Risk factors for the prognosis of pediatric medulloblastoma: a retrospective analysis of 40 cases

OBJECTIVES: In this study, we evaluated the association of molecular subtypes, clinical characteristics and pathological types with the prognosis of patients with medulloblastoma. METHODS: We analyzed forty patients with medulloblastoma who underwent surgical resection at our center between January 2004 and June 2014. Risk factors associated with survival, disease progression and recurrence were analyzed with a univariate Cox regression analysis, and the identified significant risk factors were further analyzed by Kaplan-Meier survival curves. RESULTS: Factors associated with overall survival included M stage (p=0.014), calcification (p=0.012), postoperative treatment, postoperative Karnofsky Performance Scale (KPS) score (p=0.015), and molecular subtype (p=0.005 for WNT and p=0.008 for SHH). Number of symptoms (p=0.029), M stage (p ;2 and ≥M1 stage without postoperative radiotherapy. The risk of recurrence increased with advanced M stage. Protective factors for recurrence included M0 stage and a combination of chemotherapy and radiotherapy. CONCLUSION: We identified the risk factors associated with survival, disease progression and recurrence of medulloblastoma patients. This information is helpful for understanding the prognostic factors related to medulloblastoma.


' INTRODUCTION
Medulloblastoma is a malignant, invasive embryonal tumor in the cerebellum or fourth ventricle and accounts for 12-25% of all central nervous system tumors. Medulloblastoma is the most common malignancy affecting children with an annual incidence of five per 100,000 among children o15 years of age (1). Although surgery remains the major treatment for medulloblastoma, there is still controversy regarding the impact of resection on the prognosis of patients with medulloblastoma. Furthermore, in cases of adherent medulloblastoma to the brainstem, complete resection is extremely difficult. In addition, metastasis via the cerebrospinal fluid is common; thus, medulloblastoma patients often have a poor prognosis and a high mortality rate (2).
Clinically, the prognosis of patients with medulloblastoma is often determined according to the pathological type, which also provides a reference for the application of adjunctive therapies, such as radiotherapy and chemotherapy (3,4). Currently, the World Health Organization (WHO) classification system for medulloblastoma is based on histomorphology. However, patients with the same pathological type of medulloblastoma still have distinct genetic backgrounds. Therefore, the prognosis of patients with medulloblastoma may vary even within the same WHO pathological type (2).
Recent studies on medulloblastoma have revealed that it is more accurate to stratify risk based on the molecular phenotype, which is also helpful to guide clinical treatment and determine clinical prognosis (5,6). Currently, medulloblastoma is divided into several subtypes according to the molecular phenotypes: WNT, Sonic hedgehog (SHH) and non-SHH/WNT (7,8). In addition, Ellison et al. (2) differentiated medulloblastoma subtypes by immunohistochemistry, and the authors validated their findings using microarray analyses. Ellison et al. found that it is feasible to differentiate different subtypes of medulloblastoma by immunohistochemistry (2). In the present study, the molecular phenotyping methods for medulloblastoma reported by Ellison et al. (2) were utilized to detect the expression of Yes associated protein 1 (YAP1) and GRB2-associated protein 1 (GAB1) by immunohistochemistry.
YAP1 is an auxiliary initiator of oncogene transcription and can promote cellular proliferation and transformation (9). Gene expression profiling of medulloblastoma reveals that it is highly expressed in the WNT and SHH subtypes of medulloblastoma and is not observed in the other two medulloblastoma subtypes (9,10). In contrast, GAB1 belongs to the Gab family and has specificity in the SHH signaling pathway of medulloblastoma cells (2). In the present study, YAP1 protein served as a specific marker of the WNT and SHH medulloblastoma subtypes while GAB1 was used as a specific marker of the SHH medulloblastoma subtype.
In this retrospective observational study, we evaluated the association of molecular subtypes, clinical characteristics and pathological types with the prognosis of patients with medulloblastoma. The primary outcome was overall survival (OS). Risk factors associated with survival, disease progression and recurrence were analyzed with a univariate Cox regression analysis, and the identified significant risk factors were further analyzed via Kaplan-Meier survival curves. YAP1 served as a specific marker of the WNT and SHH subtypes; GAB1 served as a specific marker of the SHH subtype. This analysis may provide information pertinent to treatment decisions for patients with medulloblastoma.

Study design
In this retrospective analysis, we reviewed the medical records of 40 children with pathologically proven medulloblastoma who underwent surgical resection at the Affiliated Children's Hospital of Fudan University between January 2004 and June 2014. Medulloblastoma was diagnosed and classified into the following subtypes according to the 2016 WHO classification system of central nervous system tumors (11): classic subtype, desmoplastic/nodular subtype, extensive nodularity subtype or large cell/anaplastic subtype.
Inclusion criteria included the following: absence of another severe disease diagnosis, a complete medical record with followed-up data, and medulloblastoma tissues that were fixed in 10% neutral formalin and embedded in paraffin following resection. All deaths were a result of disease progression or recurrence; however, one patient who died within 3 days after surgery was excluded. Two other patients were excluded due to a lack of follow-up data. This study was approved by the institutional review board of Fudan University.

Radiotherapy and chemotherapy protocols
After 28 days following the resection procedure, patients underwent postoperative craniospinal irradiation (CSI) delivering a median craniospinal dose of 36 Gy with additional boosts to the posterior fossa up to 54.0-55.8 Gy weekly for 8 weeks. Chemotherapy was initiated 6 weeks after radiotherapy in eight 6-week courses consisting of 4 weeks of chemotherapy followed by 2 weeks of rest. Specifically, patients received an intravenous infusion of cisplatin (75 mg/m 2 ) on Day 0; intravenous bolus infusion of vincristine (1.5 mg/m 2 , max of 2 mg/dose) on Days 1, 7, and 14; and intravenous infusion of cyclophosphamide (1,000 mg/m 2 ) on Days 21 and 22. Examinations of the patients' skulls and spines by MRI were performed once every 12 weeks.

Survival analysis
The OS time was defined as the time interval between surgery and death or the last follow-up and was expressed in months. The disease progression-free survival (PFS) duration was defined as the time interval between date of surgery and date of progression-free, last follow-up, or death. The recurrence-free duration was calculated from the date of surgery to the date of recurrence, last follow-up, or death. Censored data were considered if the patient survived at the last follow-up and was marked in the survival curve.

Karnofsky performance scale
Postoperative Karnofsky performance scale (KPS) scores were determined for all patients during hospitalization. In this scale, a score of 100 indicates that the physical condition of the patient is normal without evidence of disease, while a score of 10 indicates rapid, fatal disease progression.

Immunohistochemistry
Immunochemistry analysis of YAP1 and GAB1 expression was performed as previously described (2). After surgical resection, medulloblastoma tissues were fixed in 4% neutral formalin and embedded in paraffin. The paraffin-embedded tissues provided by the Department of Pathology in the Affiliated Children's Hospital of Fudan University were cut into 4-mm sections. After routine processing with xylene, graded ethanol solutions, and 3% H 2 O 2 for 10 min, antigen retrieval was performed in 0.05 M citrate buffer (pH=6.0) at 100 o C for 5-10 min followed by blocking in goat serum for 10 min. Immunohistochemistry analyses were performed using a EnVision two-step immunohistochemistry system (DAKO, Kyoto, Japan) with anti-GAB1 polyclonal antibodies (1:50, Abcam; Cambridge, MA, USA) and anti-YAP1 polyclonal antibodies (1:50, Abcam) for 1 h at 37 o C. After washing, sections were incubated in HRP-conjugated goat anti-rabbit IgG (H+L; Jackson ImmunoResearch; West Grove, PA, USA) for 1 h at 37 o C. Visualization was performed with DAB.
The immunohistochemistry results were semi-quantified. Five fields were randomly selected from each section at a magnification of 200 Â , and the positive cells were counted and averaged. Cells positive for YAP1 exhibited brown granules in the nucleus and cytoplasm; GAB1-positive cells exhibited granules in the nucleus. Positive cells were counted to obtain an average. Sections undergoing hematoxylin and eosin (HE) staining also served as controls. At a magnification of 400 Â , the proportion of positive cells X30% at the strong positive area was regarded as positive (+) while the proportion of positive cells o30% was regarded as negative (-).

Statistical analysis
All data were presented as frequencies and percentages and were assessed with a Chi-squared test. A Cox proportional hazard model was performed to identify effectors of poor survival outcome. To quantify the strength of the association between a potential risk factor and death, disease progression, or medulloblastoma recurrence, hazard ratios (HR) and 95% confidence intervals (CI) were estimated and reported. Given the limited sample size, only significant variables with p-values o0.01 in the univariate analysis were used for further multivariable analyses. If both postoperative radiotherapy and postoperative chemotherapy met the criterion, the variable ''both postoperative radiotherapy and chemotherapy'' was used instead. In cases where no variable with p-values o0.01 was identified in the univariate analysis, a multivariable analysis was not performed. A p-value o0.05 was considered significant. All statistical analyses were twosided and performed using IBM SPSS statistics for Windows version 22.0 (IBM Corp., Armonk, NY, USA). '

Patient demographics
Patient characteristics are summarized in Table 1. Patients with partial or complete resection exhibited similarities in most characteristics except a higher percentage of patients with partial resection received radiotherapy after surgery than did patients with complete resection (100% vs. 50%, p=0.022). Among the 40 patients, most were males (n=29, 72.5%) and X3 years of age (n=29, 72.5%). Clinical features of central medulloblastoma (n=36, 90%), complete tumor resection (n=34, 85%) and classic subtype (n=33, 82.5%) were commonly observed among the patients. Other subtypes included the desmoplastic/nodular (D) subtype in 3% (n=1) and the large cell/anaplastic (LC/A) subtype in 15% (n=6). At least 80% of the patients presented with an M stage at M0 and T stage at T3 or above (Table 1). Calcification (n=14) or preoperative cerebral tonsillar herniation (n=9) was observed in 23%-35% of patients. Whereas 57.5% of patients (n=23) received radiotherapy, only 35% (n=14) were treated with chemotherapy after surgery. The postoperative KPS score was X80 in 80% of the patients (n=32), suggesting near to normal activity and either the absence of disease signs/symptoms or the presence of only mild disease signs/symptoms.

Kaplan-Meier survival analysis
The Kaplan-Meier curves displaying factors associated with OS are shown in Figure 1. Four of five patients with M1 stage or above died within 6 months after surgery, with OS rates of 20% at 6 months and 0% at 20 months ( Figure 1A).
OS rates among patients with M0 stage were 61.8% at 1 year, 54.1% at 2 years, 20.3% at 3 years, and 6.8% after 44 months ( Figure 1A). Eleven of 14 patients with calcification died within 13 months after surgery (OS rates of 35.7% at 6 months and 14.3% after 20 months) ( Figure 1B). OS rates in

A R T I C L E
patients without calcification were 67.5% at 1 year, 61.8% at 2 years, 23.2% at 3 years, and 7.7% after 44 months ( Figure 1B). The OS rate among patients not receiving radiotherapy was 23.5% at 1 year and 8.8% after 27 months versus 82.6% at 1 year, 56.5% at 27 months, and 0% at 44 months among patients receiving radiotherapy ( Figure 1C). OS rates among patients not treated with chemotherapy were 42.3% at 1 year, 24.7% at 2 years and 0% at 27 months, compared to 83.6% at 1 year, 74.3% at 2 years, 31.8% at 3 years and 10.6% after 44 months among patients treated with chemotherapy ( Figure 1D). The OS rate among patients receiving both chemotherapy and radiotherapy were 100% at 2 years and 40% at 3 years versus 44.4% at 1 year and 29.0% at 2 years among patients without chemotherapy and radiotherapy ( Figure 1E).

Univariate analysis of potential predictors of recurrence
The univariate analysis indicated that recurrence was associated with a higher risk of death (HR=2.49, 95% CI: 1.07-40.80, p=0.040; Table 2); the 1-and 3-year OS rates of patients without recurrence were 70% and 42%, compared to 43.7% and 0% among patients with recurrence. We next determined which factors were associated with recurrence via a univariate analysis. As shown in Table 4, the risk of recurrence increased with advanced M stage (HR=30.71; 95% CI: 3.92-240.44, p=0.001). In contrast, patients receiving both chemotherapy and radiotherapy were less likely to experience recurrence than patients with only one therapy or without either radiotherapy or chemotherapy (HR=0.21, 95% CI: 0.05-0.93, p=0.040).

' DISCUSSION
In this retrospective analysis, we identified the clinical characteristics, including molecular subtypes, and treatment outcomes associated with the prognosis of pediatric patients with medulloblastoma in China. M stage, calcification, postoperative treatment (radiotherapy, chemotherapy, and both), postoperative KPS score, and molecular subtype were all associated with the OS of medulloblastoma patients. Factors associated with disease progression included number of symptoms, M stage and postoperative radiotherapy. M stage and postoperative radiotherapy or chemotherapy were associated with recurrence. Considered together, molecular subtyping of medulloblastoma was more predictive of survival than histopathology in patients undergoing adjuvant therapy. This is the first study to report the clinical features, prognoses, and risk factors of patients with pediatric medulloblastoma among a Chinese Han population. As a single-center study in China, this report has inherent, unique ethnic characteristics, which could be regarded as important supplementary information for global studies regarding pediatric medulloblastoma. In addition, this is the first study to compare the prognosis obtained using molecular typing compared to pathological classification in a single-center study. Specifically, this study highlights the advantages of molecular typing, which provides a more intuitive and reliable indicator of molecular classification for prognosis than pathological classification.
In the present study, no differences among patient outcomes were detected between the pathological types. Because the prognosis of patients with the same pathological type of medulloblastoma can be drastically different due to varying genetic backgrounds (12), the development of new molecular subtyping of medulloblastoma is necessary. In the present study, molecular subtyping analyses revealed that almost half of the children presented with the non-SHH/WNT subtype. Furthermore, our univariate and multivariable analyses both indicated that the prognoses of patients with the WNT subtype was the best followed by the SHH subtype of medulloblastoma. These findings are consistent with another study of medulloblastoma in China (13). Our results further confirmed the prognostic superiority of determining molecular subtypes over pathological types. However, the molecular subtypes (as determined by YAP1 and GAB1) were not associated with disease recurrence or progression. Therefore, further studies are required to identify additional markers, such as glutamate (a predictive marker for patient survival for pediatric medulloblastoma (14)), to improve molecular subtyping of medulloblastoma among children. In addition, consensus regarding the method for identifying medulloblastoma subtypes (e.g., immunohistochemistry, CTNNB1 mutation analysis, or quantitative PCR) (15) should be reached through additional studies. Finally, larger studies will permit the patients to be further divided into those having Group 3 and Group 4 tumors in order to more completely subtype the tumors and their prognostic impact.
In the present study, disease progression was associated with the presence of 42 symptoms, which might be related to the special location of medulloblastoma in children. Medulloblastoma is usually present in the midline of the posterior fossa and may cause disordered cerebrospinal fluid circulation resulting in cerebellar dysfunction characterized by intracranial hypertension and cerebellar tissue destruction (16). The clinical symptoms mainly include headache, vomiting, ataxia, nystagmus, cranial nerve palsy, an increase in head circumference, cerebral hernia and secondary epilepsy. Nervous system injury caused by the cancer or cerebral hernia due to intracranial hypertension can directly threaten the life of the patient. Previous studies have confirmed that the time interval between disease onset and surgery may directly affect the prognosis of a patient with medulloblastoma (17), which may be related to greater symptom severity.

A R T I C L E
The staging for medulloblastoma is mainly based on the Chang staging system, which is based on the pre-operative imaging and intra-operative findings to determine M stage and T stage. M stage is better for assessing the prognosis of children with medulloblastoma than T stage (1,15), which is consistent with the present study in which OS as well as disease progression and recurrence were significantly associated with M stage (M0 vs. XM1). However, no such associations were observed with T stage (T1-2 vs. T3-4).
The postoperative KPS score has also been used in the determination of postoperative prognosis. In the present study, postoperative KPS scores X80 were associated with significantly longer OS. In addition to the KPS score, OS was also associated with tumor calcification that could be

A R T I C L E
visualized with an imaging examination. Our univariate and multivariable analyses both indicated that patients receiving postoperative radiotherapy or chemotherapy exhibited significantly better OS rates than patients not receiving postoperative radiotherapy or chemotherapy, which is consistent with previous studies (12,13). Given the toxicity of radiotherapy and chemotherapy to the nervous system in children (18), in depth studies are necessary to examine individualized therapies according to the risk stratification of medulloblastoma patients. For example, the dose of radiation or chemotherapeutics may be reduced in children with a low risk for recurrence, which may minimize the associated toxicity without compromising the therapeutic effectiveness.

A R T I C L E
An age of o3 years has been identified as a factor associated with poor prognosis in medulloblastoma patients (19). However, no such association was observed in the present study. In addition, previous studies have shown that the extent of resection was a key factor affecting the prognosis of medulloblastoma patients (20,21). Pogorzala et al. (22) reported that incomplete surgical resection was associated with poor outcomes. However, OS times were similar in children with total resection and in those with subtotal resection. Furthermore, OS time was comparable between patients with and without tumors that were adherent to the brainstem. Although these differences may be due to the small sample size in the present study, we speculate that total resection should not be performed if it is difficult to completely remove the cancer, especially given the evidence showing that the residual cancer cells will be cleared by postoperative radiotherapy (23).
This study is limited in that the results are from a single institution and the study size was small, which was due in part due to patients not seeking therapy as a result of a poor prognosis or financial burden. In addition, the neurosurgical department at our hospital is relatively new. Thus, the results need to be confirmed with larger sample sizes. In addition, the precise types of chemotherapy and radiotherapy and their influence on patient prognosis were not determined. Moreover, only 35% of the patients were treated with chemotherapy, and 57.5% of the patients were treated with radiotherapy, which was due, in part, to the large proportion of patients under 3 years of age (27.5%). However, this factor may have affected the OS rate, which was only 32.5%. Finally, limitations regarding the immunohistochemistry method utilized for subtype classification did not permit the separation of medulloblastoma groups 3 and 4 molecular, which could be identified via mRNA analysis (24). Therefore, differences between the subtypes would not have been identified.
Molecular subtypes are better determinants of the prognoses of medulloblastoma patients than pathological types, which may be used to guide the therapy of medulloblastoma. Further studies are necessary to validate our results with a larger sample size and to identify and improve the novel markers of different medulloblastoma subtypes to make this approach more reliable.

' ACKNOWLEDGMENTS
We thank Professor Zheng Shan, the vice president of Children's hospital of Fudan University, for supporting this study. This study was supported by grants from the Shanghai Municipal Health and Family Planning Commission (NSFS No. 20164Y0086) and the Natural Science Foundation of Shanghai (NSFS No. 16ZR1403700).

' AUTHOR CONTRIBUTIONS
Li H guarantees the integrity of the entire study and was responsible for study concepts and manuscript review. Shi W participated in the study design, definition of intellectual content, data analyses, statistical analyses and manuscript editing. Yu J participated in the literature research, clinical studies, experimental studies, data acquisition and manuscript preparation. Zhao R participated in literature research and provided theoretical guidance during the revision.