Molecular International Prognostic Scoring System for Myelodysplastic Syndromes
- Study type
- Observational study
- Study name
- IPSS-M
- Design
- Development and external validation of a clinical-molecular prognostic model (feature selection and a robust multivariable Cox model) based on gene mutation profiles in pre-treatment samples
- Sample size
- 2957 patients (development cohort) and 754 Japanese patients (external validation cohort)
- Source
- NEJM Evid · 2022
In plain words
Myelodysplastic syndrome (MDS) is a bone marrow disease whose severity varies greatly from one person to another; this is why doctors use a "risk score" to decide when and how intensively to treat. The previous common score (IPSS-R) was based only on blood and bone marrow tests and a check of the chromosomes. In this research, by studying gene mutations in nearly 3000 patients, a new score called IPSS-M was built that also takes into account mutations in 31 genes and places patients in six risk groups. This score made predictions about the course of the disease more accurate and changed the risk group of 46% of patients. If you have MDS, you can ask your doctor whether testing for gene mutations would help to estimate your risk more precisely.
Key results
Study population
Patients with myelodysplastic syndrome (MDS) whose pre-treatment diagnostic sample, or a sample taken close to diagnosis, was profiled for mutations in 152 genes; including primary MDS and secondary/therapy-related MDS. External validation was done in an independent cohort of 754 Japanese patients with MDS. Other inclusion criteria are not stated in the abstract.
Endpoints
Leukaemia-free survival, leukaemic transformation and overall survival; identification of independent prognostic variables and their relative weights, and comparison of the prognostic discrimination of IPSS-M with IPSS-R.
Safety and side effects
This was not an interventional study and has no safety data.
Clinical significance and place in guidelines
According to the authors, IPSS-M, by adding the genomic profile to haematological and cytogenetic parameters, improves risk stratification of patients with MDS and is a valuable tool for clinical decision-making. Each patient's score is calculated with a freely available web calculator that also accommodates missing values, and patients are placed in six risk groups: very low, low, moderate-low, moderate-high, high and very high.
In a large real-world validation (Sauta et al., J Clin Oncol 2023; a retrospective analysis of 2876 patients with primary MDS from the GenoMed4All consortium), IPSS-M improved the concordance index compared with IPSS-R (0.81 versus 0.74 for overall survival and 0.89 versus 0.76 for leukaemia-free survival) and changed the risk group of 46% of patients (23.6% upstaged and 22.4% downstaged). In patients undergoing haematopoietic stem cell transplantation, IPSS-M predicted the risk of relapse and the probability of survival after transplantation significantly better than IPSS-R (concordance 0.76 versus 0.60 for overall survival and 0.89 versus 0.70 for probability of relapse); however, in higher-risk patients treated with hypomethylating agents (HMA) it could not predict the probability of an individual response to treatment, although response duration and survival probability decreased with increasing IPSS-M risk. In the same study, mutation information on at least 15 genes (ASXL1, CBL, DNMT3A, ETV6, EZH2, FLT3, IDH2, MLLPTD, NPM1, NRAS, RUNX1, SF3B1, SRSF2, TP53multihit and U2AF1) was needed to reach a predictive accuracy of 80% and 70% (in the GenoMed4All and IWG-PM cohorts, respectively).
Limitations
The abstract does not describe the study's limitations. Calculating IPSS-M requires gene mutation testing (the 31 genes of the model); the calculator accommodates missing values, but in the validation by Sauta et al. (J Clin Oncol 2023), predictive accuracy fell significantly when the number of available genes was reduced to 10 or fewer. External validation in this paper itself was done in a Japanese cohort (754 patients). IPSS-M is a prognostic tool and, in the same validation, did not predict individual response to hypomethylating agents.
This summary was prepared by the editorial team from the published paper; the original paper is the final reference. Read the original abstract on PubMed.