Odds of death after glioblastoma diagnosis in the United States by chemotherapeutic era

Bevacizumab (BZM) and temozolomide (TMZ) have been shown to be beneficial in the treatment of patients with glioblastoma. We sought evidence for the benefit of BZM in the general patient population at large. The Surveillance, Epidemiology, and End Results SEER database was queried for patients diagnosed with glioblastoma between 2000 and 2009, divided into a pre-TMZ era (January 2000-June 2003), a transitional era (July 2003-March 2005), a TMZ era (April 2005-October 2007), and a BZM-TMZ era (November 2007-December 2009). Binomial logit regression analyzed odds of death, taking into account age at diagnosis, tumor size, gender, race, marital status, radiotherapy, and extensive surgery. Compared with the pre-TMZ era, odds of death were decreased in the TMZ era by 12% (97.5% CI [confidence interval] 3-20%) 6 months after diagnosis and 36% (30-42%) a year after diagnosis; corresponding values for BZM-TMZ were 31% (24-37%) and 50% (45-55%). For era comparisons, decreases in odds of death were larger at 12 than 6 months; the opposite was true for extensive surgery and radiotherapy (P < 0.025, Wald chi(2) test, for each analysis). For both 6 and 12 month comparisons, odds of death in the BZM-TMZ era were lower than in the TMZ era (P < 0.025, Wald v 2 test, for each analysis). The results provide evidence that TMZ positively impacted survival of glioblastoma patients and that the addition of BZM further improved survival, this lends support to the addition of BZM to the chemotherapeutic armamentarium. Evaluation of odds of death is an attractive alternative to Cox regression when proportional hazards assumptions are violated and follow-up is good.