Figure 1 shows the flow of studies through the review process and the reasons for exclusion. Database searches resulted in 2020 potentially relevant titles. The full text of 110 articles was assessed for eligibility, and 18 studies (reported in 33 publications) met inclusion for the review (Table 1). Studies were grouped and reported by intervention topic: PA only (ten trials), diet only (one trial), or multiple health behavior (PA and diet) (seven trials).
Risk of bias assessment
There was initially 75 % agreement between authors on the study assessment criteria and full consensus was achieved after discussion. Risk of bias results are reported in Table 2. Of the ten PA-only studies, five were classified as strong methodological quality [48–52], three as moderate [53–56], and two as weak [57, 58]. The diet-only trial was classified as moderate . In the seven multiple behavior studies, one was classified as strong , four were moderate [61–65], and two weak [66, 67]. Three trials were excluded from the meta-analysis [57, 58, 67] due to being weak.
The most common areas with a high risk of bias were selection bias, confounders, and blinding. Eleven studies [53, 54, 56–58, 61–67] were rated as weak in selection bias category with less than 60 % of potentially eligible participants recruited. Two studies [59, 66] were rated as weak as the control of confounders was not described. No studies received a strong rating for blinding as (understandably given they are behavior change trials), all participants were aware of the research question, and if the outcome assessor was also aware of the intervention status of participants, studies were rated as weak methodological quality [57, 58, 67].
Physical activity trials
There were ten trials that targeted PA alone [48–54, 56–58].
Ten PA trials reported a total of 960 participants (range 36–330). Six trials targeted breast cancer survivors [48, 50–52, 54, 58, 68–72], one targeted colorectal cancer survivors , one targeted both breast and bowel cancer survivors , and two included cancers of mixed diagnoses [56, 57]. Mean time since diagnosis was 3.1 years (range 0.9–4.9 years) [50, 51, 53, 54, 56–58]. Time since diagnosis was not reported in two studies [49, 52]; however, one trial reported participants were scheduled to begin chemotherapy . All other trial participants had completed active cancer treatment (excluding hormone treatment). In five studies, only cancer survivors who were inactive or insufficiently active were eligible to participate [48–50, 57, 58]. Three trials used a wait-list control group design [48, 51, 58], three had attention control groups [50, 56, 57], and three had usual care control groups [49, 52, 54]. One trial used an attention control design, with the control group offered a limited intervention (written materials only) at the end of the study .
One intervention was delivered by email , one delivered by mail [54, 55], and one delivered using Facebook , and all others used a combination of delivery formats, including telephone [49, 50, 52, 53, 57, 58], mail [50, 53], and face-to-face counseling [51–53, 57, 58]. The majority were home-based, with only one intervention reporting supervised PA sessions . Three were walking interventions [51, 52, 58], and four had PA goals that were based on duration [49, 54, 56, 57] and/or moderate intensity [50, 53, 54, 56, 57]. One targeted resistance training .
Interventions were commonly 12 weeks in duration [48, 50, 51, 53, 54, 56, 58] and ranged from 6  to 18 weeks . The average number of intervention contacts was 15, and ranged from 1  to 52 . Intervention adherence was high, ranging from 94 % compliance with home exercise logs  to 99 % of total contacts completed . Telephone counseling adherence was also high with a median of 9 (of 11) calls completed  and a mean of 11 (of 12) calls completed [50, 53]. The intervention delivered using Facebook reported lower adherence, with 81 % of intervention participants who reported receiving ten or more messages from Facebook, and 49 % had made two or more Facebook posts .
Two trials used an objective measure (accelerometer) to assess PA behavior change [50, 51]. All others relied on self-report measures [48, 49, 52–54, 56–58] or used an objective measure in a subsample only . Effect sizes for PA behavior change were reported in four studies (d = 0.55–1.93) [51, 53, 54, 57]. Three home-based walking interventions reported significant improvements postintervention (6–12 weeks) to total PA (d = 1.02; P = 0.004)  and walking [52, 58]. Three moderate intensity interventions reported significant postintervention increases in PA (d = 0.55; P < 0.05) (d = 1.93; P = 0.02) [50, 53, 57], and two reported nonsignificant increases to aerobic and moderate-vigorous PA [54, 56]. One trial that targeted resistance training reported significant improvements and that the odds of meeting the resistance training guidelines had increased by 3.38 in the tailored intervention group .
Of the four trials that reported follow-up assessments of 6 months or longer [51, 53, 57, 68], only two reported behavior change 3 months after intervention completion [53, 68]. One trial reported that accelerometer-assessed behavior was maintained , and one reported that there were significant postintervention changes that were not maintained at 6 and 12 month follow-ups . Study retention was high, with a mean retention rate of 86 % (range 71 %  to 95 % [50, 53]). One trial  reported adverse events involving two participants that experienced anemia, shortness of breath, and dizziness. Participants in this trial were undergoing active treatment at the time of intervention.
Meta-analysis of SCT intervention effects on physical activity
Meta-analysis was conducted with 12 trials, which reported PA outcomes [48–54, 56–58, 60, 63–65] in Fig. 2. Six trials were not included in the meta-analysis [57–59, 61, 66, 67]. Reasons for exclusion were as follows: diet-only study ; did not report adequate information from the modified PA recall ; and did not report duration of PA ; or had a high risk of bias [57, 58, 67].
Two trials consisted of three study arms, which compared two PA interventions to a standard recommendation control [54, 55], and a PA intervention to a PA and diet intervention, compared to an attention control group . The results for each intervention compared to the control group are reported separately in the meta-analysis. Results were pooled to establish the effects of interventions on total PA at intervention completion. As there was moderate heterogeneity among interventions (χ
2 = 22.71, df = 13 [P = 0.05]; I
2 = 43 %), the random effects models were used. The impact of interventions on PA immediately postintervention was significant (SMD = 0.33 [0.23, 0.44], Z = 6.34 [P < 0.00001]) (Fig. 2). Sensitivity analysis was undertaken that compared this analysis to a meta-analysis which included the three trials with high risk of bias [57, 58, 67]. There was no change to the impact of interventions on total PA (SMD = 0.34 [0.24, 0.44]). However, there was an increase in heterogeneity (χ
2 = 30.31, df = 16 [P = 0.02]; I
2 = 47 %).
One trial reported dietary outcomes only . The trial targeted men receiving active surveillance for prostate cancer, with a dietary counseling intervention delivered by telephone over 6 months. Men completed self-report measures and provided blood samples for objective assessment of carotenoid intake at completion of the intervention. Retention rate was 97 %, with significant increases to vegetable consumption. There was no change to fruit, whole grains, beans, or fat consumption .
Multiple behavior trials
Seven studies focused on multiple behaviors (PA and diet) [60, 61, 63–66].
A total of 1107 participants were randomized (range 40–543). Three studies recruited newly diagnosed patients [60, 62–64], and patients were diagnosed with breast [60, 62–64, 66], prostate [62, 63], colorectal , or endometrial [65, 67, 73] cancer. Only three trials reported time since diagnosis, which varied from a mean of 3.8 months , to a mean of 20.6 months [65, 67]. In two trials, participants were scheduled to start chemotherapy [60, 64]. Three trials were aimed exclusively at overweight or obese breast  or endometrial cancer survivors [65, 67], with the aim of achieving weight loss through changing PA and diet behaviors. Two trials reported PA  or PA and diet  outcomes for the purpose of intervention adherence.
Three trials had a usual care control group [65–67], and one used a usual care comparison group with tailored newsletters at study completion . Two had attention control groups [60, 63], and one used attention control with the same written materials and pedometer as the intervention group .
All of the interventions were home-based and did not include any supervised PA. All trials targeted both PA and diet behaviors. Four of these targeted weight loss [60, 64–67], or prevention of weight gain [60, 64–66], through changing PA and diet behaviors [60, 64–66]. Four trials were aimed at increasing fruit and vegetables and reducing fat [60, 61, 63, 65, 67], one targeted only energy and fat , two included both aerobic and strength activity [60, 67], two targeted moderate or moderate-to-vigorous PA [64, 66], and the remaining three targeted PA [61, 63, 65].
Interventions were delivered using telephone counseling [60, 61, 64–67], written materials [60–62, 64, 65, 67], or face to face [65–67]. Most trials used multiple delivery modes, with one that used tailored newsletters . Intervention duration was 6 months [60, 65], 9 months , 10 months , or 12 months [64, 66, 67]. The average number of contacts was 27 (ranging from 4  to 88 ). Studies reported high adherence with all intervention components ranging from 73 to 100 %, with no difference between delivery modes.
Objectively assessed weight or body fat was reported as the primary outcome in five studies [60, 64–67]. All trials assessed diet using a range of self-reported measures [60, 61, 63–67]. Only one trial used an objective measure of PA (accelerometer) , and the remaining trials assessed PA by self-report only [61, 62, 64–67].
Follow-up periods were 6 months in one study , 12 months [61, 64–67], or 2 years . However, only two studies reported follow-up beyond postintervention time point [63, 65]. At 12 months, one trial reported significant differences in PA levels, with no difference in diet . After 2 years of follow-up, both study groups had maintained increased fruit and vegetable consumption, decreased saturated fat, and improved overall diet quality . The mean retention rate was 84 % (range 75 %  to 96 %). No adverse events were reported in two studies [60, 61, 63, 64]. In two studies that reported adverse events, 10–13 % (n = 4; n = 74) [63, 64] of the total sample reported serious adverse events that led to withdrawal.
Five of the seven studies [60, 61, 63, 64, 67] reported significant improvements in one or more aspects of diet quality, as assessed by self-report, over the medium to long term (6 months to 2 years). The remaining two studies reported nonsignificant decreases in energy [65, 66] and fat intake in the intervention groups . Inconsistent improvements in fruit and vegetable consumption were reported using a two-item screening question; however, these improvements were not found when using the comprehensive Food Frequency Questionnaire results . At 6 months, significant improvements were reported for vegetables , fruit , combined fruit and vegetables , and decreased fat ; however, there was no change for energy . At 12 months, significant improvements were reported for fruit and vegetables, by a mean of 0.5 , 0.9  to 3.1 serves per day . There was also a significant reduction by 2.3 % in the proportion of energy from fat , which is supported by similar reductions in other trials [64–66]. There were two trials that reported no effect on PA at 3, 6, or 12 months [60, 61]. At 12 months, four trials reported improvements in PA, ranging from 20 min per week (P = 0.02)  to 89 min per week  to 117 min per week , and a difference of 17.8 (P = 0.002) on the Leisure Score Index .
Recent development of behavior change taxonomies [28, 75, 76] has encouraged consistent reporting of behavior change techniques. We have mapped the SCT constructs to identify the behavior change techniques that align with each construct, using the taxonomy designed to change PA and eating behaviors  in Table 3.
Self-efficacy was the most commonly assessed construct [48–50, 52, 55–57], with four PA studies that assessed other SCT constructs [48, 51, 53, 55] (Table 4). Two studies [52, 57] reported that the study was based on Bandura’s self-efficacy theory, and the only construct operationalized was self-efficacy. Four PA interventions that used telephone or face-to-face counseling reported that the counseling principles were based on SCT [49, 51, 53, 58]. Five studies reported that the study was based on SCT and the transtheoretical model (TTM), or elements of TTM, such as stages of change [50, 53, 57, 61, 62]. Stage of change was assessed in four studies [50, 51, 53, 57] with reference to both the TTM and social cognitive theories. However, one trial assessed stage of change, despite not providing any reference to TTM or rationale for why stage of change was assessed . Most studies reported using goal setting [48, 51–54, 56–58, 77]; however, few specifically reported action plans  or review of goals [56–58]. The most common strategy to increase self-efficacy was to provide a pedometer [49, 50, 52, 53, 56, 57] and/or a log sheet for self-monitoring of PA behavior [49–53]. Social support or social comparison was the most common outcome expectancy targeted [48, 51, 54, 56, 58], and two trials reported strategies targeting environment [51, 54] or relapse prevention [49, 53]. Five trials incorporated identification and discussion of barriers and how to overcome them [50, 51, 53, 57, 58], but only one prompted a focus on past successful strategies .
In PA-only trials, improvements in self-efficacy were associated with increased PA in three studies [49, 50, 52]. Moderation analyses identified that intervention participants with high self-efficacy increased their PA levels faster over the 6 month assessment period compared to intervention participants with low self-efficacy . Mediation analyses identified that improvements in barrier interference and barrier self-efficacy mediated 39 and 19 % of the intervention effect on PA maintenance 3 months after the intervention . There were no significant changes in decisional balance pros, cons, or experiential processes of change [71, 72], or task self-efficacy, social support, outcome expectations, or fear of exercise . Two trials assessed but did not report results for self-regulation, outcome expectancy values, exercise self-efficacy, exercise role identity, behavioral capability, or social support [48, 54, 55]. Intervention effects on stage of change results were mixed, with one trial that reported a medium-to-large effect , one reported significant postintervention improvements that declined over subsequent follow-ups , and one reported no effect on stage of change .
The diet-only trial reported that the telephone counseling protocol “used strategies adopted from SCT”; however, no further detail was provided . In multiple behavior studies, one did not report how SCT constructs were operationalized . All other trials reported goal setting, self-monitoring, building self-efficacy (for PA and diet) [60–63, 67], or diet [65, 73], or for maintaining a healthy lifestyle , overcoming barriers, and social support [61–66, 73]. Goal setting [48, 63–67] and review of goals [63–66] were commonly operationalized. Self-monitoring was commonly operationalized through providing a pedometer [63, 64, 66, 67] or log sheet [63, 64, 66]. Few trials reported how they operationalized outcome expectations, with only three that reported social support [63, 66, 67], one that included environment , and one that reported relapse prevention . Four studies included identification of barriers and how to overcome them [48, 61, 63, 66]. Self-efficacy did not appear to be related to PA behavior change [60, 63]. Improvements to diet quality were partially mediated by changes in self-efficacy for fat restriction and eating more fruit and vegetables [62, 63, 78]. Only the social pressure subscale of self-efficacy was significantly related to eating behaviors (P = 0.03) [65, 73]. Two trials reported that self-efficacy was not associated with diet changes  or fruit and vegetable consumption . Social support for healthy eating, perceived barriers to behavior change, and knowledge of recommendations were assessed, but none mediated fruit and vegetable consumption .