Voukelatos 2007.

Methods	Study design: RCT Number of study arms: 2 Length of follow‐up: 6 months
Participants	Setting: Sydney, Australia Number of participants: 702 Number analysed: 684 Number lost to follow‐up: 18 Sample: community‐dwelling Age (years): mean 69 (SD 6.5), range 69 ‐ 70 Sex: 84% female Inclusion criteria: aged > 60; community‐dwelling Exclusion criteria: degenerative neurological disease; severely debilitating stroke; metastatic cancer; severe arthritis; unable to walk across a room independently; unable to use English
Interventions	1. Group‐based Tai Chi: style of Tai Chi differed between classes depending on Tai Chi instructor; 1‐hour class, 1 a week for 16 weeks. Cost AUD 44 2. Control: instructed not to take part in a Tai Chi programme and placed on 24‐week waiting list, then offered Tai Chi programme
Outcomes	1. Rate of falls 2. Number of people who experienced 1 or more falls (risk of falling)
Duration of the study	24 weeks
Adherence	Adherence measured by retention/attrition rate 1. Group‐based Tai Chi: dropout: 6. 76 participants provided falls data but did not complete the 16‐week balance assessment 2. Control: dropout: 12. 81 participants provided falls data but did not complete the 16‐week balance assessment
Notes	Source of funding: New South Wales Health Department Economic information: Mean cost per person (intervention): AUD 245 plus charged AUD 44 per participant. Healthcare service costs: Tai Chi group AUD 55, control group AUD 17. Incremental cost per fall prevented/per QALY gained: AUD 1683 per fall prevented (includes cost offset by charging AUD 44 per instruction course).Cost‐effectiveness analysis reported in Haas 2006
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomization list ... was prepared for each venue using randomly permuted blocks of four or six"
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Participants and instructors conducting classes in intervention group were not blinded. Control participants were asked not to take classes during the study period, but may have accessed other fall‐prevention interventions. Insufficient evidence to make judgement on impact of lack of blinding.
Blinding of outcome assessment (detection bias) Falls	Low risk	Falls were recorded using the same method in both groups. Outcome assessors were blinded to group assignment
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	Not applicable
Blinding of outcome assessment (detection bias) Health related quality of life (self report)	Unclear risk	Not applicable
Incomplete outcome data (attrition bias) Falls and fallers	Low risk	Less than 20% of fall data were missing (3%). Missing data were balanced across groups, with 6/347 participants missing from the intervention group and 12/249 missing from the control group. The reasons for missing data were balanced between groups
Selective reporting (reporting bias)	Unclear risk	Minimum set of expected outcomes not reported (adverse events not reported)
Method of ascertaining falls (recall bias)	Low risk	Quote: "Participants were given falls calendars and were instructed to record on the calendar each day for 24 weeks whether they had had a fall." Pre‐paid postage calendars returned at the end of each month, with telephone call if not returned within 2 weeks.