The International Prognostic Scoring System (IPSS) has been a mainstay of assessing prognosis of patients with newly diagnosed myelodysplastic syndrome (MDS) since 1997.1 The model was initially based on the number of cytopenias, type of cytogenetic abnormalities identified by conventional karyotype, and percentage of bone marrow blasts. The system describes four prognostic groups with distinct overall survivals and risk of progression to acute myeloid leukemia. The 2012 International Prognostic Scoring System revision (IPSS-R)2 incorporated both the number and depth of cytopenias, provided a more detailed assessment of cytogenetic risk based on the Comprehensive Cytogenetic Scoring System (CCSS),3 and introduced a new 2 percent bone marrow blast threshold. The IPSS-R comprised five rather than four risk groups and estimated patient prognosis more precisely than the IPSS.4 However, neither scheme took into account the assessment of mutations by next-generation sequencing (NGS), which has become routine in the workup of patients with MDS and has a profound impact on MDS patient outcomes as shown in several studies.5-7
Dr. Elsa Bernard and colleagues have developed the Molecular International Prognostic Scoring System (IPSS-M), a new version of the IPSS that incorporates mutation data. The IPSS-M was trained on a series of 2,957 previously untreated patients with MDS from 24 institutions across four continents and validated in a separate cohort of 754 patients with MDS from Japan. After interrogating 152 genes implicated in myeloid neoplasms, the investigators found 16 genes that were individually associated with leukemia-free survival (“main effect” genes), as well as an additional 15 genes that collectively contributed to prognosis if one or two were mutated (“residual effect” genes). The IPSS-M integrates mutation data from the 16 main effect and 15 residual effect genes with blood count and blast values (now assessed as continuous rather than categorical values to improve precision) and the original CCSS cytogenetic risk group. Based on these data, a specific risk score is calculated for each patient based on their hazard ratio for acute myeloid leukemia–free survival compared to the average patient with MDS in the series, and the patient is assigned to one of six risk groups ranging from very low to very high. The IPSS-M risk score calculation is more complex than that of the IPSS and IPSS-R: While the authors provide the equation in the supplementary material, the IPSS-M score and associated predicted clinical outcomes can be determined by entering patient data into a publicly available IPSS-M web calculator (mds-risk-model.com; Figure) that is straightforward to use.
Calculation of the Molecular International Prognostic Scoring System (IPSS-M) risk of a newly diagnosed “real-life” patient with myelodysplastic syndrome using the IPSS-M web calculator (mds-risk-model.com). The patient is an 82-year-old man with pancytopenia (white blood cells, 4.4 × 109/L [absolute neutrophil count, 0.8 × 109/L]; hemoglobin, 9.0 g/dL; platelets, 59 × 109/L) diagnosed with MDS with multilineage dysplasia (1% bone marrow blasts), normal karyotype, and pathogenic CBL, TET2, STAG2, SRSF2, ASXL1, RUNX1, and NRAS mutations. The “density” curve represents the overall number of patients in each of the six risk score groups, while the patient is shown on the vertical black dotted line. The patient’s revised IPSS (IPSS-R) score is 2.5 (low); however, the patient is scored as very high in the IPSS-M scheme (hazard ratio for leukemia-free survival, 1.93 compared to the average MDS patient) owing to the presence of multiple high-risk mutations that are not accounted for in the IPSS-R. The patient’s overall and leukemia-free survivals are estimated on the web calculator to be 1 year and 0.76 years, respectively (not shown). The grey area flanking the patient’s estimated risk score reflects the range of uncertainty due to some mutations that were not assessed on the next-generation sequencing panel (MLL-PTD, ETNK1, GNB1, and PRPF8); however, despite these missing data, this patient’s range still falls entirely within the very high risk category.
Calculation of the Molecular International Prognostic Scoring System (IPSS-M) risk of a newly diagnosed “real-life” patient with myelodysplastic syndrome using the IPSS-M web calculator (mds-risk-model.com). The patient is an 82-year-old man with pancytopenia (white blood cells, 4.4 × 109/L [absolute neutrophil count, 0.8 × 109/L]; hemoglobin, 9.0 g/dL; platelets, 59 × 109/L) diagnosed with MDS with multilineage dysplasia (1% bone marrow blasts), normal karyotype, and pathogenic CBL, TET2, STAG2, SRSF2, ASXL1, RUNX1, and NRAS mutations. The “density” curve represents the overall number of patients in each of the six risk score groups, while the patient is shown on the vertical black dotted line. The patient’s revised IPSS (IPSS-R) score is 2.5 (low); however, the patient is scored as very high in the IPSS-M scheme (hazard ratio for leukemia-free survival, 1.93 compared to the average MDS patient) owing to the presence of multiple high-risk mutations that are not accounted for in the IPSS-R. The patient’s overall and leukemia-free survivals are estimated on the web calculator to be 1 year and 0.76 years, respectively (not shown). The grey area flanking the patient’s estimated risk score reflects the range of uncertainty due to some mutations that were not assessed on the next-generation sequencing panel (MLL-PTD, ETNK1, GNB1, and PRPF8); however, despite these missing data, this patient’s range still falls entirely within the very high risk category.
The IPSS-M has several unique features that can positively impact the treatment of MDS patients going forward:
The IPSS-M demonstrates superior prognostic power compared to the IPSS-R, with a significant improvement in the C-index for both leukemia-free and overall survival.
Almost half of the patients in the study were re-stratified into different risk strata from the IPSS-R, of which three-quarters were upstaged to a higher risk group based on adverse-risk mutations that are not accounted for in the IPSS-R (see example case illustrated in the Figure).
Outcomes in therapy-related or secondary MDS (t-MDS) were similar to those of patients with primary MDS within the same risk stratum, emphasizing that t-MDS is a heterogeneous disease that is skewed toward higher risk but warrants individualized prognosis determination for each patient.
The effects of two mutations, TP53 and SF3B1, depended on their context — multi-hit status in the case of TP53 mutation and co-occurrence with del(5q) and other mutations in the case of SF3B1 mutation. While this study determined TP53 multi-hit status by loss-of-heterozygosity, which is not available in many clinical NGS platforms, the TP53 status allelic can be imputed in the IPSS-M by accounting for the TP53 variant allele fraction and cytogenetics background.
In Brief
The arrival of the IPSS-M has elevated prognostication of patients with MDS to a new level commensurate with the currently available comprehensive genetic analyses in clinical practice. Although not all of the IPSS-M genes may be included in a given NGS platform, the model provides a range of prognostic scores to reflect missing data (Figure). Hopefully, this work will stimulate more widespread inclusion of prognostically relevant mutations such as KMT2A partial tandem duplication (KMT2A/MLL-PTD) in clinical NGS testing. It should also be noted that only 30 percent of the patients underwent treatment-modifying therapy, and less than 10 percent received hematopoietic stem cell transplantation in this study. Although the authors conducted subset analysis to show that the IPSS-M improved prognostic discrimination over IPSS-R in treated as well as untreated patients, further study is needed to evaluate the application of the model to patients receiving specific treatment modalities, including hematopoietic stem cell transplantation.8 Most importantly, the IPSS-M will improve the accuracy of making risk-adapted treatment decisions and foster molecularly informed clinical trial design, thereby advancing personalized medicine for patients with MDS.
Competing Interests
Dr. Hasserjian indicated that he is a coauthor of the Bernard et al article.
