Does stereotactic eligibility for the treatment of glioblastoma cause selection bias in randomized studies?

The purpose of this study was to evaluate the potential selection bias using stereotactic eligibility as a criteria for participation in studies of glioblastoma multiforme. Radiation Therapy Oncology Group (RTOG) 90-06 comparing 60 Gy in 30 fractions with BCNU and 72 Gy in 60 fractions with BCNU was analyzed based on eligibility criteria used to enter patients on RTOG 93-05 using a stereotactic boost for patients with glioblastoma. Five hundred nine patients with histopathologically confirmed glioblastoma multiforme were analyzed; of these, 137 met criteria for 93-05 and 372 did not. Recursive partitioning analysis (RPA) was used to evaluate for differences. The RPA distribution in stereotactic radiosurgery (SRS)-eligible and -ineligible patients was similar. The median survival for RPA class 3 SRS-eligible patients was 1.4 years and -ineligible patients 1.4 years. For RPA class 4, the median survival was 1.0 years for eligible patients and 0.9 years for ineligible patients (P = 0.0421). For class 5 patients, the median survival was 8.3 months versus 7.2 months (P = 0.09). RPA class 6 patients had a median survival of 1.7 months versus 2.7 months for ineligible patients (P = 0.199). By analyzing previously randomized patients in a study not using a stereotactic boost, there does not appear to be a survival benefit for those patients who fit the criteria for consideration of a stereotactic boost in patients with glioblastoma multiforme.