Fifteen RCTs (8,745 participants) were included in the review. All except two RCTs were considered to be at high risk of bias in at least one assessment area. The main methodological weaknesses related to blinding and completeness of data.
Small to medium improvements in self reported physical activity at 12 months were found using dichotomous data (OR 1.42, 95% CI 1.17 to 1.73; 11 RCTs; Ι²=43%) and continuous data (SMD 0.25, 95% CI 0.11 to 0.38; nine RCTs; Ι²=70%). The number needed to treat was 12 (95% CI 7 to 33; nine RCTs). The pooled effect for improvements in cardiorespiratory fitness was not statistically significant (four RCTs). No statistically significant effect was found for exercise referral programmes (three RCTs for dichotomous data; two RCTs for continuous data). Adverse events were not increased as a result of physical activity promotion in five RCTs; one RCT found an 11% increase in falls and a 6% increase in injuries following the intervention.
In subgroup analyses, physical activity interventions were significantly more effective than no intervention in terms of self reported behaviour at 12 months (six RCTs) but there was no statistically significant effect when they were evaluated against other comparators (seven RCTs). Increases in those meeting recommended physical activity levels at 12 months were comparable between active study groups. Where the control group received no intervention, increases were smaller.
Funnel plots did not suggest evidence of publication bias.