Clinical and radiographic characteristics of diffuse astrocytic glioma, IDH-wildtype, with molecular features of glioblastoma: a single institution review

Abstract

Purpose

Genetic analyses of gliomas have identified key molecular features that impact treatment paradigms beyond conventional histomorphology. Despite at-times lower grade histopathologic appearances, IDH-wildtype infiltrating gliomas expressing certain molecular markers behave like higher-grade tumors. For IDH-wildtype infiltrating gliomas lacking traditional features of glioblastoma, these markers form the basis for the novel diagnosis of diffuse astrocytic glioma, IDH-wildtype (wt), with molecular features of glioblastoma (GBM), WHO grade-IV (DAG-G). However, given the novelty of this approach to diagnosis, literature detailing the exact clinical, radiographic, and histopathologic findings associated with these tumors remain in development.

Methods

Data for 25 patients matching the DAG-G diagnosis were obtained from our institution’s retrospective database. Information regarding patient demographics, treatment regimens, radiographic imaging, and genetic pathology were analyzed to determine association with clinical outcomes.

Results

The initial radiographic findings, histopathology, and symptomatology of patients with DAG-G were similar to lower-grade astrocytomas (WHO grade 2/3). Overall survival (OS) and progression free survival (PFS) associated with our cohort, however, were similar to that of IDH-wt GBM, indicating a more severe clinical course than expected from other associated features (15.1 and 5.39 months respectively).

Conclusion

Despite multiple features similar to lower-grade gliomas, patients with DAG-G experience clinical courses similar to GBM. Such findings reinforce the need for biopsy and subsequent analysis of molecular features associated with any astrocytoma regardless of presenting characteristics.

Introduction

Describing the genetic underpinnings of gliomas has made a critical impact on the definition of tumor type and patient prognostication [1]. Recent research demonstrates that genetic markers of tumor type relate to overall survival more strongly than previously-employed histopathological definitions [1–3]. Since 2016, the World Health Organization (WHO) has utilized key markers such as IDH mutation status to decipher lineage and grading of glioma [1]. Still, subsequent studies have noted further prognostic differences based upon genetic aberrations within these IDH-wildtype (IDH-wt) and IDH-mutant (IDH-mu) glioma types [2, 4]. Evidence suggests that despite more “benign” histopathological status, lower-grade infiltrating astrocytomas with molecular characteristics of glioblastoma (GBM, IDH-wt) behave similarly to classical, histopathologically-defined GBM [3, 5].

In 2018, the consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy (cIMPACT-NOW) released a third update, establishing the novel diagnosis of diffuse astrocytic glioma, IDH-wildtype, with molecular features of glioblastoma, WHO grade-IV (DAG-G) [2]. This diagnosis addresses the clinical differences observed in histologic grade 2 and 3 IDH-wt diffuse astrocytic gliomas containing at least one of the following criteria: (1) telomerase reverse transcriptase (TERT) promoter mutations; (2) the combination of whole chromosome 7 gain and whole chromosome 10 loss (+ 7/− 10); and/or (3) epidermal growth factor receptor (EGFR) amplification. Tumors with such components demonstrate clinical courses similar to that of a WHO grade 4 diagnosis [5, 6]. Subsequently, the strength of this association led to the suggestion by the cIMPACT-NOW to incorporate these molecular markers into the diagnosis of glioblastoma itself [7]. In recent publications, both the cIMPACT-NOW sixth update, released in 2020, as well as the 2021 WHO Classification of CNS Tumors now utilize the singular diagnosis of “Glioblastoma, IDH-Wildtype” to include tumors previously diagnosed as DAG-G. Notably, authors of both manuscripts admit that this nomenclature, while simplified, leaves room for confusion when considering cases of glioblastoma with no evidence of histologic features traditionally associated with the condition [8, 9]. For this reason, the presenting study will continue to use the DAG-G classification to avoid ambiguity. Given the novelty of this diagnostic approach, additional clinical description of this subset of patients is important. This is because providers remain reluctant to analyze molecular features associated with tumors demonstrating lower-grade histologic and radiographic features.

Here, we present a single-institution, retrospective series of 25 patients diagnosed with DAG-G. We detail related pathologic reports, as well as associated clinical and radiographic findings and draw comparisons to similar features reported in patients diagnosed with GBM and lower grade (2 and 3) IDH-mu astrocytomas.

Methods

Patient data

Individual patient data was obtained from our retrospective and prospectively-collected data base, central nervous system (CNS) tumor outcomes registry at Emory (CTORE) (STUDY00000332) and CLInical Neurosurgery Outcome Investigations Database (CLINOID) (IRB00117860). This data collection and subsequent analysis was approved by our institutional review board (IRB) and was included under the CTORE project (STUDY00000332). Patients who underwent surgical biopsy or resection of a diagnosed WHO grade 2 or 3 astrocytoma were examined for associated next-generation sequencing (NGS) and single nucleotide polymorphism microarray (SNP-m) results indicating a diagnosis of DAG-G. This diagnosis was made by evidence of at least one of the following diagnostic criteria: (1) TERT promoter mutations; (2) (+ 7/− 10) gain/loss, and/or (3) EGFR amplification. At our institution, TERT mutation testing was not available until recently; this testing was hence performed at reference laboratories on cases suspected to be DAG-G but lacking other diagnostic molecular findings.

We collected patient demographics, treatment regimens, radiographic imaging, histopathology results, NGS/SNP-m, and outcome data. Patterns of contrast enhancement on MRI were listed with designations made for “scant” (< 25% of tumor volume) and “moderate” (> 25% of tumor volume) classifications. Progression-free survival (PFS) was calculated as the difference between the date of initial surgery (biopsy or resection) and the date of first imaging showing unmistakable tumor progression as defined by any one of ten neuroradiologists associated with our department and confirmed at a multi-disciplinary neuro-oncology tumor board. Unmistakable progression is determined by a set of radiographic criteria systematically established by our institution’s Brain Imaging Collaboration Suit (BrICS), which is a cloud platform for integrating whole-brain spectroscopic data into a workflow for radiation therapy planning [10]. Overall survival (OS) was calculated as the difference between the date of first surgery and patient’s recorded date of death.

Molecular analysis

Each biopsied specimen was analyzed using the following techniques in order to assess for relevant abnormalities.

1. Immunohistochemistry (IHC) was used to identify IDH1, p53, and ATRX abnormalities. When indicative of a likely mutation, the p53 IHC, a surrogate marker for TP53 mutations, is reported as “positive,” “overexpressed,” or “lost / null” – all of which would suggest a mutation. Otherwise, “negative”, “focally positive”, “rare”, “scant”, and “variable” results are not suggestive of a p53mutation.

2. SNaPshot—a multiplex quantitative PCR-based platform with an estimated 5% limit of detection – was used to assess a variety of point mutations [11]. It particularly detects those mutations commonly found in IDH1/2 mutant tumors.

3. Chromosomal microarray – a whole genome copy number analysis with an estimated 20% limit of detection-was used to detect + 7/−10 and EGFR amplifications [11]. TERT mutations were assessed in cases when neither + 7/−10 nor EGFR amplifications were detected.

4. MGMT promoter methylation was used to detect the methylation status of the MGMT promoter. In this study, Bisulfite modification of tumor DNA and real-time PCR are used to quantify CpG methylation within the MGMT gene promoter with an estimated 5% limit of detection [11]. Percentage of methylated DNA (compared to total DNA) is reported for positive results. Negative results indicate a lack of methylation, while “low level” results indicate a small degree of methylation present. Total methylation is calculated as an average across listed CpG sites. Total methylation of 0–9% is reported as “Unmethylated,” 10–29% as “Low level,” and equal or more than 30% as “Methylated.”

Statistical analysis

Time-to-event analyses of PFS and OS were performed for each of the following biomarkers: loss of CDKN2A, EGFR amplification, TERT promoter mutation, (+ 7/−10) gain/loss, PTEN mutation, and chromosome 9 abnormality. Chromosome 9 was of particular interest given its relation to the CDKN2A/B genes. Kaplan–Meier curves for both PFS and OS were computed and plotted for those with and without the occurrence for each of the biomarkers. Univariable Cox proportional hazards models were fit for PFS and OS with each of the individual biomarker variables separately as predictors. All statistical analyses were performed at an alpha level of 0.05 using R version 4.1.0.

Results

Our initial search revealed 28 patients with an astrocytoma WHO grade 2 or 3 demonstrating molecular criteria for a diagnosis of DAG-G. However, three of these patients were from outside institutions and had no associated medical records that could be obtained. Thus, these individuals were excluded from this study, leaving a total of 25 patients for final analysis. Table 1 summarizes associated demographic and key clinical information for these patients. A complete description of the cohort is found in Table S1.

Table 1.

Overview of cohort data including demographics, presenting symptoms, and initial imaging data

Cohort overview	n = ()
Gender
Male	14 (56%)
Female	11 (44%)
Average Age (years)	64 (range 37–79)
Chief symptom at presentation
Seizure	14 (56%)
Headache	3 (12%)
Aphasia	2 (8%)
AMS	2 (8%)
Incidental	2 (8%)
Ataxia	1 (4%)
Hemiparesis	(4%)
Imaging
Hemisphere
Left	13 (52%)
Right	10 (40%)
Both	2 (8%)
Region
Frontal	6 (24%)
Temporal	9 (36%)
Parietal	2 (8%)
Occipital	1 (4%)
Other (insula, basal ganglia, multiple)	7 (28%)
Contrast enhancement
None	12 (48%)
Scant (< 25% of tumor volume)	5 (20%)
Moderate (> 25% of tumor volume)	4 (16%)
Ring Enhancing	1 (4%)
Heterogenous	2 (8%)
Unknown	1 (4%)

The average age for this population was 64 years, with 14 men and 11 women. Tumors occurred in a variety of locations, with the most common being the temporal and frontal regions. No predilection occurred between hemispheres for tumor incidence. The majority of patients (56%) experienced seizures as the presenting symptom. No patient demonstrated symptoms of hydrocephalus upon presentation, nor did any patient have a history of prior resection for an intracranial neoplasm before diagnostic workup. The overwhelming majority of patients received a stereotactic biopsy as the first intervention (84%), with only two patients progressing to a subsequent craniotomy for tumor resection (Table S1). Imaging and histopathologic findings were characterized for each patient in order to determine severity of disease (Figs. 1 and 2).

Fig. 1.

Initial imaging of patient cohort often displayed characteristics of “lower” grade astrocytomas. Example patient 1: A MRI brain, T2 FLAIR demonstrates a hyperintense lesion within the left superior frontal gyrus that has expanded the gyrus and resulted on minor mass effect on surrounding brain. B MRI brain, T1 post gadolinium contrast from the same patient in (A) reveals no contrast enhancement, suggesting a diagnosis of low-grade glioma. Example patient 2: C MRI brain, T2 FLAIR demonstrates an expansile, T2 hyperintense lesion in the right mesial temporal lobe. D MRI brain, T2 post gadolinium contrast exhibits no contrast enhancement in the corresponding area of expansile, intrinsic brain lesion. Again, these imaging findings suggest a lower grade glioma

Fig. 2.

Histopathology of an Astrocytoma with molecular features of GBM, WHO grade 4. A An infiltrating glioma showing hypercellular brain parenchyma but note the absence of microvascular proliferation and palisading necrosis. B The tumor cells percolating brain show round to oblong nuclei with irregular borders and hyperchromasia with others showing a gemistocytic appearance

There were no 30-day, post-surgical complications in our cohort. Due to the high-grade features on genetic testing, and fitting with diagnosis of DAG-G, every patient in our cohort received the standard of care treatment for GBM, which included concurrent radiation with Temozolomide therapies. Five (20%) of these patients received additional adjuvant chemotherapy for eventual disease progression – three patients were placed on Bevacizumab only, while two additional patients were placed on Bevacizumab and Lomustine together. The average length of follow-up for this cohort was 8.3 months. A median KPS of 60 was recorded at time of last follow up. The median PFS for this cohort was 5.39 months. Finally, as of May 05, 2021, nine (36%) patients had succumbed to their disease, with a median OS of 15.1 months.

Histopathology and genetic analysis for each patient was analyzed and explored in this study (Table S2). Summary of these findings is displayed in Table 2. Attention was placed upon a number of associated molecular markers in addition to those required to make the diagnosis of DAG-G. In this cohort, (+ 7/−10) gain/loss was the most common criteria met for diagnosing DAG-G (64%), while five individuals (20%) presented with a combination of these criteria.

Table 2.

Summary of molecular analysis results. 9P–chromosome 9p

Molecular markers	n = ()
cIMPACT-NOW qualifying markers
EGFR amplification	7 (28%)
TERT promoter mutations	2 (8%)
Chromosome 7 gain/10 loss	11 (44%)
Multiple (EGFR and + 7/−10)	5 (20%)
Other molecular features
Unmethylated MGMT	14 (56%)
Low MGMG methylation	4 (16%)
Positive MGMT methylation	11 (44%)
Loss of CDKN2A/B	19 (76%)
PTEN mutations	2 (8%)
Retention of ATRX	25 (100%)
BRAF mutations	0 (0%)
MYB/MYNL	0 (0%)
Median methylation index	10% (2–30%)
P53 status
Variable	16 (64%)
Rare	1 (4%)
High	1 (4%)
not over expressed	2 (8%)
Focally positive	3 (12%)
Scattered positivity	2 (8%)
Chromosome 9 abnormalities
Homozygous loss of CDKN2A/B	15 (60%)
Hemizygous loss of CDKN2A/B	4 (16%)
None	5 (20%)
Unknown	1 (4%)

In this study, 14 patients (56%) demonstrated a lack of MGMT promoter methylation with another four (16%) expressing “low” methylation. Interestingly, 19 of these patients (76%) demonstrated a deletion of CDKN2A/B. Additionally, it was noted that every patient retained ATRX expression and demonstrated a relatively high average proliferative index (Ki-67 or MIB1) of 10%. Associated expression of P53 was also predominantly (64%) “variable” and lacking overexpression commonly associated with an underlying TP53 mutation. Lastly, only two PTEN mutations (8%) and zero BRAF mutations were present in this cohort.

Nineteen patients (76%) demonstrated some form of alteration to chromosome 9, independent of homozygous or heterozygous loss of the CDKN2A/B genes (Table S2). The most common alteration present in this chromosome was a focal homozygous loss of the CDKN2A/B genes. Interestingly, molecular pathology did not seem to differ by age (Table S3).

Finally, we studied the impact of various molecular markers on both overall survival and progression free survival within our cohort (Tables 3 and 4). Results are also represented with individual Kaplan Meier curves (Figs. S1 and S2). Of note, some markers listed in Table 2 were not analyzed in this manner for various reasons. For instance, the presence of BRAF mutations and retention of ATRX were all respectively marked as “no” and “yes” for every patient included in this study, allowing for no comparisons to be made. Status of P53 expression involved a large number of classifications, with several categories only listing one occurrence. Fitting models in such cases with a single observation of is often unreliable and unstable.

Table 3.

Univariable Cox proportional hazards models of progression-free survival by molecular marker

Molecular marker	HR	95% CI	(p-value)
Loss of CDKN2A/B	1.63	(0.35, 7.48)	0.533
EGFR amplification	0.66	(0.24, 1.80)	0.421
TERT promoter mutation	1.53	(0.33, 7.04)	0.583
+ 7/−10	0.87	(0.29, 2.61)	0.799
PTEN mutation	0.68	(0.15, 3.09)	0.617
Chromosome 9 abnormalities	1.61	(0.35, 7.46)	0.541

Table 4.

Univariable Cox proportional hazards models of overall survival by molecular marker

Molecular marker	HR	95% CI	(p-value)
Loss of CDKN2A/B	0.58	(0.15, 2.35)	0.499
EGFR amplification	0.50	(0.112, 2.06)	0.399
TERT promoter mutation	N/A	N/A	N/A
+ 7/− 10	1.08	(0. 27, 4.36)	0.910
PTEN mutation	1.09	(0.13, 8.97)	0.937
Chromosome 9 abnormalities	0.94	(0.19, 4.68)	0.942

Discussion

The diagnosis of DAG-G represents a novel classification by which neurosurgeons, radiation oncologists, and neurooncologists are addressing growing evidence of clinical variation among astrocytomas dependent upon their molecular features [2]. Literature on this subgroup in clinical practice remains limited. In this study, we present a retrospective case series of 25 patients diagnosed with DAG-G and detail related features including initial presentation, radiographic and histologic features, molecular markers, treatment regimens, and subsequent outcomes via progression free and overall survival. Genetic analysis of gliomas, regardless of presenting symptoms, imaging findings, or histopathological analysis, is critical to discover patients with DAG-G that warrant more aggressive treatment than their lower-grade counterparts.

In this study, the mean age at diagnosis was 64. This is similar to other studies of this population as well as those diagnosed with GBM, but represents an older age of onset when compared to IDH-mu astrocytoma patients [4, 12]. In our cohort, 56% of patients presented with seizure activity, which is congruent with existing literature, except for the study by Lee et al. [4, 13]. In Lee et al., only one of six patients presented with seizures [4]. An explanation for the variance in data is likely owing to low sample size; others have described risk for seizures in glioma relates to location of origin [14–16]. Additionally, in our cohort, > 60% of tumors presented in these epileptogenic-sensitive regions (including the insula).

Rapidly growing tumors often present with symptoms of mass effect (altered consciousness, weakness, headache, nausea/vomiting). Alternatively, lower-grade gliomas are more likely to present with seizure activity [16]. As lower-grade gliomas exhibit a slower growth rate, symptoms of mass effect are less likely to occur, and instead patients are more likely to present with seizures, owing to smoldering glial changes [4, 7, 17–19]. The predilection for presentation with seizures in our cohort is suggestive of an initial diagnosis of a lower-grade glioma. Indeed, the seemingly innocuous initial presentation course in patients with DAG-G is not indicative of overall progression/survival; patients in our cohort with the lowest overall survival (< 8 months) presented similarly to those with longer survival (> 8 months), with seizures and tumors located in the frontal and temporal regions (Table 5). Two of these patients (19 and 25) presented with fulminant disease and succumbed shortly after diagnosis (0.72 and 0.13 months, respectively).

Table 5.

Information on patients with confirmed death and overall survival < 8 months. Complete cohort detailed data may be found in the supplementary data

Patient number	First intervention	Second intervention	Age at 1st surgery	Sex	Presenting symptom	Tumor location	Enhancement pattern	Chemotherapy regimen	KPS at location last follow up	Progression-free survival (months)	Overall survival (month)
9	Needle Bx	N/A	69	M	AMS	L parietal	None	TMZ	60	2.43	7.89
13	Needle Bx	N/A	71	M	Seizures + hallucinations	R temporal	Ring Enhancing	Unknown	Unknown	2.86	2.86
19	Needle Bx	N/A	65	F	Seizures	L temporal	Scant Enhancement	Unknown	50	0.36	0.72
21	Resection	N/A	67	M	Seizures	R parieto-occipital	Moderate Enhancement	TMZ	40	0.36	2.99
25	Needle Bx	N/A	54	F	Headache	Bifrontote mporal + thalamic + Basal ganglia	None	Unknown	Unknown	0.065	0.13

Figures 1 and 2 demonstrate common radiographic and histologic findings present in our patient cohort. Of note, both modalities demonstrate characteristics found in low-grade gliomas with no evidence of gliomatosis cerebri, which was widely present in patients studied by Lee et al. Radiologic studies also display a lack of calcification and little mass effect with minimal edema, as well as a lack of contrast enhancement in most cases (48%) [20]. Similarly, histopathology demonstrates an absence of microvascular proliferation and palisading necrosis along with a number of cells displaying a gemistocytic appearance [20]. Together with presenting symptomatology, these results indicate the difficulty associated with distinguishing DAG-G from other low-grade astrocytomas. It is clear that molecular studies are warranted for any diffuse glioma.

Treatment outcomes associated with this cohort are present in Tables 6 and S1. Patients were considered for maximal safe tumor resection or biopsy, depending on the location/spread of the glioma. Given both the infiltrative nature of the lesions, as well as the tendency for tumors in this cohort to present in more eloquent brain regions, stereotactic biopsy was pursued in the majority of patients. Surgery was offered for discrete lesions in non-eloquent areas. After diagnosis of DAG-G was made, all patients received the gold standard for GBM treatment—radiation and concurrent chemotherapy with Temozolomide (TMZ) [21]. Additional chemotherapeutic agents were added to five of these patients. Specifically, Bevacizumab (Avastin) was added in each case, while two patients were on a regimen consisting of both Bevacizumab and Lomustine. Cohorts studied by Lee et al., as well as a subset of patients examined by Tesileanu et al., and Wijnenga et al., received no adjuvant radiation or chemotherapies [4, 6, 13]. In those studies, some patients presented with advanced disease and preferred hospice care to treatment. The authors also expressed reticence to employ alkylating chemotherapeutics to treat DAG-G, as most present with unmethylated MGMT promoter status (56% in our cohort) [4]. Patients in this study had a median OS of 15.1 months and a median PFS of 5.39 months (Table 6). Accounting for similarities in therapeutic agents, these values are comparable to those of patients diagnosed with GBM WHO grade 4, which demonstrate a median OS of 14.6 months and a median PFS of 7.4 months with radiation and TMZ therapies [22, 23]. Survival and progression was similar to that described by Cimino et al., who examined DAG-G patients diagnosed with TERT promoter mutations alone and found OS for these patients was fitting with IDH-wt GBM patients [5].

Table 6.

Treatment and clinical outcomes related to patient cohort

Clinical outcomes	n = ()
First surgical intervention
Stereotactic biopsy	21 (84%)
Craniotomy for resection	4 (16%)
Complications	0 (0%)
Second surgical intervention
Stereotactic biopsy	0 (0%)
Craniotomy for resection	2 (8%)
Complications	0 (0%)
Treatment
Radiation	25 (100%)
Chemotherapy
−Temozolomide	25 (100%)
+Bevacizumab	3 (12%)
+Lomustine	2 (8%)
Outcomes
KPS	60 (range 40–90)
Average length of follow up (months)	8.3 (range 0.13–17.65)
Median progression free survival (months)	5.39 (range 0.36–14.23)
Documented deaths	9 (25%)
Median overall survival (months)	15.1 (range 0.13–17.65)

Finally, we examined the association between various molecular markers and both PFS and OS. The largest proportion of our cohort (44%) demonstrated a (+ 7/−10) gain/loss. We found no association of clinical outcomes with any of the mutations needed to diagnose DAG-G (Tables 3 and 4; Figs. S1 and S2). In our analysis, no other molecular markers within the diagnosis of DAG-G predicted clinical outcomes. This was particularly surprising when considering MGMT methylation status and the presence of chromosome 9 abnormalities [24]. MGMT promoter methylation is often correlated with both improved PFS and OS in patients diagnosed with GBM and receiving treatment with alkylating agents [24, 25]. Likewise, the deletion of CDKN2A/B has been recognized as a possible prognostic indicator for both IDH-wt GBM and IDH-mu astrocytoma [26, 27]. We believe that multi-institutional studies or meta-analyses with larger populations may be required to determine additional prognosticating markers.

In our cohort, the median mitotic index is 10% (KI-67 Labeling Index or MIB1). Lower grade tumors usually display a mitotic index < 6%, indicating a minor degree of proliferation [20, 28]. High grade tumors harbor a median index of 13.9% [28]. DAG-G tumors appear to be proliferating at a degree similar to that of higher-grade tumors, fitting with their clinical course.

Recommendations and limitations

Our results suggest that while seemingly of low grade in terms of initial presentation, radiographic, and histopathologic analyses, DAG-G tumors exhibit similar OS and PFS to classic GBM. At our institution, we perform molecular analysis of all gliomas, regardless of histopathological analysis. Moreover, although few patients with DAG-G in our cohort underwent surgical resection after initial biopsy, we continue to offer maximal safe resection for all discrete gliomas in non-eloquent brain tissue. Maximal safe resection remains the gold-standard of care for high-grade gliomas. Given the aggressive nature of DAG-G as described here, consideration of maximal resection when feasible should be preferred. It is a limitation of this study that the majority of these patients received biopsy only, but this is inherent to our particular cohort.

This study is limited by the relative rarity of this newly established diagnostic approach and the subsequently small cohorts available to analyze. Relatedly, meaningful statistical power was difficult to obtain in our analyses. Furthermore, as stated by Lee et al., published studies on DAG-G in the literature largely consist of retrospective studies of non-disaggregated patient data comprising multiple histology types and treatment regimens. The authors believe that our study is strengthened by the fact that it presents a cohort of patients meeting criteria for DAG-G and who received a rather standardized treatment comparable to that of histologically-diagnosed GBM.

Conclusions

Patients with DAG-G by definition present as having histologically lower grade tumors. Yet, such patients maintain a clinical course similar to that of GBM, even when treated with concomitant radiation and TMZ. All gliomas warrant molecular pathology testing to discover prognostic markers indicating when aggressive postoperative treatment is needed. Clinicians should be aware of DAG-G when counseling patients prior to tissue diagnosis and developing both surgical and treatment plans.

Data availability

The datasets generated during and/or analyzed during the current study are available in the supplemental data and from the corresponding author on reasonable request.