Thoracic Service, Departments of Surgery, Pathology, and Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, New York

Transcription

1 Molecular Markers Help Characterize Neuroendocrine Lung Tumors Valerie W. Rusch, MD, David S. Klimstra, MD, and Ennapadam S. Venkatraman, PhD Thoracic Service, Departments of Surgery, Pathology, and Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, New York Background. The terms large cell and mixed small large cell neuroendocrine carcinoma (LCNC, MNC) have been proposed to describe distinct types of high-grade neuroendocrine lung tumors. However, cytologic appearance and neuroendocrine immunohistochemical stains cannot uniformly distinguish these from other neuroendocrine tumors, such as typical and atypical carcinoids or small cell carcinoma, or nonendocrine lung cancers such as large cell undifferentiated carcinoma. This study sought to determine the patterns of expression in LCNC and MNC of several molecular markers often abnormally expressed in lung cancers. Methods. Primary lung tumors with neuroendocrine features operated on between 1984 and 1994 were reviewed and classified as typical carcinoid (TC), atypical carcinoid (AC), LCNC, MNC, and small cell lung cancer (SCLC) based on mitotic rate, extent of necrosis, and cytoarchitectural features. Immunohistochemistry was performed using antibodies MIB-1 for Ki67, pab1801 for p53, OP-66 for Rb, 31G7 for EGFR. Staining was scored as 0to4 (0 less than 5%, 1 5% to 20%, 2 20% to 50%, 3 50% to 80%, 4 more than 80%) for p53, Ki67, and EGFR; and negative, focal, or positive for Rb. Overall survival was calculated by the Kaplan-Meier method and prognostic factors compared by log rank test. Results. Ninety-two tumors were examined: 25 TC, 7 AC, 24 LCNC, 18 MNC, 18 SCLC. The LCNC and MNC presented more frequently as stage II or III tumors (n 28, 66%) than TC and AC (n 5, 15%). Median survival for LCNC and MNC was 18.7 months, for SCLC 14.3 months, and has not been reached for TC and AC tumors. TC and AC tumors were uniformly characterized by low proliferative rate, absent p53, and normal Rb staining. LCNC, MNC, and SCLC showed a high proliferative rate, abnormal p53, and absent Rb staining. Overexpression of EGFR was frequent in all five tumor types. Conclusions. (1) Ki67, p53, and Rb help distinguish LCNC and MNC from TC and AC. (2) Small numbers of patients preclude comparison of survival rates, but LCNC/MNC have a median survival similar to comparable early stage SCLC, and clearly worse than TC/AC. These results justify a sharp separation of high-grade neuroendocrine tumors from carcinoids, and suggest a close relationship between LCNC, MNC, and SCLC. (Ann Thorac Surg 1996;62: ) Neuroendocrine lung tumors exhibit a spectrum of histology and biological behavior ranging from well-differentiated, indolent typical carcinoid tumors, to histologically high-grade, biologically aggressive small cell lung cancers. In between these two well-defined extremes exists a broad group of less differentiated tumors that have been variably termed atypical carcinoids [1], well-differentiated neuroendocrine carcinomas [2, 3], or Kulchitzky cell carcinomas [4, 5]. Recently, the terms large cell neuroendocrine carcinoma and mixed small cell large cell neuroendocrine carcinoma have been proposed to designate high-grade tumors intermediate between atypical carcinoids and small cell carcinoma. Travis and colleagues [6] described large cell neuroendocrine carcinomas as tumors composed of large cells characterized by a light microscopic neuroendocrine appearance with a low nuclear to cytoplasmic ratio, frequent nucleoli, a high mitotic rate (greater than 10 Presented at the Thirty-second Annual Meeting of The Society of Thoracic Surgeons, Orlando, FL, Jan 29 31, Address reprint requests to Dr Rusch, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York, NY mitoses per 10 high-power fields), and abundant necrosis. They did not find any advantage to electron microscopy, immunohistochemistry, or flow cytometry over light microscopy in classifying these tumors, which appeared to have a clinical prognosis intermediate between atypical carcinoids and small cell lung cancer. However, this newer grouping of typical and atypical carcinoids into low- and intermediate-grade neuroendocrine tumors, respectively, and of large cell neuroendocrine carcinomas and small cell lung cancers into high-grade neuroendocrine tumors is not widely accepted, partly because the number of patients originally reported by Travis and colleagues was small. The entity of mixed small cell large cell neuroendocrine carcinoma was proposed in 1982 for tumors consisting of small cell carcinoma with admixed large cells [7, 8]. This tumor appeared to have a poorer response to therapy than classic small cell lung cancer and was associated with a more aggressive course. The relationship of mixed small cell large cell neuroendocrine carcinoma (MNC) to large cell neuroendocrine carcinoma (LCNC) has not been fully explored. This study was performed to determine whether some 1996 by The Society of Thoracic Surgeons /96/$15.00 Published by Elsevier Science Inc PII S (96)

2 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 799 Table 1. Pathologic Features Used to Classify Pulmonary Neuroendocrine Neoplasms by Light Microscopy Neoplasm Pattern Mitoses Necrosis Cell Size Nucleoli Cytoplasm Typical carcinoid Organoid 3/10 hpf None Variable May be present Variable Atypical carcinoid Organoid 3 10/10 hpf Focal punctate Variable May be present Variable Large cell neuroendocrine Organoid 10/10 hpf Abundant Large Often prominent Abundant carcinoma Mixed small cell large cell Variable 10/10 hpf Abundant Intermediate May be present Moderate neuroendocrine carcinoma Small cell carcinoma Generally diffuse 10/10 hpf Abundant Small Absent Minimal hpf high-power field. of the molecular genetic abnormalities known to be important in both non-small cell and small cell lung cancers occur in these neuroendocrine lung tumors; and whether they can help characterize large cell and mixed small cell large cell neuroendocrine carcinomas as distinct entities, and clarify their relationship to the other neuroendocrine lung tumors. Material and Methods Identification of Pathology Specimens Specimens were obtained by retrospective review of the Department of Pathology records of primary lung tumors with neuroendocrine features, operated on from 1984 to Tumors were classified as typical carcinoids (TC), atypical carcinoids (AC), LCNC, MNC, and small cell lung cancers (SCLC) solely according to the histologic criteria shown in Table 1. The histologic definitions used for TC and small cell lung cancers are well accepted (Figs 1 and 2), whereas those used for AC (Fig 3) represent a strict interpretation of the criteria originally proposed by Arrigoni and colleagues [1]. Large cell neuroendocrine carcinomas (Fig 4) were selected according to the criteria established by Travis and co-workers [6], and MNC (Fig 5) included high-grade tumors characterized by a mixture of large cell and small cell components, or a more homogeneous cell population having features intermediate between small and large cells [7, 8]. All consecutive cases of AC, LCNC, MNC, and SCLC were included as were 25 randomly selected TC from the same time Fig 1. Histologic features of typical carcinoid tumor. At low power (A) the tumor has a nesting and trabecular architecture and lacks necrosis. At high power (B) the tumor cells have moderate to abundant cytoplasm, regular nuclei with a stippled chromatin pattern, and no mitoses.

3 800 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: Fig 2. Histologic features of small cell lung cancer. At low power (A) there is little architectural pattern. Large areas of necrosis are present, with hematoxylin staining around vessels (arrows), the so-called Azzopardi effect. At higher power (B) the cells are small with minimal cytoplasm, fusiform nuclei, and inapparent nucleoli. Fig 3. Histologic features of atypical carcinoid tumor. At low power (A) there is a nesting pattern, with punctate foci of necrosis. At higher power (B) the tumor cells are uniform and resemble those of a typical carcinoid, although scattered mitoses are present (arrows).

4 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 801 Fig 4. Histologic features of large cell neuroendocrine carcinoma. The low-power appearance (A) shows a well-defined nesting pattern with peripheral palisading of the nuclei. Necrosis is evident. At high power (B) the cells have large nuclei with prominent nucleoli and there is abundant cytoplasm. Mitoses are numerous. Fig 5. Histologic features of mixed small cell large cell neuroendocrine carcinoma. The low-power pattern (A) resembles that of small cell carcinoma, showing abundant necrosis and marked hypercellularity. At higher power (B) there is an equal admixture of smaller fusiform cells with absent nucleoli and larger cells with more cytoplasm, vesicular chromatin, and prominent nucleoli (arrows).

5 802 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: Table 2. List of Original and Revised Pathologic Diagnoses in All 92 Tumors Examined Revised Pathological Diagnosis TC AC LCNC MNC SCLC Original Pathological Diagnoses Typical carcinoid Atypical carcinoid Small cell carcinoma Large cell neuroendocrine carcinoma Mixed small cell large cell neuroendocrine carcinoma Neuroendocrine carcinoma Carcinoma with neuroendocrine features Large cell undifferentiated carcinoma AC atypical carcinoid; LCNC large cell neuroendocrine carcinoma; MNC mixed small cell large cell neuroendocrine carcinoma; SCLC small cell lung cancer; TC typical carcinoma. period. Use of these strict histologic criteria resulted in reclassifying a significant number of tumors, particularly among the LCNC and MNC. As shown in Table 2, LCNC and MNC were previously given a variety of histologic diagnoses ranging from AC to small cell carcinoma, and were sometimes simply termed neuroendocrine carcinoma. Immunohistochemical Staining Immunohistochemical staining was performed using standard methodology with the streptavidin-biotin-peroxidase technique. Primary antibodies, sources, dilutions, and pretreatments are listed in Table 3. Diaminobenzidine was used as the chromogen. Negative controls for each of the antibodies were performed using nonimmune serum instead of the primary antibody. Positive controls consisted of pulmonary non small cell carcinomas (NSCLC) known to exhibit expression of EGFR, p53, and Rb, respectively. The uniform presence of EGFR staining in the basal layer of the bronchial epithelium and its absence in other normal pulmonary tissues served as additional internal positive and negative controls, respectively. The positive staining for Rb in scattered endothelial cells also served as an internal control for that antibody. Stains for Ki67, p53, and EGFR (Figs 6, 7, and 8) were scored as 0 to 4 based on the percentage of positive tumors cells (0 less than 5%, 1 5% to 10%, 2 20% to 50%, 3 50% to 80%, 4 more than 80%). Stains for Rb (Fig 9) were scored as positive (greater than 20% of tumor nuclei positive), focal (less than 20% of nuclei positive), or negative (no nuclei staining). Clinical Data Base Clinical information was obtained by retrospective chart review. Data recorded included patient sex, age, and smoking history; presenting symptoms, preoperative therapy, type of operation; postoperative adjuvant therapy, disease status at last follow-up, and when applicable, the sites of relapse. When complete follow-up information was not available from the medical record every attempt was made to obtain this through contact with the patient, family, or primary physician. Statistical Methodology Overall survival was calculated from the date of operation, and described using the product limit method of Kaplan and Meier [9]. Overall survival was compared across the levels of prognostic factors using the log rank test [10]. All statistical tests were conducted at the twosided 0.05 level. Results A total of 92 patients operated on for primary neuroendocrine lung tumors during the period of 1984 to 1994 were studied. All of these patients underwent operation with the intent to resect the primary tumor, but in 2 patients the extent of disease at operation precluded resection and only biopsies were performed. The demographic characteristics of these patients are shown in Table 4. Twenty-five patients had TC, 7 had AC, 24 had LCNC, 18 had MNC, and 18 had SCLC. Because of the relatively small numbers of patients in each histologic category, analyses of clinical features, treatment, and outcome were performed by grouping together the TC and AC (low- and intermediate-grade tumors), and the LCNC with the MNC (both high-grade tumors). The median age was similar among these three categories and approximately half of the patients in each group were asymptomatic at the time of diagnosis, having had a pulmonary mass noted incidentally on chest radio- Table 3. Materials and Methods for Immunohistochemical Staining Antigen Ab Name Source Dilution Pretreatment Ki67 MIB-1 Immunotech, Westbrook, ME 1:50 Trypsin, miamase, citrate p53 pab1801 Oncogene Science, Uniondale, NY 1:500 Microwave, citrate buffer Rb OP-66 Oncogene Science, Uniondale, NY 1:1000 Microwave, citrate buffer EGFR 31G7 Triton, Alameda, CA 1:200 Pepsin

6 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 803 A Fig 6. Immunohistochemical staining for Ki67. In a low proliferative rate tumor (typical carcinoid) there is nuclear positivity in less than 5% of the cells (1 ) (A), whereas in a high proliferative rate tumor (small cell lung cancer), more than 90% of the cells are positive (4 ) (B). B A Fig 7. Immunohistochemical staining for p53. In these two large cell neuroendocrine carcinomas, one (A) shows 4 positivity, with nearly all of the nuclei staining; the other (B) is negative, with absent staining. B

7 804 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: A B Fig 8. Immunohistochemical staining for EGFR. Diffuse cytoplasmic positivity (4 ) is present in this large cell neuroendocrine carcinoma (A). In negative cases (B), only the basal layers of the normal respiratory epithelium stain. A B C Fig 9. Immunohistochemical staining for Rb. The normal pattern of staining varies with the proliferative rate. In a low proliferative rate tumor (atypical carcinoid), only focal staining of scattered tumor cell nuclei is present (A). In a high proliferative rate tumor (small cell lung cancer) there is more widespread positivity (B). In the presence of Rb abnormalities in a small cell lung cancer, no tumor cells stain (C), although reaction product is present in nonneoplastic endothelial cells (arrows), an essential internal positive control.

8 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 805 Table 4. Demographic Information and Preoperative Treatment of 92 Patients With Neuroendocrine Tumors Characteristic TC AC (n 32) LCNC MNC (n 42) SCLC (n 18) Median age (range) 57 (24 77) 66 (36 76) 62 (50 77) Male:female 11:21 20:22 6:12 Smoking history Never 12 (36%) 4 (9%) 1 (5.5%) Current or recent Quit 1 year Asymptomatic Preoperative Rx None 29 (91%) 33 (79%) 12 (66%) Radiation Chemotherapy AC atypical carcinoma; LCNC large cell neuroendocrine carcinoma; MNC mixed small cell large cell neuroendocrine carcinoma; SCLC small cell carcinoma; TC typical carcinoma. graph. The remainder presented with cough, hemoptysis, shortness of breath, or postobstructive pneumonia. A few of the patients with either carcinoid tumors or LCNC or MNC received preoperative radiation or chemotherapy usually because SCLC was suspected on initial biopsy. However, two-thirds of the patients with SCLC did not receive any preoperative treatment emphasizing that these patients represent a highly select group with very early tumors, often diagnosed by resection of a peripheral pulmonary nodule. Not surprisingly, the largest proportion of nonsmokers belonged to the group of patients who had TC or AC, tumors known not to be tobacco related. The surgical management and the pathologic stage of the tumors is shown in Table 5. In 4 patients the available clinical information was insufficient to encode the type of pulmonary resection in the data base. The most common form of resection in each tumor group was a lobectomy or bilobectomy consistent with the fact that most of tumors were stage I or II. However, 78% of carcinoid tumors and 50% of the SCLC were stage I, whereas only 31% of the LCNC and MNC were stage I. The largest number of stage III tumors (52%) belonged to the LCNC/MNC group. The clinical postoperative course is outlined in Table 6. Only one of the AC patients was given postoperative radiation, whereas 38% of the LCNC and MNC patients and 50% of the SCLC patients received adjuvant radiation or chemotherapy. In the case of the LCNC and MNC patients the decision to offer adjuvant therapy was made on an individualized basis and appears to have been prompted by the locally advanced status or small cell-like histologic features of the tumor. Although the small numbers of patients in each histologic category preclude formal comparison of overall survival among tumor subsets, the median survival of patients with LCNC and MNC (18.7 months) was distinctly worse than that of TC and AC patients (not yet reached) and better than the median survival of SCLC patients (14.3 months). The differences in overall survival among the histologic subsets is easily appreciated by the survival curves shown in Figure 10. There was a significant difference in overall survival for TC versus AC tumors (p 0.02) and for AC versus LCNC (p 0.01), but not for LCNC versus MNC tumors (p 0.43), LCNC versus SCLC (p 0.81), or MNC versus SCLC (p 0.47). A comparison of the low/intermediate-grade tumors (TC and AC) to the high-grade tumors (LCNC, MNC, and SCLC) shows that the difference in overall survival rates (Fig 11) is highly significant (p 0.01). In a multivariate analysis controlling for stage, histology had a significant effect on survival when comparing SCLC to TC and AC tumor (p 0.01). The difference in survival between SCLC and LCNC and MNC tumors again did not reach statistical significance (p 0.06). The results of immunohistochemical staining are shown in Table 7. Only one of the AC and none of the TC showed 3 to 4( ) staining for Ki67. The remainder exhibited a low proliferative rate. Approximately one-third of the TC and AC group overexpressed the EGFR but abnormal p53 expression and absence of Rb expression Table 5. Surgical Treatment and Pathologic Staging a Characteristic TC AC (n 32) LCNC MNC (n 42) SCLC (n 18) Procedure type Biopsy Wedge/segment Lobectomy/bilobectomy Pneumonectomy TN status T1 N T2 N T1 N T2 N T1 NX 1 b 1 b 0 T3 N T3 N T1 N T2 N T3 N T3 NX a T4 NX TX N TX NX M Stage I II IIIa IIIb IV a T and N status could not be established for four tumors: 1 AC, 1 LCNC, 1 MNC, and 1 SCLC. b Not classified by stage due to unknown N status. AC atypical carcinoma; LCNC large cell neuroendocrine carcinoma; MNC mixed small cell large cell neuroendocrine carcinoma; SCLC small cell carcinoma; TC typical carcinoma.

9 806 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: Table 6. Postoperative Adjuvant Treatment and Clinical Course of All 92 Neuroendocrine Tumors Characteristic TC AC (n 32) LCNC MNC (n 42) SCLC (n 18) Adjuvant therapy None 31 (97%) 27 (64%) 9 (50%) Radiation a Chemotherapy a Median follow-up (mo) Median survival (mo) Not reached (range) (1 72) (1 59) (1 103) No. of patients with 3 (9%) 24 (57%) 9 (50%) relapse Sites of relapse Local only Regional only Distant only Distant and locoregional a Some patients received both radiation and chemotherapy. AC atypical carcinoma; LCNC large cell neuroendocrine carcinoma; MNC mixed small cell large cell neuroendocrine carcinoma; SCLC small cell carcinoma; TC typical carcinoma. were uniformly not seen. In contrast, almost all LCNC, MNC, and SCLC exhibited a high proliferative rate as assessed by 3 to 4( ) staining for Ki67. Both the LCNC and MNC and the SCLC tumors frequently overexpressed the EGFR although this occurred more often among the LCNC and MNC tumors (42% and 39%, respectively, versus 22% for the SCLC). These differences were not statistically significant (p 0.19). Abnormal p53 expression was seen more commonly in the MNC and SCLC (each 39%) than in the LCNC (17%, p 0.11). Absent Rb expression was found in 14 (58%) of the LCNC, 16 (89%) of the MNC tumors, and 17 (94%) of the SCLC. The differences in Rb expression were statistically significant for LCNC tumors versus SCLC (p 0.01), but not for MNC versus SCLC (p 0.51). Fig 10. Overall survival by tumor histology. (AC atypical carcinoma; LCNC large cell carcinoma; MNC mixed small large cell neuroendocrine carcinoma; SCLC small cell carcinoma; TC typical carcinoma.) Fig 11. Overall survival in the low- and intermediate-grade neuroendocrine tumors (typical carcinoma and atpical carcinoma) versus the high-grade neuroendocrine tumors (large cell carcinoma and mixed small large cell neuroendocrine carcinoma and small cell carcinoma). The p value was less than Comment The relationship between carcinoid tumors and other neuroendocrine lung tumors, including SCLC was not recognized for a long time. Carcinoids were originally thought to be benign and were classified under the overall category of bronchial adenomas. However, occasional reports of histologically AC that seemed to behave more aggressively prompted Arrigoni and colleagues [1] to study in detail 23 such patients that represented 11.4% of all carcinoid tumors seen at the Mayo Clinic through They defined the term atypical carcinoid and found that these tumors indeed behaved more aggressively than typical carcinoids. They were characterized by older patient age at presentation, larger primary tumors, and distant metastases in 16 of the 23 (70%) patients as compared with 5.6% of typical carcinoids. On the basis of this review, Arrigoni and co-workers proposed the histologic criteria for AC including (1) increased mitotic activity in the presence of a recognizable carcinoid pattern (average of one mitotic figure per one to two high-power fields), (2) pleomorphism and irregularity of nuclei with prominent nucleoli, (3) areas of increased cellularity with architectural disorganization, and (4) areas of tumor necrosis. The study by Arrigoni and colleagues established AC as a distinct histopathologic entity [11] and helped identify all carcinoid tumors as low-grade malignancies rather than benign tracheobronchial tumors [12]. It is now clear that neuroendocrine lung tumors form a disease spectrum ranging from bland-appearing indolent typical carcinoids to biologically aggressive SCLC [13 17]. However, the broad range of tumors in between these two extremes have defied easy, reproducible classification [18]. Frequently all higher grade tumors not showing the characteristics of SCLC have been loosely termed AC, thereby considerably extending Arrigoni s original specific histologic criteria. Alternative classification schemas, notably that by Gould and Warren, pro-

10 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 807 Table 7. Results of Immunohistochemical Staining on All 92 Neuroendocrine Tumors, Displayed According to Histologic Category a Immunohistochemical Stain TC AC LCNC MNC SCLC Ki67 (1 2 ) 25 (100%) 6 (86%) 2 (8%) 0 (0%) 1 (6%) Ki67 (3 4 ) 0 (0%) 1 (14%) 22 (92%) 18 (100%) 17 (94%) p53 (0 1 ) 25 (100%) 7 (100%) 20 (83%) 11 (61%) 11 (61%) p53 (2 4 ) 0 (0%) 0 (0%) 4 (17%) 7 (39%) 7 (39%) Rb (, focal) 23 (100%) 7 (100%) 10 (42%) 2 (11%) 1 (6%) Rb ( ) 0 (0%) 0 (0%) 14 (58%) 16 (89%) 17 (94%) EGFR (0 1 ) 18 (72%) 3 (43%) 14 (58%) 11 (61%) 14 (78%) EGFR (2 4 ) 7 (28%) 4 (57%) 10 (42%) 7 (39%) 4 (22%) a The methods of scoring the immunohistochemical staining are described under Material and Methods. AC atypical carcinoma; LCNC large cell neuroendocrine carcinoma; MNC mixed small cell large cell neuroendocrine carcinoma; SCLC small cell carcinoma; TC typical carcinoma. pose that these tumors should be termed well-differentiated neuroendocrine carcinomas and divided into three subsets based on pleomorphism, mitotic count, and necrosis [2, 3, 19]. Another classification schema suggested calling TC and AC, Kulchitzky cell carcinomas types I and II, and SCLC, Kulchitzky cell carcinoma type III, in deference to the common cell of origin of these neuroendocrine tumors [4, 5]. Neither of these classification schemas have been widely accepted, perhaps because they do not lend themselves easily to clinical application. However, the simple grouping of neuroendocrine lung tumors into TC and AC and SCLC also remains inadequate to describe the broad range of histologic appearance and biological behavior of these tumors [20]. Recently, Travis and colleagues [6] proposed the term of LCNC to describe tumors that are high grade but lack the cytologic features of SCLC. These tumors are characterized by (1) a light microscopic neuroendocrine appearance, (2) cells of large size with low nuclear-tocytoplasmic ratio and frequent nucleoli, and (3) high mitotic rate (greater than 10 mitoses per 10 high-power field) and frequent necrosis. Large cell neuroendocrine carcinomas are considered distinct from NSCLC with neuroendocrine features in which neuroendocrine differentiation is detectable only by immunohistochemistry or electron microscopy. The latter group of tumors fall under the standard classification of NSCLC and there is currently no evidence that the neuroendocrine features of these tumors indicate a different biological behavior [21]. Although the classification by Travis and co-workers neatly describes the histologic spectrum of neuroendocrine lung tumors, their original analysis was based on only five LCNC and has not yet gained wide acceptance. Thus, previous classification schemas for neuroendocrine lung tumors are not widely used, and a more recently proposed schema appears practical but needs further validation. The entity of MNC was formally separated as a subtype of SCLC by the International Association for the Study of Lung Cancer in 1988, when it was also recognized that the intermediate cell subtype of SCLC represented a category without clinical significance [7]. The MNC contained both small cells with the typical cytoarchitectural features of SCLC as well as larger cells with nucleoli and more abundant cytoplasm. Several studies have shown a more aggressive clinical course and a reduced response to chemotherapy for MNC, relative to SCLC [8, 22, 23]. Unfortunately, recent reclassifications of pulmonary neuroendocrine tumors have not specifically explored the relationship of MNC to LCNC [6]. Immunohistochemistry for standard neuroendocrine markers (eg, neuron-specific enolase, chromogranin, synaptophysin), electron microscopy and flow cytometry have been used to determine whether they might augment the histologic classification of neuroendocrine lung tumors and assist in separating low-grade from highgrade tumors. Unfortunately, none of these methods appear clinically useful because staining for neuroendocrine markers among these tumors is highly variable, and abnormal DNA content is frequent among both TC and AC [24, 25]. Although the cases included in this study did not result from a systematic review of all lung tumors, it is apparent that the number of AC is much less than might be expected from the reported frequency of this tumor relative to TC. This most likely resulted from the use of very strict criteria for classification, with the resultant inclusion of some cases in the high-grade groups that might have been regarded as AC with more liberal criteria. We hypothesized that if LCNC and MNC belong to the group of higher grade tumors within the neuroendocrine spectrum, closely related to SCLC, they should exhibit molecular genetic abnormalities that are common in SCLC. Although the precise mechanisms that lead to the development of NSCLC and SCLC are not known, several common molecular genetic abnormalities have been identified in those tumors. Both NSCLC and SCLC seem to arise as a result of combined inactivation of tumor suppressor genes and activation of oncogenes. Frequently inactivated tumor suppressor genes include p53

11 808 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: and Rb along with a putative tumor suppressor gene located on chromosome 3p [26, 27]. Frequently activated oncogenes include K-ras [28] and the EGFR, which is thought to participate in an autocrine growth loop with one of its ligands, transforming growth factor alpha. Although K-ras activation by point mutation is almost always confined to adenocarcinomas, altered expression of p53, Rb, and EGFR is seen across NSCLC histologies and stages, and can be reliably detected by immunohistochemistry. Although Rb abnormalities may be seen in 30% of NSCLC, they are nearly universal in SCLC, perhaps representing a relative molecular distinction between endocrine and nonendocrine lung cancers [29]. The expression of EGFR is normally confined to the basal layer of the bronchial epithelium and overexpression can be detected in the earliest preneoplastic lesions, even in areas of bronchial atypia and low-grade dysplasia [30]. Overexpression of EGFR is infrequent in SCLC, but is seen in approximately half of all NSCLC [31]. The p53 immunohistochemical staining caused either by mutations of the p53 gene or inactivation of the normal protein product, is seen in approximately half of NSCLC and SCLC, and also occurs in areas of high-grade bronchial dysplasia and carcinoma-in-situ [30, 32]. Thus, EGFR and p53 overexpression and loss of Rb expression constitute frequent molecular genetic abnormalities that can be easily detected by immunohistochemical staining performed on archival lung tumor samples. Less is known about the molecular genetic abnormalities associated with carcinoid tumors [33]. Roncalli and colleagues [34] compared the p53 immunohistochemical staining in 15 TC, 22 AC, and 22 SCLC. None of the TC showed positive staining, whereas 45% of the AC and 59% of the SCLC showed immunoreactivity. However, the AC were classified according to the criteria of Gould and Warren for well-differentiated neuroendocrine carcinoma and therefore it is possible that their group of atypical carcinoids included tumors that might fit Travis criteria for LCNC. Barbareschi and colleagues [35] examined the expression of p53 and Rb in 14 TC, 10 AC, 4 borderline atypical carcinoids/sclc, and 11 SCLC [35]. They found that p53 expression was absent in all carcinoids, abnormally expressed in 8 SCLC and in 1 borderline case. Rb expression was present in TC and AC and absent in all other tumors. Although the criteria for distinguishing AC and borderline cases were not clearly stated, this study suggested that it might be possible to use immunohistochemical staining for several common molecular genetic abnormalities to help distinguish among groups of neuroendocrine tumors. Our study confirms and extends Barbareschi s report. Ki67, considered a very reliable marker of proliferative rate, indicated that all but 2 LCNC and MNC categorized by light microscopic criteria were highly proliferative tumors. p53 immunoreactivity occurred in none of the TC or AC, but was seen in 17% of LCNC, 39% of MNC tumors, and 39% of SCLC. Absence of Rb expression was found in 94% of SCLC, 89% of MNC, and 58% of LCNC. EGFR overexpression was common to all neuroendocrine tumor types, occurring in 39% of carcinoids, 42% of LCNC, 39% of MNC, and 22% of SCLC. These results allow a sharp separation of low/intermediate-grade from high-grade tumors and suggest that LCNC, MNC, and SCLC share common molecular genetic features. The relatively high frequency of Rb abnormalities is consistent with the hypothesis that high-grade neuroendocrine carcinomas share this frequent molecular abnormality and that SCLC is closely related to MNC (especially) and LCNC. The confusion about the histologic classification of neuroendocrine lung tumors creates a significant dilemma for clinicians treating these neoplasms. Surgical resection is the accepted treatment for TC and AC, whereas locally advanced NSCLC are now managed with combined modality therapy, and the small group of very early SCLC included in this series are most frequently treated by resection and postoperative adjuvant chemotherapy. Discrepancies in pathologic diagnoses between small initial biopsy specimens and subsequent larger surgical resection specimens related to tumor heterogeneity or to variable use of classification schemas makes it even harder for clinicians to develop appropriate and effective treatment plans. Within the limitations of this retrospective study the poor clinical outcome of patients with LCNC and MNC raises the issue of whether they should be considered for formal trials of combined modality therapy. Approximately half of these patients presented with locally advanced tumors (T3-4, or N2-3) and median survival was only 18.7 months, despite 40% of patients having received postoperative adjuvant radiation or chemotherapy. Moreover, distant metastases developed in 66% of the patients who relapsed. This contrasts sharply with TC and AC, as surgical resection offers a significant chance of cure even when nodal metastases are present [36]. Thus, the current study suggests that it may now be possible to define specific groups of low/intermediategrade and high-grade neuroendocrine tumors based on the combined use of light microscopic criteria, proliferative rate, and at least two molecular markers. The SCLC, MNC, and LCNC appear to represent a spectrum of related high-grade neuroendocrine neoplasms sharing molecular features as well as aggressive biological behavior. The poor overall survival and frequency of distant metastases suggest that patients with LCNC and MNC should be targeted for formal investigation of other approaches to treatment, specifically combined modality therapy. We thank Ethan Dmitrovsky, MD, for his critical review of the manuscript, Marie Gregorio, RN, for her help in data management, and Melody Owens, PhD, for her expert assistance in data management and manuscript preparation. References 1. Arrigoni MG, Woolner LB, Bernatz PE. Atypical carcinoid tumors of the lung. J Thorac Cardiovasc Surg 1972;64: Warren WH, Memoli VA, Gould VE. Immunohistochemical and ultrastructural analysis of bronchopulmonary neuroen-

12 Ann Thorac Surg RUSCH ET AL 1996;62: MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 809 docrine neoplasms. II. Well-differentiated neuroendocrine carcinomas. Ultrastruct Pathol 1984;7: Warren WH, Memoli VA, Gould VE. Immunohistochemical and ultrastructural analysis of bronchopulmoary neuroendocrine neoplasms. I. Carcinoids. Ultrastruct Pathol 1984;6: DeCaro LF, Paladugu R, Benfield JR, Lovisatti L, Pak H, Teplitz RL. Typical and atypical carcinoids within the pulmonary APUD tumor spectrum. J Thorac Cardiovasc Surg 1983;86: Paladugu RR, Benfield JR, Pak HY, Ross RK, Teplitz RL. Bronchopulmonary Kulchitzky cell carcinomas. A new classification scheme for typical and atypical carcinoids. Cancer 1985;55: Travis WD, Linnoila I, Tsokos MG, et al. Neuroendocrine tumors of the lung with proposed criteria for large-cell neuroendocrine carcinoma. An ultrastructural, immunohistochemical, and flow cytometric study of 35 cases. Am J Surg Pathol 1991;15: Hirsch FR, Matthews MJ, Aisner S, et al. Histopathologic classification of small cell lung cancer. Changing concepts and terminology. Cancer 1988;62: Radice PA, Matthews MJ, Ihde DC, et al. The clinical behavior of mixed small cell/large cell bronchogenic carcinoma compared to pure small cell subtypes. Cancer 1982;50: Kaplan EL, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc 1958;53: Peto R, Peto J. Asymptotically efficient rank invariant test procedures. J R Statist Soc 1972;135: Yousem SA. Pulmonary carcinoid tumors and well-differentiated neuroendocrine carcinomas. Is there room for an atypical carcinoid? Am J Clin Pathol 1991;95: Okike N, Bernatz PE, Woolner LB. Carcinoid tumors of the lung. Ann Thorac Surg 1976;22: McCaughan BC, Martini N, Bains MS. Bronchial carcinoids. Review of 124 cases. J Thorac Cardiovasc Surg 1985;89: Rea F, Binda R, Spreafico G, et al. Bronchial carcinoids: a review of 60 patients. Ann Thorac Surg 1989;47: Harpole Jr DH, Feldman JM, Buchanan S, Young WG, Wolfe WG. Bronchial carcinoid tumors: a retrospective analysis of 126 patients. Ann Thorac Surg 1992;54: Ribet M, Gosselin B, Gambiez L, Frere LE. Bronchial carcinoids. Eur J Cardiothorac Surg 1993;7: Todd TR, Cooper JD, Weissberg D, Delarue NC, Pearson FG. Bronchial carcinoid tumors. Twenty years experience. J Thorac Cardiovasc Surg 1980;79: Lequaglie C, Patriarca C, Cataldo I, Muscolino G, Preda F, Ravasi G. Prognosis of resected well-differentiated neuroendocrine carcinoma of the lung. Chest 1991;100: Warren WH, Gould VE, Faber LP, Kittle CF, Memoli VA. Neuroendocrine neoplasms of the bronchopulmonary tract. A classification of the spectrum of carcinoid to small cell carcinoma and intervening variants. J Thorac Cardiovasc Surg 1985;89: Warren WH, Memoli VA, Jordan AG, Gould VE. Reevaluation of pulmonary neoplasms resected as small cell carcinomas. Significance of distinguishing between well-differentiated and small cell neuroendocrine carcinomas. Cancer 1990;65: Linnoila RI, Piantadosi S, Ruckdeschel JC. Impact of neuroendocrine differentiation in non-small cell lung cancer. The LCSG experience. Chest 1994;106(Suppl):367S 71S. 22. Hirsch FR, Osterlind K, Hansen HH. The prognostic significance of histopathologic subtyping of small cell carcinoma of the lung according to the classification of the World Health Organization. A study of 375 consecutive patients. Cancer 1983;52: Fushimi H, Kikui M, Morino H, et al. Detection of large cell component in small cell lung carcinoma by combined cytologic and histologic examinations and its clinical implication. Cancer 1992;70: Jackson-York GL, Davis BH, Warren WH, Gould VE, Memoli VA. Flow cytometric DNA content analysis in neuroendocrine carcinoma of the lung. Correlation with survival and histologic subtype. Cancer 1991;68: Yousem SA, Taylor SR. Typical and atypical carcinoid tumors of lung: a clinicopathologic and DNA analysis of 20 tumors. Mod Pathol 1990;3: Chiba M, Takahashi TA, Nau MM, et al. Mutations in the p53 gene are frequent in primary, resected non-small cell lung cancer. Oncogene 1990;5: Rusch VW, Dmitrovsky E. Molecular biologic features of non-small cell lung cancer. Clinical implications. Chest Surg Clin N Am 1995;5: Rodenhuis S, Slebos RJC. Clinical significance of ras oncogene activation in human lung cancer. Cancer Res 1992;52:2665s 9s. 29. Reissmann PT, Koga H, Takahashi R, et al. Inactivation of the retinoblastoma susceptibility gene in non-small-cell lung cancer. Oncogene 1993;8: Rusch V, Klimstra D, Linkov I, Dmitrovsky E. Aberrant expression of p53 or the epidermal growth factor receptor is frequent in early bronchial neoplasia, and coexpression precedes squamous cell carcinoma development. Cancer Res 1995;55: Rusch V, Baselga J, Cordon-Cardo C, et al. Differential expression of the epidermal growth factor receptor and its ligands in primary non small cell lung cancers and adjacent benign lung. Cancer Res 1993;53: Carbone DP, Mitsudomi T, Chiba I, et al. p53 immunostaining positivity is associated with reduced survival and is imperfectly correlated with gene mutations in resected nonsmall cell lung cancer. A preliminary report of LCSG 871. Chest 1994;106 (suppl):377s 81S. 33. Lohmann DR, Fesseler B, Pütz B, et al. Infrequent mutations of the p53 gene in pulmonary carcinoid tumors. Cancer Res 1993;53: Roncalli M, Doglioni C, Springall DR, et al. Abnormal p53 expression in lung neuroendocrine tumors. Diagnostic and prognostic implications. Diagn Mol Pathol 1992;1: Barbareschi M, Girlando S, Mauri FA, et al. Tumour suppressor gene products, proliferation, and differentiation markers in lung neuroendocrine neoplasms. J Pathol 1992; 166: Martini N, Zaman MB, Bains MS, et al. Treatment and prognosis in bronchial carcinoids involving regional lymph nodes. J Thorac Cardiovasc Surg 1994;107:1 7. DISCUSSION DR WILLIAM H. WARREN (Chicago, IL): I want to congratulate Dr Rusch on a very fine presentation and for providing me with a copy of the manuscript in advance for discussion. This is a very ambitious project that attempts to bridge the disciplines of molecular biology and thoracic surgery. A great deal has been accomplished in basic biology over the past 10 years and attempts to explore its clinical relevance is timely and can only advance our understanding of tumor biology. I wish to underscore several important points made by Dr Rusch and colleagues. First is the acknowledgment of the entity of large cell carcinomas with predominantly neuroendocrine differentiation, as we had recognized in our original classifica-

13 810 RUSCH ET AL Ann Thorac Surg MOLECULAR MARKERS AND NEUROENDOCRINE TUMORS 1996;62: tion of neuroendocrine tumors in The tumors may go unrecognized were it not for the almost routine use of neuroendocrine markers in our laboratory. Up to 20% of these large cell carcinomas, currently called simply large cell undifferentiated carcinoma, in fact have predominantly neuroendocrine differentiation when studied. The majority of the remaining 80% of these tumors have either exocrine or mixed exocrine/neuroendocrine patterns of differentiation. In our experience, as in the experience of Rusch and colleagues, large cell neuroendocrine carcinomas are extremely aggressive tumors that may respond to chemotherapy currently used for small cell neuroendocrine carcinoma. It remains to be seen if these tumors, stage for stage, are more aggressive than their exocrine counterparts. My first question is, were these large cell tumors studied to confirm that they were in fact neuroendocrine and distinguish them from the other population? In any study on neuroendocrine differentiation (especially those dealing with poorly differentiated neuroendocrine carcinomas), these tumors must be confirmed as neuroendocrine in nature either by immunohistochemistry or by electron microscopy. The identification of aggressive neuroendocrine tumors behaving like small cell neuroendocrine carcinoma but currently included among tumors called non small cell lung cancer is relevant, and I propose that we adopt the more useful and more accurate classification of neuroendocrine versus nonneuroendocrine tumors rather than small cell versus non small cell lung cancer. The molecular markers used in this study were chosen presumably because at some time they were reported to be markers of proliferative activity. I submit, however, that these profiles simply confirm what pathologists already know, namely that carcinoids are indolent tumors and that the more aggressive tumors can be identified by mitotic rate, vascular invasion, and presence of necrosis. A much more useful study, from a clinical point of view, would be to study the pattern of light microscopy of these tumors if they are similar by hematoxylin and eosin staining to large cell carcinomas of a similar stage and treating them in a similar fashion and assess the prognostic significance of these molecular markers. Have you looked at any nonneuroendocrine lung cancers and compared their profile with respect to these markers with those of these neuroendocrine carcinomas? DR RUSCH: Thank you, Dr Warren, for your incisive comments. The tumors selected for this study were already clearly identified by light microscopy as neuroendocrine tumors or tumors with neuroendocrine features. Some form of neuroendocrine immunohistochemical staining was usually done to confirm the light microscopic findings. However, as we have both discussed standard neuroendocrine stains (chromogranin, synaptophysin, and neuron-specific enolase) do not uniformly distinguish among these groups of tumors. We did not do a complete retrospective review of all primary lung tumors, because the number of tumors available for review during this 10-year period at Memorial Hospital would have been prohibitively large. Moreover, the Lung Cancer Study Group found that tumors diagnosed as ordinary non small cell lung cancers by light microscopy but found incidentally to have positive neuroendocrine immunohistochemical staining do not have a different prognosis. Therefore, it seems important to distinguish between these tumors and tumors that are neuroendocrine by light microscopy. The choice of molecular markers used in this study was very specific. The expression of the retinoblastoma gene is abnormal in almost all small cell lung cancers and in about a third of non small cell lung cancers. The pattern of expression of p53 and the EGF receptor is different in small cell compared with non small cell lung cancer. P53 mutation or overexpression occurs in approximately half of non small cell lung cancers and far less frequently in small cell lung cancers. EGF receptor overexpression is seen in about half non small cell lung cancers and rarely in small cell lung cancers. In contrast to standard neuroendocrine immunohistochemical stains, staining for Rb, p53, and EGF receptor does not just identify that a tumor has neuroendocrine features. It reflects the molecular genesis of these tumors.