Sarcoma of the breast: breast cancer history as etiologic and prognostic factor—A population-based case–control study

Sarcomas are rare neoplasms arising from mesenchymal cells. There are various types of sarcomas, classified by morphology and cell-type of origin, and the classification systems are still evolving [1‐3]. The tumors can originate from any part of the body. In general, soft tissue sarcoma is reported to represent about 1% of all malignant tumors in adults [4]. Despite a large number of epidemiological studies of the different types of sarcomas, the reported incidence varies greatly [5‐13]. This may in part be explained by underreporting, but also by problems identifying cases from cancer registries. There is no uniform coding system for sarcomas, which makes identification of cases difficult. Most registers are based on ICD coding, which is topographic rather than morphological. The true incidence for sarcoma thus remains unclear. This also makes prognostic studies more difficult to perform and interpret. The present study is an attempt to address these issues by exploring the nationwide Swedish registers.

Breast sarcomas, i.e., breast tumors derived from mesenchymal cells, account for about 1% of all breast malignancies and consist of several different histopathological subtypes. The current understanding of the etiology and prognosis of these tumors is still limited. The most common histopathological subtypes are angiosarcoma and malignant phyllodes tumors [14, 15]. Phyllodes tumors are composed of both epithelial cells and connective tissue stroma and range from benign (grade 1) to overtly malignant (grade 3). The malignant phyllodes tumors have a higher proportion of mesenchymal proliferation and have a genetic profile similar to other breast sarcoma, but not to angiosarcoma [16]. Angiosarcoma is a subtype of sarcoma originating from blood- or lymphatic vessels. Previous irradiation is an established risk factor for its development [17]. Causal links to individual susceptibility dependent on gene-environment interactions have been suggested [18, 19]. Chronic lymphoedema after a mastectomy with lymph node dissection together with irradiation is another reported risk factor for lymphangiosarcoma (Stewart Treves syndrome) [20]. The aims of this national survey were to analyze the incidence of breast sarcoma during a 20-year period, to evaluate risk in relation to previous malignancies and to assess prognosis in the same individuals.

The study was performed as a matched case–control study. A database with gathered information from multiple data sources was used to identify cases and controls. The database includes information from the Swedish Cancer Register, the Swedish Cause-of-Death register and the Swedish Population register. The Swedish Cancer Register is mandatory for all malignant diagnoses and contains data from 1958 [21]. It includes data on social security number, sex, place of residency at diagnosis, reporting hospital and clinic, date of diagnosis, clinical and morphological diagnosis, and extent of tumor at diagnosis. The completeness of the Swedish Cancer Register has been estimated to be > 96% [22]. The Swedish Cause-of-Death Register contains data from 1961 and includes data on time and place of death and the cause coded according to ICD classification. The Swedish Population Register contains data on all Swedish citizens and migration in and out of the country. The patient database also contains information from national censuses including socio-economic data such as educational level. All these registers have the social security number as a common parameter and can therefore easily be cross referenced.

A combination of site-coding according to ICD 7 (170, corresponding to ICD 10 C50.X) and morphological coding was used to find cases of mesenchymal tumors/sarcomas of the breast during the period 1993 to 2013. A total of 344 cases were identified (Table 1). Each case was matched according to sex and age to up to ten controls from the general population. Angiosarcoma and lymphangiosarcoma were grouped as angiosarcoma. Phyllodes tumors were divided in three sub-classes according to histopathology (benign, borderline or malignant). Due to relatively small numbers the benign phyllodes tumors, that are not mandatory to register in the cancer registry, were omitted from further analyses. The most uncommon sarcomas were grouped as “other sarcoma.” A proportion of these were not specified with histopathological coding according to SNOMED and were classified as sarcoma not otherwise specified (NOS). For patients treated in the Stockholm county, information of previous breast cancer treatment was collected from the patients’ medical records.

During the period 1993–2013, 344 patients were registered with mesenchymal tumors in the breast (Table 1). Patients with phyllodes tumors were younger (median age 56.3 (range 17–89) than patients with angiosarcoma (median age 70.5 (range 26–91), P < 0.001). The study period was divided into four groups encompassing 5–6 calendar years each (Table 2). The incidence of angiosarcoma increased from 0.09 during the first period 1993–1998 to 0.42 in 2009–2013 (trend of increase in the incidence: 1.10, 95% CI 1.05–1.16; P < 0.001). The overall incidence of breast sarcoma increased from a mean incidence of 1.52 cases per million per year during the first half of the study period to 2.04 during the second period. (Fig. 1).

The main findings of this study were the strong association between the risk of angiosarcoma and a history of ipsilateral breast carcinoma, not observed for the other histopathological subtypes, and the prognostic similarities between malignant phyllodes tumors and breast sarcoma. Angiosarcoma, often seemingly triggered by radiotherapy following breast cancer surgery, displayed the most severe clinical course, with significantly shorter survival.

Radiotherapy has been utilized in cancer treatment since its discovery in the nineteenth century. In the 1970s, breast conservative surgery in combination with postoperative radiotherapy was introduced. During the following decades, utilization of this treatment method increased and it is now dominant [21, 23].

Breast sarcoma contributes to a very small percentage of all breast malignancies. Angiosarcoma, the most common sarcoma subtype, has been reported to account for only 0.1–0.2% of all malignant breast neoplasms [24]. The reported incidences of angiosarcoma following breast cancer treatment varies between 0.01 and 1% [25, 26]. When comparing incidence figures between studies, the population at risk is obviously a key factor. In this study, the entire population of Sweden has been covered. In contrast, although male breast cancer might be underreported [22], these tumors almost exclusively affect women. If the population at risk would be considered to be only the female population, this would double the incidence.

We observed an increasing incidence of angiosarcoma over time and we suspect that this, at least in part, may be explained by radiation as a part of breast cancer treatment. There was also a slight increase in incidence of borderline phyllodes that may be explained by a shift in diagnosis, since the reported cases of benign phyllodes tumors decreased during the same period. When the reported benign phyllodes tumors were included in analysis we observed a stable overall incidence during the study period (data not shown).

The risk of breast sarcoma was strongly associated with a history of breast cancer, displaying a peak incidence 5–10 years after the diagnosis of breast cancer. The fact that a majority of angiosarcomas (73%) occurred on the same side as the previous breast cancer is consistent with several reports of sarcomas developing in previously irradiated fields [18, 27‐30].

Individual data on previous radiotherapy was available for a subset of the patients, all of whom had received radiation in doses of 42–50 Gy. We assume that a majority of the angiosarcoma patients with previous breast carcinoma may have received radiotherapy and that irradiation was a contributing factor for subsequent angiosarcoma development. It is also possible that the risk association between other sarcomas 5–10 years after the diagnosis of breast cancer could be explained by inclusion of some misclassified angiosarcomas in this group. Radiation-induced breast sarcoma was recently reported to be a clinicopathologically distinct type of radiation-induced malignancy with significantly shorter latency from time of initial radiation, different recurrence patterns and potentially better outcomes after radical surgical treatment compared to other radiation-induced sarcomas [31].

Radiation-induced breast angiosarcoma has also been associated with a worse prognosis than primary angiosarcoma [32‐34]. Based on a recent survey from Italy, including 112 patients treated at 24 centers, only 50.5% of the patients were still alive after 5 years follow-up, and the disease-free survival was only 35 months [32]. In our cohort, no such difference was obvious, possibly due to the limited size of the cohort. Previous breast cancer per se did not affect differences in survival between the groups and neither was there any difference in survival observed between the early and the late period of the study.

The prognosis for breast sarcoma patients is poor. We observed increased mortality among patients with angiosarcoma and malignant phyllodes during the first 5 to 7 years after diagnosis compared to controls. On the contrary, borderline phyllodes tumors did not lead to shortened overall survival. Conditional regression analysis showed a weak correlation between age at diagnosis and survival time.

Concern has been raised that the incidence of radiation-associated angiosarcoma will increase in the future due to extended use of radiation therapy in treatment of breast cancer including in situ cancer. Our study is consistent with this assumption. In terms of absolute or relative risk of developing angiosarcoma after radiation to the breast, this study can not provide an exact quantification, since it is not designed to include patients at risk (all patients receiving radiotherapy). Since angiosarcoma is still very rare, we can assume that the absolute risk for an individual receiving radiotherapy is low. However, we observed a fourfold increase in the incidence of angiosarcoma in the later study period from 0.09 to 0.42, which calls for concern about future incidence of this tumor type. These findings are in accordance with recently published data from a Finnish national survey [35].

In order to study rare diagnoses and outcomes, a large study population is mandatory. The strength of this study is the completeness of the Swedish national registers, which cover the entire population and are well validated. The population analyzed is relatively large. However, studies of rare tumors are always a challenge. We have to consider risk of underreporting and the fact that the coding practice of sarcomas is not fully standardized. When investigating rare diseases such as sarcomas, each case that is misdiagnosed or missed in registration will affect the overall incidence.

Previous ipsilateral breast carcinoma was a strong risk factor for development of breast sarcoma. The observed trend of increase in the incidence of angiosarcoma during the study period could hypothetically be related to changed routines regarding the use of irradiation following breast cancer treatment, and further studies on this connection are of great clinical importance.