Levin 2011.

Methods	Randomised, double‐blind, placebo‐controlled trial
Participants	Adult patients treated with radiotherapy for brain or head and neck neoplasm completed at least 10 months prior to study entry, who have a radiological diagnosis of brain radionecrosis based on MRI criteria. Patients were allowed to be taking corticosteroids prior to study participation but they were required to be taking a stable dose for at least 1 week prior to receiving study treatment.
Interventions	Bevacizumab: 7.5 mg/kg every 3 weeks x 2 cycles Placebo: every 3 weeks x 2 cycles
Outcomes	Total of 14 patients enrolled: 7 patients in the bevacizumab arm versus 7 patients in the placebo control group Primary endpoint Radiological response: The relative risk (RR) for radiological response was significantly greater with bevacizumab versus placebo with the RR 15.00 (95% CI 1.02 to 220.92). The change in oedema (T2 FLAIR) volume from baseline to 6 weeks where 25% reduction constituted response. There was 100% response for bevacizumab versus progression +14% with placebo (P value = 0.001). Median T1gad volume change was a reduction by 63% for bevacizumab arm versus increase by 17% for placebo (P value = 0.006). Secondary endpoints Clinical improvement: clinically, 5 of 7 patients on placebo had worsening neurologic signs/symptoms from 3.1 to 8.8 weeks after randomisation versus improvement by 6 weeks in all patients receiving bevacizumab. Since there appeared to be no differences in neurocognitive test results in the trial versus cross‐over bevacizumab‐treated participants combined, all bevacizumab participants were pooled into one group. Neurocognitive tests showed that mean NCF6Z score decreased from baseline to 6 weeks by 0.07 with placebo but increased by 0.08 with bevacizumab treatment such that the MD for the bevacizumab arm was MD 0.44 higher (95% CI 0.24 higher to 1.12 higher). Bevacizumab patients had greater improvement than placebo in learning trials and delayed recognition trial on HVLT‐R but bevacizumab patients had worse performance on delayed free recall trial of memory measure (HVLT‐R delayed recall). Patients reported greater interference in their everyday activities from their symptoms (MDASI‐Interference) with bevacizumab treatment than with placebo. Corticosteroid requirement: of the 5 (38%) patients who were on dexamethasone at study registration, 1 patient had dose reduction by 12 weeks and the other 4 tolerated a slow taper. Toxicities: no patients receiving placebo experienced any AE's, whereas 6 of 11 patients who received bevacizumab as initial or cross‐over treatment experienced AE's. Three of these AE's were severe: 1 aspiration pneumonia, 1 pulmonary embolus secondary to deep vein thrombosis, and 1 superior sagittal sinus thrombosis. Quality of Life: not reported
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Although this is reported as a double‐blind, randomised study, the method of allocation to the particular study arm and blinding was not described therefore the risk of allocation concealment is unclear.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	The blinding procedure for the study participants and key study personnel was not described.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Although a double‐blind design is reported, the blinding procedure for the study participants and key study personnel was not described. Due to the small size of the study, any deficits in the blinding process for even a small number of patients potentially introduces a substantial impact on the study findings.
Incomplete outcome data (attrition bias) All outcomes	Low risk	There was complete follow‐up data of all patients for the primary outcome measure or interest.
Selective reporting (reporting bias)	High risk	One or more outcomes of interest in the review are reported incompletely so that they cannot be entered in a meta‐analysis
Other bias	Unclear risk	Insufficient information to permit judgement as to whether any other biases may be present