J Neurooncol (2014) 116:609–616 DOI 10.1007/s11060-013-1345-6
CLINICAL STUDY
Pediatric and adult malignant peripheral nerve sheath tumors: an analysis of data from the surveillance, epidemiology, and end results program E. Susan Amirian • J. Clay Goodman Pamela New • Michael E. Scheurer
•
Received: 30 August 2013 / Accepted: 29 December 2013 / Published online: 5 January 2014 Ó Springer Science+Business Media New York 2014
Abstract Malignant peripheral nerve sheath tumors (MPNSTs) are rare soft tissue sarcomas that arise predominantly from Schwann cells. Despite the fact that MPNSTs have high local recurrence rates and are generally associated with poor prognosis, little is known about prognostic factors or effective clinical management for this tumor type. The purpose of this study was to describe the distributions of patient and tumor characteristics and to identify predictors of cause-specific survival among MPNST cases reported to SEER between 1973 and 2008. Patient and tumor characteristics were compared between pediatric and adult MPNST cases. Cox regression and treebased survival analysis were used to examine factors associated with MPNST-related mortality separately among adults and children. A total of 1,315 MPNST cases were isolated from the 1973–2008 SEER dataset. Among pediatric cases, sex, race, and radiation therapy predicted MPNST survival, whereas among adults, tumor site, tumor grade, number of primary tumors, and tumor size were significant predictors. As tumor size at diagnosis/resection may be the only somewhat ‘‘modifiable’’ prognostic factor, future studies should aim to identify biological and social attributes associated with tumor size at diagnosis, E. S. Amirian M. E. Scheurer (&) Department of Pediatrics & Dan L. Duncan Cancer Center, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030, USA e-mail: [email protected] J. C. Goodman Department of Pathology & Immunology and Neurology, Baylor College of Medicine, Houston, TX, USA P. New Department of Neurosurgery, The Methodist Hospital Research Institute, Houston, TX, USA
separately among individuals with and without NF-1, in order to help identify earlier opportunities for clinical intervention. Keywords Survival
Nerve sheath neoplasms SEER Sarcoma
Introduction Malignant peripheral nerve sheath tumors (MPNSTs) are extremely rare soft tissue sarcomas that are derived from cells of the peripheral nerve sheath, such as Schwann cells [1, 2]. MPNST incidence in the general population is only 0.001 % [3], but among individuals affected by neurofibromatosis type 1 (NF-1), the incidence ranges from 2 to 29 % [4], with an estimated lifetime risk of *10 % [2, 5]. Despite their rarity in the general population, MPNSTs constitute about 5–10 % of all soft tissue sarcomas [6], and are one of the most common non-rhabdomyosarcomatous soft tissue sarcomas among children [7, 8]. Peak ages at diagnosis are 30–40 years of age among NF-1-affected populations and 70–80 years of age among the general population [2, 5]. However, pediatric cases (B20 years of age) still account for 10–20 % of all MPNSTs [1]. MPNSTs have extremely high recurrence rates (40–50 %), and are associated with poor prognosis despite aggressive treatment [2, 6, 9]. Yet, little is known about etiologic and prognostic factors or effective clinical management for this tumor type. Other than NF-1, another predisposing factor that may increase risk for these tumors is radiation exposure [10–13]. However, radiation exposure would likely explain only a very small proportion of primary MPNSTs that develop among individuals with no prior history of cancer. Similarly, factors that predict
123
610
survival also remain largely enigmatic [1, 6], and even the effect of NF-1 on patient survival is somewhat inconsistently reported in the published literature [3, 14]. With regard to clinical management, a consensus seems to be that wide resection, with possible appendicular amputation when applicable, may be necessary [1, 6, 15]. However, the enormous amount of variability in tumor site, clinical presentation, and histologic subtype can, obviously, influence the treatment plan. Despite the lack of knowledge on this tumor type, MPNSTs remain understudied, which may be partially attributable to their rarity [1]. In this study, we used the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program data to describe the characteristics of one of the largest available case series of MPNSTs and to identify predictors of MPNST survival.
Methods Study population: SEER data The full 1973–2008 SEER dataset (Nov. 2011 release) was obtained, which contains data on the following registries (with year of entry into the SEER program): Arizona (1973), Connecticut (1973), Detroit (1973), Hawaii (1973), Iowa (1973), New Mexico (1973), San Francisco-Oakland (1973), Utah (1973), Seattle-Puget Sound (1974), Atlanta (1975), Rural Georgia (1978), Los Angeles (1992), San Jose-Monterey (1992), Alaska (1999), Greater California (2000), Kentucky (2000), Louisiana (2000), and New Jersey (2000). The SEER dataset encompasses *28 % of the U.S. population [16]. From this dataset, we isolated MPNSTs (ICD-O-3 code 9540/3), and acquired information on tumor location, tumor size (available for 1988–2008), tumor grade, age at diagnosis, month/year of diagnosis, race/ethnicity, sex, cancer-directed therapy, vital status, survival time, and cause of death. Vital status was ascertained through December 31, 2008. Statistical analyses Population characteristics were compared between pediatric (\18 years of age) and adult cases by v2 or Student’s t test, as appropriate. Because MPNSTs can develop anywhere on the body, there were originally over 52 tumor sites, which were re-categorized for the survival analyses as follows: central nervous system, head and neck, limbs, trunk and core, and other/unknown. The other/unknown category was largely comprised of tumors of unknown sites, autonomic nervous system NOS (not otherwise specified), and the skin. Tumor size was only available for a subset of cases reported between 1988 and 2008. Because
123
J Neurooncol (2014) 116:609–616
tumor size may help determine the clinical course of treatment, treatment type was examined by tumor size categories (\40 mm, 4–7, and [7 cm). Tumor grade and location were also examined in relation to tumor size. Cause-specific death was used as the outcome of interest for the survival analyses. A total of 37 cases with zero or missing survival time were excluded from the survival analyses. Separate regression models were built among adult and pediatric cases. Cancer survival over time was visualized with Kaplan– Meier survival curves, and the log-rank tests were used to assess whether survival probability differed by characteristics of interest (a = 0.05). Using Cox proportional hazards regression, adjusted hazard ratios were calculated for the following covariates: tumor location, tumor grade, age at diagnosis, month/year of diagnosis, race/ethnicity, sex, radiation, and extent of surgery. Because tumor size was unavailable for a large subset of cases, separate survival analyses, which included tumor size as a covariate, were run on the subset of cases for whom this information was available. Tree-based survival analysis (TBSA) was conducted to supplement the results from the Cox model and to help identify the combinations of attributes shared among cases with the highest relative mortality risk. TBSA is a datamining tool, similar to other regression tree analyses, that partitions the study population into groups that have different survival distributions, thus helping to identify hierarchical predictive relationships (and higher-order interactions between variables) [17, 18]. The log-rank statistic was utilized to identify statistically significant (p = 0.05) splits in the dataset until either there were no more significant splits remaining or the nodes reached the minimum size (n C 10). Covariates that were included in the TBSA modeling were tumor site, sex, radiation therapy, surgical intervention, tumor grade, decade of diagnosis, age at diagnosis, and race. Individuals who had missing values or had a tumor site of other/unknown were excluded from this analysis. TBSA was conducted in Stata version 10.0 (StataCorp, College Station, TX, USA; adapted from [19]), and other analyses were conducted in SAS version 9.3 (SAS, Cary, NC, USA).
Results A total of 1,315 MPNST cases were isolated from the 1973–2008 SEER dataset. Just under a third of the study population had died of MPNST-related causes by the end of the follow-up period (31 % of adults and 31 % of children). The majority of cases were reported after the year 2000, likely as a result of the addition of four registries in that year (Table 1). There was a slight
J Neurooncol (2014) 116:609–616 Table 1 Patient and tumor characteristics among pediatric and adult MPNST patients reported to the Surveillance, Epidemiology, and End Results program (1973–2008)
611
Characteristic
Overall (n = 1,315) n (%)
Pediatric (n = 128) n (%)
Adult (n = 1,187) n (%)
p-value
Age \18
128 (9.7)
18 to 59
803 (61.1)
C60
384 (29.2)
Median ± SD
45 ± 21.4
13 ± 4.4
48 ± 19.1
721 (54.8) 594 (45.2)
69 (53.9) 59 (46.1)
652 (54.9) 535 (45.1)
Non-hispanic white
874 (66.5)
63 (49.2)
811 (68.3)
Black
180 (13.7)
24 (18.8)
156 (13.1)
Sex Male Female
0.83
Race
Hispanic
172 (13.1)
34 (26.6)
138 (11.6)
Asian and other
89 (6.8)
7 (5.5)
82 (6.9)
Debulking (gross total or radical)
425 (32.3)
48 (37.5)
377 (31.8)
Less than gross total resection
875 (66.5)
78 (60.9)
797 (67.1)
Unknown
15 (1.1)
2 (1.6)
13 (1.1)
Yes
538 (40.9)
51 (39.8)
487 (41.0)
No
744 (56.6)
74 (57.8)
670 (56.4)
Unknown
33 (2.5)
3 (2.3)
30 (2.5)
1,070 (81.4)
117 (91.4)
953 (80.3)
245 (18.6)
11 (8.6)
234 (19.7)
Central nervous system
70 (5.3)
8 (6.3)
62 (5.2)
Head & neck
185 (14.1)
20 (15.6)
165 (13.9)
Extent of surgery
0.26
Radiation therapy
Number of primary tumors 1 C2
CLINICAL STUDY
Pediatric and adult malignant peripheral nerve sheath tumors: an analysis of data from the surveillance, epidemiology, and end results program E. Susan Amirian • J. Clay Goodman Pamela New • Michael E. Scheurer
•
Received: 30 August 2013 / Accepted: 29 December 2013 / Published online: 5 January 2014 Ó Springer Science+Business Media New York 2014
Abstract Malignant peripheral nerve sheath tumors (MPNSTs) are rare soft tissue sarcomas that arise predominantly from Schwann cells. Despite the fact that MPNSTs have high local recurrence rates and are generally associated with poor prognosis, little is known about prognostic factors or effective clinical management for this tumor type. The purpose of this study was to describe the distributions of patient and tumor characteristics and to identify predictors of cause-specific survival among MPNST cases reported to SEER between 1973 and 2008. Patient and tumor characteristics were compared between pediatric and adult MPNST cases. Cox regression and treebased survival analysis were used to examine factors associated with MPNST-related mortality separately among adults and children. A total of 1,315 MPNST cases were isolated from the 1973–2008 SEER dataset. Among pediatric cases, sex, race, and radiation therapy predicted MPNST survival, whereas among adults, tumor site, tumor grade, number of primary tumors, and tumor size were significant predictors. As tumor size at diagnosis/resection may be the only somewhat ‘‘modifiable’’ prognostic factor, future studies should aim to identify biological and social attributes associated with tumor size at diagnosis, E. S. Amirian M. E. Scheurer (&) Department of Pediatrics & Dan L. Duncan Cancer Center, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030, USA e-mail: [email protected] J. C. Goodman Department of Pathology & Immunology and Neurology, Baylor College of Medicine, Houston, TX, USA P. New Department of Neurosurgery, The Methodist Hospital Research Institute, Houston, TX, USA
separately among individuals with and without NF-1, in order to help identify earlier opportunities for clinical intervention. Keywords Survival
Nerve sheath neoplasms SEER Sarcoma
Introduction Malignant peripheral nerve sheath tumors (MPNSTs) are extremely rare soft tissue sarcomas that are derived from cells of the peripheral nerve sheath, such as Schwann cells [1, 2]. MPNST incidence in the general population is only 0.001 % [3], but among individuals affected by neurofibromatosis type 1 (NF-1), the incidence ranges from 2 to 29 % [4], with an estimated lifetime risk of *10 % [2, 5]. Despite their rarity in the general population, MPNSTs constitute about 5–10 % of all soft tissue sarcomas [6], and are one of the most common non-rhabdomyosarcomatous soft tissue sarcomas among children [7, 8]. Peak ages at diagnosis are 30–40 years of age among NF-1-affected populations and 70–80 years of age among the general population [2, 5]. However, pediatric cases (B20 years of age) still account for 10–20 % of all MPNSTs [1]. MPNSTs have extremely high recurrence rates (40–50 %), and are associated with poor prognosis despite aggressive treatment [2, 6, 9]. Yet, little is known about etiologic and prognostic factors or effective clinical management for this tumor type. Other than NF-1, another predisposing factor that may increase risk for these tumors is radiation exposure [10–13]. However, radiation exposure would likely explain only a very small proportion of primary MPNSTs that develop among individuals with no prior history of cancer. Similarly, factors that predict
123
610
survival also remain largely enigmatic [1, 6], and even the effect of NF-1 on patient survival is somewhat inconsistently reported in the published literature [3, 14]. With regard to clinical management, a consensus seems to be that wide resection, with possible appendicular amputation when applicable, may be necessary [1, 6, 15]. However, the enormous amount of variability in tumor site, clinical presentation, and histologic subtype can, obviously, influence the treatment plan. Despite the lack of knowledge on this tumor type, MPNSTs remain understudied, which may be partially attributable to their rarity [1]. In this study, we used the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program data to describe the characteristics of one of the largest available case series of MPNSTs and to identify predictors of MPNST survival.
Methods Study population: SEER data The full 1973–2008 SEER dataset (Nov. 2011 release) was obtained, which contains data on the following registries (with year of entry into the SEER program): Arizona (1973), Connecticut (1973), Detroit (1973), Hawaii (1973), Iowa (1973), New Mexico (1973), San Francisco-Oakland (1973), Utah (1973), Seattle-Puget Sound (1974), Atlanta (1975), Rural Georgia (1978), Los Angeles (1992), San Jose-Monterey (1992), Alaska (1999), Greater California (2000), Kentucky (2000), Louisiana (2000), and New Jersey (2000). The SEER dataset encompasses *28 % of the U.S. population [16]. From this dataset, we isolated MPNSTs (ICD-O-3 code 9540/3), and acquired information on tumor location, tumor size (available for 1988–2008), tumor grade, age at diagnosis, month/year of diagnosis, race/ethnicity, sex, cancer-directed therapy, vital status, survival time, and cause of death. Vital status was ascertained through December 31, 2008. Statistical analyses Population characteristics were compared between pediatric (\18 years of age) and adult cases by v2 or Student’s t test, as appropriate. Because MPNSTs can develop anywhere on the body, there were originally over 52 tumor sites, which were re-categorized for the survival analyses as follows: central nervous system, head and neck, limbs, trunk and core, and other/unknown. The other/unknown category was largely comprised of tumors of unknown sites, autonomic nervous system NOS (not otherwise specified), and the skin. Tumor size was only available for a subset of cases reported between 1988 and 2008. Because
123
J Neurooncol (2014) 116:609–616
tumor size may help determine the clinical course of treatment, treatment type was examined by tumor size categories (\40 mm, 4–7, and [7 cm). Tumor grade and location were also examined in relation to tumor size. Cause-specific death was used as the outcome of interest for the survival analyses. A total of 37 cases with zero or missing survival time were excluded from the survival analyses. Separate regression models were built among adult and pediatric cases. Cancer survival over time was visualized with Kaplan– Meier survival curves, and the log-rank tests were used to assess whether survival probability differed by characteristics of interest (a = 0.05). Using Cox proportional hazards regression, adjusted hazard ratios were calculated for the following covariates: tumor location, tumor grade, age at diagnosis, month/year of diagnosis, race/ethnicity, sex, radiation, and extent of surgery. Because tumor size was unavailable for a large subset of cases, separate survival analyses, which included tumor size as a covariate, were run on the subset of cases for whom this information was available. Tree-based survival analysis (TBSA) was conducted to supplement the results from the Cox model and to help identify the combinations of attributes shared among cases with the highest relative mortality risk. TBSA is a datamining tool, similar to other regression tree analyses, that partitions the study population into groups that have different survival distributions, thus helping to identify hierarchical predictive relationships (and higher-order interactions between variables) [17, 18]. The log-rank statistic was utilized to identify statistically significant (p = 0.05) splits in the dataset until either there were no more significant splits remaining or the nodes reached the minimum size (n C 10). Covariates that were included in the TBSA modeling were tumor site, sex, radiation therapy, surgical intervention, tumor grade, decade of diagnosis, age at diagnosis, and race. Individuals who had missing values or had a tumor site of other/unknown were excluded from this analysis. TBSA was conducted in Stata version 10.0 (StataCorp, College Station, TX, USA; adapted from [19]), and other analyses were conducted in SAS version 9.3 (SAS, Cary, NC, USA).
Results A total of 1,315 MPNST cases were isolated from the 1973–2008 SEER dataset. Just under a third of the study population had died of MPNST-related causes by the end of the follow-up period (31 % of adults and 31 % of children). The majority of cases were reported after the year 2000, likely as a result of the addition of four registries in that year (Table 1). There was a slight
J Neurooncol (2014) 116:609–616 Table 1 Patient and tumor characteristics among pediatric and adult MPNST patients reported to the Surveillance, Epidemiology, and End Results program (1973–2008)
611
Characteristic
Overall (n = 1,315) n (%)
Pediatric (n = 128) n (%)
Adult (n = 1,187) n (%)
p-value
Age \18
128 (9.7)
18 to 59
803 (61.1)
C60
384 (29.2)
Median ± SD
45 ± 21.4
13 ± 4.4
48 ± 19.1
721 (54.8) 594 (45.2)
69 (53.9) 59 (46.1)
652 (54.9) 535 (45.1)
Non-hispanic white
874 (66.5)
63 (49.2)
811 (68.3)
Black
180 (13.7)
24 (18.8)
156 (13.1)
Sex Male Female
0.83
Race
Hispanic
172 (13.1)
34 (26.6)
138 (11.6)
Asian and other
89 (6.8)
7 (5.5)
82 (6.9)
Debulking (gross total or radical)
425 (32.3)
48 (37.5)
377 (31.8)
Less than gross total resection
875 (66.5)
78 (60.9)
797 (67.1)
Unknown
15 (1.1)
2 (1.6)
13 (1.1)
Yes
538 (40.9)
51 (39.8)
487 (41.0)
No
744 (56.6)
74 (57.8)
670 (56.4)
Unknown
33 (2.5)
3 (2.3)
30 (2.5)
1,070 (81.4)
117 (91.4)
953 (80.3)
245 (18.6)
11 (8.6)
234 (19.7)
Central nervous system
70 (5.3)
8 (6.3)
62 (5.2)
Head & neck
185 (14.1)
20 (15.6)
165 (13.9)
Extent of surgery
0.26
Radiation therapy
Number of primary tumors 1 C2
