Comparing central nervous system (CNS) and extra-CNS hemangiopericytomas in the Surveillance, Epidemiology, and End Results program

Analysis of 655 patients and review of current literature




Hemangiopericytomas (HPCs) are rare tumors in the central nervous system (CNS) and in extra-CNS sites. The authors of this report used the Surveillance, Epidemiology, and End Results (SEER) Program to study prognostic factors in patients with HPC.


The SEER database was analyzed for patients who were diagnosed with HPC tumors from 1973 to 2007. Patients were stratified into CNS and extra-CNS groups. Univariate and multivariate analyses were performed for the overall survival (OS) endpoint using major demographic factors (age, race, and sex) and disease factors (tumor site).


In total, 655 patients with HPC were stratified into a CNS group (n = 199) and an extra-CNS group (n = 456). The patients with extra-CNS HPC were statistically older (mean age, 53 years vs 49 years; P = .008) and were more likely to have larger tumors (median greatest dimension, 7.0 cm vs 5.2 cm; P < .001). Patients who had CNS tumors had better OS and cause-specific survival (CSS) compared with patients who had extra-CNS tumors (P < .001 for both). Negative predictors of OS on multivariate analysis included extra-CNS tumor site (hazard ratio [HR], 1.6; P = .005) and older age (ages 40-59 years: HR, 2.08; P = .032; ages 60-79 years: HR, 3.9; P < .001; aged ≥80 years: HR, 7.7; P < .001).


The current analysis demonstrated that patients with extra-CNS HPCs had worse OS and CSS than patients with CNS HPCs. Cancer 2012. © 2012 American Cancer Society.


Hemangiopericytomas (HPCs) are rare tumors that were formally defined by Stout and Murray in 1942.1 This tumor, which is believed to arise from pericytes around capillaries and postcapillary venules, is located in any site that contains capillaries, including both central nervous system (CNS) sites and extra-CNS sites. Past retrospective series have demonstrated that HPCs occur most commonly in individuals ages 20 to 70 years and are observed most often in the musculoskeletal system: namely, the extremities, retroperitoneum, and skin.2-5 Accurate diagnosis of HPCs is difficult, because up to 15% of all soft tissue sarcomas can have HPC-like features. Moreover, the exact pathologic categorization of HPCs has continued to evolve over the past several years, and many pathologists currently define HPCs more generally as solitary fibrous tumors.6, 7 Further studies have indicated that HPCs can be located virtually anywhere in the body, and their morphology can vary greatly in size and shape, including both round cells and spindled cells.2 The variety in morphology and location of HPCs makes them difficult to diagnose with certainty. HPC tumors arising in the CNS are exceptionally rare and represent <1% of all intracranial tumors.4, 8, 9 All of these factors combine to make clear guidelines in the management of HPCs limited.

Because of the rarity of HPCs, information also is scarce regarding their clinical and biologic characteristics. Currently, only retrospective series have been published with cohorts of less than 110 patients.10-17 Historically, the mainstay of HPC management has involved surgical resection.3-5, 8, 18-23 However, the overall clinical course of HPCs has proven difficult to predict, and local or distant recurrences can occur as many as 10 years after surgery. This can be despite an adequate gross total resection.3, 20-22, 24 Specific features that were correlated previously with a poor outcome were a high number of mitotic figures, cellular atypia, larger tumor size, hemorrhage, and necrosis.21, 22

To better characterize the clinical features and treatment outcomes of patients with HPCs at CNS and extra-CNS sites, we performed an analysis of all patients with HPC registered in the Surveillance, Epidemiology, and End Results (SEER) public-access database collected from various geographic areas in the United States from 1973 to 2007.25



The data used in this analysis were acquired from the SEER 17 registries database, which was released in April 2010 and includes tumors that were diagnosed between 1973 and 2007. This database was accessed using SEER*Stat software (version 6.6.2; National Cancer Institute, Bethesda, Md). The inclusion criteria required that all patients had a known age and an International Classification of Diseases for Oncology, third edition (ICD-O-3) histologic diagnostic code of 915 (hemangiopericytoma) with a malignant behavior code. Patients diagnosed on autopsy or death certificate only were excluded from the analysis.


All statistical analyses were performed using the SPSS statistical software package (version 17; SPSS Inc., Chicago, Ill). A tumor site was classified as CNS if the ICD-O-3 site code was brain (C71.x), meninges (C70.x), or spinal cord (C72.0). All other sites were classified as extra-CNS. The SEER staging system was used for staging purposes, because this categorization has standardized definitions and consistent data dating back to 1973. SEER staging defines localized cancer as that limited to the organ in which it began, without evidence of spread. SEER staging defines regional cancer as that which has spread beyond the primary site to nearby lymph nodes or organs and tissues. Distant cancer is defined as disease that has spread from the primary site to distant organs or distant lymph nodes.

Patient parameters were compared statistically between the CNS group and the extra-CNS group to assess for variable heterogeneity. Continuous variables were compared statistically using the Student t test, and categorical variables were compared using the Pearson chi-square test, except when the data set was relatively small, in which case, the Fisher exact test was used. A P value < .05 was considered statistically significant.

The outcomes of the CNS group and extra-CNS group were compared primarily based on the endpoints overall survival (OS) and cause-specific survival (CSS). The latter was obtained using the SEER cause-specific death classification defined by Howlader et al.26 The Kaplan-Meier method was used to generate survival curves. Survival estimates are presented along with standard errors. The CNS and extra-CNS groups were compared in univariate and multivariate analyses using Cox proportional hazards regression. For the purposes of the multivariate analysis, the surgery parameter was consolidated into no primary site surgery or biopsy only versus surgery. The multivariate analysis was performed using a complete case analysis technique, and all variables entered into the model were used without any stepwise variable deletion or addition.


Patient and Tumor Characteristics

The patient and tumor characteristics are summarized in Table 1. No difference was observed in the race or sex of patients, because these variables were distributed equally between the 2 groups. Patients were more likely to have larger tumors in the extra-CNS group with a median greatest dimension of 7.0 cm versus 5.2 cm in the CNS group (P < .001). There appeared to be a slight predominance of older patients in the extra-CNS group with a mean age of 53 years compared with 49 years in the CNS group (P < .008). The extra-CNS category included tumors in the soft tissues, including the heart (42.2%); retroperitoneum, including the kidneys (6.7%); head and neck (6%); lung and bronchus (3.5%); and, finally, bones and joints (1.4%). Staging of patients was performed using the SEER criteria of localized, regional, or distant disease. These data were not applicable for the patients with CNS tumors. Within the extra-CNS group, 51.3% of patients had localized disease, 18.6% had regional disease, and 15.1% had distant disease. Patient age was analyzed as a categorical variable in both univariate and multivariate analyses. The age groupings that were used are provided in Tables 2 and 3.

Table 1. Patient Characteristics
 Total (N = 655)CNS (N = 199)Extra-CNS (N = 456) 
  • Abbreviations: CNS, central nervous system; Diff, differentiated; NA, not available; SD, standard deviation; SEER, Surveillance, Epidemiology, and End Results Program of the National Cancer Institute.

  • a

    Student t test.

  • b

    Pearson chi-square test.

  • c

    Fisher exact test.

Age, y       
 Median53.00 50.00 54.00  
 Mean51.94 48.88 53.27 .008a
 Range0-98 0-92 0-98  
 SD19.5 20.03 19.25  
Sex      .882b
Race      .730c
Vital status       
Grade      <.001c
 Well diff213.430.5183 
 Moderately diff6610.861609.8 
 Poorly diff335.491.5243.9 
SEER stage      <.001c
Table 2. Univariate Analysis
VariableHR95% CIP
  1. Abbreviations: CI, confidence interval; CNS, central nervous system; HR, hazard ratio.

Age, y   
Table 3. Multivariate Analysis (N = 610)a
 All PatientsCNS SiteExtra-CNS Site
VariableHR95% CIPHR95% CIPHR95% CIP
  • Abbreviations: CI, confidence interval; CNS, central nervous system; HR, hazard ratio.

  • a

    Grade and Surveillance, Epidemiology, and End Results (SEER) stage were not included because of the large number of patients with unknown tumor grade and SEER stage.

 Male1.00Reference 1.00Reference 1.00Reference 
 Female0.8240.655-1.038.1000.6770.37- 1.256.2160.8960.684-1.175.428
 White1.00Reference 1.00Reference 1.00Reference 
Age, y         
 <201.00Reference 1.00Reference 1.00  


The median follow-up was 73.3 months (range, 0-418 months). The OS rate at 1 year, 5 years, and 10 years was 94.1%, 79.3%, and 54.1%, respectively, for patients with CNS disease and 85.6%, 57.7%, and 44.2%, respectively, for patients with extra-CNS disease. The CSS rate at 1 year, 5 years, and 10 years was 95.7%, 87.7%, and 68.5%, respectively, for patients with CNS disease and 89.5%, 66%, and 56.8%, respectively, for patients with extra-CNS disease. Patients who had CNS tumors had better OS and CSS compared with patients who had extra-CNS on the basis of the log-rank test (P < .001 for both), as illustrated in Figures 1 and 2, respectively. Tumor site was the only significant positive predictor of OS on univariate analysis. Results from the univariate analysis are provided in Table 2. Potential predictors of OS that did not reach statistical significance included race and sex. Significant negative predictors of OS on multivariate analysis included extra-CNS tumor site (hazard ratio [HR], 1.611; P = .002) and older age (ages 40-59 years: HR, 1.974; P = .043; ages 60-79 years: HR, 3.781; P < .001; and aged ≥80 years: HR, 7.372; P < .001). Results from the multivariate analysis are provided in Table 3.

Figure 1.

Overall survival (OS) is illustrated according to disease site in patients with hemangiopericytoma. CNS indicates central nervous system.

Figure 2.

Cause-specific survival (CSS) is illustrated according to disease site in patients with hemangiopericytoma. CNS indicates central nervous system.


To our knowledge, the current analysis of 655 patients is the first of its kind to clearly define outcomes of patients with HPCs in CNS and extra-CNS sites and is also the largest HPC series to date. For this analysis, we used the large SEER database to analyze various prognostic factors associated with this rare tumor. In addition to our analysis, a review of the current literature available for HPCs is presented below and is focused on the roles of chemotherapy, surgery, and radiotherapy (RT) in both extra-CNS sites and CNS sites. It is noteworthy that series and case reports with <10 patients were excluded from our discussion.

Central Nervous System Versus Extra-Central Nervous System Disease

Because of the propensity for HPC to occur in both a CNS setting and an extra-CNS setting, the outcomes in these 2 different anatomic locations provide an important comparison. The vast majority of the currently published series examine CNS and extra-CNS disease as 2 separate entities and are summarized in Tables 4 and 5. Despite the complex management associated with CNS tumors, our analysis demonstrated that patients with CNS disease fared much better with respect to OS and CSS than patients with extra-CNS disease (Figs. 1, 2). These results were statistically significant (P < .001). Several key differences exist with regard to the management of patients who have CNS disease compared with those who have extra-CNS disease. It is possible that the disparity in outcomes among patients with CNS tumors and those with extra-CNS tumors may be attributed to differences in the rates of surgical resection. Given the significant limitations of SEER with regard to pathologic data, margin status, and details of surgical resection, an analysis of these variables was not included secondary to an excess number of patients who had unknown factors. In addition, a patient with CNS disease may be expected to present symptomatically at an earlier disease stage than a patient with extra-CNS disease because of the limited intracranial and spinal canal space. This earlier presentation also may result in improved outcome for the patient with a CNS tumor. The typical age at presentation for patients with HPC ranges from 32 years to 47 years in most previously published series.3, 4, 8, 18, 19, 23, 27 The SEER cohort falls slightly above this range with a mean age of 49 years in the CNS group and 53 years in the extra-CNS group, with an overall mean age of 52 years. Other patient characteristics in the SEER cohort are similar to those reported in the existing published literature. Sex and race were distributed evenly between the 2 groups, which is somewhat contradictory to past reports, in which CNS HPCs had a slight propensity to occur more frequently in men than in women.4, 22

Table 4. Select Hemangiopericytoma Extra-Central Nervous System Series
  No. of Patients/Total No. (%) No. of Patients/Total No. (%) 
ReferenceNo. of PatientsRT: Neoadjuvant or Adjuvant With Dose Information, if AvailableSurgeryChemoOS RateLR RateMetastasisMost Common Location
  1. Abbreviations: Chemo, chemotherapy; central nervous system; Gy, grays; LR, local recurrence; OS, overall survival; NR, not reported.

McMaster 1975216023/60 (38.3)38/60 (63.3)2/60 (3.3)NR22/60 (36)64.5%Thigh and groin
Enzinger & Smith 19763106NR100%5/106 (4.7)70% Survival at 10 y17.2%15% (of 71 patients with long-term follow-up)Thigh and retroperitoneum
Jha 1987331414/14 (100); mean dose, 57.4 Gy11/14 (78.5)5/14 (35.7)NR5/14 (35.7)6/14 9(42.8)Retroperitoneum and extremity
Staples 1990291510/15 (66.7); median dose, 60 Gy12/15 (80)6/15 (40)53%NR3/15 (20)Lower extremity
Spitz 199853616/36 (44.4); median dose, 57 Gy28/36 (78)10/36 (27.8)71% Actuarial survival at 5 y9/28 (38) Treated with curative intent10/36 (27.8)Visceral and Retroperitoneum
Current series456NRNRNR57.7% OS at 5 yNRNRSoft tissue, including heart
Table 5. Select Central Nervous System Hemangiopericytomas Series
  No. of Patients/Total No. (%) No. of Patients/Total No. (%) 
ReferenceNo. of PatientsRT: Neoadjuvant or Adjuvant With Dose Information, if AvailableSurgeryChemoOS RateLocal RecurrenceExtraneural MetastasisMost Common Location
  1. Abbreviations: Chemo, chemotherapy; central nervous system; Gy, grays; KM, Kaplan-Meier; OS, overall survival; NA, not available; NR, not reported; SRS, stereotactic surgery.

Jaaskelainen 19858212/10 (20)100%0%NR8/21 (38)NRNA
Guthrie 198944417/54 (31.4); mean dose, 41.5 Gy100%0%67% at 5 y29/44 (65.910/44 (22.7)NA
Mena 1991229433/94 (35.1)100%11/94 (11.7)Median, 58.5 mo57/94 (60.6)22/94 (23.4)NA
Bastin & Mehta 1992307223/72 (31.9)NR0%Surgery alone, 65 mo; surgery + RT, 96 moSurgery alone, 90% at 9 y; surgery + RT, 48%NR: CNS-specific, 21% overall, including extra-CNSNA
Dufor 200192112/21 (57.1); dose, 5000-6400 rads100%1/21 (4.8)NRSurgery alone, 88%; surgery + RT, 12.5%4/21 (19)NA
Ecker 2003323819/38 (50); mean dose, 57 Gy100%11/38 (29)Diagnosed since 1990, 28 patients; 93% KM OS at 5 yNR11/38 (29)NA
Kim 2003233117/31 (54.4); conventional RT, 11 patients; SRS, 6 patients24/31 (77.4)4/31 (12.9)NR12/31 (38.7)4/31 (12.9)NA
Soyeur 2004242910/24 (41.7); mean dose, 54 Gy for cranial RT patients100%4/29 (13.8)85% OS at 5 y, 43% OS at 15 yNR16/29 (55.2)NA
Combs 2005343725/37 (67.6); mean dose, 54 Gy23/37 (61.2)0 (0)69% Actuarial survival at 5-yNRNRNA
Current series199NRNRNR79.3% OS at 5 yNRNRNA

In the current series, the 5-year and 10-year OS rates for patients in the CNS group were 79.3% ± 1.7% and 54.7% ± 6.4%, respectively; and the 5-year and 10-years OS rates for patients in the extra-CNS group were 57.7% ± 2.4% and 44.2% ± 2.6%, respectively. These rates are lower than those reported in other large published series, in which the 5-year OS rate ranged between 67% and 85%, and the 10-year OS rate ranged between 40% and 70% for both CNS tumors and extra-CNS tumors.3-5, 24 Unique to our series is a statistically significant worse outcome for patients who had extra-CNS disease compared with patients who had CNS disease on the basis of OS and CSS (P < .001).

Surgical Resection

Surgery is the principle modality of treatment for HPCs, as established by multiple published series.4, 5, 21, 22, 28-31 The importance of gross total surgical resection for HPCs applies to tumors located in both CNS and extra-CNS sites. The importance of a total surgical resection is emphasized in the findings by Guthrie et al, who demonstrated that total tumor resection resulted in better OS compared with subtotal resection.4 It is important to note that the significance of total resection compared with subtotal resection has not been reproduced in other major series.24, 32

Radiation Therapy

It appears from a review of many different published series that RT, especially when combined with surgical resection, may provide an important addition to the treatment of patients with HPC. In the extra-CNS setting, several series have specifically examined the role of RT.3, 5, 21, 28, 29, 33 When examining past published experiences, it is important to consider that many of the series involving RT have spanned decades, during which the delivery of RT has evolved extensively. In a series by Mira et al focused on RT in patients with extra-CNS HPCs, 26 of 29 treated HPC tumors (90%) had some measurable radiographic response to a course of external-beam RT. Remarkably, there were 14 instances of complete tumor response, which represented nearly 50% of the patients. The major conclusion of that series was that RT was beneficial in terms of tumor response and should be combined with wide local excision.28 The SEER data examining the role of RT in the management of these HPCs was not included secondary to a lack of detailed RT dose and volume information. A complete summary of the current extra-CNS series is presented in Table 4.

The role of RT in the CNS setting has been examined in several different series.4, 8, 9, 22-24, 30, 32, 34 In 1 of the largest series reported by Mena and colleagues, 94 patients with CNS HPCs are reported.22 A substantial conclusion of that series was that, regardless of pathologic diagnosis of either anaplastic or differentiated HPC, survival was better for patients who received adjuvant therapy after surgical resection. The vast majority of the adjuvant therapy in that series consisted of RT, with a small percentage of the others consisting of chemotherapy. The median survival in that series for patients who did not receive adjuvant therapy was 99 months compared with 126 months for those who did receive adjuvant therapy. In addition, in patients with anaplastic HPC, the median survival without adjuvant therapy was 44 months versus 67.5 months with adjuvant therapy. That series provided additional evidence of the importance of postoperative RT for all patients with CNS HPCs. Dufor et al presented a series with special emphasis on the importance of RT in preventing local recurrence and neural axis recurrences in patients with meningeal HPCs.9 In that series of 21 patients, 8 of 17 patients with intracranial HPC underwent surgery followed by postoperative RT (dose range, 50-64 grays), and the remaining 9 patients underwent surgery alone. Of the patients who received postoperative RT, the recurrence rate was 12.5%; and, of the patients who did not receive postoperative RT, the recurrence rate was 88% (P < .05). Major conclusions were that, in patients with intracranial HPCs, postoperative RT reduced the rate of local recurrence. However, it did not appear that postoperative RT reduced the rate of neural axis metastasis.9 It is also important to consider the role of radiosurgery in the treatment of CNS HPCs, because radiosurgery produced favorable outcomes in some series. However, it is difficult to draw firm conclusions regarding the efficacy of radiosurgery in the treatment of CNS HPCs, because reports have been limited to small, retrospective series.35-37 The current major CNS series are summarized in Table 5.


Four past series focusing on HPCs in the extra-CNS setting have used chemotherapy at some point during treatment, and conclusions from each of these series demonstrate little value to the use of chemotherapy.5, 21,29,21 Spitz et al reported the largest series involving chemotherapy to date and also observed limited benefit to the addition of chemotherapy.5 Chemotherapy has proven to be of even less value in the CNS setting; multiple series have indicated little or no benefit to a wide variety of agents.9, 22, 23, 32 A notable limitation of the SEER database is that there is no chemotherapy information available for analysis.

In conclusion, our current report represents the largest currently published series to date of outcomes in patients with HPC and serves to highlight the outcome disparity between CNS tumors and extra-CNS tumors. We observed superior OS and CSS outcomes among patients who had CNS HPCs compared with those who had extra-CNS HPCs. This outcome disparity probably is multifactorial; and, because data available in the SEER database are limited, the sources of this outcome disparity could not be firmly identified. We recognize the availability of limited surgical treatment data and limited RT treatment data in the SEER database; however, because of a substantial paucity of details surrounding these data, we elected to omit them from our analysis. The findings from this analysis generate the hypothesis that more aggressive treatment of extra-CNS HPCs may be warranted because of their worse OS and CSS outcomes. Like all SEER analyses and retrospective reviews, these data should be interpreted with caution given the unknown factors in the SEER database along with other potential biases inherent to retrospective reviews.


No specific funding was disclosed.


The authors made no disclosures.