Effectiveness of sensor-based interventions in improving gait and balance performance in older adults: systematic review and meta-analysis of randomized controlled trials.

Fifty-three of the 58 trials were rated as having some concerns for risk of bias and only one was judged low risk, so the body of evidence is less solid than the number of studies suggests. The three sensor types were never compared head to head; there were too few trials, especially for wearable sensors, which contributed only seven trials in total and five for the Timed Up and Go. On the two outcomes most used in the clinic, the Timed Up and Go and gait speed, the pooled improvements were smaller than the minimal clinically important difference, so the statistical advantage may not be something a patient feels. The benefit shrank when sensor-based training was compared with traditional exercise rather than no treatment, so part of the apparent effect comes from doing something rather than nothing. The included trials did not report how familiar with or accepting of the technology the older adults were, and the authors say this may limit how far the results generalise. Several pooled analyses had high heterogeneity (gait speed, for example, at 92%), and only outcomes reported in at least three trials could be combined.

Funding is listed from five sources: the Youth Innovation Talents in Universities of Guangdong Province, Hong Kong Polytechnic University grants, a Sun Yat-Sen University start-up fund, the Shenzhen-Hong Kong-Macao Science and Technology Project Fund, and the Health and Medical Research Fund. The supplied text contains no competing-interest statement.

Do not read the significant results as proof that sensor-based training beats ordinary exercise in the clinic. The pooled gains on the Timed Up and Go and on gait speed were smaller than the difference considered clinically important, and the advantage shrank when the comparison was against traditional exercise rather than no treatment at all.