Content, delivery and effects of physical activity (PA) interventions are heterogeneous. There is a need to identify intervention features (content and delivery) related to long-term effectiveness. Behaviour change techniques (BCTs) and modes of intervention delivery were coded in 19 randomised controlled trials included in a systematic review of PA interventions for adults aged 55–70 years, published between 2000 and 2010, with PA outcomes ≥12 months after randomisation; protocol registration: PROSPERO CRD42011001459. Meta-analysis, moderator analyses and meta-regression were conducted. Meta-analysis revealed that interventions were effective in promoting PA compared with no/minimal intervention comparators [d = 0.29, 95% CI = 0.19–0.40, I2 = 79.8%, Q-value = 89.16 (df = 18, p < 0.01)]. Intervention features often concurred and goal setting was the most commonly used BCT. Subgroup analyses suggested that interventions using the BCT feedback may be more effective, whilst interventions using printed materials or the BCTs information on where and when to perform the behaviour and information on consequences of behaviour to the individual may be less effective. Meta-regression revealed that neither the number of BCTs nor self-regulatory BCTs significantly related to effect size. Feedback appears to be a potentially effective candidate BCT for future interventions promoting long-term PA. Considering concurrence of intervention features alongside moderator analyses is important.