Intermediate Filament Proteins

Group I neuroendocrine neoplasms manifest uniform immunoreactivity for keratin, especially keratin classes 8 and 18.⁶¹ To a lesser extent, keratin 19 is also observed in NECs in general. Pulmonary NECs may label for keratin 20 as well, but only in fewer than 10% of cases.⁶² Keratin proteins are generally well recognized by several commercial monoclonal antibodies, especially CAM5.2 and MFN116. Monospecific reagents directly against only one keratin class polypeptide are also available, but are not necessary in this specific context. In our opinion, the optimal approach to the detection of keratin in any poorly differentiated neoplasm, neuroendocrine or otherwise, is to prepare a mixture of monoclonal antikeratin antibodies with partially overlapping and partially distinctive keratin class specificities. When reagents of this type are used on paraffin sections, together with properly performed proteolytic digestion or microwave (heat)-mediated epitope “retrieval” methods, the detectability of keratin in NECs should approximate 100%, even in essentially “undifferentiated” tumors (e.g., SCC).⁶³ Another feature of keratin reactivity in many NECs is a distinctive punctate or globoid perinuclear zone of positivity.⁶⁴ This result (Fig. 13-5) is simultaneously diagnostic of both epithelial and neuroendocrine differentiation in a small cell malignancy, and it obviates the need for other “neuroendocrine markers.” Keratin is not typically expected in group II neuroendocrine neoplasms, although some lesions in that group have shown its presence in an “aberrant” manner.^65–68 The foremost example of that phenomenon is represented by PNET with “divergent” differentiation (also known as “polyphenotypic small cell tumor”), an example of which is the “desmoplastic small round cell tumor.”^69–71 In most cases of keratin expression in more nondescript PNETs, reactivity for that protein is focal, unlike its pattern in NECs; in addition, vimentin expression tends to be mutually exclusive in PNETs and NECs. Nonetheless, problems in differential diagnosis may arise between those classes of neuroendocrine tumors, sometimes necessitating cytogenetic evaluation to distinguish between them.⁷²

Figure 13-5 Paranuclear “dot”-like immunoreactivity for keratin in small cell pulmonary neuroendocrine carcinoma. This pattern concurrently establishes the epithelial and the neuroendocrine nature of the tumor.

Neurofilament protein is variably coexpressed by NECs,^73–76 but that determinant is seen as the sole intermediate filament in the majority of differentiated group II neuroendocrine tumors, such as PGs and pheochromocytomas, as well as some NBs.^77–79 Unfortunately, neurofilament protein is not well visualized in paraffin sections, even with epitope retrieval technology, and it is most reliably evaluated in frozen material. Vimentin also may be evident in the cells of PGs in approximately 50% of cases,⁷⁹ and it is the only intermediate filament that can be detected in primitive group II tumors, such as NB and PNET.⁶⁸ That is also true in Ewing sarcoma, a tumor that is in the same family as PNETs. Vimentin is only exceptionally present in NECs, as stated earlier. Eusebi and coworkers⁸⁰ recently described three NECs that coexpressed keratin and desmin as a reflection of divergent (sarcomatoid) rhabdomyoblastic differentiation. This same proclivity is regularly seen in desmoplastic small round cell tumors, in which keratin, desmin, and vimentin are commonly present concomitantly in the same neoplastic cells.^69–71

Glial fibrillary acidic protein is not an expected reactant in either NEC or PNET. In the central nervous system, therefore, this marker is helpful in the differential diagnosis with small cell malignant gliomas,⁸¹ most of which are reactive for glial fibrillary acidic protein. That fact is all the more important because a proportion of glial tumors manifest aberrant keratin reactivity⁸² and also express vimentin.

Chromogranins

Chromogranins are matrical proteins that are associated with neurosecretory granules and are absolutely specific for neuroendocrine differentiation.^83–85 These polypeptides have been subdivided biochemically into two groups, A and B,^86,87 with the latter being synonymous with secretogranin-I. Conceptually, every cell that packages peptides into neurosecretory granules must synthesize either chromogranin-A (CGA) or chromogranin-B (CGB); hence, a screening reagent incorporating antibodies to both of those proteins would be extremely useful diagnostically. Unfortunately, reliable commercial anti-CGB products have not been introduced. The most widely used anti-CGA antibody is clone LK2H10.⁸⁸ This reagent was raised against pheochromocytoma cells and shows excellent reactivity with paraffin sections (Fig. 13-6). The major shortcoming of anti-CGA is twofold. First, because neuroendocrine cells or tumors may preferentially synthesize CGB, CGA obviously cannot be regarded as a “universal” marker of such elements. Secondly, the detectability of both CGA and CGB is directly related to the number of neurosecretory granules in any given neuroendocrine cell population. If one is dealing with poorly differentiated malignancies that have only scant numbers of such organelles, immunostains for CGA and CGB will be predictably negative in the majority of cases. Therefore, one may legitimately conclude that a tumor has neuroendocrine properties if it can be labeled for a chromogranin, but negative results do not exclude that possibility, particularly in high-grade neoplasms.

Figure 13-6 Immunoreactivity for chromogranin-A in neuroendocrine carcinoma of the lung.

Synaptophysin

Synaptophysin is a 38 kDa molecule that is associated with the synaptic vesicles of neurons and cells with neuroendocrine or neuroectodermal characteristics.^78,89–92 Monoclonal antibodies to this marker have been used widely in diagnostic surgical pathology and cytopathology, with good success, particularly in conjunction with epitope retrieval techniques (Fig. 13-7). In light of its subcellular associations, one might assume that synaptophysin would have a synonymous tissue distribution to that of the chromogranins; however, in practicality, that is not true. Thus, a sizable proportion of neuroendocrine neoplasms label for CGA or CGB, but not synaptophysin; the converse also applies.⁹³ Thus, anti-synaptophysins and anti-chromogranins should be conceptualized as complementary reagents.

Figure 13-7 Immunolabeling for synaptophysin in pulmonary neuroendocrine carcinoma.

Loy and colleagues⁵⁹ have challenged the specificity of chromogranins and synaptophysin as markers of neuroendocrine lineage, in a study using ultrastructure as the standard for definition of that lineage. However, in our opinion, the concept behind that analysis was flawed. Sampling bias (a significant problem in fine structural evaluations) probably accounted for those cases in which immunoreactivity was observed for the specified endocrine markers, but no neurosecretory granules were identified by electron microscopy.

CD57 (Leu-7)

CD57 was originally characterized as a marker for natural killer lymphocytes, and it is recognized by monoclonal antibody HNK-1,⁹⁴ among others. Subsequently, shared epitopes of this molecule have been detected immunohistochemically in normal constituents and tumors of the brain, peripheral nervous system, soft tissues, prostate gland, thymus, and DNS.^95–101 With specific reference to neuroendocrine neoplasms, CD57 appears to bind to a matrical component of neurosecretory granules that is distinct from both CGA and CGB.¹⁰² It both corroborates and extends the sensitivity of chromogranin immunostains in many clinical settings. Nonetheless, because of the imperfect specificity of Leu-7, it cannot be used as confidently as a “standalone” neuroendocrine marker. Moreover, it has the same failings in sensitivity as CGA and CGB that relate to the density of neurosecretory granules in poorly differentiated tumors.

Neural Cell Adhesion Molecule (NCAM; CD56)

The monoclonal antibodies 123C3 and JLP5B9 recognize formalin-preserved epitopes of CD56, or neural cell adhesion molecule, a cell membrane protein that has a role in the cohesiveness of cells in the peripheral and central nervous systems.^103–109 Like synaptophysin, NCAM is distributed among neuroendocrine and neuroectodermal cells and tumors (Fig. 13-8). Several studies have shown that NCAM is a sensitive marker of endocrine lineage in SCC of the lung as well as extrapulmonary sites.^106–108 Nonetheless, it is not absolutely specific for neuroendocrine differentiation; a minority (up to 25%) of ovarian surface carcinomas, renal cell carcinomas, nonendocrine lung cancers, and endometrial carcinomas demonstrate CD56 immunoreactivity.¹⁰⁸ Despite this caveat, antibodies to NCAM are useful additions to diagnostic panels for endocrine tumors. Reagents raised against the polysialylated form of CD56 are considered the most sensitive among this group.¹⁰⁵

Figure 13-8 Immunoreactivity for CD56 (neural cell adhesion molecule) in neuroendocrine carcinoma of the lung.

Neuron-Specific (Gamma-Dimer) Enolase

“Neuron-specific” enolase (NSE) is a gamma-dimeric form of 14-3-2 protein, a glycolytic enzyme that is present in neurons and cells of the DNS.¹¹⁰ It is thought to be associated with the formation of intercellular synapses in the nervous system. NSE was one of the first markers to be used in diagnostic immunohistochemistry as a general indicator of putative neuroendocrine differentiation.^110–112 Heteroantisera to NSE are very sensitive in that regard, labeling virtually all neuronal and neuroendocrine proliferations, regardless of the level of cellular differentiation.¹¹² Nevertheless, the specificity of those reagents is poor, owing to cross-reactivity between gamma-dimers and heterodimers (e.g., alpha-gamma, alpha-beta) of NSE that are expressed by non-neuroendocrine neoplasms of many types.¹¹³ In reaction to that problem, monoclonal antibodies to NSE have been developed and tested,¹¹⁴ but these have generally manifested a low degree of sensitivity and have not enjoyed widespread usage. An alternative approach to lessening the cross-reactivity of anti-NSE heteroantisera is to absorb them with acetone-fixed, pulverized human tissues known to contain high levels of the alpha or beta form of the molecule. However, that approach is time-consuming and somewhat demanding technically, and for this reason, has not been embraced in clinical practice. Currently, antibodies to NSE are primarily used as “screens” for endocrine differentiation,^115–117 and reactivity with them is usually pursued further by applying other neuroendocrine immunostains, as described elsewhere in this chapter.

Protein Gene Product 9.5 (PGP9.5)

PGP9.5 is a protein that removes ubiquitin (an intermediate filament) from other proteins and protects them from degradation by proteases. It is expressed widely within cells and tumors of the DNS and therefore has been applied as an immunohistochemical marker of neuroendocrine differentiation.^118–120 Nevertheless, the general distribution of PGP9.5 in human neoplasms has shown that it is not particularly specific for an endocrine lineage; in particular, non–small cell carcinomas of the lung that lack neuroendocrine morphologic patterns and contain no neurosecretory granules on ultrastructural analysis have been PGP9.5-positive in approximately 55% of cases in some series.¹²¹ Thus, the practical utility of this marker is similar to that of NSE.

Specific Neuropeptides

Specific neuropeptides, such as adrenocorticotropic hormone, gastrin, insulin, somatostatin, calcitonin, “whole” bombesin and its C-terminal flanking peptide, leu- and met- enkephalins, were the initial molecular moieties that were evaluated immunohistochemically in an effort to substantiate the neuroendocrine nature of selected epithelial human neoplasms.⁹³ Today, these markers have been largely relegated to a secondary role in the pathologic assessment of tumors in the DNS. Antibodies to such peptides are of academic interest in correlating clinical endocrinopathies with histopathologic findings, but are not nearly sensitive enough to serve as screening reagents. Another potential but limited application is in the delineation of “occult” (non–endocrinopathy-related) peptide production by neuroendocrine tumors, which may serve as the basis for serologic monitoring of tumor growth and activity.

CD99 (MIC2; p30/32 Protein)

The membranocytoplasmic protein known as “CD99” in the hematopoietic antigen cluster designation is the same molecule that has been called “MIC2” or “p30/32 protein.”¹²² The function of this moiety is uncertain, but it is empirically known to be expressed in virtually all PNETs and Ewing sarcomas^122,123 (Fig. 13-9). The specificity of CD99 antibodies for a neuroectodermal lineage is not absolute, however, because they may also label a minority (<15%) of alveolar rhabdomyosarcomas as well as the overwhelming majority (90%) of lymphoblastic lymphomas.¹²⁴ CD99 is observed in up to 20% of NECs in some body sites as well.^125,126 This is important because it may obscure the difference between NEC and PNET in selected instances; this is particularly true in light of the potential for keratin reactivity that both lesions have. On the other hand, differentiated group II neuroendocrine tumors, such as PGs, are nonreactive for this determinant.

Figure 13-9 Immunolabeling for CD99 (MIC2 protein) in primitive neuroectodermal tumor. This marker can also be observed in neuroendocrine carcinomas.

Other Reagents

Several other antibodies have been developed that have immunohistochemical properties that overlap those of LK2H10 (anti-CGA). These include both polyclonal and hybridoma reagents. A monoclonal antibody designated “HISL-19” by Krisch and coworkers¹²⁷ is believed to manifest neuroendocrine specificity. The determinant that it recognizes is proteinaceous, but seems to be nonidentical to chromogranin. This statement is based on the relative strength of reactivity of HISL-19 with several neuroendocrine tissues, which is dissimilar to that of LK2H10. Moreover, some nonendocrine tissues and tumors, such as the gallbladder epithelium and adenocarcinomas of the lung, stomach, and endometrium, are labeled by the former reagent but not by the latter. These differences notwithstanding, HISL-19 appears to bind to a neurosecretory granule-related moiety, and neoplasms with few of these granules (parathyroid adenoma, melanoma, NB, and oat cell lung cancer) are only weakly stained, but others with numerous granules (pituitary adenoma, pheochromocytoma, medullary thyroid carcinoma, carcinoid, and islet cell tumor) are labeled intensely. Western immunoblot analyses have shown that the target of HISL-19 is not neuron-specific enolase.¹²⁷

Lloyd and colleagues¹²⁸ employed monoclonal antibodies to adrenaline (epinephrine) and noradrenaline (norepinephrine) in the study of similar neuroendocrine neoplasms. In general, pheochromocytomas displayed adrenaline alone, but extra-adrenal PGs exhibited reactivity for adrenaline and noradrenaline, or noradrenaline only. NBs, carcinoids, pituitary adenomas, pancreatic endocrine tumors, and parathyroid adenomas also showed positivity for noradrenaline. These authors emphasized that catecholamine-positive neoplasms also expressed reactivity with LK2H10, implying that stains for adrenaline and noradrenaline did not add appreciably to the diagnostic information obtained with antichromogranin.

Haspel and colleagues¹²⁹ found that mice infected with reovirus type I often had autoimmune syndromes featuring polyendocrinopathies. Spleen cells from such animals were used as the substrates for hybridoma production, and two of the resulting monoclonal antibodies (5B5 and 5B8) showed reactivity with human anterior pituitary cells in frozen and Bouin’s-fixed specimens.¹²⁹ Further immunostaining results in normal and neoplastic human tissues were not provided, but it was believed that 5B5 and 5B8 recognized discrete hormonal products, such as growth hormone, based on competitive inhibition studies.

The absence of discussion of S-100 protein may seem surprising because of the still-common belief that this marker is associated with neuroendocrine tumors. In reality, that is not so. The only endocrine neoplasm that reproducibly contains cells positive for S-100 protein is PG, and the immunoreactive elements therein are actually “sustentacular” cells rather than tumor cells.⁷⁹ It has been contended that sustentacular elements decrease in density when PGs undergo malignant change,¹³⁰ but that is an uncertain tenet in reference to individual cases.

Several markers have been assessed with regard to their associations with neuroendocrine tumor grade; SOX2, PAX5, and CD117 are the principal analytes. Each demonstrates a tendency for greater reactivity in this context as the histologic grade increases, such that grade III lesions show the highest level of immunolabeling.^131–133

Yet another applicable reagent is antithyroid transcription factor-1 (TTF1).^134–140 This intranuclear protein is expressed not only by thyroid epithelium but also by glandular cells of the lung, pulmonary adenocarcinomas, and NECs of the lung (Fig. 13-10). Among the last of these lesional groups, SCC and large cell NECs have the highest incidence of reactivity (in approximately 85% to 90% of cases). Interestingly, however, extrapulmonary high-grade NECs show a tendency to be TTF1-positive in many organ sites (bladder, uterine cervix, prostate, gastrointestinal tract), and therefore this marker can be considered to represent an “adjunct” neuroendocrine determinant, if additional studies for thyroid-related and lung-associated labels are negative. Obviously, then, TTF1 cannot be employed reliably to distinguish metastatic NEC of the lung from other neuroendocrine malignancies.

Figure 13-10 Nuclear immunoreactivity for thyroid transcription factor-1 in pulmonary neuroendocrine carcinoma.

Clinical Features

Grade I NECs with typical histologic features are rarely diagnostic problems. They usually grow as polypoid intraluminal masses with an intact overlying epithelium (Fig. 13-12) or one demonstrating squamous metaplasia.^45,147,148 This pattern explains a common clinical presentation, which is localized airway obstruction. Patients manifest with wheezing, cough, or pneumonia. Carcinoid syndrome almost never develops, presumably because the tumors in question are incapable of synthesizing the biochemical substances responsible for its pathogenesis. Rare individuals with grade I NEC have associated Cushing syndrome or other endocrinopathies as a result of ectopic neuropeptide production by the neoplasm (see Fig. 13-12D).^153–159 A proposed association between pulmonary carcinoids and sarcoidosis has also been made.¹⁶⁰

Figure 13-12 A, Endoscopic photograph of low-grade neuroendocrine carcinoma, protruding as a polypoid mass into the bronchial lumen. B, Gross photograph of grade I neuroendocrine carcinoma of the lung, straddling the bronchial cartilage. C, Photomicrograph of low-grade neuroendocrine carcinoma, protruding into the bronchial lumen. D, Immunoreactivity for adrenocorticotrophic hormone in this low-grade pulmonary neuroendocrine carcinoma. The patient had Cushing syndrome.

Localized obstruction also dominates the radiographic picture, with evidence of obstructive pneumonia or atelectasis (Fig. 13-13). Occasionally, a central mass with a dumbbell-like configuration is seen on plain films, and computed tomography usually demonstrates a lesion within and adjacent to a large airway.¹⁶¹

Figure 13-13 Bronchial obstruction by central low-grade neuroendocrine carcinoma caused segmental atelectasis of the right upper lobe, as seen in this computed tomogram of the thorax.

Men and women are approximately equally affected by pulmonary “carcinoids.” Although all age groups may have grade I pulmonary NECs, young to middle-aged adults account for the majority of cases. The lesions therefore appear at substantially lower ages than those associated with other pulmonary carcinomas.

Peripheral carcinoids are subpleural, small, and well circumscribed.^148,151,152 They often present as incidental findings, lacking the propensity to produce the obstructive changes of their central counterparts. Instead, the differential diagnosis of a “coin lesion” is encountered. Radiographically, such entities as granulomas, hamartomas, or peripheral adenocarcinomas enter consideration. Some studies have shown a predominance of peripheral grade I NECs in the right middle lobe, and a greater number of women have those tumors compared with central NECs of the lung.^151,152

Rarely, grade I NECs may be synchronously multifocal.^162,163 That eventuality is troublesome because of the possibility of metastasis to the lung from an extrapulmonary source. Extensive clinical evaluations and appropriate immunohistologic studies (discussed later) are usually required.

Gross and Microscopic Pathologic Findings

“Classic carcinoids” are typically 2 to 4 cm in maximum dimension. The lesions vary in color from tan-yellow to dark red, and they lack obvious internal necrosis and hemorrhage. Central tumors display diverse growth patterns, including trabecular, ribboning, insular, and solid sheet-like configurations^11,147,148 (Fig. 13-14). A delicate fibrovascular stroma is present, sometimes with associated matrical amyloid deposition.¹⁶⁴ Rare lesions may contain metaplastic bone or cartilage.¹¹ The latter findings are postulated to reflect the production of factors related to tumor growth factor beta or various bone morphogenetic proteins.¹⁴⁸ Tumor cell cytoplasm is relatively abundant, and it may be strikingly granular and oncocytic or even clear.^165–169 Eccentricity of the nuclei can yield a plasmacytoid cellular image, particularly in fine-needle aspiration biopsy specimens (Fig. 13-15). A papillary configuration has also been reported.^11,170 Other variants of grade I NEC produce melanin,^171,172 and some are said to contain sustentacular cells immunoreactive to S-100 protein, as is more typically seen in PGs.¹⁷³ An exceptional case presented by Skinner and Ewen¹⁷⁴ featured diffuse replacement of the lung parenchyma by carcinoid tumor cells.

Figure 13-14 Grade I neuroendocrine carcinoma (classic carcinoid) may assume several growth patterns, including trabecular (A), insular (B), pseudoglandular (C), and spindle cell (D; in a cell block from a fine-needle aspiration biopsy specimen).

Figure 13-15 Fine-needle aspirates of low-grade neuroendocrine carcinomas may show dyshesive cells with plasmacytoid configurations, causing potential confusion with hematopoietic proliferations.

Peripheral carcinoids may be morphologically identical to central tumors, in which case the diagnosis is usually made without difficulty. Nonetheless, peripheral grade I NECs often demonstrate a prominent spindle cell growth pattern, which is associated with a somewhat different set of diagnostic alternatives (discussed later)^151,152,175 (Fig. 13-16).

Figure 13-16 Spindle cell growth in a peripheral low-grade neuroendocrine carcinoma of the lung.

Differential Diagnostic Considerations

The intraluminal growth pattern of grade I pulmonary NEC can be simulated by salivary gland-like tumors, such as mucoepidermoid carcinoma; mesenchymal lesions, including peripheral nerve sheath tumors and rare inflammatory pseudotumors; and some metastatic neoplasms. Most of these possibilities present little difficulty histologically. Nonetheless, problems may arise occasionally in differentiating carcinoids from well-differentiated adenocarcinomas, higher-grade NECs, primary pulmonary PGs, and “solid” adenoid cystic carcinomas, particularly in small transbronchial biopsy specimens, which may demonstrate significant artifactual distortion. Distinction from “atypical” carcinoids is discussed in more detail later. However, uniform nuclear features, the lack of significant mitotic activity, abundant cytoplasm, and the absence of necrosis are all indicative of grade I NEC rather than a grade II tumor.^3,5,149 Because of the well-differentiated nature of classic carcinoids, one can expect almost all of them to be diffusely reactive for keratin, CGA, synaptophysin, and CD57,^4,173 whereas the other diagnostic alternatives do not show that immunophenotypic profile. The p53 gene, as studied by immunohistochemical or genetic analysis, is infrequently mutated in grade I NEC,¹⁷⁶ as opposed to higher-grade neuroendocrine tumors.¹⁷⁷

Another differential diagnostic consideration in this specific context is metastatic grade I neuroendocrine NEC from a source outside the lungs. Srivastava and Hornick¹⁷⁸ immunohistologically compared pulmonary with nonpulmonary well-differentiated NECs, finding that the presence of thyroid transcription factor 1 was restricted to primary tumors of the lung. Classic pulmonary carcinoids also lacked reactivity for CDX2 and PDX1, both of which are alimentary tract–related markers.

Spindle cell growth in grade I NECs raises a dissimilar set of differential diagnostic considerations. These include fibrous and smooth muscle neoplasms, fibrous pseudotumors, primary or metastatic spindle cell melanomas, and metastatic medullary thyroid carcinomas. Careful attention to nuclear detail in spindle cell carcinoids shows retention of the typical stippled chromatin pattern, and nucleoli are small and inconspicuous. The lesions also express the majority of immunohistologic markers seen in central carcinoids, as discussed earlier. A combination of morphologic and adjunctive studies should be capable of excluding most of the differential diagnostic considerations mentioned earlier, but metastatic medullary thyroid carcinoma is virtually indistinguishable from grade I pulmonary NEC on pathologic grounds.¹⁶⁴ Likewise, Eyden and coworkers¹⁷⁹ have shown that metastatic melanomas sometimes exhibit partial neuroendocrine differentiation in ultrastructural or immunohistochemical studies.

A similar issue is the separation of tumorlets (Fig. 13-17) from peripheral carcinoids. This is an arbitrary distinction because immunohistochemical and ultrastructural studies indicate that the cells of these two lesions are essentially identical.^180–182 Also, there is debate over whether tumorlets represent true neoplasms or are instead hyperplasias of respiratory neuroendocrine cells.^15,181,183 They generally occur within and around small bronchioles, and they typically have a spindle cell composition. The small nests of tumor cells in a neuroendocrine tumorlet tend to be divided by a fibrous stroma into small packets, and this feature helps to differentiate neuroendocrine tumorlets from small peripheral carcinoid tumors, where tumor cell growth is more sheetlike, with variable fibrous stands interlaced. Although such tumorlets, which may number in the hundreds in some cases, are said to occur most often in diseased lungs with such associated lesions as bronchiectasis or pulmonary fibrosis,^180–184 they also may arise in otherwise normal lungs.¹⁶² Rizvi and colleagues¹⁸⁵ and Ferolla and colleagues¹⁸⁶ have also shown that neuroendocrine hyperplasia is much more common in patients with pulmonary NECs than in others with non-neuroendocrine neoplasms. In any event, an arbitrary size of 4 mm or less has been suggested for tumorlets; larger lesions are considered peripheral carcinoids.¹⁸¹ A continuum for such proliferations was suggested by Miller and Muller,¹⁸⁷ who found that peripheral carcinoid tumors were often associated with diffuse neuroendocrine cell hyperplasia and airway obstruction. A particular pitfall associated with tumorlets is that they may be misdiagnosed as much more aggressive lesions in small biopsy specimens, especially when clinical data are ignored or unavailable.^188,189

Figure 13-17 This neuroendocrine tumorlet (NT) has the same cytologic features as those of low-grade neuroendocrine carcinoma of the lung; however, NTs are microscopic, are associated with chronic airway disease, and tend to grow in packets separated by fibrous tissue (carcinoid tumors have a more sheet-like growth pattern).

Treatment and Outcome

The clinical outcome in cases of central grade I NECs of the lung is generally excellent.^5,11,45 Complete excision is the treatment of choice, possibly necessitating lobectomy or sleeve resection.¹⁹⁰ Conservative endobronchial excision is generally associated with an unacceptable rate of local recurrence.^190–193

A peripheral location or a spindle cell growth pattern does not, in and of itself, equate with “atypical” morphology in grade I NECs of the lung. The prognosis for low-grade spindle cell peripheral lesions is equivalent to that of their central relatives, despite a slightly higher rate of lymph node metastasis in some series.^45,151,152

Figures for the incidence of metastasis differ widely and depend on two major variables. These include the definition of peribronchial lymph node metastasis (i.e., direct invasion of nodes vs. embolic metastatic involvement) and the histologic purity of the primary lesion. Some studies with high rates of metastasis for classic “carcinoid” have improperly included higher-grade neuroendocrine lesions. Quoted metastasis rates vary from 1% to 20%; a figure of approximately 5% to 10% is probably closest to the actual incidence, and most secondary implants involve adjacent peribronchial or hilar lymph nodes.^5,192–196 It is this behavioral attribute that leads us, as well as others, to consider central pulmonary carcinoids irrefutable carcinomas.¹⁴³ Distant metastases also occur rarely, and there is a tendency for spread to the bones, liver, skin, or brain.^45,193,194 Interestingly, osseous metastases are typically blastic.¹⁹⁷ Flow cytometric measurement of DNA content in grade I NECs is not helpful in predicting their metastatic potential.^149,198

The overall survival rate for patients with grade I NEC of the lung is greater than 95% at 5 years.^{5,45,148,192–196,199} Metastatic disease may be palliated with interleukin or somatostatin analogs,^38,200 and surgical excision of secondary implants can be considered, given the slow growth of this tumor type.

Clinical Findings

“Atypical carcinoids” of the lung are usually greater than 3 cm in maximum dimension. Other characteristics include an etiologic link to cigarette smoking, differing from the epidemiologic features of grade I NECs.^5,201–216 Grade II NECs are more likely to be peripherally located in the lung (Fig. 13-18) compared with “classic” carcinoids; the former lesions have been so situated in the majority of cases in several studies.^201–216 Potential associations with Cushing syndrome²¹⁷ and Eaton-Lambert syndrome also apply to these lesions.

Figure 13-18 Grade II neuroendocrine carcinoma (atypical carcinoid) is usually peripherally located in the lung parenchyma and may contain areas of hemorrhage or necrosis, as seen in this gross photograph.

Pathologic Findings

With the exception of the possible gross identification of hemorrhage or necrosis and a tendency to be slightly larger, the lesions were difficult, if not impossible, to distinguish from typical carcinoids by clinical, radiographic, or gross pathologic evaluation (Fig. 13-19). The pathologic requirements for a diagnosis of “atypical carcinoid” have varied from study to study, but most have included mitotic activity and necrosis. We currently use the criteria of El-Naggar and colleagues¹⁴⁹ to define grade II NEC of the lung (Fig. 13-20), preferring that term to “atypical carcinoid,” but stipulating that the two are conceptually synonymous. Those requirements center on the following points:

Figure 13-19 Chest radiograph of grade II neuroendocrine carcinomas of the lung (atypical carcinoids) shows features that are essentially identical to those associated with grade I tumors.

Figure 13-20 A to C, Grade II neuroendocrine carcinomas exhibit organoid growth patterns but show notable mitotic activity and areas of spontaneous necrosis, unlike grade I tumors.

Spindle cell change in grade II NECs of the lung is potentially seen but uncommon.²¹⁵ As observed by Yousem and Taylor,^3,218 a requirement for at least two of these criteria is more reasonable than basing the diagnosis of grade II NEC on only one feature in a lesion that is otherwise a classic carcinoid of the lung.

Immunohistochemical studies show some differences between grade I NEC and grade II NEC of the lung. As expected of more poorly differentiated tumors, “atypical carcinoids” tend to express fewer markers of neuroendocrine differentiation or to demonstrate more focal labeling for them; such markers as synaptophysin, CGA, and CD57 are usually less impressive in grade II lesions.^{92,117,120,219,220} Reactivity for specific neuropeptides—such as bombesin, calcitonin, and adrenocorticotropic hormone (ACTH)—is relatively uncommon.²²¹ Mutant p53 protein is demonstrable by immunostaining in some atypical carcinoids, as opposed to its absence in the great majority of typical carcinoids but like its presence in most SCCs.^176,222,223

Ultrastructural analysis of grade II NEC shows fewer neurosecretory granules compared with grade I neuroendocrine neoplasms.²⁰⁷ Cytogenetic analysis also has demonstrated multiple karyotypic abnormalities in grade II NEC but not in classic carcinoid,^224–227 paralleling the flow cytometric incidences of aneuploidy in the two tumor types.^149,218,228

Differential Diagnosis

In addition to its distinction from other neuroendocrine lesions, the differential diagnosis of grade II NEC includes poorly differentiated non-neuroendocrine carcinomas of the lung.^44,50,53,210 In particular, selected examples of high-grade adenocarcinoma may simulate the microscopic configuration of “atypical carcinoids,” and “basaloid carcinoma” (basaloid squamous cell carcinoma)²²⁹ is regularly confused with grade II NEC. The nuclear characteristics of those lesions differ from one another in that non-neuroendocrine tumors do not manifest the dispersed chromatin seen in “atypical carcinoids,” instead showing more vesicular nuclei with often-prominent nucleoli (Fig. 13-21). Immunohistochemical studies may assist with the differential diagnosis in this context, but because some grade II NECs lack reactivity for endocrine markers and conversely selected non-neuroendocrine carcinomas of the lung demonstrate their presence, this technique is problematic. Electron microscopy is still probably the surest technique for differentiating grade II NEC from its non-neuroendocrine diagnostic simulators.

Figure 13-21 Basaloid squamous carcinoma of the lung is also composed of polygonal (A) and spindle-shaped (B) tumor cells, potentially like those of neuroendocrine carcinomas, but the former tumors have more vesicular nuclear chromatin and prominent nucleoli. Their immunophenotypes also lack positivity for endocrine markers, and no neurosecretory granules are present on electron microscopy.

Treatment and Clinical Outcome

A vexing problem is which diagnostic label to assign a neuroendocrine pulmonary tumor with only one indicator of potentially aggressive behavior, such as angiolymphatic invasion, brisk mitotic activity, aneuploidy, or large size. None of these features independently appears to warrant adjuvant therapy, but it is probably prudent to suggest that they may be associated with an adverse outcome and to monitor the patient carefully for recurrence. Special techniques may provide additional information. In a flow cytometric analysis by El-Naggar and colleagues,¹⁴⁹ approximately 80% of “atypical carcinoids” were aneuploid, whereas 80% of classic carcinoids were diploid. Also, grade II lesions were more likely to have an S-phase fraction of greater than 7%. Both of these factors were statistically linked to a worse outcome. However, in that analysis, the most significant predictor was accurate morphologic separation into grade I and grade II categories. Rush and colleagues,²⁰⁶ Jackson-York and coworkers,²²⁸ and Travis and colleagues²³⁰ concluded that accurate classification of these tumors is more important in the prognosis than DNA content. Other factors said to significantly decrease survival rates in cases of “atypical carcinoid” include lymph node metastases, vascular invasion, and overall tumor size of greater than 3 cm.^{5,148,201–209,210–216,230}

Lymph node metastases are found at diagnosis in 30% to 50% of cases, and approximately 25% of patients with grade II NEC of the lung have remote metastatic disease at presentation.^{5,215,216,231} As in cases of SCC, the brain is a common site for metastasis and recurrence. Importantly, the mortality rate for this neoplasm is 30% to 50% at 2 years.^{5,148,201–209,210–215} Rush and colleagues²⁰⁶ have reported 5- and 10-year survival rates of 60% and 40%, respectively. Obviously, these figures reflect a different biologic potential than those of classic carcinoid and SCC. The use of adjunctive therapy in the management of grade II NEC of the lung is unresolved.^232,233 Some researchers have suggested that chemotherapy, irradiation, or both can be beneficial in this setting, but there has been no consensus on that point, much less on which pharmacologic agents to use. The difficulty in securing diagnostically “pure” study populations for treatment evaluations has likely contributed to this uncertainty.

Clinical Features

Small cell carcinoma of the lung has many potential modes of clinical presentation. These include symptoms and signs similar to those of other common carcinomas of the lung, such as cough, hemoptysis, weight loss, anorexia, fatigue, and the syndrome of hypertrophic osteoarthropathy.^236,237 In addition, however, unusual findings, such as hyponatremia, hypercalcemia, or Cushing syndrome, may emanate from ectopic production by the tumor of antidiuretic hormone, parathyroid hormone-like peptides, and ACTH, respectively.^{155,159,236,238,239} Rarely, individuals with SCC manifest Eaton-Lambert (pseudomyasthenia) syndrome owing to paraneoplastic interference with neuromuscular function.^240,241 Retinopathy, central and peripheral neuropathies, encephalitis, glomerulonephritis, cutaneous reactions, sarcoid-like granulomatous lesions, and systemic vasculitis also have been reported in this context.^241–248 Massive hepatomegaly with liver failure,²⁴⁹ myelophthisic anemia,^250,251 or seizure activity and headaches²⁵² are additional clinical constellations related to the effects of distant visceral metastases at initial diagnosis.

Chest radiographic findings may be relatively unremarkable or may demonstrate a central hilar (or, more rarely, a peripheral intrapulmonary) mass (Fig. 13-22). Obstructive pneumonia is a potential complication; bulky peribronchial or mediastinal lymphadenopathy is common and may be massive.^161,236 Pleural effusions typically signify serosal metastases of SCC, and rarely, the tumor may so markedly involve the pleura that it produces a peripulmonary “rind” similar to that seen in association with mesotheliomas.²⁵³

Figure 13-22 A, Chest radiograph of small cell neuroendocrine carcinoma of the lung, showing a right hilar mass with mediastinal widening. B, Gross photograph of small cell neuroendocrine carcinoma. The tumor has a “fish-flesh” appearance, as is more often seen in lymphomas or sarcomas. C, This sputum cytology specimen demonstrates an aggregate of tumor cells from small cell neuroendocrine carcinoma on the left; their sizes can be contrasted with that of a benign squamous cell in the photograph.

Pathologic Findings

The gross features of SCC are such that the tumor bears more resemblance to a lymphoma than to other carcinomas. The cut section of the lesion is typically uniformly gray-tan and fleshy, with possible small areas of necrosis or hemorrhage.

Microscopically, SCC exhibits a spectrum of morphologic appearances.^{235,254–263} Classic “oat cell” tumors are relatively rare, accounting for only 10% to 20% of cases.^254,255 Oat cell morphology (Fig. 13-23), with bluntly fusiform, carrot-shaped cells, nuclear molding, and crush artifact, is more often seen in bronchial biopsy specimens or fine-needle aspiration biopsy specimens than in resected neoplasms or lymph node metastases. Tumor cell size is in the range of 1.5 to 3 times the diameter of non-neoplastic lymphocytes.²⁶³ The now defunct term “intermediate cell variant” of SCC (Fig. 13-24) accounts for approximately 75% of cases, and may be combined with oat cell areas.^256–259 Cell size in the former subtype is more variable, but may be as much as twice that seen in oat cell SCC. Nuclear molding and crush artifact are less conspicuous; the tumor cells also may show small nucleoli, but they maintain the marked hyperchromatism and granular chromatin that are seen in other forms of SCC. There may be blunt spindle cell change, and the cells may show pseudorosette formation or may grow in distinct ribbons. Other recognized types are combined small cell and large cell NEC^260,264,265 and mixtures of SCC with either adenocarcinoma or squamous carcinoma (Fig. 13-25).²⁷ All variations share the tendency to show prominent apoptosis, brisk mitotic activity, scant amphophilic cytoplasm, and the “Azzopardi phenomenon,” which is the accretion of basophilic nucleic acid around intratumoral blood vessels^254,255 (Fig. 13-26). Interestingly, D’Adda and coworkers²⁶⁶ found that both neoplastic components were genetically homologous in tumors showing a combination of SCC and large cell NEC.

Figure 13-23 A, Small cell carcinoma of the lung, engulfing a fragment of bronchial wall. B, Numerous mitoses and abundant apoptosis are seen in this small cell carcinoma.

Figure 13-24 “Intermediate” variant of grade III neuroendocrine carcinoma, small cell type. The tumor cells grow in a discernibly organoid fashion (A) and are more regular in shape and size (B) than those of the oat cell subtype. However, the distinction between histologic variants of small cell neuroendocrine carcinoma has no clinical importance and poor reproducibility. These variants have been eliminated from the current World Health Organization classification.

Figure 13-25 Divergent squamous differentiation is apparent in this small cell neuroendocrine carcinoma (left upper center).

Figure 13-26 The “Azzopardi phenomenon” is represented by accretion of densely basophilic, smudgy nucleic acid material adjacent to intratumoral blood vessels in small cell carcinoma.

In our opinion, differences in cytomorphology in SCC have little or no prognostic importance, based on a synthesis of the aggregated literature.^{236,256–259,267,268} Instead, it is necessary to be familiar with the microscopic subtypes to avoid misdiagnosis. Some variation in nuclear size, small nucleoli, and a modest amount of cytoplasm should not prevent one from making a diagnosis of SCC, recognizing that these features are seen regularly in the intermediate variant of SCC, especially in cytologic preparations.²⁶⁵ This is particularly true in fine-needle aspirate material.^{262,269–271}

Another difficulty that one may face is the identification of SCC in biopsy specimens of peripheral lung nodules, as seen in approximately 10% of cases.²⁷² In that setting, the fear among many pathologists is that the lesion may represent a lower-grade neuroendocrine lesion. This difficulty is best resolved by paying close attention to cytologic details. Marked nuclear hyperchromatism, brisk apoptosis, and scanty cytoplasm usually allow one to comfortably exclude a lower-grade lesion. Secondly, and more importantly, evidence has emerged that suggests that surgical therapy is appropriate for peripheral high-grade but low-stage NECs of the lung.^272–278 As stated earlier, rigid adherence to the dogma of “surgery for non–small cell carcinoma; no surgery for SCC” is probably improperly restrictive. Smit and colleagues²⁷⁵ studied 20 patients with resected SCC of stages I, II, and III. The median survival in that series was 29 months for stage I and II tumors and 20 months for stage III lesions. Another review of this topic was published by Mentzer and colleagues.²⁷⁴ Other contextual topics of debate include whether surgery should precede or follow chemotherapy or be accompanied by irradiation.^236,279,280 Nonetheless, we believe that sufficient data are available to suggest that all low-stage NECs of the lung should be resected, regardless of grade.

Another conundrum in fine-needle aspirates, limited biopsy specimens, or frozen sections is the distinction between SCC and high-grade large cell NEC of the lung.^281,282 The two forms of grade III may coexist in the same tumor mass, yielding the neoplastic variant known as “combined SCC.”^{263–265,283} As discussed earlier, the separation of these two forms of high-grade NEC has dubious significance, in our opinion, with regard to the suitability of surgical management. Use of the designation “high-grade NEC” is an equitable solution; whether the lesion is then excised should be decided by the clinical stage rather than by cytologic details.

Differential Diagnosis

The diagnosis of SCC in biopsy material involves the exclusion of lymphoid infiltrates as well as other types of carcinoma that may be composed of small basaloid cells.^{254,255,284–286} As stated previously, broadly reactive keratin antibody mixtures label essentially all SCCs, in many cases with a characteristic dot-like pattern of cytoplasmic staining. Hence, a simple two-antibody panel for cytokeratin and CD45 is effective in showing that a morphologically indeterminate small cell tumor is epithelial.⁵⁶ Extremely rare reports of apparent CD45 reactivity in SCC²⁸⁷ underscore the inadvisability of relying on one immunostain diagnostically. In our experience, a more difficult question is the exclusion of basaloid squamous cell carcinoma, particularly in a small biopsy specimen.²²⁹ The keratin labeling pattern described earlier may be helpful, and an extended panel of antibodies to CGA, neural cell adhesion molecule (NCAM; CD56), synaptophysin (Fig. 13-27), and CD57 can be applied as well. That set of reagents will identify approximately 80% of all SCCs.^56,219 Incidentally, we do not subscribe to the premise that there are “small cell neuroendocrine” and “small cell undifferentiated” carcinomas, but rather believe that some poorly differentiated tumors simply do not express overt neuroendocrine differentiation to a degree detectable by current methods.^56,288 Again, clinical stage is the best determinant of whether surgery is advisable, and in that framework, it becomes less crucial to separate SCC from basaloid squamous carcinoma in biopsy specimens.

Figure 13-27 Diffuse immunoreactivity for synaptophysin in grade III neuroendocrine carcinoma of the lung, small cell type.

Treatment and Clinical Outcome

Over the years, SCC has emerged as a clinically distinctive entity. As virtually all physicians are aware, long-term survival of patients with this neoplasm is rare; it is reported in fewer than 5% of cases in most centers,^{233,236,279,280,289,290} with rare exceptions.²⁹¹ A staging system centered on the terms “limited” and “extensive” SCC has been supplanted by American Joint Committee on Cancer staging, in which stages I, II, and III correspond to the old “limited” stages and “extensive” disease is stage IV.²³⁶ Staging is a powerful parameter and is virtually the only reliable indicator of clinical outcome for pulmonary SCC. As we have stated repeatedly, low-stage tumors are potentially resectable^272–278; chemotherapy with or without irradiation can be given postoperatively as “consolidation” treatment.²³⁶

Naturally, many factors have been investigated as possible prognostic indicators in SCC. As stated earlier, we do not believe that cytologic subdivision into “oat cell” and “intermediate cell” SCC is prognostically contributory, despite the claims of some other authors.²⁶⁸ Combined SCC (e.g., SCC admixed with squamous carcinoma, adenocarcinoma) and combined SCC (SCC admixed with large cell NEC) do, however, appear to be more refractory to therapy.^236,292 Vollmer and colleagues²⁶³ have suggested that an ultrastructural loss of desmosomes was associated with more aggressive behavior in SCC, but that observation has not been the focus of practical application. DNA ploidy analysis similarly does not seem to contribute significant information. Numerous cytogenetic abnormalities are now reported in SCC of the lung; one of the most common is a deletion of 3p,^293–296 suggesting that that area may harbor potential tumor suppressor loci. Other abnormalities that have commonly been seen include deletions of 5q, 9p, 11p, l3q, and 17p.^296,297 Proto-oncogenes implicated include c-myc, n-myc, myb, c-kit, c-src, and c-jun.^294,296 At least two tumor suppressor genes—p53 and the retinoblastoma gene—may play a role in the carcinogenesis of SCC.^25,294,296 As mentioned in the discussion of grade I and grade II NECs of the lung, the p53 gene is more often mutated in SCC than it is in lower-grade lesions. Other molecules, including p-glycoprotein (the multidrug resistance protein), CD99, and bcl-2 protein have likewise been investigated as possible prognosticators.^294,296 Duncavage and coworkers²⁹⁸ have shown that tumor cells in primary and metastatic SCC of the lung lack the Merkel cell polyoma virus, separating it biologically from primary SCC of the skin. Conversely, Ralston and colleagues²⁹⁹ found nuclear MASH1 immunoreactivity in more than 80% of cases of pulmonary SCC, but that marker was not apparent in primary cutaneous NECs.

Clinical Features

The clinical attributes of LCNCs are hybrids of those attending adenocarcinoma of the lung and SCC. In series by Travis and colleagues^5,230 and others,^{233,289,290,297,300–303} LCNCs almost always occur in heavy smokers, as does SCC. Symptoms and signs are generally most like those of non-neuroendocrine carcinoma; however, as in SCC, examples of Eaton-Lambert syndrome³⁰⁴ and paraneoplastic retinopathy³⁰⁵ also have been reported in connection with grade III large cell NEC. Although a minority of cases of LCNC present as central masses, they tend to be situated in the mid-to-peripheral lung fields (Fig. 13-28). Oddly, despite a high histologic grade, most LCNCs present as T1 or T2 tumors without lymph node metastases or evidence of systemic spread.^5,300

Figure 13-28 A, A large mass is seen in the lower right lung field, representing a large cell neuroendocrine carcinoma (LCNC). B, The gross appearance of LCNC is much like that of small cell neuroendocrine carcinoma (see Fig. 13-22B). C, Prominent infarct-like necrosis is a common finding in large cell neuroendocrine carcinoma of the lung.

Pathologic Findings

Grossly, pulmonary grade III NEC of the large cell type varies in maximum dimension from 1 to more than 10 cm and often demonstrates central necrosis, with or without dystrophic calcification.^5,297,300 Microscopically, it typically shows extensive coagulative necrosis that is obvious on low-power examination (see Fig. 13-28). At slightly higher magnification, one often sees an insular or ribboning growth pattern in the tumor between the necrotic zones. The individual neoplastic cells are larger than those of grade II lesions, with moderate-to-abundant amphophilic cytoplasm. In some instances, tumor cells are granular and eosinophilic; Chetty and coworkers³⁰⁶ have reported rare examples with “rhabdoid” cytologic features, in which globular hyaline cytoplasmic inclusions were apparent in the tumor cells, and Khalifa and colleagues³⁰⁷ documented an example of LCNC with divergent sarcomatoid (pleomorphic and spindle cell) differentiation. Nuclei in LCNC are more heterogeneous than those in SCC or grade II NEC; nucleoli are variable in prominence and nuclear chromatin is more heterogeneous, varying from granular to vesicular (Fig. 13-29). The mitotic rate in LCNC is brisk, usually measuring in excess of 10 division figures per 10 high-power microscopic fields, and sometimes being in the range of 50 to 100.⁵ All cases show reactivity for cytokeratin, and immunoreactivity for at least one neuroendocrine determinant is observed in almost all cases.^5,297 Nevertheless, we believe that a diagnosis of LCNC can be made confidently, even if all such markers are absent. In those instances, the cytomorphologic attributes of the tumor are so classically those of a neuroendocrine neoplasm that no interpretative doubt exists. Electron microscopy shows dense core granules in large cell grade III NEC.^5,230,302

Figure 13-29 A to C, Foci of geographic necrosis and relatively prominent nucleoli are seen in large cell neuroendocrine carcinoma (LCNC). D, This fine-needle aspiration biopsy specimen of LCNC shows nuclear molding by the tumor cells as well as dispersed chromatin.

Differential Diagnosis

These lesions may be challenging to recognize. In the past, examples of LCNC were probably assigned to one of three other diagnostic categories: “atypical carcinoid”; “non–small cell carcinoma, not further specified”; or “large cell undifferentiated carcinoma.” A distinction from grade II NEC can usually be made by paying attention to several features. First, the cells of LCNC tend to be larger, and there is more nuclear pleomorphism. Chromatin in “atypical carcinoid” tends to be granular and nucleoli are small³⁰⁸; in contrast, the chromatin of LCNC is not uncommonly vesicular, and nucleoli are often prominent. Nucleocytoplasmic ratios in grade II tumors are actually higher than those of LCNCs. Necrosis in the latter tumor type is generally extensive and infarct-like, in contrast to grade II NEC, in which more punctate or limited confluent necrosis is seen. Mitotic activity is another key feature in this diagnostic comparison. As outlined by Travis and colleagues,^5,230,308 LCNC has a high mitotic rate (>50/10 high-power fields). Accordingly, those authors suggested that cases with more than 10 mitoses per 10 high-power fields probably represent LCNC, whereas “atypical carcinoid” is more likely to show a figure of 10 or fewer. However, we prefer to assess the overall microscopic configuration of the lesion in making that distinction, rather than relying on one pathologic parameter. Despite that caveat, some examples of pulmonary neuroendocrine tumors are still encountered in which it is virtually impossible to assign a label of grade II NEC versus LCNC with certainty.

Differentiation of LCNC from non-neuroendocrine non–small cell carcinomas depends first and foremost on endocrine morphologic features, enumerated earlier. As referenced earlier, we believe that the low-power appearance of grade III large cell NEC is one of the most helpful indicators of the proper diagnosis. Extensive zonal necrosis, producing a “jigsaw puzzle” pattern, is often visible. Prompted by this feature, which, of course, can be mimicked by the appearance of necrotic squamous cell carcinoma or poorly differentiated adenocarcinoma, one then examines the lesion for organoid growth and neuroendocrine nuclear features. In contrast to our discussion of SCC, special techniques sometimes play an important part in the diagnosis of LCNC of the lung, and ultrastructural studies are useful in demonstrating neurosecretory granules.^5,309 By immunohistochemical analysis, all cases of grade III large cell NEC should demonstrate keratin reactivity and many are carcinoembryonic antigen-positive.^5,230,297 Virtually all LCNCs react with antibodies to NSE and CD56; among more “specific” markers, CGA is the most consistently detected antigen. Antibodies to CD57 and synaptophysin stain fewer cases. Rossi and coworkers³¹⁰ showed that a three-marker panel comprising synaptophysin, CGA, and CD56 was effective in separating LCNC from nonendocrine tumors; if any two of those determinants were seen, the diagnosis was secure.

This discussion raises several points. First, given the vagaries of immunohistochemistry and neoplasia, one could predict that cases will be encountered with morphologic features that are consistent with—but not diagnostic of—LCNC, but do not demonstrate any “neuroendocrine” markers. Electron microscopic examination can be used in such cases (Fig. 13-30A and B). It is advisable to provide some corroborating evidence of neuroendocrine differentiation before making a diagnosis of LCNC in this specific context, because some basaloid squamous cell carcinomas²²⁹ and other non-neuroendocrine large cell carcinomas may closely mimic LCNC.^26,34 Conversely, however, if no light microscopic features support the diagnosis of LCNC, immunostaining results and ultrastructural analysis should not be used to make the diagnosis. It has been well documented that generic non–small cell lung neoplasms may contain cells with endocrine characteristics,^26,34 as considered subsequently.

Figure 13-30 A, Large cell grade III neuroendocrine carcinomas of the lung may require ultrastructural study to document their endocrine nature. This example shows scattered neurosecretory granules (center). B, Immunoreactivity for chromogranin-A is present in this large cell neuroendocrine carcinoma (LCNEC) of the lung. C, This figure contrasts the survival of patients with LCNEC with that of those with nonendocrine lung cancers. D, This graph depicts the relative survival rates of patients with grade 1 (TC), grade 2 (AC), and grade 3 small cell neuroendocrine carcinoma (SCLC) and LCNEC neuroendocrine carcinomas. The data for SCLC and LCNEC are similar.

Even though they are both forms of grade III NEC, LCNC and SCC can usually be distinguished from one another readily in surgical material. The fact that they are different tumor entities is supported by the observations of Ullmann and colleagues,³¹¹ showing that LCNC and SCC have dissimilar genotypes. The larger cell size, polygonal shape, lower nucleocytoplasmic ratios, nucleolation, and more irregular chromatin in LCNCs should make this distinction relatively straightforward in most cases. Nevertheless, the two cytologic forms of high-grade NEC occasionally coexist in “combined SCCs.”^35,263,264 Moreover, in cytologic material, the cited distinction between SCC and LCNC is sometimes very challenging.³¹² Yang and coworkers²⁶⁹ have described three potential images of LCNC in fine-needle aspiration specimens, one of which closely simulates SCC. The presence of prominent nucleoli is a helpful clue to the recognition of LCNC in that setting.^269,270

Treatment and Clinical Outcome

The aggressive behavior of grade III large cell NECs of the lung cannot be overstated^313–317 (see Fig. 13-30C and D). Of 10 patients studied by Warren and colleagues,¹² only 1 was alive at 2 years’ follow-up, despite the fact that all of the cases were stage I (T1 or 2/N0/M0). All of the patients in series by Travis and colleagues^5,230 had either died of their tumors or were likely to do so at the same time point. Another report by Rush and coworkers²⁰⁶ quoted respective 5- and 10-year survival rates of 33% and 11% for LCNC. Dresler and colleagues³⁰⁰ studied a series of 40 non–small cell NECs, including 23 LCNC as defined here. The survival rate for stage I cases in that series was 18% at 40 months, significantly worse than that for those with similarly staged adenocarcinomas or squamous carcinomas. Surprisingly, the survival of patients with resected low-stage SCCs,^272–278 which are usually peripheral, is better than that of patients with LCNC. Likewise, the latter tumor type is more aggressive than “atypical carcinoid.”^{5,206,230,300}

Travis and colleagues⁵ and Rush and co-workers²⁰⁶ performed flow cytometric DNA analysis on cases of grade III large cell NEC of the lung. Both groups found no prognostic value in those studies.

The optimal therapy for these lesions is still in evolution, in large part because there has been a frustrating tendency for both pathologists and clinicians to push LCNC inappropriately into generic non–small cell carcinoma treatment protocols rather than evaluating them as a separate tumor entity. Because LCNCs tend to present as low-stage lesions,300–303,³¹⁸ most can—and should—be surgically resected.^{315–317,319,320} However, there have been few randomized trials to address the value of specific chemotherapy regimens or irradiation. In most patients with grade III large cell NEC, therapy appears to fail, with either distant metastases or intrathoracic recurrence occurring.^{5,230,297,300} Therefore, the use of adjunctive treatment modalities would appear to have merit.^315–317 Nonetheless, accrued experience suggests that LCNCs respond relatively unimpressively to standard chemotherapy regimens used for SCC.³²¹ With selected exceptions, tumors representing “combined SCCs” (SCC/LCNC in the same tumor mass) have also had only a modest response to such treatment.^{35,236,263,264} In that regard, it is intriguing that expression of the multidrug resistance protein seems to be relatively common in grade III large cell NECs.³²² Unfortunately, pulmonary NECs do not appear to manifest mutations in the epidermal growth factor receptor gene. Therefore, they are not likely to respond to epidermal growth factor receptor inhibitors.³²³ Faggiano and colleagues³²⁴ have found that a high mitotic count (>10 mitoses/high-power [×400] microscopic field), an absence of immunohistologic neuroendocrine markers, and an immunohistochemical bcl-2/bax ratio of greater than 1 were adverse prognosticators for LCNC at a pathologic level of evaluation.

Clinical Findings

Virtually any carcinoma in the lung can potentially have a neuroendocrine component, which can be grade I, II, or III NEC. Hence, there are no specific clinical features that can be ascribed to “combined” neuroendocrine/non-neuroendocrine tumors. Nonetheless, mixtures of squamous carcinoma or adenocarcinoma with NEC are most commonly seen in biopsy specimens that are taken after therapy,³³⁹ suggesting that effective treatment of the neuroendocrine cell population may allow for “overgrowth” of a minor nonendocrine component.

Pathologic Features

Unlike poorly differentiated carcinomas in which neuroendocrine differentiation is “occult,” and therefore discernible only by application of immunohistology or electron microscopy,³⁴⁰ the tumors considered in this section are recognizable as mixed by conventional microscopy. All of the neuroendocrine tumor types (i.e., grade I NEC, grade II NEC, small cell or large cell grade III NEC) that have been discussed up to this point may be components of combined carcinomas in the lung.

Special studies are likely to lead to a spectrum of findings. Some tumors demonstrate truly bifid differentiation—exhibiting, for example, true glandular and neuroendocrine features in the same neoplastic cells.^325,327 Others display admixtures of cellular elements with mutually exclusive ultrastructural or immunophenotypic properties. These differences do not appear to have any meaning from mechanistic or clinical points of view. There are discussions in the literature about whether divergent neuroendocrine lesions are “collision” tumors.³²⁵ Because virtually all neoplasms derive from transformed stem cells, regardless of anatomic site, it would appear much more logical to conclude that divergent growth simply emanates from dissimilar (and largely unknown) post-transformational modulators of differentiation.^341,342

Differential Diagnosis

Because of the unique histologic appearance of combined carcinomas with neuroendocrine elements, differential diagnostic considerations are limited. These generally concern making a distinction between poorly differentiated NEC components and high-grade portions of “pure” adenocarcinomas or squamous cell carcinomas with variable microscopic patterns that simulate neuroendocrine differentiation.²⁸⁴ Ultrastructural and immunohistochemical markers of neuroendocrine differentiation should be pursued if that question arises.

Treatment and Clinical Outcome

Combined neuroendocrine/non-neuroendocrine carcinomas generally have more adverse prognoses than histologically “pure” tumors,^{27,236,292,337,343} including lesions with grade I NEC components, which are very rare.³³⁸ Thus, therapy must be chosen to address all of the cellular elements in these composite neoplasms, but with the expectation that response to treatment will likely be blunted and survival will be worse than that of patients with pathologically homogeneous lesions.

Clinical Findings

Pulmonary carcinomas with occult neuroendocrine differentiation (OND) are not substantially different clinically than “pure” non-neuroendocrine malignancies of the lung in the symptoms and signs they produce. The only exceptional aspect of lung cancer with OND is the occasional paraneoplastic phenomenon that can be ascribed to the production of an endocrine substance. For example, we have observed several examples of poorly differentiated adenocarcinoma of the lung associated with watery diarrhea–hypokalemia syndrome, in which immunoreactivity for vasoactive intestinal polypeptide was found in the tumor cells. Similarly, hypercalcemia or hypercortisolism may relate to ectopic production of parathyroid hormone–related peptides or ACTH by such neoplasms.

Pathologic Features

As discussed earlier, pulmonary squamous cell carcinomas and adenocarcinomas with OND are no different grossly or histologically than their counterparts that lack endocrine features.^{340,345–347,357} With specific reference to large cell “anaplastic” carcinomas of the lung with OND,^{26,34,341,350,351} the tumor cells are at least twice as large as those of SCC. By definition, those lesions lack overt glandular or squamous differentiation and have rather monomorphic nuclei with dispersed or vesicular chromatin and prominent nucleoli. Small foci of necrosis may be seen, but these lesions lack the infarct-like configuration seen in LCNC. Similarly, manifestations of organoid growth (e.g., insulae, ribbons, cords, rosettes) are absent in large cell carcinomas with OND. Still other tumors with occult neuroendocrine elements may show more unusual histologic patterns, such as sarcomatoid differentiation²⁷³ or a blastomatous configuration (e.g., as in “well-differentiated adenocarcinoma simulating fetal lung”).^358,359

There is some debate over the “best” method to document OND in lung carcinomas.^59,360 Some observers argue that ultrastructure should be the standard technique,⁵⁹ but as discussed earlier, sampling problems interfere with the reliability of that procedure. Putative nonspecificity of neuroendocrine determinants, such as chromogranin and synaptophysin, is not a problem in our view, because we believe that publications that raised such concerns⁵⁹ contained conceptual flaws. They were based on electron microscopy as the “validating” technology, ignoring the effects of sampling just cited. In our view, positivity for either of these two immunomarkers (Fig. 13-32) reliably predicts neuroendocrine differentiation. With reference to the incidence of neuroendocrine reactivity in nonendocrine carcinomas, Sørhaug and coworkers³⁶¹ have posited that it is steadily increasing as immunohistochemical techniques become more sensitive.

Figure 13-32 “Occult” neuroendocrine differentiation in adenocarcinoma of the lung, manifested by immunoreactivity for chromogranin-A.

Differential Diagnosis

The differential diagnosis of squamous cell carcinomas or adenocarcinomas with OND principally includes similar tumors that lack endocrine features. On the other hand, large cell carcinoma with OND must be distinguished from both large cell “undifferentiated” lung carcinoma and LCNC, primarily by electron microscopy and immunohistology.³⁴ In addition, large cell carcinomas may be sufficiently nebulous morphologically that metastatic melanoma, poorly differentiated sarcoma with epithelioid features, and large cell lymphoma enter into diagnostic consideration. Again, adjunctive pathologic studies are usually required.³⁶²

Treatment and Clinical Outcome

Some investigators have concluded that neuroendocrine differentiation justifies the use of modified therapeutic approaches, such as those employed in SCC, whereas other authors have demurred on that point.^363–367 Similarly, there is no consensus as to whether any prognostic value may be derived from the identification of “occult” neuroendocrine differentiation in lung cancers of various pathologic types, with contradictions in published studies.^365–369 Currently, all that can be said with confidence is that surgical excision is still the foundation of treatment in low-stage cases.^34,367

Clinical Features

In sporadic cases of PG in the lung and elsewhere, men predominate and usually come to diagnostic attention in middle life (40–50 years of age).^370–380 In contrast, women outnumber men in the context of the MEN type 2 syndromes, and they are recognized as having a PG approximately 15 years earlier.³⁷⁰ The latter observation may simply reflect the fact that family members in kindreds with MEN are usually regularly screened for constituent tumors from childhood onward.³⁸² PGs have been described in virtually all organ sites, including the orbits, nasal cavity, thyroid, heart, urinary bladder, gallbladder, liver, biliary system, kidneys, prostate, urethra, spermatic cord, uterus, ovaries, vagina, vulva, cauda equina, and lungs; primary pulmonary tumors are among the rarest.³⁷¹

Functionality of PGs in these diverse locations is sporadic. Biosynthetic tumors may present with episodic or sustained hypertension; hypertensive crisis or “malignant hyperthermia” on induction of general anesthesia; episodic nausea, weakness, cardiac arrhythmias, pallor, flushing, headache, diaphoresis, anxiety, or localized pain; or cardiovascular decompensation with heart failure. Reflections of neuropeptide production may include Cushing syndrome with PGs that synthesize ACTH or watery diarrhea–hypokalemia complex (Verner-Morrison syndrome) with neoplasms that manufacture vasoactive intestinal polypeptide. Occasional examples also have apparently produced a parathyroid hormone–related peptide, with associated hypercalcemia, or an erythropoietin-like moiety in linkage with paraneoplastic polycythemia. Finally, rare patients with PG and associated systemic abnormalities may have von Recklinghausen disease (neurofibromatosis) or Beckwith-Wiedemann syndrome (hemihypertrophy and macroglossia).³⁷⁰

Nonfunctional tumors become manifest only through the appearance of a steadily enlarging mass, with other symptoms and signs dependent on the anatomic location of the lesion. In the lung, PGs that impinge on large airways produce symptoms related to obstruction (e.g., cough or stridor), but peripheral nonfunctional lesions are typically found incidentally on screening chest radiographs^371–380 (Fig. 13-33). Rarely, multiple synchronous intrapulmonary PGs may be encountered,³⁸¹ simulating metastases on chest radiographs.

Figure 13-33 A, Chest radiograph of intrapulmonary paraganglioma, represented by a right apical mass (arrow). B, Computed tomography of the chest; the lesions represent synchronous multifocal paragangliomas.

Pathologic Features

Grossly, PGs are typically spherical or slightly lobulated masses that range from a few millimeters to several centimeters in greatest dimension. They may be either centrally or peripherally located in the lung. Their cut surfaces are bloody in most cases because of dense intralesional vascularity, and the tumor tissue itself may be gray, pink, lavender, brown, or mottled (Fig. 13-34). Characteristically, immersion of fresh tissue in Bouin’s fixative or another picric acid–containing solution causes the specimen to assume a brownish appearance. In general, PG is a circumscribed lesion with a partial or complete fibrous capsule; hence, the surgeon will report that the lesion was relatively easy to dissect from contiguous structures. However, approximately 10% to 20% of tumors demonstrate local infiltration of adjacent tissues,³⁷¹ and these may be submitted to the surgical pathology laboratory in a fragmented state or with adherent structures attached to their peripheral aspects. An important step in the initial pathologic evaluation of PG is not only to perform standard three-dimensional measurement of the lesion but also to weigh it after dissection of attached extraneous soft tissue. Secondly, one should pay special attention to whether a PG appears to be multinodular or multifocal. Multicentricity of such neoplasms correlates well with a syndromic association.

Figure 13-34 Gross photograph of a resected intrathoracic paraganglioma, showing a relatively nondescript reddish-yellow lobulated tumor.

The histopathologic characteristics of pulmonary PGs are also variable. The peculiar nesting configuration of the tumor cells, separated by prominent fibrovascular stromal septa (“zellballen”; Fig. 13-35) is poorly developed in many of these lesions. In the lung in particular, this feature leads to considerable difficulty in distinguishing PG from carcinoid tumors (which, of course, are far more common). Aside from the organoid growth pattern of PGs, they are marked by a tendency toward nuclear pleomorphism, the common presence of intranuclear “pseudoinclusions” (invaginations of cytoplasm), intercellular hyaline globules, accumulations of intercellular proteinaceous material resembling thyroid colloid, potential spindle cell or oncocytic change, and elements resembling ganglion cells^383,384 (Fig. 13-36). Mitotic figures are seen in approximately 45% of benign PGs and 65% of malignant lesions, regardless of location; hence, they are not useful in and of themselves in predicting tumor behavior. Similarly, although vascular invasion is apparent in one-fifth of malignant PGs, it is also evident in 5% to 6% of benign tumors.³⁸⁴ Thus, requests for a definitive diagnosis of PG in the frozen section laboratory are impossible to satisfy, particularly with reference to pulmonary tumors.

Figure 13-35 The formation of distinct cell groups (“zellballen”) may be conspicuous (A) or relatively vague (B) in paraganglioma. C, A reticulin stain highlights the organoid growth pattern of paraganglioma.

Figure 13-36 Intranuclear pseudoinclusions (A) and nuclear pleomorphism with eosinophilic cytoplasmic inclusions (B) in paraganglioma.

Relatively few PGs have been subjected to fine-needle aspiration and cytologic assessment.^385–387 However, a report on this topic by Gonzalez-Campora and colleagues³⁸⁶ showed that such tumors commonly exhibited marked anisokaryosis, a tendency to form acini or follicles, and intranuclear invaginations of cytoplasm similar to those seen in papillary thyroid carcinoma. Nuclear pleomorphism has not been correlated with adverse behavior; in fact, malignant PGs have tended to display less nuclear variability than did their benign counterparts.³⁸⁴

Differential Diagnosis

Special studies are typically needed to bolster the histologic diagnosis of PG in visceral locations, especially the lung. Other considerations in pulmonary cases center on NEC as well as metastatic malignant PG originating at another site. Electron microscopy is still a useful modality in this context, and the neurosecretory granules of paraganglion cell tumors generally differ from those of other neuroendocrine cells and neoplasms.³⁷⁰ They feature an unusual “blister”-like configuration, where the internal submembranous “halos” of the granules are obviously eccentric and appear to emanate from the dense core in a bubble-like fashion. Otherwise, the cells of PGs are similar to those of the dispersed neuroendocrine system, showing prominent Golgi complexes and rough endoplasmic reticulum, macular intercellular junctions, and incomplete pericellular basal lamina.³⁸⁸

Immunohistologically, most examples of PG have a distinctive intermediate filament protein profile that is not shared by other neoplasms except for NBs.^76,389 This features neurofilament protein, with or without vimentin, to the exclusion of other intermediate filament proteins. Neurofilament protein may be difficult to demonstrate in formalin-fixed, paraffin-embedded tissues, so in practical terms, the majority of routinely processed PGs do not manifest detectable intermediate filament protein.³⁸⁹ This observation is useful in making the distinction between PG (and other “type 2” [neural] neuroendocrine tumors) and “type 1” (epithelial) neoplasms that often are considered in the differential diagnosis and that uniformly exhibit keratin positivity.⁵⁶ We have not found keratin proteins in PGs, in the lung or elsewhere. This is in stark contrast to the rate of 30% that has been cited by some authors for keratin positivity in PG⁶⁶—a rate we believe to be a reflection of procedural shortcomings. Occasional examples of PG may also show focal immunoreactivity for glial fibrillary acidic protein, but they uniformly lack desmin.⁷⁶ Virtually all PGs are diffusely and intensely positive for chromogranin-A and synaptophysin, whereas beta-tubulin and microtubule-associated protein are only focally seen in such lesions and are instead characteristic of neuroblastic and ganglioneuromatous neoplasms.^56,83,89,90 Other immunodeterminants that can be detected in PGs include ACTH (approximately 30%), vasoactive intestinal polypeptide (40%), Leu- or Met-enkephalins (50% to 60%), calcitonin (<5%), CD56 (90%), CD57 (50%), and beta-endorphin (10%).^75,79,93

A word is in order regarding “minute pulmonary chemodectoma” as a differential diagnostic consideration in this context. “Chemodectoma” is a term that was formerly used in reference to PG.³⁷⁰ In fact, minute pulmonary chemodectomas have no relationship to the paraganglion system. They are small peripheral pulmonary parenchymal aggregations of polygonal or bluntly fusiform cells, often with a concentric configuration. Immunohistochemical studies of minute pulmonary chemodectomas have demonstrated a similarity to meningothelial rather than neuroendocrine tissues, with reactivity for vimentin and epithelial membrane antigen.³⁹⁰ Moreover, analyses of cellular clonality in such lesions have demonstrated that they are polyclonal and probably reactive,³⁹¹ rather than neoplastic, as is true of PGs.

One other recent development should be mentioned in reference to PGs. It has now been shown conclusively that both familial (MEN2-related) and selected sporadic examples of this tumor demonstrate mutations in the ret gene on chromosome 10.³⁸² These take the form of (Cys634 → Arg) in MEN2A and (Met918 → Thr) in MEN2B. Whether this information enters the diagnostic sphere in the near future must await further technical developments and clinical correlation.

Treatment and Clinical Outcome

The most contentious aspect of PGs is the prediction of their often-capricious behavior. In the lung, the great majority of primary paraganglionic tumors have been benign biologically, but occasional examples have metastasized to regional lymph nodes or other viscera.^371–380

At one extreme, some authors have stated that a diagnosis of malignant PG can only be made after metastasis has occurred; others have claimed that other findings can be correlated with aggressive behavior. These include pathologic mitotic figures or vascular invasion,³⁸⁴ decreased immunoreactivity for selected neuropeptides (particularly neuropeptide Y),³⁹² and loss of intratumoral sustentacular cells positive for S-100 protein.^130,393 These concepts are admittedly still in evolution. Some reports concerning the biology of PG have provided additional useful information. Based on the results of logistical regression analyses, mitotic activity, nuclear atypia, and vascular or capsular invasion are of little or no use as individual parameters in the prognostication of PGs.³⁸⁴ Similarly, Lack,³⁷⁰ Linnoila and colleagues,³⁸⁴ and Gonzalez-Campora and colleagues³⁹⁴ found no statistically significant association between static or flow cytometric DNA aneuploidy and behavior in paraganglion cell tumors, with only Pang and Tsao³⁹⁵ demurring on the latter point. Multiparametric assessment of 16 nonmicroscopic and histologic features by Linnoila and associates³⁸⁴ showed that 4 of them were the most predictively useful. These included extra-adrenal location, coarse gross nodularity of the tumor, confluent tumor necrosis on microscopic examination, and absence of intercellular hyaline globules. Among 120 PGs in that series, 71% of the biologically malignant lesions showed two or three of the four specified features, whereas 89% of the benign tumors manifested zero or one of them. Accordingly, there was a greater than 95% probability that more than 70% of PGs could be correctly classified using this paradigm.³⁸⁴ In our opinion, such an approach is recommended. It can be used by any pathologist and does not require special equipment.

Clinical Features

Symptoms and signs of these neoplasms typically relate to the presence of an enlarging mass (Fig. 13-40). In the lung, they are therefore most closely allied to interference with the function of structures on which the lesions impinge, such as the major bronchi. In unusual instances, paraneoplastic phenomena such as those seen with more differentiated autonomic neural tumors—i.e., PGs—may be seen at presentation in association with NBs.^411–413 In one case of pulmonary ganglioneuroblastoma reported by Hochholzer and associates,⁴¹⁰ signs of a MEN syndrome were also observed.

Figure 13-40 A, An apical right pulmonary mass is present in this 4-year-old boy, representing a neuroblastoma. B, Gross photograph of neuroblastoma, demonstrating a homogeneous, “fleshy,” white-tan cut surface.

Pathologic Findings

The gross pathologic attributes of neuroblastic tumors are variable. Undifferentiated NB has a prototypical gray-white, encephaloid appearance. In lesions with partial ganglionic differentiation, nodules of more “fleshy” tissue are noted as discrete foci in the cut surfaces of the lesion. Other potential gross features of neuroblastic neoplasms include a hemangioma-like image because of extreme intratumoral hemorrhage, extensive cystic change, and diffuse or localized calcification that is often readily apparent on sectioning the mass.⁴¹⁴

The microscopic characteristics of these tumors likewise form a continuum.^414–418 At one pole, one encounters classic undifferentiated small round cell tumors with little or no discernible stroma and no attempts at neural rosette formation (Fig. 13-41). Further along the spectrum of differentiation, the next group of NBs begins to exhibit background “neuropil,” a fibrillary eosinophilic matrix that represents the elaboration of numerous cytoplasmic extensions by the tumor cells (Fig. 13-42). Often, Homer Wright rosettes—having a fibrillary nonluminal center—are also encountered. The separation between NB and ganglioneuroblastoma requires definable ganglion cell differentiation in the latter neoplasms, but determining where they begin and “differentiating NB” ends is more art than science. This statement is limited to “diffuse ganglioneuroblastomas” (“differentiating stromal-poor NBs” reported by Shimada and colleagues⁴¹⁸) because, as noted earlier, stromal-rich tumors look most like ganglioneuromas (composed of mature ganglion cells and spindled Schwann cell–like elements) rather than predominantly small cell proliferations. Stromal-rich neoplasms that are not nodular (i.e., composite ganglioneuroblastoma) are further subdivided into “well-differentiated” lesions, with only a few randomly dispersed neuroblastic cells punctuating the image of a ganglioneuroma and “intermixed” tumors that contain small nests of neuroblasts having sharp interfaces with the surrounding ganglionic/Schwannian tissue.

Figure 13-41 Solid, undifferentiated growth of small round anaplastic tumor cells in neuroblastoma. This appearance is similar to that of high-grade neuroendocrine carcinoma.

Figure 13-42 Differentiating neuroblastoma, exhibiting the formation of fibrillar eosinophilic cytoplasmic extensions (neuropil).

Rarely, primitive NBs exhibit a striking degree of nuclear pleomorphism; those lesions have been termed “anaplastic NBs.”⁴¹⁹ Whether the pleomorphic appearance of such neoplasms correlates with a worse prognosis is uncertain.

Differential Diagnosis

The differential diagnosis with other small round cell tumors is principally a problem in “undifferentiated” NB.⁴¹⁴ In that setting, electron microscopy demonstrates characteristically elongated cell processes containing microtubular complexes in neuroblastic tumors, with or without presynaptic vesicles or dense-core granules as well⁴²⁰ (Fig. 13-43).

Figure 13-43 Electron photomicrograph of neuroblastoma, demonstrating interdigitating cytoplasmic extensions, each of which contains microtubules (A), synaptic-neurosecretory granules (B), or both. This constellation of findings would not be expected in epithelial neuroendocrine neoplasms.

Immunohistologically, most neuroblastic neoplasms express only neurofilament protein among all of the intermediate filament types. Because that marker is usually not well preserved in formalin-fixed tissues, no intermediate filaments are demonstrable in the majority of cases.⁴²¹ In contrast, SCC, an important consideration in adults, is uniformly keratin-positive.^63,64 Synaptophysin is usually present in NB, ganglioneuroblastomas, and ganglioneuromas,^89,90 but CGA is uncommonly observed in those neoplasms.⁸³ Neuron-specific enolase and CD56 are sensitive markers for NB (and may have limited utility in identifying bone marrow micrometastases⁴²²), but they have a lack of specificity and may also be seen in PNETs as well as some cases of rhabdomyosarcoma.¹¹² NB84 is another marker showing selective reactivity with NB and PNET, to the exclusion of other small cell tumors⁴²³ (Fig. 13-44). The once-difficult distinction between NB and PNET has been facilitated by the availability of antibodies to CD99 and beta-2-microglobulin, both of which are seen in PNET but not NB.³⁶² Likewise, immunoreactivity for muscle-specific actin or desmin characterizes rhabdomyoblastic neoplasms and is not expected in NB.³⁶² This observation principally comes into play in examples of “anaplastic” NB that may resemble embryonal rhabdomyosarcoma on conventional microscopy.⁴²⁴

Figure 13-44 Immunoreactivity is present in this neuroblastoma with the antibody known as NB84.

Treatment and Clinical Outcome

With reference to the three reported cases of primary intrapulmonary NB, two patients were alive and well 1 and 2.5 years, respectively, after surgical resection and chemotherapy.^410,411 The other patient died shortly after admission to the hospital.

In much more general terms, two histologic parameters with prognostic significance should be additionally recorded in evaluating neuroblastic neoplasms. These include the level of ganglionic differentiation (<5% or ≥5%)^414–418 and the “mitotic-to-karyorrhectic index” (MKI).^418,425,426 The MKI is the number of mitotic or karyorrhectic nuclei counted in an evaluation of 5000 tumor cells in “neuropil-free” areas of any given tumor. The daunting prospect of assessing that many cells has dissuaded many pathologists from embracing the Shimada system with enthusiasm. However, common practice has shown that a reasonable estimation of the MKI can be obtained with simple pattern-matching approaches. With data on differentiation and MKI, prognostic substratification of stromal-poor neoplasms can be accomplished within appropriately age-, location-, and stage-matched tumor groups.⁴²⁷

Accurate classification and prognostication of neuroblastic tumors mandates the availability of detailed data on clinical, gross, and histologic levels. Moreover, biochemical observations—as derived from the clinical chemistry laboratory—may provide additional nuances in this context.⁴²⁸ The presence of vanillacetic acid in the urine (as opposed to homovanillic acid or vanillyolmandelic acid) worsens the clinical outlook, as do elevated serum levels of ferritin, lactate dehydrogenase, NSE, CGA, and creatine kinase isozyme BB.⁴²⁸ On the other hand, elevated somatostatin levels in serum or tumor tissue have been linked to a better prognosis.⁴²⁹

With regard to other adjunctive modalities of pathologic evaluation as applied to neuroblastic tumors, three have been used increasingly in the last decade. Cytogenetic assessment (requiring submission of fresh tissue, although current fluorescence–in situ hybridization methods may change that) has shown aberrations (deletions and rearrangements) in chromosome 1p in most cases of NB and ganglioneuroblastoma,⁴³⁰ possibly aiding in the differential diagnosis with other small cell tumors that lack such abnormalities. Prognostically, aneuploidy, hyperdiploidy, and near-tetraploidy in neuroblastic neoplasms correlate with improved prognosis, in contrast to the norm for most other malignant tumors.^414,425 Conversely, DNA diploid lesions tend to behave aggressively. Likewise, increased copy numbers of the oncogene N-myc are also biologically detrimental in NBs.^431–434 This marker can be assessed directly by the Southern blot method or indirectly by Northern/Western blot or in situ hybridization. Immunohistology for the protein product of N-myc can also be performed on fresh tissue, with generally acceptable results. The RET gene, which is important in the prognosis of other neuroendocrine tumors, also appears to correlate with neuronal differentiation in NBs and, therefore, roughly with prognosis.⁴³⁵ Lastly, the immunolabeling index for Ki-67 (an indicator of cell replication) is inversely correlated with the length of survival in NB⁴³⁶; further, loss of expression of the Trk-A gene or CD44 by the tumor cells is associated with high stage at diagnosis and a poor outcome.⁴³⁴