Novel prognostic scoring system for hematopoietic stem cell transplantation in myelodysplastic syndrome and Acute Myeloid Leukemia, myelodysplasia-related

Abstract Background: Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is the only curative option for advanced MDS and AML-MR. Given disease heterogeneity, accurate risk stratification is vital for treatment strategy. Prognostic tools like IPSS-R/M guide pre-transplant decisions but are not optimized for transplant outcomes, especially in adverse-risk cases. We therefore aimed to develop a prognostic scoring system tailored to post-transplant outcomes in MDS and AML-MR. Methods: We retrospectively analyzed patients with MDS or AML-MR who underwent first single-unit CBT at our institution from 2010 to 2024, excluding MDS/MPN or MPN, secondary AML from these entities, or PS ≥3. Primary endpoint was relapse-free survival (RFS); secondary endpoints included overall survival (OS), cumulative incidence of relapse (CIR), and non-relapse mortality (NRM). Chromosomal abnormalities of ≥5% prevalence and clinical relevance were evaluated to build a new cytogenetic score, which was then combined with clinical variables in multivariate Cox models for RFS to create a novel prognostic system. Model viability was assessed by time-dependent ROC and Harrell's C-index, with ad hoc analyses for transplant outcomes. Results: A total of 402 patients were analyzed, with a median age of 62 (19–77); 70.9% were male, and median follow-up was 1,467 days. AML transformation was present in 69.9%, and 79.1% were in non-remission status (nonCR). Cytogenetic risk by IPSS-R was Good (29.6%), Intermediate (29.4%), Poor (9.7%), and Very Poor (30.8%), with monosomal and complex karyotypes (≥3) in 30.0% and 35.1%. Three-year RFS and OS were 36.4% and 38.6%; CIR and NRM were 20.0% and 43.5%. Multivariate analysis identified male sex, pre-2017 HSCT, and IPSS-R cytogenetic risk ≥Poor as independent predictors for RFS, with same results for OS. CIR was linked to platelet transfusion dependency, AML transformation, and adverse cytogenetics, while NRM related to pre-2017 HSCT and HCT-CI≥3. NRM decreased by 10.9% after 2018, leading to improved RFS. Common chromosomal lesions included -7 (20.3%), +8 (16.3%), del(5q) (11.3%), -5 (10%), -17 (10%), del(20q) (7%), and other monosomies (25.3%; defined as the existence of monosomies excluding -5, -7, -17). CIR increased stepwise with abnormality count (HR: 2.82, 3.89, 4.59, 6.79 for 1, 2, 3, and >3 vs. normal karyotype). Multivariate analysis yielded lesion-specific HRs, which were translated into weighted points (-7:3pt, -5:2pt, -17/del(5q)/other monosomies:1pt, +8:0pt, del(20q):-1pt). These were added to points assigned for abnormality count (1:1pt, 2:1.5pt, 3:2pt, >3:3pt) to form an additive cytogenetic score, which classifies patients as Good (≤0.5), Intermediate (1.0–1.5), Poor (2.0–3.5), Very Poor (4.0–5.5), or Worst (≥6). By IPSS-R cytogenetic risk, 5-year CIR were 6.1%, 16.9%, 33.0%, and 35.1% from Good to Very Poor, showing limited separation in high-risk groups. In contrast, our score stratified 5-year CIR as 5.6%, 14.9%, 23.9%, 34.9%, and 41.1% from Good to Worst, demonstrating clearer separation. Time-dependent ROC analysis confirmed its predictive validity, with 5-year AUC of 0.77 (vs. 0.74 for IPSS-R). In multivariate Cox analysis including this score, it independently predicted RFS (HR 1.20 per point, p<0.01) along with male sex (HR 1.38, p=0.03); age≥60, HCT-CI≥3, PS≥1, AML transformation, and nonCR showed adverse trends (HRs 1.24–1.54, p=0.05–0.18). We then developed a composite prognostic score (our new cytogenetic risk group: 0–4pts + 1pt each for age≥60, male sex, HCT-CI≥3, PS≥1, AML transformation, nonCR), which classifies patients as Good (≤2), Intermediate (3–4), Poor (5–7), or Worst (≥8) groups, comprising 22, 101, 196 and 80 patients, respectively. This new score significantly stratified transplant outcomes in ad hoc anlayses: 3-year RFS was 70.9%, 56.1%, 31.2%, and 16.5% (p<0.01), and OS showed similar trends (p<0.01). CIR remained low in Good/Intermediate (5.5%, 6.1%), but rose in Poor/Worst (20.5%, 41.3%; p<0.01). NRM increased from Good to Poor (23.6-48.3%, p<0.01) but was slightly lower in Worst (42.2%), reflecting early relapses. Harrell's C-index for RFS was 0.608, confirming the score's discriminatory ability. Conclusion: We established a novel cytogenetic score and integrated transplantation-specific prognostic model for MDS/AML-MR. This score provides a robust framework for risk stratification, facilitating individualized decisions regarding pre- and post-transplant treatment strategies.