Mirelman 2016.

Methods	Study design: RCT Number of study arms: 2 Length of follow‐up: 6 months
Participants	Setting: Belgium, Israel, Italy, the Netherlands, and the UK Number of participants: 152 Number analysed: no fall data Sample: community‐dwelling Age (years): mean 82.6 Sex: 35% female Inclusion criteria: aged 60 − 90 years, able to walk ≥ 5 minutes unassisted, stable medication for the past month, self‐reported ≥ 2 falls within 6 months before screening; individuals with mild cognitive impairment were included if they had a score of 0·5 on the Clinical Dementia Rating scale Exclusion criteria: psychiatric comorbidity (e.g. major depressive disorder in accordance with DSM IV criteria); history of stroke, traumatic brain injury, or other neurological disorders (other than Parkinson’s disease and mild cognitive impairment, for those groups); acute lower back or lower extremity pain; peripheral neuropathy; rheumatic and orthopaedic diseases; or a clinical diagnosis of dementia or severe cognitive impairment (MMSE score < 21)
Interventions	1. Individual, supervised treadmill training: progressed with treadmill duration and speed; 45‐minute session, 3 a week for 6 weeks 2. Individual, supervised treadmill training plus virtual reality: as (1) plus received projected images of the virtual environment (e.g. obstacles, distractors) that necessitated continual adjustment of steps; 45‐minute session, 3 a week for 6 weeks
Outcomes	1. Health‐related quality of life
Duration of the study	26 weeks
Adherence	Adherence measured by number of completed sessions of the 18 sessions: 1. Individual, supervised treadmill training: 16·82 (SD 1·81) 2. Individual, supervised treadmill training plus virtual reality: 16·62 (SD 1·78)
Notes	At baseline 130 participants had Parkinson's disease, 43 mild cognitive impairment, 109 idiopathic falls. Falls data unavailable only for non‐Parkinson's disease participants Source of funding: European Commission Economic information: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "By use of computer‐based allocation, participants were randomly assigned"
Allocation concealment (selection bias)	Low risk	Group allocation performed by a third party not involved in the day‐to‐day running of the study; treating therapist notified by e‐mail to ensure concealed allocation
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Participants and personnel unblinded but impact of unblinding unknown
Blinding of outcome assessment (detection bias) Falls	Low risk	Falls recorded using same method in each group Quote: "Falls were recorded without knowledge of training group"
Blinding of outcome assessment (detection bias) Fractures	Unclear risk	Not applicable
Blinding of outcome assessment (detection bias) Hospital admission, medical attention and adverse events	Unclear risk	The method of recording adverse events was unclear
Blinding of outcome assessment (detection bias) Health related quality of life (self report)	High risk	Participants unblinded to intervention group
Incomplete outcome data (attrition bias) Falls and fallers	Low risk	Less than 20% had missing data (7%) for the study. Missing data were balanced between the treadmill training group (n = 12) and treadmill plus virtual reality group (n = 8), with reasons for missing data similar between groups (e.g. 2 adverse events in treadmill group, 3 adverse events in virtual reality group)
Selective reporting (reporting bias)	High risk	Falls measured, but number of fallers is not presented
Method of ascertaining falls (recall bias)	Low risk	Participants received a falls calendar, which they were provided as a paper version, web‐based calender, or a smartphone application. Research staff contacted all participants every month to maximise compliance