CN-21 * RISK MODELING FOR TEMOZOLOMIDE (TMZ)-MYELOTOXICITY IN PATIENTS WITH GLIOBLASTOMA TREATED ON RTOG 0825

Abstract

BACKGROUND: TMZ is well tolerated with a relatively low incidence of myelotoxicity, which can nonetheless result in treatment delays or death. Our goal was to evaluate risk and validate prediction models in newly diagnosed patients treated with TMZ +/- bevacizumab (BEV). METHODS: Models (from recurrent disease) in newly diagnosed patients (who participated in RTOG 0825) were evaluated. Then independent gender specific models were developed using logistic regression with backward selection (p ≤0.15), to select candidate risk factors. Treatment group (TMZ/placebo or TMZ/BEV) and age were included into the final models as potential confounders. A summary risk score was computed by counting the number of final risk factors that were present. RESULTS: Clinical data from 591 cases was included. Incidence of myelotoxicity was greater in women compared with men (35% versus 15%). Previously published recurrent models did not predict risk in the newly diagnosed patients. For men, the final model included: TMZ/BEV group (OR:1.96;p = 0.04); non-smoker (OR:2.83, p < 0.01); on corticosteroids (OR:1.96;p = 0.09); hemoglobin <13g/dl (OR:2.11, p = 0.04). Patients with risk scores 0-2 were associated with an 8% myelotoxicity risk; scores of 3-5 with 21% risk. For woman, the final model included: KPS≤ 80 (OR:1.78, p = 0.06); incomplete tumor resection (OR:2.57; p=0.01), BSA < 1.78 (OR:1.99, p = 0.07); age ≥58 yrs (OR:1.76, p = 0.07); BUN < 13 (OR:2.63, p = 0.01); Platelet <313k(OR:2.29;p = 0.02); on non-enzyme inducing anticonvulsants (OR:1.81, p = 0.06) or enzyme inducing anticonvulsants (OR 1.96, p = 0.16), or no anticoagulants (OR:5.87, p = 0.03). Risk score of 1-4 was associated with low risk (13%) while 5-9 was high risk (46%). CONCLUSIONS: TMZ myelotoxicity is a significant clinical issue, and predicting risk should allow treatment optimization. This clinical model suggests that clinical factors can be used to select a group at higher risk. Future work will attempt to improve prediction by inclusion of genetic polymorphisms.