To assess the ability of [(18)F]fluorodeoxyglucose positron emission tomography for the early prediction of response in patients with relapsed metastatic germ cell tumours undergoing salvage high-dose chemotherapy. The role of positron emission tomography was compared with established means of tumour response assessment such as CT scans/MRI and serum tumour marker changes. In addition, positron emission tomography was compared with a current prognostic score which differentiates three prognostic groups with failure-free survival rates ranging from 5-50%. [(18)F]fluorodeoxyglucose uptake of metastases from germ cell tumours as well as CT scans and serum tumour marker were acquired after 2-3 cycles of induction chemotherapy but before the start of high-dose chemotherapy and CT scans/serum tumour marker were compared with the baseline examinations in 23 patients with relapsed germ cell tumours. To evaluate the validity of early response prediction by positron emission tomography, radiological monitoring and serum tumour marker decline, histopathologic response after resection of residual masses and/or the clinical course over 6 months after the end of treatment (relapse vs freedom of progression) were used. Overall, 10 patients (43%) achieved a marker-negative partial remission, three (13%) a marker-positive partial remission, five (22%) a disease stabilization and five (22%) progressed during treatment. Nine patients (39%) remained progression-free over 6 months following treatment, whereas 14 (61%) progressed. The outcome of high-dose chemotherapy was correctly predicted by positron emission tomography/CT scan/serum tumour marker in 91/59/48%. Eight patients with a favourably predicted outcome by CT scans plus serum tumour marker but a positive positron emission tomography prior to high-dose chemotherapy, failed treatment. This results in the following sensitivities/specificities for the prediction of failure of high-dose chemotherapy: positron emission tomography 100/78%; radiological monitoring 43/78%; serum tumour marker 15/100%. The positive and negative predictive values of positron emission tomography were 88 and 100%, respectively. As compared with the prognostic score, positron emission tomography was correctly positive in all patients of the three risk groups who failed treatment. In addition, a negative positron emission tomography correctly predicted a favourable outcome in the good and intermediate group. [(18)F]fluorodeoxyglucose positron emission tomography imaging can be used to assess response to chemotherapy in patients with relapsed germ cell tumours early in the course of treatment and may help to identify patients most likely to achieve a favourable response to subsequent high-dose chemotherapy. In patients with response to induction chemotherapy according to CT scans or serum tumour marker evaluation, positron emission tomography seems to add information to detect patients with an overall unfavourable outcome. It may also be a valuable addition to the prognostic model particularly in the good and intermediate group for further selection of patients who will profit from high-dose chemotherapy.