A Randomized, Controlled Trial of Tai Chi for the Prevention of Falls: The Central Sydney Tai Chi Trial

Alexander Voukelatos, MA (Psychol); Robert G. Cumming, PhD; Stephen R. Lord, DSc; Chris Rissel, PhD


J Am Geriatr Soc. 2007;55(8):1185-1191. 

In This Article


A total of 702 people were randomized: 353 into the intervention group and 349 into the waiting list control group (Figure 1). The average age of participants±standard deviation was 69±6.5 (range 60–96); 84% were female. Participant characteristics are shown in Table 1 . The only statistically significant (P<.05) difference between study groups at baseline was sway on the foam mat (which was worse in the intervention group). The difference in adequacy of physical activity (which was better in the intervention group) approached conventional statistical significance (P=.07). More participants in the control group reported two or more falls in the previous year (13% vs 10% in the intervention group). Although this difference was not statistically significant (P=.2), history of falls was included as a covariate in adjusted models.

Figure 1.

Flow chart of study participants.

Eighteen participants dropped out of the study altogether and did not undertake the postintervention balance assessments or complete any falls calendars: six from the intervention group and 12 from the control group. A further 157 participants (76 intervention participants and 81 controls) provided falls data but did not have follow-up balance measures. These participants tended to be older and less well educated and have poorer self-rated health and more difficulties with activities of daily living than participants who completed the follow-up balance assessments ( Table 1 ).

Two hundred seven participants (58.6%) in the tai chi group attended at least 13 of 16 tai chi classes, with 278 (78.8%) attending at least half the classes. Overall, subjects attended 71% of the tai chi classes offered. Reasons given for not attending classes included losing interest in tai chi, class times clashing with other commitments, and illness.

The number of falls during follow-up in each study group after 16 and 24 weeks are shown in Table 2 . After 16 weeks, 19% of participants had at least one fall, and 3% had two or more falls. After 24 weeks, 22% had at least one fall, and 6% had two or more falls. There was no difference in the percentage of participants who had one or more falls, but there was nearly a 50% relative risk reduction of two or more falls (multiple falls) for the tai chi group at 16 and 24 weeks. The relative risk for multiple falls at 24 weeks (0.54, 95% confidence interval (CI)=0.28–0.96) was just statistically significant (P=.05) ( Table 3 ).

The fall rate was lower in the tai chi group than in the control group. Using negative binomial regression, the incidence rate ratio (IRR) after 16 weeks was 0.72 (95% CI=0.48–1.10, P=.1) and after 24 weeks it was 0.67 (95% CI=0.46–0.96, P=.03) ( Table 3 ). The reduced falls rate persisted after adjusting for age, sex, falls history, baseline balance (sway on mat), and adequacy of physical activity ( Table 3 ).

Using the Andersen-Gill extension of the Cox proportional hazard model, the hazard ratio (HR) for two or more falls in the intervention group compared with the control group was 0.33 (95% CI=0.14–0.78, P=.01) at 16 weeks and 0.33 (95% CI=0.18–0.62, P=.001) at 24 weeks ( Table 3 ). Adjustment for age, sex, falls history, baseline balance (sway on mat), and adequacy of physical activity did not diminish the magnitude and statistical significance of these HRs. HRs were weaker for one or more falls and were only statistically significant after 24 weeks ( Table 3 ).

There was only weak evidence that those who attended more tai chi classes had fewer falls than those who attended less frequently. For example, 18% of those who attended more than eight classes fell at least once during the first 16 weeks of the study, and 2% fell twice or more, compared with 20% and 4%, respectively, of those who attended eight or fewer classes.

Mean scores on the six study balance measures in the intervention and control groups at baseline and follow-up are shown in Table 4 . After adjusting for age, sex, falls history, and adequacy of physical activity at baseline, the intervention group performed statistically significantly better on five of the six balance variables than the control group: sway on floor (P=.02), sway on foam mat (P=.004), lateral stability (P=.005), coordinated stability (P<.001), and choice stepping reaction time (P<.001). All these variables except choice stepping reaction time (P=.20) were also statistically significantly different before adjustment. There was no difference between groups for the maximal leaning balance range test, before (P=.20) or after adjustment (P=.50).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.