Brown 2013

Methods	Randomised controlled trial, parallel arm, double‐blind, stratification by recursive partitioning analysis class and prior surgical therapy.
Participants	Inclusion criteria: Adult patients; pathologically proven diagnosis of solid malignancy within 5 years of registration; brain metastases visible on contrast‐enhanced MRI (or a contrast‐enhanced CT for patients unable to have an MRI) with stable systemic disease 3 months prior to study entry; receiving 37.5 Gy of WBRT via 15 fractions of 2.5 Gy; KPS ≥ 70; serum creatinine ≤ 3 mg/dL, creatinine clearance ≥ 30 mL/min, total bilirubin ≤ 2.5 mg/dL, blood urea nitrogen (BUN) 20 mg/dL; MMSE score > 18; negative serum pregnancy test. Exclusion criteria: Memantine allergy, current alcohol or drug abuse, chronic use of benzodiazepines, severe active comorbidity. No. randomised: Memantine: 278; placebo: 276. Follow‐up: 8, 16, 24 and 52 weeks. Setting: 143 centres in the United States and Canada
Interventions	Treatment arm schedule: Week 1: 5 mg oral memantine taken in the morning Week 2: 10 mg oral memantine taken in divided dosage (5 mg morning, 5 mg night) Week 3: 15 mg oral memantine taken in divided dosage (10 mg morning, 5 mg night) Week 4‐24: 20 mg oral memantine taken in divided dosage (10 mg morning, 10 mg night) Control Arm: Matched placebo The dosage was lowered to 10 mg twice daily, as week 2, if creatinine clearance decreased to 30 mL/min, and was continued at this dosage if creatine clearance fell below 5 mL/min following a weekly recheck.
Outcomes	Cognitive function (HVLT‐R, COWA, TMT)
Notes	Efficacy reported via median change and interquartile ranges.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The Zelen treatment allocation scheme was used to stratify patients according to according to recursive partitioning analysis (RPA) class and prior surgical therapy.Within each stratum, patients were randomised in a 1:1 ratio to placebo or memantine.” "A computer at RTOG headquarters randomly generated the sequences for the randomization" (obtained via correspondence).
Allocation concealment (selection bias)	Low risk	"The placebo was actually provided by the company (Forest Pharma) and it was impossible to tell which was placebo and which was active drug." (obtained via correspondence).
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinding carried out.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinding carried out.
Incomplete outcome data (attrition bias) All outcomes	Low risk	"All eligible patients randomised to the study were included (intent‐to‐treat analysis)...The multiple imputation procedure employing the Markov chain Monte Carlo method was also used to determine values for all remaining living patients missing assessments."
Selective reporting (reporting bias)	Low risk	All outcomes reported.
Other bias	Low risk	None.