Eighteen studies were included: six RCTs (n= 9,030); one non-randomised controlled study (n=58); four observational studies (n=1,284, included two surveys and two cohort studies); and six process evaluations (n>1,276) and one study (n=30) that only reported qualitative data.
Controlled studies: Study quality scores ranged from 15 to 29 (median 26) out of 34.
RCTs: All RCTs scored 19 or more points for quality. Methodological limitations included lack of generalisability of the sample, lack of blinding and inadequate allocation concealment. Participation rates ranged from 26% to 92%; fewer than half of study entrants completed the exercise programme. Exercise referral schemes were associated with a statistically significant increase in the proportion of patients who were moderately active compared to controls (relative risk based on intention-to-treat analysis 1.20, 95% CI: 1.06 to 1.35; number needed to treat 17). No significant heterogeneity was found. Results for random-effects models were similar.
Observational studies: Studies were of poor to moderate quality. Response rates were 40% to 55% in two UK surveys. Studies reported mixed findings (details were reported).
Results for specified anthropometric, physiological, biochemical and psychological outcomes for RCTs and other results from observational studies, process evaluations and qualitative studies were also reported.