1Background

Finding effective strategies that promote healthy behavior change is an ongoing challenge [1]. The use of phone based interventions, mobile computing, and communication technology [mobile electronic devices] is rapidly expanding in healthcare and has demonstrated positive outcomes across a variety of populations [2],[3]. Positive results include the creation of new paradigms for evaluation and deployment within the context of health care and health care management. “mHealth” is the overarching term used to describe the practice of medicine and public health, supported by mobile devices [4].

Use of mobile electronic devices has gained popularity in recent years as a tool to facilitate and maintain weight loss among overweight and obese populations. Worldwide, obesity has more than doubled since 1980 [5] and in several developed nations obesity accounts for 2% to 7% of total healthcare costs [6]. Although the majority of mHealth interventions are reported from high-income countries, there is emerging literature on the application of mobile technologies in low-income countries [3].

Mobile electronic devices have the potential to mimic the traditional in-person face-to-face healthcare provider consults providing a cost effective and convenient alternative. Based on the current primary literature there is potential for the use of mobile electronic interventions to facilitate weight loss in overweight and obese populations [4]. While Randomized Controlled Trials (RCTs) have demonstrated the potentially positive impact of mobile interventions, the extent to which mobile electronic devices are more effective than usual care methods is still being debated. The purpose of the present study was to perform a systematic review and meta-analysis of RCTs reporting the use of mobile electronic devices in weight loss efforts among overweight and obese adult population. Previous systematic reviews have been conducted on the topic of mHealth in the treatment of obesity and have highlighted the diverse populations and interventions that have been evaluated. To advance knowledge in this area, we conducted a meta-analysis and try to identify factors that might explain the disparate effect sizes described in these prior reviews [7]-[14].

This literature may still be young, but a systematic review and meta-analysis can help demonstrate early evidence of benefit or risk particularly when the literature is comprised of a small number of trials with limited statistical power.

2Methods

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].

2.4.1 Analysis

The mean difference was calculated by subtracting the mean body weight [in kg] at the end of follow up from the baseline body weight and was compared between groups. This allowed for a comparison of weight lost over and above what was lost in the control group, and not simply a comparison of weight lost in each study. The results were then weighted by sample size and the average taken [weighted mean difference (WMD)]. The WMD in each study was pooled using a random-effects model. Heterogeneity across RCTs was assessed visually by inspecting the I2 statistics. The I2 statistic quantifies the percentage of variability that can be attributed to between-study differences [18].

The results were stratified on study duration (less than or equal to six months and greater than six months) and by the type of mobile electronic device using the WMD for the effect of using mobile electronic devices on reduction in body weight. Finally, publication bias was assessed through visual inspection of funnel plots and Begg and Mazumdar’s [19] (rank correlation) test for asymmetry. All statistical analysis was performed using Stata, version 11.0 (Stata Corp., College Station, TX, USA).

3Results

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).

Figure 1
figure 1

Study flow chart.

Full size image

3.2 Study characteristics

The characteristics and weight outcomes of trials that met the inclusion criteria are shown in Table 1. A summary of the methodological details of the trials is presented in Table 2 and a summary of the study features including information on frequency of intervention usage, feedback, social support and adherence is shown in Table 3. Publication dates ranged from 2008 to 2013 with the number of participants per study varying from 65 to 174 and proportion of women from 65.43% to 84.80%. Mean baseline BMI ranged from 31 to 34.10 kg/m2 while mean age ranged from 38 to 53 years. Weight loss was the primary outcome in all trials; change in body weight (kilograms or pounds) was reported in all trials and two trials [20],[21] reported change waist circumference. The mobile electronic devices used varied across trials, as did the study duration (ranged from 1 to 24 months). Three trials [21],[22],[24] used text messaging via mobile phones as the intervention and were compared to a control group who were given printed material, two trials [20],[23] used PDAs and were compared to control groups who kept a paper diary and one trial [25] used a weight loss application on a smartphone and was compared to a control group who kept a paper diary.

Table 1 Study characteristics
Full size table
Table 2 Characteristics of studies by outcome measure
Full size table
Table 3 Study features
Full size table

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

The quality of trials according to the Jadad score [17] was low to moderate as shown in Table 4. Three trials [21],[22],[24] reported allocation concealment and blinding was reported in two trials [24]. However, all trials reported on intention-to-treat analysis. Common sources of potential bias included research staff not blinded to the treatment groups, and unclear description of randomization. All trials adequately described dropouts, except one [22].

Table 4 Study quality characteristics
Full size table

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.

Figure 2
figure 2

Meta-analysis of standardized change scores in body mass in mobile electronic device users group compared with control. Degree of shading corresponds with study weighting in random-effects model. WMD, weighted mean difference.

Full size image
Figure 3
figure 3

Meta-analysis of standardized change scores in body mass in mobile electronic device users group compared with control, stratified by the study duration. Degree of shading corresponds with study weighting in random-effects model. WMD, weighted mean difference.

Full size image
Figure 4
figure 4

Meta-analysis of standardized change scores in body mass in mobile electronic device users group compared with control, stratified by the type of mobile electronic device used. Degree of shading corresponds with study weighting in random-effects model. WMD, weighted mean difference.

Full size image

4Discussion

This systematic review summarizes the results for the effectiveness of mobile electronic devices in weight loss among overweight and obese populations. Our meta-analysis of these data suggests that interventions using mobile phones were effective at achieving weight loss. Moreover, the data was very consistent, with low statistical heterogeneity suggesting that to date, trials evaluating mobile phone interventions for weight loss show remarkably consistent weight loss benefit even though the content of the interventions varied from study to study. Obesity is a condition that requires lifelong management and monitoring, and interventions with sustained benefits are sorely needed. Whether mHealth interventions result in sustained weight loss is not known, but given the very high prevalence of overweight and obesity and the relative low cost and high accessibility of mHealth tools, the results of this meta-analysis are encouraging and support further investigation and evaluation of the long-term efficacy of mHealth tools in weight loss.

Our results align with those of others that have evaluated healthy behavior change in other populations. The use of mobile phone messaging intervention has been shown to be an effective technique for lifestyle modification to reduce incidence of type 2 diabetes in men, with a BMI of 23 kg/m2 or higher, in India [26]. Even though there was no significant effect of the intervention on BMI, the cumulative incidence of type 2 diabetes was lower in those who received the mobile phone messaging intervention [26]. In addition to diabetes management [27], the use of mobile electronic devices been shown to be an effective means to promote healthy behaviors including smoking cessation [28], preventive medication adherence [29] and asthma management [30]. In this meta-analysis, there are three forms of mHealth interventions being used, mobile phone text messaging, mobile phone applications and PDAs. The feasibility of using text messaging to effect behavior changes that promote healthy behavior has been demonstrated [31] however, more information is needed on what combinations of text message factors (dose, duration, complimentary technologies, etc.) produce the best results. Similarly, there is evidence to support the notion that the use of PDAs outperform paper methods for data collection and led to improved compliance to study protocol and was preferred by the participants over paper [32]. However, an increased accuracy of data entry cannot be assumed, but it is possible to make data entry more accurate by carefully structuring weight loss questions to allow only determinate types of responses and by using prompts to ensure that questions are followed in sequence and are not skipped and date stamped [33]. Our analysis provides little evidence that the use of PDAs in and of themselves is helpful in promoting weight loss.

As the use of mobile technology in physical activity research among overweight and obese population is still fairly new, this review is foundational in informing the development of appropriate and efficient mobile intervention techniques to enhance weight loss in those overweight and obese. Currently, the literature surrounding mHealth and diabetes management [24] seems to be most advanced while the evidence base for other health topics is sparse. The most unique and advantageous feature of mobile electronic devices is their increasing popularity and ease of use. In particular, smart phones and their web-applications [“apps”] provide constant connectivity to store data locally or exchange it via the Internet. Since 2008, application development has continued to grow across multiple platforms at an exponential rate and has mirrored the increase in smart phone users worldwide [34]. While our review identified only a single study that evaluated a mobile phone weight loss application [25], this study did have positive results and should encourage future work in this area.

In addition to improving healthcare outcomes, the use of mobile electronic devices may help reduce disparities [2]. In our current review the majority of included studies originate from the US, however previous systematic reviews have discussed and demonstrated the international applicability of mobile phone technology [2]. Individuals around the world are using mobile technologies to access health services and information and health professionals are formally and informally integrating mobile technologies into public health and clinical activities [35]. Today, the ownership and use of mobile phones has become just as prevalent among persons of low socioeconomic status as those from the general population and therefore, the use of mobile technology has the potential to reduce income-related disparities [2],[3].

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].

5Conclusion

This systematic review and meta-analysis suggests that mobile electronic devices have the potential to facilitate weight loss in overweight and obese populations, but further work is needed to understand if these interventions have sustained benefit and how we can make these mHealth tools most effective on a large scale. As the field of healthcare increasingly utilizes novel mobile technologies, the focus must not be on any one specific device but on the best possible use of these tools to measure and understand behavior. As mobile electronic devices continue to increase in popularity and the associated technology continues to advance, the potential for the use of mobile devices in global healthcare is enormous. More controlled studies with larger sample sizes need to be conducted to look at the cost-effectiveness, technical and financial feasibility of adapting such mHealth interventions in a real clinical setting.