Management of initial and recurrent radiation-induced contrast enhancements following radiotherapy for brain metastases: Clinical and radiological impact of bevacizumab and corticosteroids

PURPOSE: The appearance of radiation-induced contrast enhancements (RICE) after radiotherapy for brain metastases can go along with severe neurological impairments. The aim of our analysis was to evaluate radiological changes, the course and recurrence of RICE and identify associated prognostic factors. METHODS: We retrospectively identified patients diagnosed with brain metastases, who were treated with radiotherapy and subsequently developed RICE. Patient demographic and clinical data, radiation-, cancer-, and RICE-treatment, radiological results, and oncological outcomes were reviewed in detail. RESULTS: A total of 95 patients with a median follow-up of 28.8 months were identified. RICE appeared after a median time of 8.0 months after first radiotherapy and 6.4 months after re-irradiation. Bevacizumab in combination with corticosteroids achieved an improvement of clinical symptoms and imaging features in 65.9% and 75.6% of cases, respectively, both significantly superior compared to treatment with corticosteroids only, and further significantly prolonged RICE-progression-free survival to a median of 5.6 months. Recurrence of RICE after initially improved or stable imaging occurred in 63.1% of cases, significantly more often in patients after re-irradiation and was associated with high mortality of 36.6% after the diagnosis of flare-up. Response of recurrence significantly depended on the applied treatment and multiple courses of bevacizumab achieved good response. CONCLUSION: Our results suggest that bevacizumab in combination with corticosteroids is superior in achieving short-term imaging and symptom improvement of RICE and prolongs the progression-free time compared to corticosteroids alone. Long-term RICE flare-up rates after bevacizumab discontinuation are high, but repeated treatments achieved effective symptomatic control.