Ahead of Print: Reoperation for Recurrent Glioblastoma and Its Association With Survival Benefit

BACKGROUND: Glioblastoma is the most common and aggressive primary brain tumor. Despite current treatment, recurrence is inevitable. There are no clear guidelines for treatment of recurrent glioblastoma.

OBJECTIVE: To investigate factors at initial surgery predictive of reoperation, and the prognostic variables associated with survival, including reoperation for recurrence.

METHODS: A retrospective cohort study was performed, including adult patients diagnosed with glioblastoma between January 2010 and December 2013. Student t test and Fisher exact test compared continuous and categorical variables between reoperation and nonreoperation groups. Univariable and Cox regression multivariable analysis was performed.

RESULTS: In a cohort of 204 patients with de novo glioblastoma, 49 (24%) received reoperation at recurrence. The median overall survival in the reoperation group was 20.1 months compared with 9.0 months in the nonreoperation group (P = .001). Reoperation was associated with longer overall survival in our total population (hazard ratio, 0.646; 95% confidence interval, 0.543-0.922; P = .016) but subject to selection bias. Subgroup analyses excluding patients unlikely to be considered for reoperation suggested a much less significant effect of reoperation on survival, which warrants further study with larger cohorts. Factors at initial surgery predictive for reoperation were younger age, smaller tumor size, initial extent of resection >=50%, shorter inpatient stay, and maximal initial adjuvant therapy. When unfavorable patient characteristics are excluded, reoperation is not an independent predictor of survival.

CONCLUSION: Patients undergoing reoperation have favorable prognostic characteristics, which may be responsible for the survival difference observed. We recommend that a large clinical registry be developed to better aid consistent and homogenous data collection.

Full text version access for all Journal subscribers.