The database of patients with pathologically confirmed brain tumors aged 1 year and younger was queried retrospectively from January 2008 to June 2023 in the clinical data warehouse at the author’s institution. We found 70 patients with infantile brain tumors. Excluding one patient who underwent the first surgery at another institution, 69 patients were finally included in the analysis. Twenty-eight patients had congenital brain tumors, defined as tumors first identified before 3 months of age in this study.

The initial diagnosis was the time when brain tumors were first identified on magnetic resonance imaging (MRI). If there was neurological impairment, hydrocephalus, or rapid tumor growth on more than two sequential imaging (ultrasonography or MRI), surgery was performed as soon as possible, and if not, surgery was performed as an elective operation. The age, body weight, and tumor volume were measured before surgery. Tumor volume was measured using the function of volumetric analysis on MRI. Leptomeningeal seeding (LMS) was evaluated on preoperative brain and spinal MRI or microscopic cerebrospinal fluid analysis (n=67).

All patients underwent surgery, and resected tumors were diagnosed according to the 2021 World Health Organization (WHO) classification of tumors of the CNS [14]. Surgery was performed by four skilled pediatric neurosurgeons who were trained at the same institution. The treatment protocol and surgical skills are relatively well-standardized and considered not significantly different among surgeons. Tumor grade was classified according to the CNS WHO (n=52) or Norris (n=10) grading systems [15]. Germ cell tumors (GCTs; n=10) followed the Norris grade. Histiocytic tumors (n=2) and soft tissue tumor (n=1) did not follow both grading systems. Low-grade tumors referred to tumors of CNS WHO grade 1, 2, and Norris grade 0, 1, and high-grade tumors referred to tumors of CNS WHO grade 3, 4, and Norris grade 2, 3. Immature teratoma with Norris grade 2 and 3 were classified as high-grade because they showed low OSs and were warranted for adjuvant therapy [16,17].

The extent of tumor resection was classified as gross total resection (GTR; 100% resection), near-total resection (NTR; 90%-99% resection), subtotal resection (50%-90% resection), and partial resection (PR; < 50% resection) on postoperative MRI (n=68) [18,19]. The postoperative MRI was acquired within 48 hours after surgery.

The delayed surgery referred that the surgeon intended to delay surgery, the interval between the diagnosis and surgery was more than one month, and there were more than two follow-up imaging (ultrasonography or MRI) before surgery (n=14). Others were defined as early surgery (n=55).

Outcomes were assessed as in-hospital mortality and early mortality to evaluate the risk of early surgery in congenital brain tumors. In-hospital mortality was defined as expiration occurring during the period of hospitalization following surgery. Early mortality was defined as expiration occurring within 30 days following surgery, which reflected surgery-related mortality. Outcomes of recurrence-free survival (RFS) and OS according to histopathology, tumor grade, location, LMS, and the extent of resection were analyzed for infantile brain tumors.

Graphs were constructed using the ggplot2 function in R (R Foundation for Statistical Computing, Vienna, Austria). The receiver operating characteristic (ROC) analyses were performed to determine cut-off values of age and body weight to avoid early mortality using the ROC function in R. An area under the curve (AUC) of < 0.50 is considered worthless, 0.60-0.69 poor, 0.70-0.79 fair, 0.80-0.89 good, and 0.90-1 excellent. Kaplan-Meier curves were constructed using survival and survmine functions in R. RFS and OS rates were reported at 1, 2, and 3 years after surgery with 95% confidence intervals. Hazard ratios (HRs) of mortality were estimated from univariate and multivariate Cox proportional hazard survival analyses. Statistical analyses were performed by R, and a p-value less than 0.05 was considered statistically significant.

Among 69 patients with infantile brain tumors, there were 44 males and 25 females (Table 1). The median ages at diagnosis and operation were 5 months and 5.5 months. Increased intracranial pressure (n=23) was the most frequent initial presentation, followed by incidental findings (n=15), increased head circumference (n=11), and seizure (n=10). The median follow-up period was 27 months after the operation.

In infantile brain tumors (Fig. 1, S1 Table), gliomas, glioneuronal, and neuronal tumors (n=22, 31.4%), and embryonal tumors (ETs; n=22) were the most common categories, followed by choroid plexus tumors (CPTs; n=11, 15.7%), and GCTs (n=10, 14.3%). As a single pathology, atypical teratoid/rhabdoid tumor (ATRT) was the most common tumor (n=14, 20%), followed by pilocytic astrocytoma (n=7, 10%), immature teratoma (n=7), medulloblastoma (n=6, 8.6%), and choroid plexus carcinoma (n=6).

As for age distribution (Fig. 2, S1 Table), GCTs were diagnosed at exclusively early ages, which were all diagnosed before six months of age, and 80% were diagnosed before three months of age. ETs (32.1%) were the most common tumors diagnosed before three months of age. Before 3 months of age, ATRT (n=6) was the most common tumor, followed by immature teratoma (n=5), mature teratoma (n=3), circumscribed astrocytic gliomas (n=2), and pediatric-type diffuse high-grade gliomas (n=2). High-grade tumors (n=40) and supratentorial tumors (n=44) accounted for 65% and 63.8%. LMS (n=24) was identified in 35.8% of patients.

The median age, body weight, and tumor volume at the operation were 5.5 months, 7.9 kg, and 30.4 mL (Table 2). There were six in-hospital mortality (8.7%) and four early mortality (5.8%) following surgery. We constructed scatter plots and curves of body weight (kg) and tumor volume (mL) according to patients’ ages (months) before surgery (S2 Fig.). The body weight tended to increase with age, but the variation in tumor volume was independent of age. Early mortality appeared to occur in low age and body weight, ranging from 1.07 to 1.27 months and from 2.72 to 4.9 kg (S3 Table). Intraoperative estimated blood loss ranged from 3,700 to 4,000 mL, and massive transfusions were performed. All four patients underwent intraoperative cardiopulmonary resuscitation (iCPR). Three patients achieved the return of spontaneous circulation, but they expired shortly after being transferred to the intensive care unit. One patient successfully recovered spontaneous circulation, but the residual tumor progressed rapidly, making it unavoidable to undergo reoperation one month later. Unfortunately, cardiopulmonary resuscitation (CPR) occurred again during reoperation, and the patient succumbed to the bleeding.

The ROC curves were generated based on patients’ age and body weight (S4 Fig.), and the ROC analyses were performed to determine the cut-off values of age and body weight to avoid early mortality (Table 3). Cut-off values were found to be 1.3 months and 5.2 kg, with AUC of 96.9 (p < 0.001) and 95.8 (p < 0.001).

We next constructed scatter plots based on patients’ ages at diagnosis and operation, and intervals between diagnosis and operation (Fig. 3). Categories represented the timing of surgery as early (n=55) or delayed (n=14), and red dotted circles represented the patients corresponding to early mortality. In the delayed surgery group, the latest age for diagnosis was 2 months, and one patient showed early mortality (7.1%). This patient (patient 1 in S3 Table) showed an excessively large and rapidly growing tumor, and surgery could not be delayed any longer than 1 month. He underwent surgery at a weight of 2.9 kg and the age of 1.07 months with a tumor volume of 953.3 mL. On the other hand, early surgery was performed in 55 patients, and 10 patients were ≤ 2 months of age at diagnosis. Three patients showed early mortality (3/10, 30.0%) in the early surgery group (≤ 2 months of age at diagnosis). In the delayed surgery group, early mortality occurred in one exceptional case, and there was no early mortality when the age at surgery was over 1.3 months (Fig. 3B). This suggests the benefit of delayed surgery in congenital brain tumors, except for exceptionally large or rapidly growing tumors.

During follow-up, tumors recurred in 21 patients (30.4%) (Table 2). The mean time to recurrence after surgery was 100.6 months (S5 Fig.). The 1-year and 3-year RFS rates were 68.4% and 63.8%. There were statistically significant associations of histopathology (p=0.037), tumor location (p=0.014), and LMS (p < 0.001) with RFS.

Twenty patients (29.0%) expired during follow-up (Table 2). The mean OS time after surgery was 106.1 months (Fig. 4). The 1-year, 2-year, and 3-year OS rates were 89.6%, 78.3%, and 66.9%. There were statistically significant associations of histopathology (p=0.002), tumor grade (p=0.003), location (p=0.040), and LMS (p < 0.001) with OS. The glioma, glioneuronal, and neuronal tumor group showed no mortality.

Univariable and multivariable Cox proportional hazard survival analyses were performed for infantile brain tumors (Table 4). On univariate analysis, age at diagnosis, tumor grade, location, LMS, and the extent of resection were significantly correlated with mortality. Younger patients (< 3 months of age; HR, 4.1; p=0.002), high-grade tumors (HR, 11.0; p=0.018), infratentorial tumors (HR, 2.5; p=0.048), and tumors with LMS (HR, 10.0; p < 0.001) showed higher mortality. The NTR and PR groups presented higher mortality than the GTR group (HR, 3.7; p=0.026 and HR, 4.1; p=0.027, respectively). On multivariable analysis, age at diagnosis, LMS, and the extent of resection (PR) remained the independent risk factors for mortality. Younger patients (< 3 months of age) with LMS showed higher mortality than older patients (≥ 3 months of age) without LMS (HR, 7.1; p=0.018 for age at diagnosis, and HR, 30.6; p=0.002 for LMS). The PR group presented higher mortality than the GTR group (HR, 14.8; p=0.024).

In the American population, gliomas and ETs are the most common in patients who are under 19 years of age [1,7,20,21]. Glioma and ETs are predominant in infants as in the whole pediatric population. However, ETs and CPTs characteristically show peak prevalence in infants [1,21]. Our results also show that gliomas and ETs are the most common. However, there is a relatively higher prevalence of GCTs (10%) compared to that of the United States (less than 4%), demonstrating the well-known characteristics of the East Asian population [1]. Medulloblastoma (6.6%) is over four times more prevalent than ATRT (1.5%) in the whole pediatric population [20,21]. In infantile ETs, ATRT accounts for the highest proportion (63.6%), which is over 2.3 folds of the proportion of medulloblastoma (27.3%). In pediatric intracranial GCTs, germinoma is the most prevalent, which accounts for more than 50% of intracranial GCTs [22-24]. Teratoma and yolk sac tumors are known to be prevalent at infantile ages [25], and our results show that only teratoma is identified, which is the characteristic feature of infantile GCTs.

High-grade tumors are prevalent in ETs, CPTs, and GCTs. All ETs are malignant tumors of CNS WHO grade 4. High-grade tumors account for 54.5% and 70% each in CPTs and GCTs. The high prevalence of ETs and CPTs in infants and the high prevalence of GCTs in East Asian infants contribute to high mortality rates and challenges in treatments for infantile brain tumors.

On the other hand, there were tumors whose diagnoses were unclear according to the latest CNS WHO classification. Although ‘medulloblastoma, histologically defined’ is a still-existing diagnosis, ‘molecularly defined medulloblastoma’ emerged from the previous version of the classification based on methylation and transcriptome profiling. One of the most noticeable changes in the latest classification is the change in the classification of gliomas based on the state of the gene fusion and the methylation profile [14,22]. ‘Glioblastoma’ and ‘anaplastic astrocytoma’ no longer exist as pediatric tumors. Therefore, we revised past diagnoses of infantile brain tumors according to the new classification. ‘Glioblastoma’ was rediagnosed as ‘diffuse pediatric-type high-grade glioma, H3-wildtype and IDH-wildtype,’ and ‘anaplastic astrocytoma’ was rediagnosed as ‘infant-type hemispheric glioma’.

The difference in prevalence between pediatric and infantile brain tumors suggests that infantile tumors have different biological features compared to pediatric tumors. Ethnic differences imply differences in genetic backgrounds. Changes in the diagnostic classification over time reflect the importance of genetic and epigenetic features of tumors.

We first evaluate the role of early and delayed surgery on congenital brain tumors. In previous reports, the major causes of surgery-related early mortality were massive hemorrhage, myocardial infarction, severe arrhythmia, and pulmonary embolism [26]. Intraoperative bleeding was the most common factor among non-cardiac causes (52.2%). Since body fluid and blood volume are proportional to body weight, patients with low body weight have small amounts of blood volume. If patients with lower body weights experience bleeding during surgery, they are more susceptible to hypovolemia. Estimated mean circulating blood volume in full-term infants is known as 84-91.7 mL/kg [27]. Calculating with Korean standard growth data [28], circulating blood volume is 277.2-302.6 mL for males and 268.8-293.4 mL for females in newborns, and 378-412.7 mL for males and 352.8 -385.1 mL for females in 1-month-old infants. In our study, four patients who showed early mortality were around 1 month of age, and intraoperative blood loss ranged from 3,700 to 4,000 mL, which was almost 10-folds of the estimated circulating blood volume, leading to iCPR. Infants with congenital brain tumors are susceptible to hypovolemic shock while removing hypervascular tumors, and they are prone to coagulopathy and pulmonary edema after massive transfusion, which contributes to the high risk of early mortality.

To avoid this fatal event, our results can guide the timing of surgery on congenital brain tumors. Although the tumor volume showed inconsistent results, it cannot be excluded from the decision-making process for the timing of surgery. If excessively enlarged or rapidly growing tumors involve core structures and pose a risk of severe neurological injury, delayed surgery is not feasible, and the surgeon’s decision on early surgery is necessary. If a tumor does not cause severe neurological impairment without rapid growth, delayed surgery with short-term image follow-up is recommended until the patients’ age and body weight exceed the cut-off values.

In a previous study on pediatric brain tumors (age ≤ 14 years), the 1-year and 2-year OS rates were 89.9%-90.0%, and 79.2%-79.4% [7]. In our study, the 1-year, 2-year, and 3-year OS rates were 89.6%, 78.3%, and 66.9%, which showed comparable results in other reports. In another study, infants exhibited lower OS compared to the other pediatric ages at yearly intervals of up to 5 years [7]. Five-year OS for all tumors were 36.7% (< 3 months), 56.0% (3-5 months), 63.8% (6-11 months), and 74.7% (1-19 years). Patients under 3 months of age especially showed the lowest OS as in our results. Younger patients have a higher proportion of malignant brain tumors [20]. The surgery-related mortality is higher in congenital brain tumors. Chemotherapy and radiotherapy are also limited in young patients.

By multivariable analysis, younger age (< 3 months of age) at diagnosis, the presence of LMS, and the extent of resection (PR) are independent risk factors for the high mortality in infantile brain tumors. Concerning histopathology, glioma, glioneuronal, and neuronal tumors showed better survival than other categories with no mortality. However, there were no differences in survival among ETs, CPTs, and GCTs. Because each histopathological category contains various pathologies with heterogeneous malignancies, the results can be inconsistent. Especially, in the category of glioma, glioneuronal, and neuronal tumors, there are not only various benign glioneuronal tumors but also highly malignant glial tumors. Therefore, individual pathologies rather than categorical diagnosis appear to be more important. Also, among 11 patients with PR, all four patients who underwent iCPR are included, which contributes to a high mortality rate. Tumors could not be properly removed in those patients due to iCPR, and four patients subsequently expired after CPR. Therefore, in this group, the impact on mortality is considerably related to surgery rather than merely the extent of tumor resection. On the other hand, LMS is a classical poor prognostic factor associated with high mortality in brain tumors [29]. Therefore, despite the development of chemotherapy, radiotherapy, and targeted therapy, treatment for younger patients with LMS remains a challenging condition. Conquering these diseases is a lifelong challenge for mankind.

Our study had some limitations. First, this study was a retrospective analysis. Also, the timing of surgery was not constant among patients, and the number of patients who showed early mortality was small. Further studies require more patients with precise treatment strategies.