THE TESTICULAR Cancer Intergroup Study (TCIS)

Prognosis and Other Clinical Correlates of Pathologic Review in Stage I and II Testicular Carcinoma: A Report From the Testicular Cancer Intergroup Study By Isabell A. Sesterhenn, Raymond B. Weiss, F. Kash Mostofi, Donald M. Stablein, Randall G. Rowland, Geoffrey Falkson, Saul E. Rivkind, and Nicholas J. Vogelzang Purpose: The Testicular Cancer Intergroup Study (TCIS) was undertaken to evaluate the pathologic findings in early-stage testicular cancer as determined by central pathology review, to compare these findings with the interpretation by the contributing pathologists, and to make correlations with various clinical parameters and outcomes. Patients and Methods: The prospective study of nonseminomatous germ cell testicular cancer staged surgically involved 459 eligible patients with stage I (nodenegative) or stage II (node-positive) disease. Pathologic materials from both the orchiectomy and lymphadenectomy specimens were submitted to a central laboratory for evaluation. Results: Central and local pathologists differed significantly in their identification of certain cellular histologies (primarily yolk sac tumors [YST]) and recognition of invasion into vascular structures. In contrast to our prior findings with local pathologic assessment, venous/ lymphatic invasion as determined by central review predicted relapse in both stages. In pathologic stage I disease, the relapse rate was 19.4% (12 of 62 cases) for THE TESTICULAR Cancer Intergroup Study (TCIS) was begun in 1979 to evaluate the efficacy of adjuvant chemotherapy versus observation with chemotherapy at relapse in pathologic stage II, nonseminomatous, germ cell testicular tumors (NSGCTs). In addition, patients with pathologic stage I disease were observed, with chemotherapy given if relapse occurred. The results of staging, therapy, failures, and survival have been reported elsewhere. 1 ' In this report we present the pathologic findings as determined by the central laboratory, compare them with the interpretations of the contributing pathologists, correlate the pathology features with immunohistochemical markers, and evaluate pathologic risk factors for tumor recurrence. In addition, we define the relationships between serum tumor markers, pathologic findings, and immunocytochemical results. This project was a multicenter, international study, and central review of pathology slides of both the orchiectomy and retroperitoneal lymphadenectomy specimens was an integral part of the study protocol. Central pathology assessment was required because there are those with invasion versus 6.0% (10 of 168 cases) for those without invasion. In pathologic stage II disease, the respective relapse rates were 63.5% (40 of 63 cases) and 24.0% (six of 25 cases). Vascular invasion was jointly predictive with nodal stage for risk of relapse. The percentage of embryonal carcinoma (EC) in the primary tumor was predictive of nodal stage and relapse in a univariate, but not a multivariate, analysis. In a large substudy, immunohistochemical staining identified a correlation between stain intensity in YST and serum alphafetoprotein (AFP) levels. In a similar fashion human chorionic gonadotropin (HCG) staining reactivity occurred exclusively in patients with syncytiotrophoblasts and correlated with serum levels of p3-hcg. Conclusions: A number of tumor histology correlates with clinical parameters have been identified or confirmed. Careful pathologic scrutiny of the primary testicular tumor, especially for vascular invasion, provides important prognostic information. J Clin Oncol 10:69-78, 1992. This is a US government work. There are no restrictions on its use. multiple pathologic classifications of testicular germ cell tumors in existence. The oldest is the Friedman and Moore classification, 6 which was based on about 1,000 testicular tumors collected from United States Army soldiers in World War II. This classification recognized four entities: seminomas (S), embryonal carcinomas (EC), teratomas (T), and choriocarcinomas (CC). Any combination of these was classified as teratocarcinoma. Dixon and Moore 7 later secured follow-up information on these patients and divided the tumors into five From the Armed Forces Institute of Pathology, Washington; Walter Reed Army Medical Center, Washington, DC; Uniformed Services University of the Health Sciences, Bethesda, MD; The EMMES Corporation, Potomac, MD; Indiana University, Indianapolis, IN; University of Pretoria, Pretoria, South Africa; Swedish Hospital Tumor Institute, Seattle, WA; and the University of Chicago, Chicago, IL. Submitted April 3, 1991; accepted July 29, 1991. Supported by the National Cancer Institute, Bethesda, MD. Address reprint requests to Raymond B. Weiss, MD, Section of Medical Oncology, Walter Reed Army Medical Center, Washington, DC 20307-5001. This is a US government work. There are no restrictions on its use. 0732-183X192/1001-0021$0.00/0 Journal of Clinical Oncology, Vol 10, No 1 (January), 1992: pp 69-78 69

70 SESTERHENN ET AL categories: pure S; EC with or without S; T with or without S; T with EC and foci of CC and with or without S; and CC with or without EC and with or without S. Although both classifications contained a category of EC with a reticular pattern, which is the pathologic description of yolk sac tumors (YST), neither classification recognized this tumor type. This omission occurred because these investigators',7 did not have any infants and children in their study population, and YST are the most common germ cell tumors in this younger age group. Many United States pathology laboratories use the two classifications of Friedman, Dixon, and Moore interchangeably. The British classification, modified by Pugh et al, 8 ' 9 divides the tumors into S, spermatocytic seminomas, and T. T are subdivided into differentiated T and malignant T, which are further divided into intermediate, undifferentiated, and trophoblastic subtypes. Mixtures of S and T are designated as combined tumors. While this classification recognizes YST (orchioblastomas) in infants and children, it does not recognize YST in pure form or in admixtures in adult patients. YST is included in undifferentiated T. This classification also does not recognize pure CC but groups pure CC with tumors that have CC elements under the heading of malignant T, trophoblastic. This classification is used in the United Kingdom and some other countries. In addition to these three classification systems, there were also Danish, French, and Soviet classifications. None of these systems recognized intratubular malignant germ cells or syncytiotrophoblasts (ST) occurring alone without cytotrophoblasts. In response to the problems of multiple pathologic classifications and omission of some histologic subtypes, the World Health Organization (WHO) formulated an international classification that divides the tumors into those of single histologic type and those of more than one type."' The basic types are: S, spermatocytic seminoma, EC, YST, CC, and T. The latter type may be subdivided into mature, immature, and T with malignant areas. In tumors of more than one histologic type, the classification requires listing the types and estimating the relative proportions of each. This classification has strict criteria for the diagnosis of CC and excludes ST without cytotrophoblasts from this category. These histologic classifications have been loosely correlated with tumor markers. Talerman et al" and others have linked alpha-fetoprotein (AFP) production to the YST elements, while Bagshawe' 2 and many others have linked human chorionic gonadotropin (HCG) production to the CC element. Serum tumor marker elevation occurs sporadically in patients with other histologic subtypes." 3 The effect of marker production on the prognosis of stage III NSGCT has been defined' 4 but not as well in stage I and II disease." The large patient population in the TCIS also allows us to examine these issues in this report. PATIENTS AND METHODS The TCIS accrued 482 patients with pathologic stage I (ie, negative retroperitoneal lymph nodes) and stage II (ie, involved retroperitoneal nodes) nonseminomatous tumors between July 1979 and August 1985.2 Twenty-three (five stage I) ineligible patients were excluded from further evaluation, resulting in 459 patients available for analysis. There were 264 patients with stage I disease and 195 with stage II. Patients with either pure S or pure CC were ineligible for entry on this study. After orchiectomy and retroperitoneal lymphadenectomy, patients with pathologic stage I disease were registered and observed. If tumor recurred, the patients were to be treated by protocol chemotherapy. Patients with pathologic stage II disease were randomized to either immediate chemotherapy, given for two cycles, or observation until relapse followed by chemotherapy. The chemotherapy in all instances was a cisplatin-based combination regimen as previously described. 3 The patients were observed monthly for the first year, every 2 months the second year, and every 6 months subsequently. Physical examination, chest roentgenograms, and serum tumor marker determinations were performed at these follow-up visits. The marker assays were performed at the local institutions. Local pathologists performed required histologic studies and then submitted slides and tissue blocks to the Armed Forces Institute of Pathology (AFIP), the laboratory for central review, where the WHO classification" was used. The percentage of each separate histologic cell type was estimated on each section, and a summary value was recorded. No formal quantitative assessment was performed. Venous invasion was recognized when tumor was either adherent to a vessel wall or almost completely filling a space lined by flat endothelial cells containing RBCs. Lymphatic invasion was distinguished by observing tumor in spaces lined by endothelial cells without RBCs but with occasional lymphocytes. The location of these vessels and their relationship to arteries and veins were also important in identifying lymphatics. When it was not possible to distinguish between lymphatics and capillaries, venous invasion was designated. We then use the term vascular invasion generically to indicate the presence of either venous or lymphatic invasion or both. The tumors in this study were staged according to the tumor-nodemetastasis (TNM) pathologic staging system. 6 In this testis tumor system, pt1 indicates tumors confined to the testis including rete and tunica albuginea; pt2, those tumors that have invaded beyond the tunica or into the epididymis; pt 3, those that have invaded the spermatic cord; and pt4, those that have invaded the scrotum. Patients with pt1 were subdivided into ptla for rete testis invasion and ptlb for invasion of the tunica albuginea. When paraffin blocks were submitted, they were sectioned and stained with hematoxylin and eosin (H & E). In addition, all tumor-bearing slides were stained in accordance with the peroxidase antiperoxidase method developed by Sternberger 7 for HCG, AFP, and pregnancy-specific P, glycoprotein (SP,). In selected cases where duplicate slide sets and no blocks were available, the H

PATHOLOGIC REVIEW IN TESTICULAR CANCER & E stained sections were destained and then restained for HCG and AFP. The presence of tumor markers demonstrated by these immunohistochemical techniques was recorded within each cell type and graded on a four-point scale ranging from none to heavily stained. Standard univariate statistical methods including X2, Wilcoxon, and Spearman correlation were used to evaluate observed relationships. The relationships between potential prognostic variables and relapse were explored via proportional hazards models with log-linear multipliers of the underlying hazard. A restricted regression with cubic splines and knots at 10%, 25%, 50%, 75%, and 90% was used to relate the percentage EC to relapse. RESULTS Pathologic material submission for both the orchiectomy and lymphadenectomy procedures was expected to yield 918 slide sets for the 459 eligible patients. The actual submissions to the AFIP were 821 (89%) slide sets from 434 (95%) patients. In 42 cases only one slide was received from the primary tumor. Overall, a median of six slides (range, one to 44) from the primary tumor and a median of four slides (range, one to 90) from the lymphadenectomy were available for study at AFIP, including special stain slides. Slides from the orchiectomy were submitted for 229 (87%) patients diagnosed at the protocol-entry institution as stage I and 185 (95%) patients diagnosed as stage II. Retroperitoneal lymphadenectomy slide sets were submitted for 221 patients with stage I and 186 patients with stage II disease. The mean age of the patients with stage I tumor was 28.2 years (range, 15 to 63 years). In stage II patients the mean age was 27.6 years (range, 15 to 48 years). Ninety-seven percent of the entire patient group was white. The presenting symptoms were pain in 46% of the patients and testicular swelling in 73%. Among the patients with EC elements in their tumors, 56% had testicular pain, whereas only 37% of patients without EC elements had pain as a presenting symptom (P <.02). There was a significantly lower incidence of palpable mass and higher rates of testicular swelling when the primary tumors contained T (v no T elements) or YST (v no YST). No differences were observed in age distribution or presenting symptoms between pathologic stage I and II tumors. T and YST were associated with significantly larger tumor sizes. The median maximum tumor dimension was 2.5 cm for those without these histologies compared with 4 cm if either T or YST elements were present (P <.001). T and YST were frequently present together in the primary tumors. Histology of Primary in Pathologic Stage I Tumors Of the 229 available orchiectomy specimens from stage I patients (as determined by the local pathologist [LP]), the central pathologist (CP) review determined that 25 (11%) contained tumor of a single histologic type. S occurred in five (2%), EC in 14 (6%) patients, T in four (2%), and YST in two (1%) (Table 1). The remaining 204 cases contained tumor of more than one histologic type. The most common mixed tumors were EC plus YST plus T, with or without ST (51 and 44 patients, respectively) and S plus EC plus YST plus T, with or without ST (22 and 29 patients, respectively). The frequency of basic cell types, in descending order, was as follows: EC in 199 (87%) patients, YST in 181 (79%), T in 180 (79%), S in 88 (38%), ST in 83 (36%), and CC in nine (4%). The patterns of venous, lymphatic, and capsular invasion appeared to be similar in most of the histologic groupings (Table 1). This outcome is understandable, because with rare exceptions, the invading component was EC and/or YST. Note that among histologic group- 71 Table 1. Primary Tumor Histology-Stage I Disease % % % No. (%) Venous Lymphatic Capsule % of Patients Invasion Invasion Invasion* Recurrence Single-cell types 25 (11) S 5 (2) 0 0 40 (0) 20 EC 14(6) 43 21 14(0) 7 T 4 (2) 25 25 0 (0) 0 YST 2 (1) 50 0 0 (0) 100 Multiple-cell types 204 (89) EC/T/YST/ST 51 (22) 27 6 6 (0) 14 EC/T/YST 44 (19) 25 9 11 (2) 7 S/EC/T/YST/ST 22 (10) 27 14 0 (0) 5 S/EC/T/YST 29 (13) 10 7 7 (0) 10 Other 58(25) 21 9 14(7) 7 Total 229 (100) 23.6 9.2 9.6 (2.2) 9.6 *Numbers in parentheses indicate percentages with tumor invasion beyond tunica or involvement of spermatic cord.

72 ings with 10 or more subjects, pure EC had the highest rates of venous invasion, lymphatic invasion, and tunica (capsular) invasion. Histology of Lymph Nodes in Pathologic Stage I Tumors Two hundred fifteen patients were classified as stage I by both the LP and the CP. Six (2.7%) stage I patients, as determined by the LP, contained tumor in the dissected lymph nodes as determined by the CP review. This difference could be due to the fact that a small focus of tumor appeared in the recut histologic specimen. In each case only a single lymph node was involved. Three patients had necrotic tumor in the metastases, while one each had S, pure EC, and mixed EC plus YST. One of the patients with necrotic tumor in the lymph nodes had a recurrence 30 months later with EC above the orchiectomy scar and a mass on computed tomographic scan adjacent to the iliac blood vessels. Histology of the Primary Tumor in Pathologic Stage II Tumors The distribution of cell types in the primary tumor among 185 patients is listed in Table 2. No pure YST or S were seen. Pure EC was present in 32 (17%) cases, while pure T was present in five (3%). In 148 cases the tumor contained more than one histologic component. As with the stage I tumors, the most frequent combination consisted of EC plus YST plus T, with or without ST (31 and 27 cases, respectively). S plus EC plus YST plus T, with or without ST (11 and 17 cases, respectively) was the next in frequency. EC was present in 172 cases (93%), YST in 122 (66%), T in 111 (60%), S in 62 (34%), ST in 57 (31%), and CC in seven (4%). SESTERHENN ET AL We confirm with CP assessments our previous reports from LP assessments 5 that EC elements are more frequent (P <.04) in stage II patients and that T and YST are less prevalent (P <.01) in stage II disease. Venous, lymphatic, and capsular invasions were significantly more common (each P <.001) in the stage II, relative to stage I, patients. Note that tunica (capsular) invasion occurred in 19% of the stage II cases and pure EC had the highest rates of venous and capsular invasion. Histology of Lymph Nodes in Pathologic Stage II Tumors Involvement of the retroperitoneal lymph nodes was observed in 182 patients by both the LP and CP. Four patients did not have tumor in the nodes on the slides submitted for central review. This disagreement is possibly due to the fact that a small tumor focus was exhausted during recutting and was not available for central review. Comparison of the Primary Tumor and Nodal Metastases Histology Histology of the nodal metastases and the correlation between primary and metastases histology are listed in Table 3. Overall, 88% of the cell types identified in the resected lymph nodes was also found in the primary, a percentage possibly affected by uneven sampling from the primary tumor. In 35 cases, there were 40 disparities between the histology of the metastases and that of the primary (ie, the cell types were observed in the metastases but not in the primary). As an assessment of reproducibility by the CPs, F.K.M. made a second Table 2. Primary Tumor Histology-Stage II Disease % % % Recurrence No. (%) Venous Lymphatic Capsule Rate After of Patients Invasion Invasion Invasion* Lymphadenectomy (%)t Single-cell types 37 (20) S 0 (0) - - EC 32 (17) 69 53 59(34) 10/19(53) T 5 (3) 0 0 0 (0) 0/2 (0) YST 0 (0) - - Multiple-cell types 148 (80) EC/T/YST/ST 31 (17) 48 23 13(10) 6/12 (50) EC/T/YST 27(15) 41 33 15(11) 7/13(54) S/EC/T/YST/ST 11 (6) 64 64 27 (18) 3/5 (60) S/EC/T/YST 17 (9) 53 47 53(29) 6/12(50) Other 62 (33) 61 56 34(26) 14/31 (45) Total 185 (100) 55 45 32 (22) 46/94 (49) *Numbers in parentheses indicate percentages with tumor invasion beyond tunica or involvement of spermatic cord. tcalculated only in those patients not receiving two cycles of adjuvant chemotherapy.

PATHOLOGIC REVIEW IN TESTICULAR CANCER Table 3. Comparison of Histology of the Metastatic Nodes With That of the Primary Tumor No. of Cases Subset With With Nodal Cell Type in % Histology Primary Agreement* S 9 6 67 EC 164 156 95 T Mature 59 53 90 Immature 15 10 67 Malignant transformations 4 1 25 CC 3 1 33 YST 76 63 83 *Indicates the level of agreement between the histology of the metastases with that of the primary. review of each of these 35 cases. There was only one disagreement between the initial CP diagnosis and the second review, indicating that common assessment was being made by the two involved CPs. Some of these disparities are likely attributable to availability of only a few primary sections; only four slides of the primary or cord were available in 18 of the 35 cases. In six cases the primary tumor contained mature T, while the metastases had immature T or mature T with malignant foci. In the two cases with disparity where CC was seen in the lymph nodes, the primary tumors contained hemorrhage and necrosis, thus indicating probable preexisting CC. Discordance Between LP and CP Interpretations The greatest discordance between LP and CP interpretations was in the differential recognition of vascular invasion. Table 4 presents a comparison between the CP assessment of venous and lymphatic invasion and the LP assessment of vascular invasion. There was substantial disagreement between the two pathology sources as to whether invasion existed. Only 54 specimens were observed to have vascular invasion by the LP, while the CP reported 179 specimens with venous or lymphatic invasion (Figs 1 and 2). Note that the distinction between venous and lymphatic invasion was recorded only by the CP. At the central laboratory, venous invasion was more likely to be seen than lymphatic invasion. The lower portion of Table 4 shows that only 52% (81 of 156) of those cases with venous invasion had lymphatic invasion, while 78% (81 of 104) of those with lymphatic invasion also had venous invasion. The former value may be due to the fact that if the vessel involved could not be determined, then the CP coded the case as venous invasion. The second area of discordance is in the diagnoses of specific cell types (Table 5). The most notable observation is that CPs defined YST in 73% (303 of 414) of primary tumors compared with only 12% (50 of 414) at the entry institutions (Figs 3 and 4). There was disagreement on 259 of these specimens. CC was observed in 68 specimens at the local site, compared with 16 by CP (Figs 5 and 6). Similarly, teratomatous elements were seen in 236 (57%) of the cases by LP versus 291 (70%) of cases by CP. Disagreement rates for the two other tumor types were 13% for EC and 17% for S. As a form of quality control, the two CPs compared their estimates of each cell type present in a given tumor in 18% of the cases, both for interobserver and intraobserver variability. The CP estimates varied at most by 5% for the predominant cell type in these comparisons. 73 Table 4. Correlations Between LP/CP Review and Presence of Venous Lymphatic Invasion (n = 414) CP Review: Venous or Lymphatic Invasion Absent Present Total LP review: vascular invasion Absent 201 135 336 Present 16 38 54 Unknown 18 6 24 Total 235 179 414 CP Review: Lymphatic Invasion Absent Present Unknown Total CP review: venous invasion Absent 234 23 1 258 Present 75 81 0 156 Total 309 104 1 414 Fig 1. EC showing lymphatic invasion (original magnification, x100). The arrows outline the vessel (AFIP negative 91-5349).

74 SESTERHENN ET AL Fig 2. EC showing vascular invasion (original magnification, x 100). The arrows outline the vessel (AFIP negative 91-5348). Fig 3. YST in association with EC, which is often overlooked (original magnification, x 100; AFIP negative 91-5347). Correlation Between Serum Tumor Markers and Immunohistochemistry Where sufficient material was available, immunohistochemical staining was evaluated within each tumor type in a subset of primaries. AFP staining was observed in 114 of 127 cases containing YST. All but one of the remaining 13 negative-staining cases had normal or borderline elevated serum AFP levels. The one case that had a serum level of 70 ng/ml had only a minute focus of YST in the Table 5. Correlations Between LP/CP Review and Histologic Subtypes in the Primary Tumor (n = 414) Histology by LP Review Histology by CP Review Absent Present Total available sections of the primary. Since the histochemical reaction is focal, this could be a false-negative. Table 6 relates the maximum prelymphadenectomy serum AFP levels for those patients where it was recorded versus the YST in tissue sections. Significant differences between stain levels and maximum AFP levels are observed (P =.002, Kruskall-Wallis test). Cases with more reacting cells had higher serum AFP levels. Occasional AFP stain reactivity was seen in EC (12 of 146 cases or 8%), mature T (19 of 117 cases or 16%), and in immature T (11 of 77 cases or 14%). No staining was observed in any S or ST cells. Four of the 12 EC cases staining for AFP were pure EC in the primary tumor, and all had serum AFP levels greater than 25 ng/ml. YST Absent Present Total CC Absent Present Total T Absent Present Total S Absent Present Total EC Absent Present 108 3 111 342 56 398 113 10 123 238 26 264 15 28 43 256 364 47 50 303 4 346 12 68 16 65 226 291 45 105 150 28 343 371 178 236 283 131 43 371 Fig 4. AFP staining, which is in YST only (original magnification, x 100; AFIP negative 91-5346).

PATHOLOGIC REVIEW IN TESTICULAR CANCER 75 Table 6. Maximum Prelymphadenectomy Serum AFP Levels by AFP Staining of YST Foci Median AFP Stain Level No. of Cases Level (ng/ml) Quartiles (251th-75th) 0 13 25 13-27 1 17 25 20-81 2 54 93 25-241 3 37 95 25-495 Fig 5. CC showing ST and cytotrophoblasts (original magnification, x 100; AFIP negative 91-5345). Two of the four patients had YST in the metastases, and in the third case, only one slide of the primary tumor was available for central review. In the fourth case the five available sections showed no reactivity with the AFP stain, thus indicating that the source of AFP could not be documented in the available tissue. The percentage of YST correlated positively with maximum serum AFP (r = 0.46, Spearman). The other cell types have no correlation (maximum r = 0.17, Spearman) with serum AFP levels. A heavy staining reaction for HCG was observed in all of six CC and 98 of 109 cases with any ST. Of the remaining ST cases, 27 had less intense staining, and only two cases with ST had no staining reactivity for Fig 6. HCG stain of ST in CC (original magnification, x 100; AFIP negative 91-5344). HCG. No HCG staining was seen in 67 S, 146 EC, 76 immature T, 117 mature T, and 125 YST. In 47 (98%) of the 48 patients with elevation of serum j3-hcg (> 10 miu/ml) and immunohistochemical studies, ST staining was observed. The percentage of EC and the other cell types did not correlate with serum p-hcg levels. The SP, marker was observed in 44 of 70 cases with ST. In 20 of these cases, there was moderate staining and in four cases, heavy staining. Although this marker was seen in one of four CC cases, it was not observed in 39 S, 89 EC, 65 mature or 46 immature T, and 78 YST. Venous and/or Lymphatic Invasion and Tumor Recurrence The recurrence rate using the LP assessment of vascular invasion is 19% (51 of 268) for those without vessel invasion versus 35% (12 of 34; P <.05) for those patients with it (note that those patients on the adjuvant treatment arm are excluded). However, using venous or lymphatic invasion determined by the CP review, only 16 of 199 (8.0%) patients relapsed who did not have invasion, in contrast to 52 of 125 (41.6%; P <.001) patients relapsing who did have invasion. This difference was the same in both stages. In node-negative patients, 6% (10 of 168) without invasion and 19.4% (12 of 62) with invasion recurred, and in node-positive patients, the values were 24% (six of 25) and 63.5% (40 of 63), respectively. Thus, CP assessment of vascular invasion strongly discriminates between groups of patients at low versus high risk of tumor recurrence. Involvement of venous or lymphatic vessels is more common in patients with EC (175 of 371 or 47% for those with EC v four of 43 or 9% for those without). In contrast, patients with T have a lower rate of venous or lymphatic invasion (99 of 291 or 34% when any T is present versus 80 of 123 or 65% when T foci are absent). Tumor Histology and Recurrence EC was more likely to be a stage II tumor; 46% (172 of 371) of those cases containing EC were stage II, whereas only 30% (13 of 43) of those cases without EC had involved nodes. When T or YST histologies were present in the primary tumor, the lymph nodes were less often

76 involved; 38% containing T and 40% containing YST had involved nodes, compared with a rate of 60% and 57% in those cases without T or YST, respectively. In the correlation of histology and timing of tumor recurrence, EC is the only histology of potential importance (P <.01). Among the 43 patients without EC, two relapsed in the first year after treatment and four, beyond the first year. In contrast, 63 of 371 patients with EC relapsed within the first year and only six relapsed thereafter. Figure 7 shows a proportional hazards regression of the percentage of EC related to relapse risk for 321 cases with either stage I disease or stage II not treated with adjuvant chemotherapy. Note that the risk of relapse remains low until the percentage of EC exceeds 30% or 40%. One should not overinterpret the downturn after 80%, as all subjects with greater than 60% EC have a high (> 2) relative risk of relapse. However, the percentage of EC is not a significant prognostic factor after adjustment for either vascular invasion as determined by CP or nodal stage. Histology and Pathologic Stage Tumor was confined to the testis (pt1) in 319 patients, including four with involvement of the rete testis (ptla) and 33 patients with involvement of the tunica '- 0 - q F - I I I I 0 20 40 60 80 100 % EMBRYONAL Fig 7. Relationship between percentage of EC and relative risk of recurrence. Regression using a proportional hazards model with restricted cubic splines and knots at 10%, 25%, 50%, 75%, and 90% (P <.02). SESTERHENN ET AL albuginea (ptlb), with or without involvement of the rete. There were no pt4 cases and only seven pt3. In 17 patients with stage I tumor and 20 patients with stage II tumor who had invasion of the tunica, the relapse rate (after lymphadenectomy and no adjuvant chemotherapy) was 42%, compared with 16% when the tumor did not invade the tunica. These figures are in contrast with a relapse rate of 52% occurring in the patients with tumor invasion beyond the tunica (pt2) or into the spermatic cord (pt3). However, the tunica invasion alone is not a significant predictor of relapse in a proportional hazards model after adjustment for nodal status and venous/lymphatic invasion. DISCUSSION This large international trial, involving 459 eligible patients, was designed to assess issues of relapse rate, adjuvant chemotherapy and chemotherapy at relapse, and overall survival in patients with early-stage NSGCT. An integral part of the trial was the collection and central review of pathology specimens. Despite the large number and geographic spread of collaborating institutions, we were able to collect 89% of the slide sets from 95% of the patients. This large data base has allowed us to analyze a number of issues relating tumor pathology to such points as risk of nodal involvement, concordance of tumor type between the primary and nodal metastases, patterns of vessel invasion, correlations of serum tumor marker and immunohistochemistry staining, and prognostic factors affecting tumor recurrence. In addition, this study has provided a unique perspective regarding the concordance of histologic interpretations between LPs and the CPs in a large patient cohort studied prospectively. From the pathologic standpoint, this study was subject to the usual difficulties of specimen collection. The extent of local sampling was variable, and tissue blocks were not always available for CP review. Despite these problems, a number of interesting observations were made relative to LP versus CP review. One major point is the discordance in the extent to which vascular invasion was detected by CP; LP reported substantially fewer cases with vascular involvement. The low number observed by LP led us to report previously 1 ' 2 that vascular invasion was not strongly predictive of tumor relapse. However, using the CP review data, there is a wide difference in relapse rate (8.0% v 41.6%) based on absence or presence of venous or lymphatic invasion, a difference that is comparable to that obtained from nodal assessment. Vascular invasion has been similarly

PATHOLOGIC REVIEW IN TESTICULAR CANCER identified by others as a risk of both nodal and distant metastases.' 1 Our data should stimulate careful assessment for and recording of this critical histologic variable in individual patients. A second area of important disagreement related to histology of the primary tumors. The least disagreement occurred for EC, in which both groups observed EC in the same number of cases. Some disagreement occurred in 43 of 414 cases regarding whether the classification should be EC or YST. The area of largest disagreement was in the recognition of YST. CP review observed YST in 303 cases, while LP reported it in only 50 cases. Many pathologists fail to recognize YST, perhaps because Friedman and Moore' did not delineate YST in their original description of EC. They included reticular growth pattern in the definition of EC, which is now recognized as diagnostic of YST. Disagreement was less for T, and for S, it was only 17%. A possible explanation for these disagreements is that pathologists may rule out pure S as the main tumor, find EC, and then may not look for other elements. The disagreement in diagnosing CC may be caused by the fact that elevation of the serum p-hcg level alerts the pathologist to look closely for CC. If any ST elements are present, there may be a tendency not to recognize the cells or to mislabel them as CC (Fig 8). This study has identified other important findings related to histology. The linkage of AFP production to YST histology (89% of cases) confirms previous reports" " and can now be considered definitive. Pure EC is rarely, if ever, associated with elevated levels of serum AFP. In this study there were only four such cases. In Fig 8. ST in EC, often misdiagnosed as CC (original magnification, x 100; AFIP negative 91-5343). two of them, the metastases contained YST, the source of the serum AFP. Although an elevated serum AFP level indicates YST, and the intensity of positivestaining correlates with the level of serum AFP (Table 6), about 10% of cases will have YST cells that are functionally inactive. The strong relationship of HCG production to the presence of ST elements (47 of 48 cases), but not to EC, T, YST, or S histology, should stimulate investigation of the genetic features of ST. Serum levels of P-HCG are strongly predictive of outcome in stage III disease, 14 15 ' 19 which suggests that a similar relationship may exist in stage I and II testicular cancer. The clinical utility of SP, plus HCG staining may also need to be reexplored. It is conceivable that SP, and/or HCG expression may be linked to a gene coding for invasive or metastatic potential. A number of points regarding clinical presentation of testicular cancers have been reconfirmed in this study: (1) these cancers (especially those with an EC component) commonly cause pain; (2) T and YST elements are associated with large primary tumor size; (3) the histology of nodal metastases generally (about 90% of the time) reflects that of the primary tumor (but with clinically intriguing exceptions); (4) 80% to 90% of the tumors are mixed, and the most common tumor mixture in the primary is EC plus YST plus T; (5) a major factor for risk of nodal metastases is the presence of EC in the primary tumor, as 93% of the stage II cases contained this tumor component; and (6) some late relapses are seen in patients without teratoma, similar to the observations of others. 20, 2 1 This study identified vascular invasion, as defined by CPs (see Patients and Methods), and invasion of the tunica albuginea as predictors of recurrence. The relationship of histology to risk of relapse 22, 23 was also confirmed in this study. The presence of EC and its percentage correlate strongly with the risk of tumor recurrence (Fig 7). However, when jointly considered with nodal involvement, only vascular invasion is retained as a significant predictor of relapse. Unfortunately, identification of vascular invasion has variable predictive value when assessed by LPs versus CPs. Thus, this characteristic currently has limited utility in specifying strategies for care of individual patients. In any case, all physicians involved in the management of patients with testicular cancer should be especially aware of the need for careful scrutiny for the presence of vascular invasion. In summary, a number of correlates of histology with clinical findings and outcome have been confirmed and 77

78 strengthened. In addition, this study emphasizes the value of appropriate pathologic assessment because of its potential clinical implications, especially when the absence of vascular invasion is used as a criterion to obviate the need for a retroperitoneal lymph node dissection. 2 4 ACKNOWLEDGMENT SESTERHENN ET AL The investigators thank all the participating pathologists in the United States, Switzerland, and South Africa for their gracious cooperation in submitting pathologic materials for this study. The authors are also grateful to Mayra E. Gonzalez-Robles for her assistance in the preparation of this report. 'ENDIX Participating cooperative groups and individual institutions are Eastern Cooperative Oncology Group, Madison, WI; Southeastern Cancer Study Group, Birmingham, AL; Southwest Oncology Group, San Antonio, TX; Walter Reed Army Medical Center, Washington, DC; Illinois Cancer Council, Chicago, IL; Northern California Oncology Group, Palo Alto, CA; Cancer and Leukemia Group B, Lebanon, NH; National Cancer Institute Clinical Center, Bethesda, MD; Mid-Atlantic Oncology Program, Washington, DC; and Uro-Oncology Research Group, Durham, NC. 1. Williams SD, Stablein DM, Einhorn LH, et al: Pathologic stage II testis cancer: Immediate adjuvant chemotherapy versus observation with treatment at relapse. N Engl J Med 317:1433-1438, 1987 2. Williams SD, Stablein DM, Muggia F, et al: Adjuvant chemotherapy of testicular cancer, in Salmon SE (ed): Adjuvant Therapy of Cancer V. Philadelphia, PA, Grune & Stratton, 1987, pp 587-592 3. Weiss RB, Stablein DM, Muggia FM, et al: Toxicity comparisons between two chemotherapy regimens as adjuvant or salvage treatment in non-seminomatous testicular cancer. Cancer 62:18-23, 1988 4. Giguere JK, Stablein DM, Spaulding JT, et al: The clinical significance of scrotal violation in testicular cancer: A report from the Testicular Cancer Intergroup Study. J Urol 139:1255-1228, 1988 5. McLeod DG, Weiss RB, Stablein DM, et al: Staging relationships and outcome in early stage testicular cancer: A report from the Testicular Cancer Intergroup Study. J Urol 145:1178-1183, 1991 6. Friedman NB, Moore RA: Tumors of the testis: A report of 922 cases. Milit Surg 99:573-593, 1946 7. Dixon FJ, Moore RA: Tumors of the Male Sex Organs. Atlas of Tumor Pathology Fascicle 31b and 32. Washington, DC, Armed Forces Institute of Pathology, 1952 8. Pugh RCB, Cameron KM: Teratoma, in Pugh RCB (ed): Pathology of the Testes. Oxford, UK, Blackwell, 1976, pp 199-244 9. Pugh RCB: Pathology of testicular tumors-a British perspective, in Donohue JP (ed): Testes Tumors. Baltimore, MD, Williams and Wilkins, 1983, pp 3-19 10. Mostofi FK, Sobin LH: International Histological Classification of Tumors of Testis (No. 16). Geneva, Switzerland, World Health Organization, 1977 11. Talerman A, Haije WG, Baggerman L: Serum alphafetoprotein (AFP) in patients with germ cell tumors of the gonads and extragonadal sites: Correlation between endodermal sinus (yolk sac) tumor and raised serum AFP. Cancer 46:380-385, 1980 12. Bagshawe KD: Recent observations related to the chemotherapy and immunology of gestational choriocarcinoma. Adv Cancer Res 18:231-263, 1973 REFERENCES 13. Norgaard-Pedersen B, Schultz HP, Arends J, et al: Tumor markers in testicular germ cell tumors. Five-year experience from the DATECA study 1976-1980. Acta Radiol Oncol 23:287-294, 1984 14. Bosl GJ, Geller NL, Cirrincione C, et al: Multivariate analysis of prognostic variables in patients with metastatic testicular cancer. Cancer Res 43:3403-3407, 1983 15. Hesketh PJ, Krane RJ: Prognostic assessment in nonseminomatous testicular cancer: Implications for therapy. J Urol 144:1-9, 1990 16. Hermanek P, Sobin LH: TNM Classification of Malignant Tumors. New York, NY, Springer-Verlag, 1987, pp 127-129 17. Sternberger LA: Immunochemistry, 2nd ed. New York, NY, John Wiley and Sons, 1979, pp 126-127 18. Mostofi FK, Sesterhenn IA, Davis CJ: Immunopathology of germ cell tumors of the testis. Semin Diag Pathol 4:320-341, 1987 19. Stoter G, Sylvester R, Sleijfer DT, et al: Multivariate analysis of prognostic factors in patients with disseminated nonseminomatous testicular cancer: Results from a European Organization for Research on Treatment of Cancer multiinstitutional phase III study. Cancer Res 47:2714-2718, 1987 20. Roth BJ, Greist A, Kubilis PS, et al: Cisplatin-based combination chemotherapy for disseminated germ cell tumors: Longterm follow-up. J Clin Oncol 6:1239-1247, 1988 21. Borge N, Fossa SD: Late relapses of testicular cancer: A review. Cancer J 3:53-55, 1990 22. Freedman LS, Parkinson MC, Jones WG, et al: Histopathology in the prediction of relapse of patients with stage I testicular teratoma by orchiectomy alone. Lancet 2:294-297, 1987 23. Fung CY, Kalish LA, Brodsky GL, et al: Stage I nonseminomatous germ cell testicular tumor: Prediction of metastatic potential by primary histopathology. J Clin Oncol 6:1467-1473, 1988 24. Pont J, Holtl W, Kosak D, et al: Risk-adapted treatment choice in stage I nonseminomatous testicular germ cell cancer by regarding vascular invasion in the primary tumor: A prospective trial. J Clin Oncol 8:16-20, 1990