2191 Subdiaphragmatic and Intrathoracic Paraspinal Malignant Peripheral Nerve Sheath Tumors A Clinicopathologic Study of 25 Patients and 26 Tumors Helen P. Kourea, M.D.1 Mark H. Bilsky, M.D.2 Denis H. Y. Leung, Ph.D.3 Jonathan J. Lewis, M.D., Ph.D.2 James M. Woodruff, M.D.1 1 Department of Pathology, Memorial Sloan-Kettering Cancer Center, New York, New York. 2 Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York. 3 Department of Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, New York. Presented in part at the annual meeting of the U.S. and Canadian Academy of Pathology, Washington, DC, March 1996. Supported by Grant No. CA#47179 from the National Institutes of Health. The authors thank the following colleagues for providing material and information on individual cases: Dr. Manuel Banez, Dover, DE; Dr. Monica Gallivan, Washington, DC; Dr. Thomas Gouge, New York, NY; Dr. Sigurd Gundersen, La Crosse, WI; Dr. Syed Hoda, New York, NY; Dr. Karen van Hoeven, Philadelphia, PA; Dr. Ralph Hruban, Baltimore, MD; Dr. Anantha Manepalli, St. Louis, MO; Dr. Jonathan Melamed, New York, NY; Dr. Arthur Miller, New York, NY; Dr. Thomas Rengel, Wausau, WI; Dr. Susan Wester, La Crosse, WI; and Dr. John C. Whitsell II, New York, NY. Address for reprints: James M. Woodruff, M.D., Department of Pathology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021. Received October 28, 1997; accepted November 28, 1997. © 1998 American Cancer Society BACKGROUND. To determine the effects of anatomic site on the presentation and diagnosis of malignant peripheral nerve sheath tumors (MPNSTs) and on the treatment and outcomes of the patients, the authors initiated a study of these tumors at different sites. An earlier report described MPNSTs of the buttock and lower extremity, and the current series analyzes those presenting at intrathoracic (IT) and subdiaphragmatic (SD) paraspinal sites. METHODS. The authors reviewed data on patients with paraspinal MPNSTs who were seen at Memorial Hospital during the period 1960 –1995 and for whom histologic slides were available. Various clinicopathologic parameters and their effects on patient outcomes were examined. RESULTS. Twenty-five patients with 26 tumors were evaluated. Seven tumors were IT and 19 were SD; 60% of the patients had neurofibromatosis type 1 (NF1). Most patients presented with pain, and a diagnostic delay (of 3 months to 2 years) was often noted. Mean tumor sizes for SD and IT tumors were 14.3 cm and 6.6 cm, respectively. Most MPNSTs were composed of spindle cells in fascicles. Twentyseven percent exhibited divergent differentiation. Twenty-four tumors were high grade, and a low grade component was identified in 8 tumors. Surgical resection was attempted for 23 tumors (88%), but complete resection was achieved in only 6 cases (23%). Eighty percent of the patients died of their tumors, 2-year and 5-year survival rates were 35% and 16%, and median survival was 8.5 months. Significant prognostic factors were tumor size ,5 cm, the presence of a low grade component, and complete tumor resection. CONCLUSIONS. Paraspinal MPNSTs have more aggressive behavior than peripherally located tumors, mainly because of the difficulty encountered in resecting them completely. Prognoses of patients with MPNST at this site appear to be affected by resection status, tumor size, and tumor grade. Cancer 1998;82:2191–203. © 1998 American Cancer Society. KEYWORDS: malignant peripheral nerve sheath tumor, malignant schwannoma, paraspinal tumor, neurofibromatosis, peripheral nervous system, pathology, treatment, surgery, prognosis. M alignant peripheral nerve sheath tumors (MPNSTs) are neoplasms arising from or recapitulating the innate cellular constituents of the peripheral nerve sheath, excluding tumors of the epineurium or peripheral nerve vasculature.1 Although most examples are assumed to develop from Schwann cells,2 some may show evidence of fibroblastic or perineurial cell differentiation;3 therefore, the noncommittal term MPNST is preferred over malignant schwannoma. MPNSTs, which occur infrequently, are usually high grade and 2192 CANCER June 1, 1998 / Volume 82 / Number 11 clinically aggressive. A number of different possible prognostic factors have been cited, such as the presence of neurofibromatosis type 1 (NF1),4 –7 tumor size,6 –10 mitotic activity,8 –9 necrosis,7 completeness of resection,6,10,11 and postirradiation occurrence;12 however, the importance that has been attributed to each has varied among different studies. For this reason, prognostic parameters for these tumors are not well established. Tumor location, which is known to be important to the presentation, diagnosis, treatment, and prognosis of other malignant soft tissue tumors,13 has not been studied in detail for MPNSTs. We therefore undertook a review of these tumors by anatomic site to assess the role of tumor location. In the initial report in this series, we analyzed tumors of the buttock and lower extremity (BLE).9 The current study examines clinicopathologic parameters of paraspinal MPNSTs in subdiaphragmatic (SD) and intrathoracic (IT) locations and compares the outcomes of patients with MPNSTs at different locations. MATERIALS AND METHODS The files of the Department of Pathology at the Memorial Sloan-Kettering Cancer Center (MSKCC) from the period 1960 –1995 and the personal files of one of the authors (J.M.W.) were reviewed, with the aim of identifying neoplasms at paraspinal locations in the thoracic cavity, retroperitoneum, and pelvis that were diagnosed as MPNST, malignant schwannoma, neurogenic sarcoma, or neurofibrosarcoma. Tumors arising in the thoracic and abdominal walls or intra-abdominal organs were excluded. Only tumors for which histologic slides were available were included in the study. Criteria for diagnosis were the same as those used in the first study in this series.9 The diagnosis of MPNST was made for tumors having the typical morphology,8,9 with supporting immunohistochemical (S100 protein expression)14,15 or ultrastructural evidence of nerve sheath differentiation.16 Furthermore, this diagnosis was rendered for malignant tumors clearly arising in a peripheral nerve, schwannoma, or neurofibroma, and also for those arising in patients with NF1 and morphology consistent with MPNST,8,9 even if supportive immunohistochemical or ultrastructural data were lacking.6,8,17 Not included were tumors of epineurial soft tissue and endothelial tumors arising from peripheral nerve vasculature. All available medical records at MSKCC and other institutions for the patients in this study were reviewed, and further follow-up information was obtained by contacting the referring physician when necessary. A clear-cut statement that the patient had NF1 was accepted as evidence of the disorder. The clinical parameters examined were age; gender; history of NF1; type and duration of symptoms; past medical and surgical history, including history of radiation therapy; tumor location (subdiaphragmatic [SD] vs. intrathoracic [IT]); spinal tumor extension; treatment modalities; recurrence; metastasis; and outcome. All pathology reports and microscopic slides available were reexamined by two of the authors (H.P.K. and J.M.W.). For patients who underwent more than 1 operation, 4 – 47 (average, 21) slides of tumor from all surgical specimens, including those from reexcisions or recurrences, were reviewed. Excluding the biopsies, tumor samplings of each major resection specimen provided 0.3– 4 (average, 1.8) hematoxylin and eosin– stained slides per cm of tumor. In 6 of 7 consultation cases, slides from all paraffin blocks were available; the sampling provided 1–2 slides per cm of tumor. In one consultation case, only sample slides were submitted. Immunohistochemical studies were performed according to published methods,18 when indicated, to confirm nerve sheath origin and to exclude other tumors, such as leiomyosarcomas and synovial sarcomas. The antibodies used were against S-100 protein (polyclonal, DAKO, Carpinteria, CA, dilution 1:10000), desmin (monoclonal, DAKO, dilution 1:50), smooth muscle antibody (SMA) (monoclonal, DAKO, dilution 1:2000), HHF-35 (monoclonal, DAKO, dilution 1:100), and CAM5.2 (monoclonal, Becton Dickinson Labs, dilution 1:200). Histopathologic parameters surveyed were tumor size, histologic grade, mitotic rate, necrosis, and the presence of divergent differentiation. Two histologic grades, low and high, were recognized. Low grade MPNSTs are more cellular than neurofibromas with nuclear enlargement to three times the size of neurofibroma nuclei and nuclear hyperchromasia.1 Mitotic figures are often present but rarely number more than 3 per 10 high-power fields (HPF), and their presence is not necessary for the diagnosis of malignancy because small biopsies may be devoid of mitotic activity. High grade MPNSTs are characterized by nuclear enlargement, hyperchromasia, and usually marked cellular crowding. Commonly mitoses exceed 4 per 10 HPF. Necrosis is often present but is not necessary to diagnose a high grade MPNST. Tumors containing both low and high grade areas are traditionally classified as high grade, regardless of the amount of the latter component. In this study, to evaluate the effect of tumor grade on outcome, the presence of even a minor low grade component was recorded, and these Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. tumors were compared with entirely high grade MPNSTs by statistical analysis. Determination of the surgical procedure performed was based on patients’ clinical records, including operative notes and histologic evaluation of the surgical margins. Complete resection refers to cases that had both gross and microscopically negative margins. Subtotal resection was defined as such by the surgeon and confirmed by pathologic examination. Grossly complete resection indicated cases that were regarded by the surgeon as completely resected; of these cases, some had microscopically positive margins, whereas others were indeed complete resections. Open biopsy specimens were the only specimens available in unresectable cases. Statistical analysis to evaluate the effect of clinical and pathologic parameters on survival was performed. Parameters analyzed were age, gender, history of NF1, diagnostic delay, tumor location (SD vs. IT), tumor size, grade, mitotic rate, necrosis, divergent differentiation, postirradiation occurrence, and type of surgical treatment. Fisher’s exact test was used to study the associations among categoric factors and to analyze their influence on survival status. Age and mitotic index were evaluated using the Wilcoxon–Mann– Whitney test. Analysis of survival time was performed by the log rank test and the Kaplan–Meier method. Because of the relatively small number of cases in this study, multivariate analysis was not performed. Statistical significance was set at P , 0.05. RESULTS Twenty-six tumors from 25 patients satisfied the criteria for inclusion in the study. Seven cases originated from the consultation files of J.M.W. Fifteen patients underwent surgery at MSKCC and 7 at other institutions. Tumors in the remaining three patients were only biopsied. Individual patient data and a summary of clinicopathologic findings are presented in Tables 1 and 2. 2193 sarcoma’’ of the lumbar spine (Case 12), Hodgkin’s disease (Case 24), and breast carcinoma (Case 23). Detailed data regarding the clinical presentation were available for 20 patients. Pain, noted in 16 cases, was the most common presenting symptom. SD tumors often produced pain localized in the back or radiating to the leg, whereas patients with IT tumors most often presented with localized chest pain. Weakness, paresis, or paresthesia were reported in five cases. Less frequently occurring symptoms were anorexia, weight loss, fatigue, nausea and vomiting, dysphagia, shortness of breath, and leg edema during the last trimester of pregnancy. In one case the tumor was discovered incidentally during the workup for menometrorrhagia (Case 15), and in another during follow-up for an ovarian mucinous cystadenocarcinoma (Case 22); the latter patient had NF1 and two MPNSTs. Delay in making the diagnosis, ranging from 3 months to 2 years after symptomatology started, was noted in 11 (55%) of the 20 cases with detailed clinical history. The delay occurred more often with SD tumors (9 of 14 SD vs. 2 of 6 IT) and in the setting of NF1 (8 of 15 NF1 vs. 3 of 10 non-NF1). Notably, 7 of the 11 MPNSTs were associated with neurofibroma. During these intervals the patients were thought to suffer from orthopedic problems (arthritis or back strain) or cardiovascular disorders. Other neoplasms noted in the patients’ clinical history, in addition to those that prompted earlier radiation therapy, were multiple melanomas associated with familial melanoma syndrome (Case 10), ovarian mucinous cystadenoma (Case 11), ovarian mucinous cystadenocarcinoma (Case 22), osteoma of the mandible (Case 2), ‘‘neurilemoma’’ of the chest (Case 10), multiple cutaneous basal cell carcinomas (Case 19), and a thyroid adenoma. All but the last patient had NF1, a finding in keeping with the observation of other nonneural tumors occurring in patients with this disorder.19 –20 Clinical Features Sixty percent of the 25 patients had NF1. The overall mean age at presentation was 31.6 years, with NF1 patients presenting at younger ages (mean age, 26.4 years); the mean age did not significantly differ with tumor location (SD vs. IT). Both genders were equally affected overall (M:F 5 1.1:1), but a slight female predominance was noted among patients with NF1 (M: F 5 0.7:1) and a male predominance among those with non-NF1 (M:F 5 2.3:1). Five non-NF1 patients with prior radiation therapy (20%) developed MPNSTs after an interval of 7–10 years; the original tumors were seminoma (Cases 17 and 18), ‘‘reticulum cell Tumor Location Nineteen (73%) of the 26 tumors were SD (Table 1) involving the retroperitoneum, psoas muscle, and/or pelvis, and 7 (27%) were IT. Tumor involvement of a spinal nerve root or lumbosacral plexus was noted in 12 cases. The tumor encroached on the spinal column in 10 cases in the form of vertebral body involvement (Cases 15, 20, 22, and 24), tumor extension along the nerve root up to the vertebral foramen (Cases 6 and 11), epidural encroachment with or without spinal cord compression (Cases 1 and 18), and epidural and intradural disease (Cases 2 and 10). NF1 Age/ gender 17/F 32/F 33/F 38/M 24/M 29/F 33/M 31/F 1 1 1 1 1 1 1 1 4 (1982) 5 (1986) 6 (1989) 7 (1990) 8 (1990) 9 (1994) 10 (1994) 11 (1995) 27/M 68/M 2 2 2 18 (1994)a 26/M 39/M 1 2 2 2 22 (1992)c a 24 (1978)a 25 (1986) 53/F (R)sup. MED, paratracheal (L) MED (L)MED RTP (L) T1 MED (L) post. MED (R)MED, UT Sacral (R)RTP (R)psoas (R)RTP RTP (R)RTP RTP, MES (R)RTP (L)RTP (R)RTP (R)RTP (L) iliac fossa (L)RTP RTP (L) presacral (R)RTP, psoas (L)RTP, lumbosacral (L) pelvis, psoas Location 2 T9 T9 1 T1 2 T5–T8 2 S2,sacral plexus, ganglia 2 Ileo-inguinal, ileo-hypogastric L5, ganglion 2 2 2 L3,L4,L5 2 L3,L4,L5 2 Sacral plexus, sciatic L2,L3,L4 ganglion 1 Sciatic L1, ganglion 2 L3,L4, sacral plexus Nerve 2 2 1 1 2 2 1 1 2 1 2 2 1 1 2 2 1 1 2 1 1 1 2 1 1 2 NFMA .10 .5 4.5 4.5 2.0 5.0 10.5 9 2.5 18 5.2 .10 25 24 14 .16 16 6.5 40 9.0 17.8 4 .10 11.0 .25 14.5 Tumor size (cm) Focal 2 2 2 N/E Prominent Prominent N/E Prominent Prominent 2 Prominent Prominent Prominent Prominent Prominent N/E 2 N/E Prominent Prominent Focal ,5% Prominent Prominent Prominent Prominent Necrosis 50 30 3 9 16 23 17 7 46 3 15 18 39 23 19 N/E 20 59 30 49 2 27 12 16 6 MF/ 10HPF High High High 1 low Low High High 1 low High 1 low High High High High 1 low High High High 1 low High High High High High High High Low High High 1 low High High Grade b STR STR CRb CRb STR STR STR Bx. 1 1 2 2 2 2 2 2 2 2 2 CRb STR CRb 2 STR 1 2 STR 1 1 2 1 STRb STR 2 1 b STR Bx STR Bx CR 2 b STRb 1 1 1 2 2 b Chemo Rx CRb STR STR STR STR Surg Rx 1 2 2 2 2 2 1 1 2 2 2 1 1 2 1 1 2 1 2 2 2 2 1 1 2 1 Rad Rx 2 2 1 (6) 1 (13) 1 (2) 1 (UNK) 2 N/A N/A 1 (3) 1 (6) 1 (UNK) 1 (UNK) 1 (9) 1 (3) 2 1 (12) 1 (9) 1 (2) N/A 2 1 (2) 2 1 (0.5) 2 2 Recur (mos to) 1 (3) 1 (UNK) 1 (21) 2 2 1 (7) 1 (15) 2 2 1 (27) 1 (14) 1 (0.5) 1 (5) 1 (1) 2 1 (0) 1 (10) 1 (0) 1 (26) 1 (0) 2 2 1 (23) 2 2 (5) Mets (mos to) DOD (7) DOD (27) DOD (7) NED (35) DOD (27) DOD (11) DOD (3) AWD (14) NED (61) DOD (40) DOD (4) DOD (14) DOD (26) DOD (7) DOD (10) DOD (4) DOD (22) DOD (10) DOD (6) DOD (2) NED (153) DOD (5) NED (221) DOD (23) DOD (5) Outcome (mos) AWD: alive with disease (at the end of the study); Bx: biopsy only; Chemo Rx: chemotherapy; CR: complete resection; DOD: dead of disease; HPF: high power fields (340); (L): left; MED: mediastinum; MES: mesentery; Mets.: metastasis; MF: mitotic figures; Mos: months; MPNST: malignant peripheral nerve sheath tumor; N/E: not evaluable; N/A: not applicable; NED: no evidence of disease; NF1: type 1 neurofibromatosis; NFMA: neurofibroma; Post.: posterior; Rad Rx: radiation therapy; Recur: recurrence; (R): right; RTP: retroperitoneum; Surg Rx: surgical treatment; STR: subtotal resection (positive gross or microscopic margins); Sup: superior; UNK: unknown; UT: upper thorax. a The patient had a postirradiation tumor. b Gross complete resection. c Patient had 2 MPNSTs that were clinically identified simultaneously. The intrathoracic tumor was resected first. 23 (1970) 41/M 1 19/F 17/M 1 21 (1992) 20/M 20 (1982) Intrathoracic MPNSTs 19 (1976) 1 23/M 47/M 2 2 16 (1990) 17 (1991)a 40/M 33/F 37/F 2 2 14 (1986) 15 (1989) 12 (1972) 13 (1984) a 25/F 26/F 1 1 2 (1977) 3 (1982) Subdiaphragmatic MPNSTs 1 (1970) 1 11/F Case no. TABLE 1 Data on 25 Patients with Paraspinal MPNST 2194 CANCER June 1, 1998 / Volume 82 / Number 11 Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. 2195 TABLE 2 Clinicopathologic Features of Paraspinal MPNSTs No. of patients No. of tumors Age (yrs) Mean Range Gender Female Male Postirradiation tumors Associated nerve Associated neurofibroma Tumor size .5 cm Mean size Divergent differentiation Median length of FU (mos) Range of FU (mos) Recurrence Range (mos) Metastasis Range (mos) Median (mos) DOD 2-year survival 5-year survival 10-year survival NF1 No NF1 Overall 15 16 10 10 25 26 26.4 11–41 39.3 23–68 31.6 11–68 9 (60%) 6 (40%) 0 11 (69%) 11 (69%) 13 (81%) 12.7 cm 2 (12%) 3 (30%) 7 (70%) 5 (50%) 5 (50%) 3 (30%) 8 (80%) 11.5 cm 5 (50%) 12 13 5 (20%) 16 (62%) 14 (54%) 21 (81%) 12.3 cm 7 (27%) 10 2–221 6 (46%) 0.5–6 10 (67%) 0–23 5 12 (80%) 27% 20% 20% 14 4–61 9 (90%) 2–13 7 (70%) 0–27 15 8 (80%) 50% 12% 12% 11 2–221 15 (65%) 0.5–13 17 (68%) 0–27 7 20 (80%) 35% 16% 16% FIGURE 1. The cut surface of an ovoid, 18 3 12 3 7 cm subdiaphragmatic malignant peripheral nerve sheath tumor reveals a fleshy, bulging, tan, focally gelatinous, hemorrhagic, and partly necrotic tumor surrounded by a thin pseudocapsule (Case 16). DOD: dead of disease; FU: follow-up; NF1: neurofibromatosis type I. Biopsy Evaluation Sixteen patients underwent open biopsy, and a diagnosis of sarcoma or MPNST was rendered in 13 cases, atypical neurofibroma in 2, and schwannoma in 1. Three patients had unresectable tumors and received no additional surgical treatment. FIGURE 2. A low grade malignant peripheral nerve sheath tumor, arising focally in a subdiaphragmatic plexiform neurofibroma, is shown. Findings include hypercellularity, fascicular arrangement of cells, nuclear enlargement, and crisp hyperchromasia. A mitotic figure is present (Case 4). Macroscopic Pathologic Findings Tumors measured 2– 40 cm in greatest dimension (mean size, 12.3 cm). SD tumors were usually larger than IT, with mean sizes of 14.3 and 6.6 cm, respectively. NF1 status did not impact on tumor size (Table 2). Most tumors were well circumscribed, ovoid masses, often with a bosselated surface. A pseudocapsule was mentioned in only 4 cases, whereas a few tumors were partly encased by a thin, translucent membrane that likely represented serosa reflected over the tumor. Some tumors were removed piecemeal or during more than one surgical procedure. Cut surfaces were tan-white, gray, yellow-gray, or brownish, at times gelatinous or glistening, with areas of hemorrhage (Fig. 1). Necrosis, noted grossly in 12 tumors, was mostly focal but in a few cases was exten- sive. Rare tumors were partly cystic. Tumor consistency varied from firm and rubbery to fleshy. Histologic Features Twenty-two tumors (85%) involved a nerve and/or originated from a neurofibroma (Tables 1 and 2). Twenty-four tumors were high grade MPNSTs, six with a low grade component, whereas the remaining two tumors were exclusively low grade MPNSTs and arose within a plexiform neurofibroma. The first entirely low grade tumor (Case 4) consisted mostly of plexiform neurofibroma that completely encompassed two microscopic malignant foci showing hypercellularity, fascicular arrangement of the tumor cells, nuclear en- 2196 CANCER June 1, 1998 / Volume 82 / Number 11 FIGURE 3. The most common histologic pattern in malignant peripheral nerve sheath tumors is crowded, hyperchromatic, mitotically active spindle cells arranged in short and long fascicles (Case 24). Note the difference between the cellularity of this high grade tumor and that of the low grade tumor shown in Figure 2. FIGURE 4. A distinctive histologic pattern in malignant peripheral nerve sheath tumors, sometimes referred to as a ‘‘tapestry’’ pattern, is created by alternating loosely and densely cellular tumor areas (Case 8). FIGURE 6. A spindle cell tumor of T-1 nerve root has the features of a malignant peripheral nerve sheath tumor with rhabdomyosarcomatous differentiation (malignant triton tumor) (Case 23). FIGURE 5. Cells in most malignant peripheral nerve sheath tumors are spindled, with elongated hyperchromatic nuclei, a moderate amount of faintly eosinophilic cytoplasm, and indistinct cell membranes. largement, and hyperchromasia (Fig. 2). The malignant component of the second entirely low grade MPNST involved slightly less than half of the preexisting neurofibroma. The low grade component noted in six of the high grade tumors consisted of cells with enlarged, hyperchromatic nuclei embedded in a moderately fibrotic background, which therefore did not reach the cellularity of the high grade component. In one of these tumors most of the malignant component was low grade (Case 2), whereas in another it represented approximately 40% of the examined tumor (Case 22) and in the remaining cases low grade areas represented 5–10% of the examined tumor. The histologic appearance of the MPNSTs was similar to previous descriptions.8,9 The most frequent pattern resembled fibrosarcoma (Fig. 3). A second distinctive pattern, noted in 11 tumors, consisted of alternating hypocellular and hypercellular tumor fascicles (Fig. 4). Storiform and diffuse growths were less frequent. Tumor cells were spindled with usually hyperchromatic, curved, and tapered nuclei and, less frequently, elongated, blunt-ended nuclei with vesicular chromatin (Fig. 5). Nuclear pleomorphism was uncommon. Mitotic activity exceeded 4 mitoses per 10 HPF in 21 of 24 tumors. Necrosis, often geographic, was present in most cases. Vessels were thin-walled and sinusoidal but on occasion thick-walled and hyalinized. Divergent differentiation,21–22 in the forms of 3 MPNSTs with rhabdomyosarcomatous differentiation (malignant triton tumors)23–27 (Fig. 6) and 4 glandular MPNSTs,28 –32 was more frequent among non-NF1 Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. 2197 FIGURE 7. In this glandular malignant peripheral nerve sheath tumor, the glandular epithelium is histologically malignant (Case 25). FIGURE 8. A cases (Tables 1 and 2). In one case, the glandular epithelium was histologically malignant (Fig. 7), an unusual finding. Immunohistochemical studies showed reactivity for S-100 protein in 10 of 13 tumors (4 of 6 NF1 and 6 of 7 non-NF1 cases) as scattered cells in a predominantly negative tumor. Focal, weak positivity for HHF-35 was noted in 3 of 12 tumors and for SMA in 1 of 7. S-100 protein reactivity was essential in establishing the diagnosis of MPNST in two cases. Rare cells in one of nine MPNSTs were reactive for desmin. None of the malignant triton tumors were evaluated immunohistochemically. Keratins were positive only in the epithelial elements of three glandular MPNSTs (tumors that were described in detail previously)32 and negative in another two cases. received only radiation therapy. The chemotherapeutic regimens varied in terms of both the agents and the dosages used, but they included doxorubicin in seven cases. Radiation was delivered by external beam or iodine-125 implants, in dosages varying from 2000 to 5940 centigray. In 3 cases, both the radiation dosage and the chemotherapy regimen were unknown. Due to the wide variation in the additional treatment modalities, these were not evaluated by statistical analysis. Therapy The primary mode of treatment was surgical resection for 23 tumors (88%) (Table 1). Complete resection was achieved for 6 tumors from 5 patients. Seventeen tumors were subtotally resected; among these, 5 were considered gross complete resections but had microscopic involvement of surgical margins. In the remaining tumors, stated reasons for subtotal resection were 1) involvement of the lumbar plexus or spine (in 7 cases), 2) vascular encasement (in 1 case), and 3) presence of metastatic disease (in 3 cases). Three patients (12%) with unresectable tumors underwent open biopsy only; 2 received palliative chemotherapy or radiation therapy, and the third died soon after diagnosis before administration of any treatment. Additional therapy was given to 14 patients: 3 patients received only chemotherapy, 7 patients had chemotherapy and radiation therapy, and 4 patients rhabdomyosarcomatous component of the tumor in Figure 6 (Case 23) encases dorsal nerve roots and invades cord parenchyma. Outcome Follow-up information was available for all patients (Tables 1 and 2). The mean and median length of follow-up were 30 and 11 months, respectively. Local recurrence was experienced by 65% of patients within a median interval of 5 months. Tumors recurred in 46% of NF1 patients, 90% of non-NF1 patients, 65% of patients with tumors at SD sites, and 67% with tumors at IT sites. Metastasis occurred in 68% of the cases overall, after a median interval of 7 months; 63% of SD tumors and 71% of IT tumors metastasized. Lungs, brain/ spinal cord/meninges, and abdomen/peritoneum were involved in 4 cases each; liver in 3 cases; pleura, bone, and adrenals in 2 cases each; and bladder, chest wall, soft tissue, spleen, and lymph nodes in 1 case each. Extensive subarachnoid spread and spinal parenchymal involvement by embryonal rhabdomyosarcoma was found at autopsy in a patient with a malignant triton tumor (Fig. 8). Central nervous system metastases occurred by subarachnoid spread in 2 patients and by systemic spread in 2 others (who also had metastases to sites such as lungs, liver, and bone). Four patients (16%) (Cases 2, 4, 16, and 22) with five tumors were alive at last follow-up with no evi- 2198 CANCER June 1, 1998 / Volume 82 / Number 11 FIGURE 9. Overall survival is shown. dence of disease (NED) at 221, 153, 61, and 35 months, respectively, after diagnosis. All five tumors arose in neurofibromas, four in the setting of NF1. Four of the tumors were completely resected (in Cases 4 and 16, and both tumors in Case 22). One tumor (Case 2), although thought on macroscopic examination to be completely resected, had focally positive microscopic margins. This patient received preoperative and postoperative chemotherapy and radiation therapy. The sizes of the tumors were 4 –18 cm, and the mitotic rate was 2– 46 mitotic figures per 10 HPF. Two tumors were entirely low grade, two were partially low grade, and one was entirely high grade. Two of the four surviving patients with NED had spinal involvement by the tumor. The first patient (Case 2) had a large retroperitoneal tumor with epidural and intradural extension that created a complete myelographic block at the T12 level. The tumor was removed in two stages. First, the intradural spinal component was resected piecemeal via a laminectomy and was found to be a neurofibroma. The second procedure was an en bloc resection of the retroperitoneal portion of the tumor that involved the psoas muscle; this was a high grade MPNST contained within a neurofibroma. The second patient (Case 22) had a thoracic mass extending along the T9 nerve root through the T9 –10 neural foramen. Removal necessitated sacrifice of the nerve root with intradural exploration. Although part of the tumor was high grade, the portion involving the T9 nerve root was a plexiform neurofibroma. Twenty patients (80%) died of disease (DOD) within 2– 40 months (median survival, 8.5 months). One patient was alive with recurrent and metastatic disease after 14 months at the end of the study period. Statistical Analysis The overall 2- and 5-year survival rates were 35% and 16%, respectively (Table 2, Fig. 9). When survival status was used as an endpoint, complete resection or gross complete resection was highly associated with better survival (P , 0.01 for both). When survival time was used as an endpoint, tumor size .5 cm (P 5 0.04, Fig. 10) and presence of a high grade tumor (P , 0.015, Fig. 11) were associated with shorter survival, whereas complete resection (Fig. 12) and gross complete resection were associated with longer survival (P 5 0.02 and P 5 0.06, respectively). Parameters that were not shown to have an effect on survival were age, gender, history of NF1, diagnostic delay, tumor location (SD vs. IT), mitotic rate, necrosis, divergent differentiation, and postirradiation tumor occurrence. Correlation was noted between high tumor grade and 1) less likelihood for gross complete resection (34% vs. 72% with low grade tumors), 2) tumor size .5 cm (83% vs. 57% with low grade tumors), and 3) more frequent use of Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. 2199 FIGURE 10. Survival is shown by tumor size. radiation therapy (84% vs. 61% with low grade tumors). DISCUSSION MPNSTs account for 5% of malignant soft tissue tumors;13 they have an estimated incidence of 0.001% in the general population6 and 2% among patients with NF1.20 Mention of MPNSTs at paraspinal location was made in some of our earlier reports24,28,32,33 and in the reports of others.4,8,10,11,12,21,26,34 – 44 Two series reported on a combined total of 11 patients with intraspinal MPNSTs.45,46 Although previous studies noted that patients with ‘‘centrally’’ located tumors had a worse outcome, these were limited by 1) lack of detailed data, which prevented critical review of the cases, especially in cases of remarkably long survival;11 b) small numbers of tumors;10 and 3) inclusion of NF1 cases only.8 The current study focuses on the paraspinal SD and IT locations to assess specific clinical and histologic parameters and their relation to tumor behavior. In addition, paraspinal MPNSTs are compared with a large series of MPNSTs of all anatomic sites6 and tumors of the BLE.9 Patients with paraspinal MPNSTs present more frequently with pain than those with BLE tumors, who usually present with an enlarging mass and neurologic symptoms. Delay in making the diagnosis of paraspi- nal MPNST after the onset of symptoms was noted in 55% of the patients; this was more frequent among NF1 patients and often associated with the presence of neurofibroma. It is likely that the symptomatology was initiated either by the neurofibroma or by more rapid enlargement of the tumor when malignant transformation occurred within the neurofibroma. The absence of a worse outcome among these cases, although paradoxical, might be accounted for in part by ‘‘containment’’ of the malignant component within the neurofibroma for some time before the diagnosis, and in instances possibly providing a ‘‘safe margin’’ during surgery. Because diagnostic delay was more common in the setting of NF1, we suggest careful investigation of ‘‘pain’’ in such cases to identify deepseated neurofibromas undergoing malignant transformation. Biopsies of paraspinal MPNSTs present a tremendous interpretative challenge even for pathologists who have considerable experience with soft tissue tumors. Of major concern is distinction from benign tumors occurring in this region; those that most closely simulate MPNST are cellular schwannoma46 and cellular or atypical neurofibroma. Cellular schwannoma46 –51 mimics a malignant tumor with its increased cellularity and mitotic activity, but it can be differentiated by its generally bland nuclei, perivascu- 2200 CANCER June 1, 1998 / Volume 82 / Number 11 FIGURE 11. The survival of patients with entirely high grade tumors is compared with the survival of those with tumors that are partly or entirely low grade. lar cellular whorls, and diffuse, strong S-100 protein expression (in contrast to the S-100 reactivity of only scattered tumor cells in MPNST). In addition, the presence of thick-walled, hyalinized blood vessels and clusters of lipid-laden histiocytes, although infrequently noted in cases of MPNST, favors the diagnosis of cellular schwannoma. Cellular neurofibroma should not display diffuse nuclear enlargement or hyperchromasia. On the other hand, atypical neurofibromas (neurofibromas characterized by scattered hyperchromatic, pleomorphic, and sometimes multinucleated tumor cells)52 lack diffuse hypercellularity. We regard neurofibromas with both diffusely increased cellularity and extensive nuclear enlargement and hyperchromasia as low grade MPNSTs. The latter, because of their potential association with a high grade component, should be treated with complete resection and microscopic evaluation of the surgical margins. Biopsies of three tumors in our study were erroneously interpreted as benign. In two instances, the interpretation was atypical neurofibroma; in the third, it was schwannoma. In one tumor designated atypical neurofibroma, undue emphasis had been placed on the absence of mitotic figures (Fig. 13), and in the other the significance of both hypercellularity and widespread nuclear enlargement had not been appreciated. The diagnosis of schwannoma was rendered for a spindle cell tumor of a nerve that lacked marked cellularity, nuclear atypia, or mitotic activity. On re- view, a small collection of rhabdomyoblasts was noted. Although a diagnosis of MPNST would have have been difficult to make based on only the appearance of the spindle cell component, the identification of rhabdomyoblasts pointed to a diagnosis of MPNST with focal rhabdomyoblastic differentiation (malignant triton tumor) (Fig. 6). Regarding the importance of mitotic figures, we would like to emphasize that they may be found in cellular schwannomas and may be absent in a biopsy of MPNST (even high grade), and in isolation they do not indicate that a neurofibroma has undergone malignant transformation. Because of the above difficulties, accurate diagnosis is best achieved by study of a generous sample that includes tissue well beneath any apparent ‘‘capsule.’’ Distinction of MPNSTs from other malignant soft tissue tumors has been discussed in previous publications.15,31 Accurate tumor typing is important because of its implications for planning surgical treatment. The goal of treatment is surgical en bloc resection of the tumor and the achievement of negative microscopic margins. Complete resection of a tumor that extends to the lumbar plexus or spine may not be possible without causing neurologic damage; for this reason, incomplete resection was performed in some cases in this study. Discussion with the patient regarding the potential risks and use of a multidisciplinary surgical team approach may therefore be indicated in selected cases. The latter point is emphasized by one of our cases (Case 2), in Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. 2201 FIGURE 12. Survival is shown by type of resection. which two separate operations were necessary to resect the spinal and retroperitoneal tumor components. Currently, at our institution, postoperative chemotherapy in the form of MAID (mesna, doxorubicin [Adriamycin; Adria Laboratories, Columbus, OH], ifosfamide, and dacarbazine) is given to patients with high grade tumors. Chemotherapy and radiation therapy are often given to patients with incompletely resected tumors. Gross examination is inadequate for evaluating completeness of resection because of the deceptively circumscribed, pseudoencapsulated appearance of MPNSTs and their potential to extend proximally and distally along involved nerves. All macroscopic specimens should be inked and the surgical margins adequately sampled. Information relative to the completeness of tumor resection can be obtained by frozen-section examination of the margins of involved nerves. Admittedly, distinction of a low grade MPNST from either a cellular schwannoma or a neurofibroma on frozen section may not be possible, but recognition of a high grade, malignant spindle cell tumor is. A neurofibromatous component can mimic MPNST grossly by encasing blood vessels, eroding vertebrae, growing into spinal foramina, or extending into the extradural and intradural space. In such cases, although complete resection is desirable, identification of neurofibroma at the surgical margins is not an ominous finding. FIGURE 13. This biopsy of a high grade malignant peripheral nerve sheath tumor was originally misinterpreted as a neurofibroma due to the absence in the biopsy of mitotic figures, a lack of appreciation of diffuse hyperchromasia, and increased cellularity. 2202 CANCER June 1, 1998 / Volume 82 / Number 11 TABLE 3 Comparison of MPNST at Different Sites MPNST 2-yr survival 5-yr survival 10-yr survival DOD Median survival (mos) Recurrence Metastasis All sites (6) BLE (9) Paraspinal 57% 35% 34% 39% 16% 23% 16% 68% 63% 80% 35 8.5 42% 40% 65% 68% 63% 68% MPNST: malignant peripheral nerve sheath tumor; BLE 5 MPNST of buttock/lower extremity; DOD 5 dead of disease. The prognosis of paraspinal MPNSTs is favorably affected by small size (,5 cm) and complete resection, findings similar to those in previous studies.6,10 Furthermore, low tumor grade, even as a minor component, is associated with better outcome. In contrast to prior studies,4,5,6,7,12 outcome was not affected by the presence of NF1, diagnostic delay, tumor location (SD vs. IT), postirradiation occurrence, divergent differentiation, mitotic rate, or necrosis. Study of additional cases would be useful to confirm these findings. The outcomes of patients with paraspinal MPNSTs is worse than those of patients with MPNSTs from all anatomic sites and BLE (Table 3). Recurrence rate and mortality are higher, and median, 2-, 5-, and 10-year survivals are shorter. Unlike the predominance of metastases to the lung in BLE tumors, paraspinal MPNSTs metastasize at similar rates to the lung, central nervous system, liver, and abdomen. Neoplastic seeding of the subarachnoid space and even spinal cord involvement may occur by way of proximal spread along spinal nerves. Because tumor size and metastatic rates in the paraspinal and BLE groups are similar, we do not agree with Guccion and Enzinger,8 who attribute the more aggressive behavior of central tumors mainly to larger size and greater likelihood of metastasis. According to our findings, the worse outcomes are due to incomplete tumor resection, which result in more frequent and earlier recurrences. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. REFERENCES 1. 2. 3. 4. Woodruff JM. Pathology of the major peripheral nerve sheath neoplasms. In: Weiss SW, Brooks JSJ, editors. Soft tissue tumors (International Academy of Pathology monograph). Baltimore: Williams and Wilkins, 1996:129 – 61. Stout AP. Tumors of the peripheral nervous system. In: Atlas of tumor pathology. Section II, Fascicle 6. Washington, DC: Armed Forces Institute of Pathology, 1949. Hirose T, Sumitomo M, Kudo E, Hasegawa T, Teramae Y, Murase M, et al. Malignant peripheral nerve sheath tumor (MPNST) showing perineurial cell differentiation. Am J Surg Pathol 1989;13:613–20. Tsuneyoshi M, Enjoji M. Primary malignant peripheral 17. 18. 19. 20. nerve tumor (malignant schwannomas). Acta Pathol Jpn 1979;29:363–75. Nambisam RN, Rao U, Moore R, Karakousis CP. Malignant soft tissue tumors of nerve sheath origin. J Surg Oncol 1984; 25:268 –72. Ducatman BS, Scheithauer BW, Piepgras DG, Reiman HM, Ilstrup DM. Malignant peripheral nerve sheath tumors: a clinicopathologic study of 120 cases. Cancer 1986;57:2006 – 21. Meis JM, Enzinger FM, Martz KL, Neal JA. Malignant peripheral nerve sheath tumors (malignant schwannomas) in children. Am J Surg Pathol 1992;16:694 –707. Guccion JG, Enzinger FM. Malignant schwannoma associated with von Recklinghausen’s neurofibromatosis. Virchows Arch Pathol A 1979;383:43–57. Hruban RH, Shiu MH, Senie RT, Woodruff JM. Malignant peripheral nerve sheath tumors of the buttock and lower extremity: a study of 43 cases. Cancer 1990;66:1253– 65. Wanebo JE, Malik JM, Vandenberg S, Wanebo HJ, Driesen N, Persing JA. Malignant peripheral nerve sheath tumor: a clinicopathologic study of 28 cases. Cancer 1993;71:1247– 53. Gosh BC, Gosh L, Huvos AG, Fortner AG. Malignant schwannoma: a clinicopathologic study. Cancer 1973;31:184 –90. Sordillo PP, Helson L, Hajdu SI, Magill GB, Kosloff C, Golbey RB, et al. Malignant schwannoma: clinical characteristics, survival, and response to therapy. Cancer 1981;47:2503–9. Lewis JJ, Brennan MF. Soft tissue sarcomas. Curr Probl Surg 1996;33:817– 80. Weiss SW, Langloss JM, Enzinger FM. Value of S-100 protein in the diagnosis of soft tissue tumors with particular reference to benign and malignant Schwann cell tumors. Lab Invest 1983;49:299 –308. Wick MR, Swanson PE, Scheithauer BW, Manivel JC. Malignant peripheral nerve sheath tumor: an immunohistochemical study of 62 cases. Am J Clin Pathol 1987;87:425–33. Erlandson RA, Woodruff JM. Peripheral nerve sheath tumors: an electron microscopic study of 43 cases. Cancer 1982;49:273– 87. Stout AP. The malignant tumors of the peripheral nerves. Am J Cancer 1935;25:1–36. Cartun RW, Pederson CA. An immunohistochemical technique offering increased sensitivity and lower cost with streptavidin-horse raddish peroxidase conjugate. J Histotechnol 1989;12:273–7. Riccardi VM. Von Recklinghausen neurofibromatosis. N Engl J Med 1981;305:1617–26. Sorenson SA, Mulvihill JJ, Nielson A. Long-term follow-up of von Recklinghausen neurofibromatosis. N Engl J Med 1986; 314:1010 –5. Paraspinal Malignant Peripheral Nerve Sheath Tumors/Kourea et al. 21. Ducatman BS, Scheithauer BW. Malignant peripheral nerve sheath tumors with divergent differentiation. Cancer 1984; 54:1049 –57. 22. Brooks JJ. Malignant schwannomas with divergent differentiation including ‘‘triton’’ tumor. In: Wiliam CJ, Krikouan JG, Green MR, Raghavan D, editors. Textbook of uncommon cancer. New York: John Wiley & Sons, Inc., 1988:653– 68. 23. Masson P, Martin J. Rhabdomyosomes des nerfs. Bull Cancer (Paris) 1938;27:751– 67. 24. Woodruff JM, Chernik N, Smith M, Millet W, Foote F. Peripheral nerve tumors with rhabdomyosarcomatous differentiation (malignant ‘‘triton’’ tumors). Cancer 1973;32:426 – 39. 25. Daimaru Y, Hashimoto H, Enjoji M. Malignant ‘‘triton’’ tumors: a clinicopathologic and immunohistochemical study of nine cases. Hum Pathol 1984;15:768 –78. 26. Brooks JJ, Freeman M, Enterline HT. Malignant triton tumors: natural history and immunohistochemistry of nine new cases with literature review. Cancer 1985;55:2543–9. 27. Dewit L, Albus-Lutter CE, De Jong ASH, Voute PA. Malignant schwannoma with a rhabdomyoblastic component, a socalled triton tumor: a clinicopathologic study. Cancer 1986; 58:1350 – 6. 28. Woodruff JM. Peripheral nerve tumors showing glandular differentiation (glandular schwannomas). Cancer 1976;37: 2399 – 413. 29. Warner TFCS, Louie R. Malignant nerve sheath tumor containing endocrine cells. Am J Surg Pathol 1983;7:583–90. 30. Christensen WN, Strong EW, Bains MS, Woodruff JM. Neuroendocrine differentiation in the glandular peripheral nerve sheath tumor: pathologic distinction from the biphasic synovial sarcoma with glands. Am J Surg Pathol 1988;12: 417–26. 31. Wong SY, Teh M, Tan YO, Best PV. Malignant glandular triton tumor. Cancer 1991;67:1076 – 83. 32. Woodruff JM, Christensen WN. Glandular peripheral nerve sheath tumors. Cancer 1993;72:3618 –28. 33. Foley KM, Woodruff JM, Ellis FT, Posner JB. Radiationinduced malignant and atypical PNST. Ann Neurol 1980;7: 311– 8. 34. Ackerman LV, Taylor FH. Neurogenous tumors within the thorax: a clinicopathological evaluation of forty-eight cases. Cancer 1951;4:669 –91. 35. White HR. Survival in malignant schwannoma: an 18-year study. Cancer 1971;27:720 –9. 36. D’Agostino AN, Soule EH, Miller RH. Primary malignant neoplasms of nerves (malignant neurilemomas) in patients without manifestations of multiple neurofibromatosis (von Recklinghausen’s disease). Cancer 1963;16:1003–14. 37. D’Agostino AN, Soule EH, Miller RH. Sarcomas of the peripheral nerves and somatic soft tissues associated with 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 2203 multiple neurofibromatosis (von Recklinghausen’s disease). Cancer 1963;16:1015–27. Trojanowski JQ, Kleiman GM, Proppe KH. Malignant tumors of nerve sheath origin. Cancer 1980;46:1202–12. Taxy JB, Battifora H, Trujillo Y, Dorfman HD. Electron microscopy in the diagnosis of malignant schwannoma. Cancer 1981;48:1381–91. Ducatman BS, Scheithauer BW. Post-irradiation neurofibrosarcoma. Cancer 1983;51:1028 –33. Chitale AR, Dickersin R. Electron microscopy in the diagnosis of malignant schwannomas: a report of six cases. Cancer 1983;51:1448 – 61. Herrera GA, Pinto de Moraes H. Neurogenic sarcomas in patients with neurofibromatosis (von Recklinghausen’s disease). Virchow Arch Pathol A 1984;403:361–76. Daimaru Y, Hashimoto H, Enjoji M. Malignant peripheral nerve sheath tumors (malignant schwannomas): an immunohistochemical study of 29 cases. Am J Surg Pathol 1985; 9:434 – 44. Matsunou H, Shimoda T, Kakimoto S, Yamashita H, Ishikawa E, Mukai M. Histopathologic and immunohistochemical study of malignant tumors of peripheral nerve sheath (malignant schwannoma). Cancer 1985;56:2269 –79. Valdueza JM, Hagel C, Westphal M, Hänsel M, Herrmann H-D. Primary spinal malignant schwannoma: clinical, histological and cytogenetic findings. Neurosurg Rev 1991;14: 283–91. Seppälä MT, Haltia MJJ. Spinal malignant nerve sheath tumor or cellular schwannoma? A striking difference in prognosis. J Neurosurg 1993;79:528 –32. Woodruff JM, Goodwin TA, Erlandson RA, Susin M, Martini N. Cellular schwannoma: a variety of schwannoma sometimes mistaken for a malignant tumor. Am J Surg Pathol 1981;5:733– 44. Fletcher CDM, Davies SE, McKee PH. Cellular schwannoma: a distinct pseudosarcomatous entity. Histopathology 1987; 11:21–35. Lodding P, Kindblom L-G, Angervall L, Stenman G. Cellular schwannoma: a clinicopathologic study of 29 cases. Virchows Arch Pathol A 1990;416:237– 48. White W, Shiu MH, Rosenblum MK, Erlandson RA, Woodruff JM. Cellular schwannoma: a clinicopathologic study of 57 patients and 58 tumors. Cancer 1990;66:1266 –75. Casadei GP, Scheithauer BW, Hirose T, Manfrini M, Van Houton C, Wood MB. Cellular schwannoma: a clinicopathologic, DNA flow cytometric and proliferation marker study of 71 cases. Cancer 1990;75:1109 –19. Woodruff JM, Horten BC, Erlandson RA. Pathology of peripheral nerves and paragangliomas. In: Silverberg SS, editor. Principles and practice of surgical pathology. New York: Churchill Livingstone, 1990:1503–20.