Long-Term Survival of Acute Myeloid Leukemia Responding Patients Who Stopped Azacytidine and/or Venetoclax Because of Poor Tolerance or Physician Choice: A Retrospective Multicenter Study from the French Innovative Leukemia Organization (FILO)

Background: Acute myeloid leukemia (AML) patients treated upfront with venetoclax and azacitidine (VEN-AZA) achieved a 66.4% composite complete response (CRc) including 36.7% complete response (CR) and 29.7% CR with incomplete hematologic recovery (CRi) in the VIALE-A trial (DiNardo et al., 2020). Median event-free survival (EFS) and duration of response were 9.8 months (95% CI, 8.4 to 11.8) and 17.8 months, respectively. These good results are mitigated by an 82% rate of hematologic adverse events and a 42 to 54% risk of febrile neutropenia both in clinical trial and real-life studies. A recent report of a small cohort of 13 patients in CRc for ≥12 months who interrupted VEN-AZA showed that these patients maintained long treatment-free remission period (Chua et al., 2022) suggesting that treatment interruptions might be considered. To evaluate the impact and the safety of this strategy, we designed a retrospective multicentric study of responding patients who stopped VEN and/or AZA.

Methods: Eligible patients had to 1) have been treated with VEN-AZA for previously untreated or pretreated AML, 2) have stopped VEN or AZA for more than 3 months because of poor tolerance and/or physician choice, 3) be in response (CR, CRi or Morphologic leukemia-free state, [MLFS]) at the time of treatment interruption, 4) be aged ≥18 years. Main endpoints were overall survival (OS), event-free survival (EFS), and treatment-free remission (TFR). We measured the impact of prognostic factors on long-term survival by calculating hazard ratios (HR) in univariate and multivariate analyses.

Results: We included 51 patients from 12 centers of the FILO group treated with VEN-AZA upfront (n=35, 68.6%) or as salvage treatment (n=16, 31.4%), between 01/19 and 02/22. Next Generation Sequencing (NGS) and minimal residual disease (MRD) data were available for 38 (74.5%) and 28 (55%) patients, respectively. Median age was 75 (ranges, 22-89). WHO classification AML was de novo, AML with myelodysplastic-related changes and therapy-related AML for 27 (52.9%), 21 (41.2%) and 3 (5.9%) patients, respectively. Median platelet count, total white blood cell count and median absolute neutrophile count were 52 G/L (ranges, 7-259), 1.9 (ranges, 0.5-95.5) and 0.7 (ranges, 0-31.6), respectively. Median bone marrow blasts involvement was 32% (range, 7-99). Fourteen patients (27.5%) had adverse cytogenetics. NPM1 and FLT3 mutations were found in 9 (17.6%) and 4 (7.8%) patients, respectively. Twenty (39.2%) patients had a IDH mutation (IDH1, n=5 [9.8%]; IDH2R140 n=9 [17.6%] and IDH2R172, n=6 [11.8%]). TP53 mutation was mutated in 5 (9.8%) cases. Patients received a median of 4 cycles of VEN-AZA (ranges, 1-15); response were CR, CRi and MLFS for 33 (64.7%), 14 (27.5%) and 11 (21.6%) before interrupting VEN-AZA. Complete response with negative MRD was found for 24 out of the 28 evaluable patient (85.7%). Discontinuation of VEN and/or AZA treatments was based on physician choice because of hematological and/or non-hematological toxicities in 36 (71.6%) and 9 (17.6%) cases. Treatment interrupted was VEN, AZA or both in 26 (51%), 1 (2%) and 24 cases (47.1%), respectively. Median follow-up was 17.8 (15.2-20.4) months. Median EFS and OS were 16.7 (12.8-20.6) and 26.8 (14.1-39.5) months, respectively (Figure). Median duration of treatment-free remission (TFR) was 10.1 (7.1-13.1) months (Figure). Twelve patients resumed VEN and/or AZA treatment. For these patients, CRc, MLFS and SD/PD was obtained in 6 (50%), 2 (16.6%) and 5 (41.6%) cases. Obtaining a CR/CRi following the first VEN-AZA cycle and number of prior VEN-AZA cycles >4 were the main prognostic factors associated with a longer EFS and OS in univariate and/or multivariate analyses (Table). None of these factors were predictive for a better TFR, except the achievement of CR/CRi after cycle 1 (HR=2.2, p=0.05). The type of interrupted treatment (VEN versus VEN-AZA) did not influence EFS, OS nor TFS.

Conclusion: Our findings confirm the high efficacy of VEN-AZA treatment in AML. Discontinuation of VEN and/or AZA treatment in responders was associated with durable responses and survival. Strategies of de-escalation based on clinical response, MRD and/or molecular profiling may be envisaged but will require a prospective evaluation in clinical trials.