The a priori criteria for studies to be included in the review are presented in Box 1. Studies were excluded if the participants were hospital inpatients or resided in an aged care facility. Studies in which subjects had health conditions likely to significantly affect their balance were also excluded, as were studies in which healthy elderly subjects with extremes of balance (either minimal or maximal deficits) were excluded, or gait aid users were excluded. Where

there were inadequate details of methods or results, an email was sent to the author where possible to seek further information. Design • Any study MAPK inhibitor design reporting baseline data on an unselected cohort Participants • Community dwelling Outcomes measures • Berg Balance Scale mean Participants: The inclusion and exclusion criteria and the country in which the data were collected were extracted for each trial. The sample size and the mean age of the participants were also extracted, Gefitinib nmr along with whether the participants were enrolled as an observational cohort, an intervention group, or a control group. Outcome: Means and standard deviations were extracted for baseline Berg Balance

Scale scores. Where variability data were presented as other statistics, these were converted to standard deviations. Meta-regression analysis of the mean Berg Balance Scale scores was conducted. Where studies provided participant groups stratified by age, analysis was conducted using subgroups rather than pooled data. In studies where subjects were listed by age decade without provision of the mean age within the data, the mean age was assumed to be the mid-point of the decade. Where studies provided data for treatment and control groups in a trial, the baseline data for each group were included in the analysis separately. To account for differences in the statistical

power of the studies included in the meta-regression analysis, samples with larger numbers and samples with homogenous balance scores are weighted more highly when calculating the overall relationship between age and Berg Balance Scale score. Conversely, small samples and samples with highly variable balance scores were given less very weight. The relationship between the mean age of a sample and the standard deviation of the Berg Balance Scale scores of the sample was investigated using linear regression analysis, with weighting for sample size. After duplicates were removed, 859 articles were found containing the term ‘Berg Balance Scale’ in their abstract, title, or keywords. Hand searches of reference lists revealed one additional relevant paper. Of these, 17 were deemed relevant and included in the analysis. Figure 1 presents the flow of studies through the review and the reasons for exclusion.