Serum erythropoietin levels and risk by IPSS were predictive of response to erythropoiesis-stimulating agents in patients with myelodysplastic syndromes.
Brett L. Houston, MD
Serum erythropoietin (EPO) levels and risk by IPSS were predictive of response to erythropoiesis-stimulating agents (ESAs) in patients with myelodysplastic syndromes (MDS), according to a Canadian database analysis presented at the 2015 International MDS Symposium. Findings from the analysis support the predictive value of the Nordic and European ESA scores, while providing additional utility for risk stratification.
"We identified a serum EPO level of less than 100 and IPSS low-risk patients as the most significant discriminators for response," lead study author Brett L. Houston, MD, Department of Internal Medicine, University of Toronto, Canada, said during her presentation. "In patients already preclassified as intermediate or good risk of response according to the Nordic score, we've identified a clinical prognostic score which further delineates based on response."
In the retrospective analysis, 208 patients at a median age of 75 were selected from the Canadian MDS registry. According to IPSS risk assessment, 95% of patients were in the low or intermediate-1 group. By IPSS-R, 19% were very low, 51% were low, 23% were intermediate, and 7% were high or very high. The median Nordic score was 4.
Patients were treated with an ESA within a median of 5.6 months from diagnosis. Erythropoietin was the most commonly administered ESA (69%), with the remainder of patients receiving darbepoetin (31%). Despite a lack of insurance coverage, 14% of patients received concurrent G-CSF.
Overall, 63% of patients were transfusion independent and 74% required less than 2 units of red blood cells per month. EPO levels were below 100 mIU/mL in 63% of patients and the median ferritin level was 463 ug/L.
The overall response rate (ORR) was 47%, without a statistically significant difference observed between ESA therapies (P = .28). In the IPSS low group, the ORR was 64% compared with 38% in the intermediate-1 group.
"In contrast to the recent European ESA score experience, we were unable to discern the same difference in overall magnitude of response in the very-low, low, and intermediate risk groups when evaluated by IPSS-R," Houston said. "We identified very low levels of response in the very-high group."
By Nordic score, the ORR was 31% and 57%, for the intermediate and good groups, respectively. "This is slightly lower than the predicted 74% chance of response that was previously published in the Nordic score paper," Houston added.
Response rates were categorically lower across all groups compared with European ESA score expectations. The ORRs were 48%, 54%, and 49%, across the 0, 1, and 2 scores in the Canadian study compared with expectations of 85%, 80%, and 64%, respectively. While the exact cause of this discrepancy is unknown, it was postulated that dose size could have played a role.
"The vast majority of patients were initiated at the equivalent of 40,000 units of erythropoietin weekly or darbepoetin 500 units every 2 to 3 weeks," Houston noted. "The typical trend was that if patients did not respond, then they were frequently dose-escalated to 60,000 or 80,000 units."
Using univariate analysis, IPSS, IPSS-R, blast percentage, number of cytopenias, hemoglobin levels, Nordic score, serum EPO, and transfusion reliance were identified as significant predictors of response. Cytogenetics, ferritin, LDH, and ESA type were not significant predictors of response.
"There was no discernable difference between baseline and 3-month ferritin in patients who responded and did not respond," Houston noted. "This held true irrespective of analysis excluding transfusion-dependent patients."
By multivariate analysis, IPSS low/intermediate groups and serum EPO levels above and below 100 mIU/mL were identified as the most significant predictors of response (P < .0001). From these numbers, a new ESA response score was fashioned, with IPSS low = 0; Int-1/2 = 1; EPO <100 = 0; EPO >100 = 2.
For evaluable patients (n = 112), those with a score of 0 (n = 43) experienced a response of 81%. For score 1 (n = 28), the ORR was 57%. In the higher score groups, the ORR was 33% and 17%, with a score of 2 (n = 18) and 3 (n = 23), respectively.
"Limitations of our study include the retrospective acquisition of data with regards to precise transfusion frequency, serum ferritin, and EPO levels," Houston noted. "Because not all of these values were identifiable in all patients, this limited our sample size in terms of calculating Nordic score and European ESA score."
Further work is required to validate the new model, including the collection of adequate data for a thorough analysis. Additionally, the consideration of molecular attributes, flow cytometry, and microenvironmental predictors of response could be used to further refine the model, Houston added.
Houston BL, Jayakar J, Lenis M, et al. A predictive model of response to erythropoietic stimulating agents in patients with myelodysplastic syndrome. Leukemia Research. 2015;39:1s (suppl; abstr 43).