Table 1 summarizes the characteristics of the fourteen studies included in the systematic review. All studies were published between 2005 and 2024, and the majority were conducted in the USA (N = 8, 60%) [42,43,44,45,46,47,48,49], with others in Canada (N = 2) [50, 51], Australia (N = 1) [52], Germany (N = 1) [53], Norway (N = 1) [54], and China (N = 1) [55]. In total, 1169 participants were included in the fourteen studies. Study sample sizes ranged from 16 to 280 (median, 56 participants), with 8 to 140 participants per intervention arm (median, 30 participants) [50]. All studies included AYA cancer survivors, with a mean age of 29.2 (SD 5.5, range 15–48). Participants were primarily female (64%) and non-Hispanic White (80%). Mean BMI ranged from 25 to 33.4 kg/m2 (median, 29.8 kg/m2). Studies included post-treatment survivors of various cancer types, including both hematopoietic malignancies and solid tumors.

The primary objective for most studies was feasibility and acceptability (N = 8) [42,43,44,45,46,47,48, 51], and all studies had multiple behavioral and psychosocial outcomes of interest. All interventions focused on improving PA (N = 15) [42,43,44,45,46,47,48,49,50,51,52,53,54,55], and only two studies included dietary components [47]. Most studies included two-arm intervention-control comparisons; Li et al. compared two intervention groups to one control [55]. Keadle et al. compared two PA intervention groups: PA or PA plus a charity donation incentive if step goals were achieved [43]. For the meta-analysis, the intervention group testing the additional novel component was considered the intervention group. Of the remaining thirteen studies, seven compared an intervention to standard care (54%), three utilized a waitlist control, and three had a self-help control group that included the provision of a Fitbit activity tracker.

The median length of follow-up was 3 months, with a range of 10 weeks to 12 months. Most interventions were delivered remotely (N = 11) [42,43,44,45,46,47,48,49,50, 53, 55] via mHealth (N = 7) [42,43,44,45, 48, 49, 55], consisting of a combination of Fitbits (N = 5) [42, 43, 48, 49, 55], websites (N = 4) [44, 45, 48, 49], social media (N = 4) [42, 48, 49, 55], and text messages (N = 2) [47, 49] or telephone [53]. One study examined the effectiveness of a digital tool designed to help care teams deliver health behavior counseling and community resources. One study consisted primarily of mailed print materials [56], one was delivered via telephone [46], and one was delivered exclusively via text messages [47]. Three [51, 54] studies were individualized, supervised exercise programs with accredited exercise physiologists [52]. Thirteen studies cited health behavior theories as informing the study, most commonly Social Cognitive Theory (N = 7 [43, 45,46,47,48,49, 55]), Self-determination Theory (N = 4 [42,43,44, 49]), Transtheoretical Model (N = 3 [45, 46, 53]), and the Theory of Planned Behavior (N = 2) [50, 55].

Figure 2 presents details of the risk of bias assessments for individual studies [57, 58]. Most studies (N = 12, 80%) were judged to have a low risk of bias, and three were deemed to have some concerns. The main reasons for concern were differential attrition and incomplete outcome data not being analyzed using intent-to-treat methods.

Of the fourteen studies (fifteen intervention arms), eight (57% [43, 44, 46,47,48, 52, 54, 55]) were judged to be promising, four (29% [45, 49, 50,