Angiosarcomas are aggressive cancers that are difficult to treat and have high mortality. As they are ultrarare tumors, with only approximately 400 new cases per year in the USA, therapy options for patients with metastatic disease are limited.1 2 Angiosarcomas are often responsive to chemotherapy, with response rates to taxanes ranging from 18% to 89% in several studies.3–6 However, these responses are not durable; the median progression-free survival (PFS) ranges from 4 to 9.5 months.3–6 Five-year survival for all patients with angiosarcoma, including those who present with localized disease, is only 30%–40%.7

A subset of angiosarcomas are characterized by high tumor mutation burden (TMB), suggesting that they may respond to immune checkpoint inhibitors.8–11 Initial immune characterization of soft tissue sarcomas included three angiosarcomas, and showed that all three had infiltration of PD-L1-expressing lymphocytes and macrophages.12 Larger series have reported varying levels of PD-L1 expression, with Tumor Proportion Scores (TPS) ranging from 14% to 66%.13 14 A series of seven angiosarcomas showed one with high PD-L1 expression.13 Two larger series have more recently been published, one reporting that 6 of 24 angiosarcoma samples (66%) of different origins, including bone, skin, breast, soft tissue, and visceral primary tumors, had at least some membrane expression of PD-L1 without a clear correlation with site of origin.15 Although angiosarcomas are characterized by dysregulated angiogenesis, there is no correlation between expression of PD-L1 or tumor infiltrating lymphocytes and vascular endothelial growth factor related gene expression.14

Published case reports and small series have demonstrated encouraging initial clinical responses to immune checkpoint inhibition (ICI) in patients with angiosarcoma. For instance, one of three patients with angiosarcomas enrolled on a phase II study of checkpoint inhibitors in sarcoma had an objective response.16 Two separate cases are also reported of patients with durable complete responses (CR) in angiosarcomas of the scalp17 and nose.18 A larger retrospective series of seven patients suggested a response rate of over 50%.19 These anecdotal cases suggest that checkpoint inhibition is active in a subset of angiosarcomas, and prompted the addition of an angiosarcoma cohort to DART (S1609) (Dual anti-CTLA-4 and anti-PD-1 Blockade in Rare Tumors), a prospective phase II study conducted through the National Cancer Institute (NCI)-supported SWOG Cancer Research Network’s Early Therapeutics and Rare Cancers Committee.

Angiosarcoma has a poor prognosis and improved treatment strategies are urgently needed. Molecular characterization of angiosarcoma has yielded insight into the pathologic drivers of these rare endothelial tumors and the identification of clear subsets of disease. Secondary angiosarcomas were differentiated molecularly from de novo tumors by the presence of MYC amplifications23; consistent with the literature, our patient with angiosarcoma post radiation to the breast showed a MYC amplification as well as, interestingly, an NTRK1 fusion (table 2). Primary breast angiosarcomas are more likely to harbor PIK3CA mutations (as seen in our patient (table 2)).8 A subset of angiosarcomas of the scalp and face harbor high TMB and a pattern of DNA damage consistent with ultraviolet (UV) light exposure, suggesting that this subset might be uniquely susceptible to ICI.8 11 Indeed, one of the responding patients in our study with angiosarcoma of the scalp had a high TMB (table 2).

Larger studies of immune checkpoint inhibitors in sarcoma have included angiosarcoma patients but in small numbers,16 24 making assessment of efficacy in this specific histology impossible from prior studies. Most angiosarcoma responders to immunotherapy in reports where site of primary tumor is available have cutaneous disease of the face or scalp, consistent with the higher TMB seen in this group. At least one response in a radiation-associated angiosarcoma of the breast has also been reported.19 There is limited other published data to suggest whether angiosarcomas of other sites might respond to ICI. We, therefore, included all angiosarcomas in this study. To our knowledge, this represents the first prospective trial of immunotherapy in angiosarcoma.

In our cohort, three of the five patients with primary cutaneous disease of the face or scalp had objective responses for an ORR of 60% (figure 1). Surgery and radiotherapy are the current standard of care for localized disease in this specific subset, which necessarily carries high morbidity due to its location. In scalp angiosarcomas, local failure rates in spite of aggressive local treatment are high and long-term survival is poor.25 Future studies to incorporate neoadjuvant immunotherapy in this unique population are warranted, as are combination studies to incorporate immunotherapy with frontline chemotherapy. Another response was seen in a patient with a radiation-associated angiosarcoma of the breast, and reduction in tumor size was also seen in a primary tumor of the liver, suggesting that angiosarcoma phenotypes other than that of cutaneous scalp or face tumors also may respond to ICI. Two patients attained stable disease ongoing at 6+ and 13+ months; one had primary angiosarcoma of the spleen and the other had a primary angiosarcoma of the deep soft tissue of the lower extremity. Angiosarcomas that do not have a pattern of DNA mutations consistent with UV light exposure26 may have elements of viral DNA,27 which may associate with response to ICI in other tumor types.28–30

To better understand the molecular basis for response, correlative samples were collected on study and will be analyzed as per the prespecified plan at Cancer Immune Monitoring and Analysis Centers sites. Eight patients had NGS performed for clinical purposes prior to enrollment on the master DART protocol, and results were available as part of the medical record. Consistent with previously published data from other angiosarcoma cohorts, TMB in angiosarcoma was variable. One responder had high TMB (table 2). A second responding patient had strong PDL1 expression on 30% of tumor31 and TMB assessed as 8 mut/mb. One patient with a primary visceral angiosarcoma had reduction in tumor size and others had prolonged stable disease, suggesting that these patients may also benefit from ipilimumab and nivolumab. Larger studies allowing for adequate power to stratify subtypes of angiosarcoma are needed to better quantify the potential benefit. This effort will likely require multi-institutional collaboration to enroll sufficient numbers of patients.

Toxicity on this study was comparable to the toxicity seen in other trials with the ipilimumab and nivolumab combination in sarcoma.16 The potential clinical benefit of ipilimumab over nivolumab monotherapy in this setting remains uncertain and future trials may better assess the clinical activity versus toxicity profile of nivolumab monotherapy compared with nivolumab and ipilimumab combinations. Similarly, escalating cytotoxic T-lymphocyte associated protein 4 (CTLA-4) inhibition with more frequent dosing or by increasing to a dose of 3 mg/kg as is done in other cancer types, usually with a limited number of ipilimumab doses, may increase efficacy endpoints with the possibility of greater toxicity.32 33 A direct comparison would be needed to better quantify alternative dosing schema.

Strengths of this study include the inclusion of patients at both academic and community centers and support from the NCI and SWOG. Since initially conceived, the DART study served an unmet need in that it made an immunotherapy trial available to multiple cohorts of patients with rare tumors, and demonstrated that it was feasible to rapidly accrue even ultrarare tumors.20 By adding an angiosarcoma cohort to the DART study, we were able to rapidly accrue this trial for a rare cancer. Weaknesses of the study include a relatively small sample size, precluding our ability to make conclusive comparisons between primary sites or molecular subgroups. Central pathology and radiology review were not mandated; therefore, we relied on local site assessments including from sites that may not have high sarcoma volumes.

Ipilimumab and nivolumab showed activity in angiosarcoma with responses seen in cutaneous angiosarcomas. Correlative studies to better understand the molecular characteristics of these patients utilizing the same centralized platforms are underway. Further study of ipilimumab and nivolumab in angiosarcoma is warranted.