Low-dose chemotherapy for induction in pediatric AML: Retrospective outcome analysis across expanded cohorts

Abstract Background: Our phase 3 randomized trial (CALSIII-AML18; n = 497) previously demonstrated non-inferior complete remission (CR/CRi) rates with low-dose chemotherapy (LDC) compared to standard-dose chemotherapy (SDC) in pediatric acute myeloid leukemia (AML) (ASH 2023). In this expanded retrospective analysis (n = 661), we evaluated induction response, long-term outcomes, toxicity, and subgroup-specific efficacy to define the clinical utility of LDC further. Patients and Methods: We analyzed 661 patients aged 18 years with newly diagnosed AML, treated at 12 centers in China between 2018 and 2024. Patients received either LDC (n = 325)—cytarabine 10 mg/m² subcutaneously every 12 hours for 10 days (20 doses), anthracycline 5 mg/m² intravenously on days 1, 3, and 5, plus G-CSF—or SDC (n = 336)—cytarabine 100 mg/m² intravenously every 12 hours for 10 days (20 doses), combined with daunomycin and etoposide. Post-remission therapy consisted of 2 (low-risk), 3 (intermediate-risk) chemotherapy blocks and hematopoietic stem cell transplantation (HSCT) for patients at high risk of relapse. The median follow-up was 40.4 months (range, 1–80). Endpoints included CR/CRi rates, measurable residual disease (MRD), 4-year event-free survival (EFS), overall survival (OS), cumulative incidence of relapse (CIR), toxicity, and subgroup outcomes. Results: Baseline clinical characteristics were similar between treatment arms. After Induction I, CR/CRi was achieved in 71.4% (LDC) vs. 71.7% (SDC); following Induction II, rates increased to 96.3% and 95.7%, respectively (P = .923 and P = .741). MRD 0.1% by flow cytometry was observed in 55.4% (LDC) vs. 61.3% (SDC) after Induction I, and 86.6% vs. 85.3% after Induction II (P = .139 and P = .667). Grade 3–4 toxicities were significantly lower in the LDC group during induction. Early mortality was also reduced with LDC (1.5% vs. 4.2%, P = .043). Four-year EFS was 63.5% (95% CI, 57.9 – 69.6) for LDC vs. 62.8% (95% CI, 57.5 – 68.5) for SDC (P = .684); 4-year OS was 83.5% (95% CI, 79.2 – 87.9) vs. 85.0% (95% CI, 81.1 – 89.1) (P = .740). CIR was comparable between groups (29.4% 95% CI, 24.0 – 34.9 for LDC vs. 25.4% 95% CI, 20.6 – 30.5 for SDC; P = .270). In multivariate analysis, adjusting for known prognostic factors, the treatment arm was not independently associated with outcome. Subgroup analysis revealed that patients 3 years of age had superior EFS with LDC compared to SDC (80.6% vs. 59.9%; HR 0.38 95% CI, 0.18–0.79; P = .009). Improved EFS with LDC was also seen in cases with NUP-rearrangement (68.3% vs. 21.4%; HR 0.18 95% CI, 0.06–0.53; P = .002) or WT1 mutations (63.2% vs. 31.5%; HR 0.47 95% CI, 0.24–0.91; P = .025). Conversely, patients with core-binding factor (CBF) AML and co-occurring KIT mutations had superior EFS with SDC (69.2% vs. 46.4%; HR 1.85 95% CI, 1.08–3.18; P = .026). Conclusions: This expanded retrospective analysis of the CALSIII-AML18 trial confirms that LDC achieves remission rates, survival outcomes, and relapse risk comparable to SDC, with reduced induction toxicity and early mortality. Notably, LDC appears particularly beneficial in younger children and those with NUP rearrangements or WT1 mutations. In contrast, SDC may offer greater benefit for patients with CBF AML harboring KIT mutations. These findings support a risk-adapted approach to induction therapy in pediatric AML.