Extent of resection and survival in glioblastoma multiforme: identification of and adjustment for bias.

OBJECTIVE: The influence of the degree of resection on survival in patients with glioblastoma multiforme is still under discussion. The highly controlled 5-aminolevulinic acid study provided a unique platform for addressing this question as a result of the high frequency of "complete" resections, as revealed by postoperative magnetic resonance imaging scans achieved by fluorescence-guided resection and homogeneous patient characteristics.

METHODS: Two hundred forty-three patients with glioblastoma multiforme per protocol from the 5-aminolevulinic acid study were analyzed. Patients with complete and incomplete resections as revealed by early magnetic resonance imaging scans were compared. Prognostic factors that might cause bias regarding resection and influence survival (e.g., tumor size, edema, midline shift, location, age, Karnofsky Performance Scale score, National Institutes of Health Stroke Scale score) were used for analysis of overall survival. Time to reintervention (chemotherapy, reoperation) was analyzed further to exclude bias regarding second-line therapies.

RESULTS: Treatment bias was identified in patients with complete (n = 122) compared with incomplete resection (n = 121), i.e., younger age and less frequent eloquent tumor location. Other factors, foremost preoperative tumor size, were identical. Patients without residual tumor survived longer (16.7 versus 11.8 mo, P < 0.0001). In multivariate analysis, only residual tumor, age, and Karnofsky Performance Scale score were significantly prognostic. To account for distribution bias, patients were stratified for age (>60 or <or=60 yr) and eloquent location. Survival advantages from complete resection remained significant within subgroups, and age/eloquent location were no longer unevenly distributed. Reinterventions occurred marginally earlier in patients with residual tumor (6.7 versus 9.5 mo, P = 0.0582).

CONCLUSION: Treatment bias was demonstrated regarding resection and second-line therapies. However, bias and imbalances were controllable in the cohorts available from the 5-aminolevulinic acid study so that the present data now provide Level 2b evidence (Oxford Centre for Evidence-based Medicine) that survival depends on complete resection of enhancing tumor in glioblastoma multiforme.