1 INTRODUCTION

Pleomorphic xanthoastrocytomas (PXAs) are rare neuroglial tumors that occur predominantly in children and young adults.1, 2 They account for <1% of primary brain tumors and are considered World Health Organization (WHO) grade II tumors.1, 3-5 They are associated with favorable prognosis with respective 5-year and 10-year overall survival (OS) rates of about 77.8% and 68.5%,2, 3, 5 with an approximate 5-year progression-free survival (PFS) rate of 68.3%.2, 5 However, these tumors may transform into anaplastic PXAs (aPXAs) in 15%-20% of patients.3, 5, 6

aPXAs are defined as a subset of PXAs in the revised 2016 WHO classification for central nervous system (CNS) tumors. These tumors are considered WHO grade III with their increased mitotic activity of ≥5 per high-powered fields.1, 2, 5, 7 They present with a higher likelihood for necrosis, proliferation, absent pericellular reticulin, and infiltration.2, 7-9

Complete surgical resection presently serves as the primary treatment for PXAs,3, 10 while the role of adjuvant radiotherapy (RT) and chemotherapy in managing this disease remains undefined. To gain a better understanding of care patterns for pediatric PXAs, the current study aims to evaluate the effect of treatment, particularly RT, on OS using the United States National Cancer Database (NCDB).

2 METHODS

The NCDB was established in 1988 by the American College of Surgeons Commission on Cancer (CoC) and the American Cancer Society.11 It is a hospital-based cancer registry that collects information on patients receiving care from a CoC-accredited institution, which comprises 30% of 5000+ hospitals in the United States. Information reported in the NCDB include deidentified sociodemographic, treatment, tumor, and survival data of oncology patients.

Based on the International Classification of Diseases for Oncology, 3rd edition Histology code of 9424/3, a search through the NCDB brain dataset was conducted for pediatric patients (≤18 years) diagnosed with PXAs in 2004-2015. The data were exported on February 28, 2020, by authors Y.Y., C.B., and Y.C. These authors verified the data to ensure no problem occurred while data export. Descriptive analysis was conducted summarizing patient demographic and clinical characteristics. To increase transparency concerning the impact of patients with missing values in the study analyses, footnotes referring to the inclusion or exclusion of these patients were provided for each table.

Patients were categorized in two treatment groups: “surgery only” and “surgery and RT.” To assess for differences in patient characteristics between treatment groups, Kruskal–Wallis test was used for continuous variables (i.e., age and tumor size); Cochran-Armitage test was used for three-level ordinal variables (i.e., comorbidity score); and Fisher's exact test was used for other categorical variables. Analyses concerning extent of resection (EOR) were based on patients diagnosed between 2010 and 2015 as such information was made available in the database. Our variables were analyzed utilizing log rank testing and Kaplan–Meier modeling. OS was defined as the time in months from diagnosis to death or last follow-up, whichever occurs first. Univariate and multivariable Cox proportional hazard ratio models were fitted to the data. The final multivariable model was determined based on purposeful selection combined with Akaike's information criterion and Bayesian information criterion. All analyses were performed using R-version 3.6.2. p-values of <.05 were considered statistically significant. Data consisting of <20 patients were deidentified with asterisks (*) and reported only in percentages.

3 RESULTS

For the 224 eligible, pediatric patients, an almost equal occurrence was found between males and females with a median diagnosis age of 14 years (Table 1). Most patients identified as Caucasian (63.8%), followed by Black (16.1%) and Hispanic patients (14.7%). Many resided in a region where ≥14% of adults had no high school degree (60.3%) and earned a median income of <$46, 000 (63.8%). Most lived in a metropolitan area (62.1%) and had private insurance (62.5%). Almost an equal number of patients were observed in 2004-2009 and 2010-2015, with many presenting without comorbidities (90.2%). The tumor primarily occurred at the temporal lobe (34.4%) and had a median tumor size of 40 mm (range = 5-988 mm). The WHO grades were predominantly unspecified (68.8%), but WHO grade IV tumors and PXAs (i.e., WHO grade II), respectively, comprised of 12.9% and 11.6%. A limited number of patients were biopsied (8.0%). Most patients underwent surgery (97.3%). Of the 222 patients with known RT delivery, 18.8% received RT at a median dose of 54.0 Gy (range = 5.4-83.4 Gy). The preferred RT modality was intensity-modulated RT (35.7%), followed by photon therapy (31.0%). Chemotherapy was administered to 14.7% of patients.

TABLE 1. Overall characteristics of pediatric pleomorphic xanthoastrocytomas (N = 224) Variable N (%) Age (years, N = 224) Median (range) 14 (1-18) Sex (N = 224) Male 115 (51.3%) Female 109 (48.7%) Race (N = 224) Caucasian 143 (63.8%) Black 36 (16.1%) Hispanic 33 (14.7%) Other (5.4%)a* Median income (N = 215) <$46 000 143 (63.8%) ≥$46 000 72 (32.1%) (Missing)b (4.0%)a* % without high school degrees (N = 215) ≥14% 135 (60.3%) <14% 80 (35.7%) (Missing)b (4.0%)a* County population (N = 215) Metropolitan, >250 000 139 (62.1%) Nonmetropolitan 76 (33.9%) (Missing)b (4.0%)a* Insurance status (N = 215) Private 140 (62.5%) Government 68 (30.4%) None (3.1%)a* (Missing)b (4.0%)a* Year of diagnosis (N = 224) 2004–2009 113 (50.4%) 2010–2015 111 (49.6%) Comorbidities (N = 224) 0 202 (90.2%) 1 (6.7%)a* ≥2 (3.1%)a* Tumor location (N = 224) Temporal lobe 77 (34.4%) Frontal lobe 46 (20.5%) Overlapping lesion of brain 29 (12.9%) Others 27 (12.1%) Parietal lobe 25 (11.2%) Occipital lobe (8.9%)a* Tumor size (mm, N = 161) Median (range) 40 (5–988) WHO Grade (N = 224) NOS 154 (68.8%) I (2.7%)a* II 26 (11.6%) III (4.0%)a* IV 29 (12.9%) Biopsy (N = 222) None 203 (90.6%) Biopsy, primary site (8.0%)a* Surgical procedure with a bypass, no biopsy (0.4%)a* (Missing)b (0.9%)a* Surgery (N = 224) Yes 218 (97.3%) No (2.7%)a* Extent of Surgery (2010–2015 data, N = 103) GTR 72 (64.9%) STR 31 (27.9%) (Missing)b (7.2%)a* RT (N = 222) No 180 (80.4%) Yes 42 (18.8%) (Missing)b (0.9%)a* RT Dose (GY, N = 37) Median (range) 54 (5.4–83.4) RT Modality (N = 42) Intensity-modulated RT (35.7%)a* Photons (31.0%)a* External beam, NOS (19.0%)a* Conformal or 3-D therapy (7.1%)a* Stereotactic radiosurgery, NOS (2.4%)a* Gamma Knife (2.4%)a* Other, NOS (2.4%)a* (Missing)b (0.05%)a* Chemotherapy (N = 222) No 189 (85.1%) Yes 33 (14.7%) (Missing)b (0.9%)a* Abbreviations: GTR, Gross total resection; NOS, not otherwise specified; RT, radiotherapy; STR, subtotal resection.

Patients who underwent surgery and had complete RT information comprised of 216 patients, of which 81.0% had surgery only (Table 2). Based on the Fisher's exact tests, factors significantly associated with RT delivery included: WHO grade (P < 0.001) and chemotherapy administration (p < .001). Most, if not all, PXAs with WHO grades I, II, and III were typically resected without RT (100%, 96%, and 55.6%, respectively), while more WHO grade IV tumors underwent RT postsurgery (55.6%). Those receiving chemotherapy more likely received surgery with RT (83.9%).

TABLE 2. Key variables associated with surgery ± Radiotherapy in pediatric pleomorphic xanthoastrocytomas (N = 216) Surgery Surgery + RT p-value N (Row %) N (Row %) WHO grade N = 175 (81.0%) N = 41 (19.0%) <.001 I (100%)a* (0.0%)a* II 24 (96%) (4%)a* III (55.6%)a* (44.4%)a* IV (44.4%)a* (55.6%)a* NOS 128 (85.9%) 21 (14.1%) Chemotherapy N = 174 (80.9%) N = 41 (19.1%) <.001 No 169 (91.8%) (8.2%)a* Yes (16.1%)a* 26 (83.9%) (Missing)b (100%)a* 0 (0.0%) Abbreviations: NOS, Not otherwise specified; RT, radiotherapy.

The median OS time of the entire cohort was indeterminate due to the limited number of events (i.e., deaths) observed. However, 3- and 5-year OS rates were calculated for selected groups of subjects (Table 3). The 3-year OS for patients, respectively, without and with RT were 94.4% (95% confidence interval [CI] = 89.2%-97.4%) and 65.2% (95% CI = 46.7%-78.7%). Based on the log rank test and Kaplan–Meier curves, OS is significantly longer for non-RT compared to RT recipients (p < .001, Table 4, Figure 1). Adjuvant chemotherapy administration also affected OS (p < .001, Table 4), with higher 3-year OS for nonchemotherapy (94.8%, 95% CI = 89.9%-97.4%) versus chemotherapy recipients (51.4%, 95% CI = 30.1%-69.1%).

TABLE 3. Summary of overall survival rates of pediatric pleomorphic xanthoastrocytomas Variable N 3-year OS % (95% CI) 5-year OS % (95% CI) Extent of Surgeryb,c, b,c (N = 82) GTR 60 97.5 (83.5-99.6) 97.5 (83.5-99.6) STR 22 73.7 (47.6-88.2) 61.5 (29.4-82.4) Extent of Surgery + Chemotherapyb,c, b,c (N = 82) GTR 56 97.3 (82.3-99.6) 97.3 (82.3-99.6) GTR + Chemotherapy a* 100 (N/A) N/A STR a* 85.1 (52.3–96.1) 70.9 (30.9-90.4) STR + Chemotherapy a* 40 (5.2-75.3) N/A Extent of Surgery + RTb,c, b,c (N = 81) GTR 50 97 (80.4-99.6) 97 (80.4-99.6) GTR + RT a* 100 (N/A) N/A STR a* 93.3 (61.3–99.0) 77.8 (31.6-94.7) STR + RT a* 33.3 (4.6–67.6) N/A RT (N = 199)c No 161 94.4 (89.2-97.2) 88.4 (81.0-93.1) Yes 38 65.2 (46.7–78.7) 60.9 (41.7-75.5) Chemotherapy (N = 199)c No 171 94.8 (89.9–97.4) 89.2 (82.2-93.5) Yes 28 51.4 (30.1–69.1) 45 (23.6-64.2) Race (N = 201)c Caucasian 127 87.5 (79.7-92.4) 82.5 (73.4-88.7) Black 31 96.7 (78.6-99.5) 92.1 (71.3-98.0) Hispanic 32 92.6 (73.5-98.1) 81.7 (57.1-93.0) Other a* 70.1 (32.3-89.5) 70.1 (32.3–89.5) WHO Grade (N = 201)c I a* 100 (N/A) 100 (N/A) II 22 95.2 (70.7–99.3) 89.3 (63.2-97.2) III a* 75 (31.5–93.1) 60 (19.5-85.2) IV 24 70.3 (44.7-85.7) 70.3 (44.7–85.7) NOS 143 91.2 (84.7-95.1) 85.5 (77.1-91.0) Abbreviations: CI, Confidence interval; GTR, gross total resection; N/A, not available; NOS, not otherwise specified; OS, overall survival; RT, radiotherapy; STR, subtotal resection. TABLE 4. Key variables determined via log rank test for pediatric pleomorphic xanthoastrocytomas Variables N p-value GTR versus STRa,b, a,b 82 <.001 Extent of Surgery + Chemotherapya,b, a,b 81 <.001 Extent of Surgery + RTa,b, a,b 81 Not reliable ± Adjuvant RTb 199 <.001 ± Adjuvant Chemotherapyb 199 <.001 Raceb 201 .175 WHO Gradeb 201 .088 Abbreviations: GTR, Gross total resection; RT, radiotherapy; STR, subtotal resection.

Kaplan–Meier plot of Pediatric Pleomorphic Xanthoastrocytomas

The 2010-2015 cohort (n = 103) indicated that most had gross-total resection (GTR, 64.9%) rather than subtotal resection (STR) (Table 1). With only 79.6% of the 2010-2015 cohort having known survival outcomes, the log rank test revealed GTR impacted survival (p < .001, Table 4). Increased OS was observed in patients who underwent GTR rather than STR (3-year OS = 97.5%, 95% CI = 83.5%-99.6% vs. 73.7%, 95% CI = 47.6%-88.2%, Table 3). EOR and chemotherapy also affected survival (p < .001, Table 4). Whereas minor differences in the 3-year OS were observed in patients having GTR with or without chemotherapy, a major difference in 3-year OS was seen for patients having STR with and without chemotherapy (3-year OS = 40%, 95%CI = 5.2%-75.3% vs. 85.1%, 95% CI = 52.3%-96.1%, respectively). Although there were too few patients to make the log rank test reliable, similar results were observed for those receiving surgery with or without RT (p < .001, Table 4). There was little difference in 3-year OS for patients having GTR with and without adjuvant RT, but increased 3-year survival was observed in patients undergoing only STR compared to patients proceeding with adjuvant RT (3-year OS = 93.3%, 95% CI = 61.3%-99.0% vs. 33.3%, 95% CI = 4.6%-67.6%).

Univariate cox regression analyses indicated EOR (p < .001) and the administration of RT (p < .001) and chemotherapy (p < .001) independently affect OS (Table 5). STR patients had a higher mortality risk (hazard ratio [HR] = 17.44, 95% CI = 2.10-144.90) relative to GTR patients. Compared to patients without adjuvant therapies, RT and chemotherapy recipients respectively had HRs of 3.82 (95% CI = 1.85-7.90) and 6.68 (95% CI = 3.21-13.89). Only receipt of chemotherapy remained significant in the multivariable model.

TABLE 5. Key variables determined via cox regression analyses for pediatric pleomorphic xanthoastrocytomas Variable N % of Events HR (95% CI) p-value SexU(N = 201) .229 Male 104 10.6% Reference Female 97 19.6% 1.57 (0.75-3.31) RaceU(N = 201) .175 Caucasian 127 15.0% Reference Black 31 6.5% 0.4 (0.09-1.73) Hispanic 32 18.8% 1.46 (0.58-3.67) Other 11 27.3% 2.4 (0.71-8.14) WHO gradeU (N = 201) .088 I a* 0% 0 (0-Inf) II 22 9.1% Reference III a* 37.5% 4.57 (0.76-27.40) IV 24 25% 3.81 (0.77-18.96) NOS 143 13.3% 1.57 (0.37-7.75) SurgeryU(N = 201) .533 No a* 25% Reference Yes 197 14.7% 0.54 (0.07-3.95) Extent of surgeryb,U(N = 82) <.001 GTR 60 1.7% Reference STR 22 27.3% 17.44 (2.10–144.90) (Missing)c a* 16.7% N/A RTU(N = 199) <.001 No 161 10.6% Reference Yes 38 34.2% 3.82 (1.85–7.90) (Missing)c a* 0.0% N/A ChemotherapyU,M(N = 199) <.001 No 171 9.9% Reference Yes 28 46.4% 6.68 (3.21–13.89) (Missing)c a* 0.0% N/A Abbreviations: CI, Confidence interval; GTR, gross total resection; HR, hazard ratio; RT, radiotherapy; STR, subtotal resection; U, Univariate; M, Multivariable. 4 DISCUSSION

PXAs were first detailed by Kepes et al in 1979.4, 5, 12 However, due to the infrequent occurrence of this disease, current literature concerning PXAs comprise primarily of case reports and scarce case series.3, 5 The publication on PXAs by Perkins et al3 (n = 215) resembles the present study, as it examined the effects of demographic, clinical, and treatment variables on OS for patients with PXAs. Perkins et al queried the Surveillance, Epidemiology, and End Results cancer registry for children and adults diagnosed with PXAs between 1981 and 2007. Relative to Perkins et al, the present study contains twice the number of pediatric patients. The present study's focus on children over the past 12-years also limited the effect of temporal bias, as it confined diagnostic and treatment variation on a multi-institutional level.

Even though Perkins et al3 included both children and adults in their study, they similarly observed that most patients identified as Caucasian (80%). Their data also indicated that race did not impact survival (p = .40). Our analyses further indicated likelihood of RT delivery and mortality risk were not significantly associated with patient demographics. RT delivery was more impacted by WHO grade and chemotherapy delivery, while significant mortality risk was associated with EOR and administration of RT and chemotherapy.

Histologically, PXAs are characterized by astrocytes that usually have large mono- or polynucleated structures and spindle cells with a mesenchymal appearance oriented in intersecting bundles.1,