New Insights on Diagnostic Reproducibility of Biphasic Mesotheliomas: A Multi-Institutional Evaluation by the International Mesothelioma Panel From the MESOPATH Reference Center

Abstract

Introduction:

The 2015 WHO classification of tumors categorized malignant mesothelioma into epithelioid, biphasic (BMM), and sarcomatoid (SMM) for prognostic relevance and treatment decisions. The survival of BMM is suspected to correlate with the amount of the sarcomatoid component. The criteria for a sarcomatoid component and the interobserver variability between pathologists for identifying this component are not well described. In ambiguous cases, a “transitional” (TMM) subtype has been proposed but was not accepted as a specific subtype in the 2015 WHO classification. The aims of this study were to evaluate the interobserver agreement in the diagnosis of BMM, to determine the nature and the significance of TMM subtype, and to relate the percentage of sarcomatoid component with survival. The value of staining for BRCA-1-associated protein (BAP1) and CDKN2A(p16) fluorescence in situ hybridization (FISH) were also assessed with respect to each of the tumoral components.

Methods:

The study was conducted by the International Mesothelioma Panel supported by the French National Cancer Institute, the network of rare cancer (EURACAN) and in collaboration with the International Association for the Study of Lung Cancer (IASLC). The patient cases include a random group of 42 surgical biopsy samples diagnosed as BMM with evaluation of SMM component by the French Panel of MESOPATH experts was selected from the total series of 971 BMM cases collected from 1998 to 2016. Fourteen international pathologists with expertise in mesothelioma reviewed digitally scanned slides (hematoxylin and eosin – stained and pan-cytokeratin) without knowledge of prior diagnosis or outcome. Cases with at least 7 of 14 pathologists recognizing TMM features were selected as a TMM group. Demographic, clinical, histopathologic, treatment, and follow-up data were retrieved from the MESOBANK database. BAP1 (clone C-4) loss and CDKN2A(p16) homozygous deletion (HD) were assessed by immunohistochemistry (IHC) and FISH, respectively. Kappa statistics were applied for interobserver agreement and multivariate analysis with Cox regression adjusted for age and gender was performed for survival analysis.

Results:

The 14 panelists recorded a total of 544 diagnoses. The interobserver correlation was moderate (weighted Kappa = 0.45). Of the cases originally classified as BMM by MESOPATH, the reviewers agreed in 71% of cases (385 of 544 opinions), with cases classified as pure epithelioid in 17% (93 of 544), and pure sarcomatoid in 12% (66 of 544 opinions). Diagnosis of BMM was made on morphology or IHC alone in 23% of the cases and with additional assessment of IHC in 77% (402 of 544). The median overall survival (OS) of the 42 BMM cases was 8 months. The OS for BMM was significantly different from SMM and epithelioid malignant mesothelioma (p < 0.0001). In BMM, a sarcomatoid component of less than 80% correlated with a better survival (p = 0.02). There was a signicant difference in survival between BMM with TMM showing a median survival at 6 months compared to 12 months for those without TMM (p < 0.0001). BAP1 loss was observed in 50% (21 of 42) of the total cases and in both components in 26%. We also compared the TMM group to that of more aggressive patterns of epithelioid subtypes of mesothelioma (solid and pleomorphic of our large MESOPATH cohort). The curve of transitional type was persistently close to the OS curve of the sarcomatoid component. The group of sarcomatoid, transitional, and pleomorphic mesothelioma were very close to each other. We then considered the contribution of BAP1 immunostaining and loss of CDKN2A(p16) by FISH. BAP1 loss was observed in 50% (21 of 41) of the total cases and in both component in 27% of the cases (11 of 41). There was no significant difference in BAP1 loss between the TMM and non-TMM groups. HD CDKN2A(p16) was detected in 74% of the total cases with no significant difference between the TMM and non-TMM groups. In multivariate analysis, TMM morphology was an indicator of poor prognosis with a hazard ratio = 3.2; 95% confidence interval: 1.6 – 8.0; and p = 0.003 even when compared to the presence of HD CDKN2A(p16) on sarcomatoid component (hazard ratio = 4.5; 95% confidence interval: 1.2 – 16.3, p = 0.02).

Conclusions:

The interobserver concordance among the international mesothelioma and French mesothelioma panel suggests clinical utility for an updated definition of biphasic mesothelioma that allows better stratification of patients into risk groups for treatment decisions, systemic anticancer therapy, or selection for surgery or palliation. We also have shown the usefulness of FISH detection of CDKN2A(p16) HD compared to BAP1 loss on the spindle cell component for the separation in ambiguous cases between benign florid stromal reaction from true sarcomatoid component of biphasic mesothelioma. Taken together our results further validate the concept of transitional pattern as a poor prognostic indicator.

Introduction

Malignant mesothelioma (MM) is a rare cancer arising from the surface mesothelial cells of the pleural, peritoneal, and pericardial cavities as well as the tunica vaginalis. Mesotheliomas are typically aggressive and, in the pleural cavity, incurable.

Tissue biopsy is still the main approach for establishing the diagnosis.^1,2 The most recent WHO 2015 classification for malignant mesothelioma into epithelioid, sarcomatoid (including desmoplastic variants), and biphasic (being diagnosed when there is more than 10% of both epithelioid and sarcomatoid components within the tumor) types. Epithelioid MM (EMM) has a better prognosis than sarcomatoid MM (SMM) and it has been suggested that the survival of biphasic MM (BMM) correlates with the amount of sarcomatoid component.^1,2 However, this view is not universally accepted. Additionally, it is well accepted that other prognostic indicators play roles in the survival of mesothelioma such as demographics, histologic subtyping, immunohistochemistry (IHC), and immune factors.^3–5

With a natural history of 7 to 9 months if untreated and a 5-year survival rate of 5%, pleural MM has an extremely poor prognosis as none of the available systemic (cisplatin, pemetrexed, or bevacizumab) or locoregional therapies results in a cure. Platinum/multitargeted antifolate combination, giving a 3-month overall benefit, is the only treatment approved by the U.S. Food and Drug Administration; for second-line therapy, immune checkpoint inhibitors were recently made available in some countries, including anti-programmed death 1 and anti cytotoxic t-lymphocyte associated protein 4 (anti–CTL-4) combination.^6–9

However, surgery is a therapeutic option that may increase overall survival (OS) rates, particularly for epithelioid types and favorable TNM stage, with the two radical surgical procedures available for maximal tumor resection being pleurectomy/decortication and extrapleural pneumonectomy.^10–12 The identification of patients who can benefit from surgery rather than those for whom palliative options are more appropriate is based on a lack of sarcomatoid component and the evaluation of the volume and resectability of the tumor. The former is the remit of the diagnostic pathologist, but criteria relating to what constitute a sarcomatoid component and levels of interobserver variability between pathologists in relation to its identification are not well described. In ambiguous cases, a “transitional” subtype (transitional between epithelioid and sarcomatoid) has been proposed, but this subtype has not been included as a specific entity in the 2015 WHO classification because its significance remains undetermined.¹ Thus, the diagnosis of a sarcomatoid component is challenging resulting in the possibility of suboptimal treatment. Reproducibility is useful to evaluate the applicability of histologic criteria for the classification of BMM and to analyze the presence and the amount of sarcomatoid component. The second important criterion to assess before surgery is the volume of the tumor resectability. Recently, Gill et al.¹³ reported that novel magnetic resonance imaging applications can play a crucial role in assessing resectability of the tumor but also can be predictive and prognostic in the separation of benign from malignant proliferation. The authors showed that, for the diagnosis of histologic type, the multi-parametric map superimposed over computed tomography (CT) scan images was able to detect intratumor heterogeneity and to recognize the distribution of sarcomatoid and epithelioid components. Moreover, the apparent diffusion coefficient value, expressed in units as mm²/s, was suggestive to predict a biphasic component.¹⁴

The aim of the present study was to evaluate the interobserver agreement in the diagnosis of BMM, to determine the nature and the significance of a transitional histologic subtype, and to relate the amount/percentage of sarcomatoid component with survival. The value of BAP-1 staining and CDKN2A(p16) fluorescence in situ hybridization (FISH) were also assessed with respect to each of the tumoral components.

Materials and Methods

The study was conducted by the International Mesothelioma Panel supported by the French National Cancer Institute (INCA) and the French National Health Institute (Santé Publique France) in collaboration with the International Association for the Study of Lung Cancer (IASLC). The patient cases were selected from the national clinicobiological database and resource biological samples through MESOBANK. All cases had been classified as mesothelioma following a standardized procedure of certification and collected over 20 years by the national reference center MESOPATH. Authorization to use human biological samples was obtained (no. AC 2011) and by the national and local ethic committee certified (no. DC2008_586, AC-2013–1806, DR-2011–309) of Cancer Center Leon Berard.

Case Selection

A random selection of 42 biopsy specimens was identified from the MESOPATH files as having been diagnosed BMM by the French panel of MESOPATH experts, based on WHO 2015 criteria and appropriate phenotype.^1,2,15 They were selected from a total series of 5219 cases of EMM, 971 BMM, and 465 SMM specimens collected over a study period from January 1998 to June 2016. The 42 cases were required to be surgical biopsy samples and to have hematoxylin and eosin (HE) sections, formalin-fixed paraffin-embedded (FFPE) blocks, and enough materials for additional ancillary techniques along with completed clinical annotations. The demographic, clinical, histopathologic, treatment, and follow-up annotations were recorded from the national clinicobiological MESOBANK database. Occupational histories were evaluated by a group of epidemiologists and were available in 71% of the cases. The majority of the cases had had conventional and CT imaging.

IHC and Molecular Analysis

The staining was performed on 4-μm-thick tissue sections cut from the FFPE block in the biopathology department of the Cancer Center Leon Berard on the automate BENCHMARK Ventana. The cases exhibited a classical mesothelial biphasic immunophenotype, for example, positivity of both epithelioid and sarcomatoid components by a broad spectrum of pan-cytokeratin (CK) (AE1/AE3), or CK8/18, or other pan-CK, as well as positivity for CK 5/6 (when available), calretinin nuclear positivity (with a cutoff requiring more than 10% positivity), and WT1 nuclear staining mandatory on the epithelioid component. CK staining could be present either on both components or on the epithelioid component alone. We accepted, according to the literature, that the sarcomatoid component can show focal and scattered or no staining of these markers.^1,2,15–20 All the carcinoma markers including CEA, BerEP4, TTF-1, ERα, and PAX8, GATA3 were performed in variable combinations as necessary and clinically appropriate.^21–23

BAP1 (clone C64) (Santa Cruz: dilution 1 of 50) nuclear staining was considered positive (when retained nuclear expression) or negative (complete loss of staining of all tumor cells with a positive internal control on the slides (fibroblast, lymphocytes, etc.). The pattern of BAP1 loss could also be present on either both components or on the epithelioid component alone. When BAP1 was lost in the spindle cell component, it was considered as a decisive argument in favor of malignancy, and when retained it was considered indeterminate for malignancy.^24–29 p16 protein (clone E6H4) prediluted Ventana was considered positive (diffuse or focal heterogeneous nuclear staining) or negative (for cases showing absence of expression on the tumor cells with a positive internal control on the slides). p16 loss of expression was not considered as a definitive argument of malignancy in the absence of homozygous CDKN2A (p16) deletion by FISH analysis.^28–31

CDKN2A (p16) FISH.

CDKN2A (p16) FISH for detection of homozygous deletion of CDKN2A (p16) was performed using the dual-color FISH analysis for the CDKN2A locus (9p21) and the centromere of the chromosome 9 (CEP-9), using the ZytoVision (Bremerhaven, Germany) probe (ZytoLight SPEC CDKN2A/CEN 9 Dual Color Probe, # Z-2063–200) and was used from fresh serial recuts of 3 μ to 5 μ in thickness from the FFPE block stored in the correct conditions. A minimum of 100 cells was required for the evaluation of the status of CDKN2A by FISH agreed upon by two independent observers. More than 80% of nuclei had to be hybridized. The number of nuclei with no copy and those with one copy was strictly counted, recorded on a score sheet, and reported. Moreover, the quality of the nuclei morphology and the intensity of 4’,6-diamidino-2-phenylindole staining were evaluated for the selection of the epithelioid and sarcomatoid component. A rigorous assessment of the size of the neoplastic nuclei and a correct counting were performed. A cutoff value of 20% of nuclei with no CDKN2A copy was considered as positive results for the presence of a homozygous deletion in both components. In some cases, a clonal disposition of the nuclei showing a homozygous deletion can strengthen this cutoff value interpretation. In cases where an average of less than one copy was observed in less than 20% of the nuclei, the results were evaluated as negative for a homozygous deletion. The presence of homozygous CDKN2A(p16) deletion by FISH analysis on the spindle cells component was also regarded as a criterion for malignancy.^29–31 On the other hand, heterozygous deletion (HD) was not considered a criterion for malignancy. To avoid technical bias, a two-step assessment with one trained technician and one molecular biologist or pathologist was involved in the case review.

Interobserver Agreement and Spread Sheet

The diagnosis of BMM was made according to the definition of the 2015 WHO classification.^1,2,32–36 A score sheet was provided and the reviewer requested to answer the following questions on one HE stain and one pan-CK (AE1/AE3) stain scanned on a digital platform accessible through a link. Reviewers were asked to make the following assessments. First, they were to confirm a diagnosis of MM and provide the histologic type: EMM, BMM (a combination of epithelioid and frank sarcomatous spindle cell component constituting at least 10% of the tumor), or SMM (proliferation of spindle cells arranged in fascicules or with haphazard distribution showing a wide range of nuclear atypia, mitosis, and necrosis that may variate from marked to minimal (including desmoplastic). Second, reviewers were asked to determine if the diagnosis was made only on morphology alone, on the pattern of pan-CK expression alone, or on both. Third, they were to determine whether the spindle cells were benign or malignant. The percentage of sarcomatoid component was not asked on the score sheet because it had been previously evaluated by the French Mesothelioma Panel of Experts on the same cohort of 42. Fourth, to provide a grade for the tumor (low-versus high-grade based on marked cytological atypia, high mitotic activity, and the presence of atypical mitosis).^37,38 Necrosis was not considered. Fifth, they were to determine whether a transitional pattern was present. The criteria for the transitional pattern was based on expert consensus and was histologically identified on the basis of sheets of plump cells starting to lose their epithelioid morphology but not overtly spindle shaped and lacking frank sarcomatous features (Fig. 1A). The transitional pattern could be diffuse covering the entire sample or only focal, but the reviewers were not asked to evaluate this point. Additionally, in the comments’ section, the panelist had the opportunity to note the presence of other morphologic qualities such as pleomorphic features characterized by the presence of atypical giant cells with multilobulated, bizarre, hyper-chromatic nuclei mixed with large cells, large hyper-chromatic nuclei showing a high mitotic count and atypical mitoses, or solid pattern of the epithelioid component (based on the presence of monotonous, relatively noncohesive sheets of polygonal cells mimicking large cell carcinoma according to the 2015 WHO definition), and the nonsolid (all other epithelioid patterns) or presence of heterologous elements.^1,39–41

Figure 1.

(A) Image 1. Transitional pattern of cohesive large plump epithelioid cells with well-defined border, high nucleocytoplasmic ratio, and prominent nucleoli.(B) The tumor displayed a conventional feature of biphasic mesothelioma showing epithelioid solid pattern of large epithelioid cell closely mixed with spindle cells component.(C) CK pattern of staining in two-tone staining. Strong AE1/AE3 perinuclear immunostaining on conventional epithelioid component and discreet, gray images staining on the spindle cell component.(D) Conventional biphasic morphology on hematoxylin and eosin (HE) staining showing a typical epithelioid component and a high-grade frank sarcomatous component characterized by nuclear atypia, numerous mitosis, and atypical mitosis.(E) Conventional biphasic morphology on HE showing an association of typical epithelioid morphology and low grade, quiet spindle cells component without nuclear atypia and mitosis.(F) (Left) Fluorescence in situ hybridization (FISH) analysis showing CDKN2A (p16) homozygous deletion on an epithelioid component. The 4’,6-diamidino-2-phenylindole stain shows a cluster of round monomorphous nuclei with loss of p16 (without green signals). Positive control on the slide with a nonmalignant cell showing two split signals red and green.(G) (Right) FISH analysis showing CDKN2A (p16) homozygous deletion. The 4’,6-diamidino-2-phenylindole staining shows the presence of a cluster of irregular, ovoid, nuclei with high nucleocytoplasmic ratio without green signals.

Statistical Analysis

The overall agreement was calculated according to the recommendation of Landis et al.⁴² and was assessed using weighted chance corrected agreement with a linear weighting. For weighted Kappa (wK) calculation, the diagnosis was ordered as follows 1) EMM, 2) BMM, and 3) SMM. Univariate analysis was performed for age, sex, asbestos exposure, histologic subtypes, and results of BAP1 and CDKN2A(p16) HD. Finally, the survival duration in months was calculated from the date of the initial pathological diagnosis until the date of death according to Kaplan-Meier methodology. Groups were compared by the log-rank test. Multivariate analysis Cox proportional hazards regression adjusted on age included the factors affecting survival in univariate analysis (p < 0.20). Hazard ratios (HRs) and 95% confidence intervals (CIs) were computed. Data were updated on 30 June 2017. The chi-square test and Fisher exact bilateral test were used for comparisons between categorical variables. Statistical calculations were performed using SAS 9.4 from SAS Institute Inc. (Chester-brook, Pennsylvania).

Results

Patient Demographics and Interobserver Agreement

The 42 patients had a mean age of 76 years; 90% were male, with an asbestos exposure history in 71%. The clinical presentation was dyspnea (45%), decreased performance status (22%), chest pain (20%), and a pleural effusion (82%). On CT scan the following were observed: pleural thickening (86%), a localized pleural mass (9%), and hyaline fibrous plaques (14%).

Fourteen experts with a special interest in thoracic pathology provided a total of 544 recorded diagnoses. The interobserver correlation was moderate (wK = 0.45). Of these cases originally classified as BMM, the reviewers agreed in 71% of cases (385 of 544 opinions), with cases classified as pure epithelioid in 17% (93 of 544 opinions), and pure sarcomatoid in 12% (66 of 544 opinions). Some panelists indicated that when a component was less than 10% of the total amount of the tumor, they classified the tumor as either epithelioid or as sarcomatoid (Table 1) (Fig. 2).

Table 1.

Results of Interobserver Agreement on the Series of 42 Biphasic Mesothelioma Cases

	Diagnosis
Criteria	EMM	BMM	SMM	Chi-Square Test
Coexistence of two distinct patterns	n = 78	n = 339	n = 66	p < 0.0001
No	73 (94%)	7 (2%)	59 (89%)
Yes	5 (6%)	332 (98%)	7 (11%)
Presence of transitional pattern	n = 82	n = 379	n = 63	p < 0.0001
No	62 (76%)	203 (54%)	50 (79%)
Yes	20 (24%)	176 (46%)	13 (21%)
Pattern of CK expression in favor of biphasic mesothelioma	n = 80	n = 363	n = 61	p = 0.008
No	37 (46%)	123 (34%)	13 (21%)
Yes	43 (54%)	240 (66%)	48 (79%)
On both morphology and pattern of CK expression	n = 83	n = 376	n = 64	p < 0.0001
No	28 (34%)	68 (18%)	25 (39%)
Yes	55 (66%)	308 (82%)	39 (61%)
Spindle cell component malignant	n = 79	n = 388	n = 66	p < 0.0001
No	65 (82%)	1 (<1%)	0
Yes	14 (18%)	387 (100%)	66 (100%)
Grade of spindle cell component	n = 46	n = 376	n = 64	p < 0.0001
Low	34 (74%)	166 (44%)	12 (19%)
High	12 (26%)	210 (56%)	52 (81%)

CK, cytokeratin.

Figure 2.

Percentage of diagnosis made by the reviewers by histological types (epithelioid, biphasic, and sarcomatoid) according to the WHO 2015 classification. EMM, epithelioid malignant mesothelioma; BMM, biphasic malignant mesothelioma; SMM, sarcomatoid malignant mesothelioma; CK, cytokeratin.

Diagnosis of BMM was made on morphology or IHC alone in 23% of the cases (121 of 544) and with additional assessment of immunohistochemistry in 77% (402 of 544) (Table 1) (Figs. 1B and 1C).

Prognosis According to Histopathologic Grouping

The overall median survival of the series of 42 BMM cases was 8 months, 38% at 1 year with a 95% CI: 23% to 53%, and 8% at 2 years with a 95% CI: 0% to19% (Fig. 3). The overall biphasic survival curve was significantly different from the sarcomatoid and the epithelioid curves (p < 0.0001). (Fig. 3)

Figure 3.

Overall median survival curves of the series of 42 biphasic cases compared to the epithelioid and sarcomatoid conventional histologic types from the large MESOBANK cohort. EMM, epithelioid malignant mesothelioma; BMM, biphasic malignant mesothelioma; SMM, sarcomatoid malignant mesothelioma.

The OS of BMM was significantly better in cases with less than 80% of sarcomatoid component initially scored by the French panel of experts, with a median survival of 12 months, 1-year survival of 56% (95% CI: 36% to 77%) and 2-year survival of 20% (95% CI: 0% to 40%) (p = 0.02).

The spindle component was considered low grade in 45% of the diagnosis of epithelioid type and in 33% of the biphasic and of the sarcomatoid. Additionally, we observed that there was a trend for patients with a high-grade spindle cell component to have shorter survival than the low-grade group (p = 0.08) (Table 1) (Figs 1D–1E and 4).

Figure 4.

Overall median survival curves of the biphasic series when comparing biphasic cases with low grade to high grade spindle cells component. CI, confidence interval.

Transitional Pattern as a Specific Clinicopathologic Entity

The cases of transitional pattern were selected according to the observers’ responses. Cases with at least 7 pathologists of 14 experts recognizing transitional features either diffuse or focal were included as a transitional group (Fig. 1A). The clinical characteristics of the biphasic with and without transitional pattern were similar. However, we observed a transitional OS curve showing significant prognostic difference between them (p < 0.0001) (Fig. 5). Moreover, there was a significant difference in survival between cases with transitional pattern showing a median survival at 6 months compared to 12 months for those without transitional pattern (Fig. 5) (p < 0.0001). We also compared the OS of the transitional cases recognized as such by the reviewers in this series to the OS data of a large cohort of MESOPATH. We found that the transitional pattern (n = 16) showed an OS close to the sarcomatoid OS curve (n = 465) compared to the epithelioid curve (n = 5219) with a median survival, respectively, of 6 months for the transitional pattern, 4 months for the sarcomatoid, and 14 months for the epithelioid type. At 1 year the survival was of 19% for transitional, 12% for sarcomatoid, and 55% for epithelioid; at 2 years 0% versus 3% and 24%, respectively; and at 5 years 0% for both transitional and sarcomatoid compared to 4% for the epithelioid type (p < 0.0001) (Fig. 6).

Figure 5.

Overall median survival curves by histologic types when comparing transitional pattern to nontransitional pattern of the series of 42 biphasic mesothelioma cases. EMM, epithelioid malignant mesothelioma; BMM, biphasic malignant mesothelioma; SMM, sarcomatoid malignant mesothelioma; CI, confidence interval.

Figure 6.

Overall median survival curves by histologic types when comparing transitional pattern of the series of 42 biphasic mesothelioma cases with more aggressive patterns of epithelioid subtypes (solid and non-solid epithelioid subtypes). EMM, epithelioid malignant mesothelioma; BMM, biphasic malignant mesothelioma; SMM, sarcomatoid malignant mesothelioma; CI, confidence interval.

Finally, we compared the OS of the transitional pattern group to that of more aggressive patterns of epithelioid subtypes of mesothelioma such as solid and pleomorphic from our large MESOPATH cohort. The curve of the transitional type was persistently close to the OS curve of sarcomatoid type, independent of the percentage of sarcomatoid component. The OS curve of the solid epithelioid type (sheets and nests of relatively noncohesive monotonous epithelioid cells, mimicking large cell carcinoma pattern) was significantly better than the transitional OS curve and remained close to the nonsolid epithelioid (all other epithelioid patterns papillary, acinous, trabecular, micropapillary, et cetera) OS curve (Fig. 4). Finally, we compared the OS of the transitional type with the OS of a series of 40 mesotheliomas with pleomorphic features. The OS curve of the group of sarcomatoid, transitional, and pleomorphic mesotheliomas were very close to each other (Fig. 7).

Figure 7.

Overall median survival curves by histologic types when comparing transitional pattern of the series of 42 biphasic mesothelioma cases with more aggressive patterns of epithelioid subtypes (pleomorphic subtype). EMM, epithelioid malignant mesothelioma; BMM, biphasic malignant mesothelioma; SMM, sarcomatoid malignant mesothelioma; CI, confidence interval.

Molecular Assessment as a Tool for Histopathologic Grouping

We then considered the contribution of BAP1 IHC and loss of CDKN2A (p16 HD) by FISH analysis for the evaluation of the spindle cells component (Table 2). BAP1 and CDKN2A(p16) HDs are considered markers of malignancy.^28–31 When BAP1 loss was observed on the spindle cell component alone it was considered as a decisive argument for malignancy and when retained it was considered indeterminate of malignancy.^25,28,29 The presence of homozygous CDKN2A (P16) deletion by FISH analysis on the spindle cell component was also regarded as a criterion for malignancy (Figs. 1F–1G). We explored the loss of BAP1 present on only one component or on both (epithelioid and sarcomatoid). BAP1 loss was observed in 50% (21 of 42) of the total cases and in both components for 27% of the cases (11 of 41). Due to the very small number of BMM showing BAP1 loss, only on the epithelioid component it was not possible to evaluate the exact prognostic impact. BAP1 loss on the sarcomatoid component alone was not observed. We did not observe any significant difference in BAP1 loss between the transitional pattern (50%, n = 8) and the nontransitional group (50%, n = 13, p 1.00).

Table 2.

BAP1 and P16 immunoreactivity and CDKN2A (p16) homozygous deletion by FISH

Variable	Nontransitional (n = 26)	Transitional (n = 16)	Fisher Exact Test
BAP1			p = 1.00
Loss on E component	6 (23%)	4 (25%)
Loss on E and S components	7 (27%)	4 (25%)
Positive	13 (50%)	8 (50%)
p16			p = 0.59
Loss on E component	3 (12%)	2 (13%)
Loss on S component	0	1 (6%)
Loss on E and S components	21 (81%)	13 (81%)
Positive	2 (7%)	0
p16	n = 23	n = 15	p = 1.00
No deletion	6 (26%)	3 (20%)
Homozygous deletion	17 (74%)	12 (80%)
On E component	3 (18%)	2 (17%)
On S component	0	1 (8%)
On E and S component	14 (82%)	9 (75%)

E, epithelioid; S, sarcomatoid; FISH, fluorescence in situ hybridization.

p16 HD by FISH was present in 74% of the series (n = 28 of 38) of the series and in 80% (n = 15) of the transitional group (p = NS). These results must be considered with caution due to the small number of FISH analyses performed. The presence of the HD was observed on the epithelioid component alone in 18% (5 of 28) showing the value of FISH analysis in determining malignancy of the spindle cells component and on both components in 82% (23 of 28) (Figs. 1F–1G). Noticeably, when p16 HD was present on the epithelioid component alone, the median survival was 13 months, and 2-year survival was 40% (95% CI: 0% – 64%) compared to 5 months (95% CI: 0% – 19) when HD (p16) was present on both epithelioid and sarcomatoid (E and S) components. Those last results where significant by multivariate analysis (HR 4.5, 95% CI: 4.5 – 16.3, p = 0.02) confirming the usefulness of FISH analysis when the spindle cell component is doubtful for malignancy. Additionally, the grade level of spindle cell component was closely correlated with p16 HD.

We evaluated the impact of BAP1 and HD p16 for the correct evaluation of the low-grade spindle cells component in separating an exuberant stromal cell reaction from a true sarcomatous component. The presence of HD (p16) was present in 62.5% (5 of 8) of the low-grade spindle cell component of biphasic type, whereas on these cases, BAP1 loss alone on the epithelioid component (4 of 8) or retained (1 of 8) was observed. These results showed the usefulness of HD (p6) by FISH analysis to identify a low-grade spindle cell component as a true sarcomatoid feature in those difficult cases for separating exuberant reactive stromal process from sarcomatoid component of a BMM.

The multivariate analysis showed that the transitional morphology remains in multivariate analysis a worse indicator of prognosis (HR = 3.2, 95% CI: 1.6 – 8.0, p = 0.003) even when compared to the presence of p16 HD on sarcomatoid component (HR = 4.5, 95% CI: 1.2 – 16.3, p = 0.02) (Table 3).

Table 3.

Multivariate Analysis Adjusted on Age

	Kaplan-Meyer Analysis			Cox Model
Variables	Median Survival	2-Year Survival (95% CI)	Log-Rank Test p Value	Chi-Square Test p Value	HR (95% CI)
Transitional pattern			p = 0.003
Absent	12 mo	14% (0–31)		1
Present	6 mo	0%		p = 0.01	3.2 (1.3–8.0)
Homozygous deletion			p = 0.08
On E component	13 mo	40% (0–64)		1
On E and S components	5 mo	5% (0–19)		p = 0.02	4.5 (1.2–16.3)

CI, confidence interval; HR, hazard ratio; E, epithelioid; S, sarcomatoid.

Discussion

This study was designed to answer questions raised by the staging committee of the IASLC regarding the reproducibility of the diagnosis of BMM, questions which have implications for treatment decisions. In addition, we correlated histologic features with survival and explored the significance of a poorly characterized variant of mesothelioma, the transitional variant. To our knowledge, this is the first interobserver agreement study between expert pathologists from different countries, all members of the International Mesothelioma Panel.

We showed that even when expert pathologists in the field of mesothelioma evaluate a diagnosis of biphasic type, the overall agreement between the experts from various academic institutions in different countries shows only a moderate agreement with a weighted kappa value of 0.45. Our findings continue to show the difficulty associated with the identification of BMM, mostly depending on the type and extent of materials submitted and to the major heterogeneity and plasticity of this cancer.

A retrospective study evaluating the effects of the amount of sarcomatoid component in BMM on the OS, previously analyzed by the French panel of experts from the MESOPATH National Reference Center, confirmed that the percentage of sarcomatoid component (>80%) was observed in patients with shorter survival (p = 0.02). The evaluation of different levels of percentages for the sarcomatoid component on OS was not significant in the current study due to the small number of cases (n = 42) and probably to the size or extent of tissue samples, and this finding confirms the difficulties in determining the exact minimum cutoff of sarcomatoid component in routine practice for treatment decisions and for evaluation of its exact role on OS.

On the other hand, a high-grade spindle cell component based on nuclear enlargement, prominent nucleoli, and atypical mitoses was an indicator of worse prognosis (p < 0.008). These results are similar to previous observations by Suzuki et al.³⁷ on epithelioid mesothelioma showing that nuclear atypia and mitotic count were found to be independent prognostic factors in a series of 232 tumors. In that study, univariate analysis showed that atypia (p < 0.001), chromatin pattern (p = 0.031), prominence of nucleoli (p < 0.001), mitotic count (p < 0.001), and atypical mitoses (p < 0.001) were indicators of poorer prognosis. By multivariate analysis, nuclear atypia (p = 0.012) and mitotic count (p < 0.001) were strong independent prognostic factors.

In a study of 77 EMM, Habougit et al.³⁸ also reported that nuclear atypia, mitotic count, atypical mitoses, nuclear atypia, prominent nucleoli, and necrosis were predictive of OS. This nuclear grading system can also be applied to the sarcomatoid component of BMM but did not appear to be an indicator of worse prognosis when compared to the transitional pattern in our series (see below).

Our study confirmed the importance of CK expression from the reviewers for the diagnosis of BMM. The pattern of expression was helpful in the assessment of the amount of sarcomatoid component. It is accepted that keratin expression can be observed in reactive fibroblastic processes. Nevertheless, in our study, the pattern, intensity, and distribution of the expression of the spindle cell component were considered by the panelists helpful to favor malignancy of the spindle cell component and to recognize a BMM pattern compared to the lesser expression in reactive fibroblastic processes but was not definitively diagnostic of a frank sarcomatoid pattern. In ambiguous cases, it was necessary to appeal to assessment of BAP1 loss or to the loss of CDKN2A(p16).

We also showed that the presence of homozygous p16 deletion on the epithelioid component of the BMM alone in 18% strengthens the value of FISH analysis in determining the malignancy of the sarcomatoid component in ambiguous cases, thus avoiding misdiagnosis of BMM. Moreover, in 33% of our cases with low-grade spindle cells components, we showed that HD (p16) was definitively more sensitive in favor of malignancy present in 62.5% (5 of 8) of the low-grade spindle cell component of biphasic type; whereas on these cases, BAP1 loss by IHC alone on the epithelioid (4 of 8) or retained (1 of 8) component was observed. These results are in agreement with previous reports published by Hwang et al.²⁹ and Wu et al.³¹ on challenges in differentiating sarcomatoid or desmoplastic mesothelioma from a cellular organizing pleuritis. They observed that BAP1 loss was seen in 3 of 20 (15%) and deletion of p16 by FISH was seen in 16 of 20 (80%) of such difficult lesion cases. Loss of one or the other marker was observed in 17 of 20 cases (85%). Moreover, the presence of HD (p16) was a stronger and more sensitive indicator of poor prognosis when observed on both component (E and S) by multivariate analysis (p = 0.02) compared when present only in the epithelioid component showing a better outcome. Our results show once again how important in the added value of FISH analysis is in routine practice for CDKN2A(p16) HD in the diagnosis of sarcomatoid component of BMM.

A major finding in our study provides ample evidence that the transitional pattern was identified among pathologists in 40% of all the diagnoses (209 of 524) (p < 0.0001). The presence of a transitional pattern was an indicator of poor prognosis, with a median survival of 6 months, 2-year survival of 0% (p = 0.01), and HR = 3.2 (95% CI: 1.3 – 8.0) (Table 3). Additionally, the OS of the patient presenting a transitional pattern was close to the OS of the sarcomatoid type and very close to the OS of the pleomorphic subtype of mesothelioma. Thus, the sarcomatoid, pleomorphic, and transitional types are similar in terms of survival despite of very different morphologies. For some panelists, some cases that had transitional areas also had other small areas of definitive epithelioid and sarcomatoid tumor, which raised the question of how much one must consider the impact of a transitional pattern on the OS.

These new data show the need to expand our knowledge of the pleomorphic, transitional, and solid variants of EMM by performing molecular analysis using RNA sequencing on each sample to evaluate how these patterns may cluster. The 2015 WHO classification clearly distinguished the epithelioid from the nonepithelioid (sarcomatoid and biphasic) types regardless of stage as indicator of prognosis. The divergent clinical behavior observed in the literature for some epithelioid types of MM might be related, adjusted based on comorbidities and stage to the presence of a transitional component. This pattern should be evaluated. Additionally, high metabolic activity at 18-fluorodesoxyglucose positron-emission tomography in the primary tumor has been considered to be associated with a shorter OS. Previous reports by Kadota et al.⁴² support the association of very high SUV max with nonepithelioid and pleomorphic epithelioid mesothelioma. The pre-operative standard update value (SUV) max on 18-fluorodesoxyglucose positron-emission tomography should be evaluated in the near future in correlation with the transitional component to confirm if this pattern will show the same association of very high SUV max as the pleomorphic aggressive variant do. This study based on morphology is in agreement with the previous study observed by Kadota et al.⁴³ and at the molecular level by Bueno et al.,⁴⁴ who showed that BMM is also molecularly recognized in two groups. Understanding the biology of mesenchymal transition and transitional patterns should help us to better discern these histologic types and subtypes of MM and open new opportunities for this aggressive and difficult-to-treat tumor. These results should be consolidated on a more extensive series and by extensive RNA sequencing analysis to better explore tumor growth, risk factors of histologic components for resectability, and responsiveness to chemotherapy or immunotherapy.

Conclusion

The interobserver concordance among the International Mesothelioma and French mesothelioma panel suggests clinical utility for an updated definition of BMM that allows better stratification of patients into risk groups for treatment decisions, systemic anticancer therapy, or selection for surgery or palliation.

We have also shown the usefulness of FISH detection of CDKN2A (p16) HD compared to BAP1 loss on the spindle cell component for the separation in ambiguous cases between benign florid stromal reaction from true sarcomatoid component of BMM. Taken together, our results of multivariate analysis further validate the concept of transitional pattern as a poor prognostic indicator.