For the knee flexor isometric strength, the ICC was 0.95 and the %SEM was 6.1%. For the knee extensor isometric strength, the ICC was 0.97 and the %SEM was 6.1%. The different variables were analysed at baseline using descriptive statistics, and the distribution of the data was examined using the Kolmogorov-Smirnov test with Lilliefors correction. After confirming that the distribution

of all variables was parametric, the comparisons between groups were performed using a two-way analysis of variance for repeated measures. The significance level was set at p < 0.05 and all analyses followed the principle of intention to treat. Means, SDs and 95% CIs were provided to depict the change within each intervention group during the course of the study and the treatment effect. The mean and 95% CI were calculated using Student's t-test. Three linear regressions were GSK-3 cancer performed. The first was performed to determine how much of the change in fear of falling, as measured by the Falls Efficacy Scale International questionnaire, was predicted by the baseline

characteristics of see more the participants. To introduce a new variable in the prediction model, a significance level below 0.05 was required. The second linear regression was performed to determine the strength of the correlation between the change in fear of falling and the change in the Falls Risk Test. The last linear regression was performed to determine the strength of the correlation between the change in the Falls Risk Test and the change in the isometric strength of the knee extensors. A 7-day reliability study was conducted on the dynamic balance and strength variables in our study with 10 study participants. The relative reliability was determined according to the ICC3,1 obtained from two sessions (Shrout and Fleiss 1979). The absolute

reliability was determined by the SEM, which was defined as SD*√(1-ICC), where SD is the average SD of Day 1 and Day 2 (Weir 2005). We anticipated that a 5-point improvement in the Falls Efficacy Scale International score would be sufficient to move typical patients in our nursing home from their current categorisation as ‘high concern’ into the ‘moderate concern’ category (Delbaere et al 2010), which we considered a clinically important change. Anticipating enough a standard deviation of 8.5, we calculated that 47 participants would provide 80% power to detect a difference of 5 points as significant at a two-sided, 5% significance level. To allow for some loss to follow-up, we aimed to recruit 50 participants. Effect size was used to determine the magnitude of change and was calculated as the difference in the mean change in each group divided by the average of the standard deviations. Cohen’s coefficient was used to assess the change. A change from 0–0.2 was considered very small, a change of 0.2–0.6 was considered small, a change of 0.6–1.2 was considered moderate, a change of 1.