Collaborative groups in the United Kingdom, the United States of America, and Germany have used a three-drug induction regimen in their ongoing research protocols or recently published studies (Supplemental Table S13). The primary objective of our study was to compare the EFS between a 2-drug DA induction regimen with a 3-drug ADE induction regimen. We wanted to evaluate if adding a third drug to 2-drug induction improved survival outcomes and if it was associated with more toxicities and increased supportive care requirements. Knowing this is important in resource-constrained settings in LICs and LMICs where treatment of pediatric AML is challenging due to the high prevalence of gram-negative multi-drug resistant (MDR) bacterial infections, financial constraints, and restricted supportive care.

The spectrum of bacterial infections in LICs and LMICs is different from HICs [18]. Gram-negative and MDR infections are more common among patients with cancer in LICs and LMICs [18, 19]. In contrast, gram-positive infections are more common in HICs [20]. A high proportion of patients with acute leukemia in LMICs are colonized with gram-negative MDR bacteria in stool at admission, indicating that these were community-acquired [21, 22]. MDR gram-negative bacterial infections are associated with a high incidence of mortality [23, 24]. Twenty out of 25 (80%) induction mortality in our study was due to sepsis (Table 3). Twenty-five out of 35 (71%) positive blood cultures during the induction period in the study were due to gram-negative bacteria, among whom 10 (40%) had MDR bacteria (Table 3). Six out of 10 (60%) patients with MDR gram-negative bacteremia died. Addressing gram-negative infections and MDR bacteria is vital to improving survival in pediatric AML in LICs and LMICs.

Our study did not find a significant difference in EFS between the DA and ADE arms (5-year EFS: 34.4% and 34.5%, respectively, p = 0.66). There were no significant differences between the ADE and DA arms regarding toxicities and supportive care requirements. However, since we did not achieve our planned accrual, our study is not adequately powered to confirm our findings.

Options for treatment after relapse in pediatric AML are limited in developing countries due to the high cost of salvage chemotherapy, allogeneic HSCT, and limited donor availability. Though HSCT in the first CR was advised for intermediate and high-risk patients, only 6% underwent allogeneic HSCT. The participating centers performed only matched sibling donor HSCT until 2018, limiting the number of patients receiving HSCT. Other factors related to meager rates of HSCT were the high cost of salvage chemotherapy and HSCT, protracted treatment course and follow-up prohibiting long-term commitment from the family/caregiver, family’s apprehension of the risk to the donor, and overall transplant outcomes, besides other socioeconomic issues. Only 21 out of 59 (35.5%) patients who relapsed in the study received chemotherapy at relapse. The high cost of salvage chemotherapy at relapse and HSCT was the most common reason why most patients at relapse did not undergo HSCT in our study. Access to HSCT is limited to most patients with AML in LMICs and is not unique to our study. The participating centers in the trial are routinely performing haploidentical HSCT and matched unrelated donor transplants since 2018; this has increased the number of patients eligible for HSCT [25]. The inclusion of HSCT under state insurance has also reduced the cost, improving access to the procedure.

The induction mortality, the CR rates, and the 5-year EFS and OS reported from our study are inferior to the results from trials from HICs (Supplemental Table S13). Ninety percent of children with cancers are treated in LICs and LMICs, and our results reflect the reality for most children with AML [26]. A recent systematic review of literature on outcomes of pediatric AML in LMICs included 27 studies, 26 retrospective, and one prospective observational study [7]. The 5-year EFS varied from 24% to 63%, and the 5-year OS from 10% to 72% [7]. The authors concluded that outcomes of pediatric AML in LMICs are substantially inferior compared to HICs [7]. There is heterogeneity in the outcomes reported from LMICs. The inferior outcomes in LMICs are due to high abandonment rates, early deaths, treatment-related mortality, and limited treatment options after relapse [7].

Thirty-nine percent of the study patients were undernourished at diagnosis, and 1.3% were obese. On univariate analysis, undernutrition at diagnosis was associated with inferior EFS and OS in the whole cohort and the ADE arm but not in the DA arm (Supplemental Tables S5 and S6). On multivariate analysis, undernutrition at diagnosis was one of the factors associated with inferior OS (Supplemental Table S8). In contrast, in the NOPHO-AML 2004 study, only 5% of patients were undernourished [27]. In the NOPHO-AML 2004 study, there was a trend toward better overall survival in obese children (23% of patients) above ten years [27]. In the Children’s Cancer Group 2961 study, 10.9% were underweight, and 14.8% were overweight. Survival was inferior, and treatment-related mortalities were higher in underweight and overweight patients than in normal-weight patients [28]. Aggressive nutritional support and tailoring cytotoxic treatment to the patient’s nutritional status might improve the overall outcomes in pediatric AML in LICs and LMICs.

Sixty-five patients screened for enrollment in the study were not eligible as they were not fit for intensive chemotherapy. This subgroup equals 44% of patients enrolled in the trial. Many patients in LICs and LMICs present late and in moribund condition. Most are malnourished and have ongoing infections or organ dysfunction requiring intensive care. Hence, they cannot be given intensive chemotherapy due to the high risk of mortality from chemotherapy [29].

Patients who achieved CR after the first induction in our study had significantly better survival than those who did not. This was observed in the whole cohort and those who received the ADE regimen. Whether a second induction is necessary after achieving CR with the first induction needs to be explored. Omitting second induction and replacing it with HIDAC consolidation in pediatric AML can reduce the risk of cardiac toxicity and second malignancies due to a reduction in anthracycline and etoposide cumulative dose. Patients who do not achieve CR after the first induction should be considered for intensive chemotherapy followed by allogeneic HSCT, as our study shows that they have inferior outcomes even if they achieve CR after the second course of induction, especially in the ADE arm.

Based on our sub-group analysis, an etoposide-based three-drug induction might be more effective in patients less than ten years of age, with favorable cytogenetics, those with an extramedullary disease, and those who are not malnourished (Supplemental Tables S5, S6, S9 and S10).

We used etoposide as the third drug in our intervention arm as the ADE regimen is the most common three-drug induction used in pediatric AML trials (Supplemental Table 13) and routine clinical practice. Studies in adult AML have shown increased CR rates and survival with the addition of fludarabine or cladribine to induction chemotherapy regimens [10, 30, 31]. The AML-08 trial compared clofarabine and ara-C induction with ADE in pediatric AML [32]. The study showed increased MRD negative rates with clofarabine-based induction, however, this did not translate to improvements in EFS or OS [32]. The addition of fludarabine or cladribine or clofarabine and omitting etoposide in pediatric AML induction needs further evaluation.

The limitations of our study include the inability to achieve the desired sample size, the absence of MRD-based risk-stratification, and limited access to HSCT. The dose and schedule of daunorubicin and ara-C were different between the DA and ADE arms. We used a dose and schedule of DA and ADE regimens used in routine clinical practice and reported in the literature. However, our study is the first RCT on pediatric AML from LICs and LMICs. The study also challenges the dogma of three-drug induction chemotherapy in pediatric AML.

To conclude, no statistically significant difference in EFS, OS, CR, or toxicity between ADE and DA regimens in pediatric patients with AML was observed in our study. Future RCTs are needed to find the optimal number of induction chemotherapy drugs and cycles in pediatric AML.