Prognostic variables for response and outcome in patients with extragonadal germ-cell tumors

J. T. Hartmann1,+, C. R. Nichols2, J.-P. Droz3, A. Horwich4, A. Gerl5, S. D. Fossa6, J. Beyer7, J. Pont8, L. Kanz1, L. Einhorn9 and C. Bokemeyer1

1 Tuebingen University Medical Center II, Tuebingen, Germany; 2 Oregon Health Sciences University, Portland, Oregon, USA; 3 Centre Léon-Berard, GETUG, Lyon Cedex, France; 4 The Royal Marsden Hospital, Sutton, UK; 5 Klinikum Großhadern, Munich, Germany; 6 The Norwegian Radium Hospital, Oslo, Norway; 7 Virchow Klinikum, Berlin, Germany; 8 Kaiser Franz Josef Spital, Vienna, Austria; 9 Indiana University, Indianapolis, IN, USA

Received 5 September 2001; revised and accepted 11 February 2002


    Abstract
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Background:

This investigation evaluates prognostic variables in patients with seminomatous and non-seminomatous extragonadal germ-cell tumors (EGCT) in order to identify relevant factors for long-term outcome following cisplatin-based chemotherapy.

Patients and methods:

Patients from six countries treated at 11 centers in Europe and the USA from 1975 to 1996 were evaluated retrospectively. Uni- and multivariate analyses of prognostic variables for survival and for response to chemotherapy were performed.

Results:

Data were available for 635 EGCT patients, 104 with seminomatous and 524 with non-seminomatous EGCT (n = 7 not specified). For non-seminomatous EGCT the following independent adverse factors were identified: presence of either liver, lung or central nervous system metastases, primary mediastinal tumor or elevation of pretreatment ß-human gonadotropin; for extragonadal seminoma (only univariate) adverse factors were: presence of liver metastases, two or greater metastatic sites or International Germ Cell Cancer Collaborative Group (IGCCCG) grouping (intermediate versus good). Integration of these variables produced the following prognostic risk groupings: ‘excellent prognosis’, all seminomatous EGCT (89% 5-year survival rate); ‘intermediate low’, ‘intermediate high’ and ‘poor’, all non-seminomatous EGCT with a 69, 55 and 17% 5-year survival rate, respectively. The decreased survival among the different groups was due to a lower rate of favorable objective remissions and a higher rate of relapses. Classification and regression tree (CART) modeling confirmed histology and location of primary tumor as the major prognosticators. For the subgroup of patients with mediastinal non-seminoma, the 2-year survival rate ranged from 34 to 84%. Multivariate testing for the probability to respond to chemotherapy revealed non-seminomatous histology, primary mediastinal tumor site, and the presence of liver, lung or CNS metastases as independent adverse factors.

Conclusions:

In EGCT, prognostic variables for the outcome and for the response to chemotherapy could be identified, which in part differ from gonadal GCT. The proposed model might help to better understand the specific prognosis of EGCT and to tailor risk-adapted treatment strategies. In addition, CART analysis demonstrated the heterogenous prognosis of patients with mediastinal non-seminoma.

Key words: chemotherapy, extragonadal germ-cell tumors, mediastinal location, prognostic variables, response, retroperitoneal location


    Introduction
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Extragonadal germ-cell tumors (EGCT) are estimated to represent ~2–5% of all germ-cell malignancies [1]. In adults, these tumors most commonly arise in the midline of the retroperitoneum or the mediastinum [24]. While the serological and cytogenetic characteristics of EGCT are similar to those of primary testicular germ-cell tumor (GCT), differences in the distribution of histologic subtypes, associated biologic disorders and, in particular, the clinical outcome seem to exist [58].

Cisplatin-based chemotherapy has demonstrated activity in patients with EGCT and long-term survival rates approach those of patients with advanced-stage metastatic testicular cancer [911]. To better characterize the prognosis of this specific form of disease, a large multicenter database of 635 patients with EGCT has been reviewed. Patients were treated at a time when cisplatin-based chemotherapy was the treatment of choice at 11 cancer centers across six countries in Europe and the USA from 1975 to 1996.


    Patients and methods
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Data collection
The medical records of 635 patients with EGCT treated at 11 cancer centers in the United States and Europe from 1975 to 1996 were reviewed. The diagnosis of EGCT was defined as a GCT arising in the mediastinum, retroperitoneum or other site, without demonstrable testicular abnormalities as determined by physical examination and testicular ultrasonography. Testicular biopsies were performed in patients with abnormalities on palpation or testicular ultrasonography. Patients with histologically undifferentiated extragonadal tumors with markedly elevated germ-cell tumor serum markers, who were treated according to GCT protocols were also included in this analysis.

The contributing centers were internationally recognized for their experience in the treatment of germ-cell tumors: Indiana University, Indianapolis, IN (n = 216; time period 1989–1996); Institute Gustave-Roussy, Villejuif; Centre Léon-Berard, Lyon; and the Groupe d‘Etude des Tumeurs Uro-geniales (GETUG), France (n = 93 from these three centers; time period 1975–1996); Eberhard-Karls University Medical Center II, Tuebingen, Germany (n = 13; time period 1986–1993); Hannover University Medical School, Hannover, Germany (n = 88; time period 1978–1995); Norwegian Radium Hospital, Oslo, Norway (n = 48; time period 1980–1995); Klinikum Großhadern, Munich, Germany (n = 63; time period 1979–1996); The Royal Marsden Hospital, Sutton, UK (n = 65; time period 1979–1994); Kaiser-Franz-Josef Spital, Vienna, Austria (n = 19; time period 1975–1996; and Virchow-Klinikum, Berlin, Germany (n = 30; time period 1987–1994). A standardized questionnaire was sent to each center and completed by the responsible co-investigator. All patients’ data were obtained in an anonymous manner. Detailed information on patients’ characteristics, such as location and histology of the primary tumor; extent of disease including serum tumor marker concentrations of ß-human gonadotropin (ß-HCG), {alpha}-fetoprotein (AFP) and lactate dehydrogenase (LDH); history of testicular abnormalities; details on diagnostic methods, treatment, response to treatment and follow-up period; and data on secondary testicular cancer and other secondary cancers were acquired. Duration of follow-up and survival in this analysis were calculated based on the date of the first day of treatment until the date of last contact, if the patient was still alive, or the date of death. For all living patients, the status as of August 1998 was verified. Tumor response was classified as follows: complete remission (CR) was defined as a complete disappearance of all clinical, radiological and biochemical evidence of disease with normalization of ß-HCG and/or AFP and/or LDH for at least 1 month. A partial response was defined as a decrease in >=50% of the sum of the products of perpendicular diameters of measurable disease lasting for at least 1 month. If elevated markers were the only evidence of disease, a decrease of >=90% was required for a partial response. Progressive disease was defined as either residual lesions increasing in size or occurrence of new lesions and/or elevation of tumor markers at repeated controls. Patients who achieved a normalization of tumor markers but an incomplete radiographic response were submitted to postchemotherapy surgery. However, in some patients who had attained serologic complete remission but with persistent minor radiographic abnormalities, individual investigators have chosen to observe such patients without surgery. Those patients were formally coded as partial remissions if their residual abnormalities remained stable or decreased on imaging studies over a 1-year period.

Patients who underwent complete surgical resection of vital carcinoma or mature teratoma were classified as no evidence of disease (NED)carcinoma or NEDteratoma, respectively.

Statistical analysis
All statistical analyses were performed using SPSS for Windows 8.0 (SPSS, Chicago, IL, USA). The overall survival (OS) calculation used death due to any reason as the end point. Various patient characteristics were investigated as potential prognostic factors by univariate analysis, such as: (i) Categorical variables: location of the primary tumor, extent of disease at diagnosis (such as bone, lung, liver, brain and lymph node involvement), age grouping (<20, 20–29, 30–39, 40–49, >=50 years of age), tumor marker concentrations at diagnosis (elevated yes/no) or tumor marker grouping (according to the IGCCCG classification) [12], presence of additional metastatic sites, and type of (conventional versus high-dose chemotherapy) and response to first-line treatment (defined as cisplatin sensitive, in case of CR or partial remission (PR) with marker normalization; relative refractory, in case of partial remission without marker normalization or stable disease; absolute refractory, in case of progressive disease during chemotherapy); and (ii) Continuous variables: age and tumor marker concentrations at diagnosis. For survival time, the proportion of survivors was estimated by the Kaplan–Meier method [13], and the log-rank test was used for comparisons. For ordered categorial variables, the log-rank test for trend was used. A multivariate model was built to evaluate the simultaneous effects of several variables predicting for OS using a stepwise forward selection procedure of the Cox regression. All factors with P <0.05 identified in the proceeding univariate analysis were included in the multivariate analysis [14]. The results of the Cox model were reported with relative risk ratios (hazard ratio, HR) and confidence intervals (CI). In addition, a classification and regression-tree (CART) model was used to explore candidates for further homogenous prognostic subgroups for OS [15]. Advantages of this procedure compared with the proportional hazards model are that the interaction effects can be detected more readily and no prior assumptions need to be made about the form in which distributions of variables should be modeled. For this test, SAS software, version 6.11 (SAS Institute, Cary, NC, USA) was used.

A second end point of this study was to identify predictive factors for response to chemotherapy. Response to treatment was analyzed as a binary variable: patients who achieved a complete or partial remission were considered as ‘responders’, and patients with stable disease or progression were considered ‘failures’. The proportion of responders was estimated in contingency tables, and the categories were compared using the chi-square test. All factors that presented with a significant or borderline prognostic value were initially included in the Cox regression model [16]. Non-significant factors were subsequently removed. All reported P values were two-sided. The results presented here are based on unadjusted analyses of categorical variables. The final multivariate model were later adjusted by therapeutic regimen. This procedure or the replacement of categorized variables by original variables (tumor marker concentrations at initial diagnosis and at relapse) did not change any conclusion.


    Results
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Patient characteristics
Individual data on 635 patients with a median age of 30 years (range 14–79) were evaluable for this analysis. Three hundred and forty-one (54%) patients had a primary mediastinal EGCT, and 283 (45%) patients a primary retroperitoneal EGCT [n = 1 with cervical lymph node involvement (0.2%); n = 10 with both midline retroperitoneal and mediastinal disease (1.6%)].

The data set contained 104 patients with a pure seminomatous EGCT, which was previously analyzed in detail [17]. The majority [524 patients (83%)] had non-seminomatous EGCT (histology not specified, n = 7). The median age was 37 years (range 18–70) for seminoma patients and 29 years (range 14–79) for patients with non-seminomatous EGCT.

Two hundred and eighty-seven of 524 non-seminomatous EGCT patients (55%) had a primary mediastinal location. Treatment consisted of chemotherapy in 278 of 287 patients (97%) (for details see Table 1). Additionally, 227 non-seminomatous patients with retroperitoneal location have been included. The majority (n = 223, 98%) has been treated with chemotherapy. Two patients underwent primary resection of retroperitoneal tumor alone (0.9%) (radiation alone, n = 1). Aspects of diagnostic approaches, clinical symptoms, epidemiological data as well as details on treatment of EGCT are reported elsewhere [18].


View this table:
[in this window]
[in a new window]
 
Table 1.  Patients characteristics according to primary tumor location
 
Response to treatment
A favorable response to treatment CR or PR was achieved in 92% (95% CI 87% to 97%) of seminoma patients. In mediastinal and retroperitoneal non-seminoma patients the remission rates were 64% (95% CI 58% to 70%) [CR, n = 54 (19%); tumor marker normalized PR (PRm-), n = 124 (45%)] and 68% (95% CI 62% to 75%) [CR, n = 61 (28%); PRm-, n = 80 (40%)]. Forty-nine per cent of patients with mediastinal non-seminoma and 45% of patients with retroperitoneal location underwent secondary surgery of residual tumor masses after chemotherapy yielding viable tumor in 35% and 25%, mature teratoma in 26% and 16% and necrotic tissue in 37% and 59%, respectively (patients not evaluable n = 11).

Survival
After a median follow-up of 5.1 years (range 1–211 months), 12 of 104 seminoma patients (12%) had died. The 5-year OS rate was 88% (Figure 1). For mediastinal non-seminoma the calculated 5-year OS and progression-free survival (PFS) rates were 45% and 44%, respectively. After a median follow-up of 41 months (range 4–178 months), 120 patients (42%) had relapsed and 139 of 287 patients (48%) had died. The calculated 5-year OS and PFS rates for patients with retroperitoneal non-seminoma were 62 and 45%. After a median follow-up period of 52 months (range 4–203 months) 81 of 227 retroperitoneal non-seminoma patients (36%) had died (Figure 1).



View larger version (16K):
[in this window]
[in a new window]
 
Figure 1. Overall survival of patients with extragonadal germ-cell tumors according to location and histology. med, mediastinal; NS, non-seminoma; rp, retroperitoneal non-seminoma; S, seminoma.

 
Univariate analysis of survival time
Univariate analysis has been performed separately for patients with seminomatous and non-seminomatous EGCT patients. The results are summarized in Table 2. For extragonadal seminomas a significant inferior PFS was found for patients treated with primary radiotherapy compared with either chemotherapy alone or chemo- plus radiotherapy. Elevation of pretreatment ß-HCG indicated a superior PFS (data not shown), but did not influence OS. The presence of two or more metastatic sites or non-pulmonary visceral metastases were identified to worsen OS, with liver involvement being the most important adverse prognostic factor.


View this table:
[in this window]
[in a new window]
 
Table 2.  Univariate analysis of patient and treatment characteristics for their influence on overall survival in non-seminomatous and seminomatous extragonadal germ-cell tumors
 
Mediastinal location, the presence of either CNS, liver or lung metastases, elevation of ß-HCG and refractoriness to cisplatin were associated with a significant inferior OS for non-seminomatous EGCT patients. Except for the primary location of EGCT, all mentioned variables were also of significant influence regarding PFS (data not shown).

Multivariate analysis on prognostic factors for survival in EGCT
Classification and regression-tree model (CART) analysis
Primary division of prognostic subgroups was based on histology—seminomatous or non-seminomatous EGCT. Patients with seminomatous histology had a 2-year survival rate of 97% (95% CI 93% to 100%), whereas EGCT patients with non-seminomatous histology had only a 65% 2-year OS rate (95% CI 61% to 70%). Among patients with seminomatous EGCT, no further prognostic subgroups could be identified.

For patients with non-seminomatous histology the next division was performed according to the primary tumor site. Patients with retroperitoneal primaries exhibit a 2-year OS rate of 76% (95% CI 70% to 82%), whereas patients with a primary mediastinal tumor site have a lower chance of cure [2-year OS rate, 57% (95% CI 51% to 63%)]. Interestingly, in this subgroup of patients, further divisions could be determined which lead to survival rates ranging from 34% to 84% at 2 years. Patients with mediastinal germ-cell tumor and presence of visceral metastases achieved the worst 2-year OS rate of 34% (95% CI 17% to 52%). In patients with mediastinal tumor location and absence of visceral metastases [2-year OS rate, 60% (95% CI 54% to 67%)], further subgroups were determined based on the presence of lung metastases, age and serum tumor marker concentration of ß-HCG. For the subgroup of patients <=29 years of age with primary mediastinal tumor site having neither visceral metastasis nor lung metastases nor elevated ß-HCG serum values, a favorable outcome of 84% was observed (95% CI 73% to 94%) (Figure 2).



View larger version (22K):
[in this window]
[in a new window]
 
Figure 2. Classification and regression tree (CART) analysis for overall survival. All estimates are 2-year survival rates according to the Kaplan–Meier method.

 
Cox proportional hazard analysis
In a second step, multivariate analysis was performed using the Cox model. However, as shown before in the CART analysis, none of the treatment or patient variables in extragonadal seminomas was of independent influence. For extragonadal non-seminomas the presence of CNS metastases [HR = 2.5 (95% CI 1.4–4.5; P = 0.002)], primary mediastinal location [HR = 2.3 (95% CI 1.6–3.2; P <0.0001)], liver metastases [HR = 1.7 (95% CI 1.2–2.5; P = 0.006)], elevation of pretreatment ß-HCG [HR = 1.5 (95% CI 1.1–2.1; P <0.022)] and the presence of lung metastases [HR = 1.4 (95% CI 1.0–2.0; P = 0.028)] were found to be independent adverse factors for OS (Table 3). With the exception of the presence of liver metastases, all factors were also independently relevant for PFS in extragonadal non-seminoma patients: CNS metastases, HR = 1.9 (95% CI 1.1–3.1; P = 0.019); lung metastases, HR = 1.6 (95% CI 1.2–2.2; P = 0.001), primary mediastinal location, HR = 1.5 (95% CI 1.1–2.0; P = 0.005), elevation of pretreatment ß-HCG, HR = 1.4 (95% CI 1.0–1.9; P <0.027) (data not shown). Refractoriness to cisplatin was also found to be of significant influence on both OS and PFS, but was not considered in the Cox model because of its lack of prognostic relevance prior to the start of treatment for EGCT (Figure 3).


View this table:
[in this window]
[in a new window]
 
Table 3. Factors with multivariate significance for OS in extragonadal non-seminoma
 


View larger version (21K):
[in this window]
[in a new window]
 
Figure 3. Calculated overall survival rates of non-seminomatous extragonadal germ-cell tumor patients according to sensitivity to cisplatin and location of primary site. med, mediastinal non-seminoma; mos, months; OS, overall survival; rp, retroperitoneal non-seminoma.

 
Development of a prognostic classification for EGCT
Because patients with extragondal seminoma possess an excellent prognosis and multivariate analysis did not reveal any negative prognosticators, this group of patients was at first not considered for further analysis. The five variables of patients with non-seminomatous histology identified in the Cox multivariate regression analysis were used to form prognostic groups by combining patients with similar hazards. Summing the ß coefficients—obtained from multivariate analysis—has allowed the calculation of the hazards ratio (exp ß) for survival after treatment for EGCT in each individual patient. To simplify the use of a prognostic model without having to calculate individual risk functions, a score was given to each of the variables that retained significance. Because the presence of CNS metastases at diagnosis and primary mediastinal tumor site of EGCT revealed the strongest effect on an unfavorable treatment outcome, both variables were given a score of 2. Presence of liver metastases and lung metastases, as well as elevation of ß-HCG values, were each given a score of 1 (Table 3). Patients were subsequently assigned to one of three prognostic categories on the basis of their score, as follows: ‘good risk’, with a score of 0 or 1 and a corresponding hazards ratio of <2; ‘intermediate risk’, with a cumulative score of <3 and a corresponding hazards ratio of 2 to <3; and ‘poor risk’, with a score >3 and a corresponding hazard ratio of >3. The curves for OS for each score and for each prognostic category of non-seminomatous EGCT patients are shown in Figures 4 and 5. The differences in OS between patients belonging to the different prognostic categories were due to both a significantly lower rate of favorable responses and due to a significantly higher rate of relapses in the intermediate and the poor-risk categories as shown in Table 4. Finally, the data set of patients with seminomatous EGCT was added to the prognostic model, and curve comparison of the four defined subgroups—‘excellent’, ‘intermediate low’, ‘intermediate high’ and ‘poor’—yielded a highly significant value (P <0.0001). The results are demonstrated graphically in Figure 6.



View larger version (19K):
[in this window]
[in a new window]
 
Figure 4. Overall survival of patients with non-seminomatous extragonadal germ-cell tumors according to the number of prognostic factors.

 


View larger version (13K):
[in this window]
[in a new window]
 
Figure 5. Overall survival of patients with non-seminomatous extragonadal germ-cell tumors according to prognostic category.

 

View this table:
[in this window]
[in a new window]
 
Table 4. Outcome in extragonadal germ-cell tumors after chemotherapy according to prognostic categories
 


View larger version (20K):
[in this window]
[in a new window]
 
Figure 6. Final prognostic model for overall survival according to prognostic category and considering patients with seminomatous histology.

 
Multivariate analysis for refractoriness to chemotherapy in EGCT patients (n = 611)
According to this model, non-seminomatous histology (HR = 5.2, P = 0.007), presence of liver (HR = 2.6, P = 0.01), CNS (HR = 3.7, P = 0.02) and lung metastases (HR = 2.4, P <0.0001) and a primary mediastinal site (HR = 1.9, P = 0.001) were independent unfavorable factors for response in all patients treated with chemotherapy (Table 5).


View this table:
[in this window]
[in a new window]
 
Table 5.  Final logistic model for response to chemotherapy in extragonadal germ-cell tumors
 

    Discussion
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Determining the exact prognosis of patients with EGCT has been difficult to date, largely because of the small number of patients available for analysis in single trials. In addition, the outcome of patients even after the introduction of cisplatin-based chemotherapy has been variable among different research groups [10, 1923]. This report focuses on identifying specific prognostic variables relevant for treatment outcome in patients with EGCT. The data set used includes more than 600 patients with EGCT who were treated from 1975 to 1996 at 11 major cancer centers.

Over the past 15–20 years several prognostic models have been proposed for patients with metastatic gonadal GCT [2426]. In recent years, the International Germ Cell Cancer Collaborative Group (IGCCCG) classification has replaced these older classifications. One of the major adverse factors identified by the IGCCCG is the presence of a mediastinal primary site in patients with non-seminomatous GCT. Such patients appear to represent a clinically and biologically distinct disease entity [6, 7, 10], associated with lower CR rates, high rates of relapse and failure of salvage chemotherapy [11, 22, 27, 28]. In contrast, patients with mediastinal seminoma prove to have excellent prognosis indistinguishable from that of seminoma patients with bulky metastases and a testicular primary site. In addition, no significant difference in PFS and OS has been demonstrated among patients with primary mediastinal and retroperitoneal seminoma [18]. Patients with mediastinal non-seminomatous EGCT revealed adverse survival rates despite the use of modern cisplatin combination regimens. Retrospective as well as prospective trials—based on small sample sizes—yielded OS rates of 30–50% [10, 2022, 2933]. Newer analyses underlined the role of aggressive postchemotherapy surgery [34]. The 284 evaluable patients with primary mediastinal location in the current series had a long-term survival of 48%. However, this detailed analysis demonstrated that this group is composed of patients with heterogeneous prognosis which is dependent on different patient characteristics. Patients, who were of young age, having localized mediastinal disease without elevation of ß-HCG at initial diagnosis revealed a favourable survival chance compared with patients presenting with adverse prognostic variables such as visceral metastases or lung metastases, elevated ß-HCG or being >30 years of age.

The second group of patients with extragonadal non-seminoma comprised patients with retroperitoneal GCT or unknown primary sites with elevated tumor markers. These patients have a prognosis in between mediastinal and testicular primary tumors. But despite this evident difference, the IGCCCG incorporated this group with those patients with a testicular primary site for practical reasons. For these patients, most authors recommend four instead of three cycles of cisplatin-based combination followed by postchemotherapy dissection in the case of residual masses. The 5-year OS for the 227 patients in this series has been determined at 62%. When this subgroup of patients was categorized retrospectively according to the IGCCCG classification, the 5-year OS rates were 59% in the poor prognosis group (expected: 48%) and 61% in the intermediate/good group (expected: 80–92%). Taking into account that the serum tumor marker categories used by the IGCCCG were of lower prognostic relevance for EGCT (Table 2) and when the same analysis was performed considering only the variable ‘presence of visceral disease’, a significant difference in 5-year OS rates of 16% could be observed between both groups (poor: 50% versus good/intermediate: 66%, P = 0.01). Thus, primary retroperitoneal tumor patients with poor IGCCCG criteria (presence of visceral disease) achieved the expected survival rate, whereas patients with good/intermediate IGCCCG criteria still achieved an inferior outcome compared with patients with gonadal GCT of the same prognostic group.

The identification of prognostic factors in EGCT was primarily limited to the small numbers of patients included in published trials. The current investigation allows for the first time the identification of prognostic factors, due to the large population of EGCT patients included. First, a CART model was used to define a hierarchy of clinical prognostic factors. Histology—seminomatous or non-seminomatous—has been identified as the most important predictor of survival in EGCT. In seminomatous EGCT patients, no further prognostic factors could be identified using both CART and the Cox proportional hazard analysis. For non-seminomatous EGCT, the primary tumor site was the most important predictor according to CART analysis. Subsequently, patients with primary mediastinal tumor could be separated into different subgroups with markedly different OS rates. The Cox proportional hazards model identified the presence of liver, lung or CNS metastases, a primary mediastinal location site and the elevation of ß-HCG as independent factors for OS. The findings that serum tumor marker levels AFP and LDH were less predictive for survival as well as the evidence of lung metastases being relevant for both PFS and OS in non-seminomatous EGCT patients are in contrast to the findings achieved for gonadal GCT patients [13].

Based on the results of the Cox multivariate analysis, prognostic categories for non-seminomatous EGCT patients could be formed using a score that related to the number of risk factors present and to the hazard each risk factor conferred. In these categories, groups of patients with similar response rates and survival were combined. Patients in the poor prognosis category had a significantly lower probability for OS than either intermediate or good prognosis patients. Overall, the differences between these categories were due both to a lower rate of favorable responses and a higher proportion of relapses with increasing score values. In addition, the data set of patients with seminomatous EGCT has been included in the final model. Those patients had an excellent prognosis. The final model consisted of four prognostic categories, excellent (all extragonadal seminomas; overall 17% of patients; 5-year OS rate 89%), intermediate low (extragonadal non-seminomas with a score of 0 or 1; 20% of patients; OS rate 69%), intermediate high (extragonadal non-seminomas with a score of maximum 3; 52% of patients; OS rate 55%) and finally poor (extragonadal non-seminoma with a score >3; 11% of patients; OS rate 17%). Within each prognostic category, we have subsequently tried to examine the influence of different chemotherapeutical treatment regimen on the outcome. However, an unequivocal benefit in terms of increased response rate or improved OS could not be demonstrated.

The multivariate analysis on prognostic factors associated with refractoriness to therapy revealed non-seminomatous histology as the most relevant factor. With the exception of the tumor marker ß-HCG, all other variables—presence of lung, CNS or liver metastases, as well as mediastinal tumor location—have also been identified to be of prognostic relevance for OS.

In summary, this analysis demonstrates that the prognosis of EGCT patients is in general heterogeneous. Patients with seminomatous histology possess an excellent prognosis comparable to those with metastatic gonadal seminoma, confirming the classification of these patients as proposed by the IGCCCG. For non-seminomatous EGCT, relevant prognostic risk factors have been identified confirming that the outcome of EGCT patients, as suggested earlier, differs from that of gonadal GCT patients even when patients with otherwise similar prognostic characteristics are compared. The knowledge of specific prognostic factors as provided by CART analysis and the Cox proportional model, might help to tailor treatment options in the future considering the individual risk for those patients. The prognosis of mediastinal non-seminoma is worse compared with other subgroups of EGCT; however, patients with localized disease have a substantial chance of cure with the use of modern cisplatin-based combination chemotherapy combined with aggressive surgery.


    Acknowledgements
 
Presented in part at the annual meeting of the American Society of Clinical Oncology, May 2000, New Orleans.


    Footnotes
 
+ Correspondence to: Dr J. T. Hartmann, Department of Hematology, Oncology and Immunology, UKT University Medical Center II, Eberhard-Karls University, Otfried Mueller Straße 10, 72076 Tuebingen, Germany. Tel: +49-7071-298-2125; Fax: +49-7071-29-5357; E-mail: joerg.hartmann@med.uni-tuebingen.de Back


    References
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
1. Collins DH, Pugh RCB. Classification and frequency of testicular cancer. Br J Urol 1964; 36 (Suppl 1): 1–11.

2. Johnson DE, Laneri JP, Mountain CF et al. Extragonadal germ cell tumors. Surgery 1973; 73: 85–90.[ISI][Medline]

3. Lisco H. Malignant tumors developing in sacrococcygeal teratomata. Ann Surg 1942; 115: 378–389.

4. Hart WR. Primary endodermal sinus (yolk sac) tumor of the liver. First reported case. Cancer 1975; 35: 1453–1458.[ISI][Medline]

5. Nichols CR, Heerema NA, Palmer C et al. Klinefelter’s syndrome associated with mediastinal germ cell neoplasms. J Clin Oncol 1987; 5: 1290–1294.[Abstract]

6. Nichols CR, Roth BJ, Heerema N et al. Hematologic neoplasia associated with primary mediastinal germ cell tumors. N Engl J Med 1990; 322: 1425–1429.[Abstract]

7. Hartmann JT, Nichols CR, Droz J-P et al. Hematologic disorders associated with primary mediastinal non-seminomatous germ cell tumors. J Natl Cancer Inst 2000; 92: 54–61.[Abstract/Free Full Text]

8. Hartmann JT, Nichols CR, Droz J-P et al. The relative risk of secondary non-germinal malignancies in patients with extragonadal germ cell tumors. Cancer 2000; 88: 2629–2635.[ISI][Medline]

9. Bosl GJ, Motzer RJ. Medical progress: testicular germ-cell cancer. N Engl J Med 1997; 337: 242–253.[Free Full Text]

10. Nichols CR, Saxman S, Williams SD et al. Primary mediastinal non-seminomatous germ cell tumors: a modern single institution experience. Cancer 1990; 65: 1641–1646.[ISI][Medline]

11. Hartmann JT, Einhorn L, Nichols CR et al. Second-line chemotherapy in patients with relapsed extragonadal non-seminomatous germ cell tumors: results of an international multicenter analysis. J Clin Oncol 2001; 19: 1641–1648.[Abstract/Free Full Text]

12. Mead GM, Stenning SP, Cook P et al. International germ cell consensus classification: a prognostic factor-erased staging system for metastatic germ cell cancers. J Clin Oncol 1997; 15: 594–603.[Abstract]

13. Kaplan EL, Meier P. Non-parametric estimation from incomplete observation. J Am Stat Assoc 1958; 53: 457–481.[ISI]

14. Cox DR. Regression models and life tables (with discussion). J Roy Stat Soc Ser B 1987; 34: 187–220.

15. Lausen B, Sauerbrei W, Schuhmacher M. Classification and regression trees (CART) used for the exploration of prognostic factors measured on different scales. In Dirschedl P, Osermann R (eds): Computational Statistics. Heidelberg/New York: Physica-Verlag 1994.

16. Cox DR. The analysis of binary data. London: Methuen 1970.

17. Bokemeyer C, Droz JP, Horwich A et al. Extragonadal seminoma—an international multicenter analysis of prognostic factors and long term treatment outcome. Cancer 2001; 91: 1394–1401.[ISI][Medline]

18. Bokemeyer C, Nichols CR, Droz J-P et al. Extragonadal germ cell tumors of the mediastinum and retroperitoeum: results from an international analysis on 635 patients. J Clin Oncol 2002; 20: 1864–1873.[Abstract/Free Full Text]

19. Feun LG, Samson MK, Stephens RL. Vinblastine (VLB), bleomycin (BLEO), cis-diamminedichloroplatinum (DDP) in disseminated extragonadal germ cell tumors. A Southwest Oncology Group Study. Cancer 1980; 45: 2543–2549.[ISI][Medline]

20. Logothetis CJ, Samuels ML, Selig DE et al. Chemotherapy of extragonadal germ cell tumors. J Clin Oncol 1985; 3: 316–325.[Abstract]

21. Garnick MB, Canellos GP, Richie JP. Treatment and surgical staging of testicular and primary extragonadal germ cell cancer. JAMA 1983; 250: 1733–1741.[Abstract]

22. Toner GC, Geller NL, Lin SY, Bosl GJ. Extragonadal and poor risk non-seminomatous germ cell tumors. Survival and prognostic features. Cancer 1991; 67: 2049–2057.[ISI][Medline]

23. Hainsworth JD, Einhorn LH, Williams SD et al. Advanced extragonadal germ-cell tumors. Successful treatment with combination chemotherapy. Ann Intern Med 1982; 97: 7–11.[ISI][Medline]

24. Bosl GJ, Geller NL, Cirrincione C et al. Multivariate analysis of prognostic variables in patients with metastatic cancer. Cancer Res 1983; 43: 3403–3407.[Abstract]

25. Birch R, Williams S, Cone A et al. Prognostic factors for favorable outcome in disseminated germ-cell tumors. J Clin Oncol 1986; 4: 400–407.[Abstract]

26. The Medical Research Working Party on Testicular Tumours. Prognostic factors in advanced non-seminomatous germ-cell testicular tumours: results of a multicenter study. Lancet 1985; 1: 8–11.[Medline]

27. Saxman S, Nichols CR, Einhorn LH. Salvage chemotherapy in patients with extragonadal non-seminomatous germ cell tumors: The Indiana University Experience. J Clin Oncol 1994; 12: 1390–1393.[Abstract]

28. Broun ER, Nichols CR, Einhorn LH, Tricot GJ. Salvage therapy with high-dose chemotherapy and autologous bone marrow support in the treatment of primary non-seminomatous mediastinal germ cell tumors. Cancer 1991; 68: 1513–1515.[ISI][Medline]

29. Nichols CR. Testicular cancer. In Williams SD (ed.): Current Problems in Testicular Cancer. St Louis, MO: Mosby 1998; 181–274.

30. Dulmet EM, Macchiarini P, Suc B, Verley JM. Germ cell tumors of the mediastinum: a 30-year experience. Cancer 1993; 72: 1894–1901.[ISI][Medline]

31. Delgado FG, Tjulandin SA, Garin AM. Long term results of treatment in patients with extragonadal germ cell tumors. Eur J Cancer 1993; 29A: 1002–1005.

32. Ganjoo KN, Rieger KM, Kesler KA et al. Results of modern therapy for patients with mediastinal non-seminomatous germ cell tumors. Cancer 2000; 88: 1051–1056.[ISI][Medline]

33. Bukowski RM, Wolf M, Kulander BG et al. Alternating combination chemotherapy in patients with extragonadal germ cell tumors. A Southwest Oncology Group study. Cancer 1993; 71: 2631–2638.[ISI][Medline]

34. Vuky J, Bains M, Bacik J et al. Role of postchemotherapy adjunctive surgery in the management of patients with non-seminoma arising from the mediastinum. J Clin Oncol 2001; 19: 682–688.[Abstract/Free Full Text]