Efficacy of Interventions That Incorporate Mobile Apps in Facilitating Weight Loss and Health Behavior Change in the Asian Population: Systematic Review and Meta-analysis

Introduction

Background

Asian countries typically have lower rates of obesity than Western countries [1,2]. However, globalization has contributed to rapid increases in Asian obesity rates over the last 10-15 years such that between 20% and 35% of Asian adults are overweight or obese [2,3]. Higher body fat percentage, prominent central adiposity, and possible genetic factors predispose Asians to insulin resistance, type 2 diabetes, and cardiovascular diseases that further aggravate the health care burden [4,5], with direct health care costs estimated to be US $100 billion in Asian countries alone [3]. As part of the global action plan for the prevention and control of noncommunicable diseases, the World Health Organization has recommended a focus on improving lifestyle behaviors, including adopting a healthy diet and increasing physical activity to modify the risk factors for obesity and noncommunicable diseases [6].

With the exponential growth of mobile technology in the past decade, researchers have explored the potential of digital health interventions using mobile apps as a service delivery platform to reach a larger target audience [7,8]. This is particularly promising in Asian countries where smartphone adoption is estimated to reach 84% by 2025 [9,10]. The ubiquity, accessibility, multifunctionality, and scalability of apps for health intervention provide health care professionals and researchers an unprecedented avenue for treating, monitoring, and interacting with patients en masse remotely [11,12]. With technological advancement, the number of health- and fitness-related apps targeted at behavioral change has burgeoned, with at least 325,000 apps available on the commercial market in 2017 [13].

To date, several systematic reviews on smartphone efficacy to improve weight and health have concluded that interventions that incorporate apps show potential in weight management as well as in improving diet, physical activity, and chronic disease outcomes and are acceptable in the short to medium term [7,8,14-17]. However, most of the studies included in these systematic reviews have been focused on Western populations. Given the differences in genetics, culture, lifestyle, health beliefs, and health-seeking behaviors between Asian and White populations [5,18,19], it is important to assess if interventions that incorporate apps are efficacious in achieving weight loss in Asians before considering them as part of a national strategy to combat obesity.

Objective

The aim of this review is to systematically synthesize evidence to address this gap in the literature and provide recommendations for future studies. The primary outcome of this review was absolute or percentage weight loss or other surrogate measures of body fat composition such as BMI or waist circumference. The secondary outcomes included dietary intake, physical activity, self-efficacy with regard to implementing healthy lifestyle behaviors, and user engagement with apps.

Methods

Literature Search

The review protocol was prospectively registered (PROSPERO ID: CRD42020165240), with modifications made over the course of the review (Multimedia Appendix 1), and conducted according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Multimedia Appendix 2) [20]. Systematic searches were conducted in June 2020 across 6 databases: MEDLINE, CINAHL, Embase, PsycINFO, Global Health, and the Cochrane Central Register of Controlled Trials. The search strategy incorporated Medical Subject Headings, keywords, and free-text search terms. The search terms included app*, application*, mobile app*, smartphone, mHealth, weight loss, weight change, body mass index change, and Asian*. A sample search for MEDLINE is detailed in Multimedia Appendix 3. BMC Proceedings, ProQuest Dissertations & Theses, and Google Scholar were searched for conference proceedings, dissertations, and any unpublished gray literature, whereas the ISRCTN registry, ClinicalTrials.gov, and the World Health Organization International Clinical Trials Registry Platform were queried for eligible clinical trials and research. Reference lists of the eligible studies and review articles were also manually searched for additional papers that warranted inclusion. In addition, a filtered search of the Journal of Medical Internet Research and JMIR mHealth and uHealth was conducted to locate papers that were published before the respective journals were indexed in MEDLINE.

Inclusion Criteria and Study Selection

Studies were included if they were randomized controlled trials (RCTs), quasi-randomized trials, or nonrandomized controlled trials (non-RCTs). Interventions with no control group, before-and-after interventions, and observational studies (cohort, case-control, cross-sectional, and ecological) were excluded. To be included in this review, studies needed to use a mobile app either in a single-component (ie, standalone use of apps) or multicomponent (ie, apps as part of an intervention with other components, eg, face-to-face consultation, phone calls, or email reviews) intervention. Given that app engagement typically declines rapidly by the second month [21], this was chosen as the minimum study duration to ensure that the intervention effects of app use and longer-term outcomes could be assessed. Participants had to be aged 18 years or older and of Asian ethnicity. Studies were excluded if participants were reported to have eating disorders or mental health conditions, bariatric surgery, or were within the pregnancy or postpartum period. This review was limited to research published in the English language and from 2008 to date because 2008 was the year in which smartphone apps emerged [22].

Data Extraction

All titles and abstracts of the retrieved records were independently screened by 2 reviewers (SMA and JC) to identify the records that potentially met the inclusion criteria. Relevant full-text articles were retrieved and independently assessed by both reviewers using the complete inclusion and exclusion criteria. Both reviewers independently extracted data from the articles based on a standardized data extraction form, including study characteristics (author, year, country, study design, study aims, sample number, attrition rate, disease group, conflict of interest, and funding), intervention characteristics (intervention type, duration, app type, app features, and cultural adaptations), and predefined outcomes. The level of agreement between the reviewers for the main stages of screening were assessed using the Cohen κ coefficient. Discrepancies were discussed and resolved between the reviewers. Any missing data or further information required was requested by email from the corresponding authors, with a follow-up reminder sent after 2 months.

Study Quality Assessment

The Cochrane Collaboration Risk-of-Bias Tool [23] and the Risk of Bias in Non-Randomized Studies of Interventions Tool [24] were implemented independently by 2 reviewers (SMA and JC) to assess the risk of bias in the RCTs and non-RCTs, respectively. Discrepancies were discussed and resolved through a third reviewer (JJ). Each domain of the RCTs received an evaluation of low, moderate, or high risk, whereas the non-RCTs were judged as having low, moderate, serious, or critical risk.

Outcomes

The primary outcome of this review was absolute (kg) or percentage weight change. Other surrogate measures of body weight, such as BMI (kg/m²) and waist circumference (cm), were also included. The secondary outcomes included app use and changes to lifestyle behaviors, including diet, physical activity, and self-efficacy for implementing healthy behaviors.

Data Analysis

The effect sizes used in the meta-analysis were Hedges g values calculated from the mean differences in outcomes (ie, changes in absolute or percentage weight, BMI, or waist circumference before and after treatment) between the treatment arms. Separate analyses were conducted for the RCTs and non-RCTs; studies without a standardized mean and SD were excluded from the meta-analysis. A unique study identifier was assigned to each intervention-control pair included in the meta-analysis (Multimedia Appendix 4 [25-41]).

Random effects models, which control for heterogeneity between studies, were used to fit the Hedges g scores. To account for within-study dependencies, comparisons were made separately for each outcome, namely, changes in absolute or percentage weight, BMI, and waist circumference. For studies with multiple time points reported, only the final outcome within the active intervention was included to avoid pseudoreplication. Furthermore, the overall data were also divided into 2 subsets comprising single-component (ie, standalone use of apps) and multicomponent (ie, apps as a part of an intervention with other components) studies, respectively, and analyzed separately. Subsequently, subgroup analyses were conducted to analyze the effects of adding an app to usual care (intervention group) compared with usual care alone (control group).

The possible moderating effect of study duration was also tested using moderation analysis, after which the possible differences in intervention outcomes in studies conducted over 3 months or less versus studies lasting for longer than 3 months were assessed post hoc. These were done as a preliminary assessment of the importance of app engagement levels on study outcomes because adherence to app use generally tails off after 2 to 3 months [8,11,42,43]. The post hoc analyses were conducted in lieu of a more robust meta-regression approach given the dearth of quantifiable app engagement data.

Heterogeneity among the studies for each comparison was assessed using the I² statistic, with values of 30% to 60%, 50% to 90%, and 75% to 100% considered to indicate moderate, substantial, and considerable levels of heterogeneity, respectively [44]. Publication bias was also assessed using a funnel plot. All analyses were conducted using the metafor package (2.4.0) in the R statistical environment (R Foundation for Statistical Computing) [45,46].

Results

Study and Sample Characteristics

A total of 3027 electronic records were identified through the search strategy and, after removal of duplicates, 2018 (66.67%) titles were screened. From these 2018 records, 127 (6.29%) full-text articles were retrieved (Figure 1). Of the 127 papers, 21 (16.5%) met all inclusion criteria for the systematic literature review [25-41,47-50]. Of these 21 papers, 17 (81%) were included in the meta-analysis [25-41]. The Cohen κ coefficients for the initial screening stage of titles and abstracts and the full-text screening stage were 0.72 (substantial agreement) and 0.92 (almost perfect agreement), respectively, with 97% level of agreement for each screening stage.

‎View this figure

Intervention characteristics and descriptions of the studies included in this review are summarized in Table 1 and Table 2, respectively. Of the 21 papers, 17 (81%) were RCTs and 4 (19%) were non-RCTs, with study locations in China [25,37-39,48], Hong Kong [34], India [26,31], Japan [35], Singapore [30], South Korea [27-29,32,33,36,40,50], Taiwan [41,49], and the United States [47]. The total number of participants across the 21 studies was 21,173 (RCTs: 4090 and non-RCTs: 17,083), with a mean age of 45.9 (SD 9.84, range 25.8-60.5) years; on average, 45.1% (SD 20.1) of the participants were women. The mean BMI of the population was 27.1 (SD 2.47, range 23.0-30.5) kg/m². Of the 21 studies, 9 (43%) were conducted among patients who were overweight or obese [28,30-35,47,49], 6 (29%) involved mainly patients with diabetes mellitus—4 (67%) [27,36,40,48] on type 2, whereas 2 (33%) [37,38] included a mix of type 1 and 2—and 3 (14%) included participants from the general population [26,39,41], whereas the remaining 3 (14%) individually targeted patients with metabolic abnormalities [50], coronary heart disease [25], or colorectal polyps [29].

The active intervention period ranged from 8 to 52 weeks, with a mean period of 18 weeks; the most common intervention periods were 6 months [25-28,30,32,37,39,49,50], followed by 3 months [29,31,33,35,36,38,40,47]. Of the 21 studies, only 1 study (5%) included a 4-week follow-up outcome measurement after the active intervention [35]. The average attrition rate was 11.5% (SD 8.53, range 0%-28%). Of the 21 studies, 2 (10%) consisted of single-component interventions using a mobile app exclusively [35,39], whereas 19 (90%) were multicomponent interventions incorporating additional components such as face-to-face consultations [25-30,32-34,36-38,47-50], reviews through phone calls or emails [26,27,29,31,32,36,40,48], supporting webpage [41], or financial incentives [33]. Of the 16 studies with face-to-face consultations, 4 (25%) also included tutorials to familiarize users on app use [29,30,36,41].

Across interventions, of the 21 studies, 12 (57%) used a researcher-designed app [27,28,30-34,36,40,41,49,50], whereas 9 (43%) used a commercially available app [25,26,29,35,37-39,47,48]. All studies had either used a culturally adapted app or had an app that was locally developed, except for 2 (10%) studies [29,47] that used a generic app developed in the United States. All studies reported using a multifunctional app, with the most common features being interactivity with health professionals; calorie, activity, or weight tracking; health and lifestyle information; and progress feedback (Multimedia Appendix 5 [25-41,47-50]). Other features such as gamification, which included scoring of app use to gain points for reward redemption and leader board positions, were less commonly reported. The average number of app features was 8.7 (SD 3.36, range 3-15). Comparing between commercially available and researcher-designed apps, the average numbers of app features were comparable at 7.7 (SD 3.15, range 3-12) and 9.5 (SD 3.45, range 5-15), respectively.

Risk-of-Bias Assessments

RCTs and Bias

The risk-of-bias assessments of the included studies are summarized in Tables 3 and 4 for RCTs and Table 5 for non-RCTs, respectively. More than half of the RCTs included in both the systematic review and meta-analysis were rated low risk for selection bias, attrition bias, reporting bias, and other biases. Of the 17 RCTs, 1 (6%) was rated high risk for random sequence generation because of errors in patient randomization [27] and 5 (29%) were rated unclear risk for allocation concealment because of insufficient details reported by the authors [27-29,37,38]. Performance bias was rated high risk for 53% (9/17) of the studies [26-30,32,33,35,38], whereas 29% (5/17) of the studies scored unclear risk [25,31,34,36,37]. Of the 17 studies, detection bias was rated unclear for most, except for 3 (18%) that were rated low risk [25,34,35]. Of the 17 studies, 3 (18%) were rated high risk for attrition bias, with attrition rates between 21% and 24% [27,28,31]; 2 (12%) were rated unclear risk [32,38]; and 1 (6%) was rated high risk for reporting bias because the authors did not report the primary outcome registered in the trial registration [31]. Apart from these bias domains, of the 17 studies, 4 (24%) and 3 (18%) were rated high risk and unclear risk with regard to other biases, respectively.

Non-RCTs and Bias

The overall risk of bias for non-RCTs ranged from moderate to critical risk. All studies were rated low risk for selection bias, classification bias, and attrition bias. The studies were mostly rated moderate risk of bias for the other domains, except for 1 study that was rated serious risk for confounding bias and detection bias, leading to the overall rating of serious bias for the study [40]. A non-RCT that was excluded from the meta-analysis had an overall critical risk score because of critical risk of bias for confounding and serious risk of bias for deviation from the intended intervention [50].

Intervention Efficacy

Of the 21 studies, 4 (19%; 3 RCTs and 1 non-RCT) were excluded from the meta-analysis because they did not provide specific values for weight-related outcomes that could be pooled [47-50]. In Bender et al [47], the authors reported significant weight loss and higher percentage of intervention patients achieving 5% weight loss, whereas the remaining 3 studies did not report any significant results.

Among the 14 selected RCTs for the quantitative analysis, 9 (64%) reported a positive effect for the intervention on weight loss between the treatment arms [26,28-33,35,47]. In the 8 RCTs that did not report significant differences between the treatment groups, 7 (88%) nonetheless found that the intervention arms contributed to greater weight loss compared with the controls [27,34,38,39,41,49,50]. In all, 3 RCTs assessed the proportion of participants achieving clinically significant weight loss (ie, 5% weight loss) and found that between 15% and 44% of the participants in the intervention groups achieved 5% weight loss, whereas 6%-9% of the control participants achieved clinically significant weight loss [30,31,47].

The pooled weighted effect size across 14 RCTs for weight change (Figure 2) was small to moderate (Hedges g=–0.26; 95% CI –0.41 to –0.11; P<.01) with substantial heterogeneity (I²=68.3%), whereas similar effect sizes were also observed for BMI (Hedges g=–0.21; 95% CI –0.42 to –0.01; P=.04; I²=69.9%) and waist circumference (Hedges g=–0.24; 95% CI –0.45 to –0.02; P=.03; I²=65.5%; Figures S1 and S2 of Multimedia Appendix 6). In terms of absolute weight, BMI, and waist circumference reduction, the raw mean difference revealed that the intervention group lost 1.16 kg (95% CI 0.81-1.52), 0.42 kg/m² (95% CI 0.16-0.68), and 1.21 cm (95% CI 0.22-2.21) more than the control group, respectively (Figures S3-S5 of Multimedia Appendix 6). Visual assessment of the funnel plots revealed no obvious asymmetry, suggesting a low risk of publication bias (Multimedia Appendix 7).

‎View this figure

In a separate analysis of non-RCTs, 2 studies were included for each outcome: weight and BMI (Figures S6 and S7 of Multimedia Appendix 6). The effect size for weight change was statistically significant but small, with Hedges g=–0.09 (95% CI –0.13 to –0.05; P<.01; I²=0%). In contrast, the effect size for BMI was not statistically significant (Hedges g=0.06; 95% CI –0.27 to 0.39; P=.74; I²=0%). No analysis was conducted for waist circumference because only 1 study reported this outcome [39]. Overall, the results for the non-RCT meta-analysis should be interpreted with caution because there were very few data points included in the analyses and the data from the studies included were highly variable.

In the subgroup analyses for single-component (standalone app interventions) studies, no meta-analysis was conducted because of a lack of data points to assess the outcomes for RCTs and non-RCTs separately. Tanaka et al [35] reported a statistically significant weight loss between the groups but not He et al [39], although the intervention patients achieved a greater weight loss.

For interventions with the addition of apps to usual care, the effect size for weight (Figure 3) was statistically significant with a small to moderate Hedges g=–0.28 (95% CI –0.47 to –0.09; P<.01; I²=67.6%); however, this was not the case for BMI and waist circumference outcomes (Figures S8 and S9 of Multimedia Appendix 6).

‎View this figure

Moderation analysis suggested that study duration was not a significant moderator of intervention effects on weight, BMI, and waist circumference, with P values of .72, .67, and .69, respectively. All studies included in this meta-analysis had an intervention period of 6 months or less. Post hoc analyses (Figures S10 and S11 of Multimedia Appendix 6) revealed that changes in body weight were significantly different between the intervention and control groups, both in studies with a duration of 3 months or less (Hedges g=–0.28; 95% CI –0.52 to –0.05; P=.02; I²=75%), as well as those between 3 and 6 months (Hedges g=–0.29; 95% CI –0.51 to –0.08; P=.01; I²=72.2%). Conversely, changes in BMI and waist circumference did not differ significantly between the treatment arms in both data subsets (Figures S12-S15 of Multimedia Appendix 6).

Secondary Outcomes

Across all 21 studies, 17 (81%) reported at least one secondary outcome measured through a range of tools and approaches, as described in Multimedia Appendix 8 [25-41,47-50]. In all, 8 out of 11 studies (73%) showed that an app intervention improved dietary outcomes, with 5 studies [26,32,38,48,50] reporting concomitant within-group and between-groups differences, whereas Suen et al [34] and Lee et al [29] reported within-group differences. Physical activity outcomes were mostly reported as changes in exercise frequency, intensity, or duration. A total of 7 out of 11 studies (64%) showed an increase in physical activity level or exercise scores, with significant within-group and between-groups differences reported in 3 [29,38,50] and 6 studies [25,33,38,41,49,50], respectively. Significant improvements in self-efficacy scores were reported in 4 out of 6 studies (67%) [26,38,48,50], whereas 2 studies reported no significant change [27,41].

Among the 9 studies reporting significant weight change, 2 (22%) [26,32] reported significant improvements in diet, whereas 1 (11%) [33] reported significant increment in physical activity. Of the 9 studies reporting significant weight change, 3 (33%) studies did not find a significant improvement in diet [29,33,35], whereas the others did not report dietary outcomes [28,30,31,47]. Similarly, for physical activity, 3 (33%) studies [28,29,32] did not find any significant improvement, whereas the remaining studies did not report this outcome [26,30,31,35,47].

Of the 21 studies, 9 (43%) reported on app use and engagement [25,29,30,33,35,37-39,50]; however, only 6 offered use statistics, albeit by using various measures [25,30,35,37-39]. These statistics included number of daily log-ins [30]; meal, activity, or weight logging [30]; uploading of meal photos [35]; reading of messages [25]; frequency of app use [37]; interactivity with coaches [37,38]; and number of questions asked and cumulative app use scores [39]. Of the 9 studies, 4 (44%) reported that increased app use or adherence was associated with greater weight loss and health outcomes [29,33,35,39], whereas the remaining 5 (56%) did not explore any associations. In addition, of the 9 studies, 2 (22%) found that app use declined over time [30,35], whereas the remaining 7 (78%) did not report app engagement trends.

Discussion

Principal Findings

The interventions that incorporated apps produced a small to moderate effect in reducing weight, BMI, and waist circumference in Asian populations, although substantial heterogeneity was present. It was unclear if single-component standalone app studies were efficacious for weight loss; however, supplementing usual care with an app seemed to be beneficial for enhancing weight loss compared with usual care alone. However, the results may not be representative of long-term studies because of a lack of data. This review also found that app interventions may be beneficial for improving diet and increasing physical activity and self-efficacy for healthy behaviors. In these interventions, apps made for Asian populations were largely culturally adapted and multifunctional, with the most common app features being communication with health professionals and self-monitoring of behaviors and outcomes. Overall, the quality of the studies ranged from low to unclear risk of bias for most domains, apart from performance bias where most of the studies were graded high risk because of the lack of blinding, which is challenging in interventions that incorporate apps. Therefore, the results should be interpreted with caution.

Among the studies in our review that reported a significant difference in either between groups or within group for weight loss favoring the intervention or a greater likelihood among the intervention groups for clinically significant weight loss, most were multicomponent. They typically included face-to-face consultations and reviews through phone calls or emails in addition to the app component.

Our review found that supplementing multicomponent usual care practices with an app was successful in achieving greater results. Reviews of studies in Western populations have observed similar findings [8,15]. This could be attributed to the provision of social support, accountability, and increased opportunities for patients to be reviewed beyond the confines of the app [51], thus underscoring the importance of additional components to raise contact frequency, enhance self-monitoring, and maximize outcomes.

Monitoring of weight, diet, and physical activity behaviors was a common feature of apps in the interventions reviewed. The addition of an app to usual care aligns with the understanding that self-monitoring can improve self-regulation of behaviors and weight loss [52,53]. By enhancing convenience for users, apps thus encourage more consistent self-monitoring [54] to promote treatment adherence and weight loss [15,42]. This review also observed that the outcomes of healthy eating and increased physical activity, which are key determinants of weight loss, tend to occur alongside improved self-efficacy for implementing healthy behaviors. As self-efficacy was previously found to be positively associated with self-monitoring, it is likely that increased self-monitoring may account for the behavioral improvements seen [55].

It was also apparent from our review that multifunctional, all-in-one apps were common among Asian interventions. The features included direct communication with health professionals through the app, in addition to functionalities for calorie, activity, and weight tracking as well as provision of health information and progress feedback, thereby matching the features found in effective digital weight loss interventions reported in a recent review [54].

In contrast with apps designed in Western countries, which tend to focus more on independent learning [8], apps designed for Asians frequently include accessibility to health professionals. It is plausible that having health coaches within apps reduces the barriers for Asians to seek health information, validation, and support from their clinicians [18,19,56], while conferring increased credibility to the coaching and support given to users [54,57]. App users recognize the benefit of health professionals such as dietitians providing support, particularly as they offer effective, evidence-based, culturally appropriate, and tailored dietary counseling to participants [58,59]. Correspondingly, most studies that employed a dietitian or nutrition-trained professional reported a greater weight loss with the intervention [28,30-32,35,37,50].

Furthermore, all but 2 studies [29,47] either used a culturally adapted app or an app that was locally developed; employed the respective country’s native language; and incorporated localized educational content, food databases, and recommendations. Research supports that using culturally appropriate content, engaging local facilitators, and offering the app in the native language are important factors that may promote app use and outcomes [57,60].

Only 6 of the 21 studies reported app use statistics, whereas 2 others provided associations between app adherence and outcomes without reporting actual app engagement data. Meyerowitz-Katz et al [61] reported in a recent meta-analysis that the pooled estimate of app nonuse (defined as attrition rate) was 43%, indicating a serious limitation of app-based interventions if strategies for maintaining long-term engagement with the intervention (longer than a year) are not considered. In this review, the interventions that incorporated an app achieved statistical significance for weight change in studies with durations that were 6 months or less. However, the efficacy of these interventions in long-term studies remains unclear because none of the studies reviewed included durations longer than a year. App engagement levels in this review varied across studies, making comparison difficult. Nevertheless, evidence from this review echoed the results reported by previous reviews that increased app use is associated with greater adherence and weight loss [8,15], notwithstanding the fact that app engagement typically declines over time [8,61].

Strengths and Limitations

This is the first meta-analysis to report on the efficacy of apps incorporated into interventions targeting weight loss with or without healthy behavioral change in populations of Asian ethnicity. The review and meta-analysis were conducted according to best practice and followed PRISMA guidelines with a comprehensive search strategy and assessment of risk of bias using Cochrane Collaboration tools. The study selection, data extraction, and quality assessment were conducted independently by 2 reviewers.

This review is not without some limitations. The heterogeneity observed across studies was substantial, making it challenging to effectively interpret the results. Substantial heterogeneity could be due to the differences in study aims, targeted outcomes, methods, populations, and interventions. The lack of consistent and detailed reporting among the studies limited our ability to assess the true dose of intervention received, user engagement levels, and behavior change techniques that may have been employed in the apps. Therefore, the results of the meta-analysis should be interpreted with caution. Future studies that incorporate apps should consider using a standardized tool such as the Behavior Change Technique Taxonomy to code app features in a systematic and replicable manner and report user engagement statistics to evaluate app use and outcomes. This review was also limited to studies that were published in the English language. This may be problematic, given that we were studying apps in Asian populations, and English may not have been the first language in many countries; hence, some articles in other languages would not have been captured in this review. Finally, as most of the studies were multicomponent in nature, components apart from the app, such as in-person education or review calls, may have more strongly influenced the outcomes; however, it was not possible to identify the contribution of these components to the weight loss outcomes.

Conclusions

This review contributes to the literature by presenting quantitative evidence that multicomponent interventions that incorporate apps produce a small to moderate effect toward weight loss in studies of Asian populations with intervention periods of 6 months or less. It is unclear if single-component standalone-app studies are efficacious for weight loss; however, adding apps to multicomponent usual care confers better outcomes. More evidence is required to determine the efficacy of apps in the long term. Cultural adaptation and offering the app in the native language of the users seem to be a priority in Asian apps. Multifunctional apps with features that include self-monitoring, individualized feedback, and a link to health professionals within the apps may be key to raising awareness and engagement, as well as promoting weight loss. Finally, it is recommended that investigators monitor and address the low uptake of apps and attempt to enhance engagement level before using apps as part of national health strategies for reducing obesity.