Method and limitations
How the ESA-PSS-M score is computed, where the coefficients come from, and what the tool does not do.
1. The score
ESA-PSS-M was derived in a European multicentre cohort of 535 patients with lower-risk MDS and validated in an external cohort of 223 patients. It combines three variables available before the treatment decision.
TD equals 1 if the patient is transfusion dependent at baseline, 0 otherwise. The score is oriented favourably: the higher it is (closer to 0), the greater the probability of erythroid response and the longer the time to transfusion dependence.
2. IPSS-M
IPSS-M (Bernard et al., NEJM Evidence 2022) is a Cox model combining haemoglobin, platelets, bone-marrow blast percentage, the IPSS-R cytogenetic category and 31 genes. Each variable is mean-centred on the derivation cohort, multiplied by its β coefficient and divided by ln(2): the score is therefore expressed as a log₂ hazard ratio, where 0 is the average patient.
| Category | Score | Median OS |
|---|---|---|
| Very low | ≤ −1.5 | 11.7 yrs |
| Low | −1.5 à −0.5 | 7.1 yrs |
| Moderate low | −0.5 à 0 | 4.4 yrs |
| Moderate high | 0 à 0.5 | 3.1 yrs |
| High | 0.5 à 1.5 | 2.3 yrs |
| Very high | > 1.5 | 1.3 yrs |
Special rules
- TP53 multi-hit: ≥ 2 TP53 mutations, or one mutation with del(17p) / LOH / VAF ≥ 50%. A mono-allelic TP53 mutation only counts in the residual-gene channel.
- SF3B1: favourable (β −0.079) only in the absence of del(5q) and of an adverse comutation (FLT3, MLL-PTD, RUNX1, NRAS, EZH2, IDH2, BCOR, BCORL1, STAG2, SRSF2, U2AF1, TP53 multi-hit). With isolated del(5q), SF3B1 becomes adverse (β +0.504).
- Residual genes: 15 genes aggregated, capped at 2 counted mutations.
- Karyotype: the IPSS-R category is derived from the entered abnormalities per Greenberg 2012. del(17p) is not an IPSS-R defining lesion and counts as an “other” abnormality.
3. Cut-offs
Two cut-offs are displayed: −1.75 and −1.30. They come from the analysis of the PRETAR cohort (Park S., Grenoble Alpes University Hospital) and place the patient in one of three zones. The cut-off published in Blood maximises specificity (76%) at the cost of a 57% sensitivity: the score rules out non-responders better than it selects responders.
4. What the tool does not do
- It does not replace clinical judgement or the tumour board.
- It is not a validated medical device; no CE marking is claimed.
- It does not predict response duration or individual survival.
- It does not include prior treatments, ferritin, or the detailed transfusion profile.
5. Data
No identifying data is requested or stored. Each calculation records only the clinical values, the resulting scores, the language and a timestamp, in order to produce aggregate usage statistics and an anonymous CSV export.
References
- Raddi MG, Park S, Fontenay M, et al. IPSS-M risk and specific sex-associated somatic mutations predict response to ESA therapy in LR-MDS: building a new score. Blood. 2025;146(14):1693-1706. doi:10.1182/blood.2024027540
- Bernard E, Tuechler H, Greenberg PL, et al. Molecular International Prognostic Scoring System for Myelodysplastic Syndromes. NEJM Evid. 2022;1(7):EVIDoa2200008.
- Greenberg PL, Tuechler H, Schanz J, et al. Revised International Prognostic Scoring System (IPSS-R). Blood. 2012;120(12):2454-2465.