Received: 9 October 2021 Revised: 4 January 2022 Accepted: 17 January 2022 DOI: 10.1111/vco.12802 ORIGINAL ARTICLE Standardisation of canine meningioma grading: Inter-observer agreement and recommendations for reproducible histopathologic criteria Sara Belluco1 | Giuseppe Marano2 | Kerstin Baiker3 | Andreas Beineke4 | Anna Oevermann5 | Frauke Seehusen6 | Patrizia Boracchi2 | Marti Pumarola7 | Maria Teresa Mandara8 1 Université de Lyon, VetAgro Sup, ICE UPSP 2016.A104, Axe Cancérologie, Marcy l'Etoile, Abstract Lyon, France The human grading system is currently applied to canine meningioma, although it has 2 Department of Biomedical and Clinical not been validated in dogs. The present study focused on standardising the human Sciences “L. Sacco”, University of Milan, Milan, Italy grading system applied to canine meningioma. Four veterinary neuropathologists 3 School of Veterinary Medicine and Science, graded 186 canine meningiomas as follows: Grade I tumour, with <4 mitoses/ University of Nottingham, Sutton Bonington Campus, Sutton Bonington, UK 2.37 mm2; Grade II tumour, with ≥4 mitoses/2.37 mm2, brain invasion or at least 4 Stiftung Tierärztliche Hochschule Hannover, three of the following criteria: sheeting architecture, hypercellularity, small cells, Institut für Pathologie, Hannover, Germany macronucleoli, necrosis; Grade III tumour, with ≥20 mitoses/2.37 mm2 or anaplasia. 5 Division of Neurological Sciences, Vetsuisse Faculty, University of Bern, Bern, Switzerland Slides with grading disagreement were reviewed to define a consensus diagnosis and 6 FTA Pathologie, Universität Zürich, Vetsuisse- to assess reproducible criteria. Concordance between histologic grade and the con- Fakultät, Institut für Veterinärpathologie, sensus diagnosis, as well as intra- and inter-observer agreements for each criterion, Zürich, Switzerland 7 were statistically analysed. Concordance between histologic grade and consensus Department Medicina i Cirurgia, Animals, Facultat de Veterinària, Campus UAB, diagnosis ranged from 59% to 100%, with lower concordance for Grade I and II Barcelona, Spain tumours. The lowest inter-observer agreement was recorded for macronucleoli, small 8 Laboratorio di Neuropatologia, Dipartimento di Medicina Veterinaria, Università degli Studi cells, hypercellularity and sheeting architecture. Tumour invasion and necrosis di Perugia, Perugia, Italy displayed fair agreement, while moderate agreement was reached for mitotic grade Correspondence and anaplasia. The following recommendations were issued to improve the reproduc- Sara Belluco, Laboratoire d'histopathologie ibility of canine meningioma grading: (1) Assess mitotic grade in consecutive HPFs Vétérinaire, VetAgro Sup, Campus Vétérinaire, 1, avenue Bourgelat, 69280 Marcy l'étoile, within the most mitotically active area; (2) Define invasion as neoplastic protrusions France. within central nervous tissue without pial lining; (3) Report spontaneous necrosis; Email: sara.belluco@vetagro-sup.fr (4) Report prominent nucleoli when visible at 100; (5) Report pattern loss Funding information when visible at 100 in >50% of the tumour; (6) Report necrosis, small cells, The authors disclosed the following financial support for the research, authorship, and/or hypercellularity and macronucleoli, even when focal; (7) Report anaplasia if multifocal. publication of this article: Impulsion Santé Globale 2019 grant from VetAgro Sup, Lyon. KEYWORDS central nervous system, dog, grading, meningioma, standardisation Sara Belluco and Giuseppe Marano contributed equally to this study. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2022 The Authors. Veterinary and Comparative Oncology published by John Wiley & Sons Ltd. Vet Comp Oncol. 2022;20:509–520. wileyonlinelibrary.com/journal/vco 509 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 510 BELLUCO ET AL. 1 | I N T RO DU CT I O N a track record of training and employment in neuropathology labora- tories and with extensive publication in the field (from now on defined Canine meningioma is one of the most commonly diagnosed tumours in as neuropathologists). The previous consensus on histologic criteria the central nervous system, representing up to 51.5% of intracranial was not assessed. The neuropathologists defined the tumour subtype, tumours diagnosed at necropsy.1,2 Intracranial and spinal meningiomas pro- the mitotic grade and the histologic grade for each case. The mitotic duce a variety of clinical signs including seizures, altered mentation, ataxia grade, purely based on the number of detected mitoses, was calcu- 1,3–5 and paresis. The median survival time is more than 1 year following lated as follows: mitotic grade 1: tumours with <4 mitoses in surgical resection and increases if adjunctive radiotherapy is applied.4,6–8 2.37mm2; mitotic grade 2: tumours with a mitotic count between Due to the striking pathological, immunological and MRI similari- 4 and 20 in 2.37mm2; mitotic grade 3: tumours with ≥20 mitoses in ties between human and canine meningiomas,9 and since the veteri- 2.37mm2. The histologic grade, based on cellular morphology, archi- 10 nary WHO classification is outdated, canine meningioma can be tecture and mitotic count, was evaluated as follows: Grade I: tumours classified into three grades according to the 2016 WHO human histo- lacking histologic criteria of Grade II and Grade III; Grade II: tumours logical grading system.11–15 Currently, no study has addressed the with a mitotic grade of 2 or tumours with central nervous tissue inva- translatability/transferability of the human grading system to the sion or tumours displaying at least three of the following criteria: canine patient, in terms of accuracy and reproducibility. sheeting architecture, small cells, hypercellularity, macronucleoli and In order to standardise a grading system for canine meningioma, spontaneous necrosis; Grade III: tumours with extreme anaplasia the specific goals of this study are: (1) to evaluate veterinary neuropa- and/or tumours with a mitotic grade of 3. Since canine meningioma thologists' inter-observer agreement when applying the human grad- subtypes have not been definitely correlated to prognosis, the tumour ing system to canine meningioma; (2) to evaluate the concordance of subtype was not considered for the histologic grade. each neuropathologist's grading to a consensus diagnosis; (3) to evalu- For each slide, the tumour area of the largest sample fragment ate the reproducibility of each evaluated histologic criterion, in order was calculated by each pathologist, using NDP view.2 free software. to identify its impact in creating histologic grade disagreement; The consensus diagnosis for the histologic grade and the mitotic (4) to propose useful amendments to the grading applied to canine grade was obtained by reviewing the cases with histologic disagree- meningioma in order to increase reproducibility. ment by all the participants at roundtables. Consensus diagnosis was defined as the diagnosis given by at least three out of four patholo- gists (majority consensus diagnosis). 2 | MATERIAL AND METHODS 2.1 | Samples 2.3 | Statistical analysis The minimum sample size was calculated using a 95% confidence interval According to the methods used to calculate sample size,16 the inter- for the Kappa index in the presence of multiple raters and multinomial observer and the intra-observer agreements were estimated using outcomes, with a precision of 10%.16 The expected a priori agreement respectively the Fleiss' Kappa index and the Cohen's Kappa coeffi- was considered 90%17; the number of veterinary neuropathologists cient. The concordance correlation coefficient (CCCTOT), accounting (defined as observers) was five and the expected frequencies, based on both inter- and intra-observer agreement, the CCCINTER, measuring the relative frequencies of meningioma in the laboratory archive of the the inter-observer agreement and CCCINTRA, measuring intra-observer participants, were 54%, 37%, 9% for Grade I, II and III, respectively. agreement, were calculated. For each of the three indices, a unilateral The protocol was reviewed and approved by the ethical commit- 97.5% lower confidence limit was calculated. tee of VetAgro Sup. For each grade, a representative number of For histologic and mitotic grades, the concordance between each paraffin-embedded samples of canine meningioma was selected by observer's second reading classification and the consensus diagnosis two veterinary pathologists with over 20 years of expertise in neuro- was calculated by evaluating the accuracy, quantifying the number pathology and retrieved from their laboratory archive. For each and respective percentage of the observer's classifications that mat- tumour, a 4-μm-thick section was stained with haematoxylin and ched the consensus diagnosis. eosin. The slides were digitised with an NDP scanner (NDP scan In order to evaluate the possible impact of the histologically avail- 2.5.90, Nanozoomer HT, Hamamatsu) with a magnification of 20 able tumour size on agreement about the histologic and mitotic grade, (454 nm/pixel) and visualised with the free NDP.2 viewer (NDP.view2 estimates of CCCTOT, CCCINTER and CCCINTRA were calculated sepa- Viewing software U12388-01jHamamatsu Photonics). rately for samples measuring arbitrarily less than or equal to 10 mm2 (small samples) and for samples of more than 10 mm2 (regular samples). 2.2 | Histologic evaluation of the tumours Analyses were performed using R software version 4.0.418 with the Agreement19 and irrCAC20 packages added, and the Knime Ana- Each case was analysed twice, with an interval of at least 2 weeks lytics Platform release 4.2.3.21,22 For κ and CCC-based agreement between analyses, by four board-certified veterinary pathologists with evaluation, the classification given by Landis and Koch23 was used: κ 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License BELLUCO ET AL. 511 from 0.00 to 0.20 for slight agreement; from 0.21 to 0.40 for fair 3.3 | Concordance with consensus diagnosis agreement; from 0.41 to 0.60 for moderate agreement; from 0.61 to 0.80 for substantial agreement; and over 0.80 for almost perfect In the slide review process, majority consensus was achieved on the agreement. histologic grade in 156/186 samples (83.9%) and on mitotic grade in 135/186 samples (72.6%). For the histologic grade, 73/156 (51.7%), 72/156 (41.1%) and 11/151 (7.3%) samples were classified as Grade I, 3 | RESULTS II and III, respectively. For the mitotic grade, 92/135 (68.1%), 34/135 (34.8%) and 9/135 (6.7%) were classified as Grade I, II and III, respec- 3.1 | Samples tively. Each neuropathologist's concordances of histologic grade to the consensus diagnosis at the second reading are reported in The calculated minimum sample size was 168, which was increased Table 2. For histologic grade, the total accuracy ranged from 72.4% to up to 186 samples (+10%) to account for potential sample losses 90.3%: the highest accuracy was found for Grade III, with agreement related to technical issues. Since one of the five neuropathologists ranging from 9/11 (81.8%) to 11/11 (100.0%) of samples, while the previously recruited for slide evaluation was involved in selecting the percentages of concordant diagnosis ranged from 58.9% to 90.4% for tumour samples, he was excluded from the observer group. The confi- Grade I and from 73.6% to 93.1% for Grade II. dence interval precision for the Fleiss' Kappa recalculated on four Similar results were obtained for the mitotic grade (Table 3). The observers was 10.8%, which was considered a negligible loss due to total accuracy between a neuropathologist's second reading and the the reduction in number of observers. consensus diagnosis ranged from 82.2% to 95.6%. The highest con- cordance was recorded for Grade III, with the percentage of concor- dant diagnosis close to or equal to 100.0% (except for observer C). 3.2 | Inter-observer agreement For Grade I, the percentage of agreement ranged from 85.9% to 97.8%, while for Grade II the range was from 67.6% to 88.2%. For histologic grade, the inter-observer agreement was moderate at the first and second reading (first reading: κ = 0.52, 95% CI: 0.42– 0.61; second reading κ = 0.52, 95% CI: 0.42–0.61). At the first read- 3.4 | Histologic criteria ing, 66/186 samples (35.9%) received the same histologic grade by the four neuropathologists: 26 of these were classified as Grade I, Considering each histologic evaluated criterion, the agreement was 33 as Grade II and 7 as Grade III. At the second reading, 73/186 sam- moderate for mitotic grade and extreme anaplasia (CCCTOT = 0.51 for ples (39.7%) received the same histologic grade by the 4 neuropathol- both) and fair for tumour invasion and spontaneous necrosis ogists; 29 of these were classified as Grade I, 36 as Grade II and 8 as (CCCTOT = 0.43 and 0.44, respectively) (Table 1). For each of these Grade III. More details are provided in Table S1. four criteria, the intra-observer agreement was almost perfect For histologic grade, CCCTOT accounting for intra- and inter- (CCCINTRA ranging from 0.86 to 0.93), while the inter-observer agree- observer agreement was 0.52 (one-sided 97.5% CI: 0.42–1) indicating ment was moderate (CCCINTER ranging from 0.44 to 0.55). moderate agreement (Table 1). The CCCINTRA was almost perfect The agreement was slight to fair for macronucleoli, small cells, (0.91, one-sided 97.5% CI, 0.88–1), while the CCCINTER was moderate hypercellularity and sheeting architecture (CCCTOT ranging from 0.18 (0.54, one-sided 97.5% CI, 0.45–1). to 0.34) (Table 1). For these histologic criteria, intra-observer TABLE 1 Estimates of the concordance correlation coefficients for histologic grade and histologic criteria Histologic Mitotic Tumour Macro Hyper Sheeting Extreme Index grade grade invasion Necrosis nucleoli Small cells cellularity architecture anaplasia CCCTOT 0.52 (0.42) 0.51 (0.40) 0.43 (0.33) 0.44 (0.36) 0.34 (0.27) 0.23 (0.15) 0.18 (0.13) 0.30 (0.23) 0.51 (0.36) CCCPREC 0.52 (0.43) 0.52 (0.41) 0.47 (0.38) 0.45 (0.37) 0.38 (0.32) 0.25 (0.17) 0.27 (0.22) 0.36 (0.29) 0.53 (0.40) χ 0.99 (0.97) 0.98 (0.96) 0.91 (0.86) 0.98 (0.96) 0.88 (0.84) 0.91 (0.85) 0.66 (0.59) 0.83 (0.76) 0.97 (0.91) CCCINTER 0.54 (0.45) 0.55 (0.44) 0.44 (0.34) 0.47 (0.39) 0.37 (0.29) 0.26 (0.17) 0.19 (0.14) 0.33 (0.25) 0.53 (0.39) CCCPREC 0.55 (0.46) 0.56 (0.45) 0.49 (0.40) 0.48 (0.40) 0.42 (0.35) 0.29 (0.20) 0.31 (0.25) 0.40 (0.33) 0.55 (0.42) χ 0.99 (0.96) 0.98 (0.95) 0.90 (0.86) 0.98 (0.95) 0.87 (0.83) 0.90 (0.83) 0.63 (0.56) 0.81 (0.74) 0.96 (0.90) CCCINTRA 0.91 (0.88) 0.87 (0.83) 0.93 (0.90) 0.86 (0.82) 0.83 (0.78) 0.70 (0.65) 0.77 (0.71) 0.78 (0.73) 0.90 (0.84) Note: Estimates of the total, inter-observer and intra-observer concordance correlation coefficients and respective sub-indices (precision and accuracy) for each histologic parameter are reported. The lower boundary of the 97.5% one-sided confidence interval is included within parentheses. CCCTOT, total concordance correlation coefficient considering the intra and inter-observer agreement; CCCPREC, precision sub-index of the concordance correlation coefficient; χ, accuracy; CCCINTER, concordance correlation coefficient for inter-observer agreement; CCCINTRA, concordance correlation coefficient for intra-observer agreement. 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 512 BELLUCO ET AL. T A B L E 2 Concordance of each Observer Histologic grade Well classified samples Total accuracy neuropathologist's second reading A Grade I 66/73 (90.4%) 135/156 (86.5%) diagnosis to the consensus classification Grade II 58/72 (79.5%) CI: (80.2%, 91.5%) for histologic grade Grade III 11/11 (100.0%) B Grade I 63/73 (86.3%) 141/156 (90.3%) Grade II 67/72 (93.1%) CI: (84.6%, 94.5%) Grade III 11/11 (100.0%) C Grade I 43/73 (58.9%) 113/156 (72.4%) Grade II 60/72 (83.3%) CI: (64.7%, 79.3%) Grade III 10/11 (90.9%) D Grade I 58/73 (79.5%) 120/156 (76.9%) Grade II 53/72 (73.6%) CI: (69.5%, 83.3%) Grade III 9/11 (81.8%) Abbreviation: CI, confidence interval. T A B L E 3 Concordance of each Observer Mitotic grade Well classified samples Total accuracy neuropathologist's second reading A Grade I 79/92 (85.9%) 114/135 (84.4%) diagnosis to the consensus classification Grade II 26/34 (76.5%) CI: (77.2%, 90.1%) for mitotic grade Grade III 9/9 (100.0%) B Grade I 90/92 (97.8%) 129/135 (95.6%) Grade II 30/34 (88.2%) CI: (90.6%, 98.4%) Grade III 9/9 (100.0%) C Grade I 82/92 (89.1%) 111/135 (82.2%) Grade II 23/34 (67.6%) CI: (74.7%, 88.3%) Grade III 6/9 (66.7%) D Grade I 83/92 (90.2%) 115/135 (85.2%) Grade II 23/34 (67.6%) CI: (78.1%, 90.7%) Grade III 9/11 (88.9%) Abbreviation: CI, confidence interval. agreement was substantial to almost perfect (CCCINTRA ranging from for small samples was lower than CCCTOT for regular samples (small sam- 0.70 to 0.83), while inter-observer agreement was slight to fair ples: CCCTOT = 0.38; regular sample: CCCTOT = 0.55). Regarding intra- (CCCINTER ranging from 0.19 to 0.37) (Table 1). observer agreement, CCCINTRA for small and regular samples was similar The most diagnosed histologic subtypes were transitional and (small samples: CCCINTRA = 0.92; regular sample: CCCINTRA = 0.87). Simi- meningothelial (50 and 46 cases, respectively) (Table S2). Meningothelial lar results were obtained for mitotic grade. The CCCINTER was lower for subtype was diagnosed at least by one pathologist in 97 cases out of small samples than for regular samples (small samples: CCCINTER = 0.41; 186 (52.2%); in 15 samples out of 97 (15.5%) the subtype was unani- regular samples: CCCINTER = 0.57). No substantial differences for mous. The transitional subtype was diagnosed at least by one pathologist CCCINTRA between small and regular samples were observed in 116 out of 186 (62%); in 25 samples out of 116 (21.6%) the subtype (CCCINTRA = 0.83 and 0.88 respectively). was unanimous. In 52 cases, the subtype was not collegially defined. 4 | DI SCU SSION 3.5 | Agreement estimates according to the size of each biological sample The goal of the present study was to evaluate the reproducibility of criteria used in the human grading when applied to canine meningi- Sample size was available for 185 cases. Of those, 47/185 (25.4%) had an oma, since it is an essential step before assessing the correlation 2 area less than 1 0mm (small size) and 138/185 (74.6%) were larger than between grading and prognosis. 10 mm2 (regular size). For these two groups, the CCCTOT, CCCINTRA and Histologic grade. The concordance between individual neuropa- CCCINTER are reported in Table 4. Concerning histologic grade, CCCTOT thologist's histologic grade and the consensus diagnosis ranged from 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License BELLUCO ET AL. 513 T A B L E 4 Estimates of the concordance correlation coefficients This difficulty was exacerbated by slide digitisation at 20 magnifica- for histologic grade and mitotic grade for small and regular biological tion, resulting in a not optimal resolution, as well as by different moni- samples tor size, resolution and colour calibration, used by the participants.26 Histologic grade Mitotic grade No recommendations on scanner magnification or monitors have been published yet for diagnostic pathology, though it is common use in Index Regular Small Regular Small diagnostic laboratories to digitise at 40 and use at least a 28 in. CCCTOT 0.55 (0.44) 0.38 (0.25) 0.54 (0.42) 0.37 (0.18) monitor. CCCPREC 0.56 (0.44) 0.41 (0.28) 0.55 (0.44) 0.38 (0.18) Another cause of mitotic grade disagreement in this study was χ 0.99 (0.97) 0.94 (0.87) 0.99 (0.97) 0.99 (0.95) the lack of mitotic count standardisation. Mitoses were counted in CCCINTER 0.58 (0.47) 0.41 (0.27) 0.57 (0.45) 0.41 (0.20) three different ways, depending on the observer's interpretation of CCCPREC 0.58 (0.48) 0.44 (0.30) 0.59 (0.47) 0.41 (0.21) the criterion: (1) in 10 consecutive fields of 0.237 mm2 within the χ 0.99 (0.97) 0.94 (0.87) 0.97 (0.94) 0.99 (0.94) most mitotically-active area; (2) in 10 different fields of 0.237 mm2 CCCINTRA 0.92 (0.89) 0.87 (0.81) 0.88 (0.83) 0.83 (0.73) dispersed throughout the tumour; (3) in a randomly chosen area of Note: Estimates of the total, inter-observer and intra-observer correlation 2.37 mm2. The most recent WHO human meningioma grading system coefficients and respective sub-indices (precision and accuracy) for each indicates to count mitoses in consecutive high power fields,27 in vet- histologic parameter are reported; the lower boundary of the 97.5% erinary literature it is preconize to count mitoses in consecutive fields one-sided confidence interval is included within parentheses. CCCTOT, of the most mitotically-active area.24 Therefore, in order to standard- total concordance correlation coefficient considering the intra and inter-observer agreement; CCCPREC, precision sub-index of the ise mitotic count in canine meningioma, we propose that mitoses concordance correlation coefficient; χ, accuracy; CCCINTER, Concordance should be counted in 10 consecutive fields of 0.237 mm2 within the correlation coefficient for inter-observer agreement, CCCINTRA, most mitotically-active area, as recommended by the current Concordance correlation coefficient for intra-observer agreement. literature.24,28 The WHO human meningioma classification cut-off of 4 and 20 mitoses is assessed in 1.60 mm2, consisting of the area obtained 59% to 100%, while the agreement among human neuropathologists by counting 10 fields at 400 magnification with a microscope field for the histologic grade is 87.2%.17 In contrast, whilst in human number equaling 18.27–29 Thus, if mitoses are counted in an area of meningiomas the agreement is high for each tumour grade, in the pre- 2.37 mm2 in dogs, the cut-offs should be readjusted to 8 and 41. Fur- sent study, an almost perfect concordance was registered only for thermore, the authors strongly encourage further studies on canine Grade III tumours. Mitotic grade and anaplasia, criteria on which mitotic count cut-off values, since, to the best of our knowledge, Grade III is assessed, displayed the highest agreement among there is no evidence that the human cut-offs have a prognostic signifi- observers, explaining the high concordance for histologic Grade III. cance in canine meningioma. Grade I and Grade II tumours had lower concordance, suggesting that In the present study, inter-observer disagreement between histologic criteria for grading tumours as I or II (mitotic index, invasion, mitotic Grade 1 and 2 was often due to a difference of only 1–3 coun- sheeting architecture, small cells, hypercellularity, macronucleoli and ted mitoses. An oversight of 1–3 mitoses could be associated with the spontaneous necrosis) are more susceptible to subjective interpreta- different amount of time spent by each observer reading slides, as tion. Considering all the examined cases, low inter-observer agree- suggested by Rogers,30 and/or with a different chosen area to count ment was recorded for macronucleoli, small cells, hypercellularity and mitoses. To reduce discrepancies, when the mitotic count is 1–2 mito- sheeting architecture. These criteria represent four out of five criteria ses lower than the cut-off, mitoses should be re-counted in another used to grade tumours as I or II, and they also represent the less highly mitotic area. 17 agreed criteria among human neuropathologists (κ < 0.5). Tumour Invasion (Figures 1 and 2). Agreement on tumour invasion was invasion and necrosis displayed moderate agreement. Since tumour fair, and, hence, much lower than in human meningiomas (κ = 0.76).30 invasion is sufficient to classify a meningioma as Grade II, the inter- In human medicine, tumour invasion is correlated to tumour recur- observer disagreement for invasion could have contributed to the rence, therefore invasion is considered sufficient to upgrade a menin- lower concordance for tumour histologic Grades I and II. gioma from Grade I to Grade II.29,31 Despite the fact that no Mitotic grade. Mitotic grade is one of the three main criteria (along prognostic studies have been published in dogs, it seems biologically with invasion and anaplasia) directly linked to histologic grade. In reasonable that invasion could have a similar prognostic impact in human medicine, mitotic grade is considered a more reproducible cri- dogs. Thus, the presence of brain tissue adjacent to the tumour sam- terion, and therefore, up to 75% of meningiomas are graded based on ple is strongly recommended. mitotic count.24 In the present study, inter-observer agreement was A standardised definition of nervous system invasion is often lac- moderate, which is comparable to reported inter-observer agreement king in the recent veterinary literature.28 We recommend using the between human neuropathologists (for mitoses between 4 and 20 in pathology human definition: ‘an extension of tumour cells beyond the 10 HPF, κ = 0.51). 17 As described in several studies, mitotic cells are pial surface into the adjacent brain parenchyma’.29,32 In most cases, often difficult to differentiate from pyknotic and karyorrhectic cells, invasion is represented by tongue/finger-like protrusions of the infiltrating leukocytes and fixation-related artefacts.24–26 tumour into the central nervous system, while invasion is less 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 514 BELLUCO ET AL. F I G U R E 1 Invasive, meningioma, dog, case 43. Tongue-like F I G U R E 3 Spontaneous necrosis, meningioma, dog, case 72. A protrusion of meningioma tumour cells into the brain tissue. large area of spontaneous necrosis (asterisk) is visible. It is composed Haematoxylin and eosin by a centre of eosinophilic amorphous material, lined by dying cells (arrowhead), that lack cell borders and have a hypereosinophilic cytoplasm and condensed chromatin. Viable tumour cells are present at the periphery. Haematoxylin and eosin F I G U R E 2 Not invasive tumour, meningioma, dog, case 28. Tumour nest, though within brain parenchyma, is surrounded by a visible layer of eosinophilic spindle pial cells (arrowhead). Haematoxylin and eosin F I G U R E 4 Artefactual necrosis, brain, dog, case 23. At the edge of this small sample, measuring approximately 2.4 mm2, a focal area of necrosis admixed with fibrin is visible (asterisk). It is close to a group frequently represented by migration of single neoplastic cells or of extravasated erythrocytes (arrowhead) in the absence of islands of tumour cells into brain tissue.29,31 Peripheral nervous sys- hemosiderin-laden macrophages, indicating a recent haemorrhage, probably caused by biopsy technique or tissue manipulation before tem invasion, as well as bone and dura infiltration, even if they have fixation. Few tumour cells are also present (arrows). Haematoxylin an impact on complete surgical resection, are not considered proper and eosin tumour invasion of the central nervous system.29,32 Moreover, tumour cells in Virchow-Robin spaces are not considered an invasion, since this space is lined by pia cells.27,29,32 Discerning the presence/absence of a layer of fusiform meningeal cells around the tumour could be Spontaneous necrosis (Figure 3). In this study, inter-observer challenging, especially in dogs where specific canine meningeal agreement on spontaneous necrosis was moderate, compared to our markers are lacking. Since tumour invasion into the central nervous human counterparts (κ = 0.66).17 Although small, rare and/or solitary 29 system is associated with reactive astrocytosis, in human medicine, foci could have been difficult to detect for some participating neuro- GFAP-staining is recommended to assess the presence of an astro- pathologists, the main reason for disagreement was different interpre- cytic reaction on the invasive front.25,27,31,33 However, the authors tation of necrotic foci. Small necrotic foci, located at the edge of think that while helpful, astrocytosis should not be considered a suffi- samples, were reported by some observers as necrosis, while others cient parameter to assess invasion, since it can be observed in interpreted these foci as artefacts related to sample handling response to other stimuli, like tumour compression and perilesional (Figure 4). Similarly, large haemorrhages were interpreted as necrotic edema. foci by some participants, because of tissue loss due to tissue tearing 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License BELLUCO ET AL. 515 F I G U R E 5 Neutrophilic accumulation in necrotic foci, brain, dog, F I G U R E 7 Macronucleoli, meningioma, dog, case 161. Neoplastic case 72. In the centre of a neoplastic lobule, tumour cells are replaced cell nuclei are vesicular and contain a large eosinophilic nucleolus, by small areas of amorphous eosinophilic material, consistent with evaluated as a macronucleolus necrotic material and a moderate number of neutrophils (arrow). Some neutrophils infiltrate the tumour lobule towards the necrotic centre F I G U R E 8 Macronucleoli, meningioma, dog, case 108. Neoplastic cell nuclei are vesicular and contain a easily visible nucleolus, F I G U R E 6 Neutrophilic accumulation in necrotic foci, brain, dog, evaluated as a macronucleolus case 72. Tumour is infiltrated by a large amount of neutrophils, forming a microabscess (arrow). Haematoxylin and eosin In veterinary literature, canine macronucleoli are neither defined nor measured, leading to a subjective interpretation of this parame- and secondary traumatic or ischemic cell necrosis. Furthermore, some ter.11,15 The vast majority of macronucleoli identified in the canine participants interpreted intra-tumoral abscesses as necrotic foci, meningiomas in this study were characterised by clumped chromatin because neutrophils may have induced tumour cell necrosis, or neu- or eosinophilic staining with different sizes and shapes. Nucleoli have trophil chemotaxis could have been triggered by previous spontane- a vast range in appearance, which contributed to the different inter- ous tumour necrosis (Figures 5 and 6). In order to increase pretations of macronucleolus by the neuropathologists involved in this reproducibility in necrosis evaluation, the authors suggest: reporting study. In human meningioma, nucleoli are defined as macronucleoli spontaneous necrosis when present inside the sample; ruling out arte- when they are easily visible at 100 magnification. For meningioma fact if necrosis is located at the sample margins or in hemorrhagic grading purposes, the criterion is considered positive even when areas; considering small and large foci equally; considering abscesses macronucleoli are focally observed (but are present in more than one as necrotic foci if cell necrosis is evident inside or around the abscess. cell).32 Although this definition sounds obsolete in the image analysis Macronucleoli (Figures 7–10). In the present study, macronucleoli era, and is reported to have only moderate agreement even among represented one of the soft criteria with the least reproducibility in human neuropathologists (κ = 0.49), macronucleolar identification is agreement. During the review process, neuropathologists identified easily applicable by light microscopy.17,25 two main reasons for macronucleoli misinterpretation: the exact For some of the neuropathologists participating in this study, meaning of macronucleoli is unclear and there are no standardised 100 magnification seemed too low to identify nucleoli, even when methods for screening macronucleoli. prominent; thus, 200 magnification was suggested. Nevertheless, in 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 516 BELLUCO ET AL. F I G U R E 9 Small nucleoli, meningioma, dog, case 5: Tumour cells F I G U R E 1 1 Small cells, meningioma, dog, case 49. Among contain small nucleoli. Chromatin is finely dispersed. Haematoxylin tumour whorls of a meningothelial meningioma, there is a cluster of and eosin small cells (encircled), characterised by reduced size and condensed chromatin. Despite the small size, cellularity is not increased F I G U R E 1 0 Small nucleoli, meningioma, dog, case 6: Tumour cells contain almost not visible nucleoli. Chromatin is finely dispersed. Haematoxylin and eosin F I G U R E 1 2 Small cells, meningioma, dog, case 69. Most of the tumour cells are small (asterisk), with condensed chromatin mimicking lymphocytes. The high number of small cells in the cluster is also suggestive of foci of hypercellularity. Haematoxylin and eosin the absence of validated studies, we still recommend using the WHO human definition for macronucleoli identified at 100 magnification when grading canine meningioma. became synonymous causing redundancy of these two evaluated Another potential reason for disagreement regarding criteria. In the WHO human meningioma classification, small cells are macronucleoli was the different screen size, ranging from 19 to 27 in., considered as having a high nucleus-cytoplasmic ratio and/or resem- used by observers in this study. Therefore, we recommend adopting a bling lymphocytes.11,32 In the present study, where no immunohisto- standardised magnification to evaluate macronucleoli, and if working chemical stains were performed, the similarity between small cells and on digitalized slides, we recommend calculating the magnification lymphocytes could have increased the disagreement, especially for using the scale bar (not the viewer magnification). Specifically, for a cases with a lymphocytic-rich pattern. Therefore, in routine diagnosis, 100 magnification, the scale bar on the screen should be 100 times T and B cell immunolabelling could help to discriminate lymphocytes longer than the displayed length (e.g., for a correct 100 magnifica- from tumour cells. tion, the scale bar indicating 500 μm should measure 5 cm on the Hypercellularity (Figure 14). Hypercellularity displayed the lowest screen). agreement among all evaluated histologic criteria. This was lower than Small cells (Figures 11–13). In the present study, the criterion of the agreement reported in human medicine (κ = 0.45).17 Several fac- small cells showed fair agreement, as in human medicine (κ = 0.39).17 tors could have contributed to this result. First, in the veterinary liter- For participating veterinary neuropathologists, as for human ature, details defining hypercellularity are lacking, leaving observers to pathologists,17,24 it was difficult to interpret the term ‘small cells’, make a subjective interpretation of the criterion. In human medicine, a especially in hypercellular areas. Hypercellularity and small cells often number of methods have been applied to make the identification of 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License BELLUCO ET AL. 517 F I G U R E 1 3 Lymphocytic-rich pattern, meningioma, dog, case 97. F I G U R E 1 5 Anaplasia and mitoses, meningioma, dog, case 179. In this area, a great number of lymphocytes, mimicking tumour small In this Grade III meningioma, no specific tumour architecture is cells, obscures the meningothelial pattern. Haematoxylin and eosin recognisable, since cells are growing in sheets F I G U R E 1 6 Anaplasia and mitoses, meningioma, dog, case 179. A F I G U R E 1 4 Hypercellularity, meningioma, dog, cases 62. An area high number of mitotic cells (arrows) are visible. Tumour cells are of basophilic hypercellularity (asterisk) in an otherwise clear increased in number (hypercellularity) and anaplastic. Haematoxylin eosinophilic tumour (white star). Haematoxylin and eosin and eosin hypercellularity less subjective and provide more reproducible results.24 One method defines increased cellularity as the presence of Sheeting architecture (Figures 15 and 16). Sheeting architecture 2 more than 53 cells in 0.058 mm (corresponding to 217 cells in displayed a fair agreement, comparable to the agreement among 0.237 mm2).24,32 This cell density seems to be unrealistic in canine human pathologists (κ = 0.41).17 meningioma, since in the present study the recorded cell number in Canine meningioma often expresses more than one growth pat- 0.237 mm2 varied between 538 and 3567 (data not shown). Second, tern, making it challenging to recognise sheeting, especially at high in the present study, hypercellularity could have been interpreted in magnification. For example, the syncytial architecture of relation to the mean cellularity of that specific tumour subtype and meningothelial meningioma can mimic sheeting at high magnifica- not as an absolute parameter. And finally, in daily routine grading of tion.17 Moreover, some histologic subtypes, like clear cell, rhabdoid or human meningioma, hypercellularity is semi-quantitatively evaluated microcystic, are characterised by cell sheeting,24,25 but this architec- 24,32 by neuropathologists at low magnification. Semi-quantitative ture is subtype-specific and does not indicate dedifferentiation and evaluation of increased cellularity remains subjective,25 which increased malignancy. For these reasons, in order to clarify the crite- should be taken into account when a grading system for canine rion and increase reproducibility in canine meningioma grading, we meningioma is correlated to prognosis. In conclusion, though sub- suggest replacing ‘sheeting architecture’ with the term ‘pattern loss’, jective, hypercellularity should be evaluated as an absolute criterion meaning the lack of a typical meningioma growth pattern.24 Pattern (not subtype-related) at low magnification and should be reported loss should be evaluated at low magnification and reported when pre- in canine meningioma, even when focally observed.29,32 sent in more than 50% of the tumour.24,32 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 518 BELLUCO ET AL. T A B L E 5 Recommendations of the consortium for reproducible criteria for canine meningioma grading Histologic criteria Definition Mitotic count Evaluated: • in the most mitotic area • in consecutive high power fields to cover 2.37 mm2 area • Grade II is defined as tumours with ≥8 mitoses in 2.37 mm2 • Grade III is defined as tumours with ≥41 mitoses in 2.37 mm2 • when mitotic count is closely lower than the cut-off (1–2 mitoses of difference), a supplemental count should be performed in F I G U R E 1 7 Anaplasia, meningioma, dog, case 41. Anaplastic another highly mitotic area cells, showing different shape and size. The grow pattern is not Invasion Presence of tumour cells into the brain or the recognisable, rendering meningioma diagnosis hard to make without spinal cord, not surrounded by a pial layer special stains. Some mitotic cells are also visible (arrows), indicating a Necrosis Focal or multifocal presence of spontaneous high mitotic grade. Haematoxylin and eosin necrosis Small and large foci are equally considered When located at sample margins or in hemorrhagic areas, artefactual necrosis should be ruled out In abscesses, reported if tumour cell necrosis is evident Macronucleoli Focal or multifocal presence of nucleoli visible at 100 Small cells Focal or multifocal presence of cells with a high nucleus/cytoplasmic ratio or with a lymphocytic appearance Evaluated at low magnification Hypercellularity Focal or multifocal Evaluated at low magnification Evaluated separately from small cells Pattern loss Not identifiable architectural pattern in more than 50% of the tumour surface F I G U R E 1 8 Anaplasia, meningioma, dog, case 162. Anaplastic Evaluated at low magnification cells, showing different shape and size. The grow pattern is not recognizable, rendering meningioma diagnosis hard to make without Anaplasia Multifocal or diffuse presence of anaplastic cells, special stains. Some mitotic cells are also visible (arrows), indicating a whose meningeal origin is not evident high mitotic grade. Hematoxylin and eosin Anaplasia (Figures 16–18). Anaplasia displayed a moderate agree- have more relevance in diagnostic routines. For human meningioma, ment among observers in this study, comparable to agreement among six blocks per tumour are recommended to achieve 95% accuracy in human pathologists (κ = 0.53).17 Anaplasia is defined as overtly malig- grading Grade II tumours.34 In the present study, reduced size contrib- nant cytology, rendering it impossible to differentiate meningioma uted to the decreased inter-observer agreement, since samples less from carcinoma, melanoma or high-grade sarcoma.29,32 Terms such as than 10 mm2 in size obtained only a fair agreement for both histologic ‘extreme’, taken from veterinary literature, 15,28 should be avoided, and mitotic grade (CCCTOT for histologic grade = 0.38; for mitotic because they are misleading and may increase the variability in inter- grade = 0.37). Small samples are more easily affected by artefactual pretation of anaplasia intensity. In the present study, anaplasia was changes due to handling, which impedes the assessment of some observed focally, multifocally or diffusely. In the absence of prognos- criteria (necrosis, mitotic figures). Moreover, since most criteria can be tic studies, the authors concluded that anaplasia, when present focally expressed multifocally, it is important to evaluate as much tissue as or affecting small clusters of cells, should not be considered to indi- possible.34 cate Grade III, but should be reported to clinicians as a comment; Tumour subtype. In human medicine, though 15 meningioma sub- however, when present multifocally or diffusely, anaplasia should be types are described, few of them present a real prognostic value. considered sufficient to upgrade the tumour to Grade III. Canine meningiomas display different tumour subtypes, often more Sample size. With the spreading of image techniques, in the near than one in the same sample. This heterogeneity makes sometimes future biopsies are going to increase in number, and sample size will difficult to attribute a prevalent subtype to the tumour, increasing the 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License BELLUCO ET AL. 519 inter-observer disagreement, as demonstrated in the present study. 2. Song RB, Vite CH, Bradley CW, Cross JR. Postmortem evaluation of To date, studies on wide cohorts correlating the subtype to prognosis 435 cases of intracranial neoplasia in dogs and relationship of neo- plasm with breed, age, and body weight. J Vet Intern Med. 2013;27(5): have not been reported yet.27,29 1143-1152. doi:10.1111/jvim.12136 For all the evaluated criteria, intra-observer agreement was 3. Foster ES, Carrillo JM, Patnaik AK. Clinical signs of tumors affecting substantial to almost perfect showing that the disagreement was the rostral cerebrum in 43 dogs. J Vet Intern Med. 1988;2(2):71-74. almost all due to inter-observer disagreement.35 As in human neuro- doi:10.1111/j.1939-1676.1988.tb02796.x 4. Petersen SA, Sturges BK, Dickinson PJ, et al. Canine intraspinal pathology, a regular discussion on different cases should be encour- meningiomas: imaging features, histopathologic classification, and aged in veterinary pathology, to better define the more subjective long-term outcome in 34 dogs. J Vet Intern Med. 2008;22(4):946-953. features of a disease entity, to homogenously educate pathologists doi:10.1111/j.1939-1676.2008.0106.x and, ultimately, to increase diagnostic accuracy.17,24,30 5. Schwartz M, Lamb CR, Brodbelt DC, Volk HA. Canine intracranial neoplasia: clinical risk factors for development of epileptic seizures. Robust and reproducible histologic tumour diagnosis and grading J Small Anim Pract. 2011;52(12):632-637. doi:10.1111/j.1748- is an essential part of providing adequate patient care and treatment. 5827.2011.01131.x In the present study, a lack of precise definitions for histologic grading 6. Lacassagne K, Hearon K, Berg J, et al. Canine spinal meningiomas criteria for canine meningioma forced pathologists to make subjective and nerve sheath tumours in 34 dogs (2008-2016): distribution and long-term outcome based upon histopathology and treatment interpretations. This was considered the main reason for the observed modality. Vet Comp Oncol. 2018;16(3):344-351. doi: variation in grading among pathologists. Therefore, the authors issued 10.1111/vco.12385 recommendations to standardise definitions of histologic criteria 7. Platt SR, Scase TJ, Adams V, et al. Vascular endothelial growth factor aimed to improve the reproducibility of canine meningioma grading expression in canine intracranial meningiomas and association with (Table 5). Unfortunately, in both human and canine meningioma patient survival. J Vet Intern Med. 2006;20(3):663-668. doi: 10.1892/0891-6640(2006)20[663:vegfei]2.0.co;2 grading, several criteria are still not completely standardised, leading 8. Suñol A, Mascort J, Font C, Bastante AR, Pumarola M, Feliu- to subjective interpretation by observers. Pascual AL. Long-term follow-up of surgical resection alone for pri- The use of digitalized slides probably contributed to the low mary intracranial rostrotentorial tumors in dogs: 29 cases inter-observer agreement, because of the lack of validated process, (2002-2013). Open Vet J. 2017;7(4):375-383. doi:10.4314/ovj. v7i4.14 the lack of users training on digital pathology and the lack of 9. Motta L, Mandara MT, Skerritt GC. Canine and feline intracranial standardised workstation.35 It is out of the scope of the present man- meningiomas: an updated review. Vet J Lond Engl. 2012;192(2):153- uscript to give recommendations on digital pathology, but readers are 165. doi:10.1016/j.tvjl.2011.10.008 strongly encouraged to read the most recent specialised literature on 10. Koestner A. Histological Classification of Tumors of the Nervous System of Domestic Animals; Armed Force Institute of Pathology; 1999. the topic. 11. Higgins RJ, Bollen AW, Dickinsons PJ, Siso-Llonch S. Tumors of the Although our results and recommendations can contribute to nervous system. In: Meuten DJ, ed. Tumors in Domestic Animals. Vol standardise the grading system for canine meningioma, further work 1. John Wiley & Sons, Inc.; 2017:834-891. is needed to generate a reproducible grading system for canine 12. Mandara MT, Pavone S, Mandrioli L, Bettini G, Falzone C, Baroni M. Matrix metalloproteinase-2 and matrix metalloproteinase-9 expres- meningioma that is correlated to prognosis. sion in canine and feline meningioma. Vet Pathol. 2009;46(5):836- 845. doi:10.1354/vp.08-VP-0185-M-FL ACKNOWLEDGEMEN TS 13. Miller AD, Miller CR, Rossmeisl JH. Canine primary intracranial can- The authors thank all the technicians and pathologists involved in cer: a Clinicopathologic and comparative review of glioma, meningi- oma, and choroid plexus tumors. Front Oncol. 2019;9:1151. doi: their daily work and Dr. Leah Cannon for editing support. Without 10.3389/fonc.2019.01151 their excellent work, studies like this one would never be completed. 14. Sturges BK, Dickinson PJ, Bollen AW, et al. Magnetic resonance imag- ing and histological classification of intracranial meningiomas in CONF LICT OF IN TE RE ST 112 dogs. J Vet Intern Med. 2008;22(3):586-595. doi:10.1111/j.1939- The authors declare no potential conflict of interest. 1676.2008.00042.x 15. Vandevelde M, Higgins RJ, Oevermann A. Neoplasia. Veterinary Neu- ropathology: Essentials of Theory and Practice. John Wiley & Sons, Inc; DATA AVAI LAB ILITY S TATEMENT 2012. The data that support the findings of this study are available from the 16. Rotondi MA, Donner A. A confidence interval approach to sample corresponding author upon reasonable request. size estimation for interobserver agreement studies with multiple raters and outcomes. J Clin Epidemiol. 2012;65(7):778-784. doi: 10.1016/j.jclinepi.2011.10.019 ORCID 17. Rogers CL, Perry A, Pugh S, et al. Pathology concordance levels for Sara Belluco https://orcid.org/0000-0002-7209-6879 meningioma classification and grading in NRG oncology RTOG trial 0539. Neuro-Oncology. 2016;18(4):565-574. 18. R Foundation for Statistical, Computing, Vienna, Austria. R: The R RE FE R ENC E S Project for Statistical Computing. Accessed June 10, 2021. https:// 1. Snyder JM, Shofer FS, Van Winkle TJ, Massicotte C. Canine intracra- www.r-project.org/ nial primary neoplasia: 173 cases (1986-2003). J Vet Intern Med. 19. Lin L, Hedayat AS, Wu W. Statistical Tools for Measuring Agreement. 2006;20(3):669-675. Springer-Verlag; 2012. doi:10.1007/978-1-4614-0562-7 14765829, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vco.12802 by Readcube (Labtiva Inc.), Wiley Online Library on [28/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 520 BELLUCO ET AL. 20. Gwet KL. irrCAC: computing chance-corrected agreement coeffi- 30. Rogers L, Gilbert M, Vogelbaum MA. Intracranial meningiomas of cients (CAC). Accessed June 10, 2021. https://CRAN.R-project.org/ atypical (WHO grade II) histology. J Neurooncol. 2010;99(3):393-405. package=irrCAC doi:10.1007/s11060-010-0343-1 21. Berthold MR, Cebron N, Dill F, et al. KNIME: the Konstanz informa- 31. Brokinkel B, Hess K, Mawrin C. Brain invasion in meningiomas— tion miner. In: Preisach C, Burkhardt H, Schmidt-Thieme L, Decker R, clinical considerations and impact of neuropathological evaluation: a eds. Data Analysis, Machine Learning and Applications. Studies in Classi- systematic review. Neuro-Oncology. 2017;19(10):1298-1307. doi: fication, Data Analysis, and Knowledge Organization. Springer; 2008: 10.1093/neuonc/nox071 319-326. doi:10.1007/978-3-540-78246-9_38 32. Arie Perry DJB. Practical Surgical Neuropathology a Diagnostic 22. Preisach C, Burkhardt H, Schmidt-Thieme B, Decker R, eds. Data Approach. Vol 1. 2nd ed. Elsevier; 2018. Analysis, Machine Learning and Applications: Proceedings of the 31st 33. Banan R, Abbetmeier-Basse M, Hong B, et al. The prognostic signifi- Annual Conference of the Gesellschaft Für Klassifikation e.V., Albert- cance of clinicopathological features in meningiomas: microscopic Ludwigs-Universität Freiburg, March 7–9, 2007. Springer-Verlag; 2008. brain invasion can predict patient outcome in otherwise benign doi:10.1007/978-3-540-78246-9 meningiomas. Neuropathol Appl Neurobiol. 2021;47:724-735. doi: 23. Landis JR, Koch GG. The measurement of observer agreement for 10.1111/nan.12700 categorical data. Biometrics. 1977;33(1):159-174. doi:10.2307/ 34. Dal Col P, Garaix T, Massard A, et al. Meningioma sampling: how 2529310 much is enough for the accurate grading of atypical meningiomas? 24. Backer-Grøndahl T, Moen BH, Torp SH. The histopathological spec- Pathology. 2021;53:602-607. doi:10.1016/j.pathol.2020.10.024 trum of human meningiomas. Int J Clin Exp Pathol. 2012;5(3): 35. Bertram CA, Stathonikos N, Donovan TA, et al. Validation of digital 231-242. microscopy: review of validation methods and sources of bias. Vet 25. Barresi V, Caffo M, Tuccari G. Classification of human meningiomas: Pathol. 2022;59(1):26-38. doi:10.1177/03009858211040476 lights, shadows, and future perspectives. J Neurosci Res. 2016;94(12): 1604-1612. doi:10.1002/jnr.23801 26. Donovan TA, Moore FM, Bertram CA, et al. Mitotic figures-normal, SUPPORTING INF ORMATION atypical, and imposters: a guide to identification. Vet Pathol. 2021; Additional supporting information may be found in the online version 58(2):243-257. doi:10.1177/0300985820980049 of the article at the publisher's website. 27. Sahm F, Perry A, von Deimling A, et al. Meningioma. WHO Classifica- tion of Tumours Editorial Board. Central Nervous System Tumours [Inter- net]. Vol 6. 5th ed. International Agnecy for Research in Cancer; 2021. https://tumourclassification.iarc.who.int/chapters/45 How to cite this article: Belluco S, Marano G, Baiker K, et al. 28. Meuten DJ, Moore FM, George JW. Mitotic count and the field of Standardisation of canine meningioma grading: Inter-observer view area: time to standardize. Vet Pathol. 2016;53(1):7-9. doi: agreement and recommendations for reproducible 10.1177/0300985815593349 histopathologic criteria. Vet Comp Oncol. 2022;20(2):509-520. 29. Louis DN, Ohgaki H, Wiestler OD, et al. In: Jaffe ES, Lakhani SR, Ohgaky H, eds. WHO Classification of Tumours of the Central Nervous Sys- doi:10.1111/vco.12802 tem. Vol 1. International Agency for Research on Cancer (IARC); 2016.