Specificity of Balance Training in Healthy Individuals: A Systematic Review and Meta-Analysis

Jakob Kümmel, Andreas Kramer, Louis-Solal Giboin, Markus Gruber

It has become common practice to incorporate balance tasks into the training program for athletes who want to improve performance and prevent injuries, in rehabilitation programs, and in fall prevention programs for the elderly. However, it is still unclear whether incorporating balance tasks into a training program increases performance only in these specific tasks or if it affects balance in a more general way. The objective of this systematic literature review and meta-analysis was to determine to what extent the training of balance tasks can improve performance in non-trained balance tasks. A systematic literature search was performed in the online databases EMBASE, PubMed, SPORTDiscus and Web of Science. Articles related to balance training and testing in healthy populations published between January 1985 and March 2015 were considered. A total of 3093 articles were systematically evaluated. Randomized controlled trials were included that (i) used only balance tasks during the training, (ii) used at least two balance tests before and after training, and (iii) tested performance in the trained balance tasks and at least one non-trained balance task. Six studies with a total of 102 subjects met these criteria and were included into the meta-analysis. The quality of the studies was evaluated by means of the Physiotherapy Evidence Database (PEDro) scale. A random effect model was used to calculate the between-subject standardized mean differences (SMDbs) in order to quantify the effect of balance training on various kinds of balance measures relative to controls. The tested balance tasks in each study were classified into tasks that had been trained and tasks that had not been trained. For further analyses, the non-trained balance tasks were subdivided into tasks with similar or non-similar body position and similar or non-similar balance perturbation direction compared to the trained task. The effect of balance training on the performance of the trained balance tasks reached an SMDbs of 0.79 [95 % confidence interval (CI) 0.48-1.10], indicating a high effect in favor for the trained task, with no notable heterogeneity (I (2) = 0 %). The SMDbs in non-trained categories reached values between -0.07 (95 % CI -0.53 to 0.38) and 0.18 (95 % CI -0.27 to 0.64), with non-notable to moderate heterogeneity (I (2) = 0-32 %), indicating no effect of the balance training on the respective non-trained balance tasks. With six studies, the number of studies included in this meta-analysis is rather low. It remains unclear how the limited number of studies with considerable methodological diversity affects the outcome of the SMD calculations and thus the general outcome of the meta-analysis. In healthy populations, balance training can improve the performance in trained tasks, but may have only minor or no effects on non-trained tasks. Consequently, therapists and coaches should identify exactly those tasks that need improvement, and use these tasks in the training program and as a part of the test battery that evaluates the efficacy of the training program. Generic balance tasks-such as one-leg stance-may have little value as overall balance measures or when assessing the efficacy of specific training interventions.