Kenneth R Fox

Background Mobility limitation in older age reduces quality of life, generates substantial health- and social-care costs, and increases mortality. Objective The REtirement in ACTion (REACT) trial aimed to establish whether or not a community-based active ageing intervention could prevent decline in physical functioning in older adults already at increased risk of mobility limitation. Design A multicentre, pragmatic, two-arm, parallel-group randomised controlled trial with parallel process and health economic evaluations. Setting Urban and semi-rural locations across three sites in England. Participants Physically frail or pre-frail older adults (aged ≥ 65 years; Short Physical Performance Battery score of 4–9). Recruitment was primarily via 35 primary care practices. Interventions Participants were randomly assigned to receive brief advice (three healthy ageing education sessions) or a 12-month, group-based, multimodal exercise and behavioural maintenance programme delivered in fitness and community centres. Randomisation was stratified by site and used a minimisation algorithm to balance age, sex and Short Physical Performance Battery score. Data collection and analyses were blinded. Main outcome measures The primary outcome was change in lower limb physical function (Short Physical Performance Battery score) at 24 months, analysed using an intention-to-treat analysis. The economic evaluation adopted the NHS and Personal Social Services perspective. Results Between June 2016 and October 2017, 777 participants (mean age 77.6 years, standard deviation 6.8 years; 66% female; mean Short Physical Performance Battery score 7.37, standard deviation 1.56) were randomised to the intervention arm ( n = 410) or the control arm ( n = 367). Data collection was completed in October 2019. Primary outcome data at 24 months were provided by 628 (80.8%) participants. At the 24-month follow-up, the Short Physical Performance Battery score was significantly greater in the intervention arm (mean 8.08, standard deviation 2.87) than in the control arm (mean 7.59, standard deviation 2.61), with an adjusted mean difference of 0.49 (95% confidence interval 0.06 to 0.92). The difference in lower limb function between intervention and control participants was clinically meaningful at both 12 and 24 months. Self-reported physical activity significantly increased in the intervention arm compared with the control arm, but this change was not observed in device-based physical activity data collected during the trial. One adverse event was related to the intervention. Attrition rates were low (19% at 24 months) and adherence was high. Engagement with the REACT intervention was associated with positive changes in exercise competence, relatedness and enjoyment and perceived physical, social and mental well-being benefits. The interventio