Population-based analysis of CNS tumor diagnoses, treatment, and survival in congenital and infant age groups

Abstract

Background

Congenital (< 3 months) and infant (3 to 11 months) brain tumors are biologically different from tumors in older children, but their epidemiology has not been studied comprehensively. Insight into epidemiological differences could help tailor treatment recommendations by age and increase overall survival (OS).

Methods

Population-based data from SEER were obtained for 14,493 0–19-year-olds diagnosed with CNS tumors 1990–2015. Congenital and infant age groups were compared to patients aged 1–19 years based on incidence, treatment, and survival using Chi-square and Kaplan–Meier analyses. Hazard ratios were estimated from univariate and multivariable Cox proportional hazards survival analyses.

Results

Between the < 3-month, 3–5-month, 6–11 month, and 1–19-year age groups, tumor type distribution differed significantly (p < 0.001). 5-year OS for all tumors was 36.7% (< 3 months), 56.0% (< 3–5 months), 63.8% (6–11 months), and 74.7% (1–19 years) (p < 0.001). Comparing between age groups by tumor type, OS was worst for < 3-month-olds with low-grade glioma, medulloblastoma, and other embryonal tumors; OS was worst for 3–5-month-olds with ependymoma, < 1-year-olds collectively with atypical teratoid-rhabdoid tumor, and 1–19-year-olds with high-grade glioma (HGG) (log rank p < 0.02 for all tumor types). Under 3-month-olds were least likely to receive any treatment for each tumor type and least likely to undergo surgery for all except HGG. Under 1-year-olds were far less likely than 1–19-year-olds to undergo both radiation and chemotherapy for embryonal tumors.

Conclusions

Subtype distribution, treatment patterns, and prognosis of congenital/infant CNS tumors differ from those in older children. Better, more standardized treatment guidelines may improve poorer outcomes seen in these youngest patients.

Introduction

Congenital (< 3 months) and infant (3 to 11 months) CNS tumors are biologically different from tumors in older children, but epidemiology of these tumors has not been studied comprehensively [1]. Despite being less common, brain cancers recently overtook leukemia as the number one cause of childhood cancer fatalities. Congenital and infant brain tumors are typically aggressive in nature, difficult to see symptomatically, and place patients at a higher risk of early death after diagnosis [2–4]. However, survival has improved for pediatric brain tumors because of improved surgical techniques, the rational use of postoperative radiation and chemotherapy, and advances in molecular-based classification of tumor types [5]. Factors affecting prognosis in patients less than one year of age include tumor type, the size of the neoplasm at diagnosis, and a lack of appropriate age-based therapeutic approaches [2]. Infants less than one year of age present a management challenge for clinicians, partly due to hesitancy to offer treatment at the same level as older children due to concern for augmented adverse effects [1, 5]. It is therefore critical to understand the epidemiology of congenital and infant CNS tumors to determine populations in need.

While key survival differences have been extensively studied between pediatric and adult tumors, we set out to study these differences between congenital and infant age groups and older children. In this study, we use population-based data from the NCI’s Surveillance, Epidemiology, and End Results (SEER) registry to analytically compare demographic, clinical, and outcomes data between congenital, infant, and older pediatric CNS tumor groups.

Materials and methods

Data source

Incidence and survival data were collected from the SEER Program of the National Cancer Institute. The years included in this study, between 1990 and 2015, encompass varying groupings of participating registries, with population coverage ranging from 9% in 1990–1991 (SEER-9 registries) to 28% during 2000–2015 (SEER-18 registries) [6], which is the reason for the increased number of patients in later periods of the study. SEER*Stat Version 8.3.6 was used to access data from the specialized SEER database that contained information on months since last birthday and primary treatment based on registry data updated in November 2017.

Study population

Data were obtained from patients between the ages of 0–19 years diagnosed with a brain or other central nervous system (CNS) tumor between 1990 and 2015. These patients were identified using the extended classification scheme of the International Classification of Childhood Cancer, Third Edition (ICCC-3) site recodes III and X(a). [6]

Variable definitions

Patient demographics, diagnosis year, tumor type, and primary treatment were included as covariates. Months since last birthday were used to categorize patients age < 1 year at diagnosis into the congenital (< 3 months) and infant age groups (3 to < 6 months and 6 to < 12 months), while patients age 1–19 years were grouped together to serve as a comparison group to the congenital and infant cohorts. Other patient characteristics included sex, race (White/Black/Other), Hispanic/non-Hispanic ethnicity (derived from the North American Association of Central Cancer Registries Hispanic Identification Algorithm), and diagnosis year (1990–1994, 1995–1999, 2000–2004, 2005–2009, and 2010–2015). Using the ICCC-3 site recodes, SEER histology recodes of brain groupings, and tumor grade, tumor type was categorized as high-grade glioma (HGG), low-grade glioma (LGG), ungraded glioma (UGG), ependymoma, medulloblastoma, primitive neuroectodermal tumor (PNET)/pineal tumor, and atypical teratoid/rhabdoid tumor (ATRT), choroid plexus tumor, germ cell tumor, teratoma, and other and unspecified tumors. All tumors were included in overall analyses, and all subtype-specific analyses were carried out for HGG, LGG, ependymoma, medulloblastoma, ATRT, and PNET; some subtype-specific analyses are also included for other tumor types. Primary treatment variables included surgery at the primary site, radiation, and chemotherapy; of note, SEER does not distinguish between no chemotherapy and unknown chemotherapy. Observed all-cause survival was measured as number of months from diagnosis to death from any cause. Patients were censored at the date of last follow-up if alive at that time. Patients surviving more than five years were censored at 60 months.

Statistical analyses

Analyses were conducted using SAS 9.4 software. Chi-square tests (or Fisher’s Exact tests for expected cell counts < 5) were used to evaluate differences in the proportion of each tumor type occurring among the congenital and infant age groups compared to the age 1–19 group, as well as bivariate associations between age and receipt of treatment within each tumor type. Kaplan–Meier survival curves and the log-rank test were used to compare all-cause survival between the age groups. Hazard ratios (HR) were estimated from univariate and multivariable Cox proportional hazards survival analyses. The proportional hazards assumption was evaluated using scaled Schoenfeld residuals correlated with time, and violations were addressed by including time-dependent interaction terms in final multivariable models (model-specific violations shown in tabled results). We included covariates in the multivariable analysis that had a signifcant association in overall univariate survival models (p < 0.05). The exception was sex, which we included due to differences in survival by sex in our age and tumor subtype analyses, and given known differences in pediatric brain tumor incidence by sex. All statistical tests were two-sided, with p < 0.05 considered significant.

Results

15,125 patients were originally obtained from SEER, but only 14,493 were used in the analysis, as those with primary tumors outside of the brain were excluded (N = 205), as were patients for whom this tumor was not their first primary malignancy (N = 294), patients with zero survival days (N = 72), and cases reported through death certificate or autopsy only (N = 67).

Of the 14,493 patients diagnosed with a primary brain tumor between 1990 and 2015, there were 786 infants less than 1 year old (5.4% of study population) and 13,707 children between the ages of 1–19 years which served as a comparison group. Patient characteristics, demographics, and disease categories are listed in Table 1. LGG was the most common tumor type overall (N = 4649, 32.1%), followed by medulloblastoma (N = 1768, 12.2%), and ependymoma (N = 1088, 7.5%). Of note, we had many ungraded gliomas (UGG; N = 3164, 21.8%), which are difficult to define and not often diagnosed clinically. This may be attributed to SEER having incomplete pathological data on these tumors. To verify that this large quantity of UGG was not disproportionately coming from earlier years in the registry because of potential poorer histological documentation, we examined the distribution of UGG by diagnosis year (Supplemental Table 1), which showed no significant difference over time. Looking at the entire study population, most patients underwent surgery for treatment (N = 10,430, 72.0%), while a minority received radiation (N = 5726, 39.5%) and/or chemotherapy (N = 6024, 41.6%).

Table 1.

Characteristics of CNS tumor patients, ages 0–19 years, SEER Registries, 1990–2015 (N = 14,493)

Variable	Category	All patients		< 3 Months		3 Months to < 6 Months		6 Months to < 12 Months		1–19 Years
		N	%	N	%	N	%	N	%	N	%
Overall		14,493		171	1.18	180	1.24	435	3.00	13,707	94.58
Sex	Female	6467	44.62	80	46.78	99	55.00	239	54.94	7608	55.50
	Male	8026	55.38	91	53.22	81	45.00	196	45.06	6099	44.50
Race	White	11,475	79.18	137	80.12	138	76.67	336	77.24	10,864	79.26
	Black	1653	11.41	17	9.94	22	12.22	51	11.72	1563	11.40
	Other	1178	8.13	10	5.85	19	10.56	39	8.97	1110	8.10
	Unknown	187	1.29	7	4.09	1	0.56	9	2.07	170	1.24
Ethnicity	Non-Hispanic	11,245	77.59	119	69.59	132	73.33	317	72.87	10,677	77.89
	Hispanic	3248	22.41	52	30.41	48	26.67	118	27.13	3030	22.11
Year of diagnosis	1990–1994	1362	9.40	16	9.36	23	12.78	37	8.51	1286	9.38
	1995–1999	1527	10.54	17	9.94	13	7.22	52	11.95	1445	10.54
	2000–2004	3599	24.83	47	27.49	40	22.22	109	25.06	3403	24.83
	2005–2009	3580	24.70	42	24.56	53	29.44	118	27.13	3367	24.56
	2010–2015	4425	30.53	49	28.65	51	28.33	119	27.36	4206	30.69
Tumor type	Low-grade glioma	4649	32.08	6	3.51	34	18.89	108	24.83	4501	32.84
	High-grade glioma	1509	10.41	33	19.30	26	14.44	25	5.75	1425	10.40
	Ungraded glioma	3164	21.83	25	14.62	24	13.33	77	17.70	3038	22.16
	Ependymoma	1088	7.51	5	2.92	16	8.89	54	12.41	1013	7.39
	Choroid plexus tumor	121	0.83	12	7.02	8	4.44	21	4.83	80	0.58
	Medulloblastoma	1768	12.20	13	7.60	16	8.89	53	12.18	1686	12.30
	PNET/Pineal	844	5.82	20	11.70	17	9.44	29	6.67	778	5.68
	ATRT	285	1.97	17	9.94	25	13.89	52	11.95	191	1.39
	Germ Cell Tumor	672	4.64	2	1.17	2	1.11	1	0.23	667	4.87
	Teratoma	81	0.56	30	17.54	6	3.33	2	0.46	43	0.31
	Other brain/CNS tumor	312	2.15	8	4.68	6	3.33	13	2.99	285	2.08
Surgery performed	Yes	10,430	71.97	109	63.74	140	77.78	316	72.64	9865	71.97
	No	3995	27.57	61	35.67	40	22.22	117	26.90	3777	27.56
	Unknown	68	0.47	1	0.58	0	0.00	2	0.46	65	0.47
Radiation administered	Yes	5726	39.51	3	1.75	15	8.33	45	10.34	5663	41.31
	No	8638	59.60	168	98.25	164	91.11	385	88.51	7921	57.79
	Unknown	129	0.89	0	0.00	1	0.56	5	1.15	123	0.90
Chemotherapy administered	Yes	6024	41.56	68	39.77	106	58.89	276	63.45	5574	40.67
	No/Unknown	8469	58.44	103	60.23	74	41.11	159	36.55	8133	59.33

PNET primitive neuroectodermal tumor; ATRT atypical teratoid rhabdoid tumor

Tumor type by age

Chi-square analysis showed the proportion of tumor types differed significantly between age groups (p < 0.001, Table 2). Although LGG was the most common tumor type in the cohort overall, they are rare in the congenital group (3.5%). HGG was more common among patients < 3 months old but less common among infants 6–11 months old, compared to patients 1–19 years old. The most common tumors identified in the < 3-month age group were HGG (19.3%), teratoma (17.5%), and UGG (14.6%); in the 3 to < 6-month age group LGG (18.9%), HGG (14.4%), and ATRT (13.9%); and in the 6 to < 12-month age group LGG (24.8%), UGG (17.7%), and ependymoma (12.4%). LGG (32.8%), UGG (22.2%) and medulloblastoma (12.3%) were the most common tumor types in patients 1–19 years old.

Table 2.

Distribution of CNS tumor patients (ages 0–19), by tumor type and age group, SEER Registries, 1990–2015

Tumor type	Age group
	< 3 Months (N = 171)		< 3 Months to < 6 Months (N = 180)		6 Months to < 12 Months (N = 435)		1–19 Years (N = 13,707)
	N (%)	P	N (%)	P	N (%)	P	N (%)
High-grade glioma	33 (19.3)	< 0.001	26 (14.4)	0.078	25 (5.8)	0.002	1425 (10.4)
Low-grade glioma	6 (3.5)	< 0.001	34 (18.9)	< 0.001	108 (24.8)	< 0.001	4501 (32.8)
Ungraded glioma	25 (14.6)	0.018	24 (13.3)	0.005	77 (17.7)	0.027	3038 (22.2)
Ependymoma	5 (2.9)	0.026	16 (8.9)	0.446	54 (12.4)	< 0.001	1013 (7.4)
Choroid plexus tumor	12 (7.0)	< 0.001	8 (4.4)	< 0.001	21 (4.8)	< 0.001	80 (0.6)
Medulloblastoma	13 (7.6)	0.063	16 (8.9)	0.166	53 (12.2)	0.942	1686 (12.3)
PNET/Pineal	20 (11.7)	< 0.001	17 (9.4)	0.031	29 (6.7)	0.381	778 (5.7)
ATRT	17 (9.9)	< 0.001	25 (13.9)	< 0.001	52 (12.0)	< 0.001	191 (1.4)
Germ cell tumor	2 (1.2)	0.025	2 (1.1)	0.019	1 (0.2)	< 0.001	667 (4.9)
Teratoma	30 (17.5)	< 0.001	6 (3.3)	< 0.001	2 (0.5)	0.650	43 (0.3)
Other brain/CNS tumor	8 (4.7)	0.029	6 (3.3)	0.283	13 (3.0)	0.194	285 (2.1)

Bold text indicates statistical significance

P-values for Chi-square test of proportions (or Fisher’s Exact test when expected cell count < 5) using 119 year olds as comparison group

PNET primitive neuroectodermal tumor; ATRT atypical teratoid rhabdoid tumor

Treatment by age and tumor type

Distribution of the first course of treatment was compared by age and tumor type (Supplemental Table 2). Under 3-month-olds were least likely to receive any treatment for CNS tumors overall (70.2%), compared to both the 3 to < 6-month and 6 to < 12-month age groups (90.6% and 86.0%, respectively) and the control population of 1–19-year-olds (87.7%). Under 3-month-olds were also less likely to undergo surgery for all tumor types except HGG, for which almost all patients underwent surgery across age groups. In medulloblastoma, for which surgery is standard of care, 46.2% of patients < 3-months old did not receive surgery, whereas 100% of patients 3 months and older did. Of the patients in the congenital age group diagnosed with medulloblastoma, it is notable that if they did not receive surgery, they were not likely to receive radiation (7.7%) or chemotherapy (38.5%), whereas almost all patients in older groups received chemotherapy. Conversely, for PNET/Pineal and ATRT tumors in the < 3-month-old group, most patients received surgery (60.0% and 82.4%, respectively), like older age groups. Patients in the < 1-year age group were far less likely than 1–19-year-olds to undergo radiation for embryonal tumors, as expected due to concern for severe neurodevelopmental side effects, but were also less likely to undergo chemotherapy.

Survival by age and tumor type

Next, we estimated all-cause survival for patients by age and tumor type at yearly intervals up to 5 years (Supplemental Table 3) and visualized these differences through Kaplan–Meier curves (Fig. 1a). The < 3-month group exhibited the lowest OS each year. 5-year OS for all tumors was 36.7% (< 3 months), 56.0% (< 3–5 months), 63.8% (6–11 months), and 74.7% (1–19 years) (log rank p < 0.001). Looking at worst outcomes by age for each disease type, OS was worst for < 3-month-olds with LGG, medulloblastoma, and other embryonal tumors. OS was worst for 3–5-montholds with ependymoma and < 1-year-olds collectively with ATRT. Conversely, the worst outcomes in HGG were seen in the 1–19-year-old group (log rank p < 0.02 for all tumor types, Supplemental Fig. 1 and Supplemental Table 3). We further broke down the ages of the congenital group into < 1-month, 1 to < 2-months, and 2 to < 3-months to identify if differences in survival could be largely attributed to newborns who are born too sick and die quickly after birth (Fig. 1b). Survival estimates were significantly different between these expanded age groups and were worst for < 1-month olds (Fig. 1b). Since UGG is not a standard pathological diagnosis, we also conducted age-based analyses excluding this group and found similar results (Supplemental Fig. 2a, b).

Fig. 1.

All-cause Kaplan–Meier overall survival curves for all CNS tumors in A pediatric age groups and B expanded congenital and infant age groups

Univariate hazard ratio model

On univariate analysis, all congenital and infant age groups had increased mortality risk overall compared to the control population of 1–19-year-olds, the risk being the highest for the congenital group (HR 4.78; 95% CI 3.93–5.82, Supplemental Table 4). We found an increased risk of mortality for male patients in the 3 to 6-month age group. Black race was associated with significantly increased risk compared to White patients in 1- to 19-year-olds, but not in infants. Risk of death was also increased for patients of other races in the 6 to 12-month age group. The risk of death for patients of Hispanic ethnicity was significantly higher than for non-Hispanic patients for the < 1-year-old group (HR 1.43; 95% CI 1.14–1.80), as well as 6- to 12-month-olds and 1- to 19-year-olds. For year of diagnosis, the risk was especially high in the first period of study (1990–1994) for both infants and older children as compared to the most recent period. In analyzing treatment, lack of radiation and lack of chemotherapy were both risk factors for death in the < 1-year-old age group, as well as for 1–19-year-olds.

Stage at diagnosis was only available since 1998, so we analyzed it separately for this subpopulation on a univariate basis (Supplemental Table 6). CNS tumor stage in SEER is categorized as localized, regional, or distant, as it is for many other solid tumors. Since pediatric primary CNS tumors rarely metastasize outside the CNS, it is likely the regional and distant categorization both indicate metastatic disease. All of the congenital and infant age groups had a higher rate of metastatic disease (25–29%) than the 1–19-year-old group (13%).

Multivariable hazard ratio model

From the crude hazard of death data gathered from all CNS tumor patients (Supplemental Tables 4, 5), we then analyzed an adjusted multivariable analysis, controlling for patient characteristics, year of diagnosis, tumor type, and treatment. Looking at all diseases combined, patients younger than 1 year had poorer outcomes than 1 to 19-year-olds. Sex was not a significant risk factor for death, while Black race and Hispanic ethnicity both were (Supplemental Table 7). Each five-year time period since 1990 carried a significant risk of death compared to the most recent period, except for 2005–2009. Lack of surgery or chemotherapy predicted a significantly higher risk of death, whereas receiving radiation carried a higher risk (Supplemental Table 7).

Patients < 1 year of age who did not receive chemotherapy, radiation, or surgery had a significantly higher risk of mortality compared to those who did receive these treatments (Table 3). Differences in treatment associated with decreased survival were noted for < 3-month-olds who did not receive either surgery or chemotherapy, and 3 to < 6-month-olds who did not receive chemotherapy. In the 6 to < 12-month-old age group, absence of surgery, radiation, or chemotherapy conferred a higher risk for death.

Table 3.

Hazard of death from all causes among CNS tumor patients (ages 0–19) adjusted for demographics, year of dx, tumor type, and treatment, SEER Registries, 1990–2015, by age group, SEER Registries, 1990–2015

Variable	Category	< 1 Year								1–19 Years
				< 3 Months		3 Months to < 6 months		6 Months to < 12 months
		aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P
Age group	< 3 Months	2.26 (1.60, 3.18)	<.001
	3 Months to < 6 months	1.34 (0.97, 1.83)	0.072
	6 Months to < 12 months	Ref
Sex	Female	Ref		Ref		Ref		Ref		Ref
	Male	1.06 (0.85, 1.33)	0.593	0.67 (0.44, 1.03)	0.069	2.25 (1.25, 4.06)	0.007	1.03 (0.73, 1.45)	0.874	1.01 (0.95, 1.09)	0.69
Race	White	Ref		Ref		Ref		Ref		Ref
	Black	1.08 (0.76, 1.55)	0.657	1.03 (0.52, 2.04)	0.939	1.06 (0.43, 2.60)	0.899	1.59 (0.92, 2.76)	0.098	1.33 (1.20, 1.47)	<.001
	Other	1.28 (0.88, 1.87)	0.202	0.52 (0.17, 1.58)	0.25	1.18 (0.51, 2.70)	0.703	1.88 (1.11, 3.18)	0.019	1.32 (1.17, 1.50)	<.001
	Unknown	0.40 (0.13, 1.27)	0.122	0.41 (0.10, 1.74)	0.227	–	–	0.65 (0.09, 4.81)	0.672	0.29 (0.14, 0.60)	<.001
Ethnicity	Non-Hispanic	Ref		Ref		Ref		Ref		Ref
	Hispanic	1.39 (1.09, 1.77)	0.007	0.83 (0.52, 1.32)	0.427	1.62 (0.91, 2.88)	0.103	1.62 (1.12, 2.34)	0.01	1.34 (1.23, 1.45)	<.001
Year of diagnosis	1990–1994	1.67 (1.11, 2.52)	0.015	0.62 (0.26, 1.53)	0.303	0.72 (0.26, 1.97)	0.517	3.53 (1.85, 6.72)	<.001	1.42 (1.25, 1.61)	<.001
	1995–1999	1.17 (0.77, 1.79)	0.457	0.70 (0.32, 1.57)	0.392	1.23 (0.42, 3.65)	0.706	1.45 (0.76, 2.74)	0.258	1.21 (1.07, 1.38)	0.003
	2000–2004	1.54 (1.12, 2.11)	0.008	1.10 (0.60, 2.00)	0.757	1.81 (0.88, 3.71)	0.108	1.79 (1.10, 2.91)	0.02	1.22 (1.10, 1.35)	<.001
	2005–2009	1.07 (0.77, 1.49)	0.667	1.29 (0.70, 2.35)	0.415	1.03 (0.50, 2.15)	0.93	1.23 (0.75, 2.02)	0.41	1.08 (0.98, 1.20)	0.127
	2010–2015	Ref		Ref		Ref		Ref		Ref
Tumor type	Low-grade glioma	Ref		Ref		Ref		Ref		Ref
	High-grade glioma	3.47 (1.99, 6.07)	<.001	2.03 (0.57, 7.28)	0.278	2.57 (0.82, 8.04)	0.106	3.70 (1.57, 8.76)	0.003	10.67 (8.99, 12.68)	<.001
	Ungraded glioma	1.06 (0.57, 1.96)	0.848	0.88 (0.24, 3.27)	0.853	0.82 (0.19, 3.56)	0.786	0.65 (0.25, 1.73)	0.389	4.90 (4.15, 5.80)	<.001
	Ependymoma	3.05 (1.68, 5.55)	<.001	0.57 (0.06, 5.71)	0.633	4.02 (1.20, 13.45)	0.024	3.51 (1.63, 7.57)	0.001	4.22 (3.43, 5.20)	<.001
	Choroid plexus tumor	2.83 (1.41, 5.69)	0.003	1.16 (0.24, 5.57)	0.851	3.59 (0.79, 16.32)	0.098	4.06 (1.58, 10.42)	0.004	9.10 (6.00, 13.83)	<.001
	Medulloblastoma	6.47 (3.80, 11.03)	<.001	4.79 (1.21, 18.89)	0.025	5.56 (1.80, 17.20)	0.003	7.47 (3.69, 15.10)	<.001	4.34 (3.54, 5.33)	<.001
	PNET/Pin e al	8.29 (4.88, 14.08)	<.001	4.04 (1.14, 14.33)	0.031	8.68 (2.88, 26.15)	<.001	9.08 (4.29, 19.23)	<.001	7.59 (6.14, 9.40)	<.001
	ATRT	13.60 (8.14, 22.72)	<.001	3.24 (0.86, 12.25)	0.083	11.38 (3.94, 32.88)	<.001	27.16 (13.68, 53.89)	<.001	19.51 (14.87, 25.59)	<.001
	Germ cell tumor	3.92 (1.12, 13.72)	0.032	–	–	34.19 (4.27, 274.00)	<.001	40.36 (4.61, 353.70)	<.001	1.07 (0.78, 1.47)	0.67
	Teratoma	3.20 (1.60, 6.40)	<.001	1.58 (0.42, 5.86)	0.498	1.10 (0.12, 9.88)	0.933	9.31 (1.17, 74.15)	0.035	4.87 (2.37, 10.04)	<.001
	Other brain/CNS tumor	8.03 (4.28, 15.06)	<.001	3.42 (0.83, 14.03)	0.088	8.83 (2.38, 32.77)	0.001	7.35 (2.76, 19.60)	<.001	6.63 (4.87, 9.01)	<.001
Surgery performed	Yes	Ref		Ref		Ref		Ref		Ref
	No	2.02 (1.45, 2.82)	<.001	2.17 (1.35, 3.47)	0.001	1.29 (0.64, 2.60)	0.475	2.43 (1.33, 4.45)	0.004	3.16 (2.80, 3.57)	<.001
	Unknown	1.39 (0.18, 10.74)	0.754	–	–			5.24 (0.56, 49.49)	0.148	7.04 (3.65, 13.56)	<.001
Radiation administered	Yes	Ref		Ref		Ref		Ref		Ref
	No	2.08 (1.29, 3.36)	0.003	1.19 (0.26, 5.33)	0.823	2.99 (0.97, 9.23)	0.056	2.23 (1.24, 4.03)	0.008	0.52 (0.46, 0.58)	<.001
	Unknown	0.42 (0.05, 3.14)	0.395	–	–	–	–	1.08 (0.14, 8.43)	0.94	1.56 (0.97, 2.51)	0.066
Chemotherapy administered	Yes	Ref		Ref		Ref		Ref		Ref
	No/Unknown	2.93 (2.12, 4.04)	<.001	4.01 (2.12, 7.60)	<.001	2.57 (1.31, 5.07)	0.006	3.08 (1.83, 5.18)	<.001	1.12 (0.99, 1.27)	0.081
Time-dependent covariates	Age Group * Time	1.02 (1.00, 1.03)	0.007
	Dx Year * Time			0.97 (0.95, 0.99)	0.007
	Tumor Type * Time									1.00 (1.00, 1.00)	<.001
	Surgery * Time	0.97 (0.94, 0.99)	0.015					0.96 (0.92, 1.00)	0.047	0.98 (0.97, 0.99)	<.001
	Radiation * Time									0.99 (0.99, 1.00)	<.001
	Chemotherapy * Time	0.94 (0.92, 0.97)	<.001	0.90 (0.84, 0.97)	0.005	0.94 (0.89, 1.00)	0.036	0.95 (0.92, 0.99)	0.006	0.97 (0.97, 0.98)	<.001

aHR adjusted hazard ratio; CI confidence interval; PNET primitive neuroectodermal tumor; ATRT atypical teratoid rhabdoid tumor; “demographics” includes age, sex, race, and ethnicity

Analyzing our multivariable analysis results by age (Table 3), we identified patients less than 6 months as having poorer outcomes than 6 to 12-month-olds. The only racial group that had a significantly higher risk of death was 6–12-month-olds of other races compared to White patients. Hispanic ethnicity placed patients at higher risk for mortality between the ages of 6 to 12-months, whereas this was not the case in the congenital group. Male sex was a significant risk factor in the 3 to 6-month age group, but not overall. Patients diagnosed in periods prior to 2004 were at higher risk as well. Lack of surgery, radiation, and chemotherapy were all risk factors for death.

We then performed a disease-specific multivariable analysis of the < 1-year-old group (Table 4). For HGG, Hispanic ethnicity and lack of chemotherapy placed patients at a higher risk of mortality, while there were no significant findings for LGG. Male sex, Hispanic ethnicity, diagnosis during earlier time periods, and lack of surgery placed patients with ependymomas at higher risk of death. Black race, diagnosis in the earliest period, and lack of radiation or chemotherapy carried significantly higher risks of death from medulloblastoma. Significant risk factors for death in PNET included Black race, male sex and lack of chemotherapy. Lack of surgery, radiation, or chemotherapy were significantly associated with death in ATRT. Compared to infant age groups, the congenital age group showed a higher risk for mortality in LGG, medulloblastoma, and PNET/Pineal but not in HGG, ependymoma, and ATRT. Compared to 1–19-year-olds, each age group < 1 year had a significantly higher risk for death from nearly every tumor type; notable exceptions were < 3-month-olds with ependymoma, and all the < 1-year-old groups with ATRT (Supplemental Table 8).

Table 4.

Hazard of death from all causes among CNS tumor patients (ages < 1 year) adjusted for demographics, year of dx, and treatment, by tumor type, SEER Registries, 1990–2015

Variable	Category	High-grade glioma		Low-grade glioma		Ependymoma		Medulloblastoma		PNET/Pineal		ATRT
		aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P	aHR (95% CI)	P
Age group	< 3 Month s	1.26 (0.53, 2.99)	0.607	10.27 (2.47, 42.68)	0.001	0.29 (0.02, 3.65)	0.338	2.77 (1.10, 6.99)	0.031	3.69 (1.55, 8.80)	0.003	0.62 (0.30, 1.29)	0.204
	3 Months to < 6 months	0.44 (0.15, 1.28)	0.131	1.12 (0.38, 3.26)	0.842	2.02 (0.70, 5.83)	0.195	1.29 (0.57, 2.92)	0.547	2.54 (1.13, 5.72)	0.024	0.93 (0.52, 1.65)	0.794
	6 Months to < 12 months	Ref		Ref		Ref		Ref		Ref		Ref
Sex	Female	Ref		Ref		Ref		Ref		Ref		Ref
	Male	0.65 (0.30, 1.43)	0.284	0.84 (0.32, 2.23)	0.728	5.74 (1.68, 19.59)	0.005	0.91 (0.47, 1.75)	0.768	2.31 (1.23, 4.34)	0.009	0.88 (0.52, 1.48)	0.626
Race	White	Ref		Ref		Ref		Ref		Ref		Ref
	Black	1.59 (0.57, 4.47)	0.379	1.24 (0.33, 4.68)	0.747	8.40 (0.64, 110.06)	0.105	3.61 (1.12, 11.66)	0.032	3.34 (1.20, 9.29)	0.021	0.94 (0.40, 2.24)	0.898
	Other	0.17 (0.02, 1.46)	0.106	1.67 (0.20, 14.01)	0.637	2.28 (0.53, 9.79)	0.269	1.57 (0.58, 4.25)	0.377	1.58 (0.50, 4.99)	0.435	1.47 (0.68, 3.20)	0.326
	Unknown	–	–	–	–	–	–	–	–	–	–	0.27 (0.03, 2.23)	0.226
Ethnicity	Non-Hispanic	Ref		Ref		Ref		Ref		Ref		Ref
	Hispanic	2.35 (1.01, 5.45)	0.046	2.24 (0.82, 6.13)	0.117	5.71 (1.72, 18.92)	0.004	1.87 (0.93, 3.79)	0.08	1.48 (0.79, 2.78)	0.221	1.09 (0.62, 1.93)	0.77
Year of diagnosis	1990–1994	0.41 (0.10, 1.72)	0.223	0.61 (0.06, 6.71)	0.688	14.10 (2.26, 87.85)	0.005	3.95 (1.21, 12.82)	0.022	0.67 (0.20, 2.24)	0.517	–	–
	1995–1999	0.69 (0.20, 2.35)	0.548	1.49 (0.23, 9.50)	0.675	2.87 (0.21, 39.22)	0.429	3.04 (1.00, 9.21)	0.049	0.38 (0.11, 1.25)	0.111	–	–
	2000–2004	1.50 (0.55, 4.12)	0.428	1.16 (0.27, 4.97)	0.839	2.86 (0.62, 13.09)	0.177	3.15 (1.06, 9.39)	0.039	0.53 (0.18, 1.60)	0.258	0.99 (0.51, 1.93)	0.979
	2005–2009	2.55 (0.82, 7.94)	0.107	1.37 (0.35, 5.32)	0.65	1.03 (0.16, 6.73)	0.974	0.44 (0.10, 1.98)	0.283	0.41 (0.13, 1.28)	0.124	0.83 (0.46, 1.49)	0.527
	2010–2015	Ref		Ref		Ref		Ref		Ref		Ref
Surgery performed	Yes	Ref		Ref		Ref		Ref		Ref		Ref
	No	1.05 (0.42, 2.66)	0.911	0.64 (0.17, 2.37)	0.506	18.38 (3.00, 112.46)	0.002	2.13 (0.84, 5.41)	0.11	1.86 (0.88, 3.91)	0.102	3.41 (1.42, 8.19)	0.006
	Unknown	–	–	–	–	–	–	–	–	–	–	–	–
Radiation administered	Yes	–		Ref		Ref		Ref		Ref		Ref
	No	–		0.57 (0.07, 4.87)	0.611	1.09 (0.30, 4.05)	0.893	5.33 (1.16, 24.51)	0.032	3.44 (0.80, 14.81)	0.097	3.57 (1.61, 7.91)	0.002
	Unknown	–		–	–	–	–	2.41 (0.18, 32.67)	0.508	–	–	–	–
Chemotherapy administered	Yes	Ref		Ref		Ref		Ref		Ref		Ref
	No/Unknown	4.18 (1.91, 9.15)	<.001	0.99 (0.39, 2.54)	0.986	1.57 (0.54, 4.52)	0.405	2.68 (1.17, 6.15)	0.02	7.64 (2.21, 26.46)	0.001	2.55 (1.39, 4.70)	0.003
Time-dependent covariates	Chemotherapy * Time									0.70 (0.56, 0.89)	0.003

Discussion

Between the < 3-month, 3 to < 6-month, 6 to < 12-month, and 1–19-year age groups, tumor subtype distribution differed significantly. LGG are the most frequent brain tumors overall in children [7]. While our study shows that this is true in the 3 to < 6-month and 6 to < 12-month age groups, as well as the 1–19-year-olds, LGG was one of the least diagnosed tumors in the < 3-month age group, in which HGG is most common. These differing tumor distributions by age further support our conclusion that congenital and infant tumors differ from those in older children and add valuable context for a previously understudied age group. Survival differed significantly by age between tumor subtypes and was lowest overall for < 1-month-olds, although comparison of overall survival between age groups is challenging due to differences in tumor distribution. Even when comparing survival between age groups for specific tumor types, however, younger age was associated with decreased survival, except in HGG. When adjusted for diagnostic time period and treatment variables, survival differences in some tumor types were attenuated.

By multivariable analysis, lack of surgery or chemotherapy remained an independent risk factor for mortality. Infants younger than 1 year with brain tumors were significantly less likely to receive surgery or chemotherapy, and those patients display an increased risk of mortality. Surgery is typically the first approach to treatment of pediatric brain tumors, while use of radiotherapy is less common under the age of 3, due to worsened long-term neurodevelopmental deficits in this already developmentally vulnerable population. Although radiation differences by age are not novel findings, surgery and chemotherapy differences are notable. Dressler et al. reported that in a population of 0–19-year-olds diagnosed with medulloblastoma, patients younger than 4 years were more likely to receive chemotherapy and less likely to receive radiation, while survival was the lowest in patients < 1 year of age [8]. Our study suggests that patients < 3 months old bear the brunt of this treatment and survival disparity. Most patients < 3 months old diagnosed with PNET/Pineal and ATRT did receive surgery, however. We believe that the higher proportion of these tumors seen in infants could be contributing to the comfort level of surgeons in operating on these patients. These findings support our overall conclusion that patterns of treatment in the congenital and infant age groups differ from those in older patients.

Pediatric brain and CNS cancer survival has been shown to differ by race and ethnicity, with studies reporting that both Hispanic and non-Hispanic Black patients have a higher risk of death than non-Hispanic White patients [9, 10]. There is ample evidence to suggest that social factors, including race, ethnicity, and socioeconomic status are associated with survival disparities in pediatric [9] and adolescent and young adult (AYA) cancer survival [11, 12]. In our patient population, we observed a Hispanic ethnicity-based mortality risk in only 6 to < 12-month-olds, not the < 3-month and 3–6-month age groups. These racial and ethnic group analyses should be regarded as exploratory and interpreted cautiously, as many minority groups were limited in sample size, especially for specific tumors.

While our study highlights the role of treatment differences in survival, biological differences across tumor types are also relevant to our discussion. Infant medulloblastomas are known to be biologically different from those in older children, notably in the Sonic Hedgehog and Group 3 types. These biological differences often make these tumors more aggressive and more likely to be metastatic [13]. Similarly, ependymomas in infants are very likely to be anaplastic and therefore more aggressive in nature [14]. Studies of HGG have revealed that infant HGG have different and often targetable genetic drivers from other pediatric HGG and thus merit different clinical management [15, 16]. In ATRT, three molecular subgroups have been established, with infant tumors disproportionately falling in the TYR (overexpression of tyrosinase) subgroup, which has unique potential therapeutic targets [17] and a better prognosis than other subgroups [18]. Some of the differences in outcomes we found in our study may therefore be based on tumor biological differences in these youngest patients. These biological differences have variable positive and negative prognostic implications, however, while nearly all of our findings demonstrated more negative outcomes for the congenital and infant age groups compared to older children. None of these biological findings have yet been determined to obviate the need for standard treatment modalities. Therefore, we believe that while tumor biology likely plays some role in our findings, differences in treatment for these young patients is a crucial factor in their poorer outcomes that necessitates attention. Patients in the congenital and infant age groups were also found to be more likely to present with metastatic disease; this could be due to biological differences in tumors for this age group and/or greater difficulty in timely diagnosis of CNS tumors at these ages, leading to delays and metastatic spread by the time of diagnosis. We know from previous trials, though, that metastatic pediatric CNS tumors are still often curable, including in very young patients [19], so metastatic status in infants should not prevent attempts at treatment.

The strengths of this study include the population-based design, allowing us to obtain large datasets to comprehensively study the rare event of congenital and infant brain tumors. Despite the obvious advantages to this approach, there are limitations in the use of SEER data, including lack of individual-level socioeconomic status data, detailed information on cause of death, limited coverage with respect to the US population especially for earlier periods, missing pathology data required to accurately classify CNS tumors, and lack of information on treatment at disease recurrence. For the untreated infants, there are clinical factors that cannot be considered in our analysis, including how ill they were upon presentation and whether they could realistically tolerate treatment. As above, we do know that the congenital and infant patients were more likely to present with metastatic disease, but we did not incorporate this factor into our multivariable analysis because stage at diagnosis was only available for a subset of patients. While inclusion of tumor subtypes creates a large number of comparisons, we felt that the heterogeneity of pediatric brain tumors by biology, treatment, and prognosis necessitates these analyses.

As research efforts continue to improve classification of brain tumors based on novel genetic and molecular insights, changes in therapy will follow. For these advances to be made, we must better understand congenital and infant CNS tumors and their response to therapy across tumor types and age categories. Here, we have shown that survival outcomes and treatment distribution are notably different between specific age groups in infants. These findings support our conclusion that congenital and infant CNS tumors differ pathologically, therapeutically, and prognostically from those in older children. These youngest groups of patients were less likely to receive standard treatment modalities, including surgery and chemotherapy and had poorer overall survival for most tumor types than older children. Therefore, the use of surgery and chemotherapy in these patients should be increased, and these approaches should be standardized through clinical studies, including trials specifically aimed at tumors in very young patients [20] (e.g. NCT04655404 targeting NTRK-fused HGG). These strategies have the potential to improve survival for congenital and infant tumors toward the survival rates seen in older children.

Data availability

All data analyzed for this study is included in main tables, figures, and supplemental tables.