Purpose: Temozolomide (TMZ) is known to induce thrombocytopenia but no early predictive test has yet been clearly established. The aim of the study was to retrospectively identify and validate a threshold of early platelet variation predicting TMZ-induced thrombocytopenia during the TMZ phase in patients treated according to the Stupp protocol for glioblastoma.
Methods: A training set was used to analyze variations in platelet count occurring from the first week (W1) to week 6 (W6) during radiotherapy. Our aim was to identify the most relevant platelet decrease associated with TMZ-induced thrombocytopenia ≤ 100 G/L at day 28 during the TMZ phase. The performance of the threshold was confirmed in an independent validation set.
Results: Overall, 147 patients were included, 85 and 62 in the training and validation sets, respectively. Twenty-seven patients (18%) experienced at least one TMZ-induced thrombocytopenia in the TMZ phase. A platelet decrease at W6 ≥ 35% (∆W6 ≥ 35%) was identified as the best predictive variation with an AUC of 0.83, a sensitivity of 65%, and a specificity of 96%. In the validation set, ∆W6 ≥ 35% platelet variation was identified as an independent marker of TMZ-induced thrombocytopenia during the TMZ phase (OR 15.23 (95% CI 3.5-107.5)) corresponding to sensitivity of 77% (66-87%), specificity of 73% (62-84%), a positive predictive value of 42% (29-54%), and a negative predictive value of 92% (86-99%).
Conclusion: Platelet decrease at W6 ≥ 35% during the RT-TMZ phase is an early and simple predictive marker of clinically relevant TMZ-induced thrombocytopenia during TMZ maintenance.
Keywords: Early platelet variation; Glioblastoma; Temozolomide; Thrombocytopenia.