Early studies with high-dose chemotherapy for treatment of relapsed cerebral PNET had shown modest efficacy but considerable toxicity. The HIT97 national trial tested a nonrandomized but stratified relapse protocol using either intensive chemotherapy, potentially high dose, or oral chemotherapy. 72 patients (59 disseminated) whose primary treatment had been surgery (97 %), radiotherapy (88 %), and/or chemotherapy (95 %) were enrolled in the intensive chemotherapy arm at diagnosis of relapse or resistance. As a window for this study they received two courses of a 96-hour infusion with carboplatin and etoposide. A response (complete or partial remission) was documented by MRI. Responders received two more cycles of this therapy and stem cell collection, before they received HDC (carboplatin, etoposide, thiotepa) and stem cell support. All possibilities of local therapy were to be explored and applied. After two courses of chemotherapy there was a 52 % response rate (41/72 patients). The median PFS and OS for all 72 patients were 11.6 and 21.1 months. Patients with medulloblastoma had a longer PFS and OS (12.6 and 22.6 months) than those with other PNETs (3.1 and 12.3 months). Favourable prognostic features were no new signs of clinical impairment and localised disease at relapse diagnosis. For the 27 patients who received HDC the median PFS and OS were 8.4 and 20.2 months, respectively. HDC did not benefit patients with resistant cerebral PNET and was associated with profound haematological and mucosal toxicity (90-100 % grade III, IV), infections (50 % grade III and IV) and severe ototoxicity (50 % grade III, 12.5 % grade IV). Treatment related mortality was 8 %. There was low long-term survival and only 2/72 patients are in continuous remission. Adding HDC in patients who responded to the initial courses of chemotherapy did not improve survival. Patients with relapsed cerebral PNET who respond to conventional chemotherapy do not profit from further augmentation to HDC.