Real‐world experience with <scp>CPX</scp> ‐351 for secondary acute myeloid leukaemia: Comparison with <scp>FLAG</scp> ‐ <scp>IDA</scp> in a propensity score matching analysis

The treatment landscape of acute myeloid leukaemia (AML) has evolved considerably over the past decade, driven by advances in translational research and precision medicine.1, 2 More than 10 novel agents have received regulatory approval, not only expanding therapeutic options but also introducing challenges in selecting the optimal front-line regimen.3-7 Among these, CPX-351 (CPX)—a liposomal formulation of cytarabine and daunorubicin at a fixed 5:1 molar ratio—was approved by the US Food and Drug Administration in 2016 for secondary AML (sAML), encompassing therapy-related AML (tAML) and AML with myelodysplasia-related changes (AML-MRC). This approval was based on a phase 3 trial comparing CPX with standard 7 + 3 induction chemotherapy in older adults (≥60 years), which demonstrated improved outcomes with CPX, including superior overall survival (hazard ratio [HR] 0.69; p = 0.005), higher Complete Remission (CR) + Complete Remission with Incomplete Hematologic Recovery CRi rates (48% vs. 33%; p = 0.016) and increased rates of allogeneic haematopoietic stem cell transplantation (Allo-HSCT) (34% vs. 25%).8 Despite these advances, the optimal induction strategy for fit, high-risk AML patients—particularly when compared with FLAG-IDA—remained uncertain. The UK NCRI AML19 trial directly compared CPX and FLAG-IDA, enrolling mainly patients under 60 years, though older fit individuals were eligible. The trial reported similar CR and overall survival rates between the two regimens; however, a subgroup analysis revealed a survival advantage for CPX among patients harbouring myelodysplasia-related (MR) gene mutations (median OS 38.4 vs. 16.3 months with FLAG-IDA).9 These findings raised important questions: Do they hold true in real-world practice, and can mutational profiles identify patients most likely to benefit from CPX or FLAG-IDA? To address these questions, we conducted a retrospective, multicentre study comparing CPX and FLAG-IDA in adult sAML patients who underwent next-generation sequencing (NGS) profiling and received either regimen as induction therapy. Patients were selected from all consecutive adult sAML patients who met eligibility criteria and received induction with FLAG-IDA at the Princess Margaret Cancer Centre or CPX at the 10 participating academic centres during the study period. A total of 163 patients were included (73 CPX; 90 FLAG-IDA). Propensity score matching (PSM) using age, cytogenetic complexity and TP53 mutation status yielded 43 matched pairs (n = 86). Patients were stratified according to European LeukemiaNet (ELN) 2022 risk categories and disease subtypes, including recurrent genetic abnormalities, TP53 mutations, MR gene or cytogenetic features and AML not otherwise specified (NOS).10 Mutations were also grouped by functional pathways, including DNA methylation, signalling, transcription factors, chromatin remodelling, spliceosome and cohesin complexes.11 The median age was 63 years (interquartile range [IQR] 56–69), with 61% male. The median number of somatic mutations was two (range 0–8), most commonly TP53 (n = 36), DNMT3A (n = 27), TET2 (n = 25), RUNX1 (n = 24) and ASXL1 (n = 23). As shown in Table 1, FLAG-IDA patients were younger and had higher risk cytogenetics; baseline imbalances were addressed using propensity score matching. Median follow-up was slightly shorter with CPX (269 vs. 288 days, p = 0.015). In the matched population, 60-day mortality was similar between groups: 2.3% (1/43) for CPX versus 4.6% (2/43) for FLAG-IDA (p = 0.55). CR rates were 69% (29/42) with CPX and 81% (34/41) with FLAG-IDA (p = 0.20). Allo-HSCT was performed in 40.5% (n = 17) and 46.5% (n = 20) of patients respectively (p = 0.58). Among CR patients proceeding to transplant, median time to HSCT was longer with CPX: 99 days (76–154) versus 78 days (47–107) with FLAG-IDA (p = 0.08). Median OS was 453 days (95% CI 235–581) for CPX and 397 days (95% CI 237–882) for FLAG-IDA (p = 0.99). One-year OS was comparable: 54.8% (95% CI 37.2–69.3) for CPX and 52.8% (95% CI 36.3–66.8) for FLAG-IDA. Outcomes improved following HSCT, with 1-year OS of 77.8% (95% CI 47.0–92.7) for CPX and 67.7% (95% CI 41.6–84.2) for FLAG-IDA (p = 0.13; Figure 1). The apparent disconnect between FLAG-IDA's higher CR rate and its lack of OS advantage suggests that post-remission factors may be key determinants of long-term outcomes. Despite comparable transplant rates, fewer FLAG-IDA patients in CR proceeded to HSCT (16/28: 57.1% vs. 17/34: 50%, p = 0.57). Infections differed significantly: Invasive fungal infections (IFI) occurred in 42.3% with FLAG-IDA versus 17.6% with CPX (p = 0.036), potentially impairing transplant readiness. Post-induction recovery patterns reflected expected regimen-specific kinetics. Median hospitalization was longer with CPX: 36 days (31–44) vs. 30 days (27–35) with FLAG-IDA (p = 0.006). Median WBC recovery was 26.5 vs. 31 days, and platelet recovery was 35 vs. 27 days for CPX and FLAG-IDA respectively (p = 0.65 and 0.39). IFI rates—encompassing suspected, probable and confirmed cases—were significantly lower with CPX (17.6% vs. 42.3%, p = 0.036), potentially facilitating a smoother transition to transplant. Among non-transplanted CR patients, relapse occurred in 28% (7/25) with CPX and 34.8% (8/23) with FLAG-IDA (p = 0.61). Median time to relapse trended longer with CPX: 142 days (51–264) vs. 76.5 days (63.5–105) with FLAG-IDA (p = 0.14). The shorter remission duration in the FLAG-IDA cohort, despite higher CR rates, raises questions about remission depth; however, inconsistent MRD data limited further insight. Although differences did not reach statistical significance, higher IFI rates, fewer transplants in CR++ and earlier relapse may have mitigated FLAG-IDA's apparent CR advantage. Conditioning intensity at HSCT varied significantly, with 58.8% (10/17) of CPX patients receiving myeloablative conditioning (MAC) versus 5% (1/20) of FLAG-IDA patients (p < 0.01), likely reflecting physician assessment of post-induction fitness. In multivariate analysis adjusting for induction therapy and conditioning regimen, neither was independently associated with OS (HR 1.44 [95% CI 0.36–5.68], p = 0.60 for CPX vs. FLAG-IDA; HR 4.76 [95% CI 0.55–40.9], p = 0.15 for MAC vs. Reduced-Intensity Conditioning [RIC]). Allo-HSCT was associated with a substantial survival benefit, with a 1-year OS of 75.3% (95% CI, 56.5–86.9) in transplanted patients compared to 37.9% (95% CI, 23.4–52.3) in non-transplanted patients. Median OS was 588 days (362–NR) versus 266 days (149–581) for non-transplanted patients (p = 0.0014). The HR for death with Allo-HSCT was 0.39 (95% CI 0.21–0.71; p = 0.002), indicating a 61% reduction in mortality risk. Among transplanted patients, CPX showed a trend towards higher 1-year OS (77.8% vs. 67.7%, p = 0.13). The 1-year CIR post-HSCT was 32.3% (95% CI 16.0–58.4) for CPX versus 44.9% (95% CI 25.8–71.3) for FLAG-IDA (p = 0.85). No baseline clinical or molecular characteristics—including age, sex, ELN-2022 risk or disease subtype—predicted differential OS. TP53-mutated patients had uniformly poor outcomes (1-year OS 16.7% vs. 44.4%, p = 0.65). Among DNMT3A-mutated cases, CPX showed a trend towards improved 1-year OS (80.0% [95% CI 6.0–91.0]) vs. FLAG-IDA (46.2% [95% CI 19.2–69.6], p = 0.06). Similar patterns were observed for TET2 (66.7% vs. 52.1%, p = 0.49), ASXL1 (64.8% vs. 58.3%, p = 0.51) and RUNX1 (68.6% vs. 25.9%, p = 0.13). Pathway analyses did not reveal significant differences. Real-world practice often diverges from controlled trial conditions. Off-label use and extrapolation of newer agents to untested populations are common but not always feasible or appropriate across health systems. FLAG-IDA is often preferred over standard 3 + 7 for fit patients with high-risk or sAML. Evidence from the MRC AML15 trial supports this approach, showing that patients with sAML, FLAG-IDA achieved superior 5-year overall and relapse-free survival compared with DA/ADE, suggesting that the conventional ‘3 + 7’ may not represent the optimal comparator in this population.12 We note that both CPX-351 and FLAG-IDA are increasingly combined with venetoclax,13 but our study focuses on settings where venetoclax is not available, highlighting the real-world applicability of these regimens without it. As azacitidine–venetoclax remains approved only for patients aged ≥75 or unfit for intensive therapy, younger fit sAML patients in our setting are typically treated with either CPX or FLAG-IDA. Our findings demonstrate that CPX and FLAG-IDA achieve comparable remission and survival outcomes in sAML, with long-term prognosis primarily determined by successful Allo-HSCT, consistent with German real-world data, which also highlights post-CPX transplant outcomes.14 While FLAG-IDA yielded slightly higher CR rates, this did not translate into superior survival, likely due to higher infectious morbidity and reduced transplant conversion. Conversely, CPX was associated with lower infection rates, more durable remissions and equivalent survival. Mutation-specific subgroup analyses of OS are limited by small patient numbers, resulting in wide confidence intervals; therefore, these findings should be interpreted cautiously and considered hypothesis-generating rather than definitive. Exploratory trends suggested potential benefit with CPX among patients harbouring DNMT3A, TET2, ASXL1 or RUNX1 mutations—mutations often associated with antecedent MDS or clonal haematopoiesis biology—supporting observations from the UK AML19 trial. In summary, this multicentre, propensity-matched analysis demonstrates that CPX and FLAG-IDA yield equivalent efficacy and survival in secondary AML, with allo-HSCT as the principal determinant of outcome. Differences in toxicity and mutation-associated trends underscore the need for individualized treatment and prospective, molecularly guided strategies for this high-risk population. M.A.P. is the corresponding author. The study was designed by C.A., H.S. and M.A.P. Data analysis and manuscript writing were performed by M.A.P. All authors contributed patient data, reviewed the manuscript and approved the final version for submission. The authors declare no competing interests. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.