The management of children with hypothalamic (H) and/or chiasmatic (C) tumors remains controversial. We evaluated the impact of clinical and neuroimaging parameters and primary therapy on overall (OS) and progression-free (PFS) survival and on neuroendocrine and neurocognitive outcome in children with H and/or C tumors.
Records were reviewed for 73 children with H and/or C tumors treated at St. Jude Children's Research Hospital between October 1981 and December 1999.
Thirty-six patients received irradiation or chemotherapy immediately postdiagnosis and 37 were observed. The 6-year OS and PFS rates were 86 ± 5%; and 36 ± 7%, respectively. The 6-year PFS rates for the irradiation, chemotherapy, and observation groups were 69 ± 16%, 12 ± 11%, and 37 ± 9%, respectively. In multivariate analysis, intracranial NF1 lesions (P = 0.015) and initial irradiation (P = 0.056) led to better PFS rates. There was no difference in OS between those initially treated or observed. Mean serial intelligence quotient (IQ) scores were 86 and 86 at diagnosis and at 6 years later, respectively. Patients younger than 5 years old had a lower mean IQ score at diagnosis (79.1) than older patients (96.3; P = 0.003). Patients who were irradiated at diagnosis had a significantly higher cumulative incidence of endocrinopathy at 3 years (P = 0.008).
The most appropriate management for hypothalamic (H) and/or chiasmatic (C) tumors in children remains uncertain. Although these tumors are generally of low grade histology, their behavior is unpredictable and most patients eventually require treatment.1, 2 Recommendations for initial treatment have included observation, primary surgery,3, 4 radiotherapy (RT),5, 6 or chemotherapy.2, 7, 8, 9 These recommendations were based on patient series spanning several decades. Therefore, they lack a uniform approach to diagnosis, neuroimaging, surgery, and therapy. A further limitation is that many series combine all tumors of the visual pathway, including isolated optic nerve gliomas. The more favorable prognosis of the latter tumors may confound attempts to assess the impact of treatment and ultimate outcome of H and/or C tumors. We report the clinical features, treatment, survival, and neuroendocrine and neurocognitive outcome of 73 pediatric patients with H and/or C tumors diagnosed and treated at a single institution in the era of modern neuroimaging.
MATERIALS AND METHODS
Between October 1981 and December 1999, 73 consecutive patients with H and/or C tumors were evaluated and treated at St. Jude Children's Research Hospital. All patients had neuroradiologic confirmation of their tumor. A chart review documented the clinical features at diagnosis, therapy, pattern of recurrence, and outcome of this cohort.
Initial diagnostic imaging included magnetic resonance imaging (MRI) scans (n =66) or computed tomographic (CT) scans (n=7). Regular follow-up included serial MRI scans in all patients. The radiologic location and extent of disease at diagnosis were reviewed by a neuroradiologist (JWL) and a neurooncologist (MF). Tumors were defined as 1) chiasmatic only (C) if they involved the optic chiasm and the anterior third ventricle was well visualized and free of tumor, 2) hypothalamic only (H) if a normal chiasm was visualized, 3) hypothalamic and chiasmatic (H/C) if both structures were involved with no obvious distinction. Patients with isolated optic nerve tumors were excluded. The neurosurgical approach remained generally consistent and conservative for all patients, including those with NF1 lesions. Patients who had H or globular H/C tumors without involvement of the optic nerve and/or tracts underwent biopsies for disease confirmation. Surgical debulking was reserved for the relief of ventricular obstruction from tumors or cysts. Postoperative imaging was used systematically to confirm the extent of tumor resection. The histologic diagnosis was confirmed through independent review of the original diagnostic tissue by two pathologists (JJJ and PB).
In addition to neuroimaging follow-up, regular ophthalmologic, neuroendocrine, and neuropsychologic evaluations were performed. Before initiation of therapy, all patients, parents or guardians, as appropriate, gave written informed consent according to the institutional guidelines.
Overall survival (OS) was measured from the date of diagnosis to the date of death or last contact. Progression-free survival (PFS) was measured from the date of diagnosis to the date of disease progression, death, or last contact. Time at risk for secondary PFS was measured from the date of initial progression until the date of second progression, death, or last contact. Distributions of PFS and OS were estimated using the Kaplan–Meier method.10 Associated standard errors were calculated by the techniques proposed by Peto et al.11 Univariate comparisons between distributions of PFS or OS were made using the exact log rank test or the log rank test when exact computation was not feasible. Multiple comparisons of PFS and OS among the three treatment groups were adjusted using the Bonferroni method. The relationship between tumor location and histology was assessed using the exact chi-square test. Only two-sided P values are reported. Exact tests were accomplished using Proc-StatXact (Cytel Software, Cambridge, MA). A multivariate stepwise Cox regression analysis12 was performed to assess independent predictors of PFS. All variables with a P value less than 0.05 from the univariate log rank tests were included as candidates in the stepwise analysis. A multivariate stepwise Cox regression analysis was not performed using survival as the outcome because of the few number of deaths.
The neuropsychological testing included IQ testing. Patients were administered standardized and age-appropriate instruments with a normative IQ of 100 and a standard deviation of 15.
Random coefficient models were constructed to estimate and compare the rate of change in general IQ scores. The methods described in Laird and Ware13 and Jennrich and Schluchter14 were used for model construction, estimation, and testing as implemented in Proc Mixed (SAS/STAT software, Version 8.0, SAS Institute, Cary, NC).
Cumulative incidence estimates for the detection of first endocrinopathy overall and by initial therapy were calculated considering death and progression as competing risks.15 Statistical tests developed by Gray16 were used to statistically compare the differences in the cumulative incidence functions.
The 73 patients with H and/or C tumors comprised 7.7% of the 947 patients with newly diagnosed primary central nervous system neoplasms admitted to our institution during the study period. Table 1 summarizes their clinical characteristics. There were 28 patients with phenotypic NF1 lesions (38%). Of these 28 (54%) patients, 15 had typical intracranial NF1 lesions on MRI scans (areas of increased T2-weighted signal intensity which were not visualized on T-1 imaging with or without contrast) in addition to the primary H and/or C lesions. Of note, one patient with intracranial NF1 lesions on MRI did not have other phenotypic evidence of the lesions.
Table 1. Clinical and Pathologic Patient Characteristics
Age at diagnosis (yrs)
Astrocytoma, not otherwise specified
None (coded as glioma, no biopsy)
Surgery extent (initial)
Tumor Location and Extent
No attempt was made to classify origin by the “predominant” site of involvement. The primary site of disease was H alone in 3 patients (4%), C alone in 19 patients (26%), and both structures in 51 patients (70%). Overall, 53 patients (73%) showed extension beyond the H/C, i.e., 6 to the optic nerves only, 28 to the optic tracts only, and 19 to both.
Twenty-three patients had spinal MRIs at diagnosis. Only the three patients initially found to have disseminated disease on cranial imaging had evidence of spinal leptomeningeal metastases at diagnosis. Sixteen patients had cerebrospinal fluid cytologic evaluations at diagnosis, including one of the three patients with intracranial and spinal leptomeningeal disease. All were negative.
Degree of Resection and Tumor Histology
Forty-two patients (58%) underwent a biopsy (n = 39), defined as less than 50% resection, or a subtotal resection (STR; defined as > 50%, < 90% resection; n=3), at the time of diagnosis. Of these 42 patients, 34 had H/C, 6 had C, and 2 had H tumors. Five other patients had a biopsy or resection at a later point during therapy. Twenty-six patients (36%) never had a tissue diagnosis. There was no statistical difference in the extent of surgery between patients younger than 2 years and older patients. Forty-two percent of patients younger than 2 years old at diagnosis had no surgery compared with 43% in the older age group.
Among the 47 patients for whom tissue diagnosis was available, 32 patients had juvenile pilocytic astrocytomas, 10 had low-grade astrocytomas not otherwise specified, 2 had gangliogliomas, 1 had a fibrillary astrocytoma, 1 had an anaplastic astrocytoma, and 1 had a neurocytoma. There was no correlation between location and histology (P = 0.20).
Treatment and Outcome Postdiagnosis
Thirty-six patients (50%) commenced active treatment immediately after diagnosis (median time, 0.8 months): 15 were irradiated, 20 received chemotherapy, and 1 patient who initially received chemotherapy, developed chronic neutropenia, and went on to receive irradiation. Thirty-seven patients (50%) were observed without initial therapeutic intervention. There was a significant difference in age at diagnosis among the three groups (P < 0.0001). Patients treated with irradiation were significantly older (median age, 7.9 years; range, 0.5–15.0 years) than those who were observed (median age, 4.2 years; range, 0.3–14.8 years; P = 0.0003) or those who received chemotherapy (median age, 2.0 years; range, 0.3–12.0 years; P = 0.002).
The median follow-up time for the 63 patients alive at last contact is 6.3 years (range, 6.5 months–14.1 years). The 6-year OS and PFS estimates are 86 ± 5% and 36 ± 7%, respectively. Figures 1, 2, and 3 summarize OS and PFS distributions by age group, initial treatment, and NF1 status, respectively.
The median age for the 37 patients followed without initial therapy was 4.2 years (range, 0.3–14.8 years). One patient had an STR at diagnosis. Eight were asymptomatic with incidental tumors found on surveillance brain MRI scans for NF1 (n = 7) or posttrauma (n = 1). Other reasons for observation included minimal signs and symptoms (n = 12), improvement after biopsy and ventriculoperitoneal (VP) shunt (n = 13), parental refusal of recommended treatment (n = 2), and unknown (n = 2). Twenty-three patients progressed at a median of 1.2 years (range, 0.1–8.6 years) after diagnosis. The 6 year-PFS and OS estimates were 37 ± 9% and 94 ± 5%, respectively (Fig. 1). Among those with progressive disease (PD), 3 were followed, 12 received chemotherapy, and 8 received RT.
The median age for the 20 patients in this group was 2 years (range, 0.3–12 years). Two underwent STR. Eighteen patients received alkylator or platinating (carboplatin or cisplatin) agent-based chemotherapy regimens and two received cyclophosphamide and vincristine. A partial response, lasting a median of 12 months, was noted in 4 patients and 13 had stable disease (SD). Overall, 14 of the 20 patients progressed at a median of 2 years (range, 0.3–5 years) from diagnosis. The 3 and 6-year PFS estimates weres 23 ± 9% and 12 ± 11%, respectively, with a 6-year OS estimate of 85 ± 12% (Fig. 2). Salvage therapy consisted of irradiation in nine patients, further chemotherapy in one patient, observation in two patients, and irradiation and chemotherapy in two patients.
Among the 15 (21%) patients who received radiation as their initial treatment, the median dose was 52.2 Gy (fraction size, 150–180 cGy once daily). Their median age was 7.9 years (range, 0.6–15.0 years). The 6-year PFS and OS estimates were 69 ± 16% and 68 ± 15%, respectively (Fig. 1). Four patients had PD at a median of 0.9 years (range, 0.7–2.2 years) after initial irradiation. Salvage therapy consisted of chemotherapy (n = 3) and local intracystic P32 instillation (n = 1).
Among the primary treatment groups, patients treated with irradiation had better PFS estimates than those treated with chemotherapy (P = 0.012). There was no difference in OS estimates among the three treatment groups (Fig. 1).
After removing the four patients with anaplastic astrocytoma, ganglioglioma, and neurocytoma, we found no statistically significant difference in PFS (P = 0.064) and OS (P = 0.94) between those diagnosed with Juvenile pilocytic astrocytama (JPA) and patients with low-grade astrocytoma, not otherwise specified.
Age younger than 2 years at diagnosis was a significant predictor of poor PFS (P = 0.01). There was no difference in OS between the two age groups (P = 0.42; Fig. 2).
Among the three tumor locations, patients with C tumors had a better PFS (6-year PFS, 59 ± 12%) than those with H/C tumors (6-year PFS, 24 ± 8%, P = 0.045).
Patients with a phenotypic diagnosis of NF1 had a slightly better OS statistically (6-year OS, 96%) than those without NF1 (6-year OS, 80% ± 8%, P = 0.056; Fig. 3). Only 1 of the 10 patients who died had NF1. There was no difference in PFS between the two groups (P = 0.18; Fig. 3).
A subgroup (n = 16) of the 28 patients with the phenotypic diagnosis of NF1 also had typical bright intracranial NF1 lesions present on T2-weighted MRI images. The presence of these lesions was an even better predictor of PFS (P = 0.017) than the phenotypic diagnosis of NF1. Only a borderline advantage in OS was evident for patients with these intracranial lesions (P = 0.059).
Multiple Regression Analysis of Prognostic Factors
We could not perform a multiple regression analysis for factors affecting OS because there were only 10 deaths. An exploratory stepwise regression evaluating factors related to PFS was performed on the subset of 69 patients who remained after eliminating the three patients with only H disease and the one patient who received both chemotherapy and irradiation. Age at diagnosis, treatment, tumor location, and the presence of NF1 lesions on MRI scans were available for the regression analysis (P < 0.05 according to the univariate log rank test).
The stepwise procedure identified the presence of NF1 lesions on MRI scans as a significant prognostic factor for PFS (P = 0.015). Initial treatment with irradiation was borderline significant in a stepwise Cox regression model (P = 0.056) after adjusting for the presence of NF1 lesions on MRI scans. Age at diagnosis and tumor location were not significant at the P is less than 0.05 level and were not included in the final stepwise regression model.
Patterns of Failure
Forty-one patients had PD and 30 remain alive without progression. Thirty-eight of 41 patients demonstrated local progression at the time of first recurrence, including three with progressive neuraxis dissemination. Three patients with localized disease at diagnosis demonstrated leptomeningeal progression: one developed intracranial metastases and two developed intracranial and distant neuraxis dissemination. The estimated PFS at 2 years following first progression is 62 ± 9%.
Of the 73 patients in this cohort, 43 had neuropsychologic testing. Thirty-one of these 43 patients had at least two measurements and were included in the analysis. Each patient had a median of four neuropsychologic evaluations (range: 2–8) There was no significant difference between the distribution of characteristics (gender, age, diagnosis, and treatment) among patients included or excluded from the IQ analysis, suggesting that the IQ analysis subset is representative of the entire cohort. The mean estimated IQ score at diagnosis was 86 (95% confidence interval = 79, 93). The mean observed IQ scores at approximate time points are provided in Table 2. There is no evidence that patients' IQ scores changed significantly over time (P = 0.81).
Table 2. The Mean Observed IQ Scores at Approximate Time Points
IQ: intelligence quotient.
1 yr postdiagnosis
2 yr postdiagnosis
3 yr postdiagnosis
4 yr postdiagnosis
5 yr postdiagnosis
6 yr postdiagnosis
There was a significant difference in the estimated IQ score at diagnosis between preschool children (children younger than 5 years of age at diagnosis, IQ = 79) and school-aged children (children older than 5 years of age at diagnosis, IQ = 96; P = 0.003). No difference in the change of the IQ score over time was observed (P = 0.21) between the two age groups.
There was no significant difference in the change in general intelligence over time between those irradiated (n = 17) and those not irradiated at any time (n = 14, P = 0.84). Many patients (58%) progressed at least once. The potential effects of the tumor progression as well as the timing of the progression on intellectual outcome could not be addressed in this analysis.
Of the 31 patients with general IQ scores, 16 had NF1. There was no difference in the change in IQ over time by NF1 status (P = 0.98). Furthermore, there was no difference in the estimated IQ at diagnosis between patients with/without NF1.
Of the 73 patients, 66 underwent serial endocrinologic evaluations. Forty-seven (64%) developed at least one endocrinopathy during the study period. Nineteen of the 66 patients (29%) had endocrinopathies requiring treatment at diagnosis or before the initiation of any therapy. Twenty-eight patients (42%) without initial evidence of endocrinopathies developed one or more such events following therapy. Overall, 85% of irradiated patients versus 53% nonirradiated patients developed endocrinopathies. The 3-year cumulative incidence of endocrinopathy was high in the radiation group (66 ± 16%) compared with patients who were observed or treated with chemotherapy (19 ± 6%; P = 0.008). Endocrinopathies occurred in 87% of patients without phenotypic NF1 compared with 54% of patients with NF1.
The location and prevalence of H and/or C tumors in patients younger than 5 years of age place these patients at high risk for potential intellectual, endocrine, visual, and other sequelae of tumor and therapy.17, 18 The current series represents the largest series of H and/or C tumors reported. It is unique in its use of an anatomically assigned neuroimaging classification, a consistent and conservative surgical and adjuvant treatment approach, and a median follow-up of more than 6 years. Despite the number of patients included in this study, a potential criticism is that the limited number of patients in various treatment and age groups makes the conclusions regarding prognostic factors and neuropsychological function somewhat difficult.
For the entire group, PFS was 36 ± 7% at 6 years, with an OS of 86 ± 5%, similar to the 85–93% OS reported in smaller series at similar time points.2, 5, 7, 19, 20
Several studies have suggested that younger patients with H and/or C tumors have a poorer prognosis.4, 21, 22 Although the current study also identified age as a significant prognostic factor for PFS in univariate analysis (P = 0.01), age was not a significant prognostic factor for PFS in multivariate analysis. Overall survival was independent of age.
Leptomeningeal spread is a rare but documented feature of low-grade gliomas, occurring in 3–12% of patients.23–26 Mamelak et al.23 reported that 30% of patients with H/C tumors developed leptomeningeal spread compared with only 2% of patients with non-H primary tumors. In the current study, 3 of 73 patients (4%) had clinical or imaging evidence of intracranial dissemination at diagnosis and 3 others had evidence of leptomeningeal dissemination at recurrence, for a total of 6 (8%). This is similar to the 6% incidence of dissemination in other supratentorial low-grade gliomas.23, 24 It is noteworthy that spinal dissemination was only detected in patients with documented intracranial metastatic disease.
Some investigators have advocated resection for large exophytic and/or cystic tumors in this location.3, 4 In Wisoff et al.,4 11 of 16 children with H/C tumors remained alive at 0.3–4.5 years following radical (60–95%, n = 9) or limited (25–50%, n = 5) tumor resection. Despite the favorable OS, surgery carries significant risks in younger children with deep midline tumors.4 Wisoff et al.4 achieved limited resections in three infants, all of whom subsequently died of PD. In one case, the infant died following significant neurosurgical sequelae with postoperative infarction.
In contrast, Sutton et al.20 reported a series of 33 patients who would have been eligible for radical surgery using Wisoff et al.'s criteria. However, these patients were treated with conservative surgery and adjuvant therapy. The OS rate was 85% at 10.9 years. Our study is similarly marked by a conservative surgical approach. For example, only three patients had STR at diagnosis and 31 had no surgery at diagnosis. The outcome of our patients was at least as favorable as those following aggressive surgical intervention,4 suggesting that the role for radical surgery is limited in most newly diagnosed patients.
Some reports suggest that H and/or C tumors in children with a clinical diagnosis of NF1 behave in a more benign manner,1, 2, 21, 27 whereas others suggest that NF1 status has no influence.28 In our study, patients with the phenotypic diagnosis of NF1 had a slightly better OS (P = 0.056) in univariate analysis. An even more robust prognostic factor was the presence of the typical nonenhancing, bright T2-weighted lesions often noted on MRI scans of patients with NF1. On multivariate analysis, patients with intracranial lesions had a significantly better PFS (P = 0.015).
Effect of Primary Treatment on Outcome
There was no statistically significant difference in OS (P = 0.23) based on the initial treatment approach (i.e., observation, chemotherapy, radiation; Fig. 2). Therefore, observation of 37 newly diagnosed, relatively asymptomatic patients was warranted.
Of the 20 patients who received chemotherapy initially, 18 were treated with standard alkylator or platinum-based regimens. An objective response was seen in only four patients (20%). However, 65% had SD, delaying the need for RT or other salvage treatment for a median of 24 months (range, 0.3–5 years). These observations confirm the findings of Packer et al.,29 Janss et al.,2 and Silva et al.19 These authors reported that chemotherapy may significantly delay the use of radiation in younger patients.
Radiotherapy has often been the standard treatment approach for older patients with newly diagnosed midline gliomas. As much as a 90% PFS has been reported following irradiation. However, this result is related to series composed largely of optic chiasmatic tumors, making a direct comparison to the current group of H/C tumors difficult.6, 30–33 The current cohort of patients who initially underwent irradiation had a 6-year PFS rate of 69%.5 There was no difference in OS among patients initially treated with radiation, chemotherapy, or observation. However, in multivariate analysis, initial treatment with irradiation was a borderline significant factor for PFS (P = 0.056). In addition, 6 (55%) of 11 patients who progressed on chemotherapy and were salvaged with irradiation are disease free at a median of 3.4 years from first progression (range, 0.52–7.8 years). Therefore, irradiation achieves durable disease control in a significant cohort of patients.6, 30–33 There was no difference in OS among those initially treated with radiation, chemotherapy, or observation. A delay in utilizing irradiation, following initial observation or chemotherapy, is justifiable and supports the generally accepted current practice of delaying irradiation in favor of observation or chemotherapy, especially in younger patients.
Effect of Tumor Location on Outcome
Scott and Mickle34 were the first to anatomically define tumor localization by CT analysis in suprasellar astrocytomas, demonstrating the apparent prevalence of lesions involving the H and C regions. Similar to Hoffman et al.'s report,3 univariate analysis of our cohort showed patients with C tumors had a better PFS than those with H/C tumors (P = 0.045).
Because OS rates are excellent among these patients (86 ± 5%), attention to the impact of tumor and treatment on functional outcome is important. Previous reports have only made limited attempts to document the neurocognitive outcome of these patients. Janss et al.2 reported that among nine children who received serial cognitive testing, IQ scores decreased a median of 12 points from baseline (range, 10–30 points) in patients who underwent radiation. Similarly, Sutton et al.,20 Pierce et al.,6 and Tao et al.30 reported that patients with progressive optic pathway gliomas were more likely to have learning difficulties after RT. However, these data are difficult to interpret as there have been few attempts at serial assessment of IQ in large cohorts.
The current study is notable for serial neuropsychological testing in 31 patients. The IQ analysis subset was representative of the entire cohort. The estimated measurement of the IQ at diagnosis was age related. Preschoolers (younger than 5 years old) had an estimated mean IQ score of 79.1 compared with 96.3 for school-aged children (P = 0.0003). In addition, IQ scores did not change significantly over time as patients underwent treatment. In contrast to others, our analysis also suggests that the use of local irradiation alone did not significantly affect IQ scores over time. There was no significant difference in the change in general intelligence over time between irradiated (n = 14) and nonirradiated patients (n = 17; P = 0.84). Our data suggest that the tumor itself is an important cause of the patients' decline in IQ and that the problem is present at diagnosis rather than being exclusively a result of treatment. Similar findings are apparent in an ongoing prospective series including intracranial astrocytomas and ependymomas, preliminary results of which have been recently published by Merchant et al.35
Collet-Solberg et al.36 reported that 42% of children with H/C tumors develop endocrinopathies as a result of field irradiation and surgery. However, these attributions neglect to account for the likely and continued evolution of various endocrinopathies that would otherwise occur in the absence of treatment. The tumor itself is a significant contributor to endocrine dysfunction. Because the majority of patients eventually need treatment, there is little or no published serial endocrine data in an untreated population to serve as a basis of discussing the effects of treatment. At the very least, irradiation may hasten the onset of new endocrinopathies, but it is difficult to suggest that it is the cause. Other factors (age at diagnosis, extent of disease, timing, and the number of episodes of PD) are also likely to play a role in the development of these problems. Therefore, the effects of treatment on the incidence of endocrinopathies are difficult to evaluate independently of other contributing factors.
In the current study, the incidence of endocrinopathy was 71%. Patients initially treated with irradiation had a significantly higher cumulative incidence of endocrinopathy at 3 years compared with others (P = 0.008), suggesting that irradiation may accelerate the occurrence of these events. Hypothyroidism is the most common endocrinopathy observed without treatment (18%). Growth hormone deficiency is the most common endocrinopathy observed after irradiation (60%).
In conclusion, the most appropriate management for each patient depends on many factors (e.g., age, NF1 status, tumor location, endocrine, ophthalmologic, and neuropsychological status, and previous therapy). We recommend that asymptomatic patients, especially those with NF1 lesions, be followed until they become symptomatic or have objective imaging progression. For younger symptomatic patients, a platinum-based chemotherapy regimen delays the use of irradiation. For older symptomatic patients, RT remains the standard therapy. The OS rate is not affected by initial observation in relatively asymptomatic patients or by chemotherapy in the younger cohort. Despite the favorable OS rates, survivors of H and/or C tumors are at high risk for endocrine and neurocognitive problems as a result of the tumor itself as well as therapy.
The authors thank Jennifer Havens, Anne Marie Fraga, Jana Freeman, Laura Cooley, and Keith Hude for helping with the collection of the data; Jason Schroeder for assistance with the statistical analysis; Patsy Burnside for typing the manuscript; and Dr. Peter Burger for review of pathology.