Molecular Biomarker Testing for the Diagnosis of Diffuse Gliomas: Guideline from the College of American Pathologists in collaboration with the American Association of Neuropathologists, Association for Molecular Pathology, and Society for Neuro-Oncology

Abstract

Context. –

The diagnosis and clinical management of patients with diffuse gliomas have evolved rapidly over the past decade with the emergence of molecular biomarkers that are used to classify, stratify risk and predict treatment response for optimal clinical care.

Objective. –

To develop evidence-based recommendations for informing molecular biomarker testing for pediatric and adult patients with diffuse gliomas and provide guidance for appropriate laboratory test and biomarker selection for optimal diagnosis, risk stratification and prediction.

Design. –

The College of American Pathologists convened an expert panel to perform a systematic review of the literature and develop recommendations. A systematic review of literature was conducted to address the overarching question, “What ancillary tests are needed to classify diffuse gliomas and sufficiently inform the clinical management of patients?” Recommendations were derived from quality of evidence, open comment feedback, and expert panel consensus.

Results. –

Thirteen recommendations and 3 good practice statements were established to guide pathologists and treating physicians on the most appropriate methods and molecular biomarkers to include in laboratory testing to inform clinical management of patients with diffuse gliomas.

Conclusions. –

Evidence-based incorporation of laboratory results from molecular biomarker testing into integrated diagnoses of diffuse gliomas provides reproducible and clinically meaningful information for patient management.

INTRODUCTION

Diffuse gliomas (DG) are primary central nervous system (CNS) neoplasms characterized by the widespread infiltration of individual tumor cells displaying cytologic and histologic features of glial differentiation.¹ These tumors affect patients of all ages and arise throughout the neuro-axis but are most common in older adults and occur most frequently in the cerebral hemispheres. Clinical course and treatment varies based on tumor type and grade. Since the early 1900s, and until the 2016 revised 4^th edition of the World Health Organization (WHO) Classification of Tumours of the Central Nervous System,¹ DG were classified based upon the morphologic features of neoplastic cells, with molecular testing playing an ancillary role.

Over the past decade, numerous investigations have uncovered molecular genetic alterations that can be used to reliably and reproducibly classify DG into clinically meaningful subsets, leading the 2016 WHO 4th Edition update to incorporate diagnostic entities based on the integration of morphologic features with molecular biomarkers.^1-6 More recent advances in our understanding of the pathogenesis and clinical behavior of specific DG subtypes has led to the inclusion of additional molecular biomarkers into clinical practice and the 2021 WHO 5^th Edition relies even more on molecular test results for diagnosis and grading.^7-11 DNA methylome profiling continues to identify numerous tumor types with specific methylation patterns that have characteristic genetic alterations and clinical behavior.¹² The increasing complexity and rapid pace of change in diagnostic criteria, relevant molecular biomarkers, laboratory testing platforms, and clinical practice warrant the development of evidence-based recommendations on biomarker testing for DG.

DESIGN

This evidence-based guideline was developed following the standards endorsed by the National Academy of Medicine.¹³ A detailed description of the methods and the systematic review (including the quality assessment and complete analysis of the evidence) used to create this guideline can be found in the supplemental digital content (SDC) (Copy editors should make the text “supplemental digital content” the hyperlink).

PANEL COMPOSITION

The College of American Pathologists (CAP) in collaboration with the American Association of Neuropathologists (AANP), Association for Molecular Pathology (AMP), and Society for Neuro-Oncology (SNO) convened a multidisciplinary expert panel (EP) consisting of 13 practicing pathologists from a variety of specialties, institution types, and other professional groups, two oncologists who were representatives of the American Society of Clinical Oncology (ASCO), one patient advocate, and a research methodologist consultant to develop this guideline. An advisory panel assisted the expert panel at specific key stages in the development of the guideline. All panel members, except for the methodologist consultant, volunteered their time and were not compensated for their involvement. Detailed information about the panel composition can be found in the SDC.

CONFLICT OF INTEREST POLICY

The collaborators agreed upon a conflict of interest (COI) policy (effective June 2017) and members of the expert panel disclosed all financial interests from three years prior to appointment through the development of the guideline. Individuals were instructed to disclose any relationship that could be interpreted as constituting an actual, potential, or apparent conflict. Complete disclosures of the expert panel members are listed in the Appendix. Disclosures of interest judged by the oversight group to be manageable conflicts are as follows: HC – Consultancies with F. Hoffman-La Roche, Ltd (Basel, Switzerland), Genentech USA, Inc. (South San Francisco, California), Upsher-Smith Laboratories, LLC (Grove, Minnesota), Novocure (St. Helier, Jersey), Insys Therapeutics (Phoenix, Arizona), Mateon Therapeutics (formerly OxiGENE) (Agoura Hills, California), CytRx Corp. (Los Angeles, California), Omniox Inc. (San Carlos, California), AbbVie Inc. (North Chicago, Illinois); EH – Consultancies with E.R. Squibb&Sons, LLC (New Brunswick, New Jersey), Honorarium from Arbor Pharmaceuticals, LLC (Atlanta, Georgia); MvdB – Consultancies with Bristol Myers Squibb (New York, New York), Celldex Therapeutics, Inc. (Hampton, New Jersey), Vaximm AG (Basel, Switzerland), Carthera, Nerviano, Agios, Honorarium from Merk Sharp & Dohme Corp (Kenilworth, New Jersey), Research Grants from AbbVie Inc (North Chicago, Illinois).

The majority of the EP (14 of 17 members) were assessed as having no relevant conflicts of interest. The CAP provided funding for the administration of the project; no industry funds were used in the development of the guideline. All panel members volunteered their time and were not compensated for their involvement, except for the contracted methodologist. See the SDC for complete information about the COI policy.

GUIDELINE OBJECTIVES

The expert panel addressed the overarching question, “What ancillary tests are needed to classify DG in order to sufficiently inform the clinical management of patients?” To answer this, a number of more pointed key questions (KQs) were developed:

KQ 1a: What genetic and molecular alterations should be included for optimal classification of DG?

KQ1b: What are the acceptable techniques/methods for molecular genetic testing of DG? What are the expected turn-around-times for individual assays?

KQ1c: What are the acceptable techniques/methods for assessing whole genome copy number alterations?

KQ 2: What are the core molecular tests/findings that provide sufficient classifying information in the setting of discrete clinicopathologic entities?

KQ 3: What are the acceptable techniques/methods/criteria for determining MGMT promoter methylation status?

OUTCOMES OF INTEREST

The panel established outcomes of interest for both clinical and pathology-based studies/articles. The clinical outcomes of interest included survival rates (overall, 1- and 3- year survival, progression free), recurrence rates, response to treatment, and accuracy of diagnosis. The pathologic outcomes of interest included sensitivity, specificity, positive predictive value, negative predictive value, concordance, turnaround time, reproducibility of the various tests, as well as mutation/alteration/deletion status (percent, presence, frequency, and association with other alterations) for the molecular targets of interest.

The target audience of this review are those within the neuro-oncology community who care for patients with DG, including pathologists, neuroradiologists, neurosurgeons, radiation oncologists, neuro-oncologists, medical oncologists and other members of the patient care team. Recommendations will also be highly relevant to brain tumor investigators, epidemiologists and cancer registrars.

LITERATURE SEARCH AND COLLECTION

Literature search strategies were developed in collaboration with a medical librarian for the concepts of DG, molecular markers or gene alterations, and laboratory test methods. In consultation with the expert panel, the search strategies were created using standardized database terms and keywords. Databases searched included Ovid MEDLINE and Embase.com. Additional searches for unindexed literature were conducted in ClinicalTrials.gov, Cochrane Library, Guidelines International Network, National Guideline Clearinghouse, Trip search engine, University of York Centre for Reviews and Dissemination-PROSPERO, and applicable U.S. and international organizational websites. In addition, EP members were surveyed for any relevant unpublished data at the onset of the project. Initial database searches were completed on November 13, 2017 and refreshed on September 3, 2019 and September 23, 2020. All searches were limited to English language and publication dates of January 1, 2008 to the date of search. Case reports, commentaries, editorials, conference abstracts, and letters were excluded. The Cochrane search filter for humans was applied in Ovid MEDLINE and Embase.com.¹⁴ MEDLINE and conference abstract records were excluded in the Embase searches. A targeted search was performed in Ovid MEDLINE on July 24, 2020 to ensure that emerging evidence about bithalamic glioma, infantile-type hemisphere glioma, and diffuse pediatric-type high grade glioma and relevant mutations or amplifications was included in anticipation of the WHO 2021 update.⁷ The search was limited to publication dates 1/1/2016 – 7/24/2020, English language, and human studies. Case reports, commentaries, editorials, and letters were excluded. The Ovid MEDLINE and Embase search strings and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagrams outlining details of the systematic and targeted review are provided in the SDC as Supplemental Figures 1-3.

INCLUSION AND EXCLUSION CRITERIA

Studies were selected for inclusion in the systematic review of evidence if they met the following criteria:

the study included human patients; the study was a systematic review with or without meta-analysis, a randomized control trial (RCT), or a comparative or non-comparative observational study (prospective or retrospective design); the study included a minimum sample size of at least 30, except for studies evaluating B-Raf proto-oncogene (BRAF) alterations, Fibroblast Growth Factor Receptor (FGFR) alterations, and MYB proto-oncogene (MYB) and MYB-like (MYBL1) alterations, for which the expert panel determined all study sizes should be included based on a lower frequency of these alterations that occur in pediatric brain tumors; the study was published in English; the study addressed one of the key questions; the study included measurable data such as accuracy, sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), time to appropriate testing, repeat procedures, overall survival, recurrence rates, turnaround time, concordance; and the study addressed DG and included at least one of the following genetic and molecular alterations:

1. Isocitrate Dehydrogenase (NADP(+))1 (IDH1) and Isocitrate Dehydrogenase (NADP(+))2 (IDH2) mutations

2. Histone H3 gene mutations

3. BRAF alterations

4. ATRX chromatin remodeler (ATRX) alterations

5. Tumor protein p53 (TP53) alterations

6. 1p/19q co-deletion

7. Chromosome 7 gain

8. Chromosome 10 loss

9. MYB and MYBL1 alterations

10. Telomerase Reverse Transcriptase (TERT) promoter mutations

11. FGFR alterations

12. Epidermal Growth Factor (EGFR) alterations

13. Platelet Derived Growth Factor Receptor Alpha PDGFR Alpha (PDGFRA) alterations

14. C-MET (C-MET) alterations

15. Cyclin Dependent Kinase Inhibitor 2A (CDKN2A) alterations

16. O-6-Methylguanine-Deoxiribose Nucleic Acid Methlytransferase (MGMT) promoter alterations

17. Phosphatase and Tensin Homolog (PTEN) alterations

18. Neurofibromin 1 (NF1) alterations

19. Microsatellite instability (MSI) status

20. MDM2 Proto Oncogene (MDM2) alterations

21. Cyclin Dependent Kinase 4 (CDK4) alteration

Articles were excluded from the systematic review if they were: meeting abstracts; case reports, consensus documents, editorials, commentaries, or letters; cell line or animal model studies; full-text articles not available in English; studies that didn’t address at least one of the key questions with outcomes of interest as agreed upon.

QUALITY ASSESSMENT

A risk of bias assessment was performed for all fully published studies meeting inclusion criteria by the research methodologist. The methodologist assessed key indicators based on study design and methodological rigor and a rating for the quality of evidence was designated. See Supplemental Tables 1-4 for the quality assessment for included studies by study design. See Supplemental Table 5 for the Quality of Evidence definitions. Using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach,¹⁵ an overall GRADE rating was given for each recommendation by outcome (Supplemental Table 6). Refer to the SDC for further details.

ASSESSING THE STRENGTH OF RECOMMENDATIONS

Following the quality of evidence assessment, completion of the GRADE Evidence to Decision framework,¹⁶ and discussion of the definitions and implications of strength of recommendation (Table 1), the expert panel designated the recommendations as either strong or conditional. Refer to the SDC for further details.

Table 1.

Definitions for Strength of Recommendation

Category	Definition	Implication
Strong Recommendation	One for which the guideline panel is confident that the desirable effects of an intervention outweigh its undesirable effects (strong recommendation for an intervention) or that the undesirable effects of an intervention outweigh its desirable effects (strong recommendation against an intervention)	Implies that most or all individuals will be best served by the recommended course of action
Conditional (Weak) Recommendation	One for which the desirable effects probably outweigh the undesirable effects (weak recommendation for an intervention) or undesirable effects probably outweigh the desirable effects (weak recommendation against an intervention) but appreciable uncertainty exists	Implies that not all individuals will be best served by the recommended course of action. There is a need to consider more carefully than usual the individual patient’s circumstances, preferences, and values.

Data derived from Grading of Recommendations, Assessment, Development and Evaluation (GRADE) Working Group materials.¹⁹⁶

GUIDELINE REVISION

This guideline will be reviewed every four years, or earlier in the event of publication of substantive and high-quality evidence that could potentially alter the original guideline recommendations. If necessary, the entire panel will reconvene to discuss potential changes. When appropriate, the panel will recommend revision of the guideline to the CAP and its collaborators for review and approval.

RESULTS

A total of 4821 studies met the search term requirements. Based on review of these abstracts, 703 articles met the inclusion criteria and continued to full-text review. A total of 188 articles were included for data extraction, and 86 articles informed the recommendations. Excluded articles were available as discussion or background references. Additional information about the systematic review is available in the SDC. Note that only the studies listed in the quality summary were used to inform the recommendations. These are studies from our systematic review of the literature. All other references mentioned in the write-up of each recommendation were brought in to provide context or further support. For most recommendations, the evidence base consisted of studies reporting on the molecular alteration frequency of the recommended gene in defined WHO diffuse glioma subtypes. To highlight this evidence as it relates to establishing an accurate diagnosis while maintaining brevity for publication, each recommendation is accompanied by a table of representative studies. These representative studies chosen for inclusion in the tables are those that the EP believed most important in summarizing the evidence for each recommendation. Refer to the diffuse gliomas guideline web page on www.cap.org for the complete set of data tables.

The expert panel convened 17 times (15 times by teleconference and two in-person meetings) to develop the scope, draft recommendations, review and respond to solicited feedback, and assess the quality of evidence that supports the final recommendations. A nominal group technique was used for consensus decision making to encourage unique input with balanced participation among group members. An initial open comment period was posted on the CAP website (www.cap.org) from September 9-30, 2019, during which the draft recommendations were posted for public feedback. To allow for more responses, the CAP re-opened the comment period from October 11-31, 2019. Refer to the SDC for further details, including a list of organizations encouraged to participate. The expert panel approved the final recommendations with a supermajority vote.

An independent review panel, masked to the expert panel and vetted through the COI process, recommended approval by the CAP Council on Scientific Affairs. The manuscript was also approved by AANP, AMP, and SNO. The final recommendations are summarized in Table 2. In addition, a visual representation of the guideline recommendations is provided to help laboratories and clinicians understand the testing involved that ultimately drives WHO-defined categorization. (Figure 1 shows the testing algorithm where IDH1/2 mutations are present. Figure 2 shows the testing algorithm where IDH1/2 mutations are not present.)

Table 2.

Summary of the Recommendations

Recommendation	Strength of Recommendation
1. IDH mutational testing must be performed on all diffuse gliomas (DG).	Strong Recommendation
2. ATRX status should be assessed in all IDH-mutant DG unless they show 1p/19q codeletion.	Strong Recommendation
3. TP53 status should be assessed in all IDH-mutant DG unless they show 1p/9q codeletion.	Conditional Recommendation
4. 1p/19q codeletion must be assessed in IDH-mutant DG unless they show ATRX loss or TP53 mutations.	Strong Recommendation
5. CDKN2A/B homozygous deletion testing should be performed on IDH-mutant astrocytomas.	Conditional Recommendation
6. MGMT promoter methylation testing should be performed on all glioblastoma (GBM), IDH-wild type (WT).	Strong Recommendation
7. For IDH-mutant DG, MGMT promoter methylation testing may not be necessary.	Conditional Recommendation
8. TERT promoter mutation testing may be used to provide further support for the diagnosis of oligodendroglioma and IDH-WT GBM.	Conditional Recommendation
9. For histologic grade 2-3 DG that are IDH-WT, testing should be performed for whole chromosome 7 gain/whole chromosome 10 loss, EGFR amplification, and TERT promoter mutation to establish the molecular diagnosis of GBM, IDH-WT, grade IV.	Strong Recommendation
10. H3 K27M testing must be performed in DG that involve the midline in the appropriate clinical and pathologic setting.	Strong Recommendation
11. H3 G34 testing may be performed in pediatric and young adult patients with IDH-WT DG.	Conditional Recommendation
12. BRAF mutation testing (V600) may be performed in DG that are IDH-WT and H3-WT.	Conditional Recommendation
13. MYB/MYBL1 and FGFR1 testing may be performed in children and young adults with DG that are histologic grade 2-3 and are IDH-WT and H3-WT.	Conditional Recommendation

ATRX, ATRX chromatin remodeler; BRAF, B-Raf proto-oncogene; CDKN2A/B Cyclin Dependent Kinase Inhibitor 2A (CDKN2A)/ Cyclin Dependent Kinase Inhibitor 2B (CDKN2B); EGFR, Epidermal Growth Factor; FGFR1, Fibroblast Growth Factor Receptor 1; H3, histone3; IDH, Isocitrate Dehydrogenase; MGMT, O-6-Methylguanine-Deoxiribose Nucleic Acid Methlytransferase; MYB/MYBL1, MYB proto-oncogene (MYB)/MYB-like; TERT, Telomerase Reverse Transcriptase; TP53, Tumor protein p53

Figure 1. Testing Algorithm for Diffuse Gliomas with IDH1/2 Mutations.

A – refer to Figure 2.

ATRX, ATRX chromatin remodeler; CDKN2A/B Cyclin Dependent Kinase Inhibitor 2A (CDKN2A)/ Cyclin Dependent Kinase Inhibitor 2B (CDKN2B); Codel, codeletion; IDH, Isocitrate Dehydrogenase (NADP(+)); MVP, microvascular proliferation; Mut, mutation; TERTp, Telomerase Reverse Transcriptase promoter; TP53, Tumor protein p53; WHO, World Health Organization.

Blue indicates WHO-defined entities; Green indicates recommended tests; *Some institutions/laboratories may prefer to perform 1p/19q codeletion as the initial step for IDH-mutant gliomas. See recommendations 2-4.**Additional molecular biomarker testing and DNA methylation profiling maybe helpful in establishing a diagnosis for challenging cases.

Figure 2. Testing Algorithm for Diffuse Gliomas Without IDH1/2 Mutations.

A – refer to Figure 1.

ALK, ALK receptor tyrosine kinase; BRAF, B-Raf proto-oncogene; EGFR, epidermal growth factor receptor; FGFR1, Fibroblast Growth Factor Receptor; H3, histone 3 gene mutation; H3K27me3, H3 K27M trimethylation (H3K27me3) IDH, Isocitrate Dehydrogenase (NADP(+)); IHC, immunohistochemistry; Mut, mutation; TERTp, Telomerase Reverse Transcriptase promoter; CNA, copy number alteration; MAPK, Mitogen-activated protein kinase; MET, mesenchymal epithelial transition factor; MGMTp, O-6-Methylguanine-Deoxiribose Nucleic Acid Methlytransferase promoter; MYB, MYB proto-oncogene (MYB); MYBL1, MYB-like; NTRK, Neurotrophic tyrosine receptor kinase; PDGFRA, platelet derived growth factor receptor alpha; MYCN, N-myc proto-oncogene protein; ROS1, ROS proto-oncogene 1; WHO, World Health Organization.

Blue indicates WHO-defined entities; Green indicates recommended tests; Italic indicates good practice statements. **Additional molecular biomarker testing and DNA methylation profiling maybe helpful in establishing a diagnosis for challenging cases.

RECOMMENDATIONS

1. Strong Recommendation. – IDH mutational testing must be performed on all DG.

The quality of evidence to support this recommendation was assessed as moderate.

Quality Summary: The evidence base informing this recommendation comprises 38 studies.^{3-6, 17-50} Four studies were genome sequencing studies,^{5, 6, 21, 22} while the remaining 34 studies were retrospective cohort studies.^{3, 4, 17-20, 23-50} All genome sequencing studies^{5, 6, 21, 22} obtained samples retrospectively and were assessed as intermediate-low quality based on risk of selection bias, while no other forms of bias were identified. The retrospective cohort studies all suffered from risk of selection bias and were assessed as low^{3, 17, 19, 23-26, 28, 31-33, 35-50} and very low quality^{4, 18, 20, 27, 29, 30, 34} based on risk of selection bias in addition to risk in performance, detection, and reporting domains. The aggregate risk of bias across the entire evidence base was serious but quality of evidence was up-graded based on a strong association between IDH mutational testing and DG diagnostic classification. Refer to the SDC for the quality assessment of included genome sequencing studies, retrospective cohort studies, and the GRADE Quality of Evidence Assessment. See Table 3 for a summary of the mutational status of IDH across all diffuse glioma subtypes.

Table 3.

Representative Studies Reporting on IDH Mutation Testing and Status Across Diffuse Glioma Subtypes

Study, Study Design	Number DG Cases	Testing Method	Mutation Frequencya	Study Conclusion
Brat et al,6 2015, GSS	n = 293	Genome sequencing	A: 67.37% (64/95)	Genome-wide data delineated three molecular classes of histologic grade 2/3 DG that were more concordant with IDH, 1p/19q, ATRX and TP53 status than with histologic class and correlated with clinical outcome.
			OA: 85.13% (63/74)
			OD: 88.07% (96/109)
			GBM: NR
Ceccarelli et al,5 2016, GSS	n = 1132	Genome sequencing	A: 69.05% (116/168)	DG classification based on IDH and 1p/19q status was further refined using DNA methylation profiles to identify clinically relevant genetic subsets.
			OA: 86.84% (99/114)
			OD: 89.02% (154/173)
			GBM: 7.51% (34/453)
Chan et al,23 2015, RCS	n = 237	Sequencing	A: 58.33% (98/168)	When combined with IDH and 1p/19q status, TERTp mutation contributed to prognostic subgroups of lower-grade DG.
			OA: 75.00% (36/48)
			OD: 95.24% (20/21)
			GBM: NR
Killela et al,34 2014, RCS	n = 473	Sequencing	A: 78.41% (69/88)	Genetic signatures of DG based on TERTp and IDH status stratifies patients into prognostically distinct cohorts
			OA: 86.21% (50/58)
			OD: 96.55% (84/87)
			GBM: 10.00% (24/240)
Mellai et al,39 2011, RCS	n = 287	Sequencing, IHC	A: 38.89% (7/18)	IDH mutation was specific for DG, was correlated with MGMTp methylation and was anti-correlated with EGFR amp.
			OA: 100% (4/4)
			OD: 53.22% (33/62)
			GBM: 10.22% (19/186)
Sanson et al,47 2009, RCS	n = 404	Sequencing	A: 63.33% (19/30)	IDH1 mutation was closely linked to prognosis, MGMTp status and molecular profile of DG, grades 2-4.
			OA: 68.18% (60/88)
			OD: 63.11% (65/103)
			GBM: NR
Labussiere et al,35 2010, RCS	n = 764	Sequencing	A: NR	IDH1/IDH2 mutation was tightly linked to the finding of t(1;19) translocation in DG, but not with other focal 1p and 19q losses.
			OA: NR
			OD: NR
			GBM: NR
Lee et al,36 2017, RCS	n = 168	Sequencing, MSP	A: 60.53% (23/38)	TERTp mutation strongly correlated with poor survival outcome in patients with IDH-WT GBM
			OA: NR
			OD: 100.0% (65/65)
			GBM: 20.00% (13/65)
Eckel-Passow et al,4 2015, RCS	n = 615	Sequencing	A: NR	DG can be classified into five clinically relevant groups on the basis of IDH, TERT, and 1p/19q
			OA: NR
			OD: NR
			GBM: NR
Korshunov et al,3 2015, RCS	n = 202	450K BeadChip array	A: NR	Pediatric GBM showed low frequency of IDH1 mutation and high frequency of histone H3 mutations, leading to refined prognostic model.
			OA: NR
			OD: NR
			GBM: 4.95% (10/202)
Ebrahimi et al,26 2016, RCS	n = 1064	IHC	A: 54.93% (156/284)	ATRX is a potential marker for predicting IDH/H3F3A mutations and substratification of DG into prognostic groups
			OA: 92.96% (66/71)
			OD: 97.59% (81/83)
			GBM: 7.41% (27/364)
Boots-Sprenger et al,20 2013, RCS	n = 561	MLPA	A: 72.37% (55/76)	IDH, 1p/19q and MGMTp status were correlated with patient outcome in DG, yet some exceptions were noted.
			OA: NR
			OD: NR
			GBM: 15.92% (36/226)
Dubbink et al,25 2016, RCS	n = 133	NGS	A: 100.0% (20/20)	DG can be subclassified into prognostic groups based on the molecular status of IDH, 1p/19q, TERTp, 7+/10q−, and H3F3A.
			OA: 0.0% (0/21)
			OD: 97.96% (48/49)
			GBM: 0.0% (0/55)

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

Abbreviations: 7+/10q−, chromosome 7 gain/chromosome 10 q loss; amp, amplification; A, astrocytoma histology; ATRX, ATRX chromatin remodeler; BRAF, B-Raf proto-oncogene; DG, diffuse glioma; EGFR, Epidermal Growth Factor; GBM, glioblastoma; GSS, genome sequencing study; mut, mutation; H3, histone3; IDH1, Isocitrate Dehydrogenase (NADP(+))1; IDH2, Isocitrate Dehydrogenase (NADP(+))2; IHC, immunohistochemistry; MGMT, O-6-Methylguanine-Deoxiribose Nucleic Acid Methlytransferase; MGMTp, MGMT promoter; MLPA, multiplex ligation-dependent probe amplification; MSP, methylation specific polymerase chain reaction; NGS, next generation sequencing; NR, not reported; OA, oligoastrocytoma histology; OD, oligodendroglioma histology; RCS, retrospective cohort study, t(1;19), translocation (1;19); TERT, Telomerase Reverse Transcriptase; TERTp, TERT promoter; TP53, Tumor protein p53; WT, wild-type.

The identification of an IDH mutation within a DG is required for the diagnosis of specific neoplastic types recognized by the World Health Organization (WHO), including astrocytoma, IDH-mutant and oligodendroglioma, IDH-mutant, 1p/19q codeleted.^{1, 7} More than 70% of histologic grades 2 and 3 infiltrating astrocytomas in adults are IDH-mutated. By definition, all oligodendrogliomas IDH-mutant, 1p/19q codeleted harbor IDH mutations.^{4-6, 34} IDH-mutant DG are distinct diseases that have specific genetic profiles, clinical courses and therapeutic options that differ from other forms of DG, such as IDH-wild type (WT) and histone H3-mutant DG.^{3-6, 18, 20, 23, 25, 35, 36, 47} Therefore, testing for IDH mutations must be performed on DG in order to diagnose these tumor types for proper clinical care. Testing may not be necessary if a definitive diagnosis of an IDH-mutant glioma or another diagnosis that is mutually exclusive with IDH-mutant glioma has been previously established. IDH-mutant gliomas are uncommon in pediatric patients and testing may not be necessary in younger children.⁵¹ Testing for IDH mutations also may not be necessary if genetic alterations that are mutually exclusive with IDH mutations have been identified in a DG, such as histone H3 mutations. It should be noted that EGFR amplifications, gain of chromosome 7/loss of chromosome 10, and TERT promoter mutations have been documented in IDH-mutant astrocytomas, albeit at much lower frequency than GBM, IDH-WT, and their identification in a DG should not automatically rule out IDH testing.^{3, 26, 39, 47, 52} Other clinical or pathologic settings may be encountered in which the diagnosis of an IDH-mutant DG can be formally excluded without testing for the mutation.

IDH1 and IDH2 mutations result in a substitution for a key arginine at codons R132 and R172, respectively.⁵³ The most frequent IDH1 mutation is R132H, which accounts for 89-93% of all IDH1 and IDH2 mutations.^{5, 6, 27, 35, 47, 53} A highly sensitive and specific monoclonal antibody that recognizes the IDH1-R132H mutant protein is widely used in diagnostic practice. Immunohistochemistry (IHC) for IDH1-R132H is a cost-effective and reliable first line test for IDH1 mutation in supratentorial DG.^{39, 50, 54-57} IDH1 R132H mutations are followed in frequency by R132C, R132S, R132G and R132L. IDH2 mutations represent approximately 3% of all IDH mutations and are more frequent in oligodendrogliomas than IDH-mutant astrocytomas. R172K is the most frequent, followed by R172M, and R172W. Testing for these non-IDH1-R132H mutations is accomplished by DNA sequence analysis and is necessary when IHC for IDH1-R132H is negative in appropriate settings, with specific exceptions (see above). Recent studies of primary infratentorial IDH-mutant astrocytomas have shown that they have a different spectrum of IDH-mutations, with about 80% harboring non-IDH R132H mutations,⁵⁸ suggesting that IDH sequencing may be required more frequently in this clinical setting.

Sequence analysis for IDH mutations can be performed by pyrosequencing, Sanger sequencing, polymerase chain reaction (PCR), or next generation sequencing (NGS).^{27, 34, 59-61} Since patients older than 55-years only rarely develop de novo glioblastomas (GBMs) that are IDH-mutant (4-7%) and the IDH1-R132H antibody recognizes the large majority of IDH mutations, the WHO has recommended that testing for non-IDH1-R132H mutations by sequence analysis may not be necessary in patients older than 55-years whose tumors are negative for IDH1-R132H by IHC.^{1, 26, 27, 32, 43, 55} However, non-R132H mutations are sufficiently common in DG of all grades under age 55, such that testing for these mutations is indicated when IHC does not detect the R132H mutation. As above, IDH testing may also be unnecessary if other mutually exclusive genetic alterations are identified.

The finding of an IDH mutation in the setting of glial proliferation strongly supports the diagnosis of a DG since this event does not occur in non-neoplastic diseases and is only rarely, if ever, found in CNS neoplasms other than DG.^{27, 46, 62-64} IDH mutations are almost always stable and persist through the course of disease, which can be exploited to evaluate residual/progressive tumor^{33, 65, 66}; however, loss of the IDH-mutated allele has been documented in some cases at the time of recurrence and may be associated with greater cell proliferation or high grade behavior.^{67, 68}

Public comment response to Recommendation 1: There were 92 respondents, of whom 91.31% (n=84) agreed or agreed with modifications, 6.52 % (n=6) disagreed, and 2.17% (n=2) were neutral. There were 31 written comments, including a number that suggested that the phrase “appropriate clinical and pathologic setting” needed to be further explained. There were also suggestions that the statement should recommend that IDH testing “should” be performed rather than “must” be performed. Others indicated that the methods of testing (eg, IHC versus sequencing) should be described. These comments were taken into consideration. While the recommendation remained the same, the comments were addressed in the text above.

2. Strong Recommendation. – ATRX status should be assessed in all IDH-mutant DG unless they show 1p/19q codeletion.

3. Conditional Recommendation. – TP53 status should be assessed in all IDH-mutant DG unless they show 1p/19q codeletion.

4. Strong Recommendation. – 1p/19q codeletion must be assessed in IDH-mutant DG unless they show ATRX loss or TP53 mutation.

Recommendations 2-4 are tightly linked to one another and should be considered together. Once the initial testing of a DG has revealed that the tumor is IDH-mutant, further testing for ATRX, TP53 and 1p/19q support the diagnosis of IDH-mutant astrocytoma or oligodendroglioma, IDH-mutant, 1p/19q codeleted. Testing algorithms and workflows differ across practices, institutions and countries. Some perform IHC for ATRX and p53 first, with follow-up studies for 1p/19q only if these tests are negative. Others perform 1p/19q testing as the initial step, followed by ATRX and/or p53 testing if the test is negative. Multigene panels represent another approach that provides test results simultaneously; proponents assert that it provides information on other biomarkers relevant to classification and grade.⁶⁹ Since positive test results from ATRX/p53 and 1p/19q testing are nearly mutually exclusive, it is not necessary to perform additional testing if the initial test results are sufficient to establish a diagnosis. Recommendations 2-4 are written in a manner to reflect these considerations.

The quality of evidence to support Recommendations 2 and 4 is moderate, the quality of evidence to support Recommendation 3 was assessed as low.

Quality Summary: Recommendation 2 was informed by 12 retrospective cohort studies.^{24-26, 49, 50, 70-76} Eight of these studies were assessed as low quality,^{24-26, 49, 50, 70, 74, 75} and four were assessed as very low quality.^{71-73, 76} All included studies were limited by a critical risk of selection bias, plus individual studies were further limited by risk of bias in performance,^{26, 72, 73} detection,^{24, 26, 71-76} and reporting,^{50, 71-74, 76} domains. Although the aggregate risk of bias across the evidence base was very serious, the evidence was upgraded based on a strong association between ATRX assessment and DG WHO classification. Recommendation 3 focuses on the need for TP53 assessment and was informed by two genome sequencing studies,^{6, 22} two prospective cohort studies,^{77, 78} and 15 retrospective cohort studies.^{19, 24, 25, 29, 42, 49, 71, 75, 79-85} The retrospective cohort studies were assessed as low^{19, 24, 25, 42, 49, 75, 79-81, 83-85} and very low quality^{29, 71, 82} based on risk of bias in selection,^{19, 24, 25, 29, 42, 49, 71, 75, 79-85} performance,^{19, 29, 79} detection,^{24, 29, 42, 71, 75, 79-82, 84, 85} and reporting^{29, 42, 71, 80, 82-85} domains. The aggregate risk of bias for the evidence base was very serious and evidence was not further downgraded or upgraded based on any domain. The evidence base supporting Recommendation 4 comprises one genome sequencing study⁶ and 11 retrospective cohort studies.^{4, 20, 42, 49, 50, 63, 74-76, 82, 86} All retrospective cohort studies carry a critical risk of selection bias, plus individual studies were further limited by risk of bias in performance,⁴ detection,^{4, 20, 42, 63, 74-76, 86} and reporting^{4, 20, 42, 50, 63, 74, 76, 82, 86} domains. Although the aggregate risk of bias across the evidence base was very serious, the evidence was upgraded based on a strong association between 1p/19q codeletion status and DG WHO classification. Refer to the SDC for the quality assessment of included genome sequencing studies, retrospective cohort studies, and the GRADE Quality of Evidence Assessment. See Table 4 for a summary of the mutational status of ATRX, TP53, and 10/19q across all diffuse glioma subtypes.

Table 4.

Representative Studies Reporting on ATRX, TP53, and 1p/19q Status Across Diffuse Glioma Subtypes

Study, Study Design	Number DG Cases	Marker	Testing Method	Mutation Frequencya	Study Conclusion
Brat et al,6 2015, GSS	n=293	1p19q, IDH, TP53, ATRX	Genome sequencing	A: IDH + 1p19q, 2.11% (2/95)	Genomewide data delineated three molecular classes of histologic grade 2/3 DG that were more concordant with IDH, 1p/19q, ATRX and TP53 status than with histologic class and correlated with clinical outcome.
				OA: IDH + 1p19q, 17.57% (13/74)
				OD: IDH + 1p19q, 63.30% (69/109)
				GBM: NR
McClendon et al,22 2008, GSS	n=91	TP53	Genome sequencing	A: NR	Highlighted core molecular pathways consistently altered in GBM, including p53, RTK and RB networks.
				OA: NR
				OD: NR
				GBM: NR
Cai et al,70 2014, RCS	n=169	ATRX	Sequencing	A: ATRX, 61.76% (42/68)	ATRX expression is tightly correlated with IDH1/2 mutation and can be used to define prognostic DG subgroups.
				OA: NR
				OD: NR
				GBM: ATRX, 41.58% (42/101)
Rajmohan et al,74 2016, RCS	n=91	ATRX, IDH, 1p19q	Sequencing, IHC, FISH	A: ATRX, 83.33% (15/18) 1p19q, 5.56% (1/18)	Histologic grade 3 DG can be stratified into prognostic groups based on IDH, 1p/19q and ATRX status
				OA: ATRX, 24.24% (8/33) 1p19q, 57.58% (19/33)
				OD: ATRX, 10.00% (4/40) 1p19q, 70.00% (28/40)
				GBM: NR
Mukasa et al,42 2012, RCS	n=250	TP53, 1p19q, IDH	Sequencing	A: 1p19q, 6.90% (4/58) TP53, 39.66% (23/58)	IDH mutation in DG was tightly correlated with 1p/19q codeletion and the combination was associated with prolonged survival.
				OA: 1p19q, 8.33% (1/12) TP53, 41.67% (5/12)
				OD: 1p19q, 72.50% (29/40) TP53, 7.50% (3/40)
				GBM: 1p19q, 0.82% (1/122) TP53, 20.49% (25/122)
Sahm et al,82 2014, RCS	n=43	IDH, ATRX, TP53, 1p19q	Sequencing, IHC	A: NR	Histologically defined oligoastrocytomas can be reclassified as astrocytoma or oligodendroglioma based on IDH, TP53, ATRX, and 1p19q
				OA: NR
				OD: NR
				GBM: NR
Eckel-Passow et al,4 2015, RCS	n=615	IDH, TERT, 1p19q	Sequencing	A: NR	DG can be classified into five clinically relevant groups on the basis of IDH, TERTp, and 1p/19q
				OA: NR
				OD: NR
				GBM: NR
Shao et al,75 2016, RCS	n=135	ATRX, TP53, 1p/19q, IDH	IHC	A: ATRX, 32.79% (20/61) TP53, 50.88% (29/57) 1p19q, 16.00% (4/25)	DG with IDH mutations and ATRX loss also had p53 overexpression and MGMTp methylation, and were mutually exclusive with 1p/19 codeletion.
				OA: NR
				OD: NR
				GBM: ATRX, 20.00% (2/10) TP53, 54.54% (6/11) 1p19q, 0.00% (0/4)
Cryan et al,71 2014, RCS	n=108	ATRX, TP53	IHC, NGS	A: ATRX, 53.57% (15/28) TP53, 71.41% (20/28)	Oligodendroglioma and astrocytoma can be distinguished by their IDH, 1p/19q, ATRX and TP53 profile. Histologic oligoastrocytomas harbored TP53 and ATRX mutations at frequencies similar to astrocytomas.
				OA: ATRX, 60.00% (9/15) TP53, 80.00% (12/15)
				OD: ATRX, 3.08% (2/65) TP53, 6.15% (4/65)
				GBM: NR
Ebrahimi et al,26 2016, RCS	n=1064	ATRX, IDH	IHC	A: ATRX, 47.89% (136/284)	ATRX is a potential marker for predicting IDH/H3F3A mutations and substratification of DG into prognostic groups
				OA: ATRX, 32.39% (23/71)
				OD: ATRX, 1.20% (1/83)
				GBM: ATRX, 7.42% (27/364)
Ikemura et al,73 2016, RCS	n=193	ATRX, IDH, TP53, 1p19q	IHC	A: ATRX, 54.45% (24/44) TP53, 40.63% (13/32) 1p19q, 10.00% (3/30)	In grade 2/3 DG, ATRX loss was strongly correlated with IDH mutation and p53 overexpression, while mutually exclusive of 1p/19q co-deletion.
				OA: ATRX, 30.77% (4/13) TP53, 22.22% (2/9) 1p19q, 63.64% (7/11)
				OD: ATRX, 0.00% (0/18) TP53, 0.00% (0/15) 1p19q, 94.44% (17/18)
				GBM: ATRX, 12.71% (15/118) TP53, 21.95% (18/82) 1p19q, 3.53% (3/85)
Wiestler et al,76 2013, RCS	n=133	ATRX, 1p19q	IHC	A: ATRX, 44.62% (29/65)	Among patients with anaplastic DG, ATRX loss strongly correlated with IDH mutations, anti-correlated with1p/19q co-deletion, and provided prognostic value.
				OA: ATRX, 27.08% (13/48)
				OD: ATRX, 20.00% (2/10)
				GBM: NR
Hewer et al,72 2016, RCS	n=54	ATRX, IDH, 1p19q	IHC	A: NR	Histologic oligoastrocytomas that had IDH mutations showed either ATRX loss or 1p/19q LOH, but not both alterations.
				OA: NR
				OD: NR
				GBM: NR
Bienkowski et al,49 2018, RCS	n=165	ATRX, IDH, TP53, 1p19q	IHC, PCR. MLPA	A: ATRX, 30.77% (24/78) TP53, 47.50% (38/80) 1p19q, 1.47% (1/68)	For assessing IDH, 1p/19q and TERTp status in DG, a combination of IHC, direct sequencing, and MLPA is a practical approach.
				OA: ATRX, 47.17% (25/53) TP53, 56.60% (30/53) 1p19q, 32.08% (17/53)
				OD: ATRX, 9.68% (3/31) TP53, 29.03% (9/31) 1p19q, 62.07% (18/29)
				GBM: NR
Alentorn et al,19 2014, RCS	n=126	TP53	IHC	A: TP53, 35.00% (7/20)	Among IDH-mutant DG, p53 overexpression is mutually exclusive with 1p19q co-deletion.
				OA: TP53, 38.78% (19/49)
				OD: TP53, 15.69% (8/51)
				GBM: NR
Wang et al,83 2016, RCS	n=670	TP53, IDH	IHC	A: TP53, 64.62% (42/65)	IDH mutation was strongly correlated with high p53 expression among histologically defined astrocytic neoplasms.
				OA: TP53, 62.04% (85/137)
				OD: TP53, 13.75% (11/80)
				GBM: TP53, 1.90% (57/300)
Rageswarie et al,50 2018, RCS	n=449	IDH, ATRX, 1p19q	IHC, FISH	A: NR	In a resource-limited set-up, histology with IHC for IDH1-R132H and ATRX form the baseline testing to derive DG subgroups.
				OA: NR
				OD: NR
				GBM: NR
Dubbink et al,25 2016, RCS	n=133	ATRX, 1p19q, IDH	NGS	A, 1p19q intact: ATRX, 65.00% (13/20) TP53, 95.00% (19/20)	DG can be subclassified into prognostic groups based on the molecular status of IDH, 1p/19q, TERTp, 7+/10q−, and H3F3A.
				OA: NR
				OD, 1p19q co-del: ATRX, 2.04% (1/49) TP53, 8.16% (4/49)
				GBM: ATRX, 0.00% (0/55) TP53, 20.00% (11/55)
Chaurasia et al,24 2016	n=163	TP53	NR	A: NR	IHC for ATRX, IDH1, and p53 can be used to stratify GBM patients as individual markers and in combination.
				OA: NR
				OD: NR
				GBM: ATRX, 15.34% (25/163) p53, 49.08% (80/163)

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content SDC for all outcomes reported by included studies

Abbreviations: 7+/10q−, chromosome 7 gain/chromosome 10 q loss; A, astrocytoma histology; ATRX, ATRX chromatin remodeler; co-del, co-deletion; DG, diffuse glioma; FISH, Fluorescence in situ hybridization; GBM, glioblastoma; GSS, genome sequencing study; H3, histone3; IDH1, Isocitrate Dehydrogenase (NADP(+))1; IDH2, Isocitrate Dehydrogenase (NADP(+))2; IHC, immunohistochemistry; LOH, loss of heterozygosity; mut, mutation; MGMTp, O-6-Methylguanine-Deoxiribose Nucleic Acid Methlytransferase promoter; MLPA, multiplex ligation-dependent probe amplification; NGS, next generation sequencing; NR, not reported; OA, oligoastrocytoma histology; OD, oligodendroglioma histology; PCR, polymerase chain reaction; RB, radial basis; RCS, retrospective cohort study; RTK, receptor tyrosine kinase; TERTp, Telomerase Reverse Transcriptase promoter; TP53, Tumor protein p53

Recommendation 2

As with all cancers, DG must employ a mechanism to prevent telomeres from shortening in order to escape cellular senescence. In the case of IDH-mutant astrocytomas (WHO grades 2-4), this is primarily accomplished by alternative lengthening of telomeres (ALT) via inactivation of the ATRX gene.⁸⁷ Most oligodendrogliomas, IDH-mutant, 1p/19q-codeleted and IDH-WT GBMs accomplish this via overexpression of telomerase resulting from activating TERT promoter mutations. Due to these associations, loss of nuclear ATRX protein expression in tumor cells of an IDH-mutant DG as determined by IHC serves as a relatively (albeit not completely) sensitive and specific surrogate marker for astrocytic lineage.^{24-26, 49, 50, 70, 71, 73-75} Since strong and diffuse nuclear p53 immunoreactivity is also commonly encountered in IDH-mutant astrocytomas, the ATRX immunostain is often run simultaneously with the IDH1 R132H (recommendation 1) and p53 stains (discussed in recommendation 3). Less common astrocytoma subtypes, which are more common in children, may also show loss of ATRX expression, including diffuse midline gliomas with H3 K27M mutation (~15% of cases), diffuse hemispheric gliomas with H3 G34 mutation (nearly all cases), and anaplastic astrocytoma with piloid features.^{26, 88-93} Consistent with their associations with glial lineage, there is a strong inverse relationship among IDH-mutant DG between ATRX loss and the presence of 1p/19q codeletion (ie, molecularly defined oligodendrogliomas).^{49, 50, 72, 76, 94} Only rare examples of “dual genotype” IDH-mutant DG have been described in which ATRX loss, TP53 mutation and 1p/19q codeletion were identified in the same neoplasm.^95-97 For this reason, it has been recommended that 1p/19q testing need not be pursued in IDH-mutant gliomas with immunohistochemically identified ATRX loss or p53 overexpression.⁹⁸ By the same token, ATRX testing is not considered necessary for IDH-mutant DG with established codeletion of chromosome 1p/19q.

Beyond these basic testing recommendations, a few practical issues are worth highlighting. Firstly, the ATRX immunostain can be technically challenging and appropriate positive and negative controls are therefore critical for accurate assessment. Internal positive controls, such as entrapped neurons, nonneoplastic glia, and endothelial cells, should show retained expression within the same regions where tumor nuclei are immunonegative. If all nuclei (including nonneoplastic cells) are immunonegative, this should be interpreted as a technical failure of the immunostain rather than true loss of expression in the tumor. A mosaic reactivity pattern, wherein some tumor nuclei appear positive and others negative, often represents a technical failure as well and as such, should not be interpreted as loss of expression within just a subset of tumor cells. Regardless, in the case of equivocal immunostaining or in diagnostic practices that do not initially screen for ATRX inactivation using IHC, ATRX gene alterations can be identified instead using other molecular techniques (eg, NGS). Secondly, since only 70-80% of IDH-mutant astroctyomas (WHO grades 2-4) show ATRX inactivation, retained ATRX immunoexpression does not exclude the possibility of this diagnosis. Infratentorial IDH-mutant astrocytomas show loss of ATRX in only about 50% of cases.⁵⁸ For cases suspected to be IDH-mutant astrocytoma based on morphology, but showing retention of ATRX on IHC, evidence of astrocytic lineage should depend instead on other findings, such as p53 protein overexpression, TP53 gene mutation, or a lack of 1p/19q codeletion. Lastly, ATRX loss may also be encountered in rare IDH-WT GBMs of older adults and in several pediatric/young adult IDH-WT astrocytoma subtypes, as already discussed. As such, ATRX loss is not entirely specific for IDH-mutant astrocytomas and this finding should not be interpreted in isolation, but rather in the context of other alterations, including IDH mutation. DNA methylation profiling has rapidly emerged as a platform that is capable of reproducibly identifying IDH-mutant astrocytomas, as well as subtypes that are infratentorial, supratentorial low grade and supratentorial high grade. Although the technology is not widely available at this time, methylation profiling may be useful in diagnostically challenging cases and will likely emerge as an important diagnostic platform.^{12, 58}

Public comment response to Recommendation 2: There were 92 respondents, of whom 84.78% (n=78) agreed or agreed with modification, 13.05% (n=12) disagreed, and 2.17% (n=2) were neutral. There were 37 written comments, including a number that requested clarification of recommended testing methodologies and a specific algorithm outlining the recommended order of testing, including the relation of ATRX to 1p/19q, p53, histone H3, or TERT promoter analyses. One respondent commented that ATRX status is not needed if 1p/19q is done routinely, whereas another stated that 1p/19q results can be misleading and as such, ATRX data provide useful quality assurance. Although some of these issues are beyond the scope of this paper, a few practical guidelines for testing were nonetheless discussed in the prior paragraph. A few respondents also commented that they did not believe ATRX testing was sufficiently cost effective or diagnostically informative to warrant its routine use, whereas a few felt that Recommendation 2 did not go far enough, since ATRX testing could be potentially useful for all gliomas, including nondiffuse or pediatric astrocytoma variants. These useful comments were taken into consideration and while the final recommendation remained the same, some of these specific concerns were addressed in the text above.

Recommendation 3

Among IDH-mutant DG, TP53 mutation and p53 overexpression are associated with astrocytic lineage and test results for these biomarkers are used to support the diagnosis of IDH-mutant astrocytomas, especially in the setting of ATRX loss or mutation.^{22, 49, 75, 78, 82, 85, 99} IDH-mutant gliomas that have whole arm 1p/19q codeletion (ie, oligodendroglioma, IDH-mutant, 1p/19q codeleted) only rarely (2-3%) exhibit TP53 mutation or strong p53 overexpression.^{19, 42, 71, 83}

In one large series, TP53 mutations were noted in 94% of IDH-mutant DG that did not demonstrate 1p/19q codeletion.⁶ Thus, 1p/19q codeletion and TP53 mutations are nearly, but not entirely, mutually exclusive in IDH-mutant DG. In IDH-mutant DG with TP53 mutations, the 17p-arm typically exhibits copy neutral loss of heterozygosity (cnLOH); thus, there are generally two identical copies of the mutant TP53 gene.^{4-6, 65, 100} In a small percentage of IDH-mutant astrocytomas there are two distinct TP53 mutations; one involves each of the TP53 alleles on the two 17 p-arms.^{4-6, 65, 100} TP53 mutations that occur in IDH-mutant astrocytomas tend to activate the expression and/or enhance the stability of the p53 protein. Thus, p53 IHC is an excellent surrogate of TP53 mutation in IDH-mutant gliomas—especially if the expression of ATRX is not retained.^{101, 102} Unlike IHC for ATRX, however, p53 expression is more of a continuum, and the interpretation may not be straightforward. For higher sensitivity and specificity, >10% of tumor cells must show strong nuclear positivity, although most TP53 mutant gliomas show even more widespread (>50%) p53 expression.¹⁰² Using this threshold, only a small subset of DG with strong p53 immunoreactivity in >10% of nuclei were found to be TP53 WT by sequencing.¹⁰² If tumor DNA sequencing is performed, the combination of IDH, TP53 and ATRX mutations and the associated copy number patterns are distinctive for IDH-mutant astrocytomas.^{4-6, 65, 100} It should be noted that TP53 mutation is also encountered in other DG, especially IDH-WT GBMs.^{24, 29, 41, 42, 86, 103} In IDH-WT gliomas, one TP53 gene typically undergoes mutation, and the WT TP53 gene is lost through deletion. In IDH-WT gliomas, TP53 mutation less predictively generates p53 overexpression.^{75, 83, 104} Nevertheless, p53 overexpression is encountered in up to 30% of IDH-WT GBM, so this finding should not be used as support for IDH-mutant status in isolation.

Public comment response to Recommendation 3: There were 87 respondents, of whom 73.56% (n=64) agreed or agreed with modifications, 18.40% (n=16) disagreed, and 8.04% (n=7) were neutral. There were 35 written comments received during the public comment period. These comments could be classified as follows: 1) The word “should” could be replaced with “must” or “may.” The “should” language of the recommendation was retained based on the strength of the evidence in the literature. 2) The method of TP53 mutation assessment should be addressed. As described in the text above, IHC can be used to assess the expression of p53 protein in the context of IDH-mutant gliomas based on its correlation with TP53 mutation; sequencing methods can also be employed. 3) Some commented that TP53 status should be evaluated in all IDH-mutant DG. Based on the evidence, it was concluded that this is not necessary if whole-arm 1p/19q deletion is known to be present. All comments were taken into consideration and the recommendation remained as initially stated.

Recommendation 4

Confirmation of whole arm 1p/19q codeletion in an IDH-mutant DG is essential to render the diagnoses of oligodendroglioma, IDH-mutant and 1p/19q codeleted. Landmark genomic analyses have firmly established that whole arm 1p/19q codeletion, the product of an unbalanced translocation involving chromosomes 1 and 19, is entirely restricted to gliomas that also harbor mutations in either IDH1 or IDH2.^{4, 6} Therefore, assessment of 1p/19q codeletion status is only relevant in IDH-mutant DG. 1p/19q testing is not recommended in IDH-WT gliomas, as false positive test results (see below) may complicate effective diagnostic categorization and the results have no bearing on patient outcome or treatment.^{105, 106}

As described above, 1p/19q codeletion arises with near mutual exclusivity with respect to inactivating alterations of TP53 and ATRX.^{4, 6} Multiple studies have concluded that loss of nuclear ATRX expression by IHC is strongly predictive of a 1p/19q non-codeleted status.^{74, 76, 82} Confirmed TP53 mutation or p53 overexpression is similarly inversely associated with 1p/19q codeletion.^{74, 75} Loss of ATRX expression or TP53 mutation has been noted in a small minority of IDH-mutant gliomas that also exhibit 1p/19q codeletion (2-8%).^{4, 6, 74, 76, 82} The extent to which these cases reflect false positive testing for 1p/19q is unclear. Notably, TCGA analysis, which employed cytogenomic arrays to confirm whole arm 1p/19q codeletion, revealed low rates of coinciding ATRX or TP53 mutations (2-3%),⁶ suggesting that under strict testing criteria, the mutual exclusivity between ATRX/TP53 mutations and 1p/19q codeletion is near absolute. Nevertheless, the small degree of overlap between these biomarkers may justify 1p/19q codeletion assessment in some IDH-mutant gliomas with established ATRX loss or TP53 mutation/p53 overexpression if oligodendroglioma is suspected based on histopathologic features. As mentioned above, rare cases of carefully documented “dual genotype” IDH-mutant DG have demonstrated the co-occurrence of ATRX loss, TP53 mutation and whole arm 1p/19q codeletion.^95-97 Nonetheless, for most cases, 1p/19q testing is no longer needed if an IDH-mutant DG demonstrates ATRX loss and TP53 mutation, as highlighted in cIMPACT-NOW update 2.⁹⁸

1p/19q codeletion is optimally assessed by cytogenomic array or NGS methodologies capable of definitively demonstrating deletion of whole chromosomal arms of 1p and 19q.^{35, 107} Other approaches, such as fluorescence in situ hybridization (FISH) and PCR-based testing, have also been effectively applied to 1p/19q analysis^{4, 74, 76, 82} and have lower cost and quicker turnaround time than their genomic counterparts. DNA methylation profiling is capable of identifying a methylome highly characteristic of oligodendroglioma and has the added benefit of providing copy number alterations that can directly inform the status of 1p/19q codeletion.¹²

FISH is the most widely used test and has the advantage of assessing a limited sample of tissue (2 slides; a minimum of 50 tumor nuclei required per slide), which is often the case in small biopsy samples. One must take into account the possibility of false positive results, attributable to partial arm deletions that are detected by FISH, yet do not represent whole arm deletions associated with the unbalanced translocation. False positive test results on 1p/19q testing by FISH are also noted in IDH-WT glioblastomas, since they are genomically unstable,^{105, 106} underlying our recommendation that 1p/19q assessment be exclusively restricted to IDH-mutant gliomas.

Public comment response to Recommendation 4: There were 89 respondents, of whom 85.39% (n=76) agreed or agreed with modifications, 11.24% (n=10) disagreed, and 3.71% (n=3) were neutral. Thirty comments were received. One comment suggested that the term “codeletional status” used in the prior version of the recommendation be removed and replaced with “codeletion” for simplicity. The text was modified in response to this recommendation. Multiple comments suggested that 1p/19q testing be performed more broadly, on all DG or on all IDH-mutant DG. The panel did not reach these conclusions based on review of the evidence, and the detailed reasoning is explained in the preceding paragraph. A related series of comments was made regarding the importance of p53 IHC or TP53 mutational analysis in the classification of DG. One suggested using the phrase “ATRX loss with or without TP53 mutations” instead of “ATRX loss or TP53 mutations”. While it was acknowledged that IHC for p53 has limitations, and that TP53 mutations may rarely co-occur with 1p/19q codeletion in IDH-mutant gliomas (see above), the evidence suggests that TP53 mutations are one of the defining genetic characteristics of IDH-mutant astrocytoma (see preceding sections) and its inclusion is justified in the diagnostic workup of IDH-mutant DG. Multiple comments argued for the formal incorporation of histopathological criteria in addition to or in lieu of biomarkers for the classification of oligodendroglioma. Based on our review of the evidence, the designation of glioma subclasses based on IDH mutation and 1p/19q codeletion is more robust and clinically meaningful than histopathology alone.^{4, 6, 74, 76, 82} There is also considerable evidence showing that the interpretation of histopathological features varies considerably, even among experienced pathologists. Multiple comments suggested that pediatric gliomas be explicitly excluded from this recommendation. The sentiment underlying this suggestion is correct, given that oligodendroglioma, IDH-mutant, 1p/19q codeleted is rare in children. However, since IDH mutations occur predominantly in adults and our recommendation for 1p/19q codeletion assessment is only for IDH-mutant DG, testing in pediatric gliomas is essentially ruled out in the vast majority of cases. Finally, multiple comments urged specific recommendations regarding the most appropriate testing modalities for 1p/19q assessment. This point was addressed in the preceding paragraphs.

5. Conditional Recommendation. – CDKN2A/B homozygous deletion testing should be performed on IDH-mutant astrocytomas.

The quality of evidence to support this recommendation was assessed as moderate.

Quality Summary: The evidence base informing this recommendation comprises two prospective cohort studies^{77, 78} and nine retrospective cohort studies.^{41, 84, 85, 99, 108-112} The included studies were assessed as intermediate,⁷⁷ intermediate-low,⁷⁸ low,^{41, 84, 85, 99, 110-112} and very low quality^{108, 109} based on risk of bias in selection,^{41, 77, 78, 84, 85, 99, 108-112} performance,^{78, 99, 108, 112} detection,^{41, 77, 84, 85, 108-110, 112}, and reporting^{41, 77, 78, 84, 85, 99, 108-112} domains. Additionally, two of the studies reported statistical analyses that were underpowered,^{109, 112} and one did not report on sources of funding.¹⁰⁹ Although the aggregate risk of bias across the evidence base was very serious, the evidence was up-graded based on a strong association between CDKN2A/B deletion and diffuse glioma WHO classification. Refer to the SDC for the quality assessment of included retrospective cohort studies and the GRADE Quality of Evidence Assessment. See Table 5 for a summary of the mutational status of CDKN2A/B across all diffuse glioma subtypes.

Table 5.

Representative Studies Reporting on CDKN2A Status Across Diffuse Glioma Subtypes

Study, Study Design	Number DG Cases	Testing Method	CDKN2A Mutation Frequencya	Study Conclusion
Reis et al,99 2015, RCS	n=270	FISH	A: Loss, 48.15% (52/108)	CDKN2A loss was associated with shorter survival among grade 2/3 IDH-mutant DG that had TP53 mutations and ATRX loss.
			OA: Loss, 38.78% (19/49)
			OD: Loss, 23.96% (23/96)
			GBM: NR
Purkait et al,109 2013, RCS	n=67	FISH	A: NR	CDKN2A deletion was observed GBMs and showed strong correlation with loss of p16 expression by IHC.
			OA: NR
			OD: NR
			GBM: Loss, 43.1%b
Collins et al,108 2014, RCS	n=267	CGH array	A: Hemizygous loss, 30.56% (11/36) Homozygous loss, 2.78% (1/36)	Homozygous deletion of CDKN2A was seen at greater frequency in DG of higher histologic grade, especially GBM.
			OA: Hemizygous loss, 33.33% (2/6) Homozygous loss, 0.00% (0/6)
			OD: NR
			GBM: Hemizygous loss, 29.78% (67/225) Homozygous loss, 31.56% (71/225)
Molenaar et al,41 2014, RCS	n=98	MLPA	A: NR	CDKN2A alteration was not associated with survival in patients with histologically diagnosed GBM.
			OA: NR
			OD: NR
			GBM: Hemizygous loss, 27.6% Homozygous loss, 43.9%
Shirahata et al,110 2018, RCS	n=211	450k array	A: NR	CDKN2A/B homozygous deletion significantly associated with overall survival in IDH mutant astrocytomas.
			OA: NR
			OD: NR
			GBM: NR
Yang et al,85 2020, RCS	n=160	FISH	A: 15.00% (24/160)	CDKN2A deletion can be used to risk stratify grade 2/3 astrocytomas.
			OA: NR
			OD: NR
			GBM: NR

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

^bStudy did not report raw values for CDKN2A loss in primary GBMs.

Abbreviations: A, astrocytoma histology; ATRX, ATRX chromatin remodeler; CDKN2A, Cyclin Dependent Kinase Inhibitor 2A; CDKN2B, Cyclin Dependent Kinase Inhibitor 2B; CGH, comparative genomic hybridization; DG, diffuse glioma; FISH, Fluorescence in situ hybridization; GBM, glioblastoma; IDH, Isocitrate Dehydrogenase; IHC, immunohistochemistry; MLPA, multiplex ligation-dependent probe amplification; mut, mutation; NR, not reported; OA, oligoastrocytoma histology; OD, oligodendroglioma histology; RCS, retrospective cohort study; TP53, Tumor protein p53; w/o, without.

CDKN2A lies adjacent to CDKN2B in a region of chromosome 9 that is frequently mutated and deleted in a wide variety of human cancers. CDKN2A/B deletion has been shown by multiple investigations to be an adverse prognostic factor in IDH-mutant astrocytomas.^{41, 85, 99, 108, 109, 111, 112} Subsequent analyses of multiple cohorts of IDH-mutant astrocytomas showed that homozygous deletion of CDKN2A/B was the most relevant adverse prognostic indicator.¹¹⁰ Patients with histologic grade 3 IDH-mutant astrocytomas that harbored homozygous deletion of CDKN2A/B had shorter survivals than patients with histologic grade 4 tumors that did not have CDKN2A/B homozygous deletion. Since homozygous deletion of CDKN2A/B can be observed in other types of primary brain tumors that have highly variable clinical outcomes, including IDH-WT glioblastoma, pleomorphic xanthoastrocytoma, and anaplastic astrocytoma with piloid features, this recommendation pertains specifically to IDH-mutant astrocytomas.

In addition to studies captured in our systematic review, other studies^{1, 113} have shown similar associations between CDKN2A/B homozygous deletion and shorter survival among patients with IDH-mutant astrocytomas. Such findings culminated in the cIMPACT-NOW recommendation that CDKN2A/B homozygous deletion could be used as a grade 4 criterion in a histologic grade 2 or 3 IDH-mutant astrocytoma¹¹ and led to its inclusion as a grade 4 criterion within the WHO, 5^th Edition.⁷ While not a part of the current recommendations, CDKN2A/B homozygous deletion has also been shown to be a marker of poor prognosis in oligodendroglioma, IDH-mutant, 1p/19q codeleted, which may be relevant to future recommendations or grading schemes for these tumors.¹¹¹

Public comment response to Recommendation 5: Of the 86 total respondents to this recommendation, 76.74% (n=66) agreed or agreed with modification, 9.30% (n=8) disagreed, and 13.95% (n=12) were neutral. (Total does not equal 100% due to rounding.) Twenty-three written comments were received. Some respondents expressed a desire to strengthen the language of the recommendation from “may be performed” to “should be performed.” In response to this suggestion, the recommendation was strengthened to “should be performed” based on review of the evidence. Others requested advice regarding methods, desired reconciliation with subsequent cIMPACT-NOW 5 statements, or underscored the importance of the distinction between heterozygous and homozygous loss of CDKN2A/B. The evidence suggests that only homozygous deletion of CDKN2A/B is strongly and reproducibly associated with decreased survival that is independent of tumor grade. These comments were taken into consideration and specific concerns were addressed and clarified in the preceding paragraphs.

6. Strong Recommendation. – MGMT promoter methylation testing should be performed on all GBM IDH-WT.

The quality of evidence to support this recommendation was assessed as moderate.

Quality Summary: This recommendation was informed by two meta-analyses,^{114, 115} three genome sequencing studies,^{5, 21, 22} and five retrospective cohort studies.^{86, 108, 116-118} The included meta-analyses were assessed as high¹¹⁵ and high-intermediate¹¹⁴ quality. Neither of these two studies reported on using publication status as a study selection inclusion criteria,^{114, 115} and one did not report on conflict of interest or sources of funding.¹¹⁴ The genome sequencing studies were all assessed as intermediate-low quality based on retrospective acquisition of samples in all of them,^{5, 21, 22} plus individual moderate risk of reporting²¹ and detection bias.^{5, 21} Finally, the retrospective cohort studies were assessed as low^{86, 116, 117} and very low quality^{108, 118} based on risk of bias. The aggregate risk of bias of the evidence base was serious and the evidence was not further up- or downgraded for any domain. Refer to the SDC for the quality assessment of genome sequencing studies, retrospective cohort studies, meta-analyses, and the GRADE Quality of Evidence Assessment. See Table 6 for a summary of the promoter methylation status of MGMT in glioblastomas.

Table 6.

Representative Studies Reporting on MGMT Promoter Methylation in Glioblastomas

Study, Study Design	Number GBM Cases	Testing Method	Promoter Methylation Frequencya	Study Conclusion
Houdova Megova et al,31 2017, RCS	n=145	MSP	IDHwt GBM: 47.78% (43/90)	IDH1 mutations are closely associated with MGMTp methylation in DG. IDH1 mutations in GBMs were a stronger marker of overall patient survival than MGMTp methylation
			IDHmut GBM:87.50% (7/8)
Lee et al,36 2017, RCS	n=65	Sequencing, MSP	IDHwt GBM: 50.00% (26/52)	MGMTp methylation was much more common in IDH-mutant DG than in IDH-WT DG.
			IDHmut GBM: 76.92% (10/13)
Mulholland et al,143 2012, RCS	n=172	Pyrosequencing, MSP, RT-PCR	GBM: 52.33% (90/172)	Nearly all IDH-mutant DG (127 of 129) showed MGMTp methylation by quantitative analysis of CpG sites following pyrosequencing.
Bady et al,116 2012, RCS	n=174	BeadChip array	GBM: NR	Two sites within MGMTp were critical for gene silencing. A strong link was noted between MGMTp methylation and G-CIMP status.

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

Abbreviations: DG, diffuse glioma; GBM, glioblastoma; G-CIMP, Glioma-CpG island methylator phenotype; IDH1, Isocitrate Dehydrogenase (NADP(+))1; MGMT, O-6-Methylguanine-Deoxiribose Nucleic Acid Methyltransferase; MGMTp, O-6-Methylguanine-Deoxiribose Nucleic Acid Methyltransferase promoter; MSP, methylation-specific polymerase chain reaction; mut, mutation; NR, not reported; RCS, retrospective cohort study; RT-PCR – real time polymerase chain reaction; wt, wild type

The MGMT (O-6-Methylguanine-DNA Methyltransferase) protein, encoded by the MGMT gene, binds to DNA and repairs mutations that occur during DNA replication. Through these actions, it is a key mediator of resistance to alkylating chemotherapy in the treatment of DG. Randomized studies have shown that the clinical benefit of adding temozolomide to radiotherapy is predominantly among patients with MGMT promoter methylated GBM (“predictive marker”).^{119, 120} Temozolomide monotherapy in newly diagnosed GBM is only effective in patients with tumors that have MGMT promoter methylation.^{121, 122} In GBM patients treated with combined temozolomide chemo-irradiation, MGMT promoter methylation is an established prognostic parameter. In a large randomized trial on temozolomide dose intensification, MGMT promoter methylation was associated with improved OS (21.2 months versus 14 months; HR, 1.74; P < .001).¹²³ Clinical benefit of treatment with temozolomide or lomustine at the time of tumor progression after initial therapy is particularly observed in patients with methylated MGMT promoter at the time of first diagnosis.^124-126 Assessment of MGMT promoter methylation can be useful for making treatment decisions for patients when either radiotherapy or chemotherapy is considered contra-indicated, and in patients with recurrent tumors for whom second line alkylating chemotherapy is unlikely to be successful.

The expression of the MGMT protein is inhibited by methylation of specific CpG sites in the MGMT promoter and laboratory testing is therefore focused on these regions.^{116, 127} Despite the overall correlation between MGMT promoter methylation and clinical outcome, test characteristics remain suboptimal. MGMT promoter methylation results are affected by the choice of CpG islands and promoter regions that are tested, as well as the platform employed. Techniques used to assess MGMT promoter methylation status include qualitative or quantitative methylation specific PCR,¹²⁸ pyrosequencing,¹²⁹ multiplex ligation-dependent probe amplification,¹³⁰ genome wide methylation analysis,¹¹⁶ and IHC.¹³¹ With any of these available assays, the determination of the optimal technical cut off between methylated and unmethylated remains a challenge.¹³² In one study that compared 5 techniques for measuring MGMT promoter methylation on a series of 100 GBMs, the percentage of tumors with promoter methylation varied between 30% and 60%, a difference that was attributed to both variations between techniques and between laboratories.^{123, 133, 134}

A major obstacle for current studies on the interpretation and optimization of these techniques is the lack of randomized trials with an untreated control arm that would allow assessment of the predictive value of MGMT promoter methylation. Current studies therefore assess the prognostic significance of various MGMT promoter methylation tests in GBM patients being treated with chemo-irradiation with temozolomide^134-136 or the concordance of results between various techniques.¹²⁹ Individual detection techniques show only modest correlation when compared to one another.^{137, 138} MGMT protein expression by IHC was found unreliable in one series due to large interobserver variation, yet another study concluded that IHC results correlated with methylation-specific-PCR.^{131, 139, 140} Also, in some studies MGMT promoter status was found to vary between paired samples obtained at the time of first diagnosis and progression.¹⁴¹ With a quantitative methylation-specific-PCR technique, analysis within cohorts have revealed a clinically relevant grey zone between MGMT promoter clearly methylated and unmethylated tumors. Establishing an appropriate threshold within this test grey zone has substantial implications for patients in clinical trials in which temozolomide is excluded from therapy based on MGMT methylation status.¹⁴²

Public comment response to Recommendation 6: Of the 86 respondents, 76.74% (n=66) agreed or agreed with modifications, 9.30% (n=8) disagreed, and 13.95% (n=12) were neutral. (Total does not equal 100% due to rounding.) Twenty-five comments were received. Comments emphasized that patients with GBM that had an unmethylated MGMT promoter may also benefit from the addition of temozolomide to radiotherapy, assumed the impact was mainly prognostic and stated that the need for testing depended on clinical decision making and treatment planning. While most of the statement remained unchanged, the panel clarified that the recommendation applies to all GBM IDH-WT.

7. Conditional Recommendation. – For IDH-mutant DG, MGMT promoter methylation testing may not be necessary.

The quality of evidence to support this recommendation was assessed as low. Quality Summary: The evidence base informing this recommendation comprises one genome sequencing study⁵ and four low quality retrospective cohort studies.^{31, 36, 40, 143} The genome sequencing study⁵ was assessed as intermediate-low quality based on a serious risk of selection bias and a moderate risk of detection bias. All retrospective cohort studies were limited by critical risk of selection bias.^{31, 36, 40, 143} Individual studies were further limited by risk of bias in performance,^{31, 40} reporting,^{31, 36, 40, 143} and detection domains,^{36, 143} as well as underpowered statistical analyses in two,^{31, 143} and a lack of reporting funding sources in one.⁴⁰ The aggregate risk of bias across these studies was very serious but evidence was not further downgraded for any domain. Refer to the SDC for the quality assessment of included genome sequencing studies, retrospective cohort studies, and the GRADE Quality of Evidence Assessment.

In the section above, data supporting MGMT promoter methylation testing in GBM, IDH-WT were presented. The clinical relevance of testing for MGMT promoter methylation in IDH-mutant DG is not as firmly established. IDH mutations lead directly to global DNA hypermethylation and the establishment of the CpG island methylator phenotype (G-CIMP), as demonstrated experimentally and in numerous investigations of clinical specimens.^{5, 144, 145} Therefore, the prevalence of MGMT promoter methylation in IDH-mutant DG is very high .^{40, 144, 146, 147}

Data supporting this recommendation from our systematic review documents the tight correlation between IDH-mutant DG and MGMT promoter methylation.^{31, 36, 40, 143, 148, 149} Using bisulfite treatment, pyrosequencing, and focusing on 16 selected CpG sites, Mulholland et al found that 127/129 (98.45%) gliomas with IDH mutations were MGMT methylated, irrespective of pathologic group.¹⁴³ A strong association between IDH mutation and MGMT promoter methylation has also been demonstrated by others.^{31, 40} Thus, the high correlation between IDH-mutation and MGMT promoter methylation suggests that testing for MGMT promoter methylation in IDH-mutant DG may not be necessary.

Public comment response to Recommendation 7: There were 89 respondents, of whom 73.03% (n=65) agreed or agreed with modifications, 16.85% disagreed (n=15), and 10.11% (n=9) were neutral. (Total does not equal 100% due to rounding.) There were 17 written comments, many of which stated that MGMT testing decisions should be left to the treating oncologist and questioned whether it is worthwhile to issue a recommendation for every test that should not be performed. We have carefully taken these comments into consideration and have kept the recommendation as is, since MGMT promoter methylation testing is one of the most frequently requested tests for DG, and we did not find sufficient evidence for reflexively ordering it in the group of IDH-mutant DG. The strong association between IDH mutation and MGMT promoter methylation supports a role for IDH mutation (when present) resulting in MGMT promoter methylation, with the limitations of predominantly retrospective evidence. The recommendation is listed as conditional, which leaves room for individual and institutional treatment decisions.

8. Conditional Recommendation. – TERT promoter mutation testing may be used to provide further support for the diagnosis of oligodendroglioma and IDH-WT GBM.

The quality of evidence to support this recommendation was assessed as low.

Quality Summary: This recommendation was informed by two genome sequencing studies^{5, 6} and 11 retrospective cohort studies.^{4, 18, 23, 34, 49, 84, 150-154} The genome sequencing studies^{5, 6} were assessed as intermediate-low quality while the retrospective cohort studies were assessed as low^{23, 49, 84, 150, 151, 153, 154} and very low quality.^{4, 18, 34, 152} Included studies suffered from risk of bias in selection,^{4-6, 18, 23, 34, 49, 84, 150-154} performance,^{4, 18, 34, 150} reporting,^{4, 18, 23, 34, 84, 150-152, 154} and detection^{4, 5, 23, 34, 84, 150-152, 154} domains. The aggregate risk of bias across these studies was very serious but evidence was not further downgraded for any domain. Refer to the SDC for the quality assessment of included genome sequencing studies, retrospective cohort studies, and the GRADE Quality of Evidence Assessment. See Table 7 for a summary of the mutational status of the TERT promoter across all diffuse glioma subtypes.

Table 7.

Representative Studies Reporting on TERT Promoter, EGFR, Chromosome 7, and Chromosome 10 Status Across Diffuse Glioma Subtypes

Study, Design	Number DG Cases	Marker(s)	Testing Method	Mutation Frequencya	Study Conclusion
McLendon et al,22 2008, GSS	n=91	EGFR	Genome sequencing	A: NR	Highlighted core molecular pathways consistently altered in GBM, including p53, RTK and RB networks.
				OA: NR
				OD: NR
				GBM: NR
Brat et al,6 2015, GSS	n=293	IDH, TP53, ATRX, 1p/19q, EGFR	Genome sequencing	A: NR	IDH-WT lower grade DG had genomic alterations and clinical behavior like IDH-WT GBM.
				OA: NR
				OD: NR
				GBM: NR
Brennan et al,21 2013, GSS	n=543	Chr 7, Chr 10, TERTp, EGFR	Genome sequencing	A: NR	EGFR amp, TERTp mutation, Chr 7 gains and Chr 10 losses were frequent events in IDH-WT GBM.
				OA: NR
				OD: NR
				GBM: TP53, 27.9%b TERT, 84.0%b
Chan et al,23 2015, RCS	n=237	TERT	Sequencing	A: TERT, 19.05% (32/168)	Among histologic grade 2/3 IDH-WT DG, TERTp mutation was associated with short survival, similar to IDH-WT GBM.
				OA: TERT, 39.58% (19/48)
				OD: TERT, 76.19% (16/21)
				GBM: NR
Gao et al,150 2016, RCS	n=389	TERT	Sequencing, Pyrosequencing	A: TERT, 26.72% (66/247)	TERTp mutations a prognostic factors among DG and inversely associated with ATRX inactivation.
				OA: TERT, 41.30% (19/46)
				OD: TERT, 22.73% (10/44)
				GBM: TERT, 32.69% (17/52)
Chan et al,160 2015, RCS	n=214	TERT, EGFR	Sequencing, FISH	A: NR	Among patients with grade 2/3 IDH-WT DG, EGFR amp and TERTp mutation was associated with shorter survival, similar to IDH-WT GBM.
				OA: NR
				OD: NR
				GBM: NR
Eckel-Passow et al,4 2015, RCS	n=615	TERT	Sequencing	A: NR	DG classified on the basis of IDH, TERTp, and 1p/19q showed that those with TERTp mutation only had short overall survival.
				OA: NR
				OD: NR
				GBM: NR
Labussiere et al,151 2014, RCS	n=807	TERT	Array, PCR	A: TERT, 30.23% (13/43)	Among IDH-WT GBM, the presence of TERTp mutation or EGFR amp was associated with shorter survival.
				OA: TERT, 31.95% (54/169)
				OD: TERT, 61.00% (122/200)
				GBM: TERT, 75.70% (299/395)
Collins et al,108 2014, RCS	n=267	Chr 7, Chr 10, EGFR	CGH array	A: Chr7 tri/poly, 5.56% (2/36); mono, 0.00% (0/36) Chr10 mono, 11.11% (4/36); 10q partial del, 38.89% (14/36) EGFR amp, 11.11% (4/36)	Among patients with recurrent grade 3/4 astrocytoma, EGFR amp Chr 7 gains and Chr 10 losses were associated with shorter survival.
				OA: Chr7 tri/poly, 0.00% (0/6); mono, 0.00% (0/6) Chr10 mono, 0.00% (0/6); 10q partial del, 33.33% (2/6) EGFR amp, 0.00% (0/6)
				OD: NR
				GBM: Chr7 tri/poly, 63.56% (143/225); mono, 0.44% (1/225) Chr10 mono, 64.89% (146/225); 10q partial del, 20.00% (45/225) EGFR amp, 44.44% (100/225)
Little et al,162 2012, RCS	n=342	EGFR	FISH	A: EGFR amp, 15.38% (2/13)	Detailed FISH analysis of GBMs showed that 44% have EGFR amp, 21% have PDGFRA amp and 9% contain both EGFR amp and PDGFRA amp.
				OA: NR
				OD: EGFR amp, 9.76% (4/41)
				GBM: EGFR amp, 44.17% (106/240)
Guillaudeau et al,161 2009, RCS	n=35	EGFR	FISH, IHC	A: EGFR amp, 0% (0/8)	EGFR amp was associated with increased protein expression of EGFR by IHC in DG
				OA: NR
				OD: NR
				GBM: EGFR amp, 72.73% (8/11)
Houdova Megova et al,31 2017, RCS	n=145	EGFR	Not reported	A: EGFR amp, 7.69% (2/26)	EGFR amp and Chr 10 losses are more frequent in IDH-WT than IDH-mut DG and are most frequent in IDH-WT GBM.
				OA: EGFR amp, 0% (0/4)
				OD: EGFR amp, 0% (0/7)
				GBM: EGFR amp, 36.56% (30/82)
Aibaidula et al,18 2017, RCS	n=166	TERT	Patient records	A: NR	Among adults with IDH-WT grade 2/3 DG, EGFR amp and mutations of TERTp or H3F3A K27M were associated with shorter survival.
				OA: NR
				OD: NR
				GBM: NR

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

^bStudy did not report raw values for mutation frequency.

Abbreviations: A, astrocytoma histology; amp, amplification; ATRX, ATRX chromatin remodeler; Chr, chromosome; del, deletion; DG, diffuse glioma; EGFR, epidermal growth factor receptor; FISH, Fluorescence in situ hybridization; GBM, glioblastoma; GSS, genome sequencing study; H3, histone3; IDH, Isocitrate Dehydrogenase; IHC, immunohistochemistry; mono, monopoly; mut, mutation; NR, not reported; OA, oligoastrocytoma histology; OD, oligodendroglioma histology; PCR, polymerase chain reaction; PDGFRA, platelet derived growth factor receptor alpha; poly, polysomy; RCS, retrospective cohort study; TERT, Telomerase Reverse Transcriptase; TERTp, TERT promoter; TP53, Tumor protein p53; tri, trisomy; wt, wild-type.

As discussed for ATRX (Recommendation 2), cancerous proliferation requires escape from cellular senescence. In oligodendrogliomas, IDH-mutant, 1p/19q codeleted and IDH-WT GBM this predominantly occurs via point mutations in the TERT promoter, which leads to enhanced TERT expression. Point mutations occur in two hotspots located 124bp and 146bp upstream of the translation start site and are referred to as C228T and C250T.^{34, 151} Testing for promoter mutation is performed by DNA sequence analysis, either by Sanger methods, pyrosequencing, PCR or targeted NGS.^{25, 34, 84, 150, 154, 155}

In contrast to oligodendrogliomas and IDH-WT GBM, IDH-mutant astrocytomas primarily use the alternative lengthening of telomeres (ALT) pathway to escape senescence, which is driven by inactivation of ATRX. Notably, TERT promoter mutations and ATRX loss/mutations are mutually exclusive,^{4, 6, 150} consistent with our understanding that these two genetic mechanisms confer equivalent selective growth advantages. Thus, in the setting of IDH-mutation, TERT promoter mutations can provide additional support for the diagnosis of oligodendroglioma, IDH-mutant, 1p/19q codeleted, WHO grade 2 or 3. It can also provide diagnostic support in cases in which immunostaining for ATRX is equivocal or was not performed. By the same logic, TERT promoter sequencing may not be necessary in IDH-mutant gliomas with sufficient evidence of ATRX loss. Similarly, if there is laboratory evidence that clearly establishes the diagnosis of oligodendroglioma (IDH-mutant and 1p/19q codeleted), then TERT promoter testing may not be necessary.

TERT promoter mutations were initially noted in the majority of histologically defined GBMs.^{156, 157} Subsequently, IDH mutations and TERT promoter mutations were shown to be mutually exclusive in GBMs, with those GBMs harboring TERT promoter mutations associated with shorter survivals compared to those with IDH mutations.^{34, 151, 155} Further comprehensive investigations demonstrated that those adult DG that harbored TERT promoter mutations but did not have IDH-mutations, were tightly associated with genetic signatures of IDH-WT GBM and short overall patient survivals.^{4, 5, 158} TERT promoter mutations are not specific for IDH-WT GBM, since they have been documented in other forms of CNS neoplasia in addition to oligodendroglioma, IDH-mutant, 1p/19q codeleted.^{156, 158, 159} However, in the proper clinical, radiological and neuropathological setting, TERT promoter mutations may be used to support the diagnosis of IDH-WT GBM.

Public comment response to Recommendation 8: There were 88 respondents, of which 73.86% (n=69) agreed or agreed with modifications, 17.05% (n=9) disagreed, and 10.23% (n=10) were neutral. (Total does not equal 100% due to rounding.) There were 15 comments, including several that reiterated the importance of interpreting TERT promoter mutations in the context of IDH-status, as well as its established relationships with ATRX mutations and 1p/19q codeletions, as discussed above. The recommendation was unchanged.

9. Strong Recommendation. – For histologic grade 2-3 DG that are IDH-WT, testing should be performed for whole chromosome 7 gain/whole chromosome 10 loss, EGFR amplification, and TERT promoter mutation to establish the molecular diagnosis of IDH-WT GBM, grade 4.

The quality of evidence to support this recommendation was assessed as moderate.

Quality Summary: The evidence base informing this recommendation includes studies evaluating testing of chromosome 7, chromosome 10, EGFR and TERT promoter. One genome sequencing study²¹ and one retrospective cohort study¹⁰⁸ comprise the evidence base for chromosome 7, one retrospective cohort study was included for chromosome 10,¹⁰⁸ two genome sequencing studies,^{21, 22} and eight retrospective cohort studies^{18, 31, 49, 108, 160-163} for EGFR, and one genome sequencing study⁵ and eight retrospective cohort studies for TERT.^{4, 18, 23, 49, 150-153} Included studies were assessed as intermediate-low through very low quality and suffered from risk of bias in selection,^{4, 5, 18, 21-23, 31, 49, 108, 150-153, 160-163} performance,^{4, 18, 31, 108, 150, 163} reporting,^{4, 18, 21, 23, 31, 108, 150-152, 160, 161, 163} and detection^{4, 5, 21, 23, 108, 150-152, 160-162} domains. Quality of evidence was assessed for each target individually and overall for the recommendation. For both chromosome 7 and EGFR testing, the aggregate risk of bias for studies included in the evidence base was serious and evidence was not further downgraded, resulting in moderate quality of evidence. For chromosome 10 the risk of bias of the one included study was very serious, thus carrying a low quality of evidence. Finally, the quality of evidence for TERT was also low based on a very serious aggregate risk of bias and no further downgrading of quality. The overall recommendation was assessed as moderate. Refer to the SDC for the quality assessment of included genome sequencing studies, retrospective cohort studies, and the GRADE Quality of Evidence Assessment. See Table 8 for a summary of the mutational status of EGFR, TERT promoter, chromosome 7, and chromosome 10 across all diffuse glioma subtypes.

Table 8.

Representative Studies Reporting on Histone H3 Mutation Frequency Across Diffuse Glioma Subtypes

Study, Study Design	Number DG Cases	Marker(s)	Testing Method	Mutation Frequencya	Study Conclusion
Castel et al,168 2015, RCS	n=62	H3K27M	Sequencing, IHC	Midline Glioma: H2K27M, 95.16% (59/62)	There was a high frequency of H3-K27M mutations in DIPGs and the correlation between Sanger sequencing and IHC for H3-K37M mutations was strong.
Khuong-Quang et al,88 2012, RCS	n=42	H3K27M	Sequencing	Midline Glioma: H3K27M, 71.43% (30/42)	H3-K27M mutations were found in high frequency in DIPG and had a poor prognosis compared to DIPG without this mutation. TP53 mutations co-occurred with H3-K27M mutations with high frequency.
Gessi et al,171 2013, RCS	n=123	H3G34, H3G27	Pyrosequencing, MLPA	GBM: H3G34/ H3K27 31.71% (39/123)	Histone H3-G34R mutations were identified in a subset of pediatric tumors histologically classified as GBM and as CNS-PNET.
				CNS-PNET: H3G34/ H3K27, 12.12% (4/33)
Korshunov et al,3 2015, RCS	n=202	H3G34, H3K27M	450k BeadChip array	GBM: H3G34, 11.88% (24/202) H3G34/ H3K27M, 46.04% (93/202)	Pediatric tumors diagnosed histologically as GBM had specific genetic and clinical subsets, defined by mutations in H3-K27M, H3-G34R and IDH1.

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

Abbreviations: CNS-PNET, pediatric primitive neuroectodermal tumors of the central nervous system; DG, diffuse glioma; DIPG, diffuse intrinsic pontine glioma; GBM, glioblastoma; IDH1, Isocitrate Dehydrogenase (NADP(+))1; IHC, immunohistochemistry; MLPA, multiplex ligation-dependent probe amplification; RCS, retrospective cohort study; TP53, Tumor protein p53.

A substantial subset of IDH-WT diffuse astrocytic gliomas in adults are considered grade 2 or 3 based on histologic criteria (no microvascular proliferation or necrosis), yet have an aggressive clinical course, with overall survival times similar to patients with IDH-WT GBM. Numerous studies have attempted to identify molecular genetic biomarkers that reliably identify histologic grade 2 and 3 tumors that behave most aggressively. The strongest evidence indicates that the following markers identify IDH-WT diffuse astrocytic gliomas with grade 4 clinical behavior: 1) whole chromosome 7 gain together with whole chromosome 10 loss (+7/−10); 2) EGFR amplification; or 3) TERT promoter mutation. Studies that have assessed the frequency of +7/−10, EGFR amplification and TERT promoter mutations have demonstrated higher percentages of these alterations in histologically defined GBMs as compared to histologic grade 2 or 3 DG, lending support to the association between these markers and the molecular signature of IDH-WT GBM.^{18, 21, 22, 31, 108, 151, 161, 162}

IDH-WT diffuse astrocytic gliomas of histologic grade 2 or 3 that carry EGFR amplification are associated with significantly shorter patient survival and have outcomes similar to patients with histologically defined IDH-WT GBM.^{5, 6, 18, 158, 160} EGFR amplification has excellent specificity for gliomas with aggressive behavior and is not present in other glioma subtypes that display a more indolent clinical course. It is typically accompanied by gain of chromosome 7 and loss of chromosome 10 in adult DG, as well as in a smaller subset of pediatric high grade gliomas.^{158, 164}

Similarly, those histologic IDH-WT diffuse astrocytic gliomas of histologic grade 2 or 3 with the +7/−10 signature are associated with short patient survival, similar to histologically defined IDH-WT GBM.^{5, 158, 165} The +7/−10 signature also has excellent specificity for DG, with the rare exception of pleomorphic xanthoastrocytomas (PXAs).¹⁵⁸ Most investigations have focused on the prognostic role of whole gains of chromosome 7 and whole losses of chromosome 10 (+7/−10), since these are most frequent; the prognostic association of other, far less common combinations of imbalances, such as +7q/−10q, +7/−10q or +7q/−10, await further evaluation.

TERT promoter mutation in IDH-WT diffuse astrocytic gliomas of histologic grade 2 or 3 is associated with significantly shorter patient survival.^{4, 18, 23, 160} However, other types of IDH-WT glial neoplasms without WHO grade 4 histology or aggressive behavior also occasionally harbor TERT promoter mutations.^{156, 158, 159} Therefore, TERT promoter mutations can be considered a marker for grade 4 behavior if the clinical, radiologic and histopathologic features are those of diffusely infiltrative astrocytic glioma. Further, more recent investigations have concluded that a subset of histologic grade 2 IDH-WT DG that harbor only TERT promoter mutations (no other genetic alterations involving canonical IDH-WT GBM pathways) have longer clinical outcomes than expected for a grade 4 neoplasm.^{166, 167} The combination of TERT promoter mutation with other markers, such as EGFR amplification and +7/−10 adds specificity as a marker of grade 4 behavior.¹⁵⁸

cIMPACT-NOW update 3 has recommended that IDH-WT DG of histologic grade 2 or 3 that have +7/−10, EGFR amplification or TERT promoter mutation should be considered grade 4 and more recently, these molecular alterations have been added as criteria for the diagnosis of GBM, IDH-WT, WHO grade 4 by the WHO 5^th edition.^{7, 10} Clinical implications for the use of these biomarkers to establish the molecular diagnosis of IDH-WT GBM, grade 4 include the use of combined chemo- and radiotherapy and expanded clinical trial inclusion.

Public comment response to Recommendation 9: Of 89 respondents, 86.52% (n=77) agreed or agreed with modifications, 6.74% (n=6) disagreed, and 6.74% (n=6) were neutral. Twenty written comments were received, many expressing concern about the use of “grade IV astrocytic glioma.” As “astrocytic glioma” is not currently an official WHO class, respondents requested that the statement be changed to “molecular diagnosis of GBM.” Additional comments requested that the recommendation include ‘and’ instead of ‘or’ for the target testing to reduce turnaround time. These suggestions were incorporated into the final recommendation.

10. Strong Recommendation. – H3 K27M testing must be performed in DG that involve the midline in the appropriate clinical and pathologic setting.

The quality of evidence to support this recommendation was assessed as moderate.

Quality Summary: This recommendation was informed by two retrospective cohort studies.^{88, 168} Both retrospective studies were limited by risk of bias in selection, performance, reporting, and detection domains.^{88, 168} In addition, one study reported underpowered statistical analyses.¹⁶⁸ The aggregate risk of bias for the evidence base was very serious and the quality of evidence was up-graded based on a strong association between H3 K27M testing and DG diagnosis in the WHO classification. Refer to the SDC for the quality assessment of included retrospective cohort studies and the GRADE Quality of Evidence Assessment. See Table 8 for a summary of the mutational status of histone H3 in glioblastoma, midline glioma, and pediatric primitive neuroectodermal tumors of the central nervous system.

Diffuse midline glioma with histone H3 K27M mutation was a newly recognized diagnostic entity in the 4^th Edition update of the WHO Classification of Tumours of the Central Nervous System (2016).¹ The diagnosis relies on the histological identification of diffusely infiltrating glioma within midline CNS structures such as spinal cord, brainstem, or thalamus and the identification of K27M mutation in either H3F3A or HIST1H3B/C using anti-H3K27M IHC or by sequencing methods.^{90, 169} By histology, tumors can show a large degree of histo- and cytopathological variability, ranging from a low density infiltrate to an undifferentiated, high grade neoplasm. Regardless of histologic features, the presence of an H3 K27M mutation in a diffusely infiltrative glioma involving the midline most often predicts clinically aggressive behavior and poor prognosis, leading to its designation of WHO grade 4.^{88, 168} While the number of studies investigating this particular mutation is small, the implications for diagnosis, prognosis and treatment are critically important. This underlies the strong recommendation that H3 K27M testing must be performed in DG that involve the midline in the appropriate clinical and pathologic setting. As most of these tumors occur in the pediatric population, this recommendation for additional testing is essential for children and young adults. For older adults there is increasing evidence that the midline location of a DG is also tightly linked with H3 K27M mutation, warranting testing of all patients with midline gliomas in order to guide appropriate care.⁹¹

H3 K27M testing can be performed using anti-H3 K27M IHC, which displays strong nuclear staining with relatively high sensitivity for the mutation, ranging from approximately 71% to 100% when compared with sequencing methods.^{88, 168, 170-172} IHC can be particularly useful in small biopsy samples, when tissue is limited. Several anti-H3 K27M antibodies are commercially available for on-site laboratory development, and reference laboratories also offer testing. Importantly, the staining pattern of a positive test is nuclear, and therefore staining of cytoplasmic and/or cellular processes should not be interpreted as positive. Confirmatory molecular sequencing may be of utility in equivocal cases. DG that have H3 K27M mutations also show loss of H3 K27M trimethylation (H3K27me3), which can be detected by loss of nuclear immunoreactivity for H3 K27me3 by IHC.¹⁷³ A smaller subset of diffuse midline gliomas that show loss of H3 K27me3 do not have H3 K27M mutations, but rather have EGFR mutations (see Good Practice Statement 1) or overexpression of EZHIP.^{172, 174, 175}

Public comment response to Recommendation10: There were 88 respondents, of whom 96.59% (n=85) agreed or agreed with modifications and 3.37% (n=3) disagreed. There were 24 written comments. Originally, the statement included that the recommendation applied to “the appropriate clinical and pathologic setting.” Some commented that this phrase needed to be further explained. There were also suggestions that the statement should recommend that the H3 K27M test phrasing of “should” be performed rather than “must” be performed. Others indicated that the methods of testing (eg, IHC versus sequencing) should be described. These comments were taken into consideration. While the recommendation remained the same, these comments have been addressed in the preceding paragraphs.

11. Conditional Recommendation. – H3 G34 testing may be performed in pediatric and young adults with IDH-WT DG.

The quality of evidence to support this recommendation was assessed as low.

Quality Summary: The evidence base informing this recommendation comprises two low quality retrospective cohort studies.^{3, 171} In addition to suffering from critical risk of selection bias, both studies were limited by a moderate risk of reporting bias^{3, 171} and one was also limited by risk of performance and detection bias.³ The aggregate risk of bias for the studies was very serious and the evidence was not further downgraded for any domain. Refer to the SDC for the quality assessment of included retrospective cohort studies and the GRADE Quality of Evidence Assessment. See Table 8 for a summary of the mutational status of histone H3 in glioblastoma, midline glioma, and pediatric embryonal tumors of the central nervous system.

A hotspot mutation in the histone gene H3F3A found within a subset of pediatric and young adult high-grade gliomas confers either a G34R or G34V substitution in the gene product.^{3, 171, 176, 177} G34R/V mutant diffuse gliomas occur in a somewhat older age group in comparison with K27M cases (median age, 19-years for G34R/V mutation; median age 6 for K27M mutation) and tend to involve the cerebral hemispheres rather than midline locations.¹⁷⁶ Prognosis for G34R/V-mutant gliomas is poor but somewhat better than the K27M-mutant gliomas (median survival of 12 months for K27M and 24 months for G34R or G34V). “Diffuse hemispheric glioma H3 G34-mutant” was recently introduced as a distinct entity in the WHO 5^th Edition, underscoring the importance of testing for this molecular alteration.⁷ Importantly, these DG with G34R/V mutations should not be lumped together with other IDH WT gliomas, as they carry a disease defining genetic alteration that directs aggressive behavior corresponding to a WHO grade 4 neoplasm regardless of histologic appearance.^{3, 171} Testing by sequencing or with mutation-specific (H3 G34R) antibodies is recommended for non-midline DG in pediatric and young adult populations.

Public comment response to Recommendation 11: There were 89 respondents, of whom 83.15% (n=74) agreed or agreed with modification, 4.49% (n=4) disagreed, and 12.36% (n=11) were neutral. There were several written comments, including one that suggested that the age of the patient should be considered but that older adults might also benefit from testing for some cases. Others indicated that the methods of testing (ie, IHC versus sequencing) should be described. These comments were taken into consideration. While the recommendation remained the same, the comments were addressed in the text above.

12. Conditional Recommendation. – BRAF mutation testing (V600) may be performed in DG that are IDH-WT and H3-WT.

The quality of evidence to support this recommendation was assessed as low.

Quality Summary: The evidence base informing this recommendation consists of one intermediate quality prospective cohort study⁷⁷ and four retrospective cohort studies.^{18, 178-180} Two included retrospective studies were assessed as very low quality based on serious or critical risk of bias across more than one domain.^{18, 178} The other two retrospective studies were also assessed as very low quality and this was based on critical risk of selection bias, moderate risk of bias in other domains, underpowered statistical analyses, and a lack of funding source reporting.^{179, 180} The aggregate risk of bias for the studies was very serious but evidence was not further downgraded for any domain. Refer to the SDC for the quality assessment of retrospective cohort studies and the GRADE Quality of Evidence Assessment. See Table 9 for a summary of the mutational status of BRAF, MYB/MYBL1, and FGFR1 across diffuse glioma subtypes.

Table 9.

Representative Studies Reporting on BRAF, MYB/MYBL1, and FGFR1 Status Across Diffuse Glioma Subtypes

Study, Study Design	Number DG Cases	Marker	Testing Method	Mutation Frequencya	Study Conclusion
Korshunov et al,179 2017, RCS	n=64	BRAF V600E	450k array	E-GBM: BRAF, 56.25% (36/64)	Epithelioid GBMs resolve into three genetic entities. Those clustering with PXA and IDHwt GBM by methylation are enriched for BRAF V600E mutations.
Kleinschmidt-DeMasters et al,178 2013, RCS	n=24	BRAF V600E	IHC, PCR	E-GBM: BRAF, 53.85% (7/13)	BRAF V600E mutations were noted in a substantial percentage of epithelioid GBMs, but not rhabdoid or giant cell GBMs.
Nakajima et al,180 2017, RCS	n=14	BRAF V600E	Sequencing, CGH array	E-GBM: BRAF, 92.86% (13/14)	Epithelioid GBMs are characterized by BRAF V600E mutations, TERTp mutations, and CDKN2A/B deletions, most often in combination.
Aibaidula et al,18 2017, RCS	n=166	BRAF V600E, MYB	Sequencing, FISH	IDHwt (all DGs): BRAF, 6.02% (10/166) MYB, 19.88% (33/166)	Subset of adult IDH wild-type histologic grade 2/3 DGs harbor clinically relevant BRAF V600E mutations and MYB amps.
Tatevossian et al,183 2010, RCS	n=77	MYB	SNP array	A: MYB amp, 14.29% (2/14); poly, 7.14% (1/14)	MYB amp was described in a subset of pediatric low grade DA, leading to increased gene and protein expression. Focal truncating deletion of MYB was noted in AG.
Ramkisson et al,184 2013, RCS	n=44	MYB	Copy number analysis	A: MYBL1 rearrangement, 26.32% (5/19)	Recurrent truncating duplications of MYBL1 were present in a subset of low grade pediatric DA. Focal truncating deletion of MYB was noted in AG.
Qaddoumi et al,2 2016, RCS	n=91	FGFR1	Sequencing	A: FGFR1/ 3, 23.53% (4/17)	Alterations in BRAF, FGFR1, or MYB/MYBL1 were solitary pathogenic alterations in subsets of low grade pediatric DG, often corresponding to histologic class.
				OD: FGFR1/ 3, 61.90% (26/42)

^aMutational frequencies refer to histologic classifications. See full data tables in the supplemental digital content for all outcomes reported by included studies

Abbreviations: A, astrocytoma histology; AG, angiocentric glioma; amp, amplification; BRAF, B-Raf proto-oncogene; CDKN2A, Cyclin Dependent Kinase Inhibitor 2A; CDKN2B, Cyclin Dependent Kinase Inhibitor 2B; CGH, comparative genomic hybridization; DA, diffuse astrocytoma; DG, diffuse glioma; E-GBM, epithelioid glioblastoma histology; FGFR1, Fibroblast Growth Factor Receptor 1; FISH, Fluorescence in situ hybridization; GBM, glioblastoma; OD, oligodendroglioma histology; IDH, Isocitrate Dehydrogenase; IHC, immunohistochemistry; MYB, MYB proto-oncogene; MYB1, MYB proto-oncogene like 1; NR, not reported; PCR, polymerase chain reaction; poly, polysomy; PXA, pleomorphic xanthoastrocytomas; RCS, retrospective cohort study; SNP, single nucleotide polymorphism; TERTp, Telomerase Reverse Transcriptase promoter; wt, wild-type.

BRAF V600E mutation is a driver mutation that is present in a wide variety of neoplasms, including melanoma, hairy cell leukemia, papillary thyroid carcinoma, colorectal carcinoma, lung carcinoma, Langerhans cell histiocytosis, Erdheim Chester disease, papillary craniopharyngioma, and ameloblastoma. It is present in a minority of DG, but prevalence is enriched in epithelioid GBMs^178-180 and other IDH- and H3-WT DG.^{2, 18} Of note, BRAF V600E mutation is distinct from the KIAA1549:BRAF fusion and other types of BRAF fusion, which are especially frequent in cerebellar pilocytic astrocytomas.

In addition to these studies, the recommendation was influenced by the availability of targeted therapy¹⁸¹ and the occurrence of BRAF V600E within a variety of brain tumors other than DG, where the mutation is more prevalent; these include PXAs, ganglioglioma (GG), and a minority of pilocytic astrocytomas.^{2, 182} Although BRAF V600E mutation occurs in a higher proportion of these generally circumscribed neuroepithelial tumors, PXA, GG, and pilocytic astrocytoma fell outside of the scope and search criteria used to develop these guidelines. However, histologic overlap between these entities can be considerable; given the presence of targeted therapy, a liberal approach to testing was favored in borderline or ambiguous cases where the distinction between a diffuse and circumscribed neoplasm could not be made with certainty.

This recommendation is most relevant to IDH- and H3-WT DG in which BRAF V600E mutations have been shown to be enriched, including: epithelioid GBMs and cases overlapping with PXA or anaplastic PXA; pediatric DG or tumors overlapping with GG or pilocytic astrocytoma; and histologic grade 2 and 3 DG lacking a “GBM molecular signature”.¹⁰ Evidence for BRAF V600E testing in the setting of an IDH-WT GBM with typical histologic features or with canonical genetic alterations in older adults is currently lacking.

While many of the markers considered in this guideline pertain principally to DG in adults, it should be noted that a small number of pediatric DG with BRAF V600E mutation have been identified,² and that BRAF V600E testing should be considered in the context of IDH- and H3-WT DG in children.⁹ This is particularly important given relatively common histologic overlap between circumscribed and DG in children.

Public comment response to Recommendation 12: Of the 89 respondents to this recommendation, 84.27% (n=75) agreed or agreed with modifications, 11.23% (n=10) disagreed, and 4.49% (n=4) were neutral. (Total does not equal 100% due to rounding.) Of 24 written comments, 7 solicited advice as to the best method of testing, and 5 emphasized the importance of clinical and pathological correlation. The remaining 12 comments expressed diverse thoughts, both supportive and skeptical, including concern about the difference between the strength of the recommendation and the level of evidence (2 comments). These helpful comments were taken into account and the initial recommendation was streamlined to encompass all IDH- and H3-WT DG (within the appropriate clinicopathologic context) instead of epithelioid GBM and “other” IDH- and H3-WT DG. We also included a more specific description of the appropriate clinicopathologic context for testing in the text above.

13. Conditional Recommendation. – MYB/MYBL1 and FGFR1 testing may be performed in children and young adults with DG that are histologic grade 2-3 and are IDH-WT and histone H3-WT.

The quality of evidence to support this recommendation was assessed as low.

Quality Summary: The evidence base informing this recommendation includes three retrospective cohort studies evaluating MYB/MYBL1 testing^{18, 183, 184} and one retrospective cohort study evaluating FGFR1 testing.² The four studies were assessed as low^{183, 184} and very low quality^{2, 18} based on risk of bias in selection,^{2, 18, 183, 184} performance,¹⁸ reporting,^{2, 18, 183, 184} and detection^{2, 183, 184} domains. Quality of evidence was assessed for each target individually and then overall for the entire statement. The aggregate risk of bias for MYB/MYBL1 studies was very serious but evidence was not further downgraded for any domain. The risk of bias for the one FGFR1 study was very serious. As only one study was identified, quality of evidence is dependent only on the risk of bias. The quality of evidence for the entire recommendation was assessed as low. Refer to the SDC for the quality assessment of included retrospective cohort studies and the GRADE Quality of Evidence Assessment. See Table 9 for a summary of the mutational status of BRAF, MYB/MYBL1, and FGFR1 across diffuse glioma subtypes.

Recent studies have identified subsets of DG in children and young adults that lack alterations in IDH1, IDH2 or histone H3 genes. In addition to BRAF, mentioned above, genes that are recurrently altered in this group include v-Myb avian myeloblastosis viral oncogene homolog (MYB), v-myb avian myeloblastosis viral oncogene homolog-like 1 (MYBL1), and FGFR1. These tumors are technically IDH- and H3- WT, but in retrospective studies they seem to exhibit an indolent clinical course, especially when compared to most DG of adults.¹⁸ Genetic mechanisms of activation of these drivers are heterogeneous, which may affect the availability of specific assays for detection. These include duplications, truncations, amplifications and rearrangements resulting in fusion genes (MYB and MYBL1) and internal tandem duplications of tyrosine kinase domains and single nucleotide variants (FGFR1). The prevalence of these alterations is difficult to assert with certainty, since most of the early data was from small retrospective studies. Integrated molecular and clinical analysis of a large cohort of pediatric low grade gliomas provide insight into the spectrum and frequency of these alterations.¹⁸⁵ Specific genetic alterations seem to correlate with glioma subtypes, with MYBL1 rearrangements being reported in up to 50% pediatric diffuse astrocytoma,¹⁸⁴ MYB amplifications/copy number gains in 2/14 (14%) pediatric diffuse astrocytoma¹⁸³ and 33/166 (20%) adult IDH WT lower grade gliomas.¹⁸ Conversely, in one study FGFR1 alterations were frequent in rare low-grade neoplasms with ‘oligodendroglial’ phenotype affecting children and young adults, 39/42 (92.86%).² In addition, MYB-QKI fusions are found in nearly all angiocentric glioma (WHO grade 1), a low grade DG often associated with epilepsy.¹⁸⁶ Given the diagnostic and prognostic implications for these genetic drivers in gliomas of children and young adults, testing for them may be advisable depending on laboratory capabilities, but the quality of evidence is lower than that supporting testing for other more commonly altered genes (eg, IDH1/2 and BRAF).

Public comment response to Recommendation 13: There were 88 respondents, of whom 75% (n=66) agreed or agreed with modifications, 9.09% (n=8) disagreed, and 15.91% (n=14) were neutral. There were 19 written comments, including a number that suggested that evidence for the recommendation was not strong and the targets are not actionable; questions about testing methods; requests to clarify what is meant by “in the appropriate clinical setting” and inclusion of other genetic drivers (ie, panels). These comments were taken into consideration and the recommendation was revised (may instead of should) or addressed by the other recommendations (eg, testing for BRAF).

GOOD PRACTICE STATEMENTS

During the period of the systematic literature review and drafting of recommendations, several clinically relevant investigations were published on the topic of biomarker testing for DG subtypes not originally covered by the Patient/population, Intervention, Comparator, and Outcome (PICO) frameworks. In order to include guidance on these subtypes, we performed a focused literature review and included Good Practice Statements based on this second review. A description of the methods used to formulate these statements is included in the SDC.

The targeted search included alterations of ALK receptor tyrosine kinase (ALK), EGFR, ROS proto-oncogene 1 (ROS1), Neurotrophic tyrosine receptor kinase (NTRK), MET, PDGFRA, and MYCN proto-oncogene (MYCN) within bithalamic gliomas, infantile-type hemisphere gliomas, or diffuse pediatric-type high grade gliomas. The panel viewed the statements below as important, but there was not a formal rating for the certainty of evidence. The decision to form these statements was based on 1) emerging evidence with limited number of studies and samples and 2) anticipated updates to the WHO occurring at the time that this manuscript was being written. The panel believed that the tests would provide prognostic or treatment-related information for clinicians to consider. Future updates to the guideline will address these tests using a formal rating of the evidence and will include a grade for the strength of recommendation. Refer to Table 10 for the list of Good Practice Statements.

Table 10.

Summary of the Good Practice Statements (GPS)

GPS1. EGFR mutation testing may be performed on diffuse gliomas that involve the midline.

GPS2. Testing for alterations in ALK, ROS1, MET and NTRK genes may be performed on cerebral hemispheric diffuse gliomas of infancy that are wildtype for IDH and histone H3.

GPS3. Testing for DNA methylation class or for alterations in PDGFRA, EGFR, MYCN may be performed in pediatric high grade diffuse gliomas that are wildtype for IDH and histone H3.

Statement 1. EGFR mutation testing may be performed on diffuse gliomas that involve the midline.

A small subset of diffusely infiltrating midline gliomas that harbor EGFR mutations have a distinct methylation profile.^{174, 187} While most of these DG with EGFR mutations are histone H3-wildtype, some contain H3 K27M mutations. Similar to H3-K27 mutant gliomas involving the midline, EGFR-mutant DG in this location that are histone H3-wildtype typically show loss of H3 K27M trimethylation (H3K27me3), which can be detected by IHC. Small in-frame insertions/duplications of the tyrosine kinase domain and point mutations of the extracellular domain are detected by NGS and are potential therapeutic targets.

Statement 2. Testing for alterations in ALK, ROS1, MET and NTRK genes may be performed on cerebral hemispheric diffuse gliomas of infancy that are wildtype for IDH and histone H3.

DG that occur in infancy are usually molecularly distinct from those of older childhood and adults. These large cerebral hemispheric tumors are typically wildtype for IDH and histone H3 and often harbor fusions (rarely other activating alterations) involving ROS1, ALK, MET or the NTRK family.^{188, 189} The identification of such fusions or other activating alterations could lead to targeted therapy. Fusions can be identified by NGS and by FISH if appropriate probes are available.

Statement 3. Testing for DNA methylation class or for alterations in PDGFRA, EGFR, MYCN may be performed in pediatric high grade diffuse gliomas that are wildtype for IDH and histone H3.

Among pediatric diffuse high grade gliomas that are IDH-wildtype and histone H3-wildtype, three molecular subclasses can be identified by DNA methylation profiling that have differing molecular genetic alterations and clinical outcomes. One class is generally characterized by PDGFRA amplifications or mutations; a second class enriched for EGFR amplifications is associated with the longest survival; and the third often harbors MYCN amplifications and is associated with the shortest survival.^{164, 190}

FUTURE CONSIDERATIONS

The recommendations provided are based on evidence available at the time of literature review. It is expected that additional studies and new technologies will emerge that will need to be incorporated into future guidelines. In the category of IDH-WT diffuse astrocytic gliomas of adults, it has been recommended that histologic grade 2 and 3 DG should be tested for chromosome 7 gain/whole chromosome 10 loss, EGFR amplification, and TERT promoter mutation to establish the molecular diagnosis of GBM, grade 4. It might be expected that additional biomarkers or combinations of biomarkers will be identified and validated in histologic grade 2 and 3 DG that are also capable of predicting grade 4 clinical behavior, including alterations in PDGFRA, FGFR3, c-Met, TP53, NF1, PTEN, among others.¹⁰ For IDH-mutant astrocytomas of histologic grade 2 and 3, future studies could uncover and validate additional markers that can more optimally stratify risk or predict grade 4 clinical behavior in addition to CDKN2A/B homozygous deletion, such as PDGFRA, MYCN or CDK4 amplifications, mutations in CDKN2A/B or Rb1, among others.^{11, 65, 85} CDKN2A/B homozygous deletion have recently been shown to be a marker of poor prognosis in oligodendroglioma, IDH-mutant, 1p/19q codeleted, which along with other biomarkers, may be relevant to future recommendations or grading schemes.¹¹¹ Similarly, it is likely that genetic or epigenetic markers will be uncovered and validated that are capable of stratifying risk among patients with diffuse gliomas driven by histone H3, Mitogen-activated protein kinase (MAPK) pathway or MYB/MYBL1 genes.^{9, 191}

The laboratory findings of tumor mutational burden (TMB), mismatch repair (MMR) loss and microsatellite instability (MSI) are of interest in multiple forms of cancer, since the findings from these tests predict response to immunotherapies, including immune checkpoint inhibitors in some cancer types.^192-194 Initial randomized clinical trials of checkpoint inhibitors in newly diagnosed and recurrent glioblastoma have failed to demonstrate improved survival.¹⁹⁵ Guidelines for testing of TMB, MMR or MSI in DG will be developed if and when clinical care warrants them.

Whole genome DNA methylation profiling represents a robust and reproducible method for precisely segregating tumor types that have similar genetic alterations, and epigenetic signatures and clinical behaviors.^{5, 12} IDH-wildtype GBM, IDH-mutant astrocytomas, oligodendrogliomas, IDH-mutant, 1p/19q codeleted, and histone H3 mutant gliomas cluster tightly within their own class with little overlap based on DNA methylation profiling.^{5, 6, 22} Other DG, such as those that occur in children and are driven by MAPK pathway or MYB/MYBL1 genomic alteration also cluster together, indicating a high degree of specificity for each signature.^{164, 190} While the literature indicates this method is superior for classification purposes and could have a role in future classification and grading, it currently lacks widespread clinical implementation due to regulatory challenges, prevailing practice patterns and uncertainties in reimbursement. Currently, DNA methylation profiling is an attractive alternative for identifying DG classes that will likely be implemented into clinical practice in the near future.

LIMITATIONS

We were unable to answer KQs 1b. (What are the acceptable techniques/methods for mutation testing of DG and what are the expected turn-around-times for individual assays?) While we collected the various methods of testing, none performed poorly in our assessment. Each have varying strengths and weaknesses. For the latter question, no studies from the literature search reported turn-around time data and so, no recommendation was made. It is up to each laboratory to determine the best methodology and acceptable turn-around-times. Although the individual DG entities have prognostic characteristics that have been well defined in the literature,^2-7 a majority of the studies that comprised the evidence base for the recommendations correlated the reported outcomes to DG entities and not to survival outcomes. Additional studies that investigate outcomes related to biomarkers and DG subtypes would increase certainty of the evidence base and lead to stronger recommendations.

CONCLUSIONS

The 13 evidence-based recommendations and three good practice statements that are provided for biomarker testing of DG, together with their explanations and justifications, are intended to guide practice and improve the clinical care of patients with these diseases. For more than a century, the diagnosis of DG was based primarily on histologic appearance, yet recent clinically relevant molecular genetic discoveries have forced a re-evaluation of diagnostic definitions and criteria. We are now firmly within the molecular era in which integrated diagnoses are formulated by directly incorporating molecular biomarker test results. The complexity of establishing an integrated diagnosis has increased and pace of change has been rapid, warranting the recommendations that we have provided. Neuropathologists, molecular pathologists and treating physicians will continue to work closely together to provide optimal clinical care by utilizing molecular biomarkers for diagnostic and treatment purposes.

Appendix - Disclosed Interests and Activities from July 2017 to April 2021

Daniel J. Brat, MD, PhD

Position of Influence: Society for Neuro-Oncology, American Association of Neuropathologists

Julia A. Bridge, MD

Research Grant: Cepheid

Consulting fees or Advisory Board: Bayer, Bristol Myers Squibb/National Comprehensive Cancer Network, Epizyme, Merck, Moffitt Cancer Center, Translational Genomics Research Institute, VENTANA/Roche

Position of Influence: United States & Canadian Academy of Pathology

Howard Colman, MD, PhD

Consulting fees or Advisory Board: AbbVie, Adastra Pharmaceuticals, Bayer, Best Doctors, CytRx Corporation, F. Hoffman-La Roche, Forma Therapeutics, Foundation Medicine, Genentech, Insys Therapeutics, Karyopharm Therapeutics, NewLink Genetics, Novocure, Omniox, Orbus Therapeutics, OXiGENE, Private Health, Upsher-Smith

Speakers' bureau/lecture fees/honoraria: Merck

Institutional contracts/clinical trials: AbbVie, Array BioPharma, Bayer, BeiGene, Bristol Myers Squibb, DNAtrix, Forma Therapeutics, Global Coalition for Adaptive Research, Kadmon Holdings, Karyopharm Therapeutics, Merck, NewLink Genetics, Nuvation Bio, Orbus, Plexxikon,

Meera R. Hameed, MD

Consulting fees or Advisory Board: Novartis

Eyas M. Hattab, MD, MBA

Consulting fees or Advisory Board: Arbor Pharmaceuticals, Cell Marque, E.R. Squibb&Sons, Bristol Myers Squibb, Ventana Medical System/Roche Diagnostics

Speakers' bureau/lecture fees/honoraria: Arbor Pharmaceuticals, Roche Diagnostics

Position of Influence: College of American Pathologists, United States & Canadian Academy of Pathologists

Other: International Collaboration on Cancer Reporting member

Jason T. Huse, MD, PhD

Research Grants: Taiho Oncology, Inc.

Martin J. van den Bent, MD, PhD

Consulting fees or Advisory Board: AbbVie, Agios, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, CarThera, Celgene, Celldex Therapeutics, Genenta Science, Karyopharm Therapeutics, Nerviano Medical Sciences, VAXIMM

Speakers' bureau/lecture fees/honoraria: Merck Sharp & Dohme Corp

Research Grants: AbbVie

No Disclosures to Report

Arie Perry, MD; Brent T. Harris, MD, PhD; Michelle Hawks Yount, MS; Fausto J. Rodriguez, MD; Robert B. Jenkins, MD, PhD; William C. McDonald, MD; Kenneth Aldape, MD; Peter Canoll, MD, PhD; Dolores H. Lopez-Terrada, MD, PhD; Lesley H. Souter, PhD; Carol Colasacco, MLIS, SCT(ASCP); Nicole E. Thomas, MPH, CT(ASCP)^cm