Development and Validation of a New Prognostic Model for Myelodysplastic Syndrome (MDS) That Accounts for Events Not Considered by the International Prognostic Scoring System (IPSS)

Background. The IPSS risk model provides survival projections for patients with de novo MDS managed with supportive measures alone. For patients receiving investigational treatment, a prognostic stratification model is needed that can be applied at intervals after diagnosis and that adjusts for the impact of prior therapy, secondary forms of disease, proliferative CMML, and adverse cytogenetic subsets (e.g. 3 abnormalities, chromosome 7 abnormalities).

Aims. To develop a new MDS risk model that accounts for subsets not included in IPSS, that refines prognostic subsets, and that applies at any time during course of MDS.

Study Group. We analyzed 1915 patients with MDS referred from 1993 to 2005 (including CMML, secondary MDS, MDS with prior therapy). Only 507 patients (26%) had primary MDS without prior therapy (i.e. categorizable by IPSS). Patients were randomly divided into a study group (n=958) and a test group (n=957).

Results. A multivariate analysis of prognostic factors in the study group identified the following adverse independent factors as continuous and categorical values (p<0.001), which were given weighted points based on coefficient (score point = coefficient: 0.15). This is shown in Table 1. Cutoffs for anemia, thrombocytopenia and blasts, and cytogenetic subsets, were different for IPSS. The new MDS prognostic model divided patients into four prognostic groups with significantly different outcomes, shown in Table 2. The model was validated in the test group with excellent segregation (Table 2). It was also highly prognostic in the 507 patients with newly diagnosed MDS (as per the original IPSS groups): median survivals 4.7, 3.0, 1.2, 0.75 years. Applying the prognostic score of the new model within the four IPSS risk groups, overall and in primary MDS without prior therapy, was highly prognostic in each. Applying IPSS within each of the 4 risk groups of the new MDS model was not prognostic. The model was also prognostic for multiple MDS subsets tested (Table 3). The new model accounts for duration of MDS and prior therapy. It is applicable to any patient with MDS at any time during the course of MDS. The new risk model was also tested in the 3 arm decitabine trial (n = 124); these patients were divided by the new model into 5 (4%) low risk, 21 (17%) Intermediate 1, 45 (30%) Intermediate 2, and 53 (43%) high risk. This indicates the worse prognosis in this study group (higher risk MDS 79%) The respective median survivals were: not reached (100% at 3 years), 42, 19, and 13 months, respectively (Table 3). This indicated the applicability of the model in a different MDS study group and suggested a better survival than expected (therapy effect?). To verify this, a cumulative score for the 124 patients was calculated and an average score deducted, which was associated with a predicted historical median survival of 13 months overall, 30 months for low-Intermediate 1 and 10 months for Intermediate 2-high risk, versus median survivals of 20 months overall, 44 months for low-Intermediate 1, and 15 months for Intermediate 2-high.

Conclusions. A new prognostic model was developed and validated for MDS, which accounts for all MDS or CMML cases, regardless of prior therapy. The model has been validated in an independent test group and was shown to be superior to IPSS. It was also used to demonstrate an improved survival with decitabine compared with the expected (historical) survival by the new risk model. Further validations are needed in independent MDS populations.

Table 1

Table 2

Table 3