Elevated Pre-Treatment Fetal Hemoglobin Predicts Better Outcome in MDS/AML Patients Receiving 5-Aza-2'-Deoxycytidine (DAC)

Introduction: Aberrant DNA methylation occurs frequently in hematologic malignancies and is associated with altered gene expression. Consequently, DNA hypomethylating agents (HMAs), e.g. 5-aza-2'-deoxycytidine (DAC), have been tested for in vivo demethylation and have become accepted as first-line treatment of older MDS and AML patients (pts). While HMAs are already routinely used for the treatment of MDS and AML, only very few outcome predictors have been established thus far. Expression of the β-like globin gene locus is tightly regulated by methylation, is HMA-sensitive in vitro, and fetal hemoglobin (HbF) expression is under study as a potential biomarker for response of MDS pts to 5-azacytidine. Here, we present the first study of serial HbF measurements in MDS and AML pts receiving DAC in order to investigate a potential clinical application of HbF as a predictor of outcome to this treatment.

Methods: 16 MDS and 36 AML pts enrolled on two clinical trials, the 06011 EORTC-GMDSSG phase III trial of higher-risk MDS, and the 00331 phase II trial of older, non-fit AML pts (Lübbert M et al., Haematologica 2012), were treated with DAC. MDS pts received nine 4-hour infusions of 15 mg/m² given over 72 hours, repeated every 6 weeks for a minimum of four courses. AML pts were treated with nine 3-hour infusions of 15 mg/m² given over 72 hours, repeated every 6 weeks for four courses followed by maintenance treatment with DAC at 20 mg/m² when responding (defined by CR, PR or an antileukemic effect). HbF levels were measured in peripheral blood by HPLC before treatment and sequentially, i. e. after the end of each treatment course (every 6 weeks). HbF levels >1% of total hemoglobin were considered elevated.

Results: Baseline HbF was elevated (>1.0%) in 7/16 MDS and 12/36 AML pts. Clinical baseline characteristics were uniformly distributed between pts with normal and elevated HbF levels. HbF induction was observed in 81% of MDS pts after a median of 2 (range 2-6) courses of DAC and in 54% of AML pts after a median of 3 (2-11) courses. Four AML pts receiving standard cytarabine-based induction chemotherapy and four pts with pancreatic cancer receiving gemcitabine-based treatment did not show HbF induction to >1%. When analyzing clinical survival endpoints, elevated baseline HbF was associated with longer median overall survival (OS) for MDS: 26.6 vs. 8.6 months (HR 8.56, 95% CI 1.74-42.49, p=0.008). Similarly, median PFS and AMLFS was prolonged in "HbF high" MDS pts with 15.4 compared to 5.9 months (p=0.016) and 26.3 compared to 8.8 months (p=0.03), respectively. A statistically non-significant trend towards higher HbF baseline values (median 1.5%, range 0.3% to 3.9%) in MDS pts achieving a CR, PR or hematological improvement compared to non-responders (median baseline 0.4%, range 0.1% - 1.9%) was noted. Likewise, OS for AML pts with elevated pre-treatment HbF was prolonged with 10.0 vs. 2.9 months OS (HR 3.01, 95% CI 1.26-7.22, p=0.014). HbF baseline values were comparable in the group that did achieve any response (CR, PR, antileukemic effect) vs. the non-responding group. In a multivariate analysis including LDH and age, the predictive value of HbF was retained. Time-dependent Cox models revealed that the predictive value of treatment-induced HbF induction was markedly inferior to that of baseline HbF.

Conclusion: We provide first evidence for in vivo induction of HbF by DAC in MDS and AML pts. Notably, HbF values elevated already prior to treatment harbor highly significant predictive value for survival benefit upon DAC whereas HbF induction is inferior to a stronger effect of pre-treatment HbF. Our findings warrant incorporation of HbF as potential predictive biomarker in larger prospective studies.