Evaluation of Microscopic Extension in High-grade Glioma Using Macropathology: Determination of Optimal Clinical Target Volume Margins for Radiotherapy

Introduction: One of the most dicult steps of radiotherapy is to determine the clinical target volume (CTV) based on microscopic extension (ME). In this study, we performed a consecutive macropathologic analysis to assess ME in high-grade glioma (HGG) to determine appropriate CTV margins for radiotherapy. Material and methods: The study included 30 HGG patients with tumors located in non-functional areas, and supra total resection was performed. The ME distance from the edge of the tumor to the microscopic tumor cells surrounding brain tissue was measured. Associations between the extent of ME and clinicopathological characteristics were evaluated by multivariate linear regression (MVLR) analysis. An ME predictive model was developed based on the MVLR model. Meanwhile, in order to validate the feasibility and safety of this model, we prospectively recruited another 30 HGG patients in a 1:1 ratio to receive guideline-based radiotherapy (RT) or model-based RT. The overall response rate (ORR) was evaluated during a follow-up of 14 months. Results: Between June 2017 and July 2019, 652 pathologic slides were analyzed. The mean ME distance was 1.70cm (range, 0.63 to 2.87cm). The MVLR analysis identied that pathologic grade, subventricular zone (SVZ) contact and O 6 -methylguanine-DNA methyltransferase (MGMT) methylation, isocitrate dehydrogenase (IDH) mutation and 1p/19q co-deletion status were independent variables predicting ME (all P < 0.05). A multivariable prediction model was developed as follows: Y ME = 0.672 + 0.513X Grade + 0.380X SVZ + 0.439X MGMT + 0.320X IDH + 0.333X 1p/19q . The R-square value of goodness of t was 0.780. In our validation cohort, after a mean follow-up of 7.65months, patients in the model RT group had a higher ORR than those in the guideline RT group (66.7% vs. 20%, P = 0.01). Conclusion: ME was heterogeneously distributed across different grades of gliomas according to the tumor location and molecular marker status, which indicated that CTV delineation should be individualized. The model could predict the ME of HGG, which may help clinicians determine the CTV for individual patients.


Introduction
High-grade glioma (HGG) is the most commonly diagnosed primary brain tumor, and has a remarkable tendency to in ltrate the surrounding brain tissue. To protect brain function, gross total resection through surgery becomes almost impossible. Therefore, radiotherapy (RT) has become the main treatment for HGG patients. In National Comprehensive Cancer Network (NCCN) guidelines, for the delineation of a clinical target volume (CTV), a margin accounting for subdiagnostic tumor in ltration, of 1-2.5cm for HGG in terms of the volumetric expansion of the gross target volume (GTV) is recommended. This is empirically determined, based on data demonstrating that over 80% of recurrences occur within a 2cm margin of the contrast-enhanced lesion on computed tomography (CT) or magnetic resonance imaging (MRI) [1][2][3][4]. Thus far, this evidence is just indirect and inadequate. Direct evidence for CTV delineation should be provided by the in ltration margin of the tumor. However, assessing the microscopic extension (ME) in HGG is challenging.
Pathology, as the gold standard of diagnosis, can precisely evaluate the ME of tumors. Unfortunately, it is di cult to obtain an adequate surgical margin for HGG, since the tumor is generally removed piecemeal under microscopy. Therefore, few previous studies have revealed the extent of the peripheral in ltration margin of glioma cells (GCs). Through autopsy, Burger et al. [5] found that GCs with a high migratory capability could in ltrate beyond 2cm of the contrast-enhanced rim on CT. This two-dimensional study is limited to one brain histological section per case and lacks detailed data for CTV delineation. Two biopsied studies [6,7] further showed that GCs could deeply in ltrate the contralateral hemisphere, which revealed that HGG had a potential tendency to invade further. However, this limited information does not provide precise evidence for target delineation.
With the development of genomics, O 6 -methylguanine-DNA-methyltransferase (MGMT) promotes methylation, isocitrate dehydrogenase (IDH) mutation and the co-deletion of chromosome arms 1p and 19q (1p/19q) have been proven to be strongly associated with the clinical behavior, response to therapy and outcome of HGG [8][9][10]. Unfortunately, to the best of our knowledge, the relationship between the ME and these molecular alterations has not yet been elucidated. Therefore, the purpose of the present study was to identify the spatial ME of HGG according to consecutive macropathology, analyze its association with malignant factors including grade, tumor volume (V tumor ), location, peritumoral brain edema (PTBE) and molecular markers, and create a model, that could provide evidence for more precisely determining the ME and, hence, the individual CTV to be applied in RT.

Patient selection
This study involved 30 HGG patients who underwent tumor resection at Shandong Cancer Hospital or the First A liated Hospital of Shandong First Medical University between June 2017 and July 2019. The inclusion criteria were as follows: (1) age ≥ 18years; (2) preoperative Karnofsky Performance Status (KPS) ≥ 80; (3) tumor located in nonfunctional area and successful supra total resection; (4) tumor removal achieved with resection margins that included the neighboring normal tissue (between 2cm and 3cm away from the tumor border); and (5) no tumor progression observed during follow-up (at least ≥ 6months). The exclusion criteria included: (1) a medical history of brain chemoradiotherapy (CRT) and (2) multicentric or multifocal cerebral lesions. All tumors were graded according to the 2016 World Health Organization (WHO) classi cation [11]. This study was approved by the institutional review board. All patients provided written informed consent to participate in the trial.
Preoperative MRI was evaluated by a senior radiologist and the data (T2-FLAIR and T1-weighted sequence) were used to determine the PTBE volume (V PTBE ), V tumor and tumor localization. Based on the spatial relationship between the tumor and the subventricular zone (SVZ) and cortex, the tumor location was classi ed as follows: Type I, tumor contacting SVZ; Type II, tumor involving cortex; Type III, tumor neither contacting SVZ nor in ltrating cortex ( Figure 1).

Surgical specimen processing
After resection, the plane of the OML was marked on the specimen. Subsequently, the surgical specimen was oriented according to the in vivo geometry, marked with different colors to indicate the original orientation of the specimen in the brain and xed in 10% formalin (≥24h). The dimensions of the tumor samples, both before and after xation were documented to determine the reduction in size due to xation (Supplementary Table S1). Then, the plane of the OML was perpendicular to the table; the specimen was cut using a tissue slicer from the cranial to the caudal side in approximately 3mm thick slices, which ensured that each specimen slice could match the MRI slice. The slices were contiguous, and their individual thickness was measured with a ruler. Finally, whole-mount para n sections were made and cut into 5µm sections per slice, which were stained with hematoxylin and eosin (H&E) (Figures 2a-e). In addition, each patient underwent molecular testing, and the methods used for analyzing the methylation status of the MGMT promoter and determining the mutational status of IDH by DNA pyrosequencing have been described previously [12,13]. Deletions of chromosomes 1p/19q were evaluated by uorescence in situ hybridization analysis in tumor tissue sections [12,13].

ME analysis and measurement
The tumor-containing area and the PTBE area in the H&E sections were microscopically outlined, scanned and recorded by TissueFAXS PLUS (TissueGnostics, Austria). Subsequently, the scan images were imported into Photoshop (Adobe Systems, USA) to identify the microscopic evidence by two experienced pathologists who were blinded to the clinical data. Invasive GCs were identi ed by means of their nuclear atypia and heteropyknotic staining [14] (Figure 2f). To measure the spatial distribution of invasive GCs, pathologic slices were used to generate three-dimensional (3D) graphics through 3D-DOCTOR (Able Software Corp, USA). First, the contours of individual H&E sections were digitized and recorded to generate a 3D surface of the reconstructed specimen. Second, the 3D specimens were correct for retraction through scaling parameters (Supplementary Table S1). Then, the corrected 3D image was registered to preoperative T1-weighted MRI using the outline of tumor to perform point-based registration. After the above steps, the ME distance and direction of the GCs relative to the primary tumor bulk were established. In the in-slice direction, the nearest (Euclidean) distance [15] from the edge of the tumor to the microscopic GCs surrounding brain tissue was measured by Photoshop ( Figure 2g). In the throughslice direction, the number of slices from the invasive GCs to the lesion border was counted and multiplied by the slice thickness (×3mm). The ME of each slice is de ned as the maximal distance of the ME. The ME of each patient was de ned as the maximum ME across different slices (Supplementary Table S2).

Predictive model development and validation
Based on the multivariate linear regression (MVLR) analysis, we constructed a ME predictive model. To validate the feasibility and safety of this model, we prospectively collected another 30 HGG patients who underwent surgery at Shandong Cancer Hospital between September 2019 and January 2020. The following inclusion criteria were applied: (1) adult patients; (2) preoperative KPS ≥ 70; and (3) no medical condition that could interfere with oral medication intake. Patients who had a history of a previous cranial surgery, CRT, or contraindication for MRI were excluded. Patients were randomly split into two arms in a 1:1 ratio to receive guideline-based CRT (Group NCCN ) or model-based CRT (Group Model ).
All patients started treatment within 2 weeks after surgery and were treated with CRT, which delivered 60Gy in 30 fractions with continuous daily temozolomide (TMZ) (75mg/m 2 /d), followed by 6 cycles of adjuvant TMZ (150-200mg/m 2 for 5/28days). Radiation treatment planning was performed with the Varian Eclipse Treatment Planning System. Target volumes were delineated according to co-registration of postoperative CT and MRI obtained with each patient in the treatment position. The management of Group NCCN was designed according to the NCCN guidelines, Version 2.2019. GTV was de ned by T1weighted abnormality on the MRI, which consisted of all postoperative-enhanced MRI and the surgical cavity. The CTV was de ned by GTV plus a margin of 2cm, adjusted to anatomical borders. The CTV was expanded by 3mm to create the respective planning target volume. For Group Model , the CTV was de ned as the GTV plus a margin, which was determined by our model. The rest of the treatment design was the same as that of Group NCCN .
MRI was repeated before concurrent CRT, before the rst cycle of adjuvant TMZ, and thereafter every 2-6weeks. Tumor progression and recurrence were identi ed by both the oncologist and the radiologist. Progression criteria were described by Macdonald et al. [16]. Recurrence patterns were de ned as in-eld if ≥ 80% of the lesion resided within the prescription 95% isodose line (D 95 ) and marginal if 20-80% of the recurrence was inside the D 95 . For other cases, recurrences were de ned as out-eld [17]. The primary endpoint was the overall response rate (ORR) (the proportion of complete response and partial response).
Secondary endpoints included progression-free survival (PFS) (the date from surgery to either rst documented progression or death), and overall survival (OS) (the date from surgery to the date of death or the last follow-up).

Statistical analysis
For all analyses, we used SPSS 22.0 (IBM Armonk, NY, USA), and values for which P < 0.05 were considered statistically signi cant. Categorical variables were expressed as proportions and continuous variables were expressed as mean ± standard deviation. The difference between two groups was assessed with Student's t-test or Chi-Squared test. When comparing more than two variables, we performed one-way analysis of variance. Post-hoc analysis was used to compare pairwise differences. Spearman's rank correlation was performed to evaluate the relationship of the ME with the grade, V tumor , location, V PTBE and molecular marker status. A MVLR model was created from variables with a P < 0.05 on correlation analysis, using stepwise regression. To assess the prediction e ciency of this model, calibration was evaluated using the R-square goodness-of-t test, and discrimination was evaluated using receiver operating characteristic (ROC) curves with the corresponding area under the curve (AUC). The patient survival rates were determined by Kaplan-Meier curves and survival curves were analyzed by the log-rank test.

Patient and tumor characteristics
In total, 652 H&E slides were analyzed in this study. The characteristics of the patients are listed in Table  1

Pathologic ME characteristics
We demonstrated obvious differences in the ME among individuals (Supplementary Table S2). The GCs were heterogeneously distributed through direct invasion, skip metastases, or along neural ber tracts, pia mater and basement membranes of blood vessels. The mean ME (ME mean ) distance was 1.70cm (range, 0.63 to 2.87cm). The ME distance of AA, AO, AOA and GBM were 1.49±0.63cm, 1.16±0.30cm, 1.47±0.02cm and 2.11±0.42cm, respectively. Grade IV gliomas had signi cantly higher ME than grade III tumors (2.11 ± 0.42cm vs. 1.39 ± 0.52cm, P<0.001) (Figure 3a). A signi cant correlation was found between the extent of ME and pathologic grade (P < 0.001). However, there was no statistically signi cant correlation between the extent of ME and V tumor (P = 0.779) ( Table 2).

Relationship between extent of ME and tumor location
The typical ME distributions in the different locations are shown in Figure 1. Invasive GCs from type I were widely distributed in normal brain tissue, the ME distance was 2.00 ± 0.57cm, and in type II, subpial growth became a main pathway for GCs to distant invasion, and the ME distance was 1.97 ± 0.37cm; whereas in type III, in ltration occurred in the border of the primary lesion with an ME value of 1.12 ± 0.30cm. The ME of type I or type II was signi cantly higher than that of type III (both P < 0.001). However, the ME difference between type I and type II did not reach statistical signi cance (P = 0.865) (Figure 3b). Furthermore, a signi cant positive correlation was found between the extent of ME and SVZ contact or cortical involvement (P = 0.036 and 0.044, respectively) ( Table 2).
Relationship between extent of ME and PTBE PTBE in ltration was found in all patients. Meanwhile, we observed that GCs invaded beyond the PTBE area in 40% (12/30) of patients, including 7 with perineural spread and 5 with subpial growth (Figures 1ab). In contrast, the invasive GCs from 60% (18/30) of patients were only contiguous with the lesion (Figure 1c) and showed a much smaller ME range than the PTBE area (24.98 ± 14.80cm 3 vs. 100.75 ± 52.48cm 3 , P = 0.017). Spearman's rank correlation analysis revealed no signi cant relationship between the extent of ME and V PTBE (P = 0.751) ( Table 2).

Predictive model analysis and validation
The MVLR analysis identi ed that grade, SVZ contact and MGMT, IDH and 1p19q status were independent variables predicting ME (all P < 0.05), with grade having the largest β-coe cient (0.513) ( Table 3). A predictive model was created as follows: Y ME = 0.672 + 0.513X Grade + 0.380X SVZ + 0.439X MGMT + 0.320X IDH + 0.333X 1p/19q . The model was evaluated with good performance in terms of calibration, with the R-square value of the goodness-of-t test being 0.780 (Figure 4a). Meanwhile, we used the ME mean value (1.70) as a cutoff to evaluate the discrimination of the model, which proved that the AUC was 0.964 (95% con dence interval [CI]: 0.909-1.000, P < 0.001) (Figure 4b).

Discussion
To our knowledge, this study is the rst to determine the CTV margins of HGG based on consecutive macropathology. Our results showed that GC invasion into the surrounding brain tissue is complex and highly heterogeneous across different types of HGGs, according to grade, location and molecular markers. We built and validated an easy-to-use model to guide individualized target delineation.
To date, there is a lack of radiologic-histopathologic correlation studies upon which a consensus can be made to guide the targeted delineation of HGG. Although MRI is proposed as the rst choice for pretherapeutic and post-therapeutic evaluation of HGG due to its economic cost-effectiveness and high accuracy, its ability to determine the target volume is inconclusive. Our study found that the contrastenhanced area on T1-weighted MRI re ects only the high-density region of GCs in macropathology, that is, the areas of blood-brain barrier disruption, as described in a previous study [18]. These areas are su cient for de ning the GTV of HGG. However, for CTV delineation, the ability of MRI is limited.
Macropathology, as the gold standard of diagnosis, has inherent advantages in evaluating the ME of tumors. In our validation cohort, we found that Group Model , which determined ME by macropathology showed signi cantly higher ORR than Group NCCN , which determined ME by traditional imaging.
Meanwhile, Group Model showed signi cantly longer PFS than Group NCCN . These ndings suggest that model-based RT is effective and feasible, however, a large cohort still needed to further validate our results.
We identi ed and incorporated 5 independent clinical factors into the MVLR model, including grade, SVZ contact and MGMT, IDH and 1p/19q status. In our model, grade contributed the most to predicting ME of HGG. In agreement with the literature [13,19,20], the WHO grade system is consistently identi ed as an important factor for ME. We found that higher grade glioma was associated with stronger aggressiveness. Another signi cant factor in uencing the ME on multivariate analysis was preoperative tumor location. It is noteworthy that the invasive GC distribution was wider in tumors contacting the SVZ, which was consistent with the results of previous retrospective studies. Lim et al. [21] demonstrated that SVZ contact was signi cantly associated with multifocality. In a study by Adeberg et al. [22], glioblastoma that contacted the SVZ showed higher rates of distant progression and multifocal recurrence than noncontacting lesions. This nding may be explained by the recruitment of glioma stemlike cells in the SVZ, resulting in an aggressive glioma subtype [21][22][23]. In contrast, invasive GCs from type III were only contiguous with the lesion. In regard to the study of Adeberg et al. [22], a similar result was found glioblastoma recurrence always occurred in the border of the primary lesion in the tumor, which neither contacted the SVZ nor in ltrated the cortex. Based on these results, the determination of the CTV margin according to different locations has been proposed for the rst time, but more evidence is still needed to inform clinical practice.
Further analysis revealed a relationship between the extent of ME and MGMT, IDH and 1p/19q status. We found that MGMT methylation induced invasion in distant locations compared with unmethylated cells.
Our results con rmed two previous imaging studies [24,25] and showed that methylated glioblastoma patients with MGMT had a greater tendency to develop out-of-eld recurrence than those with unmethylated status. Interestingly, the present study also observed that IDH wild-type or 1p/19q non-codeleted patients showed increased tumor migration and invasion compared with their counterparts, which has not been previously reported. These ndings might help oncologists provide more tailored RT elds to patients with HGG.
It is highly controversial whether PTBE needs to be intentionally included in the CTV in glioma, since the relationship between the distribution of GCs and PTBE has still been unde ned. In the present study, we rst comprehensively revealed both relationships through macropathology and found that ME was not signi cantly associated with PTBE. We observed that the ME range of 60% of patients was much smaller than that of the PTBE area, whereas in ltration outside of the PTBE occurred when GCs spread along the perineural direction or subpial growth. Similar results were con rmed by Kelly et al. [26] through stereotactic biopsy and Yamahara et al. [27] through autopsy. Based on these important ndings, we suggested that RT including the entire PTBE was not necessary. PTBE might merely coexist with in ltrating GCs in what is a spatial coincidence but actually re ects two independent processes; unreasonable RT elds would increase normal tissue toxicity, thereby in uencing the prognosis of patients [28,29].
Our study has several limitations. First, only 3 typical molecular prognostic markers were detected, and more genes such as TERT and H3K27M need to be further analyzed to explore biomarkers predicting invasion. Second, the size of our study population was small; thus, a large cohort is needed to further develop and validate our model. Third, our study analyzed only patients with good performance. An inherent bias is that tumors that are amenable to supra total resection likely have anatomic, clinical, or biological characteristics that differ from the majority of tumors where subtotal resection is performed.
In conclusion, tumor cells were heterogeneously distributed in different gliomas. Pathologic grade, location, MGMT, IDH and 1p/19q status were demonstrated to be important factors contributing to ME. This suggested that the delineation of CTV should be individualized. Using these factors, we rst built an invasive risk score model, which can better provide valuable evidence to predict the ME of glioma, and this may help clinicians determine the CTV of patients. Availability of data and materials:

Abbreviations
The data used in this analysis is from publications available in the public domain.
Ethics approval and consent to participate: This study was approved by the institutional review board. All patients provided written informed consent to participate in the trial.

Consent for publication:
Not applicable.