Validation of Predictive Models for Response to Erythropoiesis-Stimulating Agents in Myelodysplastic Syndromes

Background: Erythropoiesis-stimulating agents (ESAs) increase hemoglobin, reduce transfusion requirements and improve quality of life in myelodysplastic syndromes (MDS), but many patients do not respond to treatment. Various scoring systems have been developed to predict which patients with MDS will respond to ESAs. The Nordic score derived by Hellstrom-Lindberg et al. stratifies patients based on the erythropoietin (EPO) level and transfusion frequency. Houston et al. have proposed another score which is based on EPO and the International Prognostic Scoring System (IPSS) risk group. The latter score has not yet been validated. We assessed the validity of these scoring systems in an independent cohort of patients with MDS.

Patients and methods: We conducted a retrospective cohort study including all patients with confirmed MDS based on bone marrow biopsy, aspirate or cytogenetic findings who were treated with ESAs at our institution from 2005 to 2013 and who had available follow-up information. Both the International Working Group (IWG) 2006 criteria and the Nordic study response criteria were used to assess treatment response. Statistical significance of both scores was assessed using Pearson's chi-squared test. Analysis was repeated with the score by Houston et al. dichotomized to low (0 or 1) or high (2 or 3).

Results: Of 180 patients diagnosed with MDS, 44 patients met our inclusion criteria. Forty patients were treated with epoetin alfa and 4 were treated with darbepoetin alfa. The mean age was 76.4 years and 61% were male. The mean pretreatment hemoglobin was 85.3 g/L. There was a significant difference in the treatment response between the Nordic score groups using the Nordic response criteria but the difference was insignificant using the IWG criteria (Table 1). We found no difference in treatment response between the scores by Houston et al. using either response criteria, however a low score (0 or 1) predicted a significantly better response than a high score (2 or 3). An EPO level less than 100 IU/L was a statistically significant predictor of treatment response but the IPSS designation was not.

Conclusion: Our results have validated the Nordic score using the same response criteria as the original study, but our population size was not adequate to validate the Nordic score using IWG criteria. We have confirmed that patients with EPO levels below 100 IU/L respond to treatment significantly better than patients above 100 IU/L. Our data does not support the use of the IPSS risk group in predicting treatment response, but this result may be limited by the relatively small number of patients. Further studies exploring the predictive variables in MDS treatment with ESAs are warranted.