Predictors of Response and Survival in 206 AML Patients Treated with Guadecitabine in a Phase 2 Study

Background: Guadecitabine is a next generation hypomethylating agent (HMA) resistant to degradation by cytidine deaminase which results in prolonged in vivo exposure to the active metabolite decitabine. We conducted a prospective phase 2 study testing different schedules of guadecitabine in 206 AML patients. We present here the results of multiple logistic regression, and Cox regression analyses of predictors of composite Complete Response or CRc (CR+CRp+CRi) and overall survival (OS).

Methods: Multiple logistic regression analysis of response (CRc), and Cox regression analysis of OS were conducted with inclusion of the following baseline variables: disease state (relapsed/refractory (r/r) vs. treatment naïve (TN) AML); guadecitabine schedule (10-day vs. 5-day); age (<75 vs. ≥75); ECOG PS (0-1 vs. ≥2); Cytogenetics (others vs. poor risk); baseline BM blasts (≤40% vs. >40%); baseline Peripheral blood (PB) blasts (≤30% vs. >30%); baseline WBCs count (<20,000/µL vs. ≥20,000/µL); Flt-3 ITD, NPM, and TP53 mutations (each present vs. not detected). Cutoff values for blasts % for BM and PB were chosen based on the median or mean values respectively, and cutoff value for WBCs count was chosen as a common cutoff used for proliferative AML. Backward elimination method with alpha =0.05 was used to reach the final models.

Results: We treated 206 AML patients (103 patients each for TN or r/r AML); 101 with 5-day schedule and 105 with 10-day schedule. There were 91 patients (44%) ≥75 y; 53 (26%) with ECOG PS≥2; 85 (41%) with poor risk cytogenetics; 99 (48%) with baseline BM blasts >40%; 65 (32%) with baseline PB blasts >30%; 20 (9.7%) with WBCs ≥20,000/µL. Flt-3, NPM, and TP53 mutations were present in 8%, 9%, and 4% of patients respectively. The final logistic regression model indicated that patients with ECOG PS 0-1 and those with baseline PB blasts ≤ 30% have two-fold higher odds of response than those with ECOG PS ≥2 and PB blasts >30% (Odds ratio 2.18; and 2.03 respectively; p<0.05 for both). Patients with TN AML had five-fold higher odds of response to guadecitabine than r/r AML (Odds ratio of response for r/r AML 0.22; p <0.0001). The final Cox regression model showed that both ECOG PS 0-1 and PB blasts ≤30% retained their significance for OS (HR 0.69 with p=0.03; and 0.61 with p= 0.004 respectively). However disease state (r/r AML vs TN AML) lost significance for OS while cytogenetics risk level (others vs poor risk) became significant with a HR for OS of 0.68, p=0.016.

Summary/Conclusions: In a prospective series of AML patients treated with guadecitabine in a phase 2 study, better ECOG PS 0-1 and lower baseline PB blasts ≤30% were associated with a significantly higher likelihood of response and a longer OS. Patients with TN AML had significantly higher likelihood of response than those with r/r AML but this was not a significant factor for OS when other factors are present in the model such as ECOG PS, cytogenetics risk and PB blasts. On the other hand, the presence of poor risk cytogenetics did not alter the likelihood of response to guadecitabine but still had shorter survival compared to patients with other cytogenetics risk levels. Other variables such as age, baseline BM blasts %, and baseline WBCs count did not significantly impact response or OS in AML patients treated with guadecitabine when all other factors are present in the models. The analysis of genetic mutations was limited by the small number of patients where these mutations were present. The analysis of the treatment schedule is limited by the different effect of the schedule on r/r AML compared to TN AML where the 10-day schedule did better than the 5-day schedule in r/r AML but not in TN AML patients.