Prognostic Risk Factors That Identify Patients with Clinical Stage I Nonseminomatous Germ Cell Tumors at Low Risk and High Risk for Metastasis

Transcription

1 1002 Prognostic Risk Factors That Identify Patients with Clinical Stage I Nonseminomatous Germ Cell Tumors at Low Risk and High Risk for Metastasis Axel Heidenreich, M.D. 1,3 Isabell A. Sesterhenn, M.D. 2 F. Kash Mostofi, M.D. 2 Judd W. Moul, M.D. 3 1 Department of Urology, University of Cologne, Cologne, Germany. 2 Armed Forces Institute of Pathology, Washington, DC. 3 Department of Surgery, Uniformed Services University of the Health Sciences, Bethesda, Maryland. Presented in part at the 92nd meeting of the American Urological Association, New Orleans, Louisiana, April 12 17, Supported by a grant from the German Research Council (A. H.), He 2618/1-1, Deutsche Forschungsgemeinschaft, Bonn, Germany. Address for reprints: Department of Surgery, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD Received September 2, 1997; revision received January 2, 1998; accepted March 19, BACKGROUND. The purpose of this study was to develop a reliable model to identify clinical Stage I nonseminomatous germ cell tumors (NSGCTs) associated with low risk or high risk for occult retroperitoneal metastasis, so that the model could be used to customize the therapeutic approach for patients with these tumors. The model was to be based on pathohistologic parameters and immunohistochemical expression of proliferation markers, proteases, and adhesion molecules in the primary tumor. METHODS. One hundred forty-nine patients with clinical Stage I NSGCTs underwent retroperitoneal lymphadenectomy and were included in the study. Three to five paraffin embedded, formalin fixed tissue blocks were available from each patient and were analyzed for the following histopathologic features associated with pathologic Stage I or II disease: the presence or absence of vascular invasion (VI), the presence or absence of tunic invasion, and the percentage of each histologic type present in the primary tumor. Immunohistochemical expression of MIB-1, p53, bcl-2, cathepsin D, and E-cadherin was evaluated using a semiquantitative scoring system. Statistical analysis was performed with univariate and multivariate logistic regression models. RESULTS. The percentage of embryonal carcinoma (%EC, P 0.001) and the presence of VI (P ) and tunic invasion (P 0.002) were the most significant independent risk factors associated with pathologic Stage II disease. A combination of %EC and VI allowed correct prediction of final pathologic stage for 88% of clinical Stage I patients. Cutoff values including both variables identified the correct pathologic stage for 131 of 149 patients (88%). Less than 45% EC and the absence of VI correctly identified pathologic Stage I disease in 91.5% of patients; more than 80% EC and the presence of VI correctly predicted pathologic Stage II in 88%. In univariate analysis, only p53 (P 0.03) and E-cadherin (P 0.001) expression were significantly different in the embryonal carcinoma component of pathologic Stage I and II NSGCT. To evaluate prospectively the clinical utility of the new derived cutoff points, the data were applied to 10 consecutive patients with clinical Stage I NSGCT who underwent retroperitoneal lymphadenectomy; pathologic Stage I and II were correctly predicted for 5 of 6 Stage I and 4 of 4 Stage II patients, respectively. CONCLUSIONS. %EC and the presence or absence of VI appear to be reliable prognosticators to identify patients at high risk and low risk for occult retroperitoneal disease. In cases of clinical Stage I NSGCT, p53, bcl-2, MIB-1, cathepsin D, and E-cadherin did not appear to be of prognostic significance. The authors recommend that all patients with clinical Stage I NSGCT have their primary orchiectomy specimens evaluated for %EC and the presence of VI to determine their risk for occult retroperitoneal metastasis. Cancer 1998;83: American Cancer Society. KEYWORDS: testicular cancer, nonseminomatous germ cell tumor, prognostic risk factors, vascular invasion, embryonal carcinoma American Cancer Society

2 Prognostic Risk Factors in Clinical Stage I NSGCT/Heidenreich et al Because of the inaccuracy of clinical staging methods in distinguishing pathologic Stage I disease from occult retroperitoneal lymph node metastases in low stage nonseminomatous germ cell tumors (NS- GCTs) of the testis, clinical research focusing on the development of prognostic risk factors remains one of the most important and clinical challenges. Clinical Stage I testicular cancer is highly curable with nerve-sparing retroperitoneal lymph node dissection (RPLND), 1 primary chemotherapy, 2 or surveillance, 3 and several studies have attempted to improve clinical staging and thereby identify patients at high or low risk for occult retroperitoneal disease. Over the last decade, histologic and immunohistochemical risk factors have been examined in order to help stratify the risk of occult lymph node metastases in clinical Stage I NSGCT and thereby promote more scientifically based decision-making regarding one of the above-mentioned treatment options for the individual patient. 4 In previous studies, vascular invasion and the percentage of embryonal carcinoma in an orchiectomy specimen have been identified as significant histopathologic risk factors for pathologic Stage II disease The addition of DNA flow cytometry 15 and immunohistochemical assessment of proliferation markers such as proliferating cell nuclear antigen (PCNA), 16 MIB-1, 17,18 and p53 19,20 to these histopathologic features sometimes did not improve assessment of the probability of predicting lymph node involvement, 15,16,19,20 and at other times the data reported were contradictory, making a final decision on their clinical utility impossible. 17,18 The expression of the serine protease cathepsin D and other proteases is clearly involved in the cellular biology of metastasis and correlates with worse prognosis among some but not all patients with breast carcinoma. No studies of cathepsin D expression in testicular germ cell tumors have yet been conducted. The cadherin family is a group of genetically related transmembrane glycoproteins that are involved in intercellular adhesion mechanisms. Loss of E-cadherin has been shown to play a role in the progression of human malignancies by contributing to characteristic neoplastic properties, such as local invasiveness and the ability of the neoplasm to metastasize. With regard to testicular germ cell tumors, no studies have been performed correlating loss of E-cadherin expression with the biologic properties of the testicular cancer cell. In our current study, we evaluated the immunohistochemical expression of p53 tumor suppressor gene, bcl-2, cell cycle associated proliferation marker Ki-67, serine protease cathepsin D, and cell-to-cell adhesion molecule E-cadherin in a large cohort of patients with clinical Stage I NSGCTs to determine their clinical utility in predicting occult lymph node disease. In addition, all NSGCTs were evaluated for the presence of vascular invasion and quantitation of the percentage of embryonal carcinoma. All data were compared with the results of immunohistochemical staining in univariate and multivariate analyses. PATIENTS AND METHODS Primary tumor histopathology slides stained with hematoxylin and eosin (H&E) were available for 149 clinical Stage I (CSI) NSGCTs. CSI conformed to negative radiographic staging of the chest, abdomen, and pelvis as well as appropriate tumor marker decrease after radical orchiectomy, as described previously. 16 All 149 patients had undergone RPLND defining a known pathologic stage. Although our group of CSI NSGCTs consisted of nonconsecutive patients, no selection bias was evident. The study cohort was based strictly on the availability of tissue. In a further effort to avoid bias in the interpretation of quantitative histopathology and interpretation of immunohistochemical staining results, both the urologic and the pathologic investigators were blinded to stage until complete analysis was performed. Paraffin embedded, formalin fixed tissue blocks (3 5) from the primary tumor were available for all 149 patients. Histopathologic review of the H&E stained slides for these cases was performed by a single pathologist who was blinded to the stage of disease. After review of all available H&E stained slides, the archival block corresponding to the most representative area of tumor was chosen for immunohistochemical staining for p53, bcl-2, MIB-1, cathepsin D, and E-cadherin analysis, and was defined as a block containing all cell types in a mixed tumor or the highest concentration of neoplasm in a pure tumor. Quantitative Histopathology From the review of all available H&E stained slides, the presence or absence of vascular invasion and quantitative histology was estimated for each section under low-power magnification to determine the percentage of each separate histologic cell type, and a summary value was recorded for all CSI NSGCTs, as described previously. 13 The percentage of embryonal carcinoma (%EC) was a continuous variable between 0% and 100% and represented a general impression obtained by the reviewing pathologist. Venous invasion was recognized when tumor either adhered to a vessel wall or almost completely filled a space lined with flat endothelial cells containing red blood cells (RBCs). Lymphatic invasion was distinguished by ob-

3 1004 CANCER September 1, 1998 / Volume 83 / Number 5 serving tumor in spaces lined with endothelial cells without RBCs but with occasional lymphocytes. The locations of these vessels and their relation to arteries and veins were also important in identifying lymphatics. The prognostic significance of intratumoral or peritumoral vascular invasion was not different according to the above-mentioned definition of vascular invasion, and no differentiation was made between these two terms. We use the term vascular invasion (VI) generically to indicate the presence of either venous or lymphatic invasion. Immunohistochemistry The chosen paraffin block for each case was cut into 4 m sections, which were mounted on sialynated glass slides in the standard fashion. If one of the blocks chosen for immunohistochemistry did not contain all the tumor components identified on the initial H&E stained slide, more blocks were cut, stained, and analyzed. Immunohistochemistry for p53 The sections were deparaffinized and endogenous peroxidase was blocked with 0.6% hydrogen peroxide in methanol. Antigen retrieval was achieved by microwave heating for 15 minutes in 1mM citrate buffer. Immunologic detection was performed using a commercially available anti-p53 rabbit polyclonal antibody at a dilution of 1:80 (Signet Laboratories, Dedham, MA). Binding of the antibody was visualized using a commercial streptavidin-biotin-peroxidase detection system (Vectastain Elite kit; Vector Laboratories, Burlingame, CA) as recommended by the manufacturer. Only nuclear staining was interpreted as positive. Colon carcinoma with known p53 mutation was used as a positive control, and negative controls consisted of benign testicular tissue (2 epidermoid cysts, 1 adenomatoid tumor, 1 Leydig cell tumor, 1 gonadoblastoma, and 17 testes surgically removed because of testicular atrophy due to cryptorchidism). Immunohistochemistry for bcl-2 The sections were deparaffinized and endogenous peroxidase was blocked with 0.6% hydrogen peroxide in methanol. Antigen retrieval was achieved by microwave heating for 15 minutes in 1mM citrate buffer. Immunologic detection involved the use of a commercially available anti-bcl-2 mouse monoclonal antibody at a dilution of 1:40, (DAKO, Glostrup, Denmark). Binding of the antibody was visualized using a commercial streptavidin-biotin-peroxidase detection system (Vectastain Elite kit, Vector Laboratories) as recommended by the manufacturer. According to the mitochondrial location of bcl-2, only cytoplasmic staining was interpreted as positive. Positive bcl-2 staining of lymphocytes served as a positive internal control. Immunohistochemistry for MIB-1 The sections were deparaffinized and endogenous peroxidase was blocked with 0.6% hydrogen peroxide in methanol. Antigen retrieval was achieved by microwave heating for 3 5 minutes at 750 W in 10mM citrate buffer (ph 6.0). Immunologic detection involved the use of a commercially available anti-ki-67 (clone MIB-1) mouse monoclonal antibody at a dilution of 1:50 (AMAC, Inc., Westbrook, ME). Binding of the antibody was visualized using a commercial streptavidin-biotin-peroxidase detection system (Vectastain Elite kit, Vector Laboratories) as recommended by the manufacturer. Only nuclear staining was interpreted as positive. Human tonsil tissue served as a positive control, and mouse IgG1 irrelevant primary antibody served as a negative control. Immunohistochemistry for cathepsin D The sections were deparaffinized and endogenous peroxidase was blocked with 0.6% hydrogen peroxide in methanol. Immunologic detection involved the use of a commercially available anti-cathepsin D mouse monoclonal antibody that identified both the 34kD and the 48kD forms of cathepsin D (dilution, 1:5 in TBS; Triton Diagnostics, Alameda, CA). Binding of the antibody was visualized using a commercial streptavidin-biotin-peroxidase detection system (Vectastain Elite kit, Vector Laboratories) as recommended by the manufacturer. Cathepsin D staining of macrophages served as a positive internal control, and a negative control was created by omitting the primary antibody. Immunohistochemistry for E-cadherin The sections were deparaffinized and endogenous peroxidase was blocked with 0.6% hydrogen peroxide in methanol. Antigen retrieval was achieved by heating the slides at 90 C on a heating device for 15 minutes in 1mM citrate buffer. Tissue sections were pretreated with tissue unmasking fluid (Signet, Laboratories) as recommended by the manufacturer. Immunologic detection involved the use of a commercially available anti-e-cadherin mouse monoclonal antibody at a dilution of 1:20 (Transduction Laboratories, Lexington, KY). Binding of the antibody was visualized using a commercial streptavidin-biotin-peroxidase detection system (Vectastain Elite kit, Vector Laboratories) as recommended by the manufacturer. Positive E-cadherin staining of benign testicular tissue served as a positive internal control, and omission of the primary antibody served as a negative control.

4 Prognostic Risk Factors in Clinical Stage I NSGCT/Heidenreich et al Assessment of Staining Evaluation of each immunohistochemical marker involved counting positively stained tumor cells at 400 and calculating the percentage of positively stained cells in the total number of tumor cells. Counting was performed in at least 500 cells from the area of the most intense staining in each cell type. Tumors were scored semiquantitatively for intensity of staining on a 5-point scale: 0 no staining, %, %, %, and % of malignant cells staining. Statistical Analysis To determine statistically significant risk factors associated with pathologic Stage II disease, logistic regression was used. Initially, univariate logistic regression models were developed to determine whether immunohistochemical staining of EC, %EC, and presence of VI were each significant risk factors. Multivariate logistic regression models were developed that considered all possible combinations of EC staining, %EC, and VI. Logistic regression diagnostics and assessment of sensitivity, specificity, and positive and negative predictive values were used to confirm that the best set of risk factors was determined by multivariate logistic regression procedures. The condition best was satisfied if the resulting set of multivariate risk factors passed the diagnostic tests as defined by Hosmer and Lemeshow and produced high sensitivity, high specificity, and high positive and negative values for predicting pathologic Stage I and II. In addition, cutoff values for variables demonstrated to be significant by multivariate analysis were used to predict pathologic stage and calculate sensitivity, specificity, and positive and negative predictive values. RESULTS Quantitative Histopathology In univariate and multivariate analyses, %EC and the presence of VI were statistically significant risk factors predicting pathologic Stage IIA/B in CSI NSGCT. Table 1 shows the median quantitative histology, presence or absence of VI, and tunic invasion and a combination of both parameters for pathologic Stage I versus pathologic Stage II. Tables 2 and 3 show the cutoff values for %EC and presence or absence of VI and their probability of predicting final pathologic stage. VI had the highest probability of predicting lymph node involvement (P ) among all variables, with a positive predictive value of 78.5%. %EC correctly predicted pathologic Stage II disease in 77.2% of TABLE 1 Frequency of Germ Cell Components in Pathologic Stage I and II Primary Nonseminomatous Germ Cell Tumors n Stage I Stage II Clinical Stage I NSGCT (57.8%) 63 (31.2%) Vascular invasion 67 9 (11%) 58 (92%) Tunic invasion 35 9 (10.4%) 26 (41.3%) % Embryonal carcinoma % 24.2% 81.9% 28.4% Embryonal carcinoma (52%) 58 (48%) Mixed germ cell tumor (48%) 47 (52%) Yolk sac tumor (67%) 30 (32%) Mature teratoma (72%) 24 (27%) Immature teratoma (75%) 5 (25%) Seminoma (62%) 10 (37%) TABLE 2 Cutoff Values for Percentage of Embryonal Carcinoma and Presence/ Absence of Vascular Invasion and Their Probability of Predicting Final Pathologic Stage in Clinical Stage I Nonseminomatous Germ Cell Tumors Quantitative pathology n I Pathologic stage I Staging correct 45% EC 77 68/77 9/77 88% 45% EC/VI 71 65/71 6/ % 46 79% EC 24 9/24 15/ % EC/VI 9/24 7/9 2/9 78% 46 79% EC/VI 15/24 0/15 15/15 100% 80% EC 48 7/48 41/ % 80% EC/VI 42 5/42 37/42 88% Total (57.8%) 63 (42.2%) 131/149 (87.9%) EC: embryonal carcinoma; VI : absence of vascular invasion; VI : presence of vascular invasion. TABLE 3 Sensitivity, Specificity, and Positive and Negative Predictive Values for Defined Cutoff Values in Predicting Final Pathologic Stage Quantitative pathology Sensitivity Specificity the cases (P ). The combination of both parameters correctly predicted occult retroperitoneal metastasis in 84.2% of the patients. The negative predictive value for the absence of VI was 78.9%. Based on these data, optimal cutoff values were developed for %EC and VI to predict pathologic stage in CSI NSGCT. Using a cutoff point of 80% EC in the II Predictive value Positive Negative 45% EC/VI 91.5% 93.2% 96% 91.1% 46 79% EC/VI or 46 79% EC/VI 100% 78% 85.7% 100% 80% EC/VI 89.7% 92.5% 85.4% 92.5% EC: embryonal carcinoma; VI : absence of vascular invasion; VI : presence of vascular invasion.

5 1006 CANCER September 1, 1998 / Volume 83 / Number 5 TABLE 4 Immunohistochemical Expression of p53, bcl-2, Cathepsin D, E-Cadherin, MIB-1, Vascular Invasion, and Percentage of Embryonal Carcinoma in the Primary Tumor Specimens of 149 Clinical Stage I NSGTs by Final Pathologic Stage I Pathologic stage II Univariate analysis Multivariate analysis Vascular invasion 9 (11%) 58 (92%) P P % Embryonal carcinoma % % P P p53 expression P 0.03 n. s. bcl-2 expression n. s. not done MIB-1 expression n. s. not done Cathepsin D expression P 0.11 n. s. E-cadherin expression P 0.04 n. s. n. s.: not significant. primary tumor specimen, 41 of 48 patients (85.4%) were correctly classified as having pathologic Stage II disease. The combination of 80% EC and presence of VI slightly increased the correct classification of patients with pathologic Stage II to 87%. When cutoff values of 45% EC were applied to the primary tumor specimens, 68 of 77 patients (88%) were correctly classified as having pathologic Stage I disease; the combination of 45% EC and absence of VI increased the negative predictive value to 91.5%. Twenty-four patients (16%; 9 pathologic Stage I, 15 pathologic Stage II) revealed 46 79% EC in their primary germ cell tumors and could not be staged by quantitative histopathology alone. However, when the presence or absence of VI was included, pathologic stage could be predicted for 22 of these 24 patients (92%). Seven of 9 patients (78%) with absence of VI were correctly staged as having pathologic Stage I disease, whereas all 15 patients with presence of VI were correctly predicted as having pathologic Stage II disease. Overall, true pathologic stage could be predicted for 86.7% (131 of 151) of all CSI NSGCTs by applying cutoff values for %EC combined with absence or presence of VI. p53 Expression Overall, 142 of 149 NSGCTs (95%) demonstrated at least some nuclear p53 protein expression. All benign testicular tumors (2 epidermoid cysts, 1 adenomatoid tumor, 1 Leydig cell tumor, and 1 gonadoblastoma) and all 17 testes without evidence of malignancy demonstrated absence of p53 expression. Positive p53 staining was observed in all components of the nonseminomatous tumors, such as mature teratoma, yolk sac tumor, choriocarcinoma, and EC. However, only p53 expression in EC showed a significant difference between pathologic Stage I and pathologic Stage II disease (P 0.03, Table 4) in univariate analysis. Based on multivariate logistic regression analysis, %EC and VI remained the only significant prognosticators for prediction of lymph node status in CSI NSGCT. bcl-2 Expression Overall, bcl-2 protein expression was observed in 14 of 149 NSGCTs (9.3%). Thirteen of 14 nonseminomatous tumors demonstrated bcl-2 expression only in the glandular and stromal elements of their teratomatous components, whereas embryonal carcinomas, yolk sac tumors, and choriocarcinomas were always negative for bcl-2 (Table 4). Lymphocytes used as internal positive controls always exhibited strong (3 4 ) bcl-2 expression. Benign testicular tumors were also negative for bcl-2; however, Leydig cells occasionally showed low (1 ) bcl-2 expression, whereas positive bcl-2 staining was found in cells of the rete testis (2 ) and epididymal cells lining the efferent ducts (1 2 ). Based on univariate analysis, there was no significant difference in bcl-2 expression between pathological Stage I and II disease (Table 2). Cathepsin D Expression Overall, cathepsin D expression was detected in all NSGCTs and in 117 of 120 EC (97.5%) (Table 4). By stage, mean cathepsin D expression ( standard deviation) in EC components was in pathologic Stage I disease and in pathologic Stage II disease (P 0.11). Total cathepsin D expression in pathologic Stage I NSGCT was ; in pathologic Stage II NSGCT, it was (P 0.25). It is noteworthy that 59 of 67 tumor cells (88.1%)

6 Prognostic Risk Factors in Clinical Stage I NSGCT/Heidenreich et al invading blood or lymphatic vessels were found to exhibit high cathepsin D expression of Grade 3 4. Associated testicular carcinoma in situ (CIS) was present on the same section in 52 of 149 NSGCTs (34.2%); 9 of 52 cases of CIS (17.3%) were negative for cathepsin D, whereas in 43 of 52 cases (82.7%), CIS cells demonstrated at least some degree of cathepsin D expression. Extratubular invasion of the adjacent testicular tissue was present in 4 of 52 cases of CIS (7.3%), and all 4 CIS cases demonstrated positive cathepsin D staining in both the intratubular and the extratubular components. CIS was defined as the presence of intratubular atypical germ cells. The CIS cells are larger than normal spermatogonia and have a large hyperchromatic nucleus containing several prominent nucleoli. CIS cells are typically located close to the basement membrane, although invasion of the interstitial tissue has been described as a phenomenon of early invasive growth of undifferentiated germ cell tumors. In normal tissue adjacent to the germ cell tumor, positive cathepsin D expression was found in Leydig cells and Sertoli cells (1 2 ) and in primary and secondary spermatocytes (1 ). E-Cadherin Expression Seventy-eight of 149 CSI NSGCTs (51.6%) could be evaluated for E-cadherin expression, and it was expressed in 59 of 78 NSGCTs (76.1%) (Table 4). Nineteen of 78 CSI NSGCTs (23.9%) were negative for E-cadherin expression in the presence of positive staining for benign testicular tissue (internal control); 17 of 19 CSI NSGCTs (91%) with loss of E- cadherin expression were staged as pathologic Stage II disease, and 2 of 19 (9%) were pathologic Stage I (P 0.001). Overall, mean E-cadherin expression in pathologic Stage I disease was compared with in pathologic Stage II disease. Mean E-cadherin expression for EC was and in pathologic stage I and II disease, respectively (P 0.04). MIB-1 Expression MIB-1 expression was detected in all tumor specimens examined. Associated testicular CIS was present on the same section in 52 of 149 NSGCTs (34.9%); MIB-1 expression was observed in 49 of 52 CIS (95.4%). There was no statistically significant difference in MIB-1 expression between CIS associated with pathologic Stage I ( %) or pathologic Stage II ( %) disease (P 0.161). In univariate analysis, MIB-1 expression was not useful for correctly predicting pathologic Stage II disease in CSI NSGCT patients (Table 4), nor was MIB-1 expression in EC useful for predicting occult lymph node disease ( % for pathologic Stage I vs % for pathologic Stage II, P 0.979). Neither total MIB-1 expression nor MIB-1 scores in the highest area of staining were able to predict pathologic Stage I disease correctly. Prospective Evaluation of Quantitative Histopathology in Clinical Stage I NSGCT Our results demonstrated that %EC in combination with the presence or absence of VI represents the best model for predicting microscopic lymph node involvement in clinical Stage I NSGCT. To study prospectively the clinical usefulness of our newly derived cutoff values for %EC ( 45% for pathologic Stage I and 80% for pathologic Stage II), we attempted to predict the pathologic stage for 10 consecutive CSI NSCGT patients prior to RPLND. Pathologic stage was correctly predicted for 5 of 6 Stage I and 4 of 4 Stage II patients. DISCUSSION Clinical Stage I nonseminomatous testicular cancer is highly curable with RPLND, 1 surveillance, 2 or primary chemotherapy. 3 Identification of reliable prognostic risk factors for patients with CSI NSGCT remains one of the most important and challenging issues in assigning patients to the best therapeutic options according to their individual risk profiles for metastasis. 4,20 In our current study, we systematically evaluated histopathologic features, such as %EC and VI, as potentially useful prognostic markers for predicting occult lymph node disease in patients with clinical Stage I NSGCT, and compared them with the immunohistochemical expression of several proliferation markers, proteases, and cell- to cell adhesion molecules shown to be significant prognosticators in a number of human malignancies. The most significant finding of our study was that meticulous histopathologic evaluation of the primary tumor specimen, including quantitative examination of %EC and presence or absence of VI, accurately predicted not only pathologic Stage II but also pathologic Stage I disease. We were able to develop reliable cutoff points for %EC that predicted pathologic stage with a 90% accuracy, thereby making possibly a riskadapted therapeutic approach to CSI NSGCT. Eightyeight percent of our patients with 45% EC in their primary germ cell tumors were correctly classified as pathologic Stage I, and only 9 patients were misdiagnosed. The combination of 45% EC and the absence of VI even improved the negative predictive value to 91.5%, so that only a minority of patients would have

7 1008 CANCER September 1, 1998 / Volume 83 / Number 5 been understaged. If these numbers were extrapolated to a clinical setting, only 9% of clinical Stage I NSGCT patients might have had a retroperitoneal relapse if assigned to surveillance based on their quantitative pathology results. With regard to occult retroperitoneal disease, 84% of CSI patients were correctly classified as having pathologic Stage II disease if their primary NSGCT harbored 80% EC; the combination of 80% EC and presence of VI increased positive predictive values to 87%, underscoring the importance of EC as a prognostic risk factor in CSI disease. Therefore, only 13% of all CSI NSGCT patients would have undergone unnecessary RPLND, versus 70% of the patients if other clinical staging modalities were applied. 5 Although these data must be confirmed in a prospective randomized clinical trial, the 100% positive prediction rate for our first 10 prospective consecutive CSI NSGCT patients to whom we applied our cutoff points is extremely encouraging. However, 16% (24 patients) of our cohort with %EC ranging from 46% to 79% represented a problem with regard to adequate prediction of pathologic stage and were referred to as the gray zone. When presence or absence of VI were included in determining patients individual risk profiles, 21 of 24 (88%) could be correctly staged as having pathologic Stage I or II disease. Altogether, pathologic stage could be correctly predicted for 131 of 149 (88%) CSI NSGCT patients based on careful and meticulous histopathologic analysis of the primary germ cell tumor. The mere presence of EC and the presence of VI have been cited as significant risk factors for occult retroperitoneal disease in CSI NSGCT since the early 1980s, when Peckham et al. 3 presented their initial data on the surveillance of CSI NSGCT patients. Relapse rates were significantly higher (42.8%) for patients with undifferentiated pure malignant teratoma (the British equivalent of EC) than for those with teratocarcinoma (3.4%). These preliminary data were later confirmed by Freedman et al., 8 who demonstrated that VI, absence of yolk sac tumor, and presence of EC were the most significant risk factors in multivariate analysis for predicting relapse of CSI NS- GCT. The importance of EC in the development of metastatic disease was also demonstrated in two prospective clinical trials, which demonstrated that relapse rates among patients with EC components were significantly higher than among patients without EC. Recently, Moul et al. 14 and the Testicular Cancer Intergroup Study 13 demonstrated with quantitative assessment of EC that the combination of %EC and VI are the most important prognosticators in CS I NS- GCT, predicting true pathologic Stage II disease in 86% of patients. Similar results with regard to the definition of a high risk group of CSI NSGCT were obtained by Wishnow et al., 12 who evaluated the outcomes of 82 patients under surveillance after radical orchiectomy. The authors demonstrated that 80% EC, VI, and a serum -fetoprotein level 80 ng/dl were significant risk factors for relapse. The authors concluded that patients with 80% EC and absence of VI might represent a subgroup of CSI NSGCTs with a very low rate of relapse who should undergo surveillance instead of treatment. Contrary results were obtained by Klepp et al., 21 who conducted the largest study known to us of prognostic risk factors in clinical Stage I NSGCT but could not confirm the importance of EC in occult retroperitoneal disease. However, a careful analysis of their data showed that EC still was a significant risk factor for pathologic Stage II disease in univariate analysis, with 39% of patients harboring pure EC (P 0.021), 30% with combined NSGCT containing EC, and 18% of patients without EC elements classified as pathologic Stage II. VI was shown to be a significant risk factor in univariate and multivariate analysis. That study, however, may have been limited by the problems of recognizing EC elements. Klepp et al. 21 noted EC elements in 211 of 279 cases (76%), whereas the Testicular Cancer Intergroup Study 13 noted EC in more than 90% of cases on careful review. More importantly, Klepp et al. 25 noted yolk sac tumor elements in only 17 of 279 cases (23%), which was significantly different from the results obtained in our current study (69%) and the data given by the Testicular Cancer Intergroup Study, 13 with a reported frequency of yolk sac tumor elements in 79% of pathologic Stage I and in 66% of pathologic Stage II patients. This area of disagreement might have resulted from the inclusion of a reticular growth pattern as a characteristic of EC, which is now recognized as diagnostic of a yolk sac tumor. Therefore, the frequency of EC might have been incorrectly overestimated, with inclusion of more, false positive ECs in statistical analysis, resulting in failure to demonstrate the importance of EC. Furthermore, the authors only took into account the mere presence or absence of EC and did not include semiquantitative or even quantitative analysis of %EC, which has been demonstrated to be superior in the evaluation of the tumor s biologic behavior. 14,18 These data underscore the difficulties in recognizing and distinguishing yolk sac tumor from EC, and demonstrate that the reported incidence of yolk sac tumor elements is dependent on the awareness of the individual pathologist. The subjectiveness and interpretation in identifying and quantitating the different germ cell tumor components underscore the

8 Prognostic Risk Factors in Clinical Stage I NSGCT/Heidenreich et al need for a cell-type specific stain that can quantify EC more objectively. 22 Recent advances in molecular and cellular biology have led to the identification of multiple proliferation and cell cycle regulating markers, proteases, cell-tocell adhesion molecules, and other important cellular proteins. Because it is conceivable that one or a series of these markers could be clinically useful as prognostic risk factors in CSI NSGCT, we evaluated a number of these proteins in our current study. The p53 tumor suppressor gene has become one of the most intensively studied genetic markers in human malignancies, with mutations in the highly conserved exons 4 8 commonly observed in a variety of human cancers. In the current study, we found aberrant p53 expression in all histopathologic constituents of adult testicular germ cell tumors, including EC, yolk sac tumors, mature and immature teratoma, and choriocarcinoma. For CSI NSGCT, the difference in p53 and EC expression between pathologic Stages I and II was statistically significant in univariate analysis. However, in multivariate analysis it was determined that %EC and VI represented the best model for predicting occult retroperitoneal disease. These results were consistent with those of other recent studies, demonstrating that immunohistochemical examination of p53 expression is not clinically useful in determining occult disease. 19,23 Four independent groups, including ours, recently found no p53 gene mutations among 100 testicular tumors, which is intriguing considering the immunohistochemical overexpression of p53. Lack of p53 mutations and presence of functional wild-type p53 in in vitro studies 27 (increase of the p53-dependent gene WAF-1 after etoposide treatment of testicular cancer cell lines) suggest that other mechanisms, such as posttranslational abnormalities, stabilize p53 protein for immunohistochemical detection. bcl-2 is a 26 kd proto-oncogene protein product that was first reported to be associated with t(14;18) translocations in follicular B cell lymphomas. 28 Evidence from in vitro studies suggests that bcl-2 blocks apoptosis rather than enhancing cell proliferation. Because in vitro studies of breast epithelial cells have indicated that bcl-2 overexpression does not affect the expression of the p53 gene but may inhibit p53 functional activity, we evaluated bcl-2 expression in human testicular germ cell tumors; as yet, the presence of bcl-2 has not been investigated and described in testicular tumors. We could not observe detectable bcl-2 protein expression in pathologic Stage I or II NSGCT; seminomatous components in mixed NSGCT were also bcl-2 negative. These data were consistent with recent in vitro studies of testicular cancer cell lines, which demonstrated high basal levels of p53 and bax but no detectable bcl-2 protein. Fifty percent of mature teratomatous elements of the nonseminomatous tumors expressed bcl-2 protein to a certain extent in our series; although there was no difference in bcl-2 expression between pathologic stages I and II, these data might still have clinical use in evaluating testicular cancer. The high susceptibility of testicular germ cell tumors to drug-induced apoptosis appears to be the result of the overexpression of wild-type p53 and bax as well as the lack of bcl-2 protein expression. 27 The resistance of mature teratomas to DNA-damaging agents might be reflected by the immunohistochemical expression of p53 and bcl-2; 80% of mature teratomas were either p53 negative or exhibited rare p53 staining in the presence of strong bcl-2 expression, whereas other histopathologic constituents demonstrated a strong p53 and negative bcl-2 expression. Therefore, we believe that p53 and bcl-2 might play a role in the course of disease in those testicular cancer patients who develop chemoresistance during the management of their disease, and it might be useful to evaluate p53 and bcl-2 expression in their primary tumors and metastatic locations. Ki-67 is a bimolecular protein expressed in the nucleus during the entire cell cycle from late G 1 phase to M phase, except G 0 phase. 29 Ki-67 expression has correlated well with the prognoses of a number of human solid tumors, such as breast, 30 colorectal, 31 and prostate carcinomas. 32 Two studies examining Ki-67 expression in germ cell tumors have been performed to date. Although Due et al. 33 found a slight tendency toward a greater growth fraction in seminomas at advanced histopathologic stages, they did not demonstrate a positive correlation between growth fraction and tumor size. Recently, determination of immunohistochemical expression of MIB-1 has been advocated to predict pathologic Stage I accurately for CSI NSGCT patients. 17 Using a combined approach of MIB-1 expression ( 80%) and the absolute volume of EC ( 2 ml), a group of patients at low risk for occult metastatic disease was defined. However, we were not able to confirm these data in this study or in previous studies, 18,20 and MIB-1 scores in the overall primary neoplasm and in EC components were not able to predict pathologic Stage I or II disease correctly. Therefore, we do not believe that immunohistochemical analysis of primary testicular germ cell tumors would add clinically useful prognostic information. The serine protease cathepsin D, which is an aspartyl proteolytic enzyme with a molecular weight of 52 kd, is highly expressed in some breast adenocarcinomas and has been correlated with worse prognosis

9 1010 CANCER September 1, 1998 / Volume 83 / Number 5 among some but not all breast carcinomas. 34 Because this and other proteases might be involved in the cellular biology of metastasis, cathepsin D could also be a prognostic marker in CSI NSGCT. Our study did not demonstrate a significant difference in mean overall and EC specific cathespin D expression between pathologic Stage I and II cancers. However, we were able to demonstrate that 88% of tumor cells invading blood and lymphatic vessels exhibited high cathepsin D expression; therefore, cathepsin D expression might be clinically useful in determining the risk of metastasis in NSGCT without EC components or of pure seminomas. The cadherin family is a group of genetically related transmembrane glycoproteins (E-, P-, and N- cadherins) that are involved in a calcium (Ca)-dependent cell-to-cell adhesion mechanism. 35 It has been suggested that loss of E-cadherin plays a critical role in the progression of cancer because reduced intercellular adhesion might promote metastasis by facilitating the invasion of cancer cells into vessels and by causing detachment of cancer cells from any tumor protruding into the vessel. 36 Recently, it has been demonstrated that E-cadherin expression is lost in a subset of advanced prostatic carcinomas, breast adenocarcinomas, and squamous cell carcinomas of the head and neck; 22 with regard to testicular germ cell tumors, we know of no studies. In this study, E-cadherin was shown to be expressed in most (76%) of the testicular cancer specimens examined independently of pathologic stage. We also encountered a few NSGCTs (24%) that apparently not only lacked E-cadherin expression immunohistochemically but also lacked intercellular adhesion among cancer cells; most interestingly, more than 90% of these CSI NSGCTs demonstrated microscopic lymph node disease on RPLND. Loss of E- cadherin might be considered to free a subset of NS- GCTs from their tight intercellular adhesions, thereby enhancing their invasiveness. These considerations are underscored by the fact that most intravascular tumor cells were also E-cadherin negative. CONCLUSIONS We were able to demonstrate that %EC and VI are clinically useful histopathologic parameters in identifying clinical Stage I NSGCT patients at low risk and high risk for retroperitoneal disease. We predicted that 92% of patients with clinical Stage I nonseminomatous tumors who had 45% EC and absence of VI would have pathologic Stage I disease and could be considered for surveillance protocols. In contrast, 86% of patients with clinical Stage I nonseminomatous tumors who had 80% EC and presence of VI harbored microscopic retroperitoneal lymph node disease; these patients should be treated by nerve-sparing RPLND or primary chemotherapy. Although p53, bcl-2, MIB-1, cathepsin D, and E-cadherin are clinically useful prognostic risk factors in some human malignancies, they do not appear to be of prognostic significance in CSI NSGCT compared with quantitative histopathology and assessment of VI. Based on the data from our current study, we suggest that all CSI NSGCT patients have their primary tumor pathology material evaluated for %EC and the presence or absence of VI and the approach to their therapy modified accordingly if necessary. REFERENCES 1. Donohue JP, Thornhill RS, Foster RS, Rowland RG, Bihrle R. Primary retroperitoneal lymph node dissection in clinical stage A nonseminomatous germ cell testis cancer. Br J Urol 1993;71: Culine S, Theodore C, Terrier-Lacombe MJ, Droz JP. Primary chemotherapy in patients with nonseminomatous germ cell tumors of the testis and biological disease only after orchiectomy. J Urol 1996;155: Peckham MJ, Barrett A, Husband JE, Hendry WF. Orchiectomy alone in testicular stage I nonseminomatous germ cell tumors. Lancet 1982;2: Moul JW, Heidenreich A. Prognostic factors in low stage nonseminomatous testicular cancer. Oncology 1996;10: Fujime M, Chang H, Lin CW, Prout GR. Correlation of vascular invasion and metastasis in germ cell tumors of the testis: a preliminary report. J Urol 1984;131: Hoskin P, Dilly S, Easton D, Horwich A, Hendry WF, Peckham MJ. Prognostic factors in stage I nonseminomatous germ cell testicular tumors managed by orchiectomy and surveillance: implications for adjuvant chemotherapy. J Clin Oncol 1986;7: Javadpour N, Young JD. Prognostic factors in nonseminomatous testicular cancer. J Urol 1986;135: Freedman LS, Jones WG, Peckham MJ, Newlands ES, Parkinsom MC, Oliver RTD, et al. Histopathology in the prediction of relapse of patients with stage I testicular teratoma treated by orchiectomy alone. Lancet 1987;2: Hoeltl W, Kosak D, Pont J, Hawel R, Machacek E, Schempfer M, et al. Testicular cancer: prognostic implications of vascular invasion. J Urol 1987;137: Dunphy CH, Ayala AG, Swanson D, Ro JY, Logothesis C. Clinical stage I nonseminomatous and mixed germ cell tumors of the testis. Cancer 1988;62: Fung CY, Kalish LA, Brodksy GL, Richie JP, Garnick MB. Stage I nonseminomatous germ cell testicular tumor: prediction of metastatic potential by primary histopathology. J Clin Oncol 1988;6: Wishnow KI, Johnson DE, Swanson DA, Tenney DM, Babaian RJ, Dunphy CH, et al. Identifying patients with low-risk clinical stage I nonseminomatous testicular tumors who should be treated by surveillance. Urology 1989;34: Sesterhenn IA, Weiss RB, Mostofi FK, Stablein DM, Rowland RG, Falkson G, et al. Prognosis and other clinical correlates of pathologic review in stage I and II testicular carcinoma: a report from the Testicular Cancer Intergroup Study. J Clin Oncol 1992;10:69 78.

10 Prognostic Risk Factors in Clinical Stage I NSGCT/Heidenreich et al Moul JW, McCarthy WF, Fernandez EB, Sesterhenn IA. Percentage of embryonal carcinoma and of vascular invasion predicts pathological stage in clinical stage I nonseminomatous testicular cancer. Cancer Res 1994;54: Moul JW, Foley JP, Hitchcock CL, McCarthy WF, Sesterhenn IA, Becker RL, et al. Flow cytometric and quantitative histological parameters to predict occult disease in clinical stage I nonseminomatous testicular germ cell tumors. J Urol 1993; 150: Fernandez EB, Sesterhenn IA, McCarthy WF, Mostofi FK, Moul JW. Proliferating cell nuclear antigen expression to predict occult disease in clinical stage I nonseminomatous testicular germ cell tumors. J Urol 1994;152: Albers P, Ulbright TM, Albers J, Miller GA, Orazi A, Crabtree W, et al. Tumor proliferative activity is predictive of pathological stage in clinical stage A nonseminomatous testicular germ cell tumors. J Urol 1996;155: Heidenreich A, Schenkmann NS, Sesterhenn IA, Mostofi FK, McCarthy WF, Heidenreich B, et al. Immunohistochemical expression of Ki-67 to predict lymph node involvement in clinical stage I noseminomatous germ cell tumors. J Urol 1997;158: Lewis DJ, Sesterhenn IA, McCarthy WF, Moul JW. Immunohistochemical expression of p53 tumor suppressor gene protein in adult testicular germ cell testis tumors: clinical correlation in stage I disease. J Urol 1994;152: Heidenreich A, Sesterhenn IA, Moul JW. Prognostic risk factors in low stage testicular germ cell tumors. Cancer 1997;79: Klepp O, Olsson AM, Henrikson H, Aass N, Dahl O, Stenwig AE, et al. Prognostic factors in clinical stage I nonseminomatous germ cell tumors of the testis: multivariate analysis of a prospective multicenter study. J Clin Oncol 1990;8: DeRiese WTW, Orazi A, Foster RS, Donohue JP, Messemer JE. The clinical relevance of p53 expression in early stage nonseminomatous germ cell tumors. J Urol 1993;149:311A. 23. Fleischhacker M, Strohmeyer T, Imai Y, Slamon DJ, Koeffler HP. Mutations of the p53 gene are not detectable in human testicular tumors. Modern Pathol 1994;7: Strohmeyer T, Fleischhacker M, Imai Y, Slamon DJ, Koeffler HP. Status of the p53 tumor suppressor gene and the MDM2-gene in human testicular tumors. J Urol 1993;149: 311A. 25. Peng HQ, Hogg D, Malkin D, Bailey D, Gallie BL, Bulbul M, et al. Mutations of the p53 gene do not occur in testis cancer. Cancer Res 1993;53: Schenkman NS, Sesterhenn IA, Washington L, Tong YA, Weghorst C, Buzard GS, et al. Increased p53 protein does not correlate to p53 gene mutations in microdissected human testicular germ cell tumors. J Urol 154: Chresta CM, Masters JRW, Hickman JA. Hypersensitivity of human testicular tumors to etoposide induced apoptosis is associated with functional p53 and a high bax:bcl-2 ratio. Cancer Res 1996;56: Tsujimoto Y, Cossmann J, Jaffe E, Croce CM. Involvement of the bcl-2 gene in human follicular lymphoma. Science 1985; 228: Gerdes J, Lemke H, Baisch H, Wacher HH, Schwab U, Stein H. Cell cycle analysis of a cell proliferation associated human nuclear antigen defined by the monoclonal antibody Ki-67. J Immunol 1984;133: Leonardi E, Girlando S, Serio G, Mauri FA, Perrone G, Scamoini S, et al. PCNA and Ki-67 expression in breast carcinoma: correlation with clinical and biological variables. J Clin Pathol 1992;45: Kubota Y, Petras RE, Easley KA, Bauer TW, Tubbs RW, Fazio VW. Ki-67 determined growth fraction versus standard staging and grading parameters in colorectal carcinoma: a multivariate analysis. Cancer 1992;70: Sadi MV, Barrack ER. Determination of growth fraction in advanced prostate cancer by Ki-67 immunostaining and its relationship to the time to tumor progression after hormonal therapy. Cancer 1991;67: Düe W, Dieckmann KP, Loy V. Immunohistological determinations of proliferative activity in seminomas. J Clin Pathol 1988;41: Tandon AK, Clark GM, Chamness GC, Chrigwin JM, McGuire WL. Cathepsin D and prognosis in breast cancer. N Engl J Med 1990;322: Takeichi M. Cadherin cell adhesion receptors as a morphogenetic regulator. Science 1991;251: Jiang WG. E-cadherin and its associated protein catenins, cancer invasion and metastasis. Br J Surg 1996;83: