Better see to better agree: phosphohistone H3 increases interobserver agreement in mitotic count for meningioma grading and imposes new specific thresholds

Abstract

Background

Mitotic count on hematoxylin and eosin (H&E)–stained slides is a crucial diagnostic criterion in meningioma grading. However, mitosis assessment on H&E slides can be impaired by technical factors and by pathologist's experience. Phosphohistone H3 (PHH3) serine-10 is a mitosis-specific antibody that has proven to facilitate mitotic count in various tumors.

Methods

A series of 70 meningiomas (15 grade I, 40 grade II, 15 grade III) was used to validate PHH3 intra- and interobserver reproducibility and to identify PHH3-specific mitotic thresholds. Four pathologists with different experience in neuropathology counted mitoses on both H&E- and PHH3-stained slides.

Results

H&E and PHH3 mitotic rates were highly correlated (Pearson's r = 0.92, P < .0001). PHH3 mitotic counts had both a good mean interobserver correlation (R_m = 0.83) and a good intraclass correlation (0.78), higher than H&E mitotic indices (R_m = 0.77, intraclass correlation = 0.71). After further stratification of meningiomas according to World Health Organization grade, PHH3 performed better in terms of interobserver concordance (Kendall's W = 0.761) compared with H&E (Kendall's W = 0.697). Referring to the same meningioma groups identified by World Health Organization grade as the gold standard, the volume under the receiver operator characteristic surface was 0.91, indicating a very good diagnostic ability of PHH3 scores in discriminating the 3 meningioma groups. The 2 optimal PHH3-specific cutoff values were 6.61 and 22.02.

Conclusion

PHH3 staining is a useful diagnostic complementary tool to standard H&E mitotic count, optimizing intra- and interobserver reproducibility. PHH3-specific mitotic thresholds should be adopted to avoid overgrading of meningioma when ancillary methods are employed.

Mitotic count in meningiomas, as in other primary central nervous system tumors, is a strong morphological parameter prognosticating recurrence and survival.¹ Accordingly, the current World Health Organization (WHO) classification of meningiomas² distinguishes 3 grades with increasing risk of local recurrence by means of mitotic thresholds as an objective grading criterion:¹ benign (WHO grade I) with <4 mitoses/10 high power field (HPF; 0.16 mm²),^1,2 atypical (WHO grade II) with 4–19 mitoses/10 HPF, and anaplastic (WHO grade III) with ≥20 mitoses/10 HPF. In light of such a crucial diagnostic and prognostic role for mitoses, their assessment on hematoxylin and eosin (H&E)–stained slides can be affected both by technical factors (crushing and staining artifacts, or mimickers such as apoptosis and karyorrhexis, especially in necrotic areas) and by the experience of the pathologist performing the count (for the purpose of evaluating neuropathology specimens, choosing hot spot areas, and identifying true mitoses).

The mitosis-specific antibody phosphohistone H3 (PHH3) (also known as serine-10 [Ser10])³ highlights the cell nucleus maximally during mitotic chromosome condensation in early prophase and negligibly at any other time (including apoptosis).^4,5 PHH3 immunostaining proved to be a reliable tool with both diagnostic and prognostic implications, favoring mitotic count in melanoma,^6–15 breast cancer,^16–19 astrocytoma,^20,21 lung neuroendocrine carcinoma,²² adrenocortical carcinoma,²³ gastrointestinal stromal tumors,²⁴ leiomyosarcoma,^25,26 granular cell tumor,²⁷ esophageal squamous cell carcinoma,²⁸ ovarian²⁹ and endometrial³⁰ cancer, prostate^31,32 and urothelial³³ carcinoma, as well as in cytology materials of pancreatic³⁴ and gastrointestinal³⁵ neuroendocrine tumors and urothelial carcinoma.³⁶

Four reports previously analyzed the usefulness of PHH3 in meningiomas. PHH3 facilitated a rapid reliable grading of meningiomas by focusing on the most mitotically active areas^37,38 and showed a higher sensitivity in comparison with mitotic count on H&E-stained slides.³⁹ A novel mitotic threshold of ≥6 mitoses/10 HPF slides was proposed as the most appropriate prognostic value for predicting recurrence-free survival using PHH3.³⁹ Finally, a PHH3 mitotic index counted on 1000 tumor cells rather than per unit area on H&E sections was employed to identify 3 PHH3 subgroups, which were associated with recurrence-free survival.⁴⁰

Concerning alternative techniques for evaluating the proliferative potential in meningiomas, a high Ki-67 index already proved useful in predicting tumor behavior.^41–44 However, the use of a Ki-67 index is not officially incorporated in the current WHO classification, although deserving an indicative role in the diagnostic workup of meningiomas. Two crucial items of information have to be defined: (i) the level of diagnostic agreement among pathologists using PHH3 and (ii) the need for specific new cutoffs to equalize gradings referring to the H&E WHO grade as the gold standard, and not based on the prognoses of the studied case series³⁹ or on alternative mitotic index definitions.⁴⁰ Based on the aforementioned, a reproducibility study was designed on a series of 70 meningiomas, aimed at confirming the diagnostic role of PHH3 in assessing mitotic count and at identifying appropriate PHH3-specific mitotic thresholds.

We here show that (i) PHH3 staining is confirmed as a useful tool facilitating mitosis evaluation with a high intra- and interobserver reproducibility and (ii) PHH3-specific mitotic thresholds should be adopted to equalize PHH-3–based grading of meningioma to WHO grading groups and to avoid overgrading.

Materials and Methods

Tissue Collection

Seventy meningiomas from 57 patients were retrieved from the pathology files of the University of Torino. All patients were resected at the Neurosurgery Unit of City of Health and Science Hospital of Turin (Molinette) between 2003 and 2013. Surgical treatment of patients consisted of total or incomplete resection followed by postoperative irradiation, depending on the grade and/or incomplete resection of the tumor, through a standardized therapeutic approach. Eleven patients had multiple specimens available: 6 had the primary tumor and 1 recurrence, 1 had the primary tumor and 2 recurrences, and the other 4 had 2 recurrences, whose initial surgery was performed at another facility but had not previously undergone any type of irradiation therapy.

For all cases, the clinicopathological data were obtained and analyzed (Table 1). The study received ethical approval from the local review board of our institution.

Table 1.

Clinical pathological features of 70 cases of meningioma from 57 patients analyzed for PHH3 expression

Parameter
M/F ratio	1.19
Age, y, mean (range)	61.3 (36–82)
WHO grading, n	Grade I: 15
	Grade II: 40
	Grade III: 15
Disease status (lost to follow-up: 2)	NED: 35
	DOC: 2
	DOD: 18
Median overall survival, mo (range)	49 (1–125)

Abbreviations: NED, no evidence of disease; DOC, died of other cause; DOD, died of disease.

Immunohistochemistry

All H&E-stained slides available were reviewed, and a representative paraffin block was selected for each case. Three-micrometer-thick serial paraffin sections were processed by immunohistochemistry using an automated immunostainer (Ventana BenchMark AutoStainer) with antibodies against PHH3-Ser10 (rabbit polyclonal antibody, prediluted, 0.03 µg/mL; Cell Marque-Roche) and Ki-67 (mouse monoclonal antibody, clone mouse intestinal bacteria–1, dilution 1/100; Dako). A biotin-free, dextran chain–based detection system (EnVysion, Dako) and diaminobenzidine as the chromogen (Ventana Medical Systems) were used according to standard protocols. Appropriate positive controls were included.

Morphological and PHH3-based Mitotic Index Assessment

To assess the reproducibility of mitotic count evaluation, the same set of H&E and PHH3 slides was circulated among 4 pathologists (R.S., A.C., E.D., A.P.) with different experience in neuropathology, 2 senior and 2 junior. Participants, blinded to the original diagnoses, were separately asked to manually count mitoses in 10 consecutive HPFs (each 400×), on both H&E- and PHH3-stained sections using the same microscope, with HPF set at the dimension of 0.16 mm². PHH3-labeled mitotic figures were considered in the presence of positive staining and consistent morphological features.⁴⁵ The Ki-67 proliferation index was determined counting 1000 cells in hot spots and calculated as the percentage of positive nuclei by one senior pathologist (an experienced pathologist dedicated to neuro-oncology diagnoses, called Senior 1).

Statistical Analysis

Nonparametric tests (Wilcoxon rank-sum) were used to study differences between PHH3 and H&E counts. The relationships between Ki-67 and H&E and PHH3 were studied using Pearson's r correlation, while the Kruskal–Wallis rank-sum test was employed when considering the WHO grade for H&E and PHH3. Agreement among raters was computed using a nonparametric ANOVA test (Kruskal–Wallis rank-sum test) and the mean Spearman's correlation coefficient (R_m), and interrater reliability was measured with both Cronbach's α and the intraclass correlation coefficient (ICC). The intrarater agreement was assessed by means of linear regression analyses and the concordance correlation coefficient (CCC) for quantitative scores. For qualitative scores, Cronbach's α and Kendall's W coefficient of concordance were considered. The volume under the receiver operator characteristic (ROC) surface was used to summarize the accuracy of PHH3 scores in the 3 stratified meningioma groups for WHO grade, and optimal cutoff values for PHH3 scores (based on the maximization of the Youden index) were computed.^46–48 Statistical analyses were performed using the free software R (http://www.r-project.org/); in particular the library DiagTest3Grp was used for the 3-way ROC analysis.⁴⁹ A significance level of .05 was used.

Results

PHH3 Mitotic Count: Correlation With H&E Mitotic Rate and Intraobserver Reproducibility

All cases were evaluable for PHH3 antibody, which highlighted both mitoses and atypical mitoses, allowing an easier and rapid count (Fig. 1). The Senior 1 observer was referred to as the gold standard. Manually counted mitotic indices had a mean value of 7 mitoses/10 HPF (range, 0–44) on H&E-stained slides and of 11 mitoses/10 HPF (range, 0–45) on PHH3 slides. H&E and PHH3 mitotic rates were highly correlated (Pearson's r = 0.92) and had good intrarater reliability (CCC = 0.8). The PHH3 count was systematically higher than the H&E count (Wilcoxon test = 0.034, P < .0001). This was confirmed also by a linear regression model between H&E and PHH3 scores (PHH3 = 2.4 + 1.3 HE; for both regression coefficients, P < .001; coefficient of determination R² = 0.85) (Fig. 2). These findings were confirmed in each rater, irrespective of her/his experience (Table 2).

Fig. 1.

Representative examples of PHH3 mitotic immunostaining of meningioma. (A and B) Apoptotic bodies, which equivocally can be mistaken with traditional stain, are not recognized by PHH3, further supporting its specificity. (C and D) In meningiomas with a fusocellular pattern, PHH3 immunostaining facilitated the recognition of true mitoses in comparison with tapered nuclei (H&E, original magnification 400× in [A] and [C], immunoperoxidase for PHH3, original magnification 400× in [B] and [D]).

Fig. 2.

Linear regression model for gold standard mitotic counts, with Pearson's r = 0.92 correlation between H&E and PHH3 values. The fitted model PHH3 = 2.4 + 1.3 H&E (for both regression coefficients, P < .001), with coefficient of determination R² = 0.85, indicated that PHH3 count was systematically higher than H&E count.

Table 2.

Intraobserver correlation and reliability among H&E and PHH3 evaluations

Rater	Pearson's r	Kendall's W	Cronbach's α
Senior 1	0.92	0.86	0.85
Senior 2	0.85	0.83	0.81
Junior 1	0.9	0.83	0.78
Junior 2	0.83	0.85	0.84

Focusing on the subgroup of 25 cases (35%) that changed in grade, 19 (27%) increased from grade I to grade II, and 6 (9%) increased from grade II to grade III when mitotic count performed with PHH3 was not adjusted and considered equivalent to that determined by H&E stain.

H&E and PHH3 Mitotic Count: Interobserver Reproducibility

Mitotic count on PHH3-stained slides had a good average level of interobserver agreement (R_m = 0.83, ICC = 0.78), higher than that of H&E-based mitotic indices (R_m = 0.77, ICC = 0.71). However, between the Senior 1 rater (the gold standard) and 1 of the 2 junior raters, a significant difference for H&E values was found (Kruskal–Wallis chi-squared = 8.3678, P = .039), which increased for PHH3 evaluations (Kruskal–Wallis chi-squared = 11.7816, P = .0082).

Further stratifying meningiomas according to the WHO grade, PHH3 performed better in terms of interobserver concordance (Kendall's W = 0.761) and internal consistency (Cronbach's α = 0.897, 95% CI = 0.85–0.93) in comparison with H&E evaluations (Kendall's W = 0.697, Cronbach's α = 0.867, 95% CI = 0.78–0.91). Since observer agreement is crucial in diagnostic quality assurance, this result guarantees a relevant benefit in PHH3 use.

Comparison Between PHH3 Mitotic Count and Ki-67 Proliferation Index

Ki-67 proliferation index presented a mean value of 14% (range, 1%–40%). Ki-67 proliferation indices were correlated with the mitotic counts of H&E (Pearson's r = 0.4479, P = .0001) and PHH3 (Pearson's r = 0.4588, P = .0001). In both cases, the correlations were lower than that of PHH3 with H&E. When stratifying meningiomas according to the WHO grades, the Kruskal–Wallis rank-sum test indicated that Ki-67 was significantly different with mitotic counts of both H&E (Kruskal–Wallis chi-squared = 15.7156, P = .0003) and PHH3 (Kruskal–Wallis chi-squared = 13.8631, P = .0009) (Fig. 3).

Fig. 3.

Box plots representing the significantly different Ki-67 values (Kruskal–Wallis chi-squared = 13.8631, P = .0009) stratifying meningioma PHH3 mitotic counts according to the WHO grades.

ROC Surface: Need and Accuracy of PHH3-Specific Thresholds

As reported above, PHH3 facilitated mitoses recognition and determined an apparent overgrading of the studied cases. Therefore, identification of new specific cutoffs equalizing the H&E- and PHH3-based counts was needed. Stratifying meningiomas into 3 groups according to WHO grade, the nonparametric estimate of the volume under the ROC surface, calculated on Senior 1–gold standard PHH3 scores, was 0.91 (95% CI = 0.83–0.97), indicating a very good diagnostic ability of PHH3 scores in discriminating the 3 groups of meningiomas. The optimal cutoff values for PHH3 counts, which maximized the Youden index, were 6.61 and 22.02. These thresholds yielded an internal consistency (Cronbach's α = 0.916) and interobserver concordance (Kendall's W = 0.914) better than those obtained with current WHO cutoff values on H&E slides (Cronbach's α = 0.849, Kendall's W = 0.861). All but one of the other observers obtained similar values of Cronbach's α and Kendall's W coefficient (Table 3). With the above proposed cutoff values, the empirical correct classification probabilities associated with each diagnosis group (as identified by WHO grading criteria) are 0.94 (grade I), 0.68 (grade II), and 1.00 (grade III).

Table 3.

Intraobserver reliability with PHH3-specific thresholds derived from the ROC surface

Rater	Kendall's W	Cronbach's α
Senior 1	0.92	0.92
Senior 2	0.86	0.83
Junior 1	0.82	0.8
Junior 2	0.89	0.87

Discussion

In this study, we confirmed that PHH3 immunostaining is a reliable method to highlight mitotic figures, a fundamental morphological parameter for meningioma grading, and showed that its use improves intra- and interobserver reproducibility. Moreover, we identified PHH3-specific cutoffs, which better stratify patients into the expected histological grades, according to WHO criteria.

PHH3 as a Diagnostic Tool and Intraobserver Reliability

Mitotic count is often laborious in order to obtain reproducible values among observers. Its accurate evaluation may be remarkably impaired by technical artifacts (DNA crushing or pyknosis) or by the misdiagnosis of mitoses for apoptotic bodies or dying cells (especially in areas of extensive cellular necrosis). Several reports determined that PHH3 immunostaining is capable of clearly marking all and only mitoses in several human tumors, including melanocytic lesions,^6–15 breast cancer,^16–20 astrocytoma,^20,21 lung neuroendocrine tumors,²² prostate cancer,^31,32 adrenocortical carcinoma,⁵⁰ uterine smooth muscular neoplasms,^25,26 and gastrointestinal stromal tumors.²⁴ PHH3 mitotic figures are considered in the presence of a positive staining together with consistent morphological features, whereas stained nuclei without condensed chromatin are not counted as mitotic figures. This matched immunohistochemical/morphological positivity is a great tool to avoid under- or overestimation of mitotic count, potentially reducing interlaboratory technical variation related to immunohistochemistry setup and staining conditions. The pathological diagnostic workout of meningioma recognizes mitotic count as a key histological parameter to assess the appropriate tumor grade, which is defined by the WHO classification as grade I (<4 mitoses), grade II (4–19 mitoses), and grade III (≥20 mitoses). Mitotic figures have to be counted in 10 consecutive HPFs in the area of highest mitotic activity.^1,2 Although the diagnostic usefulness of PHH3 in meningioma was already confirmed by 4 reports,^37–40 the issue of PHH3 reliability among pathologists has never been addressed. Therefore, the first aim of this study was to assess and compare H&E and PHH3 intra- and interobserver reproducibility among pathologists with different levels of expertise in neuropathology (2 senior and 2 junior). Our data confirm the reported high concordance between morphologically and PHH3-assessed mitotic counts with a high statistical significance.^37–39 PHH3 immunostaining allowed a rapid recognition of hot spot areas and was revealed to be particularly useful for the evaluation of meningiomas with a fusocellular pattern, in which tapered nuclei could be easily mistaken for mitoses in H&E-stained slides (Fig. 1). We found the PHH3 mitotic count better correlated with the H&E count than with the Ki-67 proliferation index, although this one was significantly different, stratifying meningiomas according to the WHO grades with both H&E and PHH3 mitotic counts.

Intraobserver Reliability and Interobserver Reproducibility of H&E and PHH3 Mitotic Counts

We observed good intrarater reliability, independent of pathologist's experience; however, the PHH3 count was systematically higher than the H&E count. This finding was confirmed by each rater, regardless of experience. Interobserver reproducibility of H&E and PHH3 was systematically analyzed to determine mitotic indices. H&E-based mitotic count had a good interobserver reproducibility, but PHH3 performed better. This was confirmed by stratifying cases for meningioma grades. Nevertheless, we found a significant difference between 2 raters, the Senior 1–gold standard and a junior rater, both on H&E and PHH3 evaluations. Therefore, although PHH3 immunohistochemical staining represents a useful diagnostic tool to simplify mitosis screening, pathology expertise should still be taken into account.

Meningioma PHH3-Specific Thresholds

As PHH3 highlighted a higher number of mitotic cells,²³ the use of current H&E cutoffs to score PHH3-stained slides may overestimate the meningioma histological grade. This allowed us to suggest specific PHH3 cutoffs to discriminate grade I versus grade II and grade III meningiomas. We propose 7 PHH3 mitoses/10 HPF as the lower cutoff, corresponding to 4 H&E mitoses/10 HPF. This threshold was determined with a ROC surface.⁴⁹ The value of 6 or more PHH3 mitoses/10 HPF was suggested by Kim et al³⁹ based on significantly higher risk of recurrence on their series, and not by ROC surface approach. As an upper cutoff corresponding to 20 H&E mitoses/10 HPF in H&E, the ROC surface identified 22 PHH3 mitoses/10 HPF using PHH3.⁵¹

Conclusion

PHH3 immunostaining is a valid and reliable surrogate marker of standard H&E mitotic count, with a high intra- and interobserver reproducibility. In particular, PHH3-specific mitotic thresholds are recommended for meningiomas, in order to accurately define tumor grade.

Funding

This work was supported by grants from MIUR and Compagnia di Sanpaolo, and a Cariparo Foundation Excellence grant 2011–2012 to L.V.