the end of data collection, a total of 9/50 patients (0.18) were alive (or died of reasons unrelated to the primary disease or transplantation, labeled as “other”), 31 patients (0.61) died of primary disease (DOD), and for 10 patients, TRM (0.20) was reported, all without statistically significant differences in between groups. A more detailed overview of disease outcomes is given in Table 2 . Mean follow-up time for patients grouped to alive/other was 45.8 months (mismatched: 24.5, matched: 62.8), 15.7 months for DOC (mismatched: 13.4, matched: 22.6), and 11.5 months for TRM (mismatched: 11.1, matched: 13.0), respectively. Although Oberlin risk factors were distributed without statistical significance in between HLA-mismatched and -matched patients, 0.79 of HLA-mismatched patients exhibited an Oberlin score ≥3, whereas only a fraction (0.46) in the HLA-matched group had a score ≥3 ( Table 1 , Figure 1A ). Here, the 3 longest survivors (OS in months per patient: 130, 78, and 77) either had an Oberlin score of 1, or were not metastasized at diagnosis and all three received allo-HSCT as relapse therapy. Additionally, 2/3 had favorable primary tumor localization, 2/3 were younger than 10 years, and 2/3 had VGPR/CR at the time of allo-HSCT. As for the HLA-mismatched group, there was only one patient with a comparable distribution of risk factors and response to induction therapy, who survived for 21 months before DOD. As a consequence, 2-year EFS was significantly higher in the HLA-matched vs. -mismatched group (p-value = 0.001), with a median EFS of 7 months for mismatched and 11 months for matched (p-value = 0.051; median OS—mismatched: 11 months and matched: 14 months, p-value = 0.08) ( Table 2 ). Figure 1B depicts survival curves according to Oberlin score. When analyzing the more homogeneous group of patients with primary allo-HSCT due to stage IV disease at diagnosis (n = 40), 2-year-EFS and median OS did not differ significantly (p-value = 0.46). Here, both groups showed a median EFS of 7 months and median OS for the mismatched group was 15 months versus 12 months for the matched group (p-value = 0.60). As most patients in the matched group received myeloablative conditioning (MAC, 0.86) and most HLA-mismatched cases received reduced-intensity conditioning (RIC, 0.91), we also precluded an effect due to different conditioning regimens (p-value = 0.45; Figure 1C ) in a consecutive analysis.

As TRM was relatively high, with a fraction of 0.20 in all analyzed cases, we furthermore performed a competing risk analysis by calculating the cumulative incidence function (CIF) for relapse, and separating it from TRM (as it will prevent emergence of later relapse) (19). Equality testing across groups revealed that there was no significant difference for relapse (p-value = 0.27) and TRM (p-value = 0.39). Differences in median OS were not observed even when excluding stage IV patients with TRM from the analysis ( Figures 2A, B ).

To confirm results of univariate analyses and to further identify additional factors decisive for survival, we performed multivariate regression modeling for competing risks data according to Scrucca et al. (20). Factors with possible influence on EFS and OS were analyzed, namely: age at diagnosis, age at allo-HSCT, Oberlin score (1–4, when applicable), disease status at allo-HSCT (VGPR/CR, PR, and PD), and graft type (HLA mismatch/match).

Multiple models including different combinations of these covariates were fitted and compared ( Supplementary Table 1 ). To select the best fit, the Bayesian information criteria (BIC) were applied. The ultimately selected model included disease status at allo-HSCT and Oberlin score. Only disease status correlated with outcome [p-values with respect to baseline (PD) for PR: p-value < 0.05 for EFS and OS; VGPR/CR: p-value < 0.001 for EFS and OS]. Therefore, CIF was calculated for relapse and death (DOD, TRM) depending on disease status at allo-HSCT (phase: VGPR/CR, PR, or PD, Figure 3 ). Here, disease status at allo-HSCT significantly correlated with outcomes. A summary of regression modeling for competing risks is provided in Table 3 .

To address the question whether allo-HSCT provides survival benefits compared to standard-of-care therapy, we performed a matched-pair analysis. For 29 primary transplanted patients, matched non-transplanted pairs (3 non-transplanted per 1 transplanted patient) from CWS SoTiSaR were identified and compared for OS and EFS. Matching criteria included sex, age, favorable/unfavorable tumor site, and metastatic site. Table 4 provides relevant patient characteristics according to matching criteria.

For 2 primary transplanted patients, only 2 instead of 3 matched pairs were included for survival analysis. For the 4 transplanted patients aged ≥21 years, 11/12 matched patients were younger than 21 years. Of 2 transplanted patients presented with favorable tumor site at diagnosis, only one favorable matched control was identified.

As distribution in metastatic sites indicated a trend of unequal distribution in between study and control group (tested with Chi-square test for trend, with a tendency towards worse constellation of metastatic sites in the control group), which might influence survival analyses, regression modeling for competing risks was conducted accounting for metastatic site. No significant influence on survival was found (Wald test p-value = 0.08) (20).

EFS probability for study and control groups did not differ significantly (p-value = 0.42) with a median EFS of 13 months for primary transplanted patients versus 11 months for controls ( Figure 4A ). Similar to EFS, probability to survive (i.e., OS) was not different (p-value = 0.67) with a median OS of 22 months for study group patients versus 18 months for controls ( Figure 4B ). As status at allo-HSCT was a condition for survival (CR or VGPR) in our previous analyses, we furthermore asked whether patients transplanted in CR/VGPR had a potential benefit from allo-HSCT compared to standard-of-care patients (controls). Indeed, respective patients show an increased survival probability for EFS (median 17 versus 10 months, p-value = 0.04). However, no differences were observed for OS (median 27 versus 17 months, p-value = 0.15, Figures 4C, D ), even when neglecting cases of TRM (p-value = 0.12, graph not shown) or only focusing on those who received HLA-mismatched grafts (Kaplan–Meier curve comparison for EFS—p-value = 0.063 and for OS—p-value=0.09, graph not shown).