Physical Therapist Involved Physical Activity Interventions: A Systematic Review and Meta-Analysis.

The broad inclusion criteria produced substantial variation between trials in populations, intervention types, settings, dose and follow-up, which limits how confidently the results can be pooled. Most of the included articles were at high risk of bias, only 11 of 60 (18%) were rated low to moderate, and none reached excellent reporting quality, so the individual trial results are not reliable. The sensitivity analysis for directly measured activity rested on a small number of articles with wide confidence intervals, which increases uncertainty around that estimate. The authors say they offer recommendations for research rather than practice because they did not apply the GRADE framework, and they note that heterogeneity and risk of bias raise the risk of a false positive finding. The predictive intervals for the main analyses crossed zero, meaning a future trial could show a positive or a negative effect. Twenty-five different activity measures were used, and measurement confusion was common; self-report questionnaires had only moderate reliability and limited validity compared with device-based measures, while device measures can be hard to collect in clinic and may be inaccurate for people walking slower than 0.6 m/s. Trials in which activity was a secondary rather than primary focus were included, which the authors say may have biased results away from more effective primary activity-focused interventions. Grey literature and articles in languages other than English or Spanish were excluded. Twelve of the trials had 40 or fewer participants.

Do not read the combined exercise plus behaviour change result as settled practice. It comes from an exploratory meta-regression rather than the main analysis, the predictive interval still crossed zero, and most of the included trials were at high risk of bias.