AUTHORS
Seda Cakmak, MD, Azin Keshavarz, MHSc.2, Ryan J Stubbins, MD, MSc, FRCPC3, Sergio Rodriguez-Rodriguez, MD, MSc4, Yomna Eissa, MD5, Yishan Ye, MD6, Yu Cai, MD7, Yu Wang8, Robert Zeiser, MD, PhD9, Michael Heuser, MD10, Judith Schaffrath11, Asmaa Aloufi, MBBS12, Manuel Ricardo Espinoza-Gutarra, MD13, Jae-Sook Ahn14, Hyeoung-Joon Kim, MD15, Mohsen Alzahrani Sr, MD16, Ahmad Alhuraiji, MD, MRCP17, Ali Bazarbachi, MD, PhD18, Ali Ibrahim19, He Huang, PhD20, Xianmin Song, MD21, Xiaojun Huang, MD,PhD22, Dennis Dong Hwan Kim, MD, PhD23
AFFILIATIONS
1 Princess Margaret Cancer Centre, Toronto, Ontario, Canada
2 Princess Margaret Cancer Centre, Toronto, Ontario, Canada
3 BC Cancer and Vancouver Coastal Health, Vancouver, British Columbia, Canada
4 Princess Margaret Cancer Centre, Toronto, Ontario, Canada
5 London Health Sciences Centre, London, Ontario, Canada
6 The First Affiliated Hospital, Zhejiang University, Hangzhou, China
7 Shanghai General Hospital, Shanghai Jiao Tong University, Shanghai, China
8 Peking University People’s Hospital, Beijing, China
9 University of Freiburg, Freiburg, Germany
10 Martin-Luther-University Halle-Wittenberg, Halle, Germany
11 Martin-Luther-University Halle-Wittenberg, Halle, Germany
12 BC Cancer and Vancouver Coastal Health, Vancouver, British Columbia, Canada, King Faisal Specialist Hospital & Research Centre, Madinah, Saudi Arabia
13 O’Neal Comprehensive Cancer Center at UAB, Birmingham, United States
14 Chonnam National University Hwasun Hospital, Hwasun, Korea, Rep. of South
15 Chonnam National University Hwasun Hospital, Hwasun, Korea, Rep. of South
16 Ministry of National Guard Health Affairs, Riyadh, Saudi Arabia
17 Kuwait Cancer Control Center, Kuwait, Kuwait
18 American University of Beirut Medical Center, Beirut, Lebanon
19 American University of Beirut Medical Center, Beirut, Lebanon
20 The First Affiliated Hospital, Zhejiang University, Hangzhou, China
21 Shanghai General Hospital, Shanghai Jiao Tong University, Shanghai, China
22 Peking University People’s Hospital, Beijing, China
23 Princess Margaret Cancer Centre, Toronto, Ontario, Canada, University of Toronto, Toronto, Canada
ABSTRACT
Background: FLT3-ITD mutated acute myeloid leukemia (AML) carries a high relapse risk after allogeneic hematopoietic cell transplantation (HCT). FLT3-inhibitor (FLT3i) post-transplant maintenance (PTM) is increasingly used in clinical practice to reduce this risk. Preclinical data (ASH 2025) from FLT3-ITD-expressing Ba/F3 lines identified a pan-resistant subset defined by juxtamembrane segment (JMS)-spanning insertions conferring resistance to both FLT3i and chemotherapy. Our prior single-centre analysis (n=73) suggested PTM reduced relapse in non-JMS variants whereas its benefit was not significant in JMS-spanning variants. Here we expand to a multicentre cohort of 288 patients with available next-generation sequencing (NGS) data at diagnosis, providing genomic insertion location, length and allele frequency of FLT3-ITD, to test whether PTM benefit differs by ITD pattern using immortal-time-bias-corrected methods.
Aims: To quantify the effect of FLT3i PTM on post-HCT outcomes within JMS-spanning (Pattern 1) and non-JMS (Pattern 2) variant subgroups of FLT3-ITD AML, whilst accounting for the time-dependent nature of the PTM exposure.
Methods & Patients: We retrospectively analysed 288 FLT3-ITD AML patients transplanted in CR1/CR2 across 11 international centres with available NGS FLT3-ITD profiles. A rule-based Python classifier assigned each patient to Pattern 1 (any ITD clone spanning JMS aa579–592) or Pattern 2 (all clones outside JMS), based on protein-level HGVS annotation (cDNA fallback where protein annotation absent). PTM was defined as receipt of post-transplant FLT3i (sorafenib, gilteritinib or midostaurin) whilst in remission for any duration. Endpoints were overall survival (OS), relapse-free survival (RFS), cumulative incidence of relapse (CIR; primary), and non-relapse mortality (NRM). As PTM is initiated around day 100 post-HCT, we used time-dependent Cox regression as the primary analysis to avoid immortal-time bias. All analyses were performed in R.
Results: 288 patients were classified to Pattern 1 (n=85) or Pattern 2 (n=203). The median allele frequency of FLT3-ITD was 24.5% (interquartile range 9.0–37.0%).
Among 86 patients who received PTM (sorafenib n=67, gilteritinib n=31, midostaurin n=2) and 202 no-PTM patients, there were 83 deaths, 71 relapses and 32 NRM events over a median follow-up of 27 months. Median PTM start was day +90–100 post-HCT; median PTM duration was 250 days (Pattern 1) and 315 days (Pattern 2).
In time-dependent Cox analysis over the whole cohort, PTM significantly improved OS (HR 0.43, 95% CI 0.24–0.79, p=0.006). RFS and CIR trended in the same direction (RFS HR 0.63, 0.37–1.06, p=0.08; CIR HR 0.57, 0.30–1.08, p=0.08) but didn’t reach significance. NRM was unchanged (HR 0.78, 0.32–1.89, p=0.58). The CIR did not differ significantly between the Pattern 1 and 2 groups, with estimated 2-year CIR rates of 17.3% (Pattern 1) vs 29.6% (Pattern 2); Gray’s p=0.256. Between the 2 patterns, there was no significant difference in OS (p=0.10), RFS (p=0.16), or NRM (p=0.52).
In within-pattern time-dependent Cox analyses, in Pattern 2 patients, PTM significantly improved OS (HR 0.42, 0.22–0.82, p=0.011) and CIR (HR 0.47, 0.22–0.98, p=0.044). RFS showed a similar trend favouring PTM (HR 0.58, 0.32–1.04, p=0.069) while NRM was unaffected (HR 0.91, 0.34–2.47, p=0.86). However, in Pattern 1 patients, PTM did not reduce relapse (CIR HR 0.93, 0.25–3.42, p=0.91) or RFS (HR 0.73, 0.24–2.20, p=0.57). OS and NRM favoured PTM (OS HR 0.35, 0.08–1.54, p=0.17; NRM HR 0.43, 0.05–3.65, p=0.44) but did not reach significance.
Conclusion: In this multicentre cohort of 288 FLT3-ITD AML patients, FLT3i PTM was associated with a ∼58% reduction in the hazard of death and ∼53% reduction in relapse hazard in non-JMS variants (Pattern 2) under immortal-time-bias-corrected analysis. In JMS-spanning variants (Pattern 1), PTM did not reduce cumulative incidence of relapse (HR 0.93) or relapse-free survival, consistent with the preclinical hypothesis that JMS-spanning FLT3-ITDs are relatively FLT3i-resistant. Larger JMS-spanning cohorts and prospective validation, as well as preclinical confirmation of these findings are required before ITD pattern can guide PTM decisions. Updated and expanded data will be presented at the ASH meeting.
DISCLOSURES
Seda Cakmak, MD: No relevant conflicts of interest to declare.