Abstract 3803

A revised form of the International Prognostic Scoring System (IPSS-R) has recently been derived from a huge retrospective patient series (Greenberg et al, 2012), and a biologically upgraded version (IPSS-R “molecular”) is being worked out by the same group. The aim of this study was to evaluate the potential additive contribution of patient-related, as well as readily accessible peripheral blood disease-related prognosticators, to the IPSS-R prediction capability for estimating overall survival (OS) and progression into acute myeloid leukemia (AML) in MDS patients.


We prospectively recruited 266 MDS patients (Pts), from June 2006 to June 2010, in eight GESMD sites. The study was approved by the IRB at each study site and all Pts gave written informed consent. Cytological diagnosis and cytogenetic analysis followed standard operating procedures of the GESMD, ISCN guidelines and Schanz's categorization. Sixty-six Pts were excluded (MDS not confirmed 3, secondary MDS 6, CMML 40, lack of a valid karyotype 14 – w/o mitoses 10, not attempted 4-, duplicate 1, consent revocation 1). Finally 200 primary MDS cases (125 M/75 F; median age 76, range 31–91) were classified according to WHO-2008 (RC 11, RARS 13, RCMD 101, RAEB-1 30, RAEB-2 24, 5q- syndrome 15, hypoplastic MDS 1, unclassifiable 5), followed up until June 2012 (median follow-up 2.6 years, range 0.06–6.3) and categorized according to IPSS-R (Very Low 50, Low 80, Intermediate 33, High 25 and Very High 12). Fifty-five Pts received disease-modifying therapeutic strategies (DMTS) such as azacitidine (38), intensive chemotherapy w/o allo-BMT (17; in 6 Pts after AZA), or allo-BMT (13; in 7 after intensive chemotherapy). Forty-two Pts (21%) progressed into AML, 87 (43.5%) died and 13 (6.5%) were lost to follow-up. Median OS was 4.2 years. Age, comorbidity (as measured by Lee et al, 2006), performance status (ECOG), transfusion-dependence (according to Malcovati), serum LDH at diagnosis, ferritin, beta2-microglobulin, albumin, erythropoietin, plasma soluble p53-protein and interleukin-10 levels (ELISA), as well as peripheral blood WT1 gene expression (real-time PCR) were analyzed by testing the change in likelihood-ratio and the Akaike's information criterion (AIC) in Cox models after adding each individual covariate to IPSS-R. The increased discriminating power of the expanded prognostic model over that of IPSS-R alone was evaluated by the Harrell's C index and the R2explained variation. Replicability of the expanded prognostic model was tested by bootstrap re-sampling (1000 samples).


Addition of age (continuous) and ECOG (cutoff ≥2) to IPSS-R significantly improved the model's prognostic power, as measured by the likelihood ratio test and the AIC, as well as the discriminating power (Harrell's C increased from 0.70 to 0.75). Interestingly, addition of IL10 (cutoff ≥4.0 pg/mL) further improved the predictive power and reduced the residual variance (R2increased from 0.23 to 0.50). IL10 plasma levels were directly correlated with ferritin and transfusion dependence, and inversely correlated with hemoglobin. Bootstrap re-sampling predicted a replicability in eventual external validation series of 100%, 73%, 38% and 51%, respectively, for the covariates IPSS-R, age, ECOG and IL10. The IPSS-R category was the only predictor of progression into AML. Adjustment of the expanded prognostic model for exposure to DMTS, evaluated as a time-dependent covariate, had no relevant effect on the model's predictive ability.


Patient's age, ECOG and plasma levels of IL10 at diagnosis add further information to the IPSS-R risk category in the prognostication of patients with MDS. As suggested by Greenberg et al. in their paper, our study confirms that some covariates not yet included in the IPSS-R may be of additional help for predicting the ultimate fate of MDS patients.


No relevant conflicts of interest to declare.

We thank A. Pereira for expert statistical assistance. This study was supported by a grant from Celgene, S.L., Madrid (project INBIOMED HEMA 001/2006)

Author notes


Asterisk with author names denotes non-ASH members.