The exact overall incidence of sarcomas is unknown and the incidence of the different histological and molecular subtypes has not been determined precisely [7]. Most studies consider adults and children, or soft tissue and bone sarcomas separately. Results of population studies indicate that the incidence of sarcomas may actually be underestimated, possibly because of diagnostic confusion with carcinomas of the same organ. Recent reports of a 1.1–1.5/100,000/year incidence of gastrointestinal stromal tumors (GIST) [8] suggest that the overall incidence rates of 1 to 3/100,000/year generally reported are likely underestimated [9]. In addition, sarcomas encompass a wide variety of histotypes and molecular subtypes and are categorized in rapidly evolving phenotypic and molecular subgroups now used for sarcoma diagnosis with a growing impact on the management of patients [10].

The present study identified 748 new cases of sarcoma over a two-year period instead of the 200 per year expected. Of these, 98% were reviewed by regional and national experts in sarcoma pathology, all diagnoses were confirmed by immunohistochemistry, and 85% of cases with molecular alterations could be characterized using molecular techniques. On the basis of these numbers, the French and world age-standardized incidence rates were respectively 6.4 and 4.8 per 100,000 population, which is higher than the rates reported in previous publications, either in the USA or in Europe (between 1 and 3 per 100,000), even though some studies have reported higher results [20], [21]. In children under 15 years of age, the incidence of STS and bone tumors was estimated to be respectively 0.9 and 0.6 per 100,000 per year [22], [23]. In 2005, the number of new cases of cancer of all sites was estimated to 27,869 in the Rhone-Alpes region [24]. With a mean annual rate of 374 new cases in this study, sarcoma thus represents 1.3% of all new cancer cases in the region.

The work presented here differs from previous large retrospective studies which have used a different methodology, and from results published by cancer registries, from which incidence data have been extrapolated. Age-standardized incidence rates of STS are fairly constant in most areas covered by cancer registration, and range from 1–3 per hundred thousand population [25]. However, because of the rarity and heterogeneity of primary sites and presentations of STS, general registries do not provide routine data about their incidence. The main sources of incidence data for sarcoma are the world databases of the International Agency for Research on Cancer (IARC) [25], the American Surveillance, Epidemiology and End Results (SEER) Program [21], [26], the European database of the Automated Childhood Cancer Information System (ACCIS) [27] and the national coverage registry of the Nordic countries [28].

Because of the great number of cases recorded, diagnosis is not always reviewed by pathology experts, which is a serious limitation in the case of sarcomas, given the frequency of misdiagnosis with carcinoma, melanoma or benign tumor, or even between histological subtypes [29], [30]. The reproducibility of sarcoma diagnosis, and particularly of soft tissue sarcoma, is relatively poor across pathologists who are not familiar with these lesions, and the histopathological classification of this cancer in cancer registries is often inconsistent. Epidemiological studies have suffered from this misclassification of histology. A concordance study performed in Rhone-Alpes comparing primary diagnosis and systematic review by expert showed that 46% of diagnoses were modified at second reading and that in 19% of cases there was a discordance in the histological type [31].

In addition, the design of registries is not always suitable for sarcomas. The data collected are incomplete, mainly because of the broad diversity of morphological entities and of the classification of data per anatomic site which does not distinguish visceral sarcomas (e.g., GIST are counted with “digestive cancers”, uterine sarcomas with “uterine cancers”). Very few registry studies have focused and produced data on the histological classification of sarcomas; their results have shown that failing to include tumors arising in specific organs resulted in an underestimation of the overall incidence of sarcoma by 50% [21]. Pediatric registries have reported more accurate data on incidence rates of sarcoma in children and adolescents because the nosologic classifications for children malignancies is primarily based on histology [32], [33] and only few cases of visceral sarcoma arise in children or adolescents (two cases in our study).

Any comparisons with existing datasets may be affected by the specific population under study, by changes in sarcoma incidence over time, and by possible ascertainment biases. We excluded the possibility of an ascertainment bias because of the reputation and experience of the local clinician in the management of sarcoma. The place of residence at the first suspicion of sarcoma was collected for each patient and those who moved to Rhone-Alpes after the first diagnosis of sarcoma were excluded from the analysis. Moreover, our method of case ascertainment is likely to have missed some false negatives (i.e., sarcomas misdiagnosed as other cancer types) and the true incidence of sarcoma incidence might be higher.

Concerning the specific population under study, the overall distribution of sarcoma in Rhone-Alpes is not considered different from the rest of France [34]. Rhone-Alpes is one of the French regions with the youngest population (Source INSEE) though the age demographics are older than in the European and North American reference populations. Ethnic differences might account for some of the variance with published data. The overall incidence of STS is known to be higher among African Americans than among Caucasian patients [9], [21] while Ewing sarcoma showed a striking incidence by race with the great majority of cases occurring in white patients [35]. The Rhone-Alpes population is principally Caucasian, however, no ethnic information was recorded in the study. The bone sarcoma incidence rates reported in the present study and their age-specific distribution were similar to those recently reported in other countries with Caucasian population [36].

The time period under question may also account for the differences with published data. The obvious point of concern is the incidence of Kaposi's sarcoma, which has clearly varied with HIV prevalence over time and across the country. When excluding patients with Kaposi's sarcoma, the crude and world age-standardized incidence rates in our study were respectively 6.0 and 4.6 per 100,000/year. Previous reported incidence rates of soft tissue sarcoma were different amongst countries, even in neighbored European countries, with different distribution of the most common histological subtypes [37]. This was due to different inclusion criteria considering or not sarcoma of intermediate malignancy (e.g. dermatofibrosarcoma), to the small absolute number of cases and to inherent classifications. In the present study, we included GIST (a type of sarcoma with a specific biology) and all histological types of sarcoma described in 4 different WHO groups.

Because of changes and evolutions in the histological classifications of sarcoma, it is not really possible to compare the data collected over the decades. Whether the incidence of sarcoma increases worldwide is unclear: increased rates have been reported, sometimes due to increased incidence of Kaposi sarcoma [20], [38], but such increase was also observed in patients without Kaposi sarcoma [9]. Similar controversial findings are reported in childhood cancer, with some studies reporting increased sarcoma incidence rates [39] not confirmed by others [40]. It is yet unclear whether the incidence of sarcoma increases because of environmental or other behavioral changes, or whether there is only an apparent increase due to modifications in the registration process. Moreover, epidemiological studies are limited by the histopathological misclassification of these rare tumors in cancer registries, both between histotypes (sarcoma vs. other) and among histological subtypes (e.g., leiomyosarcoma vs. other). To test hypotheses regarding risk factors, one must be able to accurately measure disease incidence by age and by histological type. In parallel to the data collected by registries in large patient series, an exhaustive regional study based on morphological criteria is needed.

In addition to describing the overall incidence of sarcoma (all types, all ages), the present study also helps to further refine the estimation of the incidence of the different subsets. GIST was the more frequent histotype reported with a predominance in women, while other published series indicate a more mixed population or even sometimes a male predominance [41]. The incidence of dermatofibrosarcoma protuberans was unexpectedly high (5%) whereas that of synovial sarcoma was lower than expected (2%). In our study, 49% of all STS occurred in the limbs, which is consistent with results of previous site distribution studies. However, if all sarcoma subsets are taken into account, this site distribution changes, with the trunk becoming the primary site of occurrence (58%). The most common histological types in previous series were malignant fibrous histiocytoma (MFH), leiomyosarcoma, liposarcoma and fibrosarcoma. In the present study, the main histological types were GIST, unclassified sarcoma, liposarcoma and leiomyosarcoma. These type distribution differences may be due to recent advances in immunohistochemistry, cytogenetic and molecular biology. With the development of numerous new antibodies since the 1980s, immunohistochemistry has become the most accurate and reproducible tool for sarcoma diagnosis and has provided pathologists with new tests to distinguish between the different histotypes of sarcoma [42]. The distribution of sarcoma subtypes has significantly evolved since the publication of the latest WHO histological classification in 2002 which took into account the results of immunohistochemistry. Two large categories that existed before 2002, MFH and fibrosarcoma, have now become much rarer [43]. Furthermore, the incidence of GIST has increased from 0.2 to more than 1/100,000 in several European countries after the introduction in 2001 of anti-CD117 antibody for immunohistochemical staining [8], [41], [44]. Rates seem to have remained stable since then [44], which tends to prove that the higher number of GIST was due to the change in diagnostic methods and to the reclassification of many mesenchymal gastrointestinal tumors previously diagnosed as smooth-muscle tumors like leiomyosarcomas [45].

The development of molecular biology has also proven essential for the diagnosis of sarcoma subtypes and for the refined classification of sarcomas [46]. This technique is suitable for use in about 50% of sarcomas with specific known genomic abnormality. In addition to clinical presentation and morphology, molecular testing also contributes to distinguishing between malignant and benign tumors (e.g., liposarcomas and lipomas) or between sarcomas of similar morphology (e.g., round cell tumors). Molecular diagnosis has been used for several years in routine for pediatric round cell tumors, specific translocations being used as diagnostic markers.

Molecular biology has a growing impact and in the future, the molecular-based classification of sarcoma should be as important as the classification in histological subtypes. Molecular characterization already has clinical implications for some subtypes of sarcoma, either for prognosis (e.g., poor prognosis for patients with KIT exon 11 deletion in GIST, FKHR-PAX3 expression in metastatic rhabdomyosarcoma and SYT-SSX1 fusion type in synovial sarcoma) [47]–[49], treatment (adjuvant radiotherapy), or response to treatment (e.g. better response to imatinib for GIST with exon 11 mutation than with exon 9 mutation) [50]. In our population-based study, the incidence of PDGFRA mutated GIST was higher than previously reported in patients with advanced disease [50] indicating that tumors bearing mutant PDGFRA have a more indolent behaviour [19]. Moreover, Williamson et al. demonstrated that rather than histology, the key factor in terms of biology and clinical progression of rhabdomyosarcoma was the presence or absence of a fusion gene [51]. The molecular profiling approach used in our study was relatively standard and the technology evolves rapidly [52]. Novel molecular karyotyping techniques, such as array comparative genomic hybridization (aCGH) or gene expression analysis led to improve sarcoma classification by defining tumor-specific clusters that have potential value for resolution of differential diagnoses (e.g. wide separation of GIST from leiomyosarcoma) [53] and led to the identification of new diagnostic markers as well (e.g. DOG1 helpful in recognizing KIT-negative GIST) [54]. Full transcriptome sequencing and other genomic, proteomic and epigenetic profiling approaches should become available in the near future and will allow to better characterize the different histotypes of sarcoma and to better understand their complex genetic structure with numerous rearrangements. Anyway, molecular techniques have yet revealed that sarcomas were different entities with different biologies [54] and have allowed a better understanding of the pathogenesis of some types of sarcomas (e.g. initiating role of mutation of KIT or PDGFRA receptors) [55].

This study was based on the voluntary participation of the different pathology laboratories and the first data source was their spontaneous notifications. All efforts were made to ensure the accuracy of the results and the exhaustiveness of the collection. No list can be totally exhaustive and all offer different levels of quality, but the different cross-checks between lists tend to indicate that we have collected all the sarcoma cases diagnosed in the region during the study period. These good results can be attributed to the effective collaboration between the different specialists and referents.

Soft tissue, visceral and bone sarcoma represent three heterogeneous groups of mesenchymal neoplasms, with different methods of diagnosis, different classifications, different staging and treatment approaches and different management. Nevertheless, we have deliberately collected and grouped all histological types because only this accurate collection can ensure exhaustiveness. For example, the collection of all histological types of bone sarcoma allowed to collect cases of extraskeletal bone sarcoma (e.g. Ewing sarcoma) that would have been missed otherwise.

Sarcomas represent a heterogeneous group of malignancies which may occur at any site and any age. All types occur across the age spectrum [56] but distribution of histotypes during childhood is striking different with what is described in adults or in adolescent and young adult. Specific molecular events defined sarcoma histological types: most pediatric-type sarcoma are associated with translocation due to the high incidence of genetic factors unlike adult-type sarcoma, with more complexe karyotypes, for which environmental factors will tend to influence more often. Adult sarcomas have different prognostic factors [57] and significantly worse outcomes than children sarcomas [56], [58] due to inherent biological differences. Sarcomas present a specific epidemiology per histological and molecular subtypes leading to the conclusion that they comprise multiple aetiologically distinct entities rather than a single disease.