The study protocol was registered in the UMIN Clinical Trials Registry (ID UMIN000052290) and was published elsewhere [23]. This systematic review adhered to the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) statement [24]. The PRISMA 2020 checklist is provided in Checklist 1. In addition, the search strategy and reporting follow the PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses literature search extension) [25] to ensure full transparency and reproducibility of search methods (Checklist 2).

As described in the protocol [23], the first systematic search of articles was conducted on September 23, 2023, to retrieve articles published from inception until that date. MEDLINE (via PubMed), Web of Science Core Collection (via Web of Science), the Cochrane Library (including Cochrane Central Register of Controlled Trials [CENTRAL] and Cochrane Database of Systematic Reviews), and Ichushi-Web (Japan Medical Abstracts Society) were searched.

ClinicalTrials.gov and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) were reviewed via the CENTRAL. Proceedings papers indexed in Web of Science Core Collection and conference proceedings included in CENTRAL were reviewed. Study registries and conference proceedings were reviewed to determine whether full reports of eligible studies had been published. No published or validated search filters were used; the search strategies were developed specifically for this review. The search strategy was adapted from the protocol of a prior scoping review of digital health tools [26] and adjusted appropriately for the current systematic review. The search strategy was a combination of search strings that covered the areas of (1) mHealth technology, (2) physical activity and sedentary behavior, (3) workers (employees), and (4) work-related outcomes. The search strings were first developed within PubMed and then modified for use in the other databases. Consistency across databases was maintained where possible. Terms were searched within multiple fields where available (eg, MeSH [Medical Subject Headings], keywords, title, and abstract). The overall search terms were developed by 2 researchers (TM and KN) and medical librarians at Tokyo Women’s Medical University, who reviewed them to confirm that the required search terms were included and that the overall search approach was appropriate. No date or study design restrictions were applied in the database searches; studies were retrieved from database inception to the search dates. Language restrictions were not applied at the search stage, although studies written in languages other than English and Japanese were excluded during the screening process. In the Japan Medical Abstracts Society database, conference proceedings were excluded to focus on full-text original articles. Snowball searching was performed using the included studies to identify additional relevant literature. The full search strategies for all databases, exactly as executed, are provided in Multimedia Appendix 1. The searches were updated by rerunning the original strategies in each database using the same terms and procedures. All databases were last searched on September 12, 2025. To identify additional studies not captured by electronic database searches, we conducted backward citation searching by hand-screening the reference lists of included studies and relevant reviews. No forward citation searching (eg, using citation indexes or alerts) was performed. No additional information sources or search methods were used beyond the electronic database searches and backward citation searching described above.

Population, Intervention, Comparison, and Outcome (PICO) of studies in this systematic review were defined as follows: (P) inclusion of all workers (employees) aged 18 years or older regardless of race and ethnicity; (I) interventions using an mHealth intervention aimed at addressing one or both of the lifestyle factors of physical activity and sedentary behavior as a predominant component; (C) any type of control or comparator condition for comparison (eg, usual care, waitlist group, active control group, or pre- and post-data); and (O) work-associated outcomes (eg, absenteeism, presenteeism, work productivity, work performance, and workability). Health-related behaviors such as physical activity and sedentary behavior were also reported, although they were not definite inclusion criteria. All behavioral measurement types and units were accepted (eg, self-report and steps per day, and daily or weekly moderate-to-vigorous physical activity time).

Studies were included based on the following criteria:

Studies were excluded based on the following criteria:

Records identified from the searches were managed in Microsoft Excel. Duplicate records were identified and removed manually by TM through comparison of titles and bibliographic information (eg, DOI). After duplicate removal, 2 researchers (TM and either NH, DN, or SY) independently performed initial screening by reading the titles and abstracts, then obtained the full text of retained studies for independent full-text screening. At the first screening, the kappa statistic of each pair of researchers was 0.77. The corresponding value at the full-text review stage was 0.76. Inconsistent assessments were resolved through discussion among the authors. The reasons for excluding particular studies are recorded in Multimedia Appendix 2. During the screening and selection process, TM contacted the corresponding authors of potentially eligible studies to clarify unclear information, request additional data when necessary, and confirm eligibility for inclusion in the review.

Information from each of the included studies was extracted by 2 researchers (TM and either NH, DN, or SY) using a standardized data extraction form. The data to be extracted were predetermined, and their details were described in the protocol [23]. Briefly, information included author, year of publication, country, study design, participant information, intervention details, control or comparison group, work-related outcomes, summary results for each study, adverse events, adherence to the intervention, and source of funding. After extraction, information was confirmed through discussion among all authors until consensus in data collection was reached. If the studies did not include this information and/or contained unclear information, TM contacted the corresponding authors for clarification. All data extracted from each individual study included in the results of this systematic review are provided in a data extraction table (extracted data in Multimedia Appendix 3).

Bias risk for RCTs was assessed using the Revised Cochrane Collaboration Risk-of-Bias tool (RoB 2) [35], which assesses five risk domains, namely (1) randomization process; (2) deviations from the intended intervention; (3) missing outcome data; (4) outcome measurement; and (5) selection of reported results. For cluster RCTs, we assessed the 5 risk domains above and an additional domain for cluster RCTs [35]. Risk of bias (RoB) for NRSIs was assessed using the Cochrane ROBINS-I (Risk Of Bias In Nonrandomized Studies of Interventions) tool [36], which assesses bias based on (1) confounding factors, (2) participant selection, (3) classification of interventions, (4) deviations from intended interventions, (5) missing outcome data, (6) outcome measurement, and (7) selection of reported results. The RoB in each included study was independently evaluated by researchers (NH, DN, and SY) and confirmed by TM. Interrater agreement for RoB assessments was 73.9% (17/23 outcomes). Disagreements were resolved through discussion until a consensus was reached.

Due to substantial heterogeneity among the included studies in participant characteristics, intervention content, work-related outcome measures, and follow-up periods, quantitative synthesis through meta-analysis could not be conducted. Instead, a narrative synthesis approach was used. The certainty in the body of evidence (eg, using the Grading of Recommendations Assessment, Development and Evaluation [GRADE] system) was not formally assessed because the review adopted a narrative synthesis approach. The substantial heterogeneity among the included studies precluded the meaningful application of quantitative certainty grading systems. A narrative synthesis of the findings was conducted in line with the Synthesis Without Meta-Analysis (SWiM) guideline (Multimedia Appendix 4) [37]. We narratively synthesized the direction of results of individual studies using the vote-counting system [38] and visualized them using the modified effect direction plot [38,39]. The revised effect direction plot provides a comprehensive summary of the impact of mHealth interventions on work-related outcomes and health-related outcomes (Table S1 in Multimedia Appendix 3). Studies were ordered first by intervention category based on mHealth tools (mobile apps, web-based components, combinations of mobile apps and wearable activity monitors or trackers) and then chronologically by year of publication within each category. In addition, studies were further categorized based on the primary behavioral target of the intervention (physical activity, sedentary behavior, or both) to explore whether intervention focus influenced work-related outcomes (Table S2 in Multimedia Appendix 3). Heterogeneity was explored by comparing the direction of effects across these subgroups. Patterns of consistency or variation in effect direction were examined narratively and visually using the modified effect direction plot. All eligible studies, regardless of RoB or study design, contributed to the vote-counting synthesis. However, the overall conclusions were based on a comprehensive assessment, including methodological quality (RoB) and heterogeneity across studies. Following recommendations in the Cochrane Handbook for Systematic Reviews of Interventions, vote counting based on direction of effect was used for the synthesis of results [38,40]. In line with the guidance [38,40], we did not take statistical significance or magnitude of effect into account within this synthesis. We did not consider the statistical significance of effects as attention to this feature may lead to the exclusion of underpowered studies [40]. Owing to the highly heterogeneous study characteristics (such as the lack of a consistent effect measure), we relied on the direction of effects as a standardized binary metric. Effect estimates were categorized as favorable (favoring the intervention group) or unfavorable (favoring the control or comparison group) based on the direction of effect. For work-related outcomes, increases in work productivity, work performance, and workability were considered favorable, whereas decreases in absenteeism and presenteeism were also considered favorable. For health-related outcomes, increases in physical activity were considered favorable, whereas decreases in sedentary behavior were considered favorable. To implement vote counting based on the direction of effect, we followed the study-level step-by-step procedures detailed by Boon and Thomson [38]:

The PRISMA 2020 flow diagram [24] of the study selection process is shown in Figure 1. The search of the 4 databases identified a total of 4018 records. After removing 391 duplicates, 3627 records were included in the screening on the basis of title and abstract, during which 3345 records were excluded, and 282 records were selected to proceed to full-text screening. Subsequently, we excluded 269 studies that did not meet the eligibility criteria for participants (n=121), intervention (n=78), outcomes (n=63), or publication type (n=7). Four additional studies identified through sources other than the database search met the eligibility criteria. Finally, 17 studies [41-57] were included in this review.

An overview of the study characteristics is given in Table 1. An extended version of the table is available in Multimedia Appendix 3 (extracted data). Out of the 17 included studies [41-57], 12 (70.6%) [41-52] were RCTs (including 4 cluster RCTs [49-52]) and 5 (29.4%) used NRSIs [53-57]. These NRSIs consisted of 3 retrospective cohort studies [53-55], 1 prospective cohort study [57], and 1 single-group, pre-post, mixed methods pilot study [56]. All included studies except one [44] (16/17, 94.1%) were conducted from 2016 onward [41-43,45-57]. Out of the 17 included studies [41-57], 7 (41.2%) were conducted in European countries: 2 each in England [52,56], Spain [46,51], and Finland [44,48], and 1 in Germany [42]. Of the remaining studies, 5 (29.4%) were conducted in North America: 3 in the United States [54,55,57], 2 in Canada [43,45], and 3 (17.6%) were conducted in Asia: 2 in Japan [41,53] and 1 in Thailand [49]. Other countries included Australia [50] and Brazil [47].

Total sample sizes ranged from <100 (4/17, 23.5%) [41,43,46,47], 101‐300 (8/17, 47.1%) [45,48-52,55,56], 301‐1000 (2/17, 11.8%) [42,44], and >1000 (3/17, 17.6%) [53,54,57]. Ten studies (10/17, 58.8%) focused on apparently healthy workers; the majority recruited employees who were identified as office workers [43,44,49-52], while one study each was performed in health care workers [45], police officers and staff [56], workers registered with an internet survey company [53], and military workers [48]. The other 7 studies (7/17, 41.2%) targeted workers with certain health conditions, namely metabolic syndrome [42]; off-work and undergoing rehabilitation for a wrist, hand, and/or finger injury [46]; chronic low back pain [41,47]; knee or low back pain [54]; prediabetes [55]; and pelvic floor disorders [57]. Out of the 17 included studies [41-57], 8 studies (47.1%) had a higher proportion of female participants (≥60%) [43-45,49,51,52,55,57]. The mean age of participants in the included studies ranged from 38.6 (SD 10.4) to 54.3 (SD 6.5) years. Most included studies (11/17, 64.7%) [41-46,48,51-54] reported mean participant ages between 40 and 49 years.

We defined passive controls as those comprising no intervention, usual practice, or a waitlist, and active controls as comprising components intended to induce changes in one or more outcomes. Of the 17 included studies [41-57], 8 (47.1%) used passive controls [41-45,48,52,53], 4 (23.5%) used active controls [46,47,49,51], and 5 (29.4%) used pre-post comparison within the same group (pre-post design) for target work-related outcomes [50,54-57].

The included studies were evaluated for work-related outcomes such as absenteeism, presenteeism, work productivity, work performance, and workability. The most common work-related outcome was work productivity in 7 studies [41,44,49,51,54,56,57], followed in order by absenteeism in 5 [44-46,52,55], work performance in 4 [43,50,52,53], workability in 3 [42,47,48], and presenteeism in 2 [51,52]. Only 4 work-related outcomes were assessed as primary outcomes [41,44,46]; the majority of work-related outcomes were measured as secondary [41-43,45,47,48,50-57] or tertiary outcomes [49]. No overall certainty assessment for each work-related outcome was conducted due to the narrative synthesis approach.

Results summarizing the impact of mHealth interventions on work-related outcomes and health-related outcomes are presented in Table S1 in Multimedia Appendix 3. Out of the 7 studies [41,44,49,51,54,56,57], 5 studies (71.4%) [41,49,51,54,57] reported favorable effect directions with work productivity outcomes in the intervention group compared with the control or comparison groups. Of the 5 studies that showed favorable effect directions [41,49,51,54,57], 3 [41,49,51] were RCTs (including 2 cluster RCTs [49,51]) and 2 [54,57] were NRSIs. Among these studies, 3 targeted physical activity [41,54,57], whereas 2 targeted both physical activity and sedentary behavior [49,51]. Regarding the mHealth components, one study [41] used mobile apps as part of the intervention, whereas another [51] conducted a web-based mHealth intervention. The remaining 3 studies [49,54,57] implemented mHealth tools combining mobile apps with wearable activity monitors or activity trackers. The remaining 2 studies (2/7, 28.6%) [44,56] showed no clear direction of effect for this outcome.

Out of the 5 included studies [44-46,52,55] that evaluated absenteeism outcomes, 3 (60%)—2 RCTs [45,46] and a retrospective cohort study [55] targeting physical activity—demonstrated favorable effect directions with absenteeism outcomes in the intervention group compared with the control or comparison groups. Of the 3 studies that showed favorable effect directions [45,46,55], 2 used mobile apps as the mHealth tools for the intervention [45,46], whereas one used mobile apps and wearable activity trackers as mHealth technologies as part of the intervention [55]. The remaining 2 studies (2/5, 40%) [44,52] showed no clear direction of effect for this outcome.

Out of the 4 studies [43,50,52,53], 2 studies (50%)—an RCT [43] and a cluster RCT [52] targeting sedentary behavior—reported favorable effect directions on work performance outcomes in the intervention group compared with the control group. These studies used either mobile apps [52] or web-based components [43] as mHealth technologies in their interventions. However, one retrospective cohort study (25%) [53] that used a mobile app showed a slightly unfavorable effect direction for work performance outcomes. The remaining study (1/4, 25%) [50] showed no clear direction of effect for this outcome.

Out of the 3 RCTs targeting physical activity [42,47,48], 2 studies (66.7%) [42,48] reported favorable effect directions on workability outcomes in the intervention group compared with the control group. These 2 studies used mobile apps and wearable activity monitors as mHealth technologies as part of the intervention [42,48]. The remaining study (1/3, 33.3%) [47] showed no clear direction of effect for this outcome.

Both 2 included cluster RCTs (2/2, 100%) [51,52] demonstrated favorable effect directions on presenteeism outcomes in the intervention group compared with the control group. Among these, one study [52] targeted sedentary behavior, while the other [51] targeted both physical activity and sedentary behavior. Regarding the mHealth components, one [52] used mobile apps as part of the intervention, whereas the other [51] conducted a web-based mHealth intervention.

Ten included studies reported results related to physical activity [42-44,48-52,55,56] (Table S1 in Multimedia Appendix 3). Among these, 8 studies (80%) [42,43,49-52,55,56] demonstrated favorable effect directions on physical activity–related outcomes, whereas one study (10%) [48] showed no clear effect direction and another (10%) [44] showed an unfavorable effect direction. The majority of studies with favorable effect directions on physical activity–related outcomes (5/8, 62.5%) [42,49,50,55,56] used mHealth interventions combining mobile apps with wearable activity monitors or trackers. Of the remaining 3 studies, 1 [52] used mobile apps and 2 [43,51] used web-based components as mHealth tools for the intervention. Regarding sedentary behavior, 7 included studies reported relevant outcomes [43,48-52,56]. Of these, 6 studies (85.7%) [43,48,49,51,52,56] demonstrated favorable effect directions for sedentary behavior outcomes, whereas one study (14.3%) [50] showed no clear effect. Of these 6 studies [43,48,49,51,52,56], 1 [52] used mobile apps, 2 [43,51] used web-based components, and 3 [48,49,56] used combinations of mobile apps and wearable activity monitors or trackers as mHealth tools in the intervention. Of the 9 studies showing favorable effect directions on physical activity and/or sedentary behavior–related outcomes [42,43,48-52,55,56], all except one [51] implemented blended intervention strategies that included intervention components beyond mHealth technologies or tools.

Our RoB assessment for the included studies is shown in Figures 2-4. RoB was evaluated for each work-related outcome. In cases where a single study reported multiple work-related outcomes (eg, work absence evaluated using both organizational records and a self-reported questionnaire), the RoB for each outcome is presented individually. In the 12 RCT studies [41-52], 18 work-related outcomes were identified (work productivity [n=5], absenteeism [n=5], work performance [n=3], workability [n=3], and presenteeism [n=2]). Regarding the overall RoB in the included RCT studies (Figures 2 and 3), 15 of 18 outcomes were rated as having a high RoB [41-45,47-52] and 3 as raising some concerns [44,46,52]. Of the 18 work-related outcomes, 14 were evaluated as either secondary [41-43,45,47,48,50-52] or tertiary outcomes [49], and only 4 were assessed as primary outcomes [41,44,46]. Consequently, the main domains of high risk of bias in many RCT studies were missing outcome data, outcome measurement, and selection of the reported result. In addition, given the nature of mHealth interventions, the included studies were necessarily conducted as open-label trials, raising concerns about the RoB due to deviations from the intended intervention. Domain 2 of the RoB 2, which assesses bias due to deviations from the intended intervention, was assessed as high risk or of some concerns in all RCT studies [41-52]. The overall risk of bias assessment for the 5 studies with an NRSI design [53-57] is summarized in Figure 4. In these 5 studies [53-57], a total of 5 work-related outcomes were identified (ie, work productivity [n=3], absenteeism [n=1], and work performance [n=1]). Regarding the overall risk of bias in these 5 studies [53-57], 3 outcomes were rated as at critical risk of bias [54-56] and 2 as at serious risk of bias [53,57]. The main domain of critical risk of bias was bias due to confounding.

Of the 17 included studies [41-57], adverse events were reported in 8 studies (47.1%; in one study [49], adverse event information was obtained from the corresponding author) [42,44,47,49,50,54,56,57], whereas no information on adverse events was provided in the remaining 9 studies (52.9%) [41,43,45,46,48,51-53,55]. The most frequent adverse events included skin reactions (including rashes, itching, and mild irritation) associated with the use of activity monitors and trackers, with the majority of adverse events being relatively minor. There were no reports of serious adverse events attributable to the intervention.

This systematic review is, to our knowledge, the first to comprehensively examine the impact of mHealth interventions designed to encourage physical activity and decrease sedentary behavior on a wide range of work-related outcomes, including absenteeism, presenteeism, work productivity, work performance, and workability in working populations. A total of 17 studies (12 RCTs and 5 NRSIs) [41-57] met the eligibility criteria. Of these, 12 studies (70.6%) [41-43,45,46,48,49,51,52,54,55,57] reported favorable effect directions on at least one of the target work-related outcomes relative to control or comparison groups. However, the majority exhibited a high or critical risk of bias, raising concerns about the quality of the research. The considerable heterogeneity and overall low methodological quality of the included studies suggest that confidence in the reported effects of mHealth interventions aimed at promoting physical activity and reducing sedentary behavior on work-related outcomes remains limited at this stage. Considering these issues, it was not possible to reach definitive conclusions about the effectiveness of mHealth interventions on work-related outcomes.

Overall, 12 studies reported favorable effect directions for at least one targeted work-related outcome [41-43,45,46,48,49,51,52,54,55,57], 4 reported no clear effect direction [44,47,50,56], and 1 reported an unfavorable effect direction [53]. One study [53] observed that work performance decreased to a slightly greater extent among Pokémon GO players than among nonplayers, although the difference was not statistically significant. However, the majority of included studies indicated that mHealth intervention strategies targeting physical activity and/or sedentary behavior showed a favorable effect direction on work-related outcomes compared with control or comparison groups. Generally similar findings were observed across mHealth intervention types (ie, physical activity, sedentary behavior, or both). These findings suggest that health promotion efforts that incorporate mHealth components to encourage physical activity and/or decrease sedentary behavior are unlikely to have a substantial negative impact on work-related outcomes and have the potential to positively impact these outcomes. This finding is important as it provides useful evidence for stakeholders faced with decision-making about health promotion efforts incorporating mHealth components which encourage physical activity and decrease sedentary behavior among workers, in consideration of aspects of work-related outcomes.

In addition to the above issues, the use of various questionnaires to assess work-related outcomes such as presenteeism, work productivity, work performance, and workability has made it difficult to compare results across studies [60-62], and thereby hindered the ability to draw clear conclusions. Most of the included studies did not use the same questionnaires, and in some cases, identical questionnaires were used to assess different outcomes. For example, the WHO-HPQ was implemented to measure work productivity, presenteeism, and work performance [19]. Work productivity was classified as the efficiency and effectiveness of employees in performing their work, in consideration of the time (presenteeism) or resources (performance) required to perform a specific task, and the problems that required resolution in achieving their objectives [63]. Accordingly, the work productivity outcomes included both performance and presenteeism variables as a result of the two variables expressing a similar concept, albeit with the use of different tools [63,64]. In the context of workability, three included studies [42,47,48] assessed this outcome, but each used a different measurement tool. The use of different tools to measure these work-related outcomes makes it difficult to compare study results and draw definitive conclusions about the impact of mHealth intervention strategies.

Of the five included studies [44-46,52,55] that evaluated absenteeism outcomes, three studies targeting physical activity (two RCTs [45,46] and one retrospective cohort study [55]) demonstrated a favorable effect direction for absenteeism compared to the control or comparison groups. Absenteeism is defined as a period during which employees are absent from work due to illness or disability and is measured as the number of work hours or days missed [64]. Thus, this outcome can be objectively evaluated with materials such as workplace personnel records, making it generally less heterogeneous and more consistent in evaluation compared to other work-related outcomes. In fact, the risk of bias was less severe in those included studies that objectively evaluated absenteeism (eg, using organizational records) [44,46,52] compared to those that assessed it subjectively (eg, using self-reported measures) [45,52,55]. Among the three studies reporting a favorable direction of effect [45,46,55], only one [46] objectively assessed absenteeism, with the risk of bias assessed as being of some concerns. In contrast, the other two, which relied on subjective assessments, were judged to have a high [45] and a critical [55] risk of bias. In sum, mHealth interventions that promote physical activity may have the potential to decrease absenteeism; however, existing evidence on this issue is limited and raises concerns about bias risk. Therefore, further high-quality research is required before any definitive conclusions can be reached.

Since this systematic review is the first to explore the impact of mHealth interventions targeting physical activity promotion and sedentary behavior reduction with work-related outcomes, our results cannot be directly compared with those of previous systematic reviews. Nevertheless, the results are supported by evidence from prior systematic reviews on related topics, which were not restricted to mHealth-based intervention strategies. Consistent with this review, a prior systematic review investigating the effect of workplace nutrition and physical activity interventions on work-related outcomes suggested their potential effectiveness, and 14 studies showed statistically significant favorable changes in absenteeism (n=7), work performance (n=2), workability (n=3), work productivity (n=1), and both workability as well as work productivity (n=1) [19]. Likewise, another systematic review confirmed that a number of studies have demonstrated the effectiveness of worksite programs based on physical activity in improving work-related outcomes, such as workability (n=5), absenteeism (n=3), and work productivity (n=5) [20]. The consistency between our findings and those of previous studies supports the reliability of the present results and strengthens the notion that health promotion efforts incorporating mHealth technologies to promote physical activity and reduce sedentary behavior may positively impact work-related outcomes. Notably, earlier systematic reviews [19,20] primarily focused on traditional workplace-based interventions and did not consider mHealth technologies as key components of intervention. Implementation barriers commonly reported for workplace health promotion programs include financial constraints, limited personnel or expertise, and insufficient physical space [21]. mHealth technologies may help address these challenges by reducing reliance on on-site resources and enabling scalable, flexible, and context-independent intervention delivery beyond conventional workplace boundaries. The importance of scalability for broad implementation has been emphasized in health promotion research [14]. By synthesizing recent evidence, this review contributes to the existing literature by expanding the scope of intervention settings (eg, not only workplace settings but also home and outdoor contexts) and by emphasizing studies in which mHealth technologies constituted key components of the interventions, thereby enhancing their scalability and potential for broader implementation.

The included studies covered a wide range of mHealth strategies. The most frequently reported mHealth technology was mobile apps, used either alone or in combination with wearable activity monitors or trackers (14/17, 82.4%) [41,42,45-50,52-57]. Even when categorized under the same type of mHealth technology, the content and design of each mHealth intervention varied substantially. In addition, many intervention strategies included not only mHealth components but also other components, such as human support (eg, personal counseling; 13/17, 76.5%) [41-44,46,48-50,52,54-57], which further contributed to heterogeneity across studies. This makes it difficult to attribute the effects of the interventions solely to the mHealth technologies. It has been suggested that adherence to interventions is low when mHealth is used as a standalone methodology [30]. In contrast, personal communication with and integration of health care professionals has been reported to positively influence adherence to mHealth interventions focusing on physical activity [30]. In fact, in one included study using an mHealth-only intervention [45], app use declined over the 12-week intervention, with only 33 participants (33/142, 23%) continuing to use the apps regularly by the end of the trial. In the included studies, no information on adherence—defined as the extent to which users followed the program as designed—was reported for mHealth-only interventions, except for one study [45]. Consequently, adherence could not be compared between mHealth-only and blended interventions, and the details remain unclear. Nevertheless, the potentially low adherence to mHealth-only interventions may partly explain why many studies incorporated additional components rather than relying solely on mHealth technologies, which is consistent with previous literature [30]. Likewise, the NICE committee recommended considering digital and mHealth interventions as options to increase physical activity alongside existing services, while noting that their effectiveness is variable [32]. Taken together, these results indicate that mHealth interventions should be viewed as a complement to existing services, such as human support, rather than a replacement for them, both in future research and in the deployment of mHealth technologies in real-world settings.

This systematic review has several strengths. First, it used a comprehensive search strategy across major electronic databases, facilitating an evidence-based approach to the literature search. Second, it included literature on studies using not only RCT but also NRSI designs. Incorporating non-RCT study designs provides broader evidence to guide the implementation of interventions in real-world settings, particularly in situations where conducting an RCT is difficult or inappropriate. Third, each study included in this review was selected and assessed in a comprehensive manner, and its data were extracted and assessed for quality by 2 researchers acting independently in order to remove potential biases in the review process. Additionally, we used 2 methodological quality assessment tools recommended by the Cochrane Collaboration to assess the risk of bias in the included RCTs and NRSIs.

We also note a number of limitations of this review. First, a major limitation of this review is the substantial heterogeneity across study populations, interventions, and outcomes, which precluded meta-analysis and prevented a reliable assessment of the overall certainty of the evidence using the GRADE approach. Instead, this review relied on a narrative synthesis based on vote counting of effect directions across studies. While this approach can indicate effect direction, it cannot estimate the overall effect size across studies. This limitation may reduce the strength of the conclusions and impede the ability to identify which interventions are most effective for work-related outcomes, and the conditions under which they are most beneficial. Future studies using standardized outcome measures and comparable interventions would enable quantitative synthesis of the evidence through meta-analysis, and contribute to addressing this limitation. Second, the review was limited to papers written in English and Japanese, introducing a degree of language bias. However, we confirmed that there were no studies within the screened literature other than those in Japanese and English that met the inclusion criteria for this systematic review. Additionally, some studies that were potentially both useful and relevant may have been excluded because they were not full-text original articles, such as conference proceedings, and unpublished manuscripts. To minimize this possibility, we reviewed the content of non–full-text original articles identified during screening, such as conference proceedings, trial registry records, and protocol papers, to confirm whether any corresponding full-text original studies relevant to this review had been published. Moreover, only 2 of the included studies [41,47] used the CONSORT-EHEALTH checklist [65] or the reporting guidelines for mHealth interventions [66], which may hinder the successful implementation of effective interventions in real-world settings. Finally, many of the included studies in this review investigated relatively short-term effects, with the duration of the interventions ranging from 6 weeks to 12 months, and the long-term impact on work-related outcomes therefore remains unclear.

What future research is necessary to advance this field beyond the results of this systematic review? First, a key priority is to promote further high-quality research that ensures a low possibility of bias risk. Many of the included studies were evaluated as having a high or critical risk of bias, including not only NRSIs [54-56] but also RCTs [41-45,47-52]. Among the included studies, the majority of work-related outcomes were evaluated as either secondary [41-43,45,47,48,50-57] or tertiary outcomes [49], and only 4 outcomes were evaluated as primary outcomes [41,44,46]. This may have partly contributed to a high risk of bias across the domains of missing outcome data, outcome measurement, and selection of the reported results. A high proportion of missing data may reduce the reliability and generalizability of the findings. In addition, selective reporting of outcomes as well as outcome measurement based on self-reports could lead to an overestimation of intervention effects. Future research that assesses work-related outcomes as a primary outcome may help address these bias-related concerns. From other perspectives, efforts to increase adherence will likely help reduce the risk of bias due to deviation from the intended intervention. The degree to which the study intervention was implemented as intended is one of the most challenging issues in the field of mHealth interventions [31]. Addressing low adherence rates is essential, as low adherence can reduce statistical power and introduce bias, particularly by making a potentially efficacious intervention appear less effective, thereby leading to misleading recommendations in mHealth interventions [31]. While low adherence rates in mHealth interventions have recently been widely recognized, strategies that may address these difficulties are emerging [30,31]. Examples that have been shown to positively influence adherence to mHealth interventions focusing on physical activity include customizable push notifications, gamification, social features (eg, competition, social comparison, and challenges), app personalization and customization, personal communication with, and integration of, health care professionals [30]. Future studies incorporating this emerging evidence on adherence to mHealth tools may help establish effective intervention strategies that encourage physical activity and decrease sedentary behavior, thereby contributing to improvements in work-related outcomes. Finally, in addition to the challenges mentioned above, a major limitation of the current evidence is the substantial heterogeneity among the included studies, which prevents quantitative synthesis through meta-analysis. Addressing such heterogeneity might require innovative approaches such as mega-trials, where various interventions are tested simultaneously in identical populations with uniform measurement periods and consistent outcome assessments [67]. Alternatively, promoting future research that adheres to frameworks such as the CONSORT-EHEALTH checklist [65] and the reporting guidelines for mHealth interventions [66] may contribute to standardization across studies, and thereby help overcome heterogeneity among them. As an additional benefit, compliance with these checklists and guidelines could enhance research quality and, furthermore, ensure the reproducibility of the intervention, which is itself essential for the successful implementation of intervention strategies in real-world settings. It will also be valuable to observe which apps developed in research settings successfully transition to the market, while evaluating their impact and quality using tools such as the Mobile App Rating Scale (MARS) [68]. In one included study [56], a clear association was observed between engagement (defined by usage) and perceived usability and usefulness, suggesting that incorporating user feedback is essential for developing effective interventions in real-world settings.

In conclusion, this systematic review suggests that health promotion initiatives incorporating mHealth components to promote physical activity and reduce sedentary behavior may have a positive impact on work-related outcomes. However, the current evidence is limited and generally of low quality, due to a lack of high-quality studies and substantial heterogeneity in study populations, intervention content, outcome measures, and study designs. To draw more reliable conclusions, future research should include large-scale, high-quality studies with long-term follow-up and objective or validated outcome assessments. This review is the first to comprehensively synthesize evidence on work-related outcomes, extending prior reviews that primarily focused on traditional workplace-based interventions without mHealth components. The findings of this review may provide guidance for future studies, including the potential value of megastudy designs that evaluate multiple interventions within the same population over the same period on the same outcomes. In real-world settings, these findings suggest that mHealth technologies may serve as a complementary strategy for supporting occupational health and productivity.