HR, hazards ratio; CI, confidence interval; WBC, white blood cell; SCT, stem cell transplantation; CR, complete remission; GO, gemtuzumab ozogamicin; EMI, extramedullary infiltration; OS, overall survival; EFS, event-free survival.

In the current study, we assessed the clinical and prognostic characteristics of pediatric AML patients who exhibited EMI at different time points. Out of the 713 pediatric AML patients, 123 (17.3%) were diagnosed with EMI at initial diagnosis, which is consistent with previous reported⁴. However, certain patients with EMI may remain asymptomatic or lack bone marrow involvement, and the routine use of lumbar puncture, incorporation of flow cytometry as a part of cerebrospinal fluid assessment, and variations in the type and frequency of imaging examinations could impact the detection rate of EMI^12,19-21.

In this cohort, patients with EMI, both at diagnosis and at relapse, were more commonly found to be ≤2 years old and M5 morphology. Conversely, only patients with EMI at relapse exhibited a higher prevalence of WBC count ≥100×10⁹/L. These factors have also been demonstrated to be correlated with EMI in previous studies^8-11,22,23. Donna L. Johnston et al. conducted a study involving 886 pediatric patients and observed that individuals experiencing CNS relapse exhibited a higher prevalence of age <2 years, FAB-M5 subtype, and higher white blood cell (WBC) counts²⁴. In the realm of cytogenetics and molecular biology, a comprehensive investigation involving 315 children participating in the NOPHO-AML 2004 trial revealed an association between extramedullary infiltration (EMI) and 11q23/MLL(KMT2A ) rearrangements, which corroborated our own findings¹¹. Another study, comprising 240 cases registered with the Japanese Childhood AML Cooperative Study Group, identified a higher prevalence of inv16 and 11q23 abnormalities among patients with EMI¹⁰. Chromosome 11 abnormalities were also found to be related to CNS relapse in pediatric AML patients²⁴. In the context of pediatric low-risk AML, Guan-hua Hu et al. observed that abnormalities such as t(8;21), t(1;11), and c-KIT mutation were linked to EMI²⁵. Our investigation indicated that patients presenting with EMI, both at diagnosis and at relapse, exhibited a higher incidence of KMT2Arearrangements and a lower incidence of normal karyotype. NPM1and FLT3-ITD mutation was found to be associated with EMI in adult AML patients while CEBPA mutation decreased the occurrence of EMI^5,14. In this cohort, the mutation of theCEBPA showed unrelated to the presence of EMI. Intriguingly, we observed that patients with EMI at the initial diagnosis presented a lower frequency of FLT3 ITD-/NPM1+ , whereas EMI relapse patients displayed a lower frequency of FLT3-ITD mutation. Further research is needed to explore the cytogenetic and molecular biological characteristics in pediatric AML patients with extramedullary infiltration.

Although EMI is typically considered a presentation of advanced disease, there is still no consensus on the prognosis of EMI in pediatric patients. In the cohort, patients with EMI at initial diagnosis had a low rate of CR1 and a higher rate of relapse. For initially diagnosed patients, EMI appeared to independently predict both shorter OS and EFS. Sorts of researches insisted our results. In the NOPHO-AML 2004 trial, pediatric patients with EMI at diagnosis had significantly lower 5-year OS (64% vs. 73%, p=0.04) but not 5-year EFS (54% vs. 45%, p=0.57) and a higher risk of induction death (8% vs. 1%, p=0.002)¹¹. Xiaoli Xin et al. also found that EMI at initial diagnosis [HR 3.313, 95% CI 1.748-13.664, p <0.001] was an independent risk factors affecting the prognosis¹⁴. The appearance of EMI at diagnosis was also proven to be a significant adverse factor in pediatric RUNX1-RUNX1T1 (+) AML by Jae Wook Lee et al.¹⁵. Guanhua Hu et al. found myeloid sarcoma instead of CNS leukemia was a predictor of adverse prognosis for RFS and OS in low-risk pediatric AML²⁵. However, sorts of clinical trials have described no significant association between extramedullary infiltration and outcome in pediatric AML patients^9,10. When analyzing the outcomes of different subgroups of EMI, Donna L. Johnston et al. collected data from patients who participated in CCG trials 2861, 2891, 2941 and 2961 and found patients with orbital-MS and CNS-MS experienced better survival compared to patients with other types of EMI, or without EMI¹⁶. Another study adopted CCG AML trials 213 or 213P, 2861 and 2891 revealed that the non-skin EML group had the best outcome compared to the skin EML group and the no EML group¹². In the cohort, patients with myeloid sarcoma at diagnosis showed similar EFS and OS compared to those with exclusive CNS involvement at diagnosis. Due to the limitation of available data, we were unable to further explore the impact of myeloid sarcoma at different site on prognosis.

Bone marrow is the most common site for relapse, while extramedullary relapse also accounts for a considerable proportion²⁶. Without the limitation of age, the prognosis of extramedullary relapse in AML patients is controversial. Most previous studies have indicated that the survival of patients who developed extramedullary relapse was slightly better than that of patients who developed bone marrow relapse^26-28. There is a notable lack of studies focusing on pediatric patients. A retrospective study enrolled 1527 acute leukemia patients (983 ALL and 544 AML) by Volkan Hazar et al. and found that post-HSCT patients with isolated extramedullary relapse (without bone marrow involvement) remained poor prognosis but had slightly better survival than those who developed bone marrow relapse⁷. According to our results, patients with EMI at relapse had similar overall survival compared to those without EMI. This result indicated that the presence of EMI may not contribute to a worse prognosis in patients experiencing relapse. However, when further analyzed the prognosis of different types of EMI, we observed that relapse patients with MS showed a significant shorter OS than those with exclusive CNS involvement. Further studies requiring more cases of EMI relapse are essential to validate the findings and investigate whether the difference in prognosis is associated with special clinical characteristics.

HSCT is a pivotal and widely utilized treatment modality in the context of pediatric AML, but it also carries the risk of additional treatment-related mortality due to graft-versus-host disease, infection, and organ toxicity. Identifying the specific patient population that will derive benefits from HSCT is of paramount importance^29,30. There is limited research exploring the mutual prognostic significance of stem cell transplantation and EMI in AML patients, especially in pediatric patients. A report of 51 patients (both children and adults were adopted) from the SFGM-TC registry showed that allo-HSCT is a potentially efficient therapy for myeloid sarcoma, with a notable proportion of patients achieving long-term remission³¹. On contrast, a retrospective study enrolled multi-center clinical trials showed that allo-HSCT did not improved the survival⁵. A previous study by Lu‑Hong Xu et al. also used data from the TAGART dataset and found that stem cell transplantation did not improve either OS or EFS in patients with myeloid sarcoma³². Our studies included patients with myeloid sarcoma as well as those with CNS involvement, and we obtained similar results. Given the limitations in the number of EMI-positive patients who had also underwent SCT, further investigation is needed to assess the prognostic association of between EMI and SCT.

Gemtuzumab Ozogamicin is an antibody-drug conjugate (ADC) composed of an anti-CD33 monoclonal antibody and a cytotoxic agent N-acetyl gamma calicheamicin³³. The incorporation of GO in the induction treatment was proven to improve the RFS and EFS in AML³⁴. Taofeek Owonikoko et al. reported a case involving GO therapy for a 19-year-old AML patient with isolated extramedullary relapse after HSCT³⁵. This patient experienced extramedullary relapse at multiple sites, and radiation therapy showed no discernible effect. However, extramedullary remission was achieved after GO treatment. Regarding patients with EMI at initial diagnosis, we evaluated the prognosis between those who underwent GO treatment or not and found possibly significant difference in OS. The immunohistochemical stains in myeloid sarcoma reveled a positivity rate of CD33 ranging from 55% to 94%^36,37. GO emerges a potential treatment for extramedullary AML, particularly in cases with CD33 positive upon biopsy.

In conclusion, EMI at diagnosis independently predicts a worse prognosis. While for patients experiencing relapse, the presence of EMI does not significantly impact survival. Moreover, SCT does not lead to a prognosis improvement for patients with EMI at initial diagnosis, whereas GO treatment has the potential to prolong overall survival. This study enhances our comprehension of EMI in pediatric AML patients and demonstrates the potential significance of GO treatment for AML with EMI. With the advancement of flow cytometry and imaging techniques, the diagnostic rate of EMI is expected to increase, and further exploration of treatment approaches is needed to improve the survival of EMI patients.