|Home | About | Journals | Submit | Contact Us | Français|
Sarcomas represent a heterogeneous group of tumors. Accurate determination of histological diagnosis and prognostic factors is critical for the delineation of treatment strategies. The contribution of second opinion (SO) to improve diagnostic accuracy has been suggested for sarcoma but has never been established in population-based studies.
Histological data of patients diagnosed with sarcoma in Rhone-Alpes (France), Veneto (Italy) and Aquitaine (France) over a 2-year period were collected. Initial diagnoses were systematically compared with SO from regional and national experts.
Of 2016 selected patients, 1463 (73%) matched the inclusion criteria and were analyzed. Full concordance between primary diagnosis and SO (the first pathologist and the expert reached identical conclusions) was observed in 824 (56%) cases, partial concordance (identical diagnosis of connective tumor but different grade or histological subtype) in 518 (35%) cases and complete discordance (benign versus malignant, different histological type or invalidation of the diagnosis of sarcoma) in 121 (8%) cases. The major discrepancies were related to histological grade (n = 274, 43%), histological type (n = 144, 24%), subtype (n = 18, 3%) and grade plus subtype or grade plus histological type (n = 178, 29%).
More than 40% of first histological diagnoses were modified at second reading, possibly resulting in different treatment decisions.
Sarcomas are malignant tumors developing in soft tissue, bone, skin or internal organs. The large majority of soft tissue tumors are benign and 100 times more common than malignant lesions [1, 2]. The 2002 World Health Organisation (WHO) classification has identified >50 histological subtypes of soft tissue sarcoma (STS), all of them being rare tumors . Accurate diagnosis is therefore difficult for nonspecialized pathologists and, together with the rarity of the disease, this results in inappropriate medical management in >70% of the patients [4–6]. Careful pretreatment evaluation is therefore essential for accurate diagnosis and appropriate treatment decision making [7, 8].
Discrepancies between pathologists have been reported frequently [9–12]. Although immunohistochemistry, FISH and molecular biology are important tools for diagnosis, these techniques are not routinely available in all laboratories and their use requires pathologists with expertise in molecular biology.
Second opinion (SO) in diagnostic pathology has recently received considerable attention as a result of efforts to enhance institutional performance and reduce medical errors . However, the mechanisms by which SO is obtained greatly influence the results . SOs given by another institution or a specialty panel at the time of patient referral produce highly discordant results compared with analysis of cases referred to experts for review . In the case of expert review, discrepancies are not viewed as ‘misdiagnosis’ but as an acknowledged need for assistance. In a previous analysis conducted in a French region, we showed that only 54% of included patients had full concordance between primary diagnosis and SO . In view of these results, we decided to extend this population-based study for one more year and to two additional European regions. The aim of the present work was therefore to evaluate the benefit of SO by regional/national experts for all sarcoma cases diagnosed in Aquitaine (A) and Rhone-Alpes (RA) in France and Veneto (V) in Italy.
Central review was done within each of the three regions by regional/national experts. In some rare cases, the experts wished the diagnosis to be either reviewed by another expert or discussed at panel meetings (FSG-GETO; CONTICANET) and a final consensus was reached after this third level of review. Before proceeding with the workshop review, all the pathologists from the three regions involved in this ‘central review’ attended training sessions in order to homogenize the review process.
A is a western French region of ~3.1 million inhabitants and 5 ‘departments’ (i.e. the administrative and geographical unit in France), with 9 public health structures (2 university hospitals, 1 comprehensive cancer center, 6 general hospitals) including 37 pathologists. There are also 55 pathologists employed in 19 private centers. RA is the second largest region in France, with ~6.1 million inhabitants and 8 ‘departments’. It has 15 public health structures (3 university hospitals, 1 comprehensive cancer center, 11 general hospitals) with 59 pathologists and 28 private structures with 80 pathologists.
V is the eighth largest region in Italy and the fifth region in terms of population, with ~4.9 million inhabitants. There are 56 public health structures (2 university hospitals, 1 cancer center, 53 general hospitals) with 25 pathologists and 10 private structures with no pathologist.
All pathologists working in these regions agreed to cooperate in this scientific study. All suspected cases of sarcoma (soft tissue and visceral tissue sarcomas diagnosed during the different study periods) were collected [2, 17, 18] and reviewed by regional and national sarcoma experts (members of the group of pathologists of the GSF-GETO for France and pathologists for V).
For each patient, a copy of the original histology report and histopathological specimens (hematoxylin eosin saffron-stained sections and paraffin-embedded tissue) representative of the tumor sample were provided. For all included patients, immunohistochemistry (and/or molecular biology analysis) was carried out by the referent pathologist.
The French Sarcoma Group/French Federation of Cancer Centers grading system was used in this study , as now proposed in the WHO classification. Grade was rated ‘not applicable’ for some specific histological types (e.g. Kaposi sarcoma) or when the grade could not be determined (e.g. on biopsy specimens).
For all included cases, the pathologist investigators were systematically offered expert SO and discrepancies between readings were analyzed. Two groups of patients were distinguished. The first group (requested SO) corresponded to patients examined by a ‘nonexpert’ pathologist who requested a SO from experts to confirm initial diagnosis at the time when the case was declared. The second group (‘systematic review’ or control group) included patients whose tumor samples were analyzed by a ‘nonexpert’ pathologist who did not request confirmation of diagnosis by experts and whose findings were reviewed only in the context of the present project. The final results were released to the initial pathologist for all cases identified as ‘requested SO’ but not for those included in the systematic review group. The final results of the study (poster presentation at the Connective Tissue Oncology Society meeting) were made available to all participants.
Ethical approval was obtained in both countries according to national laws and a signed informed consent to participate was obtained from Italian patients.
To be eligible for the study, patients had to meet the following inclusion criteria: first, diagnosis of connective tumor, according to the 2002 WHO definition, and no previous treatment, second, sarcoma diagnosis between 1 March 2005 and 28 February 2007 in RA, between 1 January 2007 and 31 December 2008 in A and V. No limitation was placed on disease stage or patient age. Exclusion criteria were as follows: diagnosis of relapse or a diagnosis other than soft tissue or visceral sarcoma (i.e. low-grade phyllodes tumor, bone sarcoma). Similarly, patients were not included when the primary diagnosis had been established in one of the reference centers by one of the experts or if the tumor specimen was not sent to the expert or when there was not enough tumor material available for immunohistochemistry or molecular analysis.
To ensure exhaustiveness, several controls were established in each region, throughout the study, by comparing the list of registered patients with (i) the list of all sarcoma cases reviewed by the expert pathologist during the inclusion period, (ii) the list of all medical records collected by the multidisciplinary sarcoma committee and of all patients registered by the Medical Information Department of the reference center (only in RA), (iii) the list of pediatric patients with sarcoma obtained from the pediatric registry of the RA region  in order to confirm the number of patients aged 15–18 years and (iv) the list of all sarcomas diagnosed by initial pathologists. Then we excluded patients with exclusion criteria.
Differences between the first diagnosis established by the nonexpert pathologist and the SO given by experts were evaluated and scored on a three-point scale:
In some rare cases, the expert wished that the diagnosis was reexamined by another expert or discussed at panel meetings and a final consensus was reached.
From 2007 to 2009, several meetings were organized between the different experts to specify the methodology for confirming review results and, according to the definition, to homogeneously classify the reviews.
All sarcoma cases diagnosed by pathologists of the three regions were reported, and data were statistically analyzed to quantify inter-observer differences and determine their nature. To evaluate patient characteristics and diagnostic concordance, categorical data were analyzed using Pearson's χ2 test or Fisher's exact test, as appropriate. Continuous data were analyzed with Student's t-test. The statistical significance level was set at P = 0.05 in a two-sided test.
Comparisons were also made between the requested SO group and controls. The χ2 test was used to determine the level of concordance and the type of discordance. Correlations between the most frequent causes of error and groups were analyzed using the Kappa test. All analyses were carried out using SPSS® (version 17.0) and SAS® software (SAS Institute Inc., Cary, NC).
All values reported hereafter for grade, histology and type or site of sarcoma are those obtained after expert review.
Of 2016 patients initially screened by pathology laboratories, 209 (10%) were excluded, either because they corresponded to a local or metastatic relapse (n = 11, 5%) or because the patient did not live in the regions under study (n = 90, 43%) or had not been diagnosed during the periods of inclusion (n = 108, 52%). Among the 1807 selected patients, 202 (11%) were further excluded because initial diagnosis had been established in the reference center by one of the experts. Of the 1605 remaining patients, 139 (8%) were excluded because there was no or not enough tumor tissue available for a second histological analysis and 3 were excluded (1%) because the initial histological report was not available. Finally, 1463 (73%) patients were eligible for analysis.
The first diagnosis of sarcoma was carried out in private practice for 675 (47%) patients and in public laboratories for 758 (53%) patients. In V, all first diagnoses were carried out in public structures, compared with 25% in Rhone-Alpes and 43% in A (P < 0.001).
Patients' characteristics are reported in Table Table1.1. Median age was 62 years (range, 15–96): 64 years (range, 15–96) for males and 63 years (range, 16–96) for females. Interestingly, while the incidence rate was similar between the three analyzed regions (previously reported in ), the characteristics of the included cases were statistically different between the three regions concerning site, histology, histological grade and type of tumor.
Concordance analysis was carried out on 290 (20%) sarcoma biopsy specimens (micro biopsy or open biopsy) and 1167 (80%) surgical samples. Data were missing for six patients (0.5%). Concordance data regarding grade, type of tumor sample (biopsy versus surgery) and type of laboratory (private versus public) are given in Table Table2.2. Concordance analysis showed 121 cases with zero agreement (8%), 518 with partial agreement (35%) and 824 with full agreement (56%). The ratio of total concordance was higher in grade I than in grade II–III tumors (P = 0.001) and for tumors where a SO was not requested (P < 0.001); it also varied with the type of laboratory (public versus private), the site of sarcoma (visceral versus STS) and by region (V versus the other two regions) (P = 0.001). Interestingly, no significant differences were reported according to the type of tumor sample (Table (Table22).
Causes of nonconcordance are presented in (supplemental Table S1, available at Annals of Oncology online) and concern most frequently the diagnosis of benign tumors or sarcomas which happened to be carcinomas or melanomas at SO. Partial discordance between initial diagnosis and review was essentially related to grade and histological subtype. Sixty (50%) histological diagnoses were modified from malignant to benign or conversely (more specifically in V), 49 (40%) from sarcoma to other histological type (carcinoma) or conversely (more specifically in RA) and 12 (10%) from STS to GIST or conversely (more specifically in A) (supplemental Table S2, available at Annals of Oncology online).
The most frequent discrepancy was related to the grade of the tumor: (i) no grading by the diagnostic pathologist (n = 479) while the expert attributed grade 1–3 (n = 235, 57%); (ii) grading error, with grade 1 attributed by the diagnostic pathologist (n = 207) versus grade 2–3 by the expert (n = 39, 19%). Other (n = 241, 73%) discrepancies were related to grade (I versus II, II versus III) and histological subtype. The major cause of discrepancy was actually the lack of grading by the first pathologist (n = 235, 57%).
The second most frequent discrepancy was related to the histological type (supplemental Table S1, available at Annals of Oncology online). There was no correlation between concordance and type of tumor sample or tumor site.
Patient characteristics in the two groups are compared in Table Table3.3. The patients for whom initial diagnosis had been made from biopsy specimens were most frequently not proposed for a SO. Likewise, female patients, visceral tumors, abdominal, unknown or high-grade tumors, GIST and Kaposi were statistically less proposed for a SO.
In fact, initial pathologists asked for a second expert opinion principally for difficult diagnoses or cases requiring molecular analysis (well-differentiated liposarcoma, GIST, primitive neuroectodermal tumor (PNET), synovialosarcoma; n = 245). The region, the type of tumor, as well as tumor site and grade also influenced the decision of the first pathologist to request a SO.
Concordance results shown in Table Table33 demonstrate that total discordance was significantly lower in cases where the primary pathologist asked for a SO as compared with cases where expert SO was not requested spontaneously (13% versus 6%; P < 0.001). Full concordance was obtained in 40% (n = 230) versus 66% (n = 594) of cases in each group, respectively.
The primary purpose of this study was to discuss the importance of highly specialized skills in pathology to ensure correct diagnosis of sarcoma. Concordance between first assessment and second expert opinion as well as the various clinical/pathological-associated factors have been analyzed. The recent progress achieved in the understanding of sarcoma (especially after the introduction of the 2002 WHO histological classification) has emphasized the need for specialized expertise for the diagnosis and management of these rare but really aggressive tumors.
SOs have been suggested to increase the chances of establishing a correct histological diagnosis , but the magnitude of the improvement had not been tested in large region-based cohorts. While an accurate diagnosis is essential to ensure appropriate management of patients with sarcoma, in particular with new targeted therapies, nonconcordance was reported in 44% of the cases included in our comprehensive 2-year cohort of sarcoma patients.
This study confirms that centralized pathological review improves the quality of diagnosis which, because sarcoma is a rare and heterogeneous tumor, is very difficult to establish. Nearly 42% of first diagnoses were declared inaccurate by the panel of experts conducting the centralized pathological review. However, <10% of initial diagnoses were considered major misdiagnoses with direct impact on patient care. The most frequent discrepancies identified were related to tumor grade and histological type. Exact determination of the tumor type and grade is crucial for making individual treatment decisions and subsequently improving patient outcome [21, 22]. In fact, determining the grade of the tumor may be important to decide between adjuvant chemotherapy or not, neoadjuvant chemotherapy or not or radiotherapy or not. Although the clinical practice recommendations published in France have confirmed that the tumor grade must be included in the histological report generated at the time of diagnosis , this information is frequently not provided; it was missing in 34% of the cases (n = 477) in the present study, as also reported in other studies . Other studies have demonstrated that the reproducibility of grade is very difficult to achieve [21, 24]. On the other hand, when nonexpert pathologists participating in the study did provide the grade of a tumor, their evaluation was generally correct. Pathologists may sometimes simply lack experience and would benefit from training sessions organized in the framework of continuous medical education .
The central question that arises is whether all sarcoma cases should be reviewed in a specialized center. Importantly, all pathologists evaluating sarcoma patients should be able to use immunohistochemistry for confirmation of diagnosis. For certain histological types (translocation sarcomas, GIST, dermatofibrosarcoma protuberans), molecular biology can also contribute to the establishment and/or the confirmation of diagnosis [26–28].
In the literature, the proportion of diagnostic errors in patients with soft tissue sarcoma ranges between 25% and 40% [7, 12, 29–31]. Surprisingly, since the first published reports on SO for sarcoma tumors in 1980 and despite the introduction of new tools (immunohistochemistry, molecular biology etc.) and the development of educational workshops, the percentage of concordance has remained remarkably unchanged . The present study, conducted over a 2-year accrual period in three different regions, describes for the first time a lower level of discordance than previously reported in the literature. Also, the difference between initial diagnosis and SO was statistically lower from 1 year to the next, possibly due to the increasing experience of initial pathologists with time.
Another concern is the proportion of nonconcordance which is similar in the two French regions whereas it is lower in Italy. A possible explanation could be that, over the past 20 years, almost all the general pathologists of V have been trained at the same specialty school (Padova), a school of pathology with particular expertise in soft tissue tumors.
Methodological biases cannot be excluded for this study. For instance, indicating ‘suspicion of sarcoma’ as an inclusion criteria may have led to underestimating the proportion of true sarcomas misdiagnosed as carcinomas or melanomas before SO. The organization of diagnosis at the regional level may also have induced a recruitment bias and prevented exhaustive inclusion (especially in V).
Some cases (rare type of sarcoma, intermediate or high histological grade, PNET) were more frequently referred for a SO by the first pathologist. On the other hand, diagnoses of Kaposi were systematically fully concordant and could thus be considered not candidates for a systematic review. Using such methods could possibly yield, in a near future, results similar to those reported for non-Hodgkin's lymphomas where the majority of B-cell non-Hodgkin's lymphoma diagnosed in the community have remained unchanged after SO review by an expert hematopathologist .
Taken together, results of this population-based study show that SO enables to improve the accuracy of diagnosis in sarcomas. Given the therapeutic implications of a misdiagnosis, in particular for benign lesions, carcinomas or sarcomas requiring chemotherapy at an early phase, it is logical to propose SO as a standard management procedure for connective tissue tumors. This is currently implemented within the RREPPS research program in France (rreppps.org), which collects all sarcomas cases for centralized review.
The authors are grateful to Marie-Dominique Reynaud and Delphine Baconnier for editing assistance, to the data manager Muriel Rogasik and to all pathologists of the Rhone-Alpes, Aquitaine and Veneto regions for their active collaboration in the study—Rhone-Alpes: Agard Catherine, Allias-Montmayeur Fabienne, Angonin Régis, Augros Marylin, Bailly Christiane, Balme Brigitte, Bancel Brigitte, Barnoud Raphaelle, BenLagha Nadia, Benabidallah Samir, Berger Gerard, Beschet Isabelle, Beurlet Jacques, Billard Françoise, Bland Vincent, Bonin Anne-Marie, Bottero Noelle, Bourloux Jocelyne, Bouvier Raymonde, Bozon Catherine, Brambilla Elisabeth, Bringeon Béatrice, Buenerd Annie, Cantero Brigitte, Cavaillés Catherine, Chalabreysse Lara, Chalabreysse Philippe, Chambonniere Marie-Laure, Chanoz Jacques, Chanoz-Poulard Geneviève, Chassagne-Clément Catherine, Chevallier Michèle, Chouvet Brigitte, Ciapa Agnès, Claret-Tournier Catherine, Clemenson Alix, Collardeau Frachon Sophie, Corrand-Faure Anne, Corsois Laurent, Crozes Carole, Cruel Thierry, David Catherine, De la Fouchardiere Arnaud, Decaussin-Petrucci Myriam, Decouvelaere Anne-Valérie, Denier Jean-François, Der Garabedian Philiberte, Derolland Philippe, Descombes-Thivolet Brigitte, Dieny Anne-Florence, Dijoud Frederique, Donne Chantal, Donsbeck Anne-Valérie, Douchet Catherine, Dumollard Jean, Dusserre Isabelle, Economides-Jarsaillon Ariane, Elbaz Nadine, Fabre Blandine, Faisant Monique, Faure Anne, Faure Claire, Faysse Michaelie, Feutry Catherine, Fontaniere Bernard, Frachon-Collardeau Sophie, Gasnier Plaweski F, Gentil-Perret Anne, Glehen Alexandra, Godard William, Godeneche Janique, Gouarderes Catherine, Gouzy-Grosjean Fabienne, Griot Annick, Guillaubey Colette, Guillaud Catherine, Herve-Nicollet Catherine, Hervieu Valérie, Isaac Sylvie, Istier Luc, Jouvet Anne, Kanitakis Jean, Kermanac'h Pascale, Khaddage Abir, Klein Laurence, Knopf Jean-François, Koeb Marie-Hélène, Labadie Michel, Lantuejoul Sylvie, Laurent Isabelle, Lauro Carole, Le Breton Frédérique, Lucht-Versini Pascale, Mabrut Marianne, Machayekhi Jean-Pierre, Maisonneuve Martine, Maisonneuve-Gilly Dominique, Mege-Lechevallier Florence, Meyronet David, Morcillo J-L, Muller Bernard, Muller Christine, Neyra Monique, Pasquier Dominique, Perrot Guy, Pialat Jean, Picchetti Nicole, Pinel Nicole, Pocachard, Pugens Gilles, Ranchère–Vince Dominique, Reis Borges Ruth, Richard Jacques, Roux Jean-Jacques, Roux-Gilly Marie-Georges, Saint Genis Luc, Saint-Pierre Ghislaine, Salameire Dimitri, Salle Monique, Salon Caroline, Serain Francois, Soubeyrand Marie-Sophie, Streichenberger Nathalie, Sturm Nathalie, Suignard Yves, Terdjman Pierre, Treilleux Isabelle, Vancina Serge, Vaunois Brigitte, Vercherin Axelle, Vitetta Franck, Vock-Bonnet Marthe, Youssef Nelly, Devouassoux Mojgan, Pasquier Basile, Peoc'h Michel, Scoazec Jean-Yves, Thivolet Bejoui Francoise. Aquitaine: Barre Isabelle, Perrodeau Françoise, Archambaud Frédéric, Familiades Pierre, Flamme Hélèna, Lechaux Marielle, Massot Bordenave Mireille, Chebrou Denis, Mazabrey Dominique, Coindre Jean-Michel, Mac Grogan Gaëtan, Soubeyran Isabelle, Stock Nathalie, Darrasse Dominique, Desrousseaux Marie, Seurat Pierre, Chassaigne Florence, Molimard Jeanne, Raynaud Olivier, Mérino Céline, Reau Patricia, Begueret Hugues, Deminière Colette, Dupin Camille, Fregeville Marie, Ip Kan Fong Harold, Lepreux Sébastien, Parrens Marie, Pellegrin Michèle, Rullier Anne, Trouette Hélène, Vergier Béatrice, Vital Anne, Beau Georges, Boisseau Annie, Laurentjoye Lise, Roche Comet Isabelle, Sanchez Henri-Pierre, Savin de Larclause Anne-Marie, Mannat-Guillon Chantal, Mercier Sylvaine, Saurel Jacques, Cambeilh Christian, Dauzan Yves, Lecat Marcel, Mendès Isabel, Beaudeau Patricia, Canihac Muriel, Augrand M, Julien Martine, Dauriac-Jeantet Anne-Marie, Kunzelamnn Valérie, Morard Jean-Louis, Nguyen Valérie, Ancelin Malbreil E, Laborie Vincent, Rouquette Pierre, Eyremendi RP, Gay G, Manent AM, Mergey Eric, Cales Valérie, Etcharry F, Levrier Anne, Bui Matthieu, Laillou Béatrice, Demoures A, Belleannée Geneviève, Chauvel Annick, Goussot Jean-François, Marty Marion, Taillat Floriane, Dachary Dominique, Merlio Christine, Penchet Isabelle, Segretin Catherine, Ganot Geneviève, Dufau Jean-Philippe, Giglio Lydia et Revault-Gaye Danièle. Veneto: Menestrina Fabio, Lestani Maurizio, Zamboni Giuseppe, Iuzzolino Paolo, Zucchelli, Guido Maria, Bittesini Lucia, Giordano Renzo, Farruggio Angelo, Barbazza Renzo, Murer Bruno, Visonà Alfonso, D'Amore Emanuele, Iannucci Antonio, Mencarelli Roberto, Sacchi Giuseppe, Poletti Alessandro, Vendraminelli Roberto, Pea Maurizio, Martignoni Guido, Chilosi Marco, Stracca-pansa Vincenzo, Rugge Massimo.
Presented in part at the 2009 CTOS meeting.
Merck Serono; the French National League against Cancer (Ain and Rhône committees); the patients advocacy group Detect, Analyse, Inform, Sarcomas (DAM'S); Infosarcome and Sarcoma Patient Euronet (SPAEN); grant of the Commission of the European Communities, Connective Tissue Cancer Network (CONTICANET FP6-018806); Lyon Integrated Research Cancer Centre (LYRIC); Sarcoma Clinical Reference Network (Netsarc); Reference Network in Sarcoma pathologies (REPPS). We also acknowledge grants from the French National Cancer Institute (INCa).
The authors declare no conflict of interest.