Physical activity (PA) interventions are generally effective in supporting short-term behaviour change, but increases are not always maintained. This review examined the effectiveness of PA interventions for behaviour change maintenance in young and middle-aged adults, and investigated which Behaviour Change Techniques (BCTs) and other intervention features were associated with maintenance. Six databases (Medline, EMBASE, PsycINFO, Cochrane Database of Systematic Reviews, CINAHL, Web of Science) were systematically searched. Eligibility criteria were controlled trials investigating the effectiveness of PA interventions with adult (mean age 18-64 years) non-clinical populations using validated measures of PA behaviour at baseline and ≥six months post-baseline. Results were pooled in meta-analyses using standardised mean differences (SMD) at five time intervals (6-9, 9-15, 15-21, 21-24, >24 months). Moderator analyses investigated the influence of sample and intervention characteristics on PA maintenance at 6-9 months. Sixty-two studies were included. PA interventions had a significant effect on behaviour maintenance 6-15 months post-baseline relative to controls. Interventions had a larger effect on maintenance at 6-9 months (SMD=0.28; 95% CI: 0.20, 0.35; I2=73%) compared to 9-15 months (SMD=0.20; 95% CI: 0.13, 0.26; I2=70%). Beyond 15 months, PA measurements were infrequent with little evidence supporting maintenance. Moderator analyses showed some BCTs and intervention settings moderated PA outcomes at 6-9 months. A multivariable meta-regression model showed interventions using the BCTs Prompt self-monitoring of behavioural outcome (b=1.46, p<0.01) and Use of follow-up prompts (b=0.38, p<0.01) demonstrated greater effectiveness at promoting PA maintenance at 6-9 months. Interventions implemented in primary care (versus community or workplace/university) settings (b=-0.13, p=0.10) tended to demonstrate less effectiveness. This review provides evidence of some effective BCTs for maintaining behaviour to 15 months. Greater consideration must be given to how future interventions encourage and measure maintenance of changes, and investigate broader psychological, social and environmental influences of PA behaviour. PROSPERO 2015:CRD42015025462.