Incidence, demographics, and survival of malignant hemangioendothelioma in the United States

Abstract Background Malignant hemangioendothelioma is an endothelial cancer with heterogeneous clinical behavior that can range from indolent to aggressive, of which the majority are epithelioid (EHE). Its incidence and demographics have not been previously well defined in a large cohort. Methods This retrospective analysis used the US Cancer Statistics National Program of Cancer Registries – Surveillance Epidemiology End Results (SEER) combined database to identify patients in the US newly diagnosed with hemangioendothelioma between the years of 2001 and 2017 (n = 1986). Survival analyses were performed on a subset of patients within the SEER‐18 database with survival information available (n = 417). Outcomes included incidence, demographics of patients newly diagnosed with hemangioendothelioma, extent of disease at presentation, and overall survival. Results The incidence of hemangioendothelioma in the US is 0.4 cases per million person‐years. Although cases rose to 122 newly diagnosed in the year 2017 (90 EHE, 32 other hemangioendothelioma), incidence rates were stable. Skin and connective tissues were the most common presenting sites (33.4%), followed by liver (24.5%), lung (17.6%), and bone (12.5%). Median age at diagnosis was 55 years; 27.2% of patients were pediatric, adolescent, or young adult (<40 years). At presentation, 36.4% of patients had localized disease; 21.6% presented with regional and 41.7% with distant metastases. Observed survival at 3 years was 79.7%, 70.7%, and 46.0% for patients presenting with local, regional, and distant disease and most deaths occurred within the first 2 years. Conclusions Malignant hemangioendothelioma is ultra‐rare but meaningfully impacts affected patients. These data may provide benchmarks for comparison of new approaches to hemangioendothelioma therapy and highlight poor survival outcomes.


| INTRODUCTION
Malignant hemangioendotheliomas are ultra-rare vascular sarcomas with variable clinical behavior that can be either indolent or aggressive; the majority of malignant hemangioendotheliomas are epithelioid (EHE). The overall incidence has previously been reported as 0.38 per 1,000,000 person-years in a large European database, 1 but has been reported to be only 0.23 per 1,000,000 personyears in the United States (US). 2 Most (83%) are molecularly characterized by the presence of a WWTR1-CAMTA1 gene fusion and have a 5-year overall survival (OS) of 59% in case series. [3][4][5] A smaller subset has a YAP1-TFE3 gene fusion, associated with better 5-year survival of 86%. [5][6][7] Due to the extreme rarity of EHE, the patterns of initial presentation have not been well characterized in a large cohort. A consensus statement of EHE experts emphasized the need for data to guide surveillance strategies and to establish clinical benchmarks for future clinical trials. 8 Herein, we report incidence, clinical presentation, and survival outcomes for US patients with malignant hemangioendothelioma.

| Population and data
IRB approval was obtained (Providence 202000564) and this research was conducted in accordance with Helsinki principles. De-identified aggregate national registry data was extracted from public use databases of the US Cancer Statistics(USCS) National Program of Cancer Registries -Surveillance Epidemiology End Results (SEER) submission for the years 2001-2017. 9 Data met high-quality data criteria. 10

| Statistical analyses
p-value for significance was set a priori at 0.05. For incidence rate calculations, data were age-adjusted to the 2000 US standard population (19 age groups; census P25-1130). Confidence interval (CI) estimation was performed using the methods of Tiwari et al. 12 Trends were calculated using 2-year averages with annual percentage change calculated by the weighted least squares method. Survival analyses were performed with actuarial methods at 1,2,3,4, and 5 years post-diagnosis with 95% CIs. Relative survival was calculated with the Ederer II method 13 and expected survival tables included age, sex, socioeconomic status, geography, and race.

| Incidence of malignant hemangioendothelioma
Between 2001 and 2017, a total of 1986 cases of hemangioendothelioma were reported in the US (1542 EHE; 444 hemangioendothelioma). New cases increased from 100 in the year 2001 to 122 in 2017 ( Figure 1). The adjusted incidence rate was ~0.4 cases per million personyears across the study period (range 0.31-0.44 cases per million person-years) and were statistically stable (annual percentage change −0.04%; 95% CI −1.06%-0.99%; Figure 1).

J E L C L A S S I F I C A T I O N
medical oncology, meta-analysis, SEER, chemotherapy

| Demographics
Among persons newly diagnosed with malignant hemangioendothelioma, 51.3% were female (48.7% male) while 83.3% were white, 10.5% Black, 3.9% Asian, and the remaining persons were unknown/unspecified, or in a   racial group containing fewer than 16 cases ( Figure 2B). 11.0% of patients were reported as Hispanic. Median age at diagnosis was 55 years ( Figure 2C) and 27.2% of patients were pediatric, adolescent, or young adult (under 40 years). Figure S1 includes data for EHE alone.

| Extent of disease at presentation
Local-regional-distant (LRD) staging information was available for 1618 of the 1986 patients (81.5%). 41.7% presented with distant metastatic disease while 21.6% of patients presented with regional disease, including 269 patients by direct extension alone, 42 patients by regional lymph nodes alone, and 39 patients regional by both modes. Approximately one-third of patients (36.7%) presented with localized disease ( Figure 2D).

| DISCUSSION
Hemangioendothelioma is an ultra-rare sarcoma with a paucity of data to guide optimal treatment and counseling of patients. To our knowledge this is the largest cohort of malignant hemangioendothelioma/EHE patients reported to date. The US incidence rate in this series (120 new cases diagnosed in the US per year) is similar to that recently reported in France and approximately double the previously reported US incidence. 1,2 Prior differences likely reflected ascertainment bias, with earlier US data underrepresenting the true incidence of EHE. We also found equal numbers of cases in men and women, whereas prior reports have suggested a female preponderance. 15 Although EHE is often considered an indolent cancer, we observed 3-year OS of only 63.5% which was far worse than expected survival based on actuarial data matched on sociodemographic and socioeconomic factors. This, combined with the fact that most deaths occurred within 5 years from diagnosis and most of those within the first 2 years, suggests that EHE is the major factor driving mortality and further supports the use of 1-or 3-year OS with comparison to historical data as a feasible and clinically appropriate clinical trial endpoint. As expected, outcomes in patients with advanced disease at presentation were especially poor. This is consistent with prior reports identifying pulmonary infiltrates, pleural effusion, and ascites as negative prognostic signs in EHE. 15 Indeed, in a recent prospective clinical trial for progressing EHE less than half of patients were still on study treatment at 6 months and 17% still on study treatment after 1 year. 16 However, even patients presenting with localized disease had mortality of one in five, highlighting the ongoing need for more research and better therapies.
A limitation of our data set is that considering the rare nature of this tumor, misclassification bias is possible and genomic or biomarker data to confirm the diagnosis were not available in this national populationbased registry. Other types of hemangioendothelioma such as those with other gene fusions are likely also included in the "hemangioendothelioma" group. A second limitation is that the USCS registry only includes malignant tumors. Several types of hemangioendothelioma, including those with aggressive potential such as kaposiform hemangioendothelioma, have diagnosis codes which are categorized as benign in ICD-O-3 categorization of malignancies. These types (Kaposiform hemangioendothelioma, papillary intralymphatic angioendothelioma, retiform hemangioendothelioma, pseudomyogenic hemangioendothelioma) were thus not recorded in USCS or possible to include in our analyses. Although there have been improvements in survival of patients with Kaposiform hemangioendothelioma in recent years due to sirolimus based regimens, we do not expect that would impact our findings. 17 However, given the high predominance of WWTR1-CAMTA1 gene fusions in patients with hemangioendothelioma and the consistency in demographics and survival outcomes that we demonstrated between the separate EHE and malignant hemangioendothelioma groups, we believe that this data still reflects a useful comparator.
As noted in a recent EHE consensus paper, 8 better survival estimates are needed to serve as a benchmark for future interventional trials. Knowing a benchmark is especially critical for this ultra-rare sarcoma where randomized studies are not feasible even with multicenter collaboration. These data effectively represent the entire US population, capturing 99% of hemangioendothelioma diagnoses in the study period and can serve as a benchmark for future single-arm interventional studies.

FUNDING INFORMATION
Work supported in part through the Paul G. Allen Research Center at Swedish Cancer Institute (K.G.P.) and by support to the Sarcoma Oncology Program of F I G U R E 3 Survival after hemangioendothelioma diagnosis. Survival data were available for 417 patients with malignant hemangioendothelioma who were included in the SEER-18 databases subset of the US Cancer Statistics. (A) Observed overall survival (OS) (red circles) as compared to expected survival based on demographic data (dotted line). Calculated relative survival (blue squares) and causespecific survival (black triangles) are also shown. (B) OS by diagnosis code (hemangioendothelioma or epithelioid hemangioendothelioma). (C) Survival by decade of diagnosis. Data are presented +/− 95% confidence interval. (D) OS by extent of disease at diagnosis. Survival if distant disease at diagnosis was significantly worsened from localized or regional disease (p < 0.05). (E) OS by reported primary site. Differences between lung/mediastinum and skin/soft tissue were significant (p < 0.05). the University of Washington from Curt and Elizabeth