Histopathology of benign versus malignant sympathoadrenal paragangliomas: Clinicopathologic study of 120 cases including unusual histologic features

Histopathology of Benign Versus Malignant Sympathoadrenal Paragangliomas: Clinicopathologic Study of 120 Cases Including Unusual Histologic Features R, ILONA LINNOILA. MD, HARRY R. KEISER, MD, SETH M, STEINBERG, PHD, AND ERNEST E. LACK, MD The clinical and pathologic features of 120 adrenal and extraadrenal paragangliomas were studied in an attempt to identify features which might predict malignant behavior. Clinical follow-up was obtained in 98 cases (82%); 64 tumors were clinically benign, and 34 were malignant as evidenced by regional or distant metastases and/or extensive local invasion. Thirty-two of the 34 malignant tumors (94%) were functionally active. Features noted more frequently in malignant tumors included male predominance (74%; P, [two-sided P value] = .002), extraadrenal location (52%; P, < .OOOl), greater tumor weight (mean 383 g versus 73 g for nonmalignant tumors), confluent tumor necrosis, and the presence of vascular invasion and/or extensive local invasion. Intracytoplasmic hyaline globules were seen in 59% and 32% of benign and malignant tumors, respectively (P, = .OOl). Logistic regression analysis of 16 nonhistologic and histologic parameters showed four of them to be most predictive of malignancyextraadrenal location, coarse nodularity of the primary tumor, confluent tumor necrosis, and absence of hyaline globules. Most malignant paragangliomas had two or three of these features (71%), while 89% of benign tumors had only one (or none: P < .OOOl). According to the statistical model developed, there was better than a 95% probability that more than 70% of tumors could be classified correctly on the basis of the four factors indicated. Although limitations still exist, results of this study provide some basis for evaluating malignant potential of these tumors. HUM PATHOL 21:1168-1180. 0 1990 by W.B. Saunders Company.

Sympathoadrenal paragangliomas are rare tumors in general clinical practice which usually arise in the adrenal glands (ie, pheochromocytomas), but may occur in extraadrenal sites anywhere from the neck tb

the pelvic floor, locations which parallel the distribution of the sympathetic nervous system. In general, these tumors usually require prompt and complete surgical resection because of potentially lethal com-

plications related to excess catecholamine secretion, and also because these tumors have a real (but largely unpredictable) potential for malignant behavior. It has been generally accepted that there are no reliable gross or histolc/gic features which can distinguish benign from malignant tumors, and furthermore, some studies give the unmistakable impression that there are virtually no detectable morphologic differences at all. As a corollary to this, some investigators have proposed that the only absolute criterion of malignancy is documentation of tumor in sites where chromaffin tissue is not normally found.’ A few recent studies (which included head and neck paragangliomas) suggest that some morphologic features are more frequently seen in clinically malignant tumors, but the number of cases in each study is too small for appropriate statistical analysis”-“; moreover, malignant tumors have been recently associated with decreased capacity to produce neuropeptides.“,7 The current study uses logistic regression analysis of 16 nonhistologic and histologic parameters to identify factors which may be useful in predicting biologic behavior. This study is notable for the relatively large number of clinically malignant paragangliomas (n = 34) which allows for statistical evaluation. Emphasis is also placed on the incidence of unusual morphologic features including tumors designated as composite pheochromocytomas. MATERIALS AND METHODS

From the National Cancer Institute-Navy Medical Oncology Bethesda, MD; the HypertensionBranch, Naval Hospital, Endocrine Branch, National Heart, Lung, and Blood Institute, and the Biostatistics and Data Management Section, National Cancer Institute, National Institutes of Health, Bethesda, MD; and the Department of. Surgical Pathology. Georgetown University School of Medicine, Washington, DC. Accepted for publication February 8, 1990. Dr Lack was formerly with the Surgical Pathology and Postmortem Section, National Cancer Institute, National Institutes of Health, Bethesda, MD. Kq\ word\: paraganglioma, pheochromocytoma, extraadrenal, malignant, histology. Address correspondence and reprint requests to K. Ilona Linnoila, MD, National Cancer Institute. Navy Medical Oncology Branch, Naval Hospital. Bldg 8, Koom 5 IO 1, Bethesda, MD 208 1-l. 0 I990 by W.B. Saunders Company.

0046~81771901211 l-0013$5.U0/0

1168

Records of cases of‘ sympathoadrenal paragangliomas (n = 120) were retrieved from the surgical pathology files of the National Cancer Institute (n = 109) and the Bethesda Naval Hospital (n = 11) for the %-year period dating from 1955 through 1985. An average of eight hematoxylin-eosin stained sections was available for review (range: one to 28 sections). Only slides of the primary tumor were evaluated, and review was done without prior knowledge of clinical outcome. Gross characteristics of the primary tumor were gleaned from surgical pathology reports and/or review of specimen photographs. Follow-up information was obtained in 98 of the 120 cases (X2%.) by review of medical records and/or personal communication with the patient’s primary physician. All tumors with clinical follow-up were classified as benign or malignant based upon the presence of regional or distant metastases and/or extensive local invasion. The presence of metastases was documented by biopsy and/or serial catecholamine levels. Logistic regression analysis was used to identify a prediction relationship between 16 nonhistologic and histo-

PATHOLOGYOF MALIGNANTPARAGANGLIOMAS (Llnnoila et al) aceous material resembling colloid. The term “phrochromoototna” is used herein for paragangliomas of adrenal medillat-? origin. with extraadrenal tumors being designated according to anatomic site of origin. Comparisons were made using Fisher’s exact test. with ;III P values being twosided (P.,). Log&tic regression analysis wx perfi)rmed with PKO(: LOGIST of S,4S (MS Institute Inc. Carev. NC) to identify a prediction relationship between the clil,icopathologic variables and biologic behavior of the tumors.‘~” T‘wo thirds of the data were randomly selected (65 of 9X observations) to develop a model (training set), and the remaining were used for validation of the results (test set). Evaluation of the predictive ability was determined by the percentage of benign and malignant tumors that were correctlv identified bv the classification rule developed. !I

II) 1I

1L’ I3 I4 1.!I Iti

RESULTS Biographic Data, Anatomic Distribution, and Gross Characteristics of Tumors (SW Table 2 f’or details.) ‘I‘his study showed equal sex distribution (60 males, 60 females). Eleven of‘ 1UX patients with known racial background were black (10%). Clinical follow-up was possible for 08

logic paranletcrs” ( I able 1) and the biologic beh;r\ iol- of’the I he weighr or sk of the tumors f’or this analysis g weight or G4 cni diameter. 2 = \v;,s swId: 1 = C30 W- 101) g \reight or -1. I-7.9 cm diameter, ?I = > IO0 g weight or ZCY.Ocnl cliameter. Based upon gross rnorpholo~y. two parameters wt’w evaluated-solid versus cystic- appearanc-r on cross src.tion and tumor IKISS which was unicentric- ver\us (-oawd~ ~~odularimul~int,dular. Histologic parameters Illc-luded rhe tollowing: predominant architec-tural pattern t toul- major- ~ategories~. tumor necrosis (ir, extent, conflut’nt versus foe-al). mitoric rate (number per 30 high-powerlields [[II’E‘] using a ,*0X objective and IOX wide-field ocv ktlars). nut lear- h\~i~ercliromasiaipleomorphis~~l (on a scale or seof. 0 [abs~nr]. IO ‘1 + , !! + 1 and Y + [mild. moderate iere]). vasc~lila~invasion and/or extensive local invasion, artd the ptwenw of’other features such as hvaline globules (in tiernatou~liri-rosin stained preparations ;,I, ;I scale of’ 0 [Asent] to 1 t , 2 $- . mcl 3 + [rare, occasional. &undant]), ( vtolo,qic Icat uwr resembling ganglion cells. and protein-

TABLE 3.

lunwrs.

1169

Clinical and Laboratory Characteristics of Extraadrenal Paragangliomas

HUMANPATHOLOGY

Volume 21, No. 11 [November1990)

‘Table 4. Of all the malignant cases, 30 patients (88%‘) developed distant metastases, two (6%) had distant metastases as well as extensive local invasion, and two (6%)) had extensive local invasion only. Malignant tumors were also larger (mean weight 383 g versus 73 g, Table 2); these tumors were more often characterized as having cystic change on gross examination (74%‘), but this leature was also noted in some of the clinically benign tumors (35%,). A significant gross morphologic feature of the malignant tumors was coarse nodularity or multiple nodules which were noted on external examination or more commonly on inspection of the tumor in cross section. This coarse nodularity was different from the often symmetrical multicentric pheochromocytomas (exclusively adrenal in this series) occurring in the setting of multiple endocrine neoplasia (MEN) syndrome, type 2. Clinically benign pheochromocytomas in a sporadic, nonfamilial setting were characteristically unilateral and solitary; grossly these tumors appeared as unicentric spheroid masses which were often grayish or duskyred on cross section (Fig 1). Some tumors (both benign and malignant) had a mottled appearance, or geographic areas of congestion or hemorrhage which imparted a beefy-red appearance.

patients (82%‘). Paragangliomas in 64 patients were clinically benign (65’S) and 34 were malignant (35%). The length of clinical follow-up ranged from 1 to 24 years for patients with benign paragangliomas (mean 7.9 years) and up to 4 1 years for those with malignant tumors (mean 8.7 years). The mean age of all patients with follow-up was 38.2 years (range: 6 to 74 years); patients with malignant tumors were diagnosed at an average age of 35.5 years, compared with 39.5 years for those with benign tumors. Malignant paragangliomas were significantly more common in men, with the male to female ratlo being 2.8 compared with 0.64 in the clinically benign group (F, = .002 by Fisher’s exact test). The clinical presentation was characterized b) catecholamine overproduction. The most common symptoms were hypertension, labile at times, sweating, headache, and tachycardia. In two out of 34 clinically malignant cases (6%‘) the tumors appeared to be causing symptoms relating to tumor mass only, while the catecholamines were minimally elevated. One of the ‘silent’ primary tumors was adrenal and the other extraadrenal by location (Table 3). With regard to both clinically benign and malignant adrenal tumors, there was equal predilection for either adrenal gland. Four of 39 sporadic benign pheochromocytomas ( 10%) were bilateral. Clinically malignant paragangliomas were more often extraadrenal in location (52% compared with 6%’ of benign tumors; I’, < .OOOl). More detailed information regarding anatomic distribution of extraadrenal tumors is shown in

TABLE 4. (Iase No.

Chromaftin status

Multiple Endocrine Neoplasia Syndrome Type 2 and Familial Pheochromocytomas (See Table 5 for details.) Pheochromocytomas in the clinical setting of YvIEN syndrome type 2 were

Pathologic Features and Clinical Outcome of Extraadrenal Paragangliomas Primarv

Neuropeptides*

Site

Stetastatic

I 2 3 -l .: ti 7

+ + ND h’D ND ND ND

o/to 4110 f/10 ND 4110 o/to 51 to

L’rinarb bfaddetKetrop&itoneum Hilum of kidnq Ilrinat-y bladder Urinal-,, bladder Ketroperitoneum Ketroperitonelclll

8 9

ND +

ND 6110

Ketroperitoneum Porta hepatis

10 I1 12 1?I 14

+ ND _

l/l0 ND oi to 2110 S/l0

Neck Organ of Zuckerkandf Ketroperitoneum NOS Organ of Zuckerkandf Organ of Zuckerkandf

NOS

Sites

Pelvic nodes Liver. lung, spine Neck Spine

h’OS NOS

Spine Lung. Spine,

neck pelvic bone\.

(:finical Outcome hlaficrnant Mat&ant Malignant Malignant Malignant Malignant hlafignant

femur

+ ND

NOS

Organ

of Zuckerkandf

Hifutn

of. kidne)

17

ND

ND

Ketroperitoneum

1X 19 20 21

ND + + ND

ND .5i IO 2110 u/19

Urinary bladder Retroperitoneum NOS Organ of’ Zuckerkandf Ketroperitoneum. left upper quadrant

NOS

Abbreviations: ND, not done; NOS. not otherwise specified. * Number of neuropeptides present as analyzed by immunohistochemistI-~.”

1170

Kib Posteriol to vena cava bv extension Femur Liver. invasion to kicfnq Lung. spine, I) mph node5 Spine Spine, skull. humerus, bilateral Widely disseminated mediastinaf and abdominal areas Widely disseminated abdominal area Widefv disseminated abdominal area

hlafignant Malignant Slalignant Malirrnant Mafi~nant Malignant Malignant Xlahgnant Malignant Malignant Benign Benign Benign Benign

PATHOLOGYOF MALIGNANTPARAGANGLIOMAS [Linnoila et al]

FIGURE 2. Bilateral pheochromocytomas resected from 31. year-old white female with MEN syndrome type 20. Cross section of left adrenal gland shows three distinct nodules arising in and expanding the medulla (arrows). The largest nodule measured 1.4 cm in diameter, thus fulfilling the arbitrary “1 cm rule” for a small pheochromocytoma. The periadrenal fat has not been dissected free.

FIGURE 1. Pheochromocytoma of left adrenal gland In a 27. year-old black female. The tumor measured 5.5 cm in diameter and was characterized as gray and focally spongy on cross section. Note unicentrlc well-circumscribed appearance. Adrenal remnant IS attached to periphery of tumor.

bilatet-al itt eight of 17 patients (47% ). and often had ;I very distinctive gross appearance with multicentric growth within the involved adrenal gland(s) (Fig 2). (Lt-eater titan 50% of’ pheocht-omoc~tomas \\.eighed Icss than :30 g or wert S 1 cm in diameter (score 1 ). ‘l-he gross morphologic tindings were usually symmetric in ternis of’ overall size of tumor nodules in t-ach Cglattcl (Fig Z3); the impression was that of two or more distinct nodules of‘ pheochromocytoma arising tvithin thtb adrenal medulla. which contrasted with

the unicentric morphology in the sporadic tumors. Where evident. the extratumoral medulla often exhibited diffuse and/or nodular hy,perplasia. but due to vagaries in tissue sampling thts feature was not evaluable in each case. The distinction between nodular medullary hyperplasia and earl\, (ot small) pheochrontocytomas was entirel!, arbitrq , and similar to other studies, a size of 1 cm or greater ~vas used to define a pheocht-otnocvtotna. None of the patients with MEN syndrome type 2 had extraadrenal or malignant para~angliomas. Five of eight patients with f‘amilial (but not MEN svridrome ryf>e 2) paragangliomas had clinically mal;Cgnattt tumors with documented tnetastases. The mean age at the time of diagnosis in this familial group was 16 l/2 years (range (i to 27 vears) which was sigttificantly,vounger than patients Faith sporadic tumors. In addttton, 9 1% of tumors weighed less than SO g or were c 4 cm in diameter. \vhile only 20% of sporadic. tutttors were in this category (score I). One patient with van Kecklinghausen’x disease had bilateral ~,heochromoc~tomas and also at1 c~xtraadrenal abdominal para,ganglioma. Histologic

Features

of histologic tindings in clinically A tabulation benign and tttalignant paragangliomas is shown in .I‘able 6. There were tto statisticall~~ significant differences in overall architectural patterns (Fig -2) or degree of nuclear hyperchromasia/plec,mc,l-phistn. ‘The most common pattern in malignant rumors was alveolat- (44%) and in the benign tumors the most common pattern ~2s a mixture of’alveolar and trabecular (38%). An alveolar pattern characterized by a more uniform rounded nesting pattern as seen in head attd neck paragangliomas (eg, carotid bode tumors with typical “zellballen”) was uncommon. ?‘\vo clinically maligttattt paragangliomas (one each adrrnal and ex1171

HUMAN PATHOLOGY

Volume 21, No. 11 (November 1990) Histologic Features of Clinically Benign and Malignant SympathoadrenalParagangliomas

TABLE 6.

Number-

of Tumors

[n(%)l Histologic

Features

Architectural pattern Alveolar (nesting, “zellballen”) Trabecular (anastomosing cords) Mixed alveolaritrabecular Diffuse/solid Confluent tumor necrosis* Mitotic rate (mean per 30 HPF. range) Nuclear hvperchromasiaipleomorphism (mean iA scale 1 to 3) Extensive local invasion or vascular invasion* Hyaline globules* Features resembling ganglion cells Proteinaceous material

hlalignant (n = 34)

15 (41) 7 (“1) 11 (33) 0 1 I (32) 3 (O-IX) I.51 11 (J?) 8 (32) 2 (6) 1(12)

Benign (n = 64)

19 19 ?4 2

(30) (30) (38) (3)

4 (6) I (I-10) 1.7 T(l11 3x (.3) I4 (22) 16 (25)

* A statistically significant difference (Pz < .O.i) between cliw ically malignant and benign tumors was noted for the following: confluent tumor necrosis (Pz = .0023), extensive local invasion (P, = .022), and hyaline globules (P, = .0013) by Fisher’s exact test. The difference in features resembling ganglion cells (6% versus 227;) was not statistically significant (P, = .07).

Unusual Histologic Patterns Including Composite Pheochromocytomas

FIGURE 3. Bilateral pheochromocytomas resected from a 30. year-old white male with MEN syndrome type 20. The left adrenal gland weighed 220 g and the right weighed 168 g. Note multicentric origin of tumors within each gland and the variability in size of each nodule. The external aspect of each tumor is seen at top with some showing cystic degeneration.

traadrenal) contained large nests of tumor cells with central degenerative change. Mitotic figures were noted in 29 of the 64 benign tumors (45%) and 22 of the 34 malignant ones (65%). Slightly higher average mitotic counts were obtained in the malignant neoplasms (mean, 3 per 30 HPF) compared with the benign ones (mean, 1 per 30 HPF), but the differences were not statistically significant. Atypical mitotic figures were extremely rare in both groups. A feature more frequently noted in benign paragangliomas was the presence of cells superficially resembling ganglion cells; these were often randomly scattered individually throughout the tumor or in small clusters (Fig 5). The clinically benign tumors also showed an increased incidence of intracytoplasmic hyaline globules (Fig 6) and eosinophilic proteinaceous material resembling colloid (Fig 7). Hyaline globules were regarded as numerous (score 3 + ) in only six of the benign tumors and in none of the malignant ones (Table 6). Vascular invasion was identified in seven of the 34 malignant tumors (21%. Fig 8), but was also apparent in four of 64 benign ones (6%). 1172

Nine of the 120 tumors studied were considered to have distinctly unusual histologic features (8%). Four of them were regarded as composite pheochromocytomas (3%‘), tumors which combined features of ganglioneuroma or ganglioneuroblastoma with typical paraganglioma. All four tumors arose in the adrenal glands as a unilateral solitary mass. In one of the clinically benign composite tumors there were areas of cystic degeneration on cross section (Fig 9). In two cases a component of ganglioneuroma was evident in focal areas (Fig 10, top left) while the other two showed a more diffuse distribution of ganglioneuroblastomatous elements in numerous areas of the primary tumor (Fig 10. top right and bottom). A conspicuous feature in both cases was a relatively abundant neurofibrillary stroma (or matrix) resembling neuropil. Cytologic features of ganglion cells included ample eosinophilic cytoplasm with distinct cell borders, Nissl substance, and eccentric nuclei with a prominent nucleolus. These well-developed cytologic features, coupled with the spindle cell (Schwannian) stroma (Fig 10, top left) or the ganglioneuroblastic component with neurofibrillary stroma, typified these composite pheochromocytomas. Only one of the four tumors was clinically malignant (Fig 10, top right, case no. 2). A spindle cell pattern was a predominant finding in three clinically benign pheochromocytomas; each tumor showed areas of more typical alveolar or trabecular histology. A similar spindle cell component was noted in a number of other tumors, but it was very limited in distribution. One tumor had welldeveloped oncocytic features throughout, and was

PATHOLOGY OF MALIGNANT PARAGANGLIOMAS (Linnolla et al)

FIGURE 4. (lop left) Anastomosing cell cord [trabecular) pattern was apparent throughout this clinically benign, sporadic pheochromocytoma. [Hematoxylin-eosin stain; magnification x 130.) pop right) More discrete alveolar or nesting pattern was a predominant feature in this clinically benign and sporadic adrenal tumor. [Hematoxylin-eosin stain; magnification x 230.) [Bottom) Pheochromocytoma shows a more diffuse or solid growth pattern. (Hematoxyllneosin stain; magnification x 220.)

characterized by cells with abundant granular eosinophilic cy~.oplasm (Fig 1 1). Another tumor had angiomatous features due to intense \ascularity with plump endothelial cells (Fig 12); reticulum stain in this case helped to accentuate the small nests and slender cords of tumor cells. In addition to these unusual patterns there was one instance of tumor to tumor metastasis (Fig 13) in a patient with MEN syndrome type 2 who had medullary thyroid carcinoma metastatlc to a pheochromocytoma. Positive immunostaining for calcitonin helped to differentiate the metastatic carcinoma from the pheochromocytoln;1. with the latter being entirely negative. Statistical

ful parameters predicting malignant behavior were extraadrenal location, coarse tumor nodularitv/multinodularity, areas of confluent tumor necrosis, and the absence of intracytoplasmic hl,aline globules. When the statistical model was e\.alu&ed on the original data set used, 50 of 61 tumors with data which were complete with respect to these foul- parameters \vere correctl!, identified as benign or malignant. This model M’as equivalent to the classification rule shown in Table 7. Although the rule was not infallible, its application on the test set enabled correct classification in nine of’ 12 clinically malignant tumors (75%) and 20 of 21 benign tumors (95%). The overall correct classification rate of 88% (29 of. :13 ‘cases) had an associated 95% confidence interval, ranging from 72% to 9’i’c1(:thus there was at least a 95’7; probability that more than 7Oc% of tumors were being classified correctl) on the basis of this model.

Analysis

The relative frequency of selected histologic parameters in the benign versus malignant paragangliomas is shown graphically in Fig 14. The most success1173

HUMAN PATHOLOGY

Volume 21. No. 11 (November 1990)

FIGURE 6. Clinically benign pheochromocytoma contained moderate numbers of eosinophilic hyaline globules [arrow). Globules range in size from only a few microns to some as large as nuclei of tumor cells, (Hematoxylin-eosin stain; magnification x 330.)

FIGURE 5. Extraadrenal paraganglioma contains diffusely scattered cells with cytologic features resembling ganglion cells, many of them being multinucleated; some cells had granular basophilic staining of peripheral cytoplasm resembling Nissl substance (arrows]. (tiematoxylin-eosin stain; magnification X 330.)

functionally active, presenting with symptoms relating to catecholamine overproduction. The case material in this study consisted mainly of pheochromocytomas and extraadrenal abdominal paragangliomas, which comprise over 95% of all tumors of the sympathoadrenal system. The relatively

There were many ways in which the conditions of the statistical classification rule in Table 7 could be met, but if the conditions were not met, then the tumor was classified as benign. The group of clinically malignant tumors had an average of 1.9 out of these four predictive factors, while the average index for the benign tumors was 0.8 (Fig 15). It should be noted that 7 1% of clinically malignant tumors were associated with two or three of the four parameters, but only 11% of the benign paragangliomas possessed as many as two factors, a difference which was statistically significant (P < .OOl-Mantel test for trend).‘O

DlSCUSSlON This study provides new observations on the predictive value of certain nonhistologic and histologic parameters in clinically malignant paragangliomas. It should be emphasized that, as documented by biopsy and/or serial catecholamine levels, nearly all (94%) of the malignant tumors in the current study had distant metastases, while the remaining two malignant tumors demonstrated extensive local growth. Moreover, 94% of the clinically malignant tumors were

FIGURE 7. Extracellular accumulation of eosinophilic proteinaceous material is evident with peripheral scalloping or clear spaces. Material vaguely resembles colloid of thyroid gland (‘tiyroidization”). Tumor was clinically benign, familial phecchromocytoma. (Hematoxylin-eosin stain; magnification x 220.)

1174

PATHOLOGY OF MALIGNANT PARAGANGLIOMAS (Linnoila et al)

FIGURE 9. Right-sided composite pheochromocytoma resected from a 42-year-old white female who had u hlstory of paroxysmal attacks of headaches and hypertenston. Tumor measured 8.8 cm in diameter and weighed 248 g. Tumor shows cystic degeneration on cross section, with irregular fibrous bands. Remaining tumor is graytan and somewhat lobulated. Tumor proved to be cltnicolly benign FIGURE 8. Cllnlcally mollgnont. sporadic extraadrenal paraganglioma of urinary bladder showed multifocal areas of vascular invasion, probobly lymphattc. Two small foci of angioinvaslon are shown [arrows). Parogongliomas of urinary bladder have been noted to show extension of nests and cords of tumor cells between bundles of smooth muscle of bladder wall, making interpretation of local invasion difficult. (Hematoxylin-eosin stain, magnification 1 220.1

pheochrolnoc.).torrl~ls high inc%tenc~r 01 malignant (2 1‘J ) and estr;utli-enal ttlnlors (L,Z’/;; iT“1 of which 51ere iiialignaiit) is largely attributed to tnatri-ial fi-om the National (ki( er Institute emphasizing treatment of c~linkall\~ rnaligmrlt tumors. ” ‘l‘he incidence of nlafigiianc \ i5 consiclerahlv lower in other serirs. The ittctdenc-c of’ maligi~an~ pheocht-omoc! tomas (iv, atlreml n~ed~dlal-v paraganglionm) in three recent syries \~a:; yC;; ,‘I’ :i
tnor tlodularit\~, confluent tutnor necrosis. .ind the absence of. h&line globules rcere tttost predictive of’ c~linicall~- tnal~gttant beha\kr. Mor~c~w~. most maliqnant tumors had two or three of these features whi‘le benign tumors usually had only one (or tto~te). ‘l‘he presence of necrosis has classicall\, kwt associated with tnaligtiancv, and results of some studies have suggested that this is also the ase with L-linktlly n1;1lignant p”t,“R’LrlBliomas.‘-” \~~%ilr i?:/r 11t the tumors in this study showing confluent necrosis in histologic exatnin;ition wwe malignant, ttec-rosis w;is prtwtit in onlv 32% of‘ malignant tumors, thtts lirrtiting its ttsefultiess as ;t sole prognostic finding. tl;tve t)een r-e Intracvtoplastnic hyaline globules ported iii *wrmali7 and hyperplastic adtwt;ll tnedutomas ;tnd cxtraadlae” as \vell .IS in pheochrotnoc! renal p~iragaiigliotiias.~~ ‘v2’) In 0~11‘ experienc.e, hvaline glol-mles were more often seen in clinicall) benign tumors (5YS7 ) l$,ith the overall incidence king higher than previously reported. Medeiros et ~1. f&r exatnple. identified them in 27% of’ clinic;tll~ Iwnigti tttof‘ the tnors with follo~v-up data.” ‘I‘he sigtiific.ance hyaline globules is not known. hut sotiit’ investigators have associated them ivith secretor! activitv since ultrastructurallv they contain retrtnattts of Mhat appeat to be dense-core granules.” I 111;I ptwious intniutiohistochetnical study. the qtoplasrnic~ hvaline globu1e.s showed ra%bIe staining mtctiort f’ot- 10 diff’et-ent neuropvptides: in most citseb the\ \\.(:rt* nytive, b.hile in some instances the ~lol~~~lr~sssc’rt’ poslt~ve for

HUMAN PATHOLOGY

Volume 21, No. 11 [November 1990)

FIGURE 10. Fop left) Composite pheochromocytoma with nests of neoplastic endocrine cells set wrthin a spindle cell (Schwannian] matrix. Typical pheochromocytoma was identified in numerous sections. Mature gongllon cells were easily seen in other areas of the ganglioneuromatous portion of the tumor. (Hematoxylin-eosin stain; magnification x 210.) Fop right] Clinically malignant composite pheochromocytoma shows areas indistinguishable from ganglioneuroblastoma replete with neurofibrillary matrix resembling neuropil: other areas of tumor showed typical pheochromocytoma. (Hematoxyin-eosin stain; magnification x 330.) [Bottom) Clinically benign composite pheochromocytoma illustrated in Fig 9 shows abundant neurofibrillary matrix with clusters of ganglion cells. Some cells show peripheral bosophilic staining of cytoplasm representing Nissl substance. A typical pattern of pheochromocytoma was present in other areas of this tumor. (Hematoxylin-eosin stain; magnification x 220.)

1176

PATHOLOGY OF MALIGNANT PARAGANGLIOMAS [linnollo et al)

FIGURE 12. Anglomatous features are seen In thts pheochromocytoma. Small nests and slender cords of tumor cells are separated by prominent vascular channels lined by plump endotheliol cells. [Hematoxylin-eosin stain; magnificotlon x 130.)

FIGURE Il. Clinically benign pheochromocyfoma showed welldeveloped oncoc@ic features throughout, with tumor cells having abundant darkly staned eosinophW granular cytoplasm. Tumor had a predominant alveolar or nesting pattern. [Hemoton/lin-eosin stain; magnihcotlon 1 330.1

the same neuropeptidr. vet the rest of’ the c>,toplasm t\‘;is negati! e.” Kesulrs of’ tlk study confirm pi-r\+oua reports that nuclear h~l)er’hI-omasia/pleornorl~hisrn and mitotic activit\, aI-> not helpful in predicting the biolog$ Ijrha\,ior of‘ 1)“‘.“gangliomas. l.‘.‘!’ Flow cytometric atlalvsis of’ nuclear DNA content has been pertijrmed on l;hetj~hI-olllocvtorna tells retrieved from paraffin embedding and the resulrs have been conflicting. Hosaka et al foulId that a normal DNA histogram app;irentl>, defined a group of’ clinically benign tumors. with DNA whereas a Ilumber of pheochroInoc~tonl~s tctraploitl~, . p~~lvpoicl~~. or aneuploid peaks were clinic-Al\ mali~~n~u~~‘.‘” The stud\. by Amberson et al, how0 e*‘, indic ;tted that anruyl&dy was trequenl in cliliic;dlv beni:gn pheochrorno~);tonl~s,~:~ and contrary to lvhar had been suggested In previous reports,g4.z’ aneuploid,\. per se was not considered to he a specific marker of‘ nialignanq hlitotic. figures wt‘re noted in 29 ofthe 64 benign tumors tumors (45si;) and in 22 of. 33 malignant (65% ), but the differences in the average counts obtained rvere not statistically significant. The differences observed could have been accounted f’or by just a ftw tumors uith counts in the higher range Iloted in some of rhr malignant rumors.

Although there were morph&+$ diff’erences between clinically benign and malignant tumors (ie, tumor weight, coarse turnr>r nr)dularitv anti c-ysticchange, invasiveness [both \~ascular and local], anti presence or absence of‘ hvaline globules) these factors lzhen used indi\~idually were not able to staGsticall>, predict whether a gi\.en tumor FV~Sclinically benign or malignant. Howevt7-, using logistic regression analysis. a statistical model was created whi& sirbstantiall~ improved the ability to predict biologic brhavior. ‘I‘he ltllmors (7 1%) observation that most of the malignant had two or three of‘ the four ftatures tloteci above pomay prokie some basis for evaluating malignant tential ot‘s,mpathoadrenal paragangliomus, although linlitatic)nsstill exist. In a pre\?ous study alAyAng the of’ 10 difterent netiilnnlunophenot~I~i~ expression ropeptides. it wx found that there IV;ISdecreased espression in ilinicallv malignant ~J;~rag:anglionias. which might form an adjun&e basis f’or plognostication.” When the data’was analvr.ed together, the predictive ability of thr decreased neuropeptide expression tz appeared greater than anv combination of the currently reported histologic 01‘ nonhistologic parameters. Decreased expression of’ neuropeptide 1 mKNA has also been reported in malignant paragangf iomas. i This study. \vhich included i’7 patients with MEN syndrome type 2, confirmed the high incidence of’ b&tteraJ pheochromocytomas (47%) and multicentric growth, sometimes within a background of recognizable adrenal medullary hyperplasia both diff’use and nodular. which is considered to be the precursor for

HUMAN PATHOLOGY

Volume 21, No. 11 (November 1990) TABLE

7.

Statistical Model for Prediction Tumor Behavior

Step No. 1 2 3 4 5 6

Scoring

of

Values

Score 116 if the location of tumor is extraadrenal. Score 161 if the tumor shows coarsely nodular growth. Score I.58 it histology reveals tumor necrosis. Score - 156 if tumor cells contain cytoplasmic hyaline globules. Sum up the scores. If the sum equals or exceeds 100. classify the tumoras malignant.

NOTE: The logistic model assumed the form: Probability ot malignant tumor = I/[1 + exp(-a-b.x.-b,x,-11,x,)] where exp(z) is the exponential function e’ = 2.71828’. x,. xp are variables being used to predict tumor status, and a. b,. , b, are regression parameters to be estimated. The most suitable logistic regression model incorporated the following parameter estimates: intercept, - 1.736; extraadrenal location. 2.024; multiple tumor nodules, 2.798; the histologic presence of necrosis, 2.742: the presence of cytoplasmic hyaline globules, -2.733.

FIGURE 13. Pheochromocytoma from patient with MEN syndrome type 2a contained metastatic medullary thyroid carcinoma (bottom]. lmmunohistochemical staining for calcitonin revealed metastatic tumor to be diffusely positive for calcitonin, while adjacent tumor was negative. (Hematoxylin-eosin stain; magnification x 130.)

these tumors.“,“” In the 41 patients recently reported by Sammaan et al, 75%’ of tumors were bilateral and none were extraadrenal.” The “1 cm rule” for distinguishing pheochromocytoma from nodular

n q 0

Malignant Benign Total

hyperplasia is a purely arbitrary oneL’H.“9 and is based upon the smallest tumor in the size range reported in 1950.“’ Some investigators have suggested that the pheochromocytomas arising in this setting of adrenal medullary hyperplasia in MEN type 2 represent an extreme mamfestation of hyperplasia.17 In the experience reported by Carney et al, pheochromocytomas were responsible for significant mortality in patients with MEN syndrome type 2; four of 19 patients developed metastases (2 1%‘) with two of them dying as a result of pulmonary involvement.“8 By comparison, it is noteworthy that of the sporadic pheochromocytomas in this study, only four of 59 tumors were bilateral (7%). In an even larger study of 69 sporadic pheochromocytomas reported by Webb et al. none was found to be bilateral.“” Nine of the 120 tumors originally evaluated in this study were considered to have unusual histologic

(34) (64)

(98)

: ’

i:

rl NeCr

rAd

Nod1

Hyal

FIGURE 14. Histopathology of malignant paragangliomas. The incidence (%) of the following features in 98 sympathoadrenal paragangliomas [open bars] with clinical follow-up is presented: necrosis (Necr). vascular and/or capsular invasion (lnv], extraadrenal location (ExAd), multinodularity of tumor mass (Mu), and the presence of cytoplasmic hyaline globules (Hyal). Solid bars represent malignant tumors with these features, and hatched bars benign tumors, Note that necrosis was seen in 15% of paragangliomas, of which 73% were malignant; 23% of tumors were in extra adrenal locations, of which 77% were malignant. In contrast, 47% of tumors had hyaline globules, but only 17% of them were malignant.

1178

Benlqn In = 64)

Malignant (rl = 3-l)

FIGURE 15. Predictive features in paragangliomas. Using regression analysis, necrosis, extraadrenal location, multinodularity, and the lack of cytoplasmic hyaline globules were found to be the most suitable features in predicting clinically malignant behaviors in sympathoadrenal paragangliomas. The 64 benign tumors (open circles) demonstrated paucity of these features with a mean of 0.3 features per tumor, while the 34 malignant tumors (closed circles) had a mean of 1.9 of these features, ranging from 0 in one tumor to 3 in nine tumors.

PATHOLOGYOF MALIGNANTPARAGANGLIOMAS (Linnoila et al)

(‘l‘alk H) and. due to the relativeI>. large number of tlunors examined, these data mav provide a useful estimate for their incidence. A spindle cell pattern h.ns been reported in 3% to 8’2 of pheochroniocytomas. cxtraadrenal paragangliomas. and in adrinal medullary h~perplasia.“-:‘,‘X.‘!‘.:” In the stud?, by Medeiros et al, a spindle cell pattern was seen f’ocall\. in a I hird of‘ their cases.” Similarl!,. a spindle cell coniponent was not uncommon in our material, hut ah a domin;mt histologic feature it was present in onI) three of 120 tumors (2..5%) each of’ which was clinically benign. Sherwin noted that a spindle cell tentienc), seemed to be a characteristic of man!- of the invas’ive tumors which had been reported, but its presence tlid not always correlate with metastasis.“’ Another unusual histologic pattern has been characterized as angiomatous featuresz” with prominent vascular channels lined by plump endothelial cells, a pattern which ma?’ lead to diagnostic problems if’ one does not focus 011 the neoplastic elements which appear as attenuated nests and slender cords of cells. Staining fol. reticulum may accentuate the small nests and slender cords of‘ neoplastic cells, thus aiding in (,orrect diagnosis. The I Ilmors designated herein as composite pheochrornoc)tolnas f’orm a very small but interesting group of neoplasms with morphologic features of ganglioneuroma or ~anglioneuroblastoma combined with typical areas of pheochromocytoma. In our exfour had c’omposperience uith 120 paragangliomas, ite ftatures (3.5%) and each arose in the adrenal gl;md. One of’ the tumors which proved to he clinically maligliant had some areas of more primitivt neuroblastoma. As evidenced by the pawit). of’ cases reported in the literatllre,:‘:‘.“!’ these composite tumors are quite rare, but are not entirely unexpected given the common embryologic ancestry of’ chromuffin cells of the adrenal glands and the s):mpawhich 1s rethetic nervous system. a relationship tlected in I he rerrn “neurocristopathy” proposed by Rolande.“” Using a transplantable rat pheochromocyoma, some investigators have reportecl cellular features common to both pheochromocytoma and neurot)lastoma. and postulated divergent differentiation from a progenitor stem cell.-” Moreover, normal and neoplastic human chromaffin cells can manifest neuronal characteristics in vitro,4y and this plastic phenotype may be the in viva counterpart of composite tumors. In addition. cultured pheochromocytoma cells ;Ire known to produce f’actors that are mitogenic for bv such gli;il cells. ,a:\ and it is possible that production teatu1-es

tumors in viva might induce compositt* tumors, including tumors where a paraganglioma is combined with a glioma as reported in the conus medullaris and cauda rquina.-” An unusual cast of’ malignant peripheral ner\‘e sheath tumor and pheochromoc~toma has also been reported, thus evidencin,g the possible ividening spectrum of composite tumors. ‘:I An interesting pathologic f’eature in several reports of‘ composite pheochrorrlocvtonl~~s is a water); diarrhea syndrome probably related to s) nthesis and secretion of vasoactive Intestinal polyl)eptide (VIP) by the tumor.“” *I‘he capacity fi)r VIP elaboration has been associated with the presence of ganglionic cells within the tumor,.“’ but this does not appeal to be an absolute requirement, since ;I number of’ pheochromoc,~tc,mas shown to be positi\,f* fi)r VIP using immunoc~tochemical stainin either lack ganglior& (or neuronal) cells, or contain cmlv sparse numbers,“:‘7 .1x In summary, results of’ this stud\. indicate that there exist some nonhistologic and histologic differences between clinically benign and malignant symlopathoadrenal pa~~~~gan~liornas. and f’urthermore, gistic regression analysis of’the data base \uggests that a combination of’ two or more selected Factors identified herein has predictive value in correc tl!, identif.jing malignant tumors. The statistical model used hcnl~ever, is not infallible, and long-term clinic.al f’ollow-up in a number of cases may prove 01 he the final arbiter in diagnosis.

REFERENCES 1. Nevillr ARI: rhe adrenal mrdulla. 111S\ m~n~ton I‘ fed). :\tirenat (;t;ul(t. Haltimorr. Fun&mat Patholog\ of the Human ML). \Villiams and Wilkins. t9ti9. pp 2 I i-324 II

1179

HUMAN PATHOLOGY

Volume 21, No. 11 (November

pression of neuropeptides in malignant paragangliomas: An immunohistochemical studv. Ilv>< PATHOL 19:47-50. 198X 7. Helman LJ, (:oh& PS, Averbuch SD. et al: Neuropeptide 1’ distinguishes betiign from malignant pheocht-c,mocytoma. ,I (:litt Oncol 7:720-725. 19X9 8. Press S,]. Wilson S: <%oosing between logistic t-egression and discriminant analysis. J Am Stat Assor 73:tiW705. 107X 9. Harrell FE Jr: SAS Supplemental Librar) Users (;uid?, 1980 Edition. (:ary. NC. SAS Institute, IWO, p X3-102 IO. Mantel N: (:hi-square tests with one deqee of freedom: Extensions of the Mantel-Haenzel Procedure. J Am Stat Assoc :ix:690-70(). I963 Il. Averbuch SD, Steaklev (1s. Young R<:, et al: Malignant pheochrontocvtotna: Effective treatment with a combination of cyclophospham’ide. vinrristine and dacarbazine. Ann lnt Met1 109:267-273. lW8 12. Melicow MXI: One hundred cases oi pheochrotnoc~tonta (107 tumors) at the t-gan of Zuckerkattdl. Ann Sut-g IX3:iiX-iX6. 1976 Ii. Dekker A, Oehrle ,JS: II,>aline globules of the adrenal medulla of man: A product of liptd peroxidation. Arch l’athol Lab \lecl 91:353-364. 1971 IX. DeLellis KA. Wolfe Ii,], (&gel RF. et al: Adrenal tnedullary hyperplasia. A morphometric anal! sis in patient\ with familial medullary thvroid carcinoma. Am J Pathol X3: 177-190. IO76 19. &erwitt RP: Histopathologv of pheochromoc~torn;t. (:ancer 12:X61-877, 19.X 20. Wilson RA, Ihanez 511.: .4 comparatiw studv ot II cases of’ tamilial and nonfamilial plreorhrotnocvton~~ts, HLX P.UHOL !I: I H I 18X, 197x 2 1. Xlendelsohn (;, Olson JL: I’heocIlt-c~mc~rvtc~mas. HYV 1’.~~rttot. 9:607-WX, 1978 (letter) 22. Hosaka Y, Rainwater Lhl, Grant C:S: l’heochrottto~~~ot~~~~: Nuclear deoxyribonucleic acid patterns studied b! flow cvtometrv. Surgery 100: lOO3- 10 IO. 1986 23. Amherson JB, Vaughan ED Jr, (Gra! (;F, et al: Flow cvtotnetric determina& of nuclear DNA content in benign adrenal pheochromocytomas. IIt-olog) 30: IO?- I O-l, 19X7 21. Lewis PD: A cytophotometric rtudy of. benign atld malignant pheochromocvtot~~as. Virchows Arch [B] 9:37 I-376. I97 I 2.5,. Klein FA, ‘Kay S. Ratliff. JE: Flow cvtometric determinations of ploidy and prolifet-ation patterns of adrenal neoplastns: An acijunct to histological classification. J Ural 134:X62-Htiti, 198.5 26. (:arne) JA, Sizemot-e GW, Tpce (;M: Bilateral adrenal medullary h);perplasia in multiple endocrine neoplasia. type 2. The precursor of bilateral I’t”ochromocytonra. Maw Clin Pror 50:310. 1975 27. Sammaan N.4. Itic-kev KC:. Shutts PE: Diagnosis. IocaliLation and management ot pheochromoc)toma. Pitfalls and followup in 4 1 patients. Cancer 62:245 l-2460, 198X 28. Carney JA. Sizemore GM’. Sheps SC:: Adrenal medullar) disease in multiple endocrine neoplasia, type 2. Pheochromocytoma and its precursors. Am 1 Clin Pathol 66:?79-290. 1976 29. Webb TA, Sheps St;, Carney JA: Differences between

1990)

sporadic pheochronlo[!tolna and pheorht-onloc~tonla in multiple endocrine neoplasia. type 2. Am J Surg Path01 4: I2 l- 126. 1980 3(1. Karsner HT: ‘Tumors of the Adrenal. First Series. Farcitle 29. Washington. DC, Armed Forces Institute or Pathologk 31. Lewis HJE. Reid R. Murci B. et al: Malignant pheocht-omocvtoma. Br J Crol 57:394-39X, 198.5 ?2. Shin \$-Y, (it-oman GS. Berkman JI: Pheuchromocytoma with attgiomatous features. A c-ase report and ul~rastructural stud!. (;ancer 40:275-283, 1977 33. Trump DL. Livingston JN. Baylitt SB: Watery diarrhea syndrome in an adult with ganglioneuronta-pt~e~~cllront~)~~torn;t. Identification of vasoactive intestinal peptide. calcitonin and catecholamines and assessment of their biologic activities. (:ancer 40:1526-1532, 1977 31. Kragel PJ, Johnston (:A: l’heochromoc~to~~~~-~at~~liotteut-oma of the adrenal. .4rch Pathol Lab Med 109:47(3-472. 1985 35. Nakagawara A, Ikeda Ii, T~uneyoshi ht. et al: h~alignant pheochromocytoma with ganglionettrohlastotna eletnents in a patient with van Recklinghausen’s disease. Cancer 35:2794-279X. 1985 36. Lakfield LJ. (Glasgow BJ, DuPuis RIH. et al: Aspiration cytology and immttnohistoc-hemistr) of a pheochromocytom;(ganglioneuroma of the adrenal gland. Acta C) rol 3 1 :?&Y!l. 1987 37. Tischler AS, Dayal 1’. Halo@ K. et al: The distribution of immttnoreacti\~e chromogt-anitts, S-100 protein, and vasoactive intestinal peptide itt cot~~pouttd tumors of the adrenal medulla. Hcht I’XI.HOI. I8%,%Yli. I%i 3X. Aiha M. Hiravama A. Fujimoto Y. et al: A compound adrenal medullary tumor (pheochronroc:ytoma and ganglioneuroma) and a rorttcal adenoma in the ipstlateral adrenal gland. A case report with enzyme histochemical and immunohistochenlical rtudies. Hc-u P.V~HOL 12:5X-566, 198X 39. Balars 51: Mixed pheochromoclitoma and ganglioneuroma of the adrenal medulla: A case report with electron microscopic examination. H~ht P.4.t HOL. 19: 13.52-1X5,5. 198X -10. Bolande RP: The tteut-orristopathies: A uttif\ing concept of disease arising in neural crest maldevelopmetrt. HL~M PxrHoL .5:-w-429. 197-l 4 1, \Iurphy GP. Abramovici A. Servadio (1. et al: Furthetstudies of a trattsplantahle rat pheochrornocvtoma. Oncolog> 44::w-3x5. 1987 42. ‘Tischler AS, Die htet- 51. Biales B. et al: Neural properties of human endocrine cells of proposed neural crest origin. Science lY2:902-904, 1976 -13. Kattter N. (ilaser L. Bunge RP: I’<: 12 cells as a source of tteurite-derived cell surface mirogen. which stimulates Schwnn cells division. ,J Cell Biol 1)X:1 I.‘()- I 15.5. I984 44. Shuangshoti S. Suwanwela N, Suwanwela (:: (:omhined pardganglioma and glioma of conus medullaris and cauda equina. J Sut-g Onrol ?.5: l62- 167. 19X-l 45. hlin K-W, Clentens A. Bell J, et al: Xlaligttant peripheral nerve sheath tumor and p~~eochtotnoc~totna. Arch Pathol I.ah Med I I‘):966-‘170 ..* * , 1988 46. %lendelsohtt (;, Eggleston JC, Olson JL, et al: Vasoactive intestinal peptide and its relationship to gan$on cell dif.f’erentiaticm in nrurohlastic tumors. Lab Invest 4 I : 144- 149. 1979 37. Sane T. Saito H, Inaha H, et al: I mtnuttoreactive somatortatin and vasoartive intestinal polypeptide in adrenal pheochromocv~oma. An imtnunochetnical and ultrastructul-al study. Cancer ‘,“:‘$j”-3#~ 198’3 ~w*_, . 4X. Wale C;, Dell’Orttr 1’. Moto E. et al: Vasoactive intestinal polypeptide-, somatostatiw. and calcitonin-protlucing adrenal pheochromocytoma associated with the water) diarrhea (WDHH) syndrome. First report with itnmunohistochemical findings. Cancer .55:1(W-1 106. 1985

1180