Concordance of fluorescence in situ hybridization, whole-exome sequencing, and chromosomal microarray for evaluating high-risk molecular features in meningiomas

Abstract

Background

Meningiomas are the most common primary intracranial tumors, classified into three WHO grades based on histologic and/or molecular criteria. cIMPACT-NOW Update 8 underscores the increasing importance of molecular testing in grading and risk stratification, noting that morphologically WHO grade 1 tumors with concurrent 1p and 22q losses demonstrate clinical behavior comparable to WHO grade 2 meningiomas. This study aimed to assess the concordance of fluorescence in situ hybridization (FISH), chromosomal microarray analysis (CMA), and whole-exome sequencing (WES) in defining the molecular risk profile of meningiomas.

Methods

A total of 113 meningiomas were analyzed by WHO grade to assess the concordance of FISH and WES with CMA in detecting CDKN2A homozygous deletion and chromosomal arm losses involving 1p, 10q, 14q, and 22q.

Results

FISH demonstrated a lower concordance with CMA compared to WES. Some discrepancies likely reflect inherent technical limitations of probe-based assays. Nonetheless, FISH remains widely used as a first-line method and in some settings the only available method—particularly in resource-limited environments such as low- and middle-income countries.

Conclusion

Our findings suggest that FISH can provide clinically informative molecular risk assessment in select contexts where more comprehensive platforms are unavailable. However, due to discordance in detecting key copy number alterations, results should be interpreted cautiously. By contrast, CMA and WES offer more reliable alteration detection, making them the preferred platforms for precise molecular risk stratification and prognostication. Refining FISH probe design and establishing optimal thresholds for defining chromosomal arm losses may mitigate some discrepancies, although further validation is needed.

Key Points.

• Copy number alterations are increasingly recognized as critical for meningioma grading and prognosis.

• CMA and WES outperform FISH in detecting losses of 1p, 10q, 14q, and 22q.

• FISH, CMA, and WES show relatively consistent results for assessing CDKN2A homozygous deletion, though interplatform reliability is suboptimal

Importance of the Study

The cIMPACT-NOW Update 8 recommendations emphasize the role of chromosomal arm loss in assessing molecular risk, particularly for meningiomas classified as “grade 1 with worrisome features.” In our retrospective review of 113 meningiomas diagnosed between July 2024 and June 2025, chromosomal microarray analysis (CMA) and whole-exome sequencing (WES) outperformed fluorescence in situ hybridization (FISH) for detecting 1p, 10q, 14q, and 22q chromosomal arm losses. Detection of CDKN2A homozygous deletion showed suboptimal inter-platform reliability due to a small number of positive cases. Integration of molecular findings influenced the final integrated grade and, consequently, patient management in this cohort. These results have implications for the diagnosis and treatment of patients without access to comprehensive molecular testing and provide a framework for future studies to refine surrogate markers or widely available platforms for optimal evaluation of meningioma molecular risk in resource-limited settings, especially as these platforms vary in turnaround time and resolution.

Meningiomas are the most common primary intracranial tumors, with grading based on both morphological and molecular features. Certain molecular alterations, such as CDKN2A/CDKN2B homozygous deletion (HD) and oncogenic TERT promoter variants, can independently designate a tumor as grade 3 by the most recent World Health Organization classification of central nervous system (CNS) tumors (WHO CNS5), even in the absence of high-grade histologic features.¹ The cIMPACT-NOW Update 8 (cIMPACT-NOW8) further refines molecular risk assessment, recommending that morphologically WHO grade 1 meningiomas with chromosomal arm 1p deletion, combined with 22q loss and/or an NF2 oncogenic variant, be considered at least WHO grade 2 due to their more aggressive clinical behavior.² Recurrent losses involving 6p/q, 10q, 14q, and 18p/q have been observed in subsets of higher-grade or recurrent meningiomas; cIMPACT-NOW8 recommends integrating these alterations into molecular risk stratification, as they may co-occur with NF2/22q loss and reflect cumulative genomic instability. Another important update is the integration of molecular findings in brain-invasive, otherwise benign (BIOB) meningiomas, as cIMPACT-NOW8 purports that brain invasion alone is not sufficient to qualify for WHO grade 2. Loss of 1p remains the most frequently observed secondary chromosome arm alteration associated with recurrence, and additional combined alterations, such as 1p loss with 1q gain, are emerging as potential high-risk markers.³ Collectively, these observations highlight the utility of broader copy number variation (CNV) profiling to complement histopathology for more precise risk stratification, even as formal guideline criteria beyond 1p and 22q continue to evolve.

Thus far, no single platform is endorsed for assessing chromosomal arm status or molecular risk. Fluorescence in situ hybridization (FISH), chromosomal microarray (CMA), next-generation sequencing (including whole-exome sequencing [WES]), and DNA methylation profiling (DNAMP) offer differing resolutions: FISH targets specific loci, whereas CMA, WES, and DNAMP provide comprehensive chromosomal assessment. The threshold percentage of a chromosome arm that must be lost in meningioma cells to affect molecular risk varies across chromosomes; cIMPACT-NOW8 defines a loss threshold of 5% for 1p, but other arms lack standardized thresholds, contributing to inter-laboratory variability.⁴ Loss thresholds above 95% are associated with shorter recurrence-free intervals, emphasizing the importance of comprehensive arm assessment.^5,6 These findings underscore the need for more precise refinement of chromosomal arm loss thresholds to ensure accurate molecular risk evaluation.

In this study, we assessed the concordance of 1p, CDKN2A HD, 10q, 14q, and 22q losses across FISH, CMA, and WES in 113 meningioma biopsies or resections that were evaluated at Washington University School of Medicine between June 2024 and June 2025, and examined the effect of cIMPACT-NOW8 and WHO CNS5 criteria on the management of WHO grade 1 meningiomas with worrisome features. DNAMP was excluded from the evaluated platforms, as it is not a current component of meningioma diagnosis at our center. Other copy number changes, such as loss of 6p/q and 18p/q, were also excluded as these are currently not part of our in-house FISH panel.

Methods

Tissue Samples and Cases

113 cases with a diagnosis of meningioma, that had been made between June 1, 2024 and June 30, 2025, were retrieved from the Washington University Division of Neuropathology department archives. The primary inclusion criterion was a neuropathologic specimen obtained from surgery performed at our center, with a diagnosis of meningioma. Retrospective review of pathology reports included histologic and immunohistochemical findings, as well as cytogenetic and molecular testing results when available. Tumor grade was recorded based on the grade assigned at diagnosis according to WHO CNS5, or modified by cIMPACT-NOW Update 8 criteria if the report was issued after the latter’s publication. No additional testing was performed on these cases. FISH testing was conducted by the Clinical Genomics Laboratory at Washington University School of Medicine (St. Louis, MO), CMA studies were performed at Mayo Clinic Laboratories (Rochester, MN), and WES was performed by Caris Life Sciences (Irving, TX). FISH and CMA results were obtained directly from the provided reports, while WES results were determined by examination of the “e-karyotype” derived from copy number analysis in the whole-exome sequencing report. WES reports were also reviewed to determine the presence or absence of an NF2 variant, and whether said variants were pathogenic or of unknown significance.

Testing Platforms

FISH assays included the use of probes developed by Abbott Molecular (Des Plaines, Illinois, USA), with laboratory-derived cutoffs for percentage of nuclei with the investigated alteration; assays were performed on unstained formalin-fixed, paraffin-embedded (FFPE) slides. For every case submitted for FISH analysis, an area was circled on the chosen slide to represent the most optimal tumor purity. Enumeration probes localizing to 1p36 (spectrum orange, 435 kilobases [kb]) and 1q25 (control, spectrum green and 618 kb) were used to evaluate for 1p deletion with a cutoff of 15% for deletion and 28% for monosomy. A CDKN2A locus-specific probe (spectrum orange, 222 kb) paired with a CEP9 enumerating probe (spectrum green) was used to evaluate for CDKN2A homozygous deletion, with a cutoff of 16%. A PTEN (10q) enumeration probe (spectrum orange, 368 kb) and control CEP10 probe (spectrum green) were used to evaluate for 10q deletion with a cutoff of 26% for deletion and 25% for monosomy. An IGH break-apart probe (250 kb, 3′ orange and 5′ green) was used to evaluate for 14q deletion with a cutoff of 24%. An EWSR break-apart probe (497 kb, 5′ orange and 3′ green) was used to evaluate for loss of 22q, with a cutoff of 30%.

CMA was performed on unstained FFPE slides by the Mayo Clinic Laboratories (Rochester, MN), using molecular inversion probes on an OncoScan (Applied Biosystems [Affymetrix]) whole-genome array, with a theoretical functional resolution of 500 kb for non-mosaic deletions and duplications. However, in practice, the functional threshold varies based on the alteration of interest, with deletions having the lowest threshold of 1.0 megabases (Mb), duplications having a threshold of 2.0 Mb, and copy-neutral loss of heterozygosity having a threshold of 10.0 Mb. As further described by the Mayo Clinic Laboratories, data were analyzed and reported using the February 2009 human genome build. In cases submitted for analysis, detected copy number variants were reported with the affected chromosomal region(s) and the size in Mb of each variant. As discussed in recently published work concerning the association between molecular alterations and survival in patients with IDH-mutant astrocytomas, the Mayo Clinic Laboratories use a holistic combination of log₂ ratio, allelic differences, and changes in smooth signal to assess potential deletions, without hard-set log₂ cutoffs that are used in every case.⁷

Caris’ MI Cancer Seek assay requires at least 25 mm² of tissue (with at least 20% tumor cellularity) for testing, with our center’s practice being to send unstained FFPE slides rather than entire FFPE tissue blocks. Nucleic acids are extracted and quantified from submitted tissue with proprietary MI Cancer Seek Kits, with a minimum of 50 ng DNA necessary for testing, and 220 ng being optimal for library construction (which also includes the use of a Cancer Seek Kit). For library construction, the extracted nucleic acids are fragmented and used to construct cDNA molecules, which are chemically labeled to allow for DNA versus RNA origin differentiation. The library molecules are then amplified using a pre-capture polymerase chain reaction, along with MI Cancer Seek Kit-driven quantification. The library is enriched for genes of interest through hybridization with a custom panel of “baits” (each 120 nucleotides long, double-stranded, and biotinylated). The library is then pooled following quantity-based sample normalization and quantified, followed by use of the MI Cancer Seek Sequencing kit and sequencing on the Illumina NovaSeq 6000 platform. Sequencing results are analyzed and reported using proprietary software, with a physician generating a report that also identifies single nucleotide variants and insertion-deletions within 228 genes. This report also includes ERBB2 copy number, microsatellite stability, and tumor mutational burden status. The sequencing data are combined with interrogation of multiple single-nucleotide polymorphisms to assess copy number and provide an “e-karyotype” along with information on genes of interest. The threshold used to show loss on e-karyotype was less than 1.4 copies of the region of interest, which could include an entire arm or a focal segment. Our review and interpretation of each e-karyotype showing alterations to the chromosomes of interest was subjective, with chromosome arm loss called if the arm showed 20% or more overall loss (including small deletions).

Statistical Analyses

For all 113 cases, results for 1p, CDKN2A HD, 10q, 14q, and 22q losses obtained by FISH, CMA, and WES were compared to assess platform concordance. Each chromosomal arm was evaluated separately for discrepancies between FISH and CMA, and between CMA and WES. Cases lacking results for both platforms being compared were excluded from that specific analysis but retained for others (Figure 1). Concordance rate was based on the percentage of cases that were concordant upon evaluation for copy number status of a given arm by either FISH & CMA or WES & CMA, with 95% confidence intervals (CIs). Interplatform reliability and the concordance of NF2 status and 22q status on WES were also evaluated with Cohen’s kappa (κ); a weighted κ was used to evaluate the concordance of FISH and CMA.

Figure 1.

Breakdown of the study cohort according to the testing platforms performed—FISH, CMA, and/or WES—and the number of results obtained for each platform.

To assess the impact of the cIMPACT-NOW8 update on meningioma grading at our institution, tumors originally classified as CNS WHO grade 1 were examined for concurrent 1p and 22q loss on ancillary studies; cases with such losses were considered prognostically equivalent to CNS WHO grade 2, while cases without those losses were left as CNS WHO grade 1.

Finally, we evaluated whether the same tissue block was used for FISH & CMA and for CMA & sequencing. These data were analyzed alongside platform concordance to estimate the effect of block selection on concordance, using CMA as the reference and reporting percentage correlation with 95% CIs. Statistical analyses were performed using GraphPad Prism v10.6 (GraphPad Software LLC, Boston, MA).

Results

Compared to FISH, WES Demonstrated Greater Agreement With CMA, With Concordance Levels Differing Between Chromosomal Arms

Of the 113 cases within the cohort, 70 cases were evaluated with CMA, 46 were evaluated with WES, and anywhere from 22 (10q loss) to 38 (CDK2NA HD) cases were evaluated with FISH (Table 1). FISH showed less overall concordance with CMA for detecting chromosomal arm losses, with variability across arms (Figure 2), compared to the concordance between WES and CMA (Figure 3). For 1p deletion, concordance was 45.83% (n = 11; 95% CI: 27.89-64.93, κ = 0.16) between FISH & CMA, compared with 92.86% (n = 39; 95% CI: 80.99-97.54, κ = 0.85) for WES & CMA. For CDKN2A HD, FISH & CMA concordance was 96.43% (n = 27; 95% CI: 82.29-99.82, κ = 0.00), closely matching WES & CMA (100%; n = 42; 95% CI: 91.62-100.00, κ not evaluable), though there was less inter-platform reliability compared to that in the four arms of interest. For 10q loss, FISH & CMA concordance was 15.00% (n = 20; 95% CI: 5.24-36.04, κ = 0.012) versus 95.24% (n = 42; 95% CI: 84.21-99.15, κ = 0.48) for WES & CMA. Concordance for 14q loss was 40.0% (n = 20; 95% CI: 21.89-61.34, κ = 0.056) for FISH & CMA and 88.10% (n = 42; 95% CI: 75.00-94.81, κ = 0.79) for WES & CMA. For 22q loss, FISH & CMA concordance was 62.50% (n = 24; 95% CI: 42.71-78.84, κ = 0.21), compared with 85.71% (n = 42; 95% CI: 72.16-93.28, κ = 0.76) for WES & CMA. These findings indicate that FISH, while appearing to agree with CMA results (limited by low κ) for CDKN2A HD, shows lower concordance than CMA & WES for whole chromosomal arm losses, underscoring its limitations for comprehensive molecular risk assessment. When separating tumors by CNS5 WHO grade, grade 1 tumors were submitted for CMA most often (26 cases submitted), with 8 submitted for WES and variable numbers (5-9 depending on the specific test) submitted for FISH. 40 grade 2 tumors were submitted for CMA, 38 for WES, and smaller numbers (ranging from 15 to 24) were submitted for FISH. Grade 3 tumors were submitted equally (2 cases each) to CMA and WES, with 3 of the 4 cases submitted for CDKN2A HD FISH.

Table 1.

Results for evaluation of 1p loss, CDKN2A homozygous deletion/9p loss, 10q loss, 14q loss, and 22q loss by FISH, CMA, and WES, showing the number of cases with and without an alteration detected by each platform

Alteration	FISH		CMA		WES
	Positive (n, %)	Negative	Positive	Negative	Positive	Negative
		(n, %)	(n, %)	(n, %)	(n, %)	(n, %)
1p loss	22	6	30	40	26	23
	(78.57%)	(21.43%)	(42.86%)	(57.14%)	(53.06%)	(46.94%)
CDKN2A homozygous deletion/9p loss	1	37	0	70 (100.00%)	2	44
	(2.63%)	(97.37%)	(0.00%)		(4.17%)	(95.83%)
10q loss	20	2	4	66	3	46
	(90.91%)	(9.09%)	(5.71%)	(94.29%)	(6.25%)	(93.75%)
14q loss	21	1	22	48	17	32
	(95.45%)	(4.55%)	(31.43%)	(68.57%)	(34.69%)	(65.31%)
22q loss	24	4	44	26	27	22
	(85.71%)	(14.29%)	(62.86%)	(37.14%)	(55.10%)	(44.90%)

Figure 2.

Concordance between FISH and CMA in evaluation of 1p loss (n = 24), CDKN2A homozygous deletion (n = 28), 10q loss (n = 20), 14q loss (n = 20), and 22q loss (n = 24).

Figure 3.

Concordance between WES and CMA in the evaluation of 1p loss, CDKN2A homozygous deletion, 10q loss, 14q loss, and 22q loss (n = 43).

WES Showed Substantial Co-occurrence of 22q Loss and NF2 Variants

Of the cases submitted for WES, 22 showed a variant in NF2, with one case showing a variant of uncertain significance and the other 21 cases showing a pathogenic variant. 95.45% (n = 21) of the NF2 variant cases showed loss of 22q in their e-karyotype, and 74.07% (n = 20) of the NF2-wildtype cases showed an intact 22q on e-karyotype (κ = 0.711). One grade 2 meningioma was submitted for WES, but while testing was able to determine NF2-wildtype status, the sample failed e-karyotype quality control, so its 22q status could not be determined.

cIMPACT-NOW8 Impacts Treatment When Oncologists Are Aware of a Tumor’s Risk Profile

Based on WHO CNS5 criteria, 51.3% of the 113 meningiomas (n = 58) were classified as grade 1 (Figure 4), 41.6% (n = 47) as grade 2, and 3.5% (n = 4) as grade 3. One grade 3 tumor was classified on the basis of CDKN2A HD, while two showed over 20 mitoses in 10 high-power fields, and the fourth represented a resection of a recurrent grade 3 meningioma, which had been diagnosed based on a prior elevated mitotic count in the initial resection (not seen in the more recent resection). No cases showed a pathogenic TERTp mutation on WES. The remaining 3.5% (n = 4) of cases could not be assigned a WHO CNS5 grade due to a paucity of evaluable tumor (often in the setting of limited overall cellularity or significant cautery artifact). Upon reevaluation of the 58 WHO CNS5 grade 1 tumors using cIMPACT-NOW 8 recommendations, 20.69% (n = 12) could be upgraded to WHO CNS grade 2 (Figure 4). Of those 12 cases, two showed concurrent 1p & 22q loss on both CMA and WES, seven only showed these on CMA with no submission for WES, and three were upgraded based on concurrent losses on FISH that were not seen on CMA or WES. We also evaluated these “upgraded” cases via chart review to assess clinical outcome: chiefly, whether the patient was treated with adjuvant radiotherapy and whether their follow-up imaging showed evidence of recurrent tumor. One case represented a partial resection (a subsequent resection shortly after showed a grade 3 meningioma and was treated accordingly with adjuvant radiation), and adjuvant radiotherapy had been incorporated into the treatment of four other cases; three of these cases’ molecular features (ie, concurrent 1p and 22q loss) were explicitly incorporated into their oncologic treatment plans. Of the remaining cases, six were treated with surgical resection, with no evidence of recurrence on follow-up imaging. The final case represented a subtotal resection in a patient with schwannomatosis and multiple meningiomas. These data highlight cIMPACT-NOW8’s relevance for grade 1 meningiomas with worrisome features.

Figure 4.

(A) Meningioma distribution by CNS5 WHO grade and (B) the proportion of grade 1 tumors reclassified as grade 2 by cIMPACT-NOW Update 8 recommendations. These graphs show that the majority of tumors were grade 1 based on guidelines at the time of diagnosis, with a subset being grade 2 by cIMPACT-NOW8.

Using the Same Tissue Block for Ancillary Testing Has an Equivocal Effect on Test Concordance

We examined whether using the same FFPE tissue block influenced concordance between platforms (Figure 5). For FISH and CMA, concordance was 37.5% (n = 12; 95% CI: 22.9-54.8) when the same block was used, compared with 50.0% (n = 4; 95% CI: 21.5-78.5) when different blocks were analyzed. For WES and CMA, concordance was 74.1% (n = 20; 95% CI: 55.3-86.8) with the same block and 84.2% (n = 16; 95% CI: 62.4-94.5) when different blocks were used. These findings suggest that block choice is not a significant factor in concordance, though this may be an artifact of small sample size.

Figure 5.

Impact of tissue block selection on test concordance. This graph shows cases with discrepancies between CMA and FISH, and CMA and WES, highlighting whether discordance occurred when ancillary tests were performed on the same versus different tissue blocks.

Discussion

The 2021 WHO CNS5 classification has increasingly emphasized the role of molecular features in meningioma grading. In particular, CDKN2A/B HD can upgrade an otherwise grade 1 or 2 tumor to grade 3, while the standalone prognostic significance of certain morphologic subtypes, such as papillary or rhabdoid variants, has been de-emphasized.^1,8 Recent evidence also clarifies that brain invasion alone is insufficient to assign a grade 2 designation. BIOB grade 1 tumors should only be considered prognostically equivalent to grade 2 when certain molecular criteria are present, such as the presence of combined 1p deletion and 22q loss/oncogenic NF2 variants.^2,9,10 These molecular features may independently upgrade tumors to grade 2 and highlight the critical importance of accurate molecular risk assessment, as a grade 2 diagnosis directly influences clinical management, including consideration of adjuvant radiation therapy.^2,3,11

In our cohort, comparison of FISH and WES with CMA indicated that WES demonstrates higher concordance with CMA than FISH for most chromosomal arm losses. Concordance varied between modality comparisons: CDKN2A HD exhibited the smallest discrepancy (though this may be an artifact of the small number of positive cases across all platforms), whereas 10q loss showed the largest. Using κ as an indicator for inter-platform reliability, with the exclusion of CDKN2A HD, CMA also showed a consistently stronger level of reliability with WES than it did with FISH (even with the use of weighted κ to account for the limited sample size of many FISH assays). These findings suggest that FISH may be reliable for detecting CDKN2A HD but less robust for assessing broader chromosomal alterations that contribute to molecular risk.

The limitations of FISH are both technical and interpretive. Probes are designed for specific loci and do not capture entire chromosomal arms, reducing sensitivity for partial arm losses. Interpretation is also subjective, introducing potential inter-observer variability. Standardized thresholds for calling chromosomal arm losses remain incomplete; while cIMPACT-NOW8 defines a >5% arm loss threshold for 1p loss in tumor cells to affect molecular risk, consensus thresholds for other arms are not established.^5,6 If a probe covers too little of a chromosomal arm, testing could miss significant deletions that fall outside of its coverage area, while an excessively large probe could lead to overcalling. Prior investigation of CDKN2A probe design suggests that a shorter probe is more effective in detecting CDKN2A HD, which sometimes arises in the setting of a microdeletion.¹² These factors can result in false-positive or false-negative results, most commonly false positives, highlighting the importance of confirming FISH findings with CMA, WES, or DNAMP when feasible. Although optimizing FISH probe design may reduce some discrepancies, intrinsic platform limitations—including locus-specific coverage and interpretive subjectivity—are likely to persist.

These findings have direct clinical implications, particularly in resource-constrained settings. Patients treated at tertiary care centers may have access to CMA, WES, or DNAMP, whereas those in low- and middle-income countries, rural regions, or resource-limited institutions may only have FISH available. In this context, FISH may provide a reliable baseline assessment for CDKN2A HD, though reliance on FISH alone may not detect additional chromosomal alterations that could molecularly upgrade histologic grade 1 tumors to grade 2. Studies of ancillary evaluation of CDKN2A HD and other molecular alterations in gliomas have also demonstrated that, depending on the size of the probe relative to the region of interest, FISH can have an intrinsic increased risk of false-positive or false-negative results that necessitate optimal tailoring of probe design.^12,13 MTAP and p16 immunostaining have also emerged as potential supplements or replacements to FISH testing, though they present similar limitations (specifically, MTAP immunostaining could give a false negative result if a deletion does not extend beyond CDKN2A) and are not as widely adopted as molecular testing for meningioma diagnosis.^13,14 Even in well-resourced settings, CMA and WES are often performed at reference laboratories, introducing delays compared with the more rapid turnaround of FISH assays. Improving FISH concordance with CMA or other comprehensive platforms could enhance diagnostic efficiency and broaden access to actionable molecular data, while recognizing that certain methodological limitations of FISH are unlikely to be fully mitigated.

Several biologic factors likely play a role in affecting test concordance, alongside technical factors. Even when tissue is microdissected for testing, the tumor cells are admixed with “normal” elements such as brain parenchyma, fibrovascular tissue, and immune cells, suggesting that the most careful selection of tissue for molecular testing may still leave room for signal interference.¹⁵ The selection of an appropriate block is also dependent on the tumor’s overall architecture: a tumor with an unusual morphology (or concurrent morphologic patterns) may show markedly different results on FISH and other testing depending on which portion of the tumor is sampled. This emphasizes the potential for pathologists’ choice of tissue for ancillary testing to bias results, though, as described, this would be better evaluated in future studies with a larger cohort or with prospective, more controlled test ordering.

The cases evaluated with WES also showed substantial reliability between the assessment of NF2 variation and the assessment of 22q loss. While WES is the only platform among the three evaluated that can show both multigene mutation status and overall copy number, it also carries the greatest potential cost for patients. While a thorough cost analysis is beyond the scope of this paper, a review of fee schedules suggests that WES pricing can exceed twice that of CMA, and without insurance coverage, could be catastrophically costly to many patients. Nevertheless, it remains a staple of oncologic care, and many laboratories offer payment plans and other mechanisms to mitigate financial burden, though these may not be available to patients in all geographic areas.

Reevaluation of WHO CNS5 grade 1 tumors using cIMPACT-NOW8 criteria revealed that approximately 21% of cases were molecularly equivalent to grade 2 (with the caveat that one grade 1 tumor was re-resected shortly thereafter, and the second specimen was classified as grade 3). With a quarter of the upgraded cases receiving adjuvant radiation, and two of those showing no recurrent tumor on surveillance imaging, cIMPACT-NOW8 has a tangible impact on patient treatment. This underscores the potential utility of cIMPACT-NOW8-guided molecular testing to identify higher-risk tumors that may benefit from closer surveillance or tailored therapeutic strategies and reinforces the importance of close and clear communication between pathologists and members of the patients’ care team so that oncologists, surgeons, and other team members have the most complete picture of a tumor’s molecular risk. When available, CMA or WES should be incorporated for comprehensive assessment, whereas FISH may serve as an interim or baseline tool.

This study has limitations that affect its generalizability and emphasize the need for continued longitudinal research. It is a retrospective, single-institution analysis, which may limit generalizability to other practice settings with different laboratory methodologies or testing algorithms. Not all tumors underwent all three testing platforms, resulting in variable sample sizes across concordance analyses and relatively small subgroups for certain comparisons. At the time of project initialization, our standard practice was to submit all morphologic grade 2 or 3 meningiomas for CMA and WES (often with CDKN2A HD FISH being ordered as well), expanding to include grade 1 tumors with worrisome features following the publication of cIMPACT-NOW8. The variation in FISH sample size can be partially attributed to individual pathologists’ preferences for ancillary testing: depending on the pathologist, FISH was sometimes omitted in favor of ordering only CMA and/or WES, especially in cases that were already grade 2 by morphology. While some grade 1 tumors with worrisome features were still evaluated with FISH, grade 1 tumors without these features were not often evaluated with any ancillary testing. While this was done to optimize the use of hospital resources, a consequence was the relatively low sample size for comparing FISH and CMA, which presents several avenues for future work. First, the inclusion period could be widened to simply review a much larger number of cases—this could also capture more grade 3 meningiomas, which are significantly more rare than grade 1 or grade 2 tumors. Additional studies could also draw on the findings from this initial study and prospectively order FISH on future cases, to validate these findings as the understanding of call thresholds becomes more nuanced. CMA was used as the reference comparator, although it is not an established gold standard, and platform-specific analytical approaches may influence copy number calls; additionally, microarray copy number plots are not provided to us by the reference laboratory, precluding comparison to e-karyotypes for evaluation of potential false negatives/positives when comparing WES and CMA. For example, just as an overly large FISH probe may miss a microdeletion that could impart significant molecular risk, CMA’s effective resolution could lead to the missed detection of copy number changes that do not meet its threshold. The methodology for assessing arm loss on e-karyotype was purely subjective and binary, classifying both focal and whole-arm losses as arm loss. Just as knowledge of ideal call thresholds can inform optimal FISH probe design, so can it make e-karyotype evaluation less subjective. Also, while it is featured in cIMPACT-NOW8 guidelines as a reliable measure of molecular risk, DNAMP was not included in this study as it is not currently a component of routine meningioma diagnosis at our institution. While it is often used to diagnose morphologically unusual meningiomas, DNAMP can also stratify molecular risk by placing a meningioma into one of several subgroups.¹⁶ These subgroups are associated with certain molecular alterations—as one might expect, tumors with TERTp or NF2 pathogenic variants tend to fall in the more aggressive subgroup, while tumors without those variants are usually classified within the more benign subgroup.2,¹⁶ However, similar to MTAP immunostaining, DNAMP is not routinely used in many centers for meningioma workup, though as it becomes more ubiquitous, this should resolve. While cIMPACT-NOW8 does include loss of multiple arms, not all chromosomal arms were assessed in our panel, with 6p/q and 18p/q excluded from analysis, which may affect comprehensive risk evaluation as more data emerge on the exact impact of each chromosomal arm on molecular risk. Variability in chromosomal arm loss thresholds across platforms and laboratories may also have contributed to discordance and underscores the need for standardization. While evaluation of the association between block selection and test concordance was not the main aim of this study, this evaluation is limited in scope by the sample size of the study. Nonetheless, when selecting a tissue block for ancillary studies, the ideal block should have maximum viable tumor cellularity and an accurate representation of overall tumor architecture. Prospective studies with larger sample sizes, with block selection being actively randomized to either the same or different blocks between tests, could help to provide future data.

Despite these limitations, we believe the current study remains informative, highlighting both inter-platform discordance and fundamental limitations of widely used first-line testing strategies, particularly FISH, in risk stratification. While some of these limitations could be addressed with improved probe design, there will always be a level of variability due to underlying biologic factors that manifest once a tissue block is selected for testing.

In conclusion, WES demonstrated higher concordance with CMA than FISH for most chromosomal arm losses, with CDKN2A HD/9p loss being an exception due to limited sample size. Applying cIMPACT-NOW8 criteria upgraded a subset of WHO CNS5 grade 1 meningiomas to molecular grade 2 designation, underscoring the potential clinical value of comprehensive molecular assessment. FISH remains a practical baseline tool, particularly for CDKN2A HD, but its locus-specific coverage and interpretive variability highlight inherent limitations. Where feasible, CMA or WES (or DNAMP) should complement FISH to provide a more complete molecular risk profile. These findings underscore the practical challenges of implementing molecular testing in resource-constrained settings and highlight the importance of developing accessible and cost-conscious strategies. They also point to the need for further studies to better define platform concordance, refine chromosomal loss call thresholds, and assess the reliability and applicability of molecular testing across diverse clinical and resource contexts.

Author Contributions

Project conceptualization: S.D. Case identification and review: B.C., E.R.-G., S.D. Statistical analysis: B.C. Manuscript drafting: B.C., S.D. Manuscript editing: B.C., E.R.-G., S.D.

Conflict of Interest Statement

The authors declare they have no conflicts of interest related to the content of this article.

Funding

The authors declare that they have received no external funding for this study.

Ethics Approval Statement

This project was approved by the Washington University Institutional Review Board.

Data Availability

Most data that support the findings of this study are available from the corresponding author upon reasonable request; however, WES data are proprietary to Caris and are not available.