Development of a Risk Stratification System to Guide Treatment for Female Germ Cell Tumors

Abstract

Objective

Due to their rarity, little is known about prognostic factors in female germ cell tumors (GCTs) or outcomes following systemic therapy. Management is largely based on studies of male GCT and epithelial ovarian cancer.

Methods

Chart review was performed for all females with GCT seen at Memorial Sloan Kettering Cancer Center (MSKCC) from 1990 to 2012. Patients receiving chemotherapy were stratified using a modification of the male IGCCCG risk system, and the classifier was correlated with outcome.

Results

Of 93 patients, 92 (99%) underwent primary surgery and 85 (92%) received chemotherapy. Modified IGCCCG classification was significantly associated with progression-free survival (PFS) and overall survival (OS), both when applied preoperatively and pre-chemotherapy (p<0.001 for all four analyses). Progression after initial chemotherapy (n=29) was detected by imaging in 14 (48%) patients, by serum tumor markers in 6 (21%) patients, and by multiple methods in the rest. Seven (29%) of 24 patients treated with salvage chemotherapy achieved long-term PFS, including 4/6 who received high-dose chemotherapy (HDCT) as initial salvage versus 3/16 treated with other initial salvage regimens. The estimated 3-year OS rate was 84% (95% CI, 76-92%), with a trend favoring dysgerminoma over non-dysgerminoma histologies (p=0.12).

Conclusions

Modified IGCCCG classification was prognostic for female GCT patients in this cohort and identified a poor-risk group who may benefit from more intensive first-line chemotherapy. Both imaging and tumor marker evaluation were important in identifying relapses after first-line chemotherapy. The majority of long-term remissions with salvage therapy were achieved with initial salvage HDCT.

Introduction

Germ cell tumors (GCTs) are a highly curable malignancy primarily affecting adolescents and young adults. Even with advanced disease, greater than 70% of patients can be cured with standard chemotherapy regimens containing etoposide and cisplatin with (BEP) or without bleomycin (EP). Furthermore, a significant proportion of the remaining patients can still be cured with salvage conventional-dose chemotherapy (CDCT) or high-dose chemotherapy (HDCT) with autologous stem-cell support (ASCT).

GCTs comprise 70% of ovarian neoplasms diagnosed in females ages 10-30, but account for only 1-3% of all ovarian cancers [1]. GCTs originating at other gynecologic sites are even less common. Due to their rarity, female GCTs are difficult to study, and management is largely based on trials of male GCTs and epithelial ovarian cancer [2]. While male and female GCT patients achieve similar responses to first-line chemotherapy, females may have inferior outcomes to salvage treatment [3,4].

The International Germ Cell Cancer Collaboration Group (IGCCCG) classification stratifies male GCT patients requiring chemotherapy into good-, intermediate-, and poor-risk groups, with estimated 5-year OS rates of 91%, 79%, and 48%, respectively [5]. The model is based on histology, primary tumor site, tumor marker levels and sites of metastasis, and not only provides useful prognostic information but also helps guide selection of initial chemotherapy. In good-risk patients, standard options include either 3 cycles of BEP or 4 cycles of EP, whereas intermediate- and poor-risk patients require 4 cycles of BEP [6].

Several adverse prognostic factors have been suggested for women with GCT, including age >40 at diagnosis [7], presence of metastatic disease [7], residual tumor on imaging after initial surgery [8], and slow rate of postoperative AFP decline among patients with yolk sac tumors [9]. However, there is no formal risk stratification system to estimate prognosis or guide treatment. Therefore, we aimed to characterize the distribution of histology, primary site, and stage among females with GCT and develop a method to estimate prognosis. We also sought to determine how recurrence was identified and evaluate responses to salvage therapy for relapsed disease.

Methods

After obtaining Institutional Review Board approval, we used our institutional tumor registry to identify all females with GCT of gynecologic origin seen at Memorial Sloan Kettering Cancer Center (MSKCC) between January 1, 1990 and December 31, 2012. Pathology review-only cases and patients with a non-gynecologic primary tumor site (e.g., mediastinum) were excluded, as were patients with gestational trophoblastic neoplasms (e.g., gestational choriocarcinomas), high-grade endometrial carcinomas with yolk sac features, mature cystic teratomas, and struma ovarii. Patients with teratoma with malignant transformation were included, with sensitivity analysis for progression-free survival (PFS) and overall survival (OS) to see how inclusion of this unique subtype affected results. Histologic classification was based upon pathologic review of tumor specimens at MSKCC.

Individual chart review was performed to document each patient's International Federation of Gynecology and Obstetrics (FIGO) [10] stage and initial management approach (surgery, chemotherapy, and/or surveillance). Patients who received chemotherapy were classified into risk groups according to a modified version of the IGCCCG [5] model for males with advanced GCT (Table 1). Dysgerminomas (termed seminomas in men) and non-dysgerminomas (termed nonseminomas in men) were considered separately, in keeping with the male IGCCCG model [11,12]. Good- and intermediate-risk dysgerminomas were segregated based on the presence or absence of nonpulmonary visceral metastases. Serum tumor marker levels and the presence or absence of nonpulmonary visceral metastases (e.g. metastases to liver, bone, brain) were used to separate non-dysgerminomas into good-, intermediate, and poor-risk groups. Peritoneal metastases were not considered a nonpulmonary visceral metastasis given the frequency of peritoneal involvement by direct extension in ovarian GCT. Patients were classified just prior to first-line chemotherapy (when the IGCCCG system is applied in men) and also prior to initial surgery. Given the more substantial tumor debulking performed during primary surgery for female GCT compared with male GCT (e.g., orchiectomy), the latter analysis was undertaken to determine if preoperative marker levels differed sufficiently from postoperative levels to affect risk classification, and if so, which level was more strongly associated with outcome.

Table 1. Modified International Germ Cell Cancer Collaborative Group (IGCCCG) Risk Model.

	Good	Intermediate	Poor
Dysgerminoma	No metastases other than lung, lymph nodes, or peritoneum	Metastases beyond lung, lymph nodes, and peritoneum	N/A
All other histologies	No metastases other than lung, lymph nodes, or peritoneum AND -AFP <1,000 ng/L -HCG <5,000 mIU/mL -LDH <1.5×ULN	No metastases other than lung, lymph nodes, or peritoneum AND ≥1 of the following -AFP 1,000 – 10,000 ng/mL -HCG 5,000 – 50,000 mIU/mL -LDH 1.5 – 10×ULN	Metastases beyond lung, lymph nodes, and peritoneum OR -AFP >10,000 ng/mL -HCG >50,000 mIU/mL -LDH >10×ULN

Abbreviations: AFP, alpha-fetoprotein; HCG, beta-human chorionic gonadotropin; LDH, lactate dehydrogenase; ULN, upper limit of normal

For patients who underwent post-chemotherapy surgery (defined as exploratory surgery performed <6 months from finishing first-line chemotherapy), pathologic findings were correlated with outcomes. For patients who recurred after first-line chemotherapy, we noted the method by which progression was identified and the salvage chemotherapy regimens received.

Follow-up and survival data were determined from chart review. OS and PFS were analyzed using the Kaplan-Meier method [13]. OS was defined as time from diagnosis until death. Patients still alive were censored on their date of last contact. Progression was diagnosed via imaging, tumor markers, or exam. In cases where progression was diagnosed by both tumor markers and imaging, the date of progression was defined as whichever occurred first. PFS was defined as time from diagnosis until documented progression or death. Patients alive and progression-free were censored at their most recent follow-up. For subset analyses of patients treated with chemotherapy and patients who underwent post-chemotherapy surgery, OS and PFS were defined from start of chemotherapy and date of post-chemotherapy surgery, respectively. The log-rank test was used to assess differences between groups. A two-sided p-value of <0.05 was considered statistically significant. Statistical analyses were performed using the ‘survival’ package in R (version 3.0.1; R Development Core Team). Multivariate analyses were not performed due to a limited number of events.

Results

Between January 1, 1990 and December 31, 2012, 93 females with GCTs of gynecologic origin (median age, 23 years; range, 0-78 years) were managed at MSKCC. Median follow-up for survivors was 63 months (range, 5-236 months).

Patient and Disease Characteristics

Patient and disease characteristics are listed in Table 2. Histology was classified as dysgerminoma, immature teratoma, endodermal sinus tumor, teratoma with malignant transformation, choriocarcinoma, or mixed GCT. In 90 patients (97%), ovary was the primary tumor site. Fifty-three patients (57%) had FIGO stage I disease, 4 (4%) had stage II disease, 29 (31%) had stage III disease, and 7 (8%) had stage IV disease at diagnosis.

Table 2. Patient and Disease Characteristics (N=93).

	N (%)
Age at diagnosis (median, range)	23 (0-78)
Race
White	70 (75%)
Black	18 (19%)
Asian	5 (5%)
Primary tumor site
Ovary	90 (97%)
Fallopian tube	1 (1%)
Vagina	1 (1%)
Vulva	1 (1%)
Histology
Immature teratoma	28 (30%)
Dysgerminoma	24 (26%)
Mixed GCT	20 (22%)
Endodermal sinus tumor	18 (19%)
Teratoma with malignant transformation	2 (2%)
Choriocarcinoma	1 (1%)
Stage
I	53 (57%)
II	4 (4%)
III	29 (31%)
IV	7 (8%)
Sites of metastasis
Peritoneum, serosa, or omentum	28
Retroperitoneal or pelvic lymph nodes	161
Supraclavicular lymph node	3
Lungs	9
Non-pulmonary visceral metastasis
Liver	2
Pleura	3
Other	4

Details of primary treatment are outlined in Table 3. Ninety-two patients (99%) underwent primary surgery and 85 (92%) received chemotherapy. One patient received primary rather than postoperative chemotherapy after biopsy confirmed the diagnosis of GCT. Nearly 80% of the patients received bleomycin, etoposide, and cisplatin (BEP) as their first-line regimen; 14% received carboplatin-containing regimens (mostly through a clinical trial); and the remaining patients received other regimens tailored to unique clinical situations. For example, a patient with teratoma with transformation to gastrointestinal-type tumor received 5-flourouracil, oxaliplatin, and leucovorin (FOLFOX); another with transformation to sarcomatoid-type histology was treated with ifosfamide-based therapy.

Table 3. Primary Treatment Characteristics.

	N (%)
Initial Treatment Characteristics
Surgery	92 (99%)
Chemotherapy	85 (92%)
Chemotherapy regimen
BEP	67 (79%)
Carboplatin-containing	12 (14%)
EP	4 (4%)
Other**	2 (3%)
Number of cycles of chemotherapy
1-2	4 (5%)
3	33 (39%)
4	43 (51%)
>4	5 (6%)
Modified Pre-chemotherapy IGCCCG Group (n=85)
Good	49 (58%)
Intermediate	12 (14%)
Poor	11 (13%)
Unable to classify	13 (15%)
Modified Preoperative IGCCCG Group (n=84)*
Good	27 (32%)
Intermediate	9 (11%)
Poor	13 (15%)
Unable to classify	35 (42%)

Abbreviations: BEP, bleomycin, etoposide, and carboplatin; IGCCCG, International Germ Cell Cancer Collaborative Group

^*One patient not included because she did not undergo primary surgery prior to chemotherapy

^**Other regimens included FOLFOX and ifosfamide/Adriamycin (in patients with teratoma with malignant transformation to gastrointestinal type and sarcoma type tumors, respectively)

Modified IGCCCG Risk Classification

Of 85 patients treated with systemic chemotherapy, 49, 12, and 11 were classified as good-, intermediate-, and poor-risk, respectively, when tumor markers just prior to chemotherapy initiation were used. Thirteen patients could not be classified due to missing tumor markers (Table 3). When preoperative tumor marker levels were considered, 27 patients were good-risk, 9 were intermediate-risk, and 13 were poor-risk. Thirty-five patients could not be classified due to missing markers. Additionally, the patient who did not undergo primary surgery was excluded from the preoperative IGCCCG classification. In 11 patients, classification differed between the two methods (Table 3).

PFS and OS with Chemotherapy Based on Initial Disease Characteristics

PFS and OS were evaluated based on histology, FIGO stage, preoperative IGCCCG group, and pre-chemotherapy IGCCCG group (Figure 1). There were 15 deaths and 25 PFS events (progressions/deaths) within the cohort (n=93); the estimated 3-year PFS and OS rates were 76% and 84%, respectively. The start date of chemotherapy was not available for one patient; therefore, 84 of 85 patients were included in the PFS and OS analyses of the pre-chemotherapy IGCCCG model and 83 of 84 in the analyses of the preoperative IGCCCG model.

Figure 1. Progression-Free Survival by Histology, Stage, Modified IGCCCG Groupˆ.

A. Pre-chemotherapy IGCCCG Group (n=84)

B. Preoperative IGCCCG Group (n=83)

C. Histology (n=90)

D. International Federation of Gynecology and Obstetrics (FIGO) Stage (n=93)

E. FIGO Stage III versus Stage IV (n=36)

ˆFigures are truncated at 60 months.

Both preoperative and pre-chemotherapy IGCCCG models were significantly associated with outcome. The estimated 3-year PFS rates for good-, intermediate-, and poor-risk patients were 88%, 78%, and 31%, respectively, using preoperative markers (log-rank p<0.001); and 85%, 83%, and 18%, respectively, using pre-chemotherapy markers (log-rank p<0.001). The estimated 3-year OS rates were 91%, 80%, and 50% for good-, intermediate-, and poor-risk patients, respectively, using preoperative markers (log-rank p<0.001); and 90%, 89%, and 47%, respectively, using pre-chemotherapy markers (log-rank p<0.001). These rates remained the same at 5 years (Table 4, Figures 1 and 2).

Table 4. Kaplan-Meier Overall Survival Estimates by Histology, Stage, and Modified IGCCCG Risk Group.

Initial Treatment Characteristics	1-year OS (95% CI)	3-year OS (95% CI)	Log-rank p-value#
Full cohort (n=93)	0.94 (0.90-0.99)	0.84 (0.76-0.92)	-
Histology (n=90)*			0.44
Dysgerminoma (24)	0.96 (0.88-1.00)	0.96 (0.88-1.00)
Endodermal sinus tumor (18)	0.94 (0.84-1.00)	0.83 (0.66-1.00)
Immature teratoma (28)	0.92 (0.83-1.00)	0.84 (0.71-1.00)
Mixed GCT (20)	1.00 (1.00-1.00)	0.80 (0.62-1.00)
			0.12
Dysgerminoma (24)	0.96 (0.88-1.00)	0.96 (0.88-1.00)
Other (66)	0.95 (0.90-1.00)	0.83 (0.73-0.93)
Stage (n=93)			0.049
I/II (57)	0.96 (0.91-1.00)	0.90 (0.81-0.99)
III/IV (34)	0.91 (0.82-1.00)	0.77 (0.63-0.94)
Late stage (n=36)			0.018
III (29)	0.97 (0.90-1.00)	0.80 (0.65-0.98)
IV (7)	0.71 (0.45-1.00)	0.57 (0.30-1.00)
Modified pre-chemotherapy ICGGGC Group (n=84)**			<0.001***
Good (48)	0.98 (0.94-1.00)	0.90 (0.80-1.00)
Intermediate (12)	1.00	0.89 (0.71-1.00)
Poor (11)	0.70 (0.47-1.00)	0.47 (0.23-0.94)
Unable to classify (13)	0.91 (0.75-1.00)	0.73 (0.51-1.00)
Modified preoperative ICGGGC Group (n=83)**			<0.001****
Good (26)	0.96 (0.89-1.00)	0.91 (0.79-1.00)
Intermediate (9)	1.00	0.80 (0.52-1.00)
Poor (13)	0.77 (0.57-1.00)	0.50 (0.28-0.89)
Unable to classify (35)	0.97 (0.91-1.00)	0.87 (0.76-1.00)

Abbreviations: GCT, germ cell tumor; IGCCCG, International Germ Cell Cancer Collaborative Group

^*Excluding teratoma with malignant transformation and choriocarcinoma due to small sample sizes.

^**One patient excluded from PFS/OS analyses for both preoperative and pre-chemotherapy models due to lack of a chemotherapy start date. One additional patient excluded from preoperative model because she did not undergo primary surgery.

^***Includes unclassifiable patients. With exclusion of unclassifiable patients, remains P <0.001.

^****Includes unclassifiable patients. With exclusion of unclassifiable patients, P=0.001.

^#Note that the log-rank test compares Kaplan Meier survival function over time.

Figure 2. Overall Survival by Modified IGCCCG Groupˆˆ.

A. Pre-chemotherapy Tumor Markers (n=71)*

B. Preoperative Tumor Markers

ˆˆFigures are truncated at 15 years.

Unclassified patients had similar OS as classified patients in both preoperative and pre-chemotherapy models, and similar PFS in the preoperative model. In the pre-chemotherapy model, PFS appeared to be shorter in unclassified patients (1-year PFS 34% vs 76%), with 13 of 14 having immature teratoma histology.

Histology was significantly associated with PFS, with dysgerminoma patients faring better than those with other histologies. The 3-year PFS rate was 89% for dysgerminoma; 73% for mixed GCT, endodermal sinus tumor, or choriocarcinoma; and 67% for immature teratoma or teratoma with malignant transformation (p=0.044). A trend toward improved OS was observed for dysgerminomas versus other histologies (p=0.12). Sensitivity analyses to assess the effect of including teratomas with malignant transformation showed no significant effect.

PFS was significantly improved for FIGO stage I-II versus III-IV disease (p=0.029); and in the subset of patients with advanced-stage disease (III/IV), PFS was significantly worse for stage IV versus stage III disease (p<0.001). Similarly, OS was significantly improved for FIGO stage I-II vs III-IV disease (p=0.049); and in advanced-stage patients, OS was significantly worse for stage IV versus stage III disease (p=0.018, Table 4).

Post-chemotherapy surgery

Post-chemotherapy surgery was performed in 25 (29%) of 85 patients treated with chemotherapy. Pathology results were divided into three categories: necrosis only (n=9), mature or immature teratoma (n=9), and viable GCT or teratoma with malignant transformation (n=7), with significantly different PFS (p=0.017) and OS (p<0.001) rates. Inferior outcomes were observed for patients with residual viable GCT or teratoma with malignant transformation (Supplemental Figure 1).

Detection of recurrence

Twenty-nine patients (31%) experienced disease recurrence; 26 relapsed after first-line chemotherapy and 3 after surgery alone. This relatively high recurrence rate was felt to be due to referral bias; many patients came to MSKCC only at the time of relapse, and some who were referred for upfront therapy were referred to MSKCC because they were felt to be higher-risk at the time of diagnosis. Of those who recurred, one patient had a localized vulvar recurrence detected by physical examination, 14 (48%) had recurrence diagnosed solely by imaging studies, 6 (21%) solely by rising tumor markers, and 8 (28%) using both markers and imaging. Of the 22 patients for whom imaging aided detection of recurrence, CT scan was used most frequently (n=18, 82%), followed by PET/CT (n=2, 9%), and then MRI abdomen/pelvis and ultrasound (n=1 each).

Outcomes to salvage chemotherapy

Twenty-four of 26 patients who relapsed after first-line chemotherapy received salvage chemotherapy, and 2 patients were managed with surgery alone. The initial salvage regimen was HDCT plus ASCT in 6 patients, cisplatin-based CDCT (VIP, TIP) in 7, repeat BEP or EP in 5 (all at outside centers), and other non-cisplatin-based therapy in 6. In total, 7 (29%) of 24 patients treated with salvage chemotherapy achieved long-term PFS; 4 of these 7 patients received HDCT as initial salvage therapy. Of the remaining 3 patients, one received salvage TIP followed by resection of residual teratoma; another achieved a durable clinical remission to ifosfamide, carboplatin and etoposide; and one recurred with teratoma with transformation to primitive neuroectodermal tumor (PNET) and received the P6 regimen directed at that histology, consisting of high-dose cyclophosphamide, doxorubicin, and vincristine alternating with ifosfamide and etoposide [14].

HDCT was attempted in 7 other patients later in their disease course. In 2 patients, it was not possible to harvest sufficient stem cells to support ASCT. Of the 5 who completed HDCT with ASCT as later salvage, only one (20%) achieved long-term PFS.

Discussion

Due to their rarity, little is known about the factors associated with outcome following systemic chemotherapy for female GCT. Since there is no established method for predicting outcomes in women with GCT, there is no reliable way for clinicians to identify specific patient subsets that require treatment with more intensive chemotherapy or closer follow-up to evaluate for recurrence. Furthermore, there are limited data on the optimal management of recurrent disease. Our goal was to address some of these issues through retrospective analysis of our large, single-institution database.

Predictors of Prognosis

Our modified version of the male IGCCCG risk classification was significantly associated with PFS and OS in both the preoperative and pre-chemotherapy settings. In particular, patients classified as poor-risk based on pre-chemotherapy markers had extremely poor 3-year PFS (18%) and 5-year OS (47%) rates. In contrast to the male IGCCCG staging system, our model did not differentiate well between good- and intermediate-risk patients; both groups achieved favorable outcomes. This observation could relate to the small number of patients in the intermediate group, a lack of importance of intermediate-level marker cutoffs in female GCT, or some other reason not yet identified.

One limitation of our study was the inability to assign a risk group to some patients, primarily because tumor markers were unavailable. Possible reasons markers were not obtained include situations in which clinicians felt compelled to start treatment urgently without time to wait for tumor marker results, cases where initial treatment occurred at a facility with less experience managing GCTs, or simply because clinicians felt the prognostic importance of serum tumor markers had not been established. Notably, 13 of 14 patients with missing pre-chemotherapy tumor markers had immature teratoma, and providers may have believed tumor markers would not help in that setting. While our findings need to be validated in an independent cohort, they suggest that tumor markers obtained before surgery and before chemotherapy can help estimate the likelihood of recurrence and survival.

In addition to aiding prognostication, our risk classification could also be used to refine recommendations for first-line chemotherapy. In male GCT, the IGCCCG risk system serves as the basis for chemotherapy decisions, with good-risk patients receiving either 3 cycles of BEP or 4 cycles of EP [15-17], and intermediate- and poor-risk patients receiving 4 cycles of BEP [18]. In contrast, without a formal method to risk-stratify patients, NCCN guidelines recommend either 3 or 4 cycles of BEP for all female GCT (except stage I dysgerminomas and stage I, grade I immature teratomas, for which chemotherapy is not recommended) [19], with no guidance as to which patients should receive 3 versus 4 cycles. As such, currently, some women with favorable disease features may be over-treated and experience unnecessary toxicity, and some women with unfavorable features may be undertreated, leading to decreased durable remission rates. Development of a prognostic classification system is the first step needed to establish risk-based treatment algorithms that minimize toxicity and maximize efficacy. This could also allow for the testing of alternative or more intensive regimens in the poor-risk population.

FIGO staging was also a significant predictor of PFS and OS in our cohort, although the small numbers of patients with stage II and stage IV disease prevented us from considering each stage independently. Instead, we combined early-stage (I and II) patients together for comparison against a combined group of advanced-stage (III and IV) patients. However, in the group of patients with advanced-stage disease, even withstanding the small numbers, we were able to demonstrate significantly worse PFS and OS for patients with stage IV vs. stage III GCT, suggesting that disease that has metastasized to other organs is a strong predictor of poor outcome in female GCTs as well as in the epithelial ovarian tumors for which the FIGO staging system was initially designed.

Histology was significantly associated with longer PFS but not OS for dysgerminomas compared to other histologies. Larger studies are needed to further assess the relationship between histology and outcome in this disease.

Second-look surgery (an exploratory surgery performed in the absence of radiographic evidence of residual disease to determine whether a patient is truly pathologically free of disease) for malignant ovarian GCT has fallen out of favor in recent years due to studies showing little benefit, except when tumor is incompletely resected at initial surgery [20,21]. However, post-chemotherapy surgery remains an important part of management for patients with residual tumor masses on imaging in the setting of normalized serum tumor markers. In our cohort, viable tumor (embryonal carcinoma, yolk sac tumor, choriocarcinoma, or dysgerminoma) found at post-chemotherapy surgery was associated with shorter PFS. This finding supports our approach to administer 2 additional cycles of adjuvant chemotherapy to patients with these histologies present in the post-chemotherapy specimen. This strategy has also previously been demonstrated to result in superior outcomes in men [22]. If only mature teratoma is removed at post-chemotherapy resection, no further treatment is necessary since mature teratoma is resistant to both chemotherapy and radiation.

Since nearly all of the patients in this cohort had primary surgery, our data do not allow conclusions regarding when or in whom administration of chemotherapy should be administered prior to debulking surgery. However, our approach to female patients with advanced GCT is to considerchemotherapy prior to surgery only in patients who are not surgical candidates at the time of diagnosis.

Recurrence

Although current guidelines [19] recommend against routine CT imaging in female GCT patients with previously elevated tumor markers, imaging was an important part of surveillance in our cohort, providing the only evidence of recurrence in 48% of relapsing patients. With limited data on the utility of imaging for surveillance in female GCT, this topic deserves further study. In male GCTs, the NCCN recommends CT imaging every 6-12 months for 5 years following a complete response to first-line chemotherapy [23]. Our data suggest that in women, CT may also enable earlier diagnosis of recurrence at a time when treatment might be more likely to lead to durable remission. Furthermore, adding CT imaging to surveillance guidelines might provide women with advanced GCT increased access to clinical trials. Currently, since insurance may not cover protocol-mandated post-chemotherapy CT scans categorized as billable by investigators some women with GCT may not be able to participate in trials that include both genders.

Salvage Chemotherapy

Literature on salvage therapy for women with relapsed GCT is even sparser than in the primary setting due to the rarity of the disease and because few women have been enrolled in salvage chemotherapy trials. To our knowledge, the 24 female patients who received salvage chemotherapy in our study is the largest group yet reported. Despite categorizing salvage regimens in several ways, no significant associations between treatment approach and outcome could be identified, likely because of the small number of patients in each subgroup. Nevertheless, there were two important findings in this analysis.

First, of 5 patients who received repeat EP or BEP (all at outside institutions), none experienced long-term PFS. We do not believe this was related to confounding (e.g., administration of EP or BEP rather than HDCT because a patient was felt to be too sick to tolerate more aggressive therapy), as all these patients went on to receive further chemotherapy at MSKCC, including later-line salvage HDCT. Given these results, we believe repeating first-line therapy should be avoided in the salvage setting. Instead, HDCT with ASCT should be strongly considered, with CDCT (TIP or VeIP) [24,25] as an alternative if HDCT is contraindicated or not felt to be in the patient's best interest at that time.

Second, the long-term PFS achieved in 4 of 6 patients who received HDCT as initial salvage shows this approach can be curative in women with recurrent GCT. In 2/7 patients in whom HDCT was attempted as third-line or later treatment, an insufficient number of stem cells were harvested due to the extent of prior treatment, and in only one of the remaining 5 patients (20%) was long-term PFS achieved. Recently reported data from Indiana University suggest a similar pattern: of 12 ovarian GCT patients who underwent HDCT for recurrent disease, long-term PFS was achieved in 3 (60%) of 5 who received HDCT as initial salvage compared with only 1 (14%) of 7 who received HDCT as second or later salvage therapy [26]. While additional research is needed to better define which relapsed patients benefit most from initial salvage HDCT, both studies highlight the importance of referring these patients to experienced transplant centers to ensure HDCT is given appropriate consideration as initial salvage therapy and underscore the need for collaborative studies to assure adequate sample size for data analysis.

Conclusion

This study describes the clinical features and treatment outcomes of a large group of women with GCT. For patients receiving first-line chemotherapy, our modified IGCCCG model was highly prognostic for PFS and OS, and identified a poor-risk group who may benefit from more intensive induction chemotherapy and post-treatment surveillance. We also found imaging to be helpful in detecting relapse. Finally, re-treatment with first-line regimens in the salvage setting led to universally poor outcomes, whereas initial salvage therapy with HDCTdemonstrated curative potential and should be strongly considered in this setting.

Research Highlights.

• This manuscript details the largest investigation of patients with malignant germ cell tumors (GCTs) of gynecologic origin ever reported

• We developed a risk stratification system for gynecologic GCTs that was highly prognostic for progression-free and overall survival

• This prognostic system may help to establish risk-based treatment algorithms that minimize toxicity and maximize efficacy in this population