Modern molecular techniques applied on tissue specimens have increased the interest in using archival material for research purposes. The risk of wrong diagnosis by doing so might be high due to the lack of immunohistochemical (IHC) evaluation at the time of diagnosis.

Malignant mesothelioma (MM) is a tumour that traditionally has been considered being difficult to diagnose histologically in a substantial subset of cases. To our knowledge, the extent of misdiagnosis of MM in archival material has not been studied systematically earlier, with the exception of one recent study 1 .

MM is a highly aggressive tumour defined as derived from mesothelial cells 2 3 . It is considered an almost incurable disease. Its molecular profile indicates that most of the known genes for radio- and chemoresistance are overexpressed 4 5 .

MM has traditionally been divided into epithelial, sarcomatous and mixed types, where the less common mixed and sarcomatous variants show a poorer survival. Sometimes a desmoplastic type is added, which may be diagnosed as either epithelial, mixed or sarcomatous type 6 7 . For the mixed type MM, International Mesothelioma Panel (IMP)/World Health Organization (WHO) require the presence of arbitrarily at least 10% of either an epithelial or a sarcomatous component 3 . According to recent studies, median survival after diagnosis is 4.5 to 17 months, depending on histological type, tumour stage, performance status and treatment, and other factors as gender and age 8 9 10 11 .

In Norway, registered new MM cases have continuously increased from five cases in the period of 1958-62 to 382 cases in the period of 2003-07. In the last period, 88% of registered cases were histologically verified. By adding the cytologically diagnosed cases the rate of morphologically diagnosed cases reached nearly 100% (97.5% of 360 cases in the period of 2001-05) 12 13 14 . However, in its recent guidelines the European Respiratory Society (ERS)/European Society of Thoracic Surgeons (ESTS) Task Force does not recommend making a MM diagnosis based on cytology alone because of the high risk of diagnostic error 15 .

Before IHC became a part of routine pathology in the end of the 1980s/beginning of the 1990s, and especially before more MM-specific positive IHC markers were introduced a few years later, it could be difficult or even impossible to differentiate MM from other malignant tumours, such as primary or metastatic adenocarcinomas (AC) in a part of the cases. Around 1995, antibodies against Wilms tumour protein (WT) 1, around 1998, against Calretinin and around 2005, against Podoplanin/D2-40 were introduced into routine diagnostics as new positive MM markers 16 17 18 19 20 21 22 .

The introduction of such more specific positive markers, in combination with distinctive negative markers has significantly improved the possible accuracy of MM diagnosis. However, there is no reliable single MM IHC marker. The IMP/WHO and the International Mesothelioma Interest Group (IMIG) recommendations on MM IHC diagnostics comprise a panel of at least two positive and two negative antibodies 3 23 .

This study aimed primarily at establishing histopathological diagnoses on archival tissue material from patients diagnosed as MM, in order to get a reliable basis for further studies using recent histological and IHC criteria. Furthermore, we wished to find out how the availability and choice of IHC markers may have influenced MM diagnosis.

The archival material in this study has been used in two previous studies where verification of the histological MM diagnoses was mandatory 11 24 . Use of the material was approved by the Regional Ethical Committee.

Paraffin-embedded tissue material from 73 patients taken between 1980 and 2002 was received from 19 Norwegian departments of pathology. Primary MM diagnoses had been given mostly based on tumour biopsies, and/or in a few cases on autopsy material. Tissue specimen volumes varied from a few mm³ in needle biopsies to some cm³ in autopsy material. Clinical and radiological data were fragmentary, and limited to those received in written form by the primarily diagnosing pathologists in the histology requisitions. After having received additional clinical information, tumour localisation was adjusted to both pleura and peritoneum in three cases, and to scrotum in one case, compared with earlier data 11 24 . From two of the four patients with MM both in pleura and peritoneum, autopsy material from both localisations was disposable, from the others the biopsies were only from one site.

From the paraffin blocks, Hematoxylin-Eosin-Saffron (HES) stained slides were made. For 61 patients (84%), the samples showed a sufficient amount of remaining tumour tissue suitable for diagnosis by the planned IHC procedure. Preliminary, exclusively HES-based tentative diagnoses were established. Basic histological parameters were determined on HES-stained slides. Mitoses were counted per so-called high-power fields (HPF), constituted by 40 times of objective and 10 times of ocular magnification.

After having determined optimal antibody dilution individually for each marker, by applying established IHC protocols from routine pathology and in the remaining cases by testing optimal dilutions, IHC was performed by the standardised Dako Cytomation protocol DAKO REAL™ EnVision™ with peroxidase/diaminobencidine and chromogen, rabbit/mouse 25 . The minimum procedure included four positive (Calretinin, Epithelial Membrane Antigen (EMA) membranous, Mesothelin, Podoplanin) and two negative (carcino-embryonic antigen (CEA), Ber-Ep4) markers, and in a subset of cases, if considered necessary, additional IHC markers (each case including, as a minimum, Hector Battifora Mesothelial Cells (HBME) 1 and Thrombomodulin as positive and CD15 and Sialosyl-TN as negative markers, before applying Mesothelin and Podoplanin) and Alcian Blue staining. In order to reach a reliable MM diagnosis, in five cases of epithelial type MM (No. 30, 39, 62, 74, 75) (12%) and in two cases of mixed MM (nr. 11 and 25) (25%), this second set of four basic IHC markers was applied. Reasons for this were too weak and/or not typical immunoreactivity (IR).

Slides were evaluated in multiple steps defined by the used basic and additional antibodies, by two pathologists (E. Larsson, H. Sandeck) independently, finally by a third pathologist (H. Willén), before establishing a diagnosis without access to the previous histological diagnoses. After each step, the diagnosis or eventual markers to add were determined by consensus. Our MM selection criteria included the histological MM variants and subtypes according to the current WHO MM classification and clearly positive IHC staining in at least a part of the tumour tissue, or absence of staining by the mentioned positive and negative markers, respectively 3 . For the diagnosis of a sarcomatous type or component, we regarded a predominant spindle cell pattern as indispensable 26 . A definitive entity diagnosis was only given when there was a high degree of likelihood for the entity based on morphology and IHC, with positive IR for at least one positive MM marker (Table 1).

Cases with negative IR were coded 0 and those with positive IR were classified into quartiles (H. Sandeck). Sensitivity (number of MM cases with positive IR divided by total number of MM cases that underwent IHC for the marker in question) and specificity (number of carcinoma reference cases with negative IR divided by total number of carcinoma reference cases for the marker in question) were calculated for positive and negative MM IHC markers.

Survival data were obtained from the Cancer Registry of Norway. These and the microscopic data were processed in the statistical software SPSS, v. 16.0 (Producer: SPSS Inc., IBM Company, Chicago, USA), with survival being a dependent variable. Survival dependent on each single histological and IHC parameter and on a few basic clinical data was determined, and visualised by Kaplan-Meier plots. Parametric linear (Pearson) and non-parametric (Kendall's tau-b, Spearman's rho) bivariate correlations to survival were analysed with regard to significance. A multivariate linear model was calculated.

In this study, we re-evaluated a Norwegian tissue material with an earlier diagnosis of MM which was regarded as potentially suitable to further molecular studies aiming at a deeper understanding of MM biology 4 11 24 .

All patients were donors to the Norwegian Janus serum bank, as the only applied selection criterion of the MM material 33 .

Our material is not representative concerning MM type distribution, as there have been diagnosed more mixed and sarcomatous types in larger newer studies (Roberts et al. 2001: 73% epithelial/11% mixed/10% sarcomatous/6% desmoplastic; Borasio et al. 2008: 67% epithelial/23% mixed/10% sarcomatous), so far as one supposes that similar diagnostic criteria were applied, which is not necessarily the case 32 9 . However, when including the four cases of possible sarcomatous type in our study, proportions would be more similar (n = 40/8/4; 77%/15%/8%). MM organ distribution seems to reflect the overall incidence 12 .

Due to scarce material, some relevant antibodies as WT-1, h-Caldesmon, and others, as well as specific histochemical reactions on glycoproteins and hyaluronic acid could not be applied, despite of their diagnostic value 34 35 36 . More extensive diagnostics including genetic testing, fluorescence in situ hybridisation (FISH) and electron microscopy (EM) in order to possibly reach a final entity diagnosis in the unclear cases, were not carried out, since this was not a concern in this study where the main point was to exclude all insecure and non-MM cases by a high degree of likelihood 37 38 23 39 .

Our choice of Calretinin, EMA, CEA and Ber-Ep4 was determined by their diagnostic value, as recently established by Brockstedt et al. who singled out CEA and Calretinin as the most informative markers in differential diagnosis between MM and AC 31 . Their next most valuable markers were CD15 and EMA, while Ber-Ep4 was fifth. The reason to choose Ber-Ep4 instead of CD15 as one basic negative marker was to test whether Ber-Ep4 really could replace CD15 in this constellation, as it according to our knowledge apparently has been used much more. However, direct comparison was only possible in seven cases where also IHC on markers no. 5-8 (HBME-1, Thrombomodulin, CD15 and Sialosyl-TN) was performed. As there was partial positive IR for Ber-Ep4 in two verified MM cases, but negative IR for CD15 in all of the seven verified cases, we too experienced that CD15 was a more reliable marker than Ber-Ep4. In four of the seven cases (no. 11, 25, 39, 62), CK5/6 and TTF-1 were considered necessary for confirming the diagnosis. Mesothelin was chosen in order to compare IR on MM tissue with serum levels and to evaluate it as an IHC marker, Podoplanin for the latter reason 4 11 30 40 20 21 .

WHO/IMP demand a minimum of 10% sarcomatous cells to define a mixed MM. In order to reduce MM typing problems, it has been proposed that at least 30-40% of a sarcomatous component should be present to call it a mixed type 41 . This would also be appropriate in order to avoid taking sarcoma-like stromal reactions as true sarcomatous MM components. However, in small biopsies, even such a component ratio may happen to be not representative at all. Which cytological and histological patterns are considered to represent a sarcomatous MM component or type is also varying 26 42 . Such variations in opinions, approaches and definitions may cause significant differences in the MM type frequencies making it difficult or impossible to compare different studies.

We detected a 15% discrepancy of former histological MM diagnosis. This could be explained by either 1) lack of IHC (n = 1), 2) use of non-specific IHC markers (n = 4), 3) differences in IR (n = 1) or 4) not appropriate conclusions based on appropriate markers (n = 3) influencing the primary diagnosis.

The case of the first group was from the period before 1998 when no specific IHC markers were available for actual differential diagnostics in routine pathology. Today it would be quite uncommon not to use IHC at all in MM diagnostics, at least in industrialized Western countries. At the same time, this illustrates the problems that may occur in countries with a less developed health care system where IHC might not be affordable.

As to the non-specific markers used in the second group that covers a period until 1998, this was due to the lack of more specific markers.

Differences in IR as in the third group and not appropriate conclusions as in the fourth group may occur also today. The former may be reduced by the use of standardized equipment and workflow in IHC laboratories, e.g., by IHC staining machines.

In two cases where there had been used non-specific markers in primary IHC (No. 31, 72), the later (1999) introduced marker TTF-1 was crucial to establish the pulmonary carcinoma diagnosis. This also happened in three of the cases that had lacked IHC (No. 3, 6, 9) (table 4) 43 .

Sarcoma, not further specified, was a differential diagnosis in five of the six cases with undetermined re-evaluation diagnosis, together with sarcomatous MM of pleura/lung in four of them. This reflects the potential diagnostic difficulty in a part of the sarcomatous lesions of pleura and lung even when IHC is available. In the one CK-negative (no. 32) and in the four CK-positive (no. 5, 49, 57, 63) spindle cell tumour cases, Comparative Genomic Hybridisation (CGH) might have allowed distinguishing between sarcomatous MM and other entities 37 .

The high percentage of cases in which university departments were involved amongst the cases not confirmed by us (8/9, 89%, with respect to histology) may be partially due to the generally high rate of MM diagnoses in which a university department was involved (52/61, 85%). Amongst these cases, there were no consultation cases sent from non-university departments. Apparently, difficult cases were sent from non-university departments to university departments for a second opinion, or primary diagnosis. However, in none of the not confirmed cases, a second opinion was asked for.

Whereas there does not seem to exist poorly differentiated epithelial type MM without any expression of one or more of such positive markers, as, e.g., Calretinin, WT-1, Mesothelin or Podoplanin, the complete absence of IR for such markers is allowed for the sarcomatous type by the above mentioned definition, as far as there is a positive IR for broad-spectrum cytokeratins. This would indeed mean that, with increasing dedifferentiation, specific mesothelial markers disappear, while cytokeratins are preserved even in poorly differentiated sarcomatous MM. However, in cases with completely negative IR for mesothelial markers, there would not be any sufficient histological evidence for the presence of MM. The clinical and radiological picture would then play a major role in tentative diagnosis.

In a study by Blobel et al., all forms of MM expressed CK 8 and 18, and most of them CK 7 and 19, the same CK profile as in AC 44 . In addition, CK 5 was expressed in all epithelial and most of the mixed types. CK 4, 6, 14 and 17 were also present in variable amounts (see also 45 ). The authors concluded that MM has to be defined as a variant of carcinoma. Histologically, this is indeed a reasonable conclusion even from today's point of view 46 . Sarcomatous MM could then be regarded as mesothelial sarcomatoid carcinoma. On the other hand, several sarcomas also express cytokeratins.

In our material, 19 MM diagnoses (39%) from the whole period (1980-2002) which were established without any use of IHC could be confirmed by us. All of them were also tentatively diagnosed by us as most likely MM in HES only slides. Apparently, the microscopic pictures in these cases were sufficient in order to recognize epithelial or mixed type MM. Related to our final diagnoses, in about 94% of the cases of epithelial MM, our preliminary, tentative HES-based entity diagnoses were correct. There would remain about 6% of diagnostic failure regarding the epithelial subtype alone. Most often this would concern the differential diagnosis between MM and AC, which may have therapeutic implications. In the differential diagnosis of sarcomatoid lesions, without IHC, there would be a much higher error rate when the pathological diagnosis is based only on HE or HES stained slides. This underlines the value of IHC in MM diagnosis, especially if there is any doubt concerning the HES picture.

Before more specific MM IHC markers were introduced, in the confirmed MM cases, marker combinations consisting of CEA and Pan-cytokeratin or Vimentin were applied. Here, negative IR on CEA and the morphological picture seem to have been crucial for the MM diagnosis, since there may be positive IR of both Pan-cytokeratin and Vimentin in carcinomas and sarcomas. Specific positive MM markers were not available until approximately 1998 when Calretinin was introduced in Norway.

All of our confirmed cases showed at least focally positive IR for Calretinin. However, there were no cases amongst the not confirmed cases in which missing Calretinin IR would have been the only exclusion criterion [see Additional file 1].

A basic IHC marker combination of two positive (e.g., Calretinin, EMA membranous) and two negative (Ber-Ep4, CEA) markers was well applicable to distinguish most of epithelial MM from other differential diagnostic possibilities. When it is necessary to distinguish MM from pulmonary AC, and when a synovial sarcoma can be ruled out, positive IR for TTF-1 would be useful for a pulmonary carcinoma diagnosis. With the exception of EMA, all of the mentioned markers are among the mesothelial and epithelial markers focused on by the WHO panel 3 . Meanwhile, positive membranous IR for EMA scored high as a positive MM, or mesothelial marker in the study of Brockstedt et al. that is based on a large MM material, and it is also mentioned as usually negative in benign mesothelial proliferations in the recent IMIG recommendations 31 23 . EMA is also a recommended marker in the recent ERS/ESTS guidelines 15 .

Based on our results, only Calretinin and CEA showed 100% sensitivity (Table 6). In two cases, only a few MM cells showed positive nuclear IR for Calretinin. One of them (no. 25) was a mixed MM of tunica vaginalis testis. In a study of Winstanley et al. on 20 tunica vaginalis testis cases, all of them showed negative IR for CEA and positive IR for Calretinin, but only 16 (80%) nuclear IR for Calretinin 47 . Our second case (no. 75) was an epithelial MM of pleura that also showed only a few cells with positive membranous IR for EMA.

Because of the predominating purpose of re-evaluation, the used tissue material was not set up against a material of non-MM tumours relevant in differential diagnosis. Thus, our results on specificity of MM IHC markers cannot be generalized, although the results do reflect the historical development of MM IHC diagnostics.

For epithelial MM, IMP and IMIG recommend a combination of at least two positive and two negative markers, and a broad spectrum CK mix. They point out that the preferred markers depend on the experience of the laboratory, and that more mesothelial, epithelial, vascular, and malignant melanoma markers can become necessary if the results are not conclusive 6 23 . In our study, in the first IHC step, we followed these recommendations, except for the broad spectrum CK antibodies. However, we used three broad spectrum CK antibodies (AE1, AE3, KL1) together with other markers in some cases of possible sarcomatous MM where negative IR on these antibodies would indicate that the tumour was not a MM. For a diagnosis of epithelial and mixed MM, more than two positive and negative markers were considered being necessary in 12% and 25%, respectively.

Concerning the choice of IHC markers, Marchevsky and Wick point out the substantial value that odds rate calculations, systematic reviews and meta-analysis may have for the development of evidence-based guidelines for MM diagnosis 48 . This is even true if there are several possible applicable "optimal" marker combinations, dependant on the experience of the laboratory and the specific case.

A recent study on accuracy of pathological MM diagnosis based on a more than 5-fold larger Japanese material (382/73) set up as interdisciplinary re-evaluation shows with 15.8% (47/298) of all cases diagnosed by IHC a similar scale of diagnoses classified as "definitively not/unlikely" MM, compared with 11.5% (7/61) in our study [1]. The authors defined 80.5% (240/298) of the IHC cases as "probable/definite" MM, while the corresponding amount in our study was virtually the same, 80.3% (49/61). However, in the Japanese study, even cytological material and not furthermore processed histological slides were included; IHC was only performed in 80.6% (308/382) of the cases. As in our study, diagnostic accuracy was lower in sarcomatoid lesions. While there was no agreement in diagnoses in as much as 22.4% of all female pleural and 72.2% of all female peritoneal cases, the respective numbers in our study are 0% (0/6) and 50% (2/4).

Detailed data on exposition to asbestos or other possible MM risk factors were not available to us. By using census data from 1970 where occupations had been classified into having high, moderate or little/no asbestos exposure and expanding them to the 1960 and 1980 censuses, we have earlier made estimations on the possible asbestos exposure of nearly the same patient group (n = 47, two patients less than in the current study). Occupational asbestos exposure was likely in 12 patients, all men. However, these data could not be validated 11 .

Survival analysis revealed no surprising data in respect of recent oncological knowledge on MM or malignant tumours in general. The presence of atypical mitoses alone, irrespective of the number of mitoses, could possibly be an independent negative prognostic factor, but this has to be verified or falsified on a larger material. Any sex preponderance in MM survival could not be seen in our material. The results of multivariate linear analysis may not be representative because of the too low number of cases for this purpose.

Our results underline the need of histopathological re-evaluation of tissue biopsies with an earlier diagnosis of MM, especially those given before the introduction of at least one of the more specific positive MM markers that showed a good diagnostic reliability in our cases. In our opinion, a reliable IHC-based definition of the sarcomatous MM type has been absent for cases with negative IR for mesothelium-specific markers. We have considered such cases not being sufficient for a definitive pathological MM diagnosis, but if clinical and radiological findings support a MM diagnosis in such a setting, nevertheless, a MM diagnosis may be established. Differential diagnoses in our material were adenocarcinoma, pleomorphic and sarcomatoid carcinoma and sarcoma. The WHO/IMP and IMIG recommendations of at least two positive and two negative IHC markers can be affirmed by our study, as far as sensitivity is concerned. The marker choice has to be adapted to the specific differential diagnostic needs in each single case. In difficult cases, it may be useful to refer to a second opinion.

In six cases, despite of relatively extensive IHC application, we were not able to give a clear-cut entity diagnosis, which may highlight the need of novel diagnostic tools as biomarkers and other molecular analyses, as well as new criteria. Such new tools will also be needed for the more personalized approaches for cancer treatment in the future.

The authors declare that they have no competing interests.

HS participated in the design of the study, established histological diagnoses, processed the data and drafted the manuscript.

ODR participated in the design of the study and reviewed the manuscript.

KK provided data from the Cancer Registry of Norway and reviewed the manuscript.

HW established histological diagnoses and reviewed the manuscript.

EL participated in the design of the study, established histological diagnoses and reviewed the manuscript.

All authors read and approved the final manuscript.