Glioblastoma multiform is one of the most devastating primary tumors in neurooncology. We analyzed prognosis factors in patients with grade IV glioblastoma treated between 1993 and 1997. The 22 long-term survival patients (survival over 26 months) were extracted from our 30 years archives and the 2 populations are compared. The incidence was 2.6/100,000h/year, 62% male and 38% female, mean age 59 years, mean survival 12 months, median survival time 9 months. Multivariate analysis showed that younger age, surgical treatment and radiotherapy were all dependent prognosis factors for better survival. Statistically, survival was best for total surgical removal of tumors, followed by tumor gross resection then biopsy. Clinical status and inextirpable tumor location were also prognosis factors. The free interval time between total surgery and tumor reappearance was strongly correlated with survival (r=0.94). This suggests that some grade IV gliomas follow a quicker course, others exhibiting slow growth. Each of the prognosis factors was confirmed in the long-survival patients. Prevalence of all glioblastomes was 4.3%. Their mean age was 42 and mean survival 62 months. A larger proportion of these patients had total surgery and radiotherapy. The time lapse before tumor reappearance was longer. Deep tumor locations were less frequent. The proportion of secondary versus primary glioblastomas was the greatest difference between the long-term and regular survivors. Secondary glioblastomas were found in only 4% of the standard population and in 23 to 41% in the long-term survivors (p<0.01). Primary glioblastomas typically show EGFR over expression and mutation (variant III). The pathway to secondary glioblastoma involves early P53 mutation. Despite the fact that the anatomopathologist regards similar tissues under the microscope, these subtypes of glioblastomas are distinct disease entities which evolve through different genetic pathways and exhibit different outcomes.