Epidemiology of Soft Tissue Sarcoma and Bone Sarcoma in Italy:AnalysisofData from15Population-BasedCancerRegistries

Sabrina Fabiano, Paolo Contiero, Giulio Barigelletti, Anna D’Agostino, Andrea Tittarelli, Lucia Mangone, Isabella Bisceglia, Salvatore Bongiorno, Lucia Elena De Lorenzis, Guido Mazzoleni, Marine Castaing, Teresa Intrieri, Antonella Puppo, Giuliano Carrozzi, Giovanna Maria Sini, llaria Fontanili, Lorenza Boschetti, Rosario Tumino, Francesca Bella, AnnaClara Fanetti, Maurizio Castelli, Antonio Ardizzone, Fabio Vittadello, Salvatore Sciacca, Gianfranco Manneschi, Claudia Casella, Claudia Cirilli, Filomena Pala, Maria Michiara, Simona Migliazza, Francesca Ferrari, Eugenia Spata, Ylenia Dinaro, Ivan Cometti, Anna Maghini, and Giovanna Tagliabue 1

frequent soft tissue sarcomas were sarcoma not otherwise specified (NOS) (29.4%) and liposarcoma (22.2%); the most common bone sarcoma was chondrosarcoma (37.6%). Soft tissue sarcomas occurred most frequently (35.6%) in lower limb connective tissue; bone sarcomas arose mainly (68.8%) in long bones. e frequencies of morphologies arising at different sites varied considerably by Italian area; for example, 20% of hemangiosarcomas occurred in the head and neck in south Italy with 17% at this site in the centre and 6% in the north. For soft tissue sarcoma, the highest ASRs of 2.6 and 2.4 contrast with the lowest ASRs 1.1 and 1.3, suggesting high-risk hot spots that deserve further investigation. e marked variations in morphology distribution with site and geography suggest geographic variation in risk factors that may also repay further investigation particularly since sarcoma etiology is poorly understood.

Introduction
Sarcomas are a heterogeneous group of malignant neoplasms of mesenchymal origin. ey are one of the major groups of rare cancers in Europe [1] accounting for around 1% of all cancers diagnosed [2]. e World Health Organization [3] recognises approximately 70 histotypes grouped into two main categories: soft tissue sarcoma and bone sarcoma. In Europe, soft tissue sarcomas are more frequent than bone sarcomas, with world age-standardised incidence rates (ASRs) of 1.5-3.0/100,000/year and 0.5-2.0/100,000/ year, respectively. Both types of sarcoma are moderately more frequent males than females [4].
Soft tissue sarcomas constitute about 84% of the sarcomas diagnosed in Europe [5]. Accurate diagnosis of these entities is challenging in relation to their rarity and morphologic heterogeneity [6]. Nevertheless, site of origin tends to correlate with certain histotypes. For example, liposarcomas arise mainly in the thigh of young adults, and myxofibrosarcomas often arise in the superficial soft tissue of elderly patients [7]. e most common soft tissue sarcoma histotypes in Europe have been reported as leiomyosarcoma (19%), liposarcoma (16%), and sarcoma not otherwise specified (NOS, 14%) [8].
Bone sarcomas constitute around 14% of sarcomas diagnosed in Europe [5]. Osteosarcoma, the most frequent bone sarcoma overall (ASR 0.3), generally arises in long bones, but the proportion arising at an axial site increases with age [9]. Chondrosarcoma and Ewing's sarcoma are the most frequent bone sarcomas in adults, with ASRs of 0.2 and 0.3, respectively [9], followed by chordoma (ASR 0.5/million/year), which arises from persisting notochord [9]. Ewing's sarcoma mainly occurs in children and adolescents but is also seen in adults; it arises in long bones (about 50% of cases), followed by the pelvis, ribs, and vertebra. e causes of most sarcomas are unknown. A recent study on 1,162 patients over 15 years with histologically confirmed sarcoma found that 55% had an excess of pathogenic (and potentially aetiological) germline variants [10]. Several risk factors are well established, including advanced age, previous radiation treatment, previous cancer, and rare genetic conditions [11]. e rarity of sarcomas makes them difficult to study. Single institution reports of incidence, survival, and prevalence are unlikely to be reliable, and pooled data from population-based cancer registries are required to obtain representative data. e aim of the present study is to provide updated estimates of the incidence of malignant soft tissue and bone sarcomas in Italy overall, by area (north, central, and south) and by cancer registry, using data from 15 population-based cancer registries affiliated to the Association of Italian Cancer Registries (AIRTUM) [12]. e presence of metastases at diagnosis and morphology in relation to site-which appears to not have been previously investigated in sarcomas diagnosed in Italy-were also assessed.

Materials and Methods
Italian cancer registries were asked to provide data on malignant soft tissue and bone sarcoma cases incidence between January 1, 2009, and December 31, 2012, that had at least five years of follow-up. Fifteen population-based cancer registries, all affiliated to AIRTUM, sent in data.
For each case, ICD-O-3 morphology, site (ICD-O-3 topography) [13], age at diagnosis, date of diagnosis, and vital status (with date and cause of death) were also ascertained. Cases were staged at diagnosis according to the 6th or 7th editions of the TNM [14,15]. Sarcomas with organs as primary site were excluded, as were Kaposi's sarcoma and sarcomas of uncertain behaviour. NOS cases were included as this was a population-based study, and completeness was of interest, not simply morphological precision. e populations covered by each cancer registry, stratified by age and sex at January 1st of each study year, were obtained from the Italian Institute of Statistics (ISTAT) [16]. e cancer registries carried out passive and active follow-up of cases from date of diagnosis to the end of follow-up (December 31, 2017). For this study, north Italy was represented by the cancer registries of Val d'Aosta, Bolzano, Sondrio, Varese, Pavia, Genova, Reggio Emilia, Modena, and Parma. Central Italy was represented by the single registry of Firenze-Prato. South Italy was represented by the registries of Brindisi, Ragusa, Siracusa, Catania-Messina-Enna, and Nuoro. e results are presented overall, by area and by cancer registry. For metastases and morphology in relation to site, data are also presented by age class.

Statistical Methods.
e analyses were descriptive and exploratory in nature. e chi-squared test was used to assess the significance of differences in distribution frequencies between sarcoma categories, using RStudio, version 3.2.5 [17]. Incidence rates (per 100,000 per year, unless otherwise stated) were age-standardised using direct methods and the world standard population 2000-2025 (ASRs) [18]. We also

Results
A total of 1,112 sarcoma cases were diagnosed in 2009-2012 in the areas covered by the 15 registries. ere were 859 soft tissue sarcomas and 253 bone sarcomas; 55% of total cases were in males and 45% in females. Mean age at diagnosis was 57 years; 44.5% were of age 65 or more (

Site and Morphology.
As regards site, soft tissue sarcomas arose mainly in the lower extremities (35.6%) and trunk (33.6%), while 35.6% of bone sarcomas arose in the lower limb long bones, 33.2% arose in the upper limb long bones, and 21.3% in the trunk (Table 1).
Among the 253 bone sarcomas diagnosed, 37.6% were chondrosarcoma-a high proportion related to the elderly population, 29.2% were osteosarcoma, 13% were Ewing's sarcoma, and 12.7% were other bone sarcoma morphologies. e most frequent soft tissue sarcoma morphologies at each site and by geographic area are shown in Table 4. In the north, leiomyosarcoma was the most frequent morphology at head and neck (31%) and abdominal (26%) sites; liposarcoma was the most frequent morphology in the lower limb (32%) and trunk (24%); and fibrosarcoma was the most common in the upper extremity (29%). In central Italy, sarcoma NOS was the most common (33%) at abdominal sites; leiomyosarcoma was the most common at the upper extremity (71%), head and neck (17%), and trunk (27%); and fibrosarcoma (29%) was the most common at the lower extremity and trunk (27%), while myxosarcoma (17%) was also common in the head and neck. In the south, leiomyosarcoma (30%) was the most common in the abdomen; liposarcoma (37%) was the most common in the lower limbs; fibrosarcoma was the most common in the upper limbs (25%) and trunk (22%); and hemangiosarcoma and fibrosarcoma were the most common in the head and neck (20% each). e most frequent bone sarcoma morphologies at each site, by geographic area, are shown in Table 5. In the north, most bone sarcomas were chondrosarcoma (57% upper limb, 47% trunk, and 28% pelvic cases). e second most frequent bone sarcoma in the north was osteosarcoma (42% of lower extremity and 38% of head and neck cases). In central Italy, the commonest morphology was chordoma which constituted 50% of pelvic cases, whereas chondrosarcoma and Ewing's sarcoma accounted for 33% and 17%, respectively, of pelvic cases, followed by chondrosarcoma which constituted 89% of lower limb and 100% of upper limb cases (data not shown). Osteosarcoma constituted 60% of head and neck cases; 33% each of trunk cases were Ewing's sarcoma and chondrosarcoma. In the south, the commonest morphology was chondrosarcoma which formed 29% of pelvic cases, 90% of upper limb cases, 37% of facial bone cases, and 45% trunk bone cases, while 64% of lower limb cases were osteosarcoma.

Metastases.
A total of 338/1,112 (30%) cases had information onstage at diagnosis, 136/338 (40%) of which had metastatic disease at diagnosis: 109/136 (80%) from soft tissue sarcoma and 27/136 (20%) from bone sarcoma. e commonest metastatic sites for soft tissue sarcoma were lung (38%) and bone (21%), with liver (15%) and pleura (6%) also fairly frequent; 4% had metastases to the central nervous system (Figure 1). e commonest metastatic sites for bone sarcoma were lung (58%), bone (18%), and pleura (6%) (Figure 2). Figure 3 shows the most important metastasis sites by age for soft tissue sarcomas. e proportion of soft tissue sarcoma cases with lung metastases increased from 23.1% in the 0-19-year age class to 42.6% in the 50-69-year age class and declined to 34.4% in those of ≥70 years. e proportions with bone metastases reduced up to the 50-69-age class to then increase in the oldest group. Metastases to liver increased from 7.7% in those of 0-19 years, peaked at 20.4% in the 50-59-year age class, and declined in the oldest group. Figure 4 shows the most important metastasis sites by age for bone sarcomas: in the 0-19-year age class metastases involved lungs in 81.8% of cases and bone in 9.1%. Presentation with lung metastases declined in intermediate ages but increased again (to 57.1%) in patients ≥70 years. Pleural metastases became more frequent with increasing age to Sarcoma 3 peak at 25% in the 50-69-year age class and declined in patients ≥70 years.

Discussion
e 15 AIRTUM cancer registries that contributed data covered an annual mean of 8,183,639 persons over the four-year study period, constituting 13.8% of the mean annual Italian population of 59,237,407 million during that period.
e data used in this study were checked for quality and completeness, by each cancer registry and by AIRTUM, using automated data checking procedures developed by the International Agency for Research on Cancer [21] and International Association of Cancer Registries, and described in Cancer Incidence in Five Continents Volume XI [22]. e     It is noteworthy that ASRs varied between cancer registries (Table 2). For soft tissue sarcomas, ASR was highest in Pavia (north Italy) followed by Varese (north), Firenze-Prato (central), and Ragusa (south), with Parma (north) having the lowest ASR. For bone sarcomas, ASR was also highest in Pavia, followed by Firenze-Prato then Brindisi (south), with no cases recorded in Modena (north). ASR variation between cancer registries is unlikely to be due to variation in completeness, since data quality controls were stringent [5].
ey are more likely to be due to variation in sarcoma risk factors, particularly, for example, variation in proportion of persons given chemotherapy or radiotherapy for another cancer [25]. e area covered by the Pavia registry is characterized by high mesothelioma incidence [26]. Mesothelioma is known to be caused mainly by exposure to asbestos, leading us to speculate that the high ASRs for sarcomas in Pavia could be related to high levels of workplace or environmental carcinogens [27].      (predominantly long bone involvement), 21.3% in trunk, and 9.9% at other sites. is site distribution is consistent with the morphology distribution (chondrosarcoma 37.6%, osteosarcoma 29.2%, and Ewing's sarcoma 13%) in our ageing population. To our knowledge, data on the subsite distribution of bone sarcomas have not been published previously, so we consider published subsite distributions according to morphology. As regards chondrosarcomas, these arise principally in long bones but can also develop in axial bones (pelvis, ribs, and scapula). us, in a population-based series from the Netherlands [30], 50.8% of chondrosarcoma arose in long bones and 49.2% in axial bones. In a SEER study, 44.5% arose in extremities, 31.1% in axial bones, 14.9% other bone site, and 9.6% in soft tissues [31].

Site and
As regards osteosarcomas, according to Savage and Mirabello [32], around 80% of them occur in lower long bones, whereas in a SEER series, 20.8% occurred on an extremity, 39.3% in axial bones (pelvis, spine, and ribs), and 19.3% at another location (mandible, skull, etc.) [33]. It has been reported that axial bone involvement becomes more frequent with advancing age [34]. A Japanese populationbased study reported that 48.9% of osteosarcomas arose on a lower extremity, 36.9% on the trunk, and 10.1% on an upper extremity [35].
As regards Ewing's sarcoma, a SEER study [9] reported the most common primary sites as extremity bones (50%) followed by pelvis, ribs, and vertebra, although any bone can be affected. Bernstein et al. [36] reported that 26% of Ewing's sarcomas arose in pelvis, 20% in femur, 16% in chest wall, 10% in tibia, 6% in humerus, and 2% in skull. A populationbased study from the Netherlands [30] reported that 56.9% of Ewing's sarcomas occurred in axial bones and 43.1% in long bones.

Soft Tissue Morphology.
We found that 29.4% of soft tissue sarcomas were NOS, followed by liposarcoma (22.2%), fibrosarcoma (20.4%), leiomyosarcoma (16.1%), rhabdomyosarcoma (4.4%), and hemangiosarcoma (4.2%). e study of Stiller at al. [4] on EUROCARE 5 cases diagnosed in 2000-2007 from 29 European countries reported that the proportion of NOS sarcomas (likely mainly soft tissue sarcomas) varied from 37% in Croatia to 4% in Slovenia, with an overall figure of 13%. Wibmer et al. [37] analysed soft tissue sarcomas diagnosed in 1984-2004 in the Austrian National Cancer Registry, finding that 36% were NOS. Interestingly, in a RARECARE study [5], 18% were sarcoma NOS, and 79% of these soft tissue sarcomas were confirmed as NOS after revision, indicating that even experienced pathologists diagnose sarcoma NOS in a substantial proportion of cases.
We found that morphology distribution by site varied with geographic area in that 20% of hemangiosarcomas occurred in the head and neck in south Italy (Table 4), compared to 6% (data not shown) in the north, suggesting geographic variation in the distribution of soft tissue sarcoma risk factors.

Bone Morphology.
e most frequent bone sarcoma in our series was chondrosarcoma (37.6%), which is generally reported as most common bone sarcoma in adults: it is typically low grade and often originated from preexisting benign bone lesions [31]. e 2016 UK guidelines for bone sarcoma management [38] noted chondrosarcoma as the most common bone sarcoma in the ageing UK population. Our population is also ageing (44.5% ≥ 65 years), and our high proportion of chondrosarcomas is consistent with UK findings. e second most frequent bone sarcoma was osteosarcoma (29.2%). is mainly affects people in the first three decades of life [32]. However, a recent study noted a second incidence peak in persons over 50 years [34], and this is consistent with the high relative proportion of osteosarcomas in our ageing Italian population.
Ewing's sarcoma, the third most common (13.0%) bone sarcoma in our population, is an aggressive yet highly treatable tumour that mainly affects adolescents and young adults [39]. Genetically Ewing's sarcoma is characterized by balanced chromosomal translocations in which a member of the FET gene family fuses with an ETS transcription factor [39]. As regards geographic variations of site and morphology of bone sarcomas (Table 5), we found that in central Italy proportionately more chordomas arose in the pelvic (sacral) region. A large (400 cases) SEER study of chordomas reported that 32% of chordomas were cranial, 32.8% spinal, and 29.2% sacral [40], suggesting that subsite occurrence within the axial skeleton is in fact more uniform than indicated by our limited data.
In another SEER study [31], 44.5% of chondrosarcomas arose in upper and lower extremities-similar to the proportion we found in north Italy, but lower than in south Italy. However, comparison is complicated by the fact that 14.9% of SEER cases had location not specified, while in our study all cases had specified locations.

Metastasis.
Only 30% of our cases had stage information at diagnosis, 40% of which were metastatic. In a populationbased study [8] of malignant adult soft tissue sarcomas diagnosed in Germany in 2003-2012, only 18.2% had stage information at diagnosis, 36% of which had metastases at diagnosis. Several US SEER-based studies have been published. In one [41] on cases of primary soft-tissue sarcoma occurring on an upper extremity and diagnosed in 1973-1998, only 10.7% had no stage information, and 4.7% of the whole population were metastatic at diagnosis. In two more recent SEER studies [42,43] on pediatric soft-tissue sarcomas, all cases had stage information (local, regional, or distant), 15% of which were metastatic at diagnosis. Clearly, SEER data on pediatric cases are more complete in terms of stage information than European data. According to the follow-up study of Patel et al. [44], disease dissemination in sarcomas occurs mainly by hematogenous as opposed to lymphatic spread (<5%); metastases primarily occur in lung, but also bone, liver, regional lymph nodes, peritoneum, distant subcutaneous tissue, and brain, although the preferred metastatic site varies with subtype [45]. We found that 8 Sarcoma the main metastatic sites for both soft tissue and bone sarcomas were lung, bone, and pleura; liver was also a frequent metastatic site for soft tissue sarcoma (Figures 3  and 4). Nevertheless, liver was the most common site in adults in the 50-69-year age class, and peritoneum was the most common site in the 20-49-year age class. Our findings are broadly similar to those of Patel et al. [44] and also to those of Shweikeh et al. [46] and Marchiori et al. [47].

Strengths and Limitations
e main strength of our study is that we used data from AIRTUM chancer registries characterized by high quality (95.7% of cases microscopically verified) and high completeness. However, the data were not perfect. e large proportions of NOS reflect the difficulties of histopathologic, and increasingly, molecular diagnosis and suggest that sarcoma diagnoses should be routinely reviewed by a pathologist specialised in sarcomas. However, this is not often possible. Another problem is that sarcoma patients are normally treated at specialist centres available outside the cancer registry area and data from centres outside registry areas are not routinely available to cancer registries. AIR-TUM is working to ensure that cases treated outside registry areas are not lost to cancer registries.
A second study strength is that we excluded visceral and Kaposi's sarcomas, thereby avoiding the misclassification that is common in sarcomas arising in organs. We emphasize that our cases were population-based, thereby avoiding selection bias, and that we analysed a relatively large number of cases so that our findings are likely to be reliable. Nevertheless, there were low numbers of cases in some soft tissue and bone sarcoma categories, and in these cases, findings should be interpreted cautiously. As a limitation, we note that stage at diagnosis and details of treatment were largely lacking: such information if present would likely have provided additional insights into these rare cancers. In addition, only a few Italian cancer registries use software to automatically generate case data from the data sources (now mostly electronic databases). Automatically generated cases are subjected to checking procedures, and flagged cases undergo manual (operator) checking. Most cancer registries, however, still generate cases manually from clinical records and various electronic databases.

Conclusions
High sarcoma incidence rates in the cancer registry areas of Pavia and Firenze-Prato suggest that these may be high-risk hot-spots deserving further investigation. Marked variations of morphology distribution with site between Italian areas suggest a more general geographic variation in risk factors that may also repay further investigation, particularly since sarcoma etiology is poorly understood. We are planning an analysis of survival for bone and soft tissue sarcomas archived in the cancer registries that participated in the present study. For this, we will collect information (treatment comorbidities) from clinical records (not available in registry records) to enable us to better interpret survival variations.
Data Availability e data on sarcoma cases used in this study were provided by cancer registries affiliated to AIRTUM and cannot be made freely available.

Conflicts of Interest
e authors declare that there are no conflicts of interest.