Use of the RTOG recursive partitioning analysis to validate the benefit of iodine-125 implants in the primary treatment of malignant gliomas

Abstract

Purpose: To date, numerous retrospective studies have suggested that the addition of brachytherapy to the conventional treatment of malignant gliomas (MG) (surgical resection followed by radiotherapy ± chemotherapy) leads to improvements in survival. Two randomized trials have suggested either a positive or no survival benefit with implants. Critics of retrospective reports have suggested that the improvement in patient survival is due to selection bias. A recursive analysis by the RTOG of MG trials has stratified MG patients into 6 prognostically significant classes. We used the RTOG criteria to analyze the implant data at Wayne State University to determine the impact of selection bias. Methods and Materials: Between July 1991 and January 1998, 75 patients were treated with a combination of surgery, radiotherapy, and stereotactic I-125 implant as primary MG management. Forty-one (54.7%) were male; 34 (45.3%) female. Median age was 52 years (range 4–79). Twenty-two (29.3%) had anaplastic astrocytoma (AA); 53 (70.7%), glioblastoma multiforme (GBM). Seventy-two patients had data making them eligible for stratification into the 6 RTOG prognostic classes (I–VI). Median Karnofsky performance status (KPS) was 90 (range 50–100). There were 14, 0, 14, 31, 12, and 1 patients in Classes I to VI, respectively. Median follow-up time for AA, GBM, and any surviving patient was 29, 12.5, and 35 months, respectively. Results: At analysis, 29 (40.3%) patients were alive; 43 (59.7%), dead. For AA and GBM patients, 2-year and median survivals were: 58% and 40%; 38 and 17 months, respectively. For analysis purposes, Classes I and II, V and VI were merged. By class, the 2-year survival for implanted patients compared to the RTOG data base was: I/II—68% vs. I—76%; III—74% vs. 35%; IV—34% vs. 15%; V/VI—29% vs. V—6%. For implant patients, median survival by class was (in months): I/II—37; III—31; IV—16; V/VI—11. Conclusion: When applied to MG patients receiving permanent I-125 implant, the criteria of the RTOG recursive partitioning analysis are a valid tool to define prognostically distinct survival groups. As reflected in the RTOG study, a downward survival trend for the implant patients is seen from “best to worse” class patients. Compared to the RTOG database, median survival achieved by the addition of implant is improved most demonstrably for the poorer prognostic classes. This would suggest that selection bias alone does not account for the survival benefit seen with I-125 implant and would contradict the notion that the patients most eligible for implant are those gaining the most benefit from the treatment. In light of the contradictory results from two randomized studies and given the present results, further randomized studies with effective stratification are required since the evidence for a survival benefit with brachytherapy (as seen in retrospective studies) is substantial.