MEASURING BIAS IN UNCONTROLLED BRAIN-TUMOR TRIALS - TO RANDOMIZE OR NOT TO RANDOMIZE

Purpose: To help investigators decide if new therapies for glioma warr ant definitive evaluation in randomized studies we have been developin g a method for assessing the degree to which patient selection may hav e enhanced the results of uncontrolled treatment trials. In this study , we analyzed the impact of case selection on the survival of patients with malignant glioma receiving adjuvant stereotactic radiosurgery, a promising therapy reserved for those with small tumors and good perfo rmance status. Methods: Following published eligibility criteria we si mulated the patient selection process for stereotactic radiosurgery gi ven as a boost at the conclusion of conventional radiotherapy. Eligibl e patients were culled from a pre-existing clinical/imaging database o f 101 consecutive conventionally-treated patients with biopsy-proven m alignant glioma and known survival times. Median durations of survival and 2- and 3-year survival rates were determined for those judged eli gible or ineligible for stereotactic radiosurgery, Results: Twenty-sev en percent of patients were deemed eligible for stereotactic radiosurg ery, eligible patients had more favorable prognostic factors and signi ficantly longer median survival than ineligible patients (23.4 vs. 8.6 months; 2-year rate, 48% vs. 15%; 3-year rate, 30% vs, 7%); eligible patients also had a longer median survival than the entire group of un selected patients (23.4 vs. 11.4 months). Radiosurgery-eligible, conve ntionally-treated patients with glioblastoma multiforme and a group of radiosurgery-treated patients at a special referral center had simila r median survival times (16.4 vs. 19.7 months), Conclusion: We provide additional evidence for selection bias in uncontrolled trials of ster eotactic radiosurgery and by simulating the selection process accurate ly have detected a larger bias effect than noted previously. Judging f rom experience with interstitial radiation and intraarterial chemother apy where substantial selection bias also occurred and randomized cont rolled trials proved disappointing, we conclude that a phase HI study of stereotactic radiosurgery for malignant glioma is unlikely to yield a positive result and may not be necessary.