Comprehensive analysis of prognostic factors in myelodysplastic syndromes with isolated deletion of chromosome 5q∗† Authors: Maria Julia Montoro, Víctor Navarro, Pamela Acha, Claudia Haferlach, Onyee Chan, Yasuo Kubota, Felicitas Schulz, Robert Briski, Najla H Al Ali, Blanca Xicoy, Felix Lopez-Cadenas, Francesc Bosch, Anna Aguilera Romero, Manja Meggendorfer, Lea Naomi Eder, Andrés Jerez, Yu-Hung Wang, Alessia Campagna, Luca Lanino, Beate Betz, Valeria Santini, Teresa Bernal, Esperanza Such, Anne Sophie Kubasch, Tariq Kewan, Carmelo Gurnari, Carla Zindel, Austin Kulasekararaj, Maria Teresa Voso, Mikkael Sekeres, Sangeetha Venugopal, Nicolas Díaz Varela, Aref Al-Kali, Uwe Platzbecker, Detlef Haase, Amer M. Zeidan, Pierre Fenaux, Maria Diez-Campelo, Guillermo Garcia-Manero, Daniel H Wiseman Published: 2026 ABSTRACT Key Points•IPSS-R and IPSS-M show limited prognostic discrimination in MDS-del(5q), despite most patients being classified into lower-risk groups.•A disease-specific model identifies high-risk patients and reclassifies a subset of lower-risk cases in MDS-del(5q). Introduction Myelodysplastic syndromes (MDS) characterized by the isolated deletion of chromosome 5q [MDS-del(5q)] constitute a unique entity within MDS. According to the 2017 World Health Organization classification, it is defined by dysplasia in 1 to 3 myeloid lineages, bone marrow (BM) blasts of <5%, circulating blasts of <1%, and del(5q), either in isolation or with 1 additional chromosomal abnormality, excluding monosomy 7 or 7q deletion.1 Patients with MDS-del(5q) have a median overall survival (OS) of 75 months, which is superior to that of the overall MDS population. Nevertheless, in ∼25% of patients, MDS evolve to acute myeloid leukemia (AML) within 5 years.2, 3, 4, 5 Risk factors associated with AML evolution in patients with MDS-del(5q) include transfusion dependency, low platelet counts, and an increased percentage of BM blasts.2,6 In addition, mutations in SF3B1, TP53, and CSNK1A1 genes further refine the prognosis of these patients.7, 8, 9, 10, 11, 12 Existing prognostic scoring systems for MDS, the International Prognostic Scoring System (IPSS), both the Revised IPSS (IPSS-R) and the Molecular IPSS (IPSS-M),13, 14, 15 were developed based on the analysis of all MDS subtypes. The broader MDS population includes substantial heterogeneity in blast percentages (range, 0%-19%), cytogenetic profiles (normal to complex karyotype), and variable molecular signatures. Importantly, many high-risk features present in these prognostic models are not applicable to MDS-del(5q), given the stringent diagnostic criteria defining this distinct MDS subtype.1,13, 14, 15 Against this background, we took advantage of the largest cohort of patients with MDS-del(5q) compiled to date to (1) assess the predictive performance of the IPSS-R and IPSS-M, (2) comprehensively evaluate the prognostic impact of clinical and biological variables, and (3) develop a risk stratification model specifically tailored to patients with MDS-del(5q). Methods Patients Patients consecutively diagnosed with de novo MDS-del(5q) according to the World Health Organization 2017 classification, comprising cases with TP53-multihit abnormalities1 from 1997 to 2022, were included. Therapy-related or those under disease-modifying therapy before the genetic analysis were excluded. Diagnosis and pathological evaluation were performed at each participating local center. Clinical data Clinical characteristics and laboratory parameters were recorded at diagnosis. Risk stratification was performed according to the IPSS-R and IPSS-M. For clinical simplicity, patients with an IPSS-R score of ≤3.5 were classified as lower risk, whereas those with a score of >3.5 were classified as higher risk. Similarly, patients in the very-low, low, and moderate-low IPSS-M categories were grouped as lower risk, whereas those in the moderate-high, high, and very-high IPSS-M categories were classified as higher risk. Clinical outcomes included OS and AML-free survival, assessed through the review of medical records. Genetic data Cytogenetic and molecular information were gathered from all patients at diagnosis or before treatment initiation. Genetic profiling was performed at each institution and included conventional G-banding analyses in all cases. The break points of the del(5q) were categorized as type 1 if they corresponded to the most frequently observed break points (q13q31, q13q33, q22q33, q12q33, and q14q34) or as type 2 for all other breakpoints, as previously defined.13 TP53 mutations were assessed in all patients. Next-generation sequencing (NGS) mutation panel analysis was performed in 626 (91.7%) cases, including targeted NGS, whole-exome sequencing, and whole-genome sequencing (supplemental Table 1). A set of 27 genes recurrently mutated in MDS were consistently assessed across all NGS platforms, including ASXL1, CALR, CBL, CEBPA, CSF3R, DNMT3A, ETV6, EZH2, FLT3, IDH1, IDH2, JAK2, KIT, KRAS, MPL, NPM1, NRAS, PTPN11, RUNX1, SETBP1, SF3B1, SRSF2, TET2, TP53, U2AF1, WT1, and ZRSR2. In addition, CSNK1A1 mutations were analyzed in 435 (63.8%) patients (supplemental Figure 1). Moreover, patients were stratified by TP53 status in non−high-risk TP53 (nonHR-TP53), encompassing cases with TP53 wild type or TP53 monoallelic mutations with a variant allele frequency (VAF) of <20%, and high-risk TP53 (HR-TP53), including MDS-del(5q) cases with TP53 multihit alterations or monoallelic mutations with a VAF of ≥20%, as previously described.8 Patient information was collected in accordance with institutional review board–approved protocols at each participating center and in compliance with the Declaration of Helsinki. Statistical analysis Clinical characteristics were summarized descriptively, and group comparisons were assessed using logistic regression models. OS, defined as time from diagnosis to last follow-up or death from any cause, and leukemia-free survival (LFS), defined as time from diagnosis to disease progression to AML or death, were estimated using the Kaplan-Meier method and compared using Cox proportional hazards models. The cumulative incidence of leukemic progression, defined as time from diagnosis to disease progression or death, was estimated by the Fine-Gray method with nonleukemic death as a competing event. Patients were categorized as progression to AML (the event of interest), death without previous progression (the competing event), or censored in the absence of these events. Variables included in the univariable analysis were sex, age (<70 years vs ≥70 years), hemoglobin (Hb; ≤10 g/dL vs >10 g/dL), leukocytes (≤4.5 × 109/L vs >4.5 × 109/L), neutrophils (≤1.8 × 109/L vs >1.8 × 109/L), platelets (≤100 × 109/L vs >100 × 109/L), BM blasts (≤2% vs >2%), karyotype [isolated del(5q) vs del(5q) plus 1 additional abnormality], break points (type 1 vs type 2), genes mutated in ≥5% of patients (TP53, SF3B1, DNMT3A, TET2, ASXL1, JAK2, and CSNK1A1), TP53 status (nonHR-TP53 vs HR-TP53), and total number of mutations (<2 mutations vs ≥2 mutations), considering only the 27 genes analyzed among all patients. Variables with a P value of <.1 in the univariable analysis were considered in the multivariable models. All statistical tests were 2-sided, and a P value of <.05 was considered statistically significant in the multivariable models. For the development of the IPSS for MDS-del(5q) [IPSS-del(5q)], all candidate predictors (Hb, platelets, BM blasts, and age at diagnosis) were initially analyzed as continuous variables and selected by least absolute shrinkage and selection operator regression, with LFS as the primary end point. Progression to AML was analyzed using Fine-Gray subdistribution hazard models with nonleukemic death as a competing event, and OS was assessed separately. Proportional subdistribution hazards were evaluated using Schoenfeld-type residuals adapted for Fine-Gray models. Because continuous modeling, including restricted cubic splines, did not meaningfully improve predictive performance, variables were ultimately categorized using clinically relevant cutoffs optimized for outcome discrimination to enhance usability in routine practice. The final least absolute shrinkage and selection operator–Cox model was developed using complete-case data. Risk score points were assigned by rounding variable coefficients. Model discrimination was evaluated using the concordance index (C-index), including for the Fine-Gray model, using the QHScrnomo R package. Internal validation was performed with 500 bootstrap iterations to assess model variability. Missing data were not imputed. Analyses were conducted in R version 4.2.2. Ethics approval statement This study was conducted in accordance with the Declaration of Helsinki and approved by the appropriate institutional review boards. Results Patient population A total of 682 patients with MDS-del(5q) from 19 international institutions were included. The median follow-up for the entire cohort was 68.5 months (95% confidence interval [CI], 62.3-81.1). Data on AML progression were available for 507 (74.3%) patients, and survival data were available for 539 (79.0%) patients. The median OS for the entire cohort was 73.8 months (95% CI, 63.2-83.4). At 60 months, MDS-del(5q) had progressed to AML in 115 of 507 (22.7%) patients, with a median time to progression of 32 months (interquartile range, 18-57). The baseline clinical characteristics of the cohort are summarized in Table 1. Type 1 break points of del(5q) were present in 513 (79.7%) patients, whereas type 2 break points were observed in 131 (20.3%) patients. The molecular landscape was assessed by NGS in 626 (91.8%) patients, with 375 (59.9%) analyzed by targeted NGS, 240 (38.3%) by whole-genome sequencing, and 11 (1.8%) by whole-exome sequencing.Table 1.Main clinical characteristics of the study cohortMedian (IQR)Age at diagnosis, y74 (66-80)Sex, female, %73.3Hb, g/dL9.2 (8.1-10.3)WBC count, ×109/L4.1 (3.0-5.5)Neutrophils, ×109/L2.0 (1.3-3.1)Platelet count, ×109/L249.0 (169.0-347.0)BM blasts, %2 (1.0-4.0)Cytogenetics, n (%)Isolated del(5q)568 (83.3)+1 additional abnormality114 (16.7)Breakpoints of del(5q), n (%) Type 1513 (79.7) Type 2131 (20.3)IPSS-R, n (%) Very low111 (18.2) Low390 (64.0) Intermediate108 (17.8)IPSS-M, n (%) Very low66 (11.8) Low312 (55.8) Moderate low100 (17.9) Moderate high43 (7.7) High31 (5.5) Very high7 (1.3)IQR, interquartile range; WBC, white blood cell. Table 1.Main clinical characteristics of the study cohortMedian (IQR)Age at diagnosis, y74 (66-80)Sex, female, %73.3Hb, g/dL9.2 (8.1-10.3)WBC count, ×109/L4.1 (3.0-5.5)Neutrophils, ×109/L2.0 (1.3-3.1)Platelet count, ×109/L249.0 (169.0-347.0)BM blasts, %2 (1.0-4.0)Cytogenetics, n (%)Isolated del(5q)568 (83.3)+1 additional abnormality114 (16.7)Breakpoints of del(5q), n (%) Type 1513 (79.7) Type 2131 (20.3)IPSS-R, n (%) Very low111 (18.2) Low390 (64.0) Intermediate108 (17.8)IPSS-M, n (%) Very low66 (11.8) Low312 (55.8) Moderate low100 (17.9) Moderate high43 (7.7) High31 (5.5) Very high7 (1.3)IQR, interquartile range; WBC, white blood cell. Main clinical characteristics of the study cohort IQR, interquartile range; WBC, white blood cell. Among the subset of 27 genes commonly assessed across all NGS analyses, the most frequently mutated genes were TP53 (20%), SF3B1 (19%), DNMT3A (17%), TET2 (13%), ASXL1 (9%), and JAK2 (7%). Beyond these 27 genes, only CSNK1A1 was mutated in ≥5% of patients, occurring in 31 of 435 (7%) patients (supplemental Figure 1). A total of 601 (88.4%) patients with MDS-del(5q) were classified as nonHR-TP53, whereas 79 (11.6%) were classified as HR-TP53. Treatment data were available for 439 (64.4%) patients. Among them, 254 (57.9%) had received lenalidomide at some point, with 186 (73.2%) achieving either a partial or complete response. Sensitivity analysis showed that patients with missing AML progression or treatment data had more adverse clinical features, although the molecular profile, including TP53, SF3B1, DNMT3A, and TET2 mutations, was similar between groups (supplemental Table 2). Performance assessment of MDS prognostic scores in patients with MDS-del(5q) IPSS-R Variables required to calculate IPSS-R score were available in 609 (89.3%) patients. In the remaining 73 (10.7%), complete blood count data were missing, precluding IPSS-R calculation. Among patients with available data, 575 (94.4%) were classified in the lower-risk group, with 111 (19.3%) in the very-low, 390 (67.8%) in the low, and 74 (12.9%) in the intermediate IPSS-R (≤3.5) subgroup. The remaining 34 (5.6%) patients were classified in the higher-risk group (intermediate IPSS-R [>3.5]). Notably, no patients were classified in the high and very–high-risk IPSS-R categories. Comparative outcome analysis between IPSS-R subgroups showed that patients classified in the lower-risk IPSS-R category exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-R category, with median OS of 76.9 months (95% CI, 62.9-85.4) vs 49.7 months (95% CI, 27.0-93.5; P = .002), LFS of 60.2 months (95% CI, 50.7-73.2) vs 31.2 months (95% CI, 26.1-31.4; P < .001), and risk of progression to AML at 60 months of 20% vs 39% (P = .002) in the lower- and higher-risk groups, respectively. However, the IPSS-R demonstrated limited predictive value for OS, LFS, and progression to AML, with C-indices of 0.53, 0.53, and 0.44, respectively (Figure 1A-C).Figure 1.Patient outcomes according to lower-risk IPSS-R (≤3.5 points) and higher-risk IPSS-R (>3.5 points). (A-C) Patients classified in the lower-risk IPSS-R category exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-R category: median OS of 76.9 months (95% CI, 62.9-85.4) vs 49.7 months (95% CI, 27.0-93.5; P = .002) (A); LFS of 60.2 months (95% CI, 50.7-73.2) vs 31.2 months (95% CI, 26.1-31.4; P < .001) (B); and risk of progression to AML at 60 months of 20% vs 39% (P = .002) (C) in the lower and higher-risk groups, respectively. However, the IPSS-R demonstrated limited predictive value for OS, LFS, and progression to AML, with C-indices of 0.53, 0.53, and 0.44, respectively. Patient outcomes according to lower-risk IPSS-R (≤3.5 points) and higher-risk IPSS-R (>3.5 points). (A-C) Patients classified in the lower-risk IPSS-R category exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-R category: median OS of 76.9 months (95% CI, 62.9-85.4) vs 49.7 months (95% CI, 27.0-93.5; P = .002) (A); LFS of 60.2 months (95% CI, 50.7-73.2) vs 31.2 months (95% CI, 26.1-31.4; P < .001) (B); and risk of progression to AML at 60 months of 20% vs 39% (P = .002) (C) in the lower and higher-risk groups, respectively. However, the IPSS-R demonstrated limited predictive value for OS, LFS, and progression to AML, with C-indices of 0.53, 0.53, and 0.44, respectively. IPSS-M Data required to calculate IPSS-M were available in 559 (82.0%) patients, whereas 123 (18.0%) lacked complete blood count or molecular data, precluding IPSS-M assessment. Among patients with available data, 478 (85.5%) were classified in the lower-risk IPSS-M category, with 66 (13.8%) in the very–low-risk, 312 (65.3%) in the low-risk, and 100 (20.9%) in the moderate–low-risk subgroup. The remaining 81 (14.5%) patients were classified in the higher-risk category, with 43 (53.1%) in the moderate–high-risk group, 31 (38.3%) in the high-risk, and 7 (8.6%) in the very–high-risk subgroup. Patients classified in the lower-risk IPSS-M category exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-M category, with OS of 81.0 months (95% CI, 63.1-89.5) vs 47.9 months (95% CI, 32.0-85.5; P = .008), LFS of 62.9 months (95% CI, 54.5-79.1) vs 32.0 months (95% CI, 26.5-57.0; P < .001), and risk of progression to AML at 60 months of 18% vs 44% (P < .001) in the lower- and higher-risk IPSS-M categories, respectively. Again, the predictive capacity of the IPSS-M for OS, LFS, and progression to AML end points remains limited and shows barely improved IPSS-R results as indicated by C-indices of 0.54, 0.55, and 0.51, respectively (Figure 2A-C).Figure 2.Patient outcomes according to lower-risk IPSS-M (very-low, low and intermediate-low IPSS-M) and higher-risk IPSS-M (intermediate-high, high and very-high). (A-C) Patients classified as lower-risk IPSS-M exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-M category: OS of 81.0 months (95% CI, 63.1-89.5) vs 47.9 months (95% CI, 32.0-85.5; P = .008) (A); LFS of 62.9 months (95% CI, 54.5-79.1) vs 32.0 months (95% CI, 26.5-57.0; P < .001) (B); and risk of progression to AML at 60 months of 18% vs 44% (P < .001) (C) in the lower and higher-risk groups, respectively. However, the predictive capacity of the IPSS-M for OS, LFS, and progression to AML end points remains limited, with C-indices of 0.54, 0.55, and 0.51, respectively. Patient outcomes according to lower-risk IPSS-M (very-low, low and intermediate-low IPSS-M) and higher-risk IPSS-M (intermediate-high, high and very-high). (A-C) Patients classified as lower-risk IPSS-M exhibited significantly favorable outcomes compared with those in the higher-risk IPSS-M category: OS of 81.0 months (95% CI, 63.1-89.5) vs 47.9 months (95% CI, 32.0-85.5; P = .008) (A); LFS of 62.9 months (95% CI, 54.5-79.1) vs 32.0 months (95% CI, 26.5-57.0; P < .001) (B); and risk of progression to AML at 60 months of 18% vs 44% (P < .001) (C) in the lower and higher-risk groups, respectively. However, the predictive capacity of the IPSS-M for OS, LFS, and progression to AML end points remains limited, with C-indices of 0.54, 0.55, and 0.51, respectively. Impact of genetic and clinical factors on outcomes in patients with MDS-del(5q) Genetic complexity We first evaluated the impact of genetic complexity, defined by the presence of additional chromosomal abnormalities and the total number of gene mutations. Although the presence of 1 additional chromosomal abnormality had a negative impact on OS (59.5 months [95% CI, 42.7-91.1] vs 76.4 months [95% CI, 66.5-85.1]; P = .03), no significant differences in outcomes were observed among patients based on the specific type of additional chromosomal abnormality (as per definition, chromosome 7 alterations were not included; supplemental Figure 2). Consistent with previous observations in MDS,14,15 the presence of ≥2 mutations in patients with MDS-del(5q) was associated with significantly worse outcomes compared with those with <2 mutations or no mutations. Specifically, OS was 60.5 months (95% CI, 64.8-76.8) and 86.0 months (95% CI, 73.7-99.3; P < .01), LFS was 44.7 months (95% CI, 34.3-63.9) and 61.4 months (95% CI, 56.2-81.2; P < .01), and the 60-month cumulative incidence of AML progression was 29% and 20% (P = .05) for patients with ≥2 vs <2 mutations, respectively (supplemental Figure 3A-C). Interestingly, ≥50% patients with TP53 and SF3B1 mutations, the most frequently mutated genes in our cohort, exhibited additional mutations. Among patients with TP53 mutations, 80 of 125 (64%) harbored additional mutations (1 mutation in 52.5%, 2 in 30.0%, and ≥3 in 17.5%). Similarly, 63 of 118 (53.4%) patients with SF3B1 mutations displayed extra mutations (1 in 54.0%, 2 in 27.0%, and ≥3 in 19.0%; supplemental Figure 4). However, no significant differences in outcomes were observed between patients with TP53 mutations alone and those with TP53 plus additional gene mutations (excluding TP53 multihit abnormalities) or between patients with SF3B1 mutations alone and those with SF3B1 plus additional gene mutations (supplemental Figures 5 and 6). Prognostic factors for OS In the univariable analysis for OS, CSNK1A1 mutations (hazard ratio [HR], 0.5) were associated with better outcomes. In contrast, age of ≥70 years (HR, 2.1), male sex (HR, 1.4), Hb of ≤10 g/dL (HR, 1.7), platelet count of ≤100 × 109/L (HR, 1.4), the presence of 1 additional chromosomal abnormality (HR, 1.4), TET2 mutations (HR, 1.4), the presence of ≥2 mutations (HR, 1.5), and HR-TP53 status (HR, 1.5), showed worse survival. In the multivariable analysis, age of ≥70 years (95% CI, 1.6-3.0; HR, 2.1), male sex (95% CI, 1.0-1.9; HR, 1.4), Hb of ≤10 g/dL (95% CI, 1.5-1.9; HR, 1.4), platelet count of ≤100 × 109/L (95% CI, 1.0-2.3; HR, 1.6), the presence of 1 additional chromosomal abnormality (95% CI, 1.1-2.2; HR, 1.5), and the presence of ≥2 mutations (95% CI, 1.2-3.4; HR, 2.0), were independently associated with a significantly worse prognosis (Table 2).Table 2.Multivariate analysis for OS, LFS, and risk of progression to AML of patients with MDS-del(5q)VariablesOSLFSProgression to AMLHR (95% CI)PHR (95% CI)PHR (95% CI)PAge, y≥70<702.1 (1.5-2.9)< .0012.1 (1.4-2.7)< .0010.60 (0.4-0.9).01GenderMaleFemale1.4 (1.1-1.9).041.4 (1.0-2.0).031.0 (0.6-1.5).9BM blasts %>2≤2————1.9 (1.2-2.8).003Hb, g/dL≤10>101.4 (1.5-1.9).021.4 (1.1-2.0)——Platelet count, ×109/L≤10>101.6 (1.0-2.3).021.9 (1.3-2.7)2.5 (1.4-4.3).0015q break pointsType 1Type 2————1.4 (0.9-2.3).1Cytogenetics+1 chromosomal abnormalitiesIsolated del(5q)1.5 (1.1-2.2).021.4 (0.9-2.0).1——Number of mutations≥2<22.0 (1.2-3.4).0071.7 (1.2-2.5).005——TET2MutatedWild type0.9 (0.6-1.5).81.4 (0.7-1.6).9——TP53HR-TP53nonHR-TP531.3 (0.9-1.8).11.4 (1.0-2.0).0091.8 (1.1-3.0).03SF3B1MutatedWild type1.0 (0.7-1.5).90.9 (0.6-1.4).71.7 (1.1-2.8).03 Table 2.Multivariate analysis for OS, LFS, and risk of progression to AML of patients with MDS-del(5q)VariablesOSLFSProgression to AMLHR (95% CI)PHR (95% CI)PHR (95% CI)PAge, y≥70<702.1 (1.5-2.9)< .0012.1 (1.4-2.7)< .0010.60 (0.4-0.9).01GenderMaleFemale1.4 (1.1-1.9).041.4 (1.0-2.0).031.0 (0.6-1.5).9BM blasts %>2≤2————1.9 (1.2-2.8).003Hb, g/dL≤10>101.4 (1.5-1.9).021.4 (1.1-2.0)——Platelet count, ×109/L≤10>101.6 (1.0-2.3).021.9 (1.3-2.7)2.5 (1.4-4.3).0015q break pointsType 1Type 2————1.4 (0.9-2.3).1Cytogenetics+1 chromosomal abnormalitiesIsolated del(5q)1.5 (1.1-2.2).021.4 (0.9-2.0).1——Number of mutations≥2<22.0 (1.2-3.4).0071.7 (1.2-2.5).005——TET2MutatedWild type0.9 (0.6-1.5).81.4 (0.7-1.6).9——TP53HR-TP53nonHR-TP531.3 (0.9-1.8).11.4 (1.0-2.0).0091.8 (1.1-3.0).03SF3B1MutatedWild type1.0 (0.7-1.5).90.9 (0.6-1.4).71.7 (1.1-2.8).03 Multivariate analysis for OS, LFS, and risk of progression to AML of patients with MDS-del(5q) Prognostic factors for LFS In the univariable analysis for LFS, the following factors were associated with worse outcomes: age of ≥70 years (HR, 1.5), male sex (HR, 1.4), Hb of ≤10 g/dL (HR, 1.6), platelet count of ≤100 × 109/L (HR, 1.8), BM blasts of >2% (HR, 1.3), SF3B1 mutations (HR, 1.5), HR-TP53 status (HR, 1.3), and the presence of ≥2 mutations (HR, 2.0). In the multivariable analysis, age of ≥70 years (95% CI, 1.4-2.7; HR, 2.0), male sex (95% CI, 1.0-2.0; HR, 1.4), Hb of ≤10 g/dL (95% CI, 1.1-2.0; HR, 1.4), platelet count of ≤100 × 109/L (95% CI, 1.3-2.7; HR, 1.9), and the presence of ≥2 additional mutations (95% CI, 1.2-2.5; HR, 1.7) were independently associated with a significantly shorter LFS (Table 2). Prognostic factors for evolution to AML In the univariable analysis for AML evolution, patients with age of ≥70 years (HR, 0.6) presented lower risk of AML progression, whereas male sex (HR, 1.4), platelet count of >100 × 109/L (HR, 2.5), BM blasts of >2% (HR, 1.7), type 2 del(5q) break points (HR, 1.5), SF3B1 mutations (HR, 1.7), and HR-TP53 status (HR, 2.0) showed higher risk of AML progression. In the multivariable analysis, age of <70 years (95% CI, 1.1-2.5; HR, 1.7), platelet count of >100 × 109/L (95% CI, 1.4-4.3; HR, 2.5), BM blasts of >2% (95% CI, 1.2-2.8; HR, 1.9), SF3B1 mutations (95% CI, 1.1-2.8; HR, 1.7), and HR-TP53 status (95% CI, 1.1-2.9; HR, 1.8) were independently associated with a significantly higher AML progression risk (Table 2). Specific risk score for patients with MDS-del(5q) Finally, given the limitations of existing MDS prognostic scores in MDS-del(5q), we developed a disease-specific risk model, the IPSS-del(5q). This model includes 6 prognostic variables: sex (male), Hb (≤10 g/dL), platelets (≤100 × 109/L), ≥2 mutations, SF3B1 mutations, and HR-TP53 status (Table 3; supplemental Table 3), that stratified patients into 2 groups, standard-risk (≤3 points; n = 320 [74.3%]) and high-risk (>3 points; n = 111 [25.7%]) groups. The median LFS was 69.2 months (95% CI, 60.9-87.3) in the standard-risk group and 32.0 months (95% CI, 26.2-42.6) in the high-risk group (95% CI, 0.34-0.58; HR, 0.44; P < .01; C-index = 0.60; Figure 3; supplemental Table 4). The IPSS-del(5q) also showed significant differences in OS, with a median OS of 86.1 months (95% CI, 79.1-92.3) in the standard-risk group and 43.4 months (95% CI, 33.4-60.5) in the high-risk group (supplemental Figure 7A), and in the risk of AML progression (17% for the standard-risk group and 39% for the high-risk group at 60 months; supplemental Figure 7B). As expected, patients classified within the high-risk IPSS-del(5q) category exhibited a more adverse molecular profile with significantly higher frequencies of mutations in SF3B1 (48.5% vs 4.6%), TP53 (50.7% vs 13.6%), and TET2 (33.7% vs 6.6%) compared with those in the standard-risk group (supplemental Figures 8 and 9). Although treatment data were not included in the model, sensitivity analyses stratified by lenalidomide exposure and treatment response confirmed that the predictive performance of score remained robust across all the subsets (supplemental Figures 10 and 11).Table 3.IPSS for patients with MDS-del(5q) [IPSS-del(5q)]VariablePointsSex, male1Hb ≤10 g/dL2Platelet count of ≤100 × 109/L2≥2 additional mutations2SF3B1 mutations1HR-TP53 status1Using least absolute shrinkage and selection operator–Cox regression analysis with LFS as the event of interest, 6 prognostic variables were identified. Based on these variables, patients were stratified into 2 risk groups: standard-risk IPSS-del(5q) (≤3 points; n = 320 [74.3%]) and high-risk IPSS-del(5q) (>3 points; n = 111 [25.7%]).Figure 3.LFS according to IPSS-del(5q). According to the IPSS-del(5q), patients were stratified into 2 risk groups: standard risk (≤3 points; n = 320 [74.3%]) and high risk (>3 points; n = 111 [25.7%]). The median LFS was 69.2 months (95% CI, 60.9-87.3) in the standard-risk group and 32.0 months (95% CI, 26.2-42.6) in the high-risk group (HR, 0.44; 95% CI, 0.34-0.58; P < .01; C-index = 0.60). Table 3.IPSS for patients with MDS-del(5q) [IPSS-del(5q)]VariablePointsSex, male1Hb ≤10 g/dL2Platelet count of ≤100 × 109/L2≥2 additional mutations2SF3B1 mutations1HR-TP53 status1Using least absolute shrinkage and selection operator–Cox regression analysis with LFS as the event of interest, 6 prognostic variables were identified. Based on these variables, patients were stratified into 2 risk groups: standard-risk IPSS-del(5q) (≤3 points; n = 320 [74.3%]) and high-risk IPSS-del(5q) (>3 points; n = 111 [25.7%]). IPSS for patients with MDS-del(5q) [IPSS-del(5q)] Using least absolute shrinkage and selection operator–Cox regression analysis with LFS as the event of interest, 6 prognostic variables were identified. Based on these variables, patients were stratified into 2 risk groups: standard-risk IPSS-del(5q) (≤3 points; n = 320 [74.3%]) and high-risk IPSS-del(5q) (>3 points; n = 111 [25.7%]). LFS according to IPSS-del(5q). According to the IPSS-del(5q), patients were stratified into 2 risk groups: standard risk (≤3 points; n = 320 [74.3%]) and high risk (>3 points; n = 111 [25.7%]). The median LFS was 69.2 months (95% CI, 60.9-87.3) in the standard-risk group and 32.0 months (95% CI, 26.2-42.6) in the high-risk group (HR, 0.44; 95% CI, 0.34-0.58; P < .01; C-index = 0.60). Notably, 19.1% of patients classified as lower risk by the IPSS-R were reclassified as high risk by the IPSS-del(5q), whereas 2.8% of patients classified as higher risk by the IPSS-R were reclassified as low risk by the IPSS-del(5q). Similarly, 14.6% of patients classified as lower risk by the IPSS-M were reclassified as high risk by the IPSS-del(5q), whereas 3.8% of patients classified as higher risk by the IPSS-M were reclassified as low risk by the IPSS-del(5q). Importantly, patients upstaged were enriched for both adverse clinical and molecular features (supplemental Table 5). Given the absence of an adequately large independent validation cohort, 500 bootstrap iterations were performed, demonstrating stable C-index values and robust reproducibility of the IPSS-del(5q) model (supplemental Figure 12). Discussion To our knowledge, in this largest reported MDS-del(5q) cohort, both the IPSS-R and IPSS-M showed limited prognostic accuracy, whereas the disease-specific IPSS-del(5q) improved risk stratification but remained suboptimal, highlighting the need for more refined prognostic approaches. Patients with MDS-del(5q) are typically considered to have a favorable prognosis.16 However, several studies have shown that 25% of these patients eventually progress to AML,2, 3, 4 a notion corroborated in our cohort. Although AML transformation occurs later in patients with higher-risk MDS, it is noteworthy that, at 5 years, the cumulative incidence of AML is comparable with that observed in patients with MDS with multilineage dysplasia and approaches the risk seen in MDS with excess blasts-1 and those classified as intermediate risk by the IPSS-R.17 These data challenge the established perception of “favorable” prognosis in MDS-del(5q) and highlight the need to identify specific factors that contribute to poorer outcomes. Although MDS-del(5q) is recognized as a unique clinicopathological entity,1,16 prognostic risk in these patients has historically been assessed using models developed for the broader MDS population,13, 14, 15 which may fail to capture high-risk features specific to this subgroup. Notably, the original IPSS-M cohort included only 142 patients with MDS-del(5q) of 2957 MDS cases.15 In our series, most patients were classified as lower risk according to the IPSS-M and IPSS-R classifications, whereas <15% fell into the higher-risk categories by these models. This observation is consistent with the findings of Giagounidis et al and Germing et al, who reported that, when applying the original IPSS, nearly all patients with MDS-del(5q) (92% and 100%, respectively) were categorized as low or intermediate-1 risk. However, the IPSS did not adequately discriminate clinically meaningful risk subgroups in their cohorts.2,6 Similarly, in our cohort, although the IPSS-R and IPSS-M formally stratified patients into lower- and higher-risk categories, their predictive performance was notably poor, with C-index values at ∼0.5. To address the need for a more refined prognostication, to our knowledge, we assembled the largest cohort of patients with MDS-del(5q) and identified the variables with prognostic impact in their outcomes. We observed that older age (≥70 years) was associated with poorer OS and LFS, whereas younger age correlated with a higher risk of progression to AML. These findings likely reflect the influence of nonleukemic mortality driven by age-related comorbidities on survival outcomes in older patients. Similarly, male sex was associated with worse survival rates, further underscoring the influence of age-related health factors, because males tend to have higher prevalence and death rates of cancer and heart diseases.18 Notably, the adverse impact of older age and male sex was also observed in another large MDS-del(5q) cohort assembled before this study.2 Cytopenias were also associated with worse outcomes in patients with MDS-del(5q). Anemia correlated with inferior survival, consistent with previous studies, linking transfusion dependency to poor prognosis,2 whereas thrombocytopenia was associated with reduced survival and an increased risk of progression to AML, in line with findings from 2 independent studies.2,4 In our series, increased genetic complexity was associated with worse outcomes. The prognostic significance of a single additional chromosomal abnormality in patients with MDS-del(5q) remains controversial. Although studies by Mallo et al and Kaivers et al found no statistically significant difference in survival between patients with isolated del(5q) and those with 1 additional abnormality,4,17 other studies reported significantly worse outcomes.2,19 In our study, a single additional chromosomal abnormality correlated with significantly worse survival, regardless of the specific accompanying cytogenetic alteration (chromosome 7 abnormalities were not included by definition). In addition, a higher mutational burden (≥2 mutations) had a consistently negative impact across all survival end points, highlighting that the presence of a more complex molecular profile outweighs, in part, the favorable prognosis associated with the single del(5q). This observation aligns with numerous findings in the general MDS population20,21 and reflects the link between increased genetic complexity and a more advanced, biologically aggressive disease phenotype. Interestingly, progression to AML was associated with HR-TP53 status and SF3B1 mutations. The association between TP53 multihit abnormalities and large TP53 clones (VAF, >20%) with an increased risk of AML evolution in patients with MDS-del(5q) was previously reported by others and supported by our results.8,10 The association between SF3B1 mutations and poorer prognosis in patients with MDS-del(5q) has also been reported in previous studies.9,15 This finding is particularly intriguing, given that SF3B1 mutations are generally associated with favorable outcomes in patients with MDS.22 Importantly, the negative prognostic impact of SF3B1 and TP53 mutations in patients with MDS-del(5q) appears to be independent of co-occurring mutations, suggesting an intrinsic biological effect. In line with this, recent MDS classifications23, 24, 25 have emphasized the increasing relevance of molecularly defined MDS entities, particularly those with TP53 multihit alterations or SF3B1 mutations. However, our recent data suggest that patients with MDS-del(5q) and TP53 multihit alterations have significantly better outcomes than patients with low-blast TP53-mutated MDS lacking isolated del(5q).26 Similarly, Komrokji et al reported that although SF3B1-mutated MDS-del(5q) was associated with inferior survival compared with isolated MDS-del(5q) or MDS-SF3B1, lenalidomide responses remained similar to those with isolated MDS-del(5q), whereas responses to luspatercept were inferior to classical MDS-SF3B1. Together, these findings suggest that co-occurring molecular lesions in MDS-del(5q) function mainly as prognostic modifiers rather than disease-defining features. Accordingly, the IPSS-del(5q) was developed as a complementary tool to refine risk stratification across the biological spectrum of MDS-del(5q) in contemporary practice, including cases with isolated del(5q) and TP53 multihit. Interestingly, although this score outperforms existing models, identifies a meaningful subset of patients at high risk, and reclassifies nearly one-fifth of patients labeled as lower-risk by the IPSS-R and IPSS-M, its overall predictive accuracy remains limited (C-index = 0.60). This may be attributable to the stringent diagnostic criteria of the entity, which require a restricted blast percentage and a uniform cytogenetic profile as well as the low genetic complexity typically observed in MDS-del(5q) (median of 1 mutation per patient). In such a homogeneous context, identifying additional high-risk features with strong prognostic impact is inherently challenging, thereby limiting further refinement of prognostic stratification. Future improvements may arise from the integration of dynamic clinical variables, such as longitudinal transfusion dependency, response to erythroid stimulating agents, or molecular residual disease monitoring during lenalidomide treatment, together with additional genetic features, including a more comprehensive characterization of the molecular architecture at diagnosis and identification of emerging subclones over time. Notably, although lenalidomide treatment was not incorporated into the model and may have attenuated the observed risk differences, the IPSS-del(5q) retained its prognostic value across treatment subgroups. The retrospective design represents the main limitation of this study, particularly regarding the availability of treatment and AML progression data for all patients. Variability in treatment strategies and response assessment across participating countries further complicates standardized data collection. The absence of an independent validation cohort is another important limitation. Nevertheless, the large multicenter cohort, the comparable molecular profiles between patients with and without missing data, and the use of computer-generated simulated cohorts support the robustness of our findings and mitigate potential bias. However, validation in large independent data sets will be required to confirm these results. In conclusion, our findings have important clinical implications: (1) it demonstrates that current MDS prognostic scores are unable to discriminate risk in patients with MDS-del(5q), which underscores the need to refine prognostic assessment for this entity, rather than relying on models developed from the broader MDS population; and (2) the IPSS-del(5q) could help identify patients who have a higher probability of disease progression and may therefore benefit from closer monitoring, earlier therapeutic intervention, or consideration of alternative treatment strategies. Future efforts should incorporate not only baseline diagnostic variables but also dynamic parameters captured throughout the course of the disease. Conflict-of-interest disclosure: M.J.M. received honoraria for consulting and scientific support from Jazz Pharmaceuticals. A.A.-K. received honoraria from Celgene, Bristol Myers Squibb (BMS), Novartis, Astex, ALX Oncology, Aprea Therapeutics, MedImmune, and AstraZeneca; received research support from H3 Biomedicine, Hemavant; and had scientific steering committee membership in Novartis. A.M.Z. participated in advisory boards, consulted, participated in clinical trial committees, and/or received honoraria from AbbVie, Akesobio, Agios, Amgen, Astellas, BioCryst, BeiGene, Boehringer Ingelheim, Celgene, BMS, Chiesi, Cornerstone Pharmaceuticals, Daiichi Sankyo, Dr Reddy, Epizyme, Faron, FibroGen, GlaxoSmithKline (GSK), GlycoMimetics, Genentech, Gilead, Geron, Janssen, Jasper Therapeutics, Karyopharm, Kyowa Kirin, Keros Therapeutics, Kura Oncology, Novartis, Notable, Orum Therapeutics, Otsuka, Pfizer, Regeneron, Rigel, Seattle Genetics, Shattuck Labs, Schrodinger, Syros Pharmaceuticals, Syndax, Servier, Takeda, Treadwell Therapeutics, Taiho, Vincerx Pharma, and Zentalis. D.V. participated in advisory boards, consulted, participated in clinical trial committees, and/or received honoraria from AbbVie, Agios, Amgen, Asofarma, Astellas, Celgene, BMS, Grifols, Janssen, Jazz Pharmaceuticals, Merck Sharp and Dohme, Novartis, Pfizer, Sanofi, Servier, Sobi, Syros Pharmaceuticals, and Takeda. U.P. received honoraria and research funding from BMS. C.H. declares equity ownership of the MLL Munich Leukemia Laboratory. A.J. received consulting fees from Novartis and BMS, and research funding from AstraZeneca. M.D.-C. received honoraria from BMS, Novartis, and Keros Therapeutics; and consulting fees from Novartis, BMS, Blueprint Medicines, Agios, Hemavant, Syros Pharmaceuticals, Keros Therapeutics, Curis, and GSK. The remaining authors declare no competing financial interests.