Abstract
Outside of clinical trials, the occurrence of objective responses (OR) to chemotherapy in patients with recurrent gliomas is poorly characterized. Further, the predictive value of OR for PFS and OS is unclear. We screened the German Glioma Network Database for patients who had received chemotherapy only for recurrent glioma. As PFS was not available for a large number of patients we used the composite endpoint time-to-treatment-failure (progressive disease, start of a new therapy or death). We included 485 patients who received 646 chemotherapy regimens for treatment of recurrent glioma. Of these, only 32 chemotherapies in 32 patients resulted in an OR (30 PR, 2 CR) after central review according to RANO criteria. OR rates were 2.8%, 11.0% and 10.6% for glioblastoma, anaplastic glioma WHO grade III and diffuse glioma WHO grade II, respectively. Temozolomide (n=232) resulted in ORs in 7.8% of the patients, while nitrosourea (n=212) and imatinib (n=27) resulted in ORs in 6.1% and 3.7%, respectively. Overall, responders showed a significantly improved OS compared to non-responders (median OS 33.3 versus 15 months, p=0.054). Yet, a Cox regression analysis adjusted for diagnosis and age did not reveal a significant association of objective responses with overall survival (relative risk 0.8, p=0.391). In addition, we generated a 1:3 cohort (n=96) of non-responders matched for diagnosis. There was no relevant difference in OS comparing responders to the matched cohort (median OS 33.3 versus 25.6 months, p=0.929). When comparing non-responders with longer time-to-treatment-failure (>14 weeks, assumed stable disease as best response) with responders (n=32) the difference in outcome was lost (median OS 24.3 versus 33.3 months, p=0.86). In this prospective trial ORs to chemotherapy in the recurrent setting were rare. Notably, when comparing responders with a matched cohort or patients with assumed stable disease as best response, ORs were not associated with improved survival.