A Review of Second-line Chemotherapy and Prognostic Models for Disseminated Germ Cell Tumors

The majority of patients who are diagnosed with germ cell tumors (GCT), including those who present with disseminated disease, has an excellent prognosis. Approximately 70% to 80% of today’s patients with advanced GCT are cured with 1^st-line cisplatin-based combination chemotherapy and adjunctive surgery.^1–5 Achievement of a complete remission (CR) to such treatment signifies an excellent chance of cure. Less than 10% of these patients relapse.⁶ However, an estimated 20% to 30% of patients presenting with metastatic GCT are either refractory to or relapse following initial treatment.

The development of the International Germ Cell Cancer Collaboration Group classification of risk in 1997⁷ improved our ability to estimate the risk of failure to 1^st-line chemotherapy. It predicts that with standard cisplatin-based chemotherapy, approximately 10% of good-risk, 25% of intermediate-risk, and more than 50% of poor-risk patients will fail such treatment. While in most malignancies, refractory or relapsed disease that is not both localized and amenable to surgical resection is considered incurable, a significant proportion of such patients with GCT can still be cured with either salvage conventional-dose chemotherapy (CDCT) or high-dose chemotherapy (HDCT) plus stem-cell rescue with or without adjunctive surgery in the 2^nd- or even 3^rd-line setting. A small proportion of patients with relapsed disease that is locally confined may achieve durable remissions with surgical resection alone.^8–10

Since only a small number of patients require salvage therapy, few large prospective studies have been done to guide management in this setting. Rather, most data derive from phase II trials and retrospective series. The optimal regimen, both for CDCT and HDCT, remains to be determined. Other central questions include 1) How to identify those patients likely to be cured with CDCT, thus avoiding the toxicity of high-dose therapy, versus 2) those patients, who are unable to achieve cure with CDCT but might be cured with 2^nd-line HDCT.

This article reviews reports of initial salvage chemotherapy, including CDCT and HDCT, in patients with disseminated GCT refractory to or relapsing after standard 1^st-line chemotherapy. We also discuss prognostic models as well as the salvage approach for special patient populations such as those with late relapse.

Conventional Dose Chemotherapy in the 2^ND-line setting

Following seminal work in the 1970s and 1980s that established cisplatin as the essential component for treatment of disseminated disease ^{11, 12} and marked the beginning of the modern era of chemotherapy for GCT, PVB (cisplatin, vinblastine and bleomycin) was regarded as the standard of care in first-line management of advanced GCT. Subsequent investigators identified salvage regimens for the 20% of patients unable to achieve CR and the 10%–20% of patients who relapsed following PVB. In the early 1980s, etoposide was found to have activity in refractory GCT,¹³ and subsequently the combination of etoposide and cisplatin (EP) was shown to exhibit incomplete cross-resistance with PVB,^{14, 15} and thus became standard salvage for patients with disease progression following PVB. Durable remissions to salvage therapy with EP, however, remained rare, and investigators continued searching for agents to combine with cisplatin and/or etoposide.

Cisplatin, Ifosfamide, and Vinblastine (VeIP) or Etoposide (VIP)

In the mid 1980s, ifosfamide was found to have efficacy in patients with refractory GCT.^{16, 17} One trial using single-agent ifosfamide in heavily pretreated refractory patients reported an objective response rate of 23%.¹⁷ However, the median duration of response was short, and median overall survival was only 3.5 months. Several groups subsequently combined ifosfamide and cisplatin with either etoposide (VIP) or vinblastine (VeIP) to treat refractory disease.^18–23

Initial reports from American and European groups included mixed populations receiving VIP or VeIP in the 2^nd-line, 3^rd-line, or later setting.^{18, 24–26} These trials established a 5-day regimen of cisplatin (100 mg/m² daily x 5 days) and ifosfamide (1.2 g/m² daily x 5 days), with either etoposide (75 mg/m² daily x 5 days) or vinblastine (0.11 mg/kg x 2 days) administered every 3 weeks as standard salvage chemotherapy, with CR rates in 25%–36% of patients and median response durations between 3.5 and 34 months. Nephrotoxicity and myelosuppression were the main toxicities with these regimens.

Building on these reports, investigators evaluated these regimens in the initial salvage setting (Table 1).^27–31 In one series, 56 patients with advanced GCT resistant to one prior cisplatin-containing regimen were treated with 4 cycles of either VIP or VeIP.²⁹ The reported CR rate was 36%. An additional 15% of patients who did not undergo resection of residual disease achieved marker-negative PR. With a median followup of 52 months, sustained CRs were seen in 23% of patients, and median overall survival (OS) was 18 months. French investigators from the Institut Gustave Roussy retrospectively reported experience with 2^nd-line VIP / VeIP in 54 patients.²⁸ The CR rate following VIP / VeIP was 44%, with sustained CR in 19% of treated patients. However, nearly one-third of responders received consolidation with HDCT, making it difficult to interpret the durability of remissions to VIP / VeIP alone. With a 30-month median follow-up, sustained CRs were observed in 63% of those who attained a CR and then received HDCT versus 26% of those with CR without subsequent consolidative treatment. Toxicity profiles were similar for VIP and VeIP; 2 treatment-associated deaths occurred in patients receiving subsequent HDCT. A series from Italy documented 36 patients who failed 1^st-line therapy with either PVB or BEP (bleomycin, etoposide, cisplatin) and went on to receive 2^nd-line treatment with modified versions of VIP or VeIP.²¹ In an attempt to reduce renal toxicity, the dosing schedule was altered to avoid simultaneous administration of platinum and ifosfamide; dosages of ifosfamide and etoposide were also slightly reduced, while cisplatin dose was increased. The CR rate for all evaluable patients was 56%; sustained CR rates were 42%. Remission rates were higher for patients with prior PVB treatment who received modified VIP than for those receiving modified VeIP for progressive disease after BEP treatment. This result was not surprising given the known superiority of BEP over PVB in the 1^st-line setting. Renal toxicity was less pronounced than in previous reports of patients with heavier pretreatment.

Table 1.

VeIP/VIP for initial salvage therapy of advanced GCT

Reference	Regimen	Retrospective vs Prospective	No. Evaluable	CR (%)	Durable CR (%)
Pizzocarro et al21	VIPa or VeIPa	R	36	20 (56)	15 (42)
Farhat et al28	VIPb or VeIPb	R	54	24 (44)	10 (19)b
McCaffrey et al22	VIP or VeIP	R	56	20 (36)	13 (23)
Loehrer et al23	VeIP	P	135	67 (50)	32 (24)
Pico et al31	VIP or VeIP	P	122	51 (42)	31 (25)

Abbreviations: No., number; CR, complete response; VIP, etoposide, ifosfamide, and cisplatin; VeIP, vinblastine, ifosfamide, and cisplatin; R, retrospective; P, prospective; NS, not stated.

^aModified dosing schedule as outlined in text

^bSeven of 24 patients with CR went on to receive consolidative high-dose chemotherapy with stem cell rescue.

No prospective trial has compared the efficacy of VIP vs VeIP in the 2^nd-line setting. With PVB as the previous standard 1^st-line regimen for disseminated GCT, however, VIP had generally been used more frequently for initial salvage, since vinblastine was part of PVB. In 1987a pivotal randomized trial by the Southeastern Cancer Study Group comparing BEP versus PVB in the 1^st-line treatment of disseminated GCT demonstrated superiority for BEP in the advanced disease subgroup and overall significantly less toxicity.³ It not only established this regimen as the standard of care for 1^st-line, but also made VeIP the 2^nd-line therapy of choice, since vinblastine had been replaced by etoposide in the 1^st-line setting. The first prospective study to exclusively evaluate VeIP in the 2^nd- line setting was reported by Loehrer et al in 1998,³⁰ and included 135 patients who were treated after failing a cisplatin/etoposide-based 1^st-line regimen, mostly standard BEP (88%). Of 135 assessable patients, 67 (50%) achieved disease-free status; 32 (24%) remained in sustained CR with a minimum follow-up of 6 years. In this single-center study, 2-, 3-, and 7-year survival rates were 38%, 35%, and 32%, respectively.³⁰ Prospective data for 2^nd-line VIP/VeIP are available from the European multicenter randomized phase III IT-94 trial,³¹ which compared CDCT (VIP or VeIP) to 3 cycles of CDCT consolidated with one cycle of HDCT. The VIP/VeIP arm treated 136 patients and yielded a CR rate of 42% with sustained CRs in 26% at a median follow-up of 45 months. The trial’s findings comparing CDCT versus HDCT are discussed in detail below.

Paclitaxel-containing Regimen

Early trials of paclitaxel in metastatic GCT were conducted after antitumor activity was seen in other malignancies, including breast and ovarian cancer,^{32, 33} and efficacy was demonstrated in platinum-refractory ovarian cancer,³⁴ making paclitaxel an attractive option for salvage therapy of disseminated GCT. Several small phase II trials of single-agent paclitaxel were reported. A study from MSKCC treated 31 patients with limited prior therapy (2^nd- and 3^rd-line setting) and poor prognostic features, such as incomplete response to prior therapy (76%) or extragonadal primaries (29%).³⁵ The results showed an overall response rate (ORR) of 26% including 3 CR (10%); all responders remained disease-free at the time of report. Three of 8 patients with treatment response had primary mediastinal NSGCT (PMNSGCT). Similarly, a German phase II trial treated 24 patients with a median of 7 prior cycles of platinum-containing therapy. The ORR was 25%, including PR in 2 patients who had previously received HDCT and one with primary mediastinal (PM) NSGCT. A CR rate of 8% was reported for this trial; the median response duration was 8 months. Additional early-phase trials in more heavily pretreated patients demonstrated slightly lower response rates, but confirmed single-agent activity for paclitaxel.^36,37

Following these encouraging results, combination regimens containing paclitaxel were undertaken. Preclinical data showing synergy with cisplatin ³⁸ led to the TIP regimen, which combined paclitaxel, ifosfamide, and cisplatin. Several phase II trials evaluated TIP in the initial salvage setting (Table 2).^39–41 The original report ⁴² was a phase I/II study of 30 patients who received one prior platinum-containing regimen. In the phase I portion, patients received escalating doses of paclitaxel with 250 mg/m² as a 24-hour continuous infusion established as the maximal tolerated dose (MTD) for the phase II portion of the trial. Paclitaxel was combined with ifosfamide 6 g/m² plus mesna and cisplatin 100 mg/m² per 21-day cycle, each administered over 4–5 days following completion of paclitaxel infusion. All patients received prophylactic growth factor support with G-CSF on days 7 to 18. Recognizing earlier reports of low response rates to CDCT salvage therapy in patients with poor -risk features, and hypothesizing that such patients would likely benefit from more aggressive salvage therapy upfront, enrollment was limited to patients with favorable features, requiring all of the following: 1) prior treatment limited to one program, or ≤ 6 prior cycles of cisplatin, 2) gonadal primary site, and 3) best prior response to 1^st-line chemotherapy of CR of any duration or marker-negative PR lasting 6 months or longer. Following encouraging results with the first 30 patients, additional subjects were accrued to the phase II portion of this trial, and the updated analysis of 46 patients with 2^nd line TIP was published in 2005.³⁹ With a median follow-up of nearly 6 years, 32 (70%) of 46 patients achieved CR and 29 (63%) were continuously disease-free. Myelosuppression, the major toxicity, was managed with growth factor support, while nephrotoxicity and neurotoxicity were not more severe than that described with other ifosfamide-based salvage CDCT regimens.

Table 2.

Phase II initial salvage therapy with TIP in advanced GCT

Reference	Drug doses (per cycle)	Selection Criteria	Evaluable	CR (%)	Additional Efficacy Data
Kondagunta et al 39	Paclitaxel: 250 mg/m2	1st line salvage in	46	32	susCR:
	Ifosfamide 6000 mg/m2	pts with		(70)	63%
	Cisplatin 100 mg/m2	favorable			2-year PFS
		featuresa			65%
					2-year OS
					78%
Mead et al 40	Paclitaxel: 175 mg/m2	1st line salvage	43	13	1-year FFS
	Ifosfamide 5000 mg/m2	following failure		(31)	38%
	Cisplatin 100 mg/m2	to BEP			1-year OS
					70%
Mardiak et al 41	Paclitaxel: 175 mg/m2	1st line salvage	17	7	2-year DFS
	Ifosfamide 6000 mg/m2	following failure		(41)	47%
	Cisplatin 100 mg/m2	to cisplatin-			2-year OS
		based regimens			64%

TIP, paclitaxel, ifosfamide, cisplatin; CR, complete response; susCR, sustained CR (median follow-up 69 months); PFS, progression-free survival; FFS, failure-free survival; OS, overall survival; DFS, disease-free survival, PR, partial response.

^aMust meet all the following: 1) gonadal primary site, 2) maximum 6 prior cycles of cisplatin 3) CR or PR with negative tumor markers following 1^st-line chemotherapy.

A second phase II study of 2^nd-line TIP was conducted as a multicenter trial through the British Medical Research Council (MRC).⁴⁰ Patients received paclitaxel 175 mg/m², ifosfamide 5 g/m², and cisplatin 100 mg/m² per cycle without prophylactic growth factor support. Fifty-one patients were included, all had received first-line therapy with BEP, and risk stratification was not part of the eligibility assessment (ie, patients with PM-NSGCT as well as patients with late relapse were included). The CR rate was 31%. With a median follow-up of 26 months, the one-year failure-free survival (FFS) and OS were 38% (95% CI, 23%–53%) and 70% (95% CI, 56%–84%), respectively. Less favorable features in this patient population (compared with those in the MSKCC study) may account for the lower CR rate. However, a subgroup analysis of the MRC trial demonstrated that patients with good-risk features had a CR rate of 35% with one-year FFS and OS of 43% (95% CI, 23%–63%) and 81% (95% CI, 64%–98%), respectively.⁴⁰ Similarly, a small phase II trial from Slovakia treated 17 patients after failure to cisplatin-based 1^st-line therapy (mostly BEP [88%]) with 4–6 cycles of 2^nd-line dose-attenuated TIP (paclitaxel 175 mg/m², ifosfamide 6 g/m², and cisplatin 100 mg/m² per cycle). The CR rate, 2-year disease-free survival (DFS), and 2-year OS were 41%, 47% (95% CI, 23%–71%), and 64% (95% CI, 37%–91%), respectively⁴¹. The authors identified the lower dose of paclitaxel as the primary reason for the inferior response rates and survival compared to the original report from MSKCC³⁹, and this may similarly explain the more modest outcomes to TIP in the MRC study..

On the basis of phase II data for TIP and VeIP, a randomized phase III trial compared these regimens in the 2^nd-line setting, but terminated early because of poor accrual. Hence, both VeIP and TIP continue to be considered standard conventional-dose regimens in the salvage setting, although the single-arm series of TIP suggest this is the more active regimen.

Salvage High-dose Chemotherapy

Following success with autologous bone marrow in relapsed hematologic malignancies during the 1970s,^43–48 studies were undertaken in several solid malignancies in the early 1980s.^49–52 This treatment was investigated in GCT because of known chemosensitivity, the dose-response phenomena of individual drugs with synergistic action, the rare occurrence of bone marrow metastasis, and a young patient population with low incidence of significant comorbidities. Characteristics of drugs chosen for these studies included: 1) antitumor activity in conventional doses without evidence that the dose-effect plateau had been reached, and 2) myelosuppression, even at high doses, being the dominant adverse effect, a toxicity that could be modulated by autologous bone marrow infusion. In the case of GCT, carboplatin, etoposide, and cyclophosphamide fulfilled these requirements. Carboplatin was chosen over cisplatin, since its predominant toxicity is myelosuppression, but other adverse effects such as neurotoxicity and nephrotoxicity were not seen. Early phase I studies established safety for high-dose carboplatin and etoposide as single agents with and without the use of stem cell rescue.^53–56

One noteworthy study of HDCT in GCT was a phase I/II trial reported in 1989.⁵⁷ The phase I portion established a regimen of etoposide 1200 mg/m² and carboplatin 1500 mg/m² per cycle as the recommended dose. Thirty-three treatment-refractory patients received 1–2 cycles of HDCT followed by autologous stem cell rescue. Toxicities were significant–all patients developed severe myelosuppression, and 7 treatment-related deaths (21%) were reported. However, the CR rate was 25% despite heavy pretreatment, and a subsequent publication of long-term followup (including additional patients treated on this program) reported sustained CR rates of 15%.⁵⁸ These results prompted further study of this treatment program in a multi-institutional phase II protocol sponsored by the Eastern Cooperative Oncology Group (ECOG). Similar results were achieved (ORR 44%, CR sustained >1 year 13%, treatment-related deaths 13%).⁵⁹ Together, these studies demonstrated that HDCT could be curative, even in the 3^rd-line setting. They established the combination of HD carboplatin and etoposide as the basis for subsequent high-dose protocols.

Later efforts focused on confirming activity, increasing the number of patients who achieved durable responses, and reducing toxicity. Better tolerability has largely been achieved through the routine use of growth factor support as well as peripheral blood stem cells (PBSC) in lieu of autologous bone marrow (BM) rescue.^{60, 61} General improvements in supportive care and antibiotics also diminished the toxicity associated with HDCT, and with all of the above measures, treatment-related mortality has improved from greater than 20% initially to typically less than 3%.^62–64 Strategies to improve efficacy have included intensification of HD carboplatin / etoposide ^{60, 63} as well as addition of other agents.^65–68 In 2007 an Indiana University group published a retrospective evaluation of 184 patients treated between 1996 and 2004,⁶³ 73% of whom were in the initial salvage setting. The HD regimen consisted of 2 cycles of carboplatin 2100 mg/m² and etoposide 2250 mg/m², both administered over 3 days and supported by autologous stem cell reinfusion, given in some patients following 1–2 cycles of VeIP. After a median follow-up of 4 years, 63% of patients were continuously disease-free. Primary mediastinal NSGCTs and late relapses were not included in this trial because of previously observed poor outcomes with HDCT in these subgroups.⁶³ Other investigators have incorporated additional agents into HDCT programs with carboplatin and etoposide. This incorporation has been tolerable in the case of ifosfamide,^{65, 68} cyclophosphamide,⁶⁶ and paclitaxel,⁶⁴ but resulted in excessive toxicity for other combinations such as the addition of thiotepa.⁶⁷

HDCT in the 2^nd-Line Setting

The choice of CDCT versus HDCT is controversial in the initial salvage setting, as there are limited data for guidance. A point made in favor of HDCT was its better tolerability in patients with less prior therapy.⁶⁹ A large single-center series of 150 patients published by Rick et al in 1998 retrospectively reported a heterogeneous population with respect to the timing of HDCT (1^st-line vs initial salvage vs later salvage).⁶⁸ A program of 3 cycles of CDCT followed by one cycle of HD VIP was used. For patients with good-risk Beyer scores (see section on Prognostic Factors),⁶⁰ the 2-year OS declined significantly depending on whether HDCT was used as 1^st-line, 1^st- salvage, or subsequent salvage (78%, 66%, and 47%, respectively; P<.05). In a retrospective report from Indiana University, 65 patients who received 2 cycles of HD carboplatin and etoposide with autologous stem cell rescue as initial salvage following relapse after standard cisplatin-based 1^st -line therapy, demonstrated a continuous DFS rate of 57% with or without surgery at a median follow-up of more than 3 years.⁷⁰ All patients on this trial had one or more favorable features such as gonadal primary tumor and CR or marker-negative PR to 1^st--line therapy.

Following these encouraging early results, European investigators retrospectively analyzed 193 patients treated with CDCT (119) versus HDCT (74) for 1^st-salvage treatment at several centers.⁷¹ A matched-pair analysis between the 2 groups was undertaken using prognostic factors recognized in prior studies,^{30, 60, 72–74} including primary site, response to 1^st -line therapy, relapse-free interval, and tumor marker levels. Thirty-eight pairs of patients from both groups were fully matched (all 5 criteria), and an additional 17 pairs were partly matched (4/5 criteria including primary site and response to 1^st-line therapy). For these 55 pairs, multivariate analyses examined event-free survival and OS as primary endpoints. Hazard ratios for OS favoring HDCT ranged between 0.77 and 0.83 (95% CI, 0.60%–0.99%), whereas those for event-free survival showed a trend ranging between 0.72 and 0.84 (95% CI, 0.59%–1.01%). Of note, salvage regimens in the CDCT group contained ifosfamide and etoposide in 36% and 63% of patients, respectively, whereas 100% of patients in the HDCT group were treated with 3 cycles of VIP followed by one cycle of HD carboplatin / ifosfamide /etoposide.

The only prospective, randomized study to address the question of the optimal initial salvage approach was IT-94, a randomized phase III trial comparing HDCT to CDCT for 2^nd-line therapy. This European multicenter study enrolled 280 patients from 43 institutions in 11 countries between 1994 and 2001.³¹ Through 1:1 randomization, the trial compared the efficacy of 4 cycles of CDCT using VIP/VeIP versus 3 cycles of the same CDCT followed by one cycle of CarboPEC (HDCT using carboplatin [200–550 mg/m²], etoposide [1800 mg/m²] and cyclophosphamide [200 mg/kg]) with stem cell rescue in patients who failed 1^st-line cisplatin-based therapy (85% standard BEP or EP regimens). Response to salvage treatment was similar in both groups, with CR rates for CDCT and HDCT of 42% and 43%, respectively; treatment-related mortality was 3% and 7%. With a median follow-up of 45 months, sustained CR rates were 26% and 35%. No survival benefit was seen for the HDCT arm. This study has recognized limitations: 1) Only one cycle of HDCT was given, whereas 2–3 cycles of HDCT are widely accepted as the standard of care, 2) In the HDCT group, only 81% of patients actually proceeded to CarboPEC following 3 cycles of CDCT, 3) Patients refractory to 1^st-line platinum-containing chemotherapy were excluded, 4) Many centers recruited as few as 1 to 3 patients in the trial, which may have compromised safety and efficacy of HDCT.

An unplanned subgroup analysis from IT-94 demonstrated that among those patients who achieved a CR to salvage chemotherapy, 2-year DFS was superior in the HDCT arm. The authors concluded that future strategies for optimizing HDCT should include multivariate prognostic analyses to identify those patients who were likely to benefit from this approach.³¹ A large, more recent international study took this approach. A retrospective analysis of initial salvage chemotherapy in almost 1600 subjects treated at multiple centers worldwide was performed to identify prognostic factors.⁷⁵ Approximately equal numbers of patients were treated with CDCT and HDCT, respectively. The use of modern 1^st-line and salvage regimens was required for inclusion in the report. On multivariate analysis, prognostic factors were identified that allowed patient stratification into 5 well-defined categories. These data have been further developed, and through inclusion of additional patients, have been used to develop a prognostic model for initial salvage therapy regardless of treatment intensity (see Prognostic Models).⁷⁶ There was a similar distribution among the 5 prognostic categories for CDCT and HDCT. Despite this, superior progression-free survival (PFS) and OS were seen for HDCT in each category with the exception of OS in the low-risk group.⁷⁵ Because of the retrospective nature of this analysis, selection bias is likely a factor behind the favorable outcomes in patients treated with HDCT. A prospective trial comparing sequential HDCT with CDCT as initial salvage is planned.

Choosing 2^nd -line therapy for relapsed or recurrent GCT varies. HDCT is considered a standard approach by some investigators, while others favor reserving HDCT for the 3^rd-line setting in order to avoid unnecessary toxicity in patients who might potentially be cured with CDCT. The approach at MSKCC is to base initial salvage chemotherapy decisions on prognostic models for CDCT and HDCT.

Prognostic Factors

While prognostic factors at the initiation of 1^st-line chemotherapy for disseminated GCT are universally accepted, there has previously been no agreement on risk stratification after failure of 1^st-line therapy. Clinical variables that could help predict response to salvage therapy had largely been based on relatively small or single-center retrospective series and early-phase clinical trials. The unresolved question of CDCT vs HDCT has further complicated stratification, since prognostic factors had typically been analyzed separately for each entity and could not be readily applied to each other.⁷⁷ Lastly, patient populations are heterogeneous as to the extent of prior therapy, as well as the 1^st-line and salvage regimen used, making development of a practical model challenging. While several groups have systematically applied their findings in selecting patients for subsequent trials and determining whether to opt for CDCT or HDCT,^{39, 63, 64} agreement had not been reached on a universal model until recently.

For CDCT salvage regimens, earlier retrospective series and phase II trials identified best response to 1^st-line therapy and location of primary tumor as predictors of response to 2^nd- and 3^rd-line therapy.^{21, 26, 29, 30} Our group retrospectively reviewed 124 patients who had been treated on 4 prospective trials and failed to achieve a sustained CR, 94 of whom went on to receive 2^nd-line therapy. It was determined that response to salvage therapy was significantly enhanced in patients with a prior CR to 1^st-line chemotherapy, testis primary site, normal LDH, normal HCG, and one site of metastasis.⁷² Similarly, a European multicenter retrospective analysis of 164 patients treated in the initial salvage setting reported 3 independent variables of prognostic significance, including progression-free interval, response to induction treatment, and serum marker levels (HCG and AFP) at relapse.⁷⁴ Multivariate analysis from a German report of 60 patients treated with initial salvage chemotherapy identified age less than 35 years, CR to primary treatment, and relapse-free interval greater than 3 months as independent predictors for successful salvage treatment.⁷⁸

For salvage HDCT, a number of specific variables predictive of a poor outcome have been reported, including pretreatment marker levels of HCG,⁶⁶ primary mediastinal (PM) NSGCT, ^{79, 80} absolute refractory disease (defined as no marker response to initial treatment).⁸¹ In addition to individual factors, several prognostic models have been developed to help stratify patients in consideration of HDCT. In 1996, Beyer and colleagues retrospectively performed a multivariate analysis of 310 patients treated with at least one cycle of salvage HDCT at 4 centers in the US and Europe.⁶⁰ Progressive disease before HDCT, PM-NSGCT, refractory or absolute refractory disease to conventional-dose cisplatin, and HCG levels greater than 1,000 U/L were identified as independent adverse prognostic indicators of survival after HDCT. A scoring system based on these risk factors, referred to as the Beyer score,⁶⁰ was established to categorize patients as good (0 points), intermediate (up to 2 points), or poor prognosis (>2 points) with reliable discrimination in regard to RR, proportion of patients that relapsed, failure-free survival (FFS), and OS (all with P<.001). Of note, more than 90% of the patients treated in this European series received a single HDCT course, and most were treated with 2 or more regimens before HDCT.

These findings could not be reproduced in a later report from Indiana University evaluating patients with less extensive prior therapy.⁸² Subsequently, Indiana investigators developed a separate prognostic model for HDCT based on 184 testicular GCT patients treated between 1996 and 2004.⁶³ Patients with late relapse (>2 years) and PM-NSGCT were not included in this analysis because of poor results seen previously in this population.⁸² Multivariate analysis identified 3 significant predictors of adverse DFS: 1) IGCCCG poor-risk classification at initial diagnosis, 2) platinum-refractory disease defined as tumor progression within 4 weeks after the most recent cisplatin-based chemotherapy, and 3) receipt of HDCT as 3^rd-line or subsequent chemotherapy. Based on these factors, patients were classified into 3 prognostic groups based on their total score (referred to as the Einhorn score). DFS was approximately 80%, 60%, and 40% for patients with low-, intermediate-, and high-risk Einhorn scores, respectively.⁶³ The Beyer score did not reliably predict DFS in this patient population (P=.25).

The authors’ group evaluated prognostic factors in a series of 107 patients treated with 2 courses of rapidly recycled (every 14 days) conventional-dose paclitaxel and ifosfamide for stem cell mobilization, followed by 3 cycles of HD carboplatin and etoposide (TI-CE) plus stem cell support.⁶⁴ This HDCT trial targeted patients predicted for a poor prognosis to conventional salvage therapy by requiring at least one unfavorable prognostic feature for enrollment, including extragonadal primary site, progressive disease (PD) following an incomplete response (IR) to 1^st-line chemotherapy, and PD after cisplatin plus ifosfamide-based CDCT salvage. Nearly half of patients (47%) achieved 5-year DFS. Factors predicting unfavorable DFS or OS to TI-CE included mediastinal primary tumor site, HCG ≥1,000 U/mL, 2 or more lines of prior therapy, 3 or more metastatic sites, and IGCCCG intermediate- or poor-risk classification at diagnosis.⁶⁴ The study also tested both the Einhorn and Beyer prognostic models for their ability to predict DFS. While findings were partially reproduced, the 2 models could not be fully applied because of the differences in eligibility criteria.⁶⁴ The lack of complete reproducibility of the Einhorn and Beyer models in the TI-CE series demonstrates the limitations of these prediction rules, each developed in specific patient populations with varying clinical features, prior management, and HDCT regimens.

The International Prognostic Factor Study Group recently presented their prognostic model for initial salvage therapy independent of regimen intensity (Table 3).⁷⁶ This series comprised a total of 1,984 patients from 38 centers throughout 14 countries in Europe and North America. Seven factors were significant for PFS on multivariate analysis including histology (seminoma vs nonseminoma), primary tumor site (mediastinal vs retroperitoneal vs gonadal), response to 1^st-line chemotherapy (CR vs PR vs other), progression-free interval following 1^st-line chemotherapy, AFP level at salvage, HCG level at salvage, and the presence of nonpulmonary visceral metastases. Each factor was assigned a point value, and a sum score was calculated for each patient. Scores were divided into 5 risk groups (very low, low, intermediate, high, and very high) with distinct PFS and OS rates regardless of treatment intensity.⁷⁶ The large, international and multicenter population in this study, the strict definition of inclusion criteria and salvage regimen, as well as this model’s ability to predict outcomes to both HDCT and CDCT initial salvage approaches, allows wider applicability than prior prognostic systems. This scoring system is regarded as the new standard predictive model in the relapsed/refractory setting.

Table 3.

International prognostic model for initial salvage therapy in the relapsed/refractory setting⁷⁶

Factors		Points
Primary Site	Gonadal:	0
	Retroperitoneal:	1
	Mediastinal	3
Response to 1st	CR/PR−:	0
	PR+/SD:	1
line	PD:	2
Progression free	>3 mo:	0
	≤3 mo:	1
Serum hCG level	≤1000 IU/L:	0
	> 1000 IU/L	1
Serum AFP level	Normal:	0
	≤1000 ng/ml:	1
	>1000 ng/ml:	2
Liver, bone or	Absent:	0
	Present:	1
Add points to determine preliminary score (0–10). Then determine
category: 0 : 0;	1–2 : 1;	3–4 : 2;	≥5 : 3
Add histology points (below) to category score to determine final risk
Histology	Seminoma: −1 NSGCT / mixed: 0
Risk Category	# of points	2-year PFS	3-year
Very low risk	−1	75%	77%
Low risk	0	51%	66%
Intermediate risk	1	40%	58%
High risk	2	26%	27%
Very high risk	3	6%	6%

Abbreviations: NSGCT: nonseminomatous germ cell tumor; CR: complete response; PR-: partial response with negative markers; PR+: partial response with positive markers; SD: stable disease; PD: progression of disease; hCG: human chorionic gonadotropin; AFP: alpha fetoprotein; PFS: progression free survival; FFS: failure free survival; OS: overall survival; DFS: disease free survival

Seminoma

Pure seminoma histology has been associated with a high cure rate because of chemosensitivity. Approximately 90% of patients requiring 1^st-line chemotherapy for disseminated disease are categorized as IGCCCG good risk,⁸³ and cure rates can exceed 80%–85%.^{84, 85} Few of these patients have an IR to initial cisplatin-based chemotherapy or relapse from a CR that requires salvage chemotherapy, and late relapses are rare. Chemotherapy regimens for seminoma have typically mirrored those used in disseminated NSGCT, including in the salvage setting. In fact, most studies of salvage regimens for recurrent disseminated GCT have included some patients with pure seminoma, but most patients in these trials had advanced NSGCT. Only a few retrospective series have specifically looked at seminoma in the 2^nd-line setting, and these have generally demonstrated favorable results with higher cure rates than what has historically been seen with NSGCT. One report included 24 patients with recurrent seminoma treated with VeIP as initial salvage, and demonstrated an 88% CR rate with 54% long-term DFS. Outcome was equally good for patients with primary testicular and extragonadal primary tumors.⁸⁶ Similarly, a different retrospective series included 27 patients with seminoma who relapsed after prior platinum-based chemotherapy. Overall, this group of patients was more heavily pretreated (15% had received 2 prior regimens). Of the 27 patients, 15 were treated with salvage cisplatin/ifosfamide-based CDCT, while 12 received HDCT. Overall CR rate in this series was 56%, with 48% of patients achieving a sustained CR with a median follow-up of 72 months. Eight patients underwent post-chemotherapy surgery following salvage treatment. None of the 2 patients with viable seminoma was alive at the time of report, while all 6 patients with necrosis remained without evidence of disease.

The superior outcome with seminoma was recognized in a report of the International Prognostic Factor Study Group.⁷⁶ The prognostic model, constructed and validated in NSGCT, was applied to a subset of patients with pure seminoma and was able to discern prognostic groups, but demonstrated significantly better survival for each group than what had been seen for patients with NSGCT. Consequently, histology was added to the score with seminoma coded as −1 and NSGCT as 0. A ”very low risk” category was created for patients with seminoma and no adverse factors (sum score −1), with an estimated 3-year PFS and OS of 75.1% and 77%, respectively.⁷⁶

Late Relapse

Late relapse, defined as disease recurrence 2 years or more after successful initial treatment, is seen in 2%–4% of patients achieving a CR with initial chemotherapy for advanced GCT,⁸⁷ i.e., comprising up to 10% of all relapses in that setting. Late relapse is also seen after management of early stage disease by means of surgery, chemotherapy, or radiotherapy, but overall remains uncommon, occurring in only 1% to 4% of GCT patients who achieve a CR after initial diagnosis.^87–89 It has long been recognized as a poor prognostic indicator for further therapy. Late relapses are more common in patients with NSGCT than seminoma,⁸⁸ in patients with bulky retroperitoneal adenopathy,^90–92 and in those with teratoma found at the time of post-chemotherapy retroperitoneal lymph node dissection (RPLND).^92–94 Outcome seems particularly poor for those patients who previously received cisplatin-containing systemic chemotherapy (vs those who recur many years after surgery, radiation, or carboplatin alone), with one report estimating the hazard ratio of death from cancer for this group at 4.0 (95% CI, 1.2%–13.6%, P=.03) on multivariate analysis.⁹⁵

Because of an increased incidence of chemotherapy resistance, surgical resection is considered the mainstay of management.^{89, 94, 95} Chemotherapy, however, is attempted in patients with disseminated or unresectable disease. Our group has reported durable CR for these patients with the use of paclitaxel-based regimens followed by surgery.^{39, 87} The previously cited phase II trial of 2^nd-line TIP included 14 patients with late relapse, all of whom were deemed ineligible for surgery. Seven of these (50%) achieved a CR, and all 7 remained in remission at the time of report (median follow-up 51 months).³⁹

Cisplatin plus epirubicin has been reported to achieve CR in patients with late- relapse GCT. A phase II study of cisplatin plus epirubicin treated 30 patients with a median of 2 prior regimens, most (70%) of whom suffered a late relapse.⁹⁶ The sustained CR rate for patients with late relapse was 29% (follow-up range 28+ – 48+ months); one-third of these patients with favorable outcome were treated in the 2^nd-line setting, the remaining 5 received more extensive prior therapy.

The benefit of HDCT in patients with late relapse is controversial. Two reports on a small number of patients suggested that a subset of patients can achieve complete response with this approach. Two of 7 (29%) patients with late-relapse GCT treated with HDT using the TI-CE regimen achieved long-term DFS.⁶⁴ German investigators recently published a secondary analysis of 35 late-relapse patients included in a randomized phase III trial of single-cycle versus sequential HDCT.⁹⁷ Five of 35 patients (14%) were disease free at the time of report, 4 of the 5 having received HDCT in the 2^nd-line setting.

Teratoma, frequently found in late relapse, is not chemosensitive. The rate of malignant transformation increased in this group.^{98, 99} Malignant transformation should be suspected if the disease does not respond to cisplatin-based chemotherapy or growth is seen on imaging without rise of disease tumor markers despite cisplatin-based chemotherapy. Isochromosome of chromosome 12, i(12p), is a specific marker of GCT.¹⁰⁰ Identification of i(12p) or excess 12p copy number can be used to establish the clonal origin of malignant transformation. If such disease is localized, outcome is typically more favorable with complete resection.¹⁰¹ Otherwise, treatment follows the histology paradigm of the secondary somatic malignancy, including chemotherapy, radiation, and/or surgery, as appropriate.^{99, 102} There are no data supporting the use of HDCT in patients with relapsed GCT comprised exclusively of malignant transformation. Following the above treatment paradigm, subjecting patients to the toxicity of such treatment is not warranted, with the exception of the rare case where the transformed malignancy is sensitive to HDCT.

Surgery in the Salvage Setting

Following cisplatin-based 1^st-line therapy for NSGCT, standard management calls for surgical resection of residual disease when serum tumor marker levels have normalized. Necrotic tissue, fibrosis, teratoma, or viable GCT may be found at the time of surgery. When viable GCT is present, 2 cycles of adjuvant chemotherapy are recommended for the 1^st-line setting.¹⁰³ In contrast, the pathologic presence of necrotic tissue, fibrosis, or mature teratoma requires no further therapy since only 5% to 10% of these patients will relapse.

In the salvage setting, pathologic specimens of resected masses after 2^nd-line VeIP or VIP chemotherapy demonstrate viable tumor in approximately 50%, teratoma in 40%, and necrosis in only 10%.^{103, 104} The rate of viable tumor at the time of surgery is significantly higher following salvage chemotherapy than after 1^st-line treatment. Presence of viable tumor in the salvage setting confers a worse prognosis compared with mature teratoma or necrotic tissue,¹⁰⁵ and the risk of relapse despite resection of residual masses is approximately 50% in such cases. In contrast to the 1^st-line setting, there is no established benefit to additional standard-dose chemotherapy in patients who have viable NSGCT in the resected specimen after salvage chemotherapy.

Surgical resection of isolated metastasis in the presence of elevated serum tumor markers is performed in highly select instances. Often referred to as “desperation surgery” or “surgical salvage,” this approach has curative potential in patients whose disease is refractory to cisplatin. Patients with a solitary retroperitoneal mass and increased AFP alone are the best candidates for this surgery. Resection of all disease seen on imaging studies and normalization of serum tumor markers predict long-term survival.⁹ As mentioned above, patients with PM-NSGCT respond poorly to conventional salvage chemotherapy. In rare cases, surgical resection of residual mediastinal masses in the setting of elevated serum tumor markers can lead to long-term survival.¹⁰⁶

Summary

Established 2^nd-line CDCT regimens are cisplatin-ifosfamide based and include either VeIP or TIP, with sustained CR rates reported in prospective clinical trials ranging from 24%–25% for VeIP ^{30, 31} and 38%–63% for TIP^39–41. Enrollment criteria varied for these studies, and the 2 regimens have never been directly compared. Other investigators routinely consider HDCT in the initial salvage setting, and depending on patient selection criteria, have reported sustained CR rates of 40%–50% for this approach.^{63, 64, 70} HDCT in 2^nd-line therapy remains controversial, and prospective data specific to the 2^nd-line setting are limited. Direct comparisons of both approaches are sparse. While retrospective analyses suggest that HDCT may be superior to CDCT,^{71, 75} this was not confirmed by the one randomized prospective trial comparing salvage CDCT versus HDCT,³¹ a study that is limited by flaws outlined in prior sections of this review. Some investigators, including our own group, use a risk-adapted approach to determine salvage strategy, reserving HDCT for those patients at highest risk of failing 2^nd-line CDCT. For a more definitive answer regarding the optimal approach to GCT salvage therapy, further prospective studies are needed. The recently developed prognostic model developed by the International Prognostic Factor Study Group for 2^nd-line chemotherapy applies to both CDCT and HDCT, and should be incorporated into any future investigative efforts.