mHealth Interventions to Address Physical Activity and Sedentary Behavior in Cancer Survivors: A Systematic Review

This review aimed to identify, evaluate, and synthesize the scientific literature on mobile health (mHealth) interventions to promote physical activity (PA) or reduce sedentary behavior (SB) in cancer survivors. We searched six databases from 2000 to 13 April 2020 for controlled and non-controlled trials published in any language. We conducted best evidence syntheses on controlled trials to assess the strength of the evidence. All 31 interventions included in this review measured PA outcomes, with 10 of them also evaluating SB outcomes. Most study participants were adults/older adults with various cancer types. The majority (n = 25) of studies implemented multicomponent interventions, with activity trackers being the most commonly used mHealth technology. There is strong evidence for mHealth interventions, including personal contact components, in increasing moderate-to-vigorous intensity PA among cancer survivors. However, there is inconclusive evidence to support mHealth interventions in increasing total activity and step counts. There is inconclusive evidence on SB potentially due to the limited number of studies. mHealth interventions that include personal contact components are likely more effective in increasing PA than mHealth interventions without such components. Future research should address social factors in mHealth interventions for PA and SB in cancer survivors.


Introduction
In 2020, it was estimated that there were about 19.3 million new cancer cases and almost 10.0 million cancer deaths worldwide [1]. Cancer affects all regions of the world and is a major cause of morbidity and mortality. This is concerning considering that it is predicted that an estimated 28.4 million new cancer cases are expected to occur worldwide in 2040. This is a 47% increase in annual cases from the 19.3 million cases reported in 2020 [1].
Although there is a trend of increasing cancer survival [2], cancer survivors must cope with cancer complications and treatments that impact health and quality of life [3]. For example, 25% and 10% of cancer survivors reported poor physical and mental health, respectively. This stands in contrast to 10% and 6% in adults without cancer [4]. Cancer survivors are also at increased risk of recurrent cancer and other diseases, such as cardiovascular disease, diabetes, and osteoporosis [5]. Increasing physical activity (PA) and reducing sedentary behavior (SB) can improve the health and quality of life of cancer survivors.
The value of PA, defined as any bodily movement produced by skeletal muscles that need energy expenditure [6], on cancer survivors have been reported by numerous studies [7]. Engaging in PA after a cancer diagnosis has been found to be associated with decreased risk of cancer-specific and all-cause mortality among individuals with breast,

Materials and Methods
This systematic review is registered with the prospective international register of systematic reviews PROSPERO network (registration no. CRD42020167694) and followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement for reporting systematic reviews [28].

Study Designs
Eligible studies employed experimental or quasi-experimental designs, including randomized controlled trials (RCT), controlled trials, pre-and-post trials, and crossover trials.

Participants
Studies were eligible for inclusion if participants were cancer survivors of any age (persons who have been diagnosed with cancer, from the time of diagnosis through the remainder of their lives) [29]. Thus, persons of any age with any type of cancer, either living with cancer, undergoing, or those having completed cancer treatment, were included.

Interventions
This review included studies that implemented an intervention that featured an mHealth component to address PA and/or SB in cancer survivors. mHealth components could have been delivered via mobile devices, such as mobile phones, smartphones, tablets, personal digital assistants, mobile apps, short messaging services, or wearable activity trackers. In this review, we did not consider pedometers as an mHealth component due to their non-interactive nature to communicate electronically with mobile devices or the Internet [30].

Comparator(s)
Studies that compared mHealth interventions with any type of control were included in this review. This may be an electronic health (eHealth) intervention (e.g., a web-based app, email, website), a non-eHealth intervention (e.g., face-to-face, pamphlets, brochures), or a no-intervention control. Studies without a comparison group (e.g., a non-controlled study) were also included.

Outcomes
This review included studies that reported changes in terms of PA or SB following mHealth interventions. For PA, this included changes in energy expenditure, step counts, PA level, daily time PA in minutes/hours, PA frequency, and PA intensity. For SB, this included sitting time/day, sedentary breaks, bouts of prolonged sitting, and screen time. Studies that measured PA and SB outcomes objectively (e.g., via accelerometers) or via self-report (e.g., questionnaires) were included.

Search Strategy
The electronic search strategy aimed to locate published scientific studies in any language from 2000 until 13 April 2020. The year 2000 was chosen because a previous bibliometric analysis of studies on e-and mHealth for PA, SB, and diet showed that almost no studies were published before that year [21]. The following six databases were systematically searched: Web of Science, PubMed, Scopus, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), and SPORTDiscus. The search strategy presented in Supplementary Material S1 consists of three main categories, namely (1) "cancer" population, (2) "mHealth" intervention method, and (3) "physical activity" or "sedentary behavior" outcome variable and their synonym keywords in each category. Reference lists of relevant articles were also searched to identify additional articles. Unpublished studies, preprints and gray literature were not included in this review.

Study Selection
Following the search, references were imported into Zotero 5.0.89 (Corporation for Digital Scholarship, Vienna, Virginia 22182, USA), and duplicates were removed. Two independent reviewers (N.M., M.A.) screened titles and abstracts of potentially relevant articles against the inclusion criteria for the review. Disagreements were resolved by a third reviewer (A.M.M.) with vast experience in the research area and systematic review methods. Full texts of potentially relevant papers were retrieved and screened similarly. As only articles published in English were included in full-text screening, no interpretation was conducted for different language articles. The number of articles at each screening stage is shown in Figure 1.
cally searched: Web of Science, PubMed, Scopus, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), and SPORTDiscus. The search strategy presented in Supplementary Material S1 consists of three main categories, namely (1) "cancer" population, (2) "mHealth" intervention method, and (3) "physical activity" or "sedentary behavior" outcome variable and their synonym keywords in each category. Reference lists of relevant articles were also searched to identify additional articles. Unpublished studies, preprints and gray literature were not included in this review.

Study Selection
Following the search, references were imported into Zotero 5.0.89 (Corporation for Digital Scholarship, Vienna, Virginia 22182, USA), and duplicates were removed. Two independent reviewers (N.M., M.A.) screened titles and abstracts of potentially relevant articles against the inclusion criteria for the review. Disagreements were resolved by a third reviewer (A.M.M.) with vast experience in the research area and systematic review methods. Full texts of potentially relevant papers were retrieved and screened similarly. As only articles published in English were included in full-text screening, no interpretation was conducted for different language articles. The number of articles at each screening stage is shown in Figure 1.

Data Extraction
Data were extracted from included studies by one reviewer (N.M.) and checked for completeness and accuracy by a second reviewer (P.A.) using a standardized data extraction form adapted from a checklist presented in the Cochrane Handbook for Systematic Reviews of Interventions [31]. The extracted information of included studies included: author, year of publication, country of study, study aim, study design, sample characteristics, intervention characteristics, comparator information, outcomes measured, and results. Any disagreements that arose were resolved through consultation with a third reviewer (A.M.M.).

Risk of Bias Appraisal
Included studies were critically appraised by two independent reviewers (S.K., N.M.) using the revised Cochrane risk-of-bias tool for randomized trials (RoB 2) [32] and the risk of bias in nonrandomized studies of interventions (ROBINS-I) [33]. RoB 2 includes the following domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of reported results. ROBINS-I contains seven domains, including bias due to confounding, bias in the selection of participants into the study, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in the measurement of outcomes, and bias in the selection of the reported result. Any disagreements that arose between the reviewers were resolved through discussion with a third reviewer (A.M.M.).

Analysis
Study and intervention characteristics are described narratively. We decided against conducting a meta-analysis due to the significant heterogeneity of eligible studies in terms of intervention length, design, comparators, and outcomes. We categorized studies based on study design: controlled and non-controlled trials. We further grouped them according to intervention components: interventions that only used an mHealth component; interventions that used mHealth in addition to non-mHealth components that involved no personal contact (e.g., printed material or a website); interventions that used mHealth in addition to non-mHealth components that involved personal contact (e.g., group meetings, phone consultations).
To establish the effectiveness of interventions based on the above categorization, we conducted best evidence synthesis [34] following guidelines by Fanchini et al. [35] and Kuijer et al. [36], which were adapted from van Tulder et al. [37]. We only considered controlled trials when evaluating the evidence due to their inherently lower risk of bias. We considered the risk of bias of these controlled studies when judging the evidence. For this, the proportion of RCT with low risk of bias and some concerns raised as well as quasi-experimental trials with low risk of bias, were the basis for judging the strength of the evidence. We decided that low-risk of bias quasi-experimental trials are likely comparable to RCT with some risk [33]. We defined the following adapted evidence categories and accompanying criteria from previous guidelines [35][36][37].
(1) Strong evidence: at least 66% (2/3) of controlled trials with low/some risk of bias show effect in the same direction; (2) Moderate evidence: 50% to 65% of controlled trials with low/some risk of bias show effect in the same direction; (3) Limited evidence: less than three low/some concerns risk of bias controlled trials available; (4) Inconclusive evidence: other findings not applicable to strong, moderate, or limited.
For example, inconsistent findings in multiple studies (two of five or 40% low-risk studies show effect in the same direction).

Study Selection
A total of 5971 articles were identified, and after the removal of duplicates, 3917 articles were screened for eligibility, of which 3839 were excluded. Following the first screening of titles and abstracts, 78 articles were eligible for full-text screening. After the full-text screening, 32 articles reporting on 31 interventions were included in this review. The study by Lynch et al. [38] was reported in another article [39]. Thus both will be reported as one study in this review. The PRISMA flowchart (see Figure 1) summarizes the study selection.

Intervention Effects on PA Outcomes in Controlled Trials
Out of eight controlled trials, seven reported significant effects of mHealth interventions with personal contact on MVPA [38,42,44,52,53,56,61]. From these trials, two interventions featuring activity trackers and apps with personal contact reported a significant increase in MVPA daily minutes [38,61] when compared to inactive control groups. Similarly, interventions that combined activity trackers with personal contact showed significant MVPA increases compared to both active [42,53,56] and inactive [52] controls. In an intervention that used activity trackers and text messages in conjunction with personal contact, MVPA improved significantly compared to the inactive control group [44].
Some controlled trials also reported PA outcomes as total PA (n = 6) [42,48,52,53,61,65], total energy expenditure [49,51], or METs [57]. From these studies, only two mHealth interventions showed a significant increase in total activity [52,53]. These interventions employed activity trackers and also had a personal contact component. However, there was no effect on total PA when activity trackers were the only intervention component [49].
Step counts were reported as an outcome in seven controlled trials [42,47,49,53,58,64,69]. Only two of these controlled trials reported significant increases in step counts of interventions. One intervention only featured mHealth components [58], and the other intervention had other non-mHealth components without personal contact [42]. Further details are reported in Table 2, where interventions are categorized by intervention components: mHealth only interventions, multicomponent interventions featuring mHealth and other non-personal contact components, multicomponent interventions featuring mHealth and personal contact components.

Not applicable
No significant between-group changes in outcome Low * (+,+,+,+,+) S : subjective; O : objective; kcal/w: kilocalories per week; METs/w: total metabolic equivalent per week; min/d: minutes per day; min/w: minutes per week; steps/d: step counts per day; steps/w: step counts per week. * Assessed using RoB2.0 (Randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, selection of the reported result); +: low-risk of bias; /: some concerns; -: high-risk of bias. ** Assessed using ROBINS-I (bias due to confounding, bias in the selection of participants, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in the measurement of outcomes, bias in the selection of reported results). +: low-risk of bias; /: moderate risk of bias; ?: serious risk of bias; -: critical risk of bias: NI: No information.

Effects on PA Outcomes in Non-Controlled Trials
In total, six non-controlled trials reported on MVPA, of which two reported a significant effect favoring the mHealth intervention [62,67]. A mHealth only intervention featuring an app to monitor and provide feedback on PA increased MVPA [67]. Another intervention that used activity trackers in conjunction with other non-personal contact intervention components showed significant MVPA improvements [62].
Total PA [50,63] and the total metabolic equivalent [59,60] as outcomes were reported in four non-controlled trials. Out of these, only one intervention that relied solely on mHealth components leads to increased total energy expenditure [60]. This finding is in contrast to a similar study, which employed a similar intervention approach and found no effects on energy expenditure [59].
Step counts were also reported in four noncontrolled trials [43,45,62,67]. Only one of these reported significant effects on weekly step counts following an intervention [62]. Further details are reported in Table 3.  Text messages for promoting behavioral change.

Intervention Effects on SB Outcomes
Ten out of 31 of included studies reported on the effects of mHealth interventions on SB outcomes [38,40,41,44,49,56,[61][62][63]65]. Overall, only four studies reported a significant decrease in SB following an mHealth intervention. Two controlled trials that used activity trackers with apps and personal contact reported decreased daily sedentary time when compared to inactive comparators [38,61]. Reduction in weekly sedentary time was also reported in a non-controlled trial in which an intervention featuring an activity tracker and web-based components was implemented [62]. Finally, one non-controlled study reported a significant reduction in weekly sitting time following an mHealth intervention [63].

Risk of Bias Assessment
Ten out of 16 RCT were considered to have a low risk of bias. Four RCT were judged to have some concerns arising from the randomization process [46], outcome measurement [53,65], and selection of the reported result [44]. A high risk of bias was present in two RCT, which had several concerns due to lack of information in the randomization process [48,64] and selective in reporting the results [64]. The study by Haggerty et al. [48] had a high risk of bias due to a lack of information on subjective measurement of outcome. Of the non-RCT included, there were three quasi-experimental [57,58,66] and 12 pre-post studies [41,43,45,50,54,55,59,60,62,63,67,68]. All the non-RCT had either high-risk of bias [45,54,59,63,66,68] or no information [41,43,50,55,58,60,62,67] except one in which risk was slightly lower [57]. Details of the risk of bias assessment are presented in Supplementary Material S2.

Best Evidence Synthesis
There is strong evidence that mHealth interventions in conjunction with personal contact components can lead to increases in MVPA among cancer survivors based on seven out of eight (87.5%) controlled trials with either low-risk or bias [38,42,52,56,61] or some concerns raised [44,53] reporting positive effects. There is inconclusive evidence on mHealth interventions with personal contact to impact total PA/activity among cancer survivors based on two out of six (33.3%) controlled trials with low risk of bias [52] and some concerns raised [53] reporting positive effects. There is limited evidence on the effectiveness of mHealth interventions with a personal contact for step counts among cancer survivors, as only one out of two controlled trials with low-risk reported an increase in step counts [42]. There is inconclusive evidence that mHealth interventions with personal contact are effective in reducing sedentary time among cancer survivors, as only two out of six (33.3%) low-risk-controlled trials showed significant effects. There is limited evidence on the effects of mHealth only interventions due to a lack of studies. In addition, there is inconclusive evidence on the effects of mHealth interventions without personal contacts components for PA due to inconsistent findings in studies. There were no studies, which examined the effects of mHealth interventions without personal contacts components for SB, and this hindered the evidence synthesis.

Discussion
The aim of this review was to identify, evaluate, and synthesize the scientific literature on mHealth interventions to promote PA or reduce SB in cancer survivors. In brief, 31 studies of mHealth interventions were included and systematically analyzed. All the studies evaluated the effects of mHealth interventions on PA outcomes, and only 10 of these also reported on SB outcomes. To evaluate the effects of mHealth interventions, we conducted the best evidence synthesis for which we only considered controlled trials due to their inherently lower risk of bias. The synthesis revealed that only for interventions that employed mHealth components in conjunction with personal contact components, there was strong evidence of effects on MVPA.
From our best evidence synthesis, it suggests that the implementation of mHealth interventions with personal contact may be effective in increasing MVPA among cancer survivors. This finding is consistent with a previous meta-analysis that reported digital interventions to increase MVPA by approximately 40 min per week among cancer survivors [25]. It is also in line with reviews on eHealth interventions in cancer survivors [24,70]. However, these earlier reviews did not examine the effects of how interventions were implemented (mHealth only, mHealth plus non-personal contact components, mHealth plus personal contact components), and as such, direct comparisons are difficult to make. This mHealth interventions that also employed personal contact components increased MVPA is intriguing because it suggests that in-person sessions, phone calls, or group consultations are to be considered when designing interventions. It is probable that such an approach will yield greater effects on health behaviors as research has shown a link between social support from others and PA in cancer survivors.
Cancer survivors whose social interactions were limited tend to engage in less PA [71]. Cancer survivors have specific support needs [27] that may not be fulfilled by purely digital interventions. Incorporating personal contact elements into mHealth interventions can be done in various ways. Personal contact in this review was defined as any person-to-person contact involving nondigital/traditional modes of communication (i.e., phone calls and face-to-face meetings). Lynch et al. [38] incorporated an in-person goal-setting session and phone call behavioral counseling to facilitate adaptation to and maintenance of PA. It is likely that these interactions provided social support because they allowed participants to receive feedback and encouragement [72]. Social support through direct interaction has repeatedly been shown to positively influence PA in various populations [73]. Other interventions included follow-up discussions [61], over the phone as well as in-person counseling [53], and goal-setting sessions [42], which all fall within the domain of social support.
Researchers that assessed an mHealth intervention and its effects on MVPA in controlled studies were often 12 weeks long. Most reported significant effects of mHealth interventions that incorporated personal contact after this time and for this outcome [38,42,52,53,56,61]. However, one 8-weeks controlled study with a similar intervention mode also showed increased MVPA [44]. Although it is not possible to provide an intervention duration that is optimal for increasing PA, 12 weeks appears to be a reasonable length that may also be feasible to implement. However, it is important to consider that the intervention duration may be less important than engaging participants effectively [74]. Interventions of short duration may be highly effective if they greatly engage participants.
There was inconclusive or limited evidence regarding the other intervention modes (mHealth only and mHealth without personal contact) and other outcomes. As such, it is not clear whether cancer survivors will benefit from mHealth interventions with or without other components in terms of total activity, step counts, and SB. It was surprising that a very limited number of studies evaluated interventions that solely relied on mHealth components. However, findings from only mHealth interventions in three non-controlled trials showed promising improvement in PA outcomes [60,67,68], warranting more research.
Despite a plethora of research on mHealth interventions targeting PA, there is still limited research on interventions targeting SB. There was inconclusive evidence in relation to mHealth interventions with personal contacts on SB. Controlled studies that incorporated this mode of intervention mainly only target PA behavioral changes and showed no effects on SB [44,49]. In contrast, findings from mHealth interventions with [63] and without [62] personal contacts targeting both PA and SB behavioral changes in non-controlled trials showed a promising reduction in sedentary time. However, SB is distinct from physical inactivity [75], whereas sitting time is unassociated with inadequate amounts of MVPA [76]. Further, it is reported that sitting time takes the most proportion of the waking day and displaces light-intensity PA [77]. Thus, a feasible approach to reduce SB among cancer survivors could be conducted in the future by replacing sitting time with light-intensity PA before gradually shifting to MVPA.
Findings from this review may not be generalizable to all cancer survivors because studies were only conducted in high-income countries. This is disappointing and indicates that the inherent potential of mHealth interventions in many lower-income countries has not yet been utilized. Using the available mobile network infrastructure in many countries to support cancer survivors in resource-constrained contexts should be considered. Although there may still be barriers in terms of the accessibility of more advanced personal mobile devices, such as fitness trackers [78], lower-end technology can also be utilized. This digital behavioral intervention can yield positive effects in middle-to low-income countries, as has been reported previously [79].
This review has several strengths. An extensive search strategy in six large databases using broad search terms was conducted. The procedures of this review were in line with the PRISMA guidelines, which strengthens the methodology of the review. We also included all experimental studies, which assessed PA and/or SB outcomes following the breadth of mHealth interventions in cancer survivors. This leads to a more comprehensive overview of mHealth as an intervention to promote PA and reduce SB for cancer survivors. In this review, the findings were grouped into intervention modes (mHealth only, mHealth plus non-personal contact components, mHealth plus personal contact components) and their effect on the outcomes. Limitations of this review are the variety of study designs, including several with a small sample size due to its pilot nature. This systematic review shows the heterogeneity in the methodology and study designs of the selected studies likely become its weakness. These issues make a comparison between selected studies difficult and why a meta-analysis could not be conducted.
In the future, larger studies with higher quality study designs are needed to generate findings that encompass the whole spectrum of movement behaviors and mHealth interventions. The effects of mHealth interventions targeting SB are still unclear. Thus more studies focusing on mHealth interventions reducing SB should be conducted. As only studies from high-income countries were included in this review, it is uncertain if the findings also apply to middle-to low-income countries. Social support when paired with mHealth intervention components has potential for promoting PA among cancer survivors, while effects on SB are still elusive. It is recommended to investigate the cost-effectiveness of mHealth interventions implementation with virtual or non-virtual social aspects in-depth. Finally, large-scale implementations, which consider a thorough cost-effectiveness analysis of mHealth interventions targeting both PA and SB in all ranges of income countries, require attention in future mHealth research involving cancer survivors.

Conclusions
We systematically reviewed the scientific literature on mHealth interventions that aimed to promote PA and/or reduce SB in cancer survivors. mHealth interventions with personal contact appeared to have a positive effect on MVPA among cancer survivors. The evidence for this observation was strong. Further, mixed findings for other PA outcomes like total activity and step counts were observed, while the evidence on SB outcomes was inconclusive due to the lack of studies. More research is needed to establish the optimal mHealth intervention mode for various PA outcomes in cancer survivors. Interventions that aim to reduce SB among cancer survivors are also highly encouraged as cancer survivors may be more likely to engage in changing lower-threshold health behavior. Finally, researchers may focus on the cost-effectiveness of interventions because mHealth interventions may need to incorporate personal contact components, which is more costly.