Effectiveness of mobile electronic devices in weight loss among overweight and obese populations: a systematic review and meta-analysis

2.1 Data sources and search strategy

This systematic review was performed using a predetermined, unpublished protocol and in accordance with standardized reporting guidelines [15]. Two reviewers (BK and JJ) performed independent searches of the following online electronic databases (Medline, PsycINFO, Embase and CENTRAL). The search of online databases was not restricted by language or date – the search is up to date to May 2014. The search was broken down into four main categories. To identify the relevant population, the first Boolean search was done using the term “OR” to explode [search by subject heading] and map [search by keyword] the following MeSH headings “overweight” or “obese”. To identify relevant interventions the second Boolean search used the term “OR” to explode and map “mobile phone” or “internet” or “computers handheld” or “wireless technology” or “text messaging” or “electronic mail” or “smartphone” or “[iPad or iPhone or iPod touch]” or “mHealth”. The third category of MeSH headings was also related to the intervention and included: “exercise” or “motor activity” or “physical fitness” or “diet”. Finally, the fourth group of key terms was used to identify study design. A Boolean search using the term “or” to explode and map the keywords “controlled clinical trials” or “randomized controlled trials” or “meta-analysis” or placebo*” or “random*” or “groups”. These four search categories were then combined using the Boolean operator “and”. In addition, two individuals (BK and JJ) searched the reference lists of prior review papers and all identified research articles were hand searched. Clinical trial registries were also consulted to identify all ongoing trials (www.clinicaltrials.gov, www.controlled-trials.com/mrct, www.isrctn.com). Table of contents of key journals [Telemedicine Journal and E-Health, Health Informatics Journal and Journal of Medical Internet Research] were also hand searched. Finally, experts identified during the review process were contacted for clarifications on their published trials.

2.2 Study selection

Two reviewers (BK and JJ) independently evaluated articles for eligibility in a two-stage procedure. In stage one, all identified titles and abstracts were reviewed. In stage two, a full text review was performed on all of the articles that met the predefined inclusion criteria as well as all articles for which there was uncertainty as to eligibility. If either reviewer defined an article as eligible it was included in the full-text review and evaluated independently by both reviewers.

2.3 Inclusion/exclusion criteria

An article was considered for inclusion in the systematic review if it met the following criteria: (i) study population (overweight or obese adults defined as having a BMI of ≥ 25.0 kg/m²); (ii) intervention (use of one or more mobile electronic devices); (iii) comparison (usual care defined as any weight loss intervention that does not use a mobile electronic device); (iv) outcome (change in BMI, weight and waist circumference) and (v) study design (RCT).

For inclusion in this review, trials had to report on original data (i.e. no review articles). Trials that used mobile electronic devices as a co-intervention to another weight loss intervention were included only if the data from the group using the mobile electronic device could be extracted independently and had a respective control. Mobile electronic devices considered for inclusion included, smartphones, Personal Digital Assistants (PDAs), portable media players, hand held video game consoles and handheld computers. Technologies that were not considered to be mobile electronic devices included: desktop personal computers, notebooks/sub-notebooks/netbooks, pagers, pedometers and landline telephones.

2.4 Data extraction and quality assessment

Two reviewers (BK and JJ) independently extracted data from all trials that satisfied the inclusion criteria. Agreement between reviewers on the inclusion or exclusion of a document was assessed using Cohen’s kappa statistic [16]. Any disagreement in data extraction and/or study inclusion was resolved through discussion between the two reviewers and, when necessary, a third reviewer [PR]. Primary outcome was change in mean body weight. Baseline and post intervention means and standard deviations for mean body weight were extracted from both the intervention and control groups. The authors of potentially eligible trials were contacted when necessary to obtain missing or incomplete data.

A number of study characteristics were also extracted including geographic location, description of study population, primary outcomes, mobile electronic device used, description of intervention and control, results and participant feedback. Other data extracted included sample size, length of intervention in months, mean age, number and percentage of female participants, and features of the weight loss intervention.

Measures of study quality were also extracted including randomization, treatment allocation and concealment, blinding and loss to follow up. Study quality measures were scored independently by each reviewer (BK and JJ) and assessed using the validated 5-point scale described by Jadad [17].

3.1 Study selection

The initial search identified 559 unique citations. Through title and abstract review, we excluded 451 articles (k = 0.86). For the remaining 108 citations, full-text articles were obtained for more detailed evaluation. We excluded 102 articles during this screening phase primarily due to the lack of use of mobile electronic devices and not being an RCT. Six RCTs [k = 0.98] were deemed appropriate for inclusion for the review [20]-[25] and the purpose of meta-analysis (Figure 1).

3.2 Study characteristics

3.3 Study features

Table 3 describes features of the intervention used in each of the six RCTs. All the trials were self-monitored interventions with the goal of reducing calorie intake to promote weight loss. Participants were required to report their weight frequently in four trials [20]-[22],[24]. Counselor [healthcare professional] feedback was provided in only one trial [22], whereas automated feedback was provided in four trials [21],[22],[24],[25]. Three trials [20],[24],[25] used reminder prompts to encourage participants to record meals or weigh themselves. Only one trial [25] provided social and peer support. Four trials [20]-[23] followed a structured program. Almost all of them were individually tailored, expect for one [25], and measured for intervention adherence, except for one [21].

3.4 Quality assessment

3.5 Risk of publication bias

The risk of publication bias of the included trials was assessed through visual inspection of funnel plots and Begg and Mazumdar’s [19] [rank correlation] test for asymmetry. There was no evidence of publication bias with Begg and Mazumdar’s test or with visual inspection of the funnel plots.

3.6 Effect of use of mobile electronic devices on body weight

A total of 632 participants were included in the six RCTs reporting a mean change in body weight. There were 320 participants who used some form of a mobile electronic device and 312 control comparator participants. All trials reported change in body weight and were pooled to assess effect estimates. All the trials demonstrated a weight loss as a result of using a mobile electronic device-assisted intervention (Table 2), but the effect sizes and the contents of the individual interventions varied greatly.

Using a random-effects model, the WMD for the effect of using mobile electronic devices on reduction in body weight was −1.09 kg (95% CI −2.12, −0.05) in the six eligible trials (Figure 2). Heterogeneity was observed in this pooled estimate (I² = 49.60%, p value = 0.077). Stratified analyses were conducted to determine whether duration of the intervention or the device used to deliver the intervention (PDA vs. mobile phone) were associated with differing effect sizes or measures of heterogeneity. When stratified on study duration, there was no appreciable difference between studies that were 6 months or less in duration (WMD = −0.97, 95% CI −2.23, 0.30) or those that were longer than 6 months (WMD = −1.20 kg, 95% CI −3.34, 0.94), with both strata demonstrating non-significant reductions in weight with a moderate degree of heterogeneity (I² = 47.1%, p value = 0.151 for trials ≤ six months and I² = 62.1%, p value = 0.071 for studies > six months) (Figure 3). However, when stratified by the type of mobile electronic device used, PDA or mobile phones, the heterogeneity reduced in the group that used an intervention delivered via mobile phones (I² = 16.40%, p value = 0.309) and studies that used mobile phones were found to have a significant and consistent benefit with respect to weight loss (WMD = −1.78, 95% CI −2.92, −0.63) (Figure 4). One of the trials [25] in the strata using mobile phones was using a weight loss application on a smart phone whereas the remaining three trials used text messaging via mobile phones. Therefore a sensitivity analysis was performed to look at the robustness of the pooled estimate. When the trial using a smart phone application was excluded, the WMD for the effect of using mobile electronic devices on reduction in body weight was −1.05 kg (95% CI −2.20, 0.10), suggesting that the benefit associated with the single study using the mobile application strongly influenced the pooled effect size.

4.1 Strengths and limitations

A number of limitations of the present review should be acknowledged. In terms of study quality, half of the included studies lacked allocation concealment and/or blinding, which may introduce bias in the estimation of the effect of the use of mobile electronic devices. The number of participants was small, ranging from 75 to 174 in the six RCTs included and this may have limited our statistical power. The population of interest for this study was ‘healthy participants’ ruling out the opportunity to address those populations with chronic diseases associated with obesity such diabetes and cardiovascular conditions. The population of interest may also belong to a certain literacy and socioeconomic status because of the use of mobile phone and Internet in the experimental groups. Our overall pooled estimates of effect were associated with a large degree of heterogeneity. However, our stratified analyses were successful in identifying type of mobile electronic device and the duration of intervention as potential sources of heterogeneity in this review.

Despite these limitations, this systematic review and meta-analysis includes the most recently published studies using mobile electronic devices using PRISMA guidelines [15]. While our results suggest a weight loss benefit with mHealth interventions, the specific elements of these interventions that affect healthy behavior change are not fully understood. To advance knowledge and enhance understanding of the benefits of using mobile electronic devices, standardization of evaluative studies in this area is important. Results from this systematic review and meta-analysis should be considered in future trial designs.

4.2 Recommendations for improving the quality and reporting of future trials

This review demonstrates the need for more systematic study design and standardized reporting to enable comparisons across different mHealth interventions. Associations between study features such as providing counseling by a healthcare professional, automated feedback, measuring adherence to use of intervention, prompting reminder to log food, providing financial incentives to participants, tailored diet and exercise prescriptions and weight loss should be studied further. Understanding these associations will enable efficient distribution of resources during trials. For example, it would be financially beneficial to understand if an expensive resource such as counseling by a healthcare professional is necessary for the trial or an automated feedback in the context of a tailored intervention, is equally beneficial. Tailored interventions have previously shown to be an appropriate model for weight management according to principles suggested by Social Cognitive Theory and the Social Marketing Model [36].