The realisation that patients relapsing after cisplatin-based initial therapy could be salvaged using further chemotherapy was first made over 20 years ago. It was realised that removal of bleomycin and adding instead ifosfamide 6 g/m², was capable of curing around 25–30 % of patients. It was initially suggested that those originally treated with BEP should receive cisplatin 100 mg/m², ifosfamide 6 g/m² and vinblastine 0.11 mg/kg on day 1 and 2 of each 3 week cycle (VeIP) whilst those who received PVB first should then receive the same regimen but with etoposide 375 mg/m² replacing the vinblastine (VIP or PEI) with etoposide often given at the relapse regardless of initial therapy as it is thought to be less toxic, particularly in terms of neurotoxicity [43, 44]. More recently the identification of paclitaxel as an active drug in heavily pre-treated patients has led to the adoption of this regimen; TIP with a dose of paclitaxel replacing either etoposide or vinblastine at a dose of 250 mg/m². Whilst the outcome with TIP appears numerically superior the results in fact are likely to be due to patient selection [45]. It is usual to give four cycles of salvage therapy.

Other treatment approaches have used dose dense cisplatin i.e. cisplatin delivered weekly or every 2 weeks. This has been combined with other drugs that often are not particularly myelosuppressive e.g vincristine, bleomycin and methotrexate. Durable responses are seen in around 30 % of cases. Both actinomycin-D and methotrexate have shown activity when combined with etoposide in cisplatin refractory disease [46]. These drugs have been combined with cisplatin in the dose dense regimen GAMEC [24]. Approximately 40 % of patients have been salvaged using this approach; however myelosuppression and mucositis were problematic particularly in older patients (over 35) and those with poor renal function. As this regimen contained methotrexate in high dose it was used in patients with cerebral metastases and was able to lead to complete remissions. Anthracyclines e.g epirubicin have shown significant activity when combined with cisplatin when given to chemo-sensitive relapse [47].

Many agents including gemcitabine (11 % response rate) [19], oxaliplatin (13 % response rate) [48], paclitaxel (11 % response rate) [17] have shown activity in relapsed germ cell tumors. Combinations of these agents have shown higher response rates (30–40 %); for example, oxaliplatin and gemcitabine, oxaliplatin and paclitaxel, gemcitabine and paclitaxel. Paclitaxel may be particularly important when treating relapsed mediastinal germ cell tumors as there do not seem to be survivors if this drug is omitted. The drug irinotecan, a semisynthetic analogue of the natural alkaloid camptothecin, is unusual in that it showed no activity when used as a single agent in a heavily pre-treated cohort [49], however in combination with cisplatin or oxaliplatin much higher response rates were seen (30–40 %) [23]. The combination of oxaliplatin, irinotecan and paclitaxel has been associated with a response rate of 71 % [15]. One of the reasons why response rates seem so variable in these settings is that the patient population varies with some studies having a high proportion of very refractory patients. The advantage of all these treatments is that they can be used in patients with poor renal function.

The inherent chemo-sensitivity of germ cell tumors and the ability to salvage relapse has led to the development of high dose chemotherapy (HDCT) in the hope that this may circumvent resistance. In order to attenuate the effect of marrow suppression stem cells are collected to re-infuse after high dose therapy has been given. Normally this is done by giving chemotherapy, often cisplatin and ifosfamide-based followed by filgrastim. When the peripheral CD34 count (a marker for hemopoetic stem cells) reaches a threshold (>10/μl), patients have these cells collected using a cell separator. Sometimes high dose filgrastim is given alone particularly if marrow reserve is expected to be good. Many factors have been looked at to stratify the response to high dose therapy, in particular the site of the primary, the response to the last cisplatin based therapy and the tumour marker level at relapse. A prognostic score has been established for a single high dose therapy based on 5 factors. These include: the presence of progression prior to high dose treatment, a mediastinal primary tumor, refractory disease to cisplatin (progression within 4 weeks of treatment having been at least stable during therapy), absolute refractory disease to last cisplatin-based therapy (progression on therapy), and an hCG of over 1,000 IU/l. These factors were weighted to provide an overall score (Table 29.3) [50].

Patients proceed to receive HDCT once an adequate number of stem cells have been collected. In some protocols this follows some conventional chemotherapy to assess chemo-sensitivity and collect stem cells while in others patients proceed directly to HDCT. The drugs used in the preparative regimen have varied. The use of carboplatin is standard. The dose varies from 1,200 to 2,200 mg/m². Increasingly carboplatin is dosed using a formula based on renal clearance e.g the Calvert Formula – Dose in mg = AUC × (A × GFR + B) [AUC: a figure that reflects the concentration of drug in the body that is to be achieved; GFR: glomerular filtration rate; A is the ratio of GFR to how quickly the kidneys get rid of carboplatin; and B is the rate at which the body gets rid of carboplatin in other ways than through the kidneys]. On this basis a dose of between AUC 15- AUC 30 is selected. The reason for this is that the correlation between effect and dose is much more closely linked with renal clearance than a dose based on body surface area. Carboplatin is most often combined with etoposide in a dose of 1,500–2,200 mg/m^{2 52}. Other drugs have been used most commonly cyclophosphamide (6 g/m²) or thiotepa [13, 14].

Thiotepa has an advantage in that it has very good brain penetration. Other drugs used have included ifosfamide, melphalan and topotecan [15, 51]. The most popular regimen is carboplatin and etoposide and this has predictable toxicity with mucositis being the most troublesome side effect. The number of cycles of HDCT has varied. Increasingly several cycles are preferred. Two cycles of high dose carboplatin and etoposide is commonly employed [52] and in some protocols three cycles have been used although the doses chosen have tended to be rather lower [53]. At present there is no reason to prefer three over two cycles. The time between two cycles of high dose therapy is generally 4–6 weeks.

There has been one randomized study of high dose chemotherapy in relapsed disease and this showed no advantage over standard second line therapy; the study, however, was small. One study comparing three cycles of VIP and one high dose carboplatin, etoposide and cyclophosphamide to one cycle of VIP and 3 cycles of high dose carboplatin and etoposide showed that the multicycled high dose approach produced better overall survival (49 % vs 39 % p = 0.06) but this was mostly due to reduced treatment related deaths in that arm. The single cycle of high dose therapy using cyclophosphamide, etoposide and carboplatin had a much higher treatment related death rate (16 % vs 4 %, p < 0.01) and this is the main cause in the difference in overall survival [54]. A recent overview from a database looking at the management of first relapse showed that the prognosis seemed much improved in those who received high dose chemotherapy. In addition patients who received sequential high dose chemotherapy had an improved overall survival (60.6 vs 46.3 %, p = 0.001) [55].

Currently on the basis of the non-randomised data presented the largest series to date has used 1 cycle of induction cisplatin and ifosfamide –based therapy followed by two cycles of high dose carboplatin and etoposide, followed in many patients by 3 months of oral etoposide although the value of this 3 month oral therapy remains unclear. Relapsed metastatic seminoma seemed to do particularly well using this approach with up to 90 % being cured [52].

Overall this carries a very poor outcome with perhaps only 5 % being cured. It is often picked up as a rise in tumor markers following the end of treatment. Patients who have completed high dose therapy should have as many residual masses removed as possible. If relapse occurs and the sites of radiological relapse are amenable to surgery this should be considered. The regimen of gemcitabine and paclitaxel used in this setting seems to give a higher relapse free survival of 14 % [21]. It should be remembered that these patients did not receive either drug prior to high dose chemotherapy. The combination of cisplatin and epirubicin has also been used in the setting of relapsing disease. It is clearly active and the regimen seems well tolerated it is clearly an option in patients who do not have disease that is absolutely refractory to cisplatin [58]. Oral etoposide has been used in this setting, easy to administer but with unpredictable toxicity. The management of these patients is inevitably more subjective- patients will tend to be given dugs that they either did not receive to date or have shown significant activity at relapse. It is likely that these patients should receive either surgery or perhaps radiotherapy to these known sites following chemotherapy if this is practical. Chronic oral etoposide has been recommended by some as it is easy to administer but has a limited activity in this setting; in addition, as its absorption is variable toxicity can be unpredictable, nevertheless durable remissions have been seen when therapy is consolidated by surgery or irradiation [59]. In this setting clearly some degree of chemo-sensitivity is maintained but in the absence of either surgery or radiation long term remission seems unlikely. For patients where this is not possible treatment is palliative in intent and therefore it is best to avoid toxic regimens.