Effect of Behavioral Weight Management Interventions Using Lifestyle mHealth Self-Monitoring on Weight Loss: A Systematic Review and Meta-Analysis

Alongside an increase in obesity, society is experiencing the development of substantial technological advances. Interventions that are easily scalable, such as lifestyle (including diet and physical activity) mobile health (mHealth) self-monitoring, may be highly valuable in the prevention and treatment of excess weight. Thus, the aims of this systematic review and meta-analysis were to estimate the following: (i) the effect of behavioral weight management interventions using lifestyle mHealth self-monitoring on weight loss and (ii) the adherence to behavioral weight management interventions using lifestyle mHealth self-monitoring. MEDLINE via PubMed, EMBASE, the Cochrane Central Register of Controlled Trials and the Web of Science databases were systematically searched. The DerSimonian and Laird method was used to estimate the effect of and adherence to behavioral weight management interventions using lifestyle mHealth self-monitoring on weight loss. Twenty studies were included in the systematic review and meta-analysis, yielding a moderate decrease in weight and higher adherence to intervention of behavioral weight management interventions using lifestyle mHealth self-monitoring, which was greater than other interventions. Subgroup analyses showed that smartphones were the most effective mHealth approach to achieve weight management and the effect of behavioral weight management interventions using lifestyle mHealth self-monitoring was more pronounced when compared to usual care and in the short-term (less than six months). Furthermore, behavioral weight management interventions using lifestyle mHealth self-monitoring showed a higher adherence than: (i) recording on paper at any time and (ii) any other intervention at six and twelve months.


Introduction
Overweight and obesity are the fifth highest risk factors for global death, which corresponds to about 3.4 million deaths yearly, making them a global public health priority and a health challenge [1]. According to the World Health Organization (WHO), the obesity rate has tripled since the 1980s. In 2014, 39% of the adult population was classified as overweight and 13% as obese [2,3], with an 2.1. Search Strategy MEDLINE (via PubMed), EMBASE, the Cochrane Central Register of Controlled Trials and the Web of Science databases were systematically searched, from their inception until March 2020. We searched for experimental studies comparing the effects of behavioral weight management interventions using lifestyle (diet and physical activity) mHealth self-monitoring on weight loss. The search strategy for the MEDLINE database is displayed in Appendix A Table A1. To complete the systematic literature search, we examined the references of the eligible articles.

Study Selection
The included studies had to meet the following inclusion criteria: (i) participants-general population; (ii) design-randomized controlled trials (RCTs), non-randomized controlled trials (non-RCTs) and pilot studies; (iii) type of interventions-studies comparing the effect of lifestyle (diet and physical activity) mHealth self-monitoring (i.e., personal digital assistants (PDAs), smartphones or web-based); and (iv) outcomes-weight change and adherence to behavioral weight management interventions using lifestyle mHealth self-monitoring. The criteria for the exclusion of studies were as follows: (i) non-eligible publication types, such as review articles, editorials, comments, guidelines or case-reports; and (ii) duplicate reports-when this was the case, we extracted the data from the different reports and included in this systematic review the one providing the most detailed data.

Data Extraction and Risk of Bias Assessment
An ad-hoc table summarized the following information from the original reports: (1) year of publication; (2) country; (3) study design; (4) sample characteristics (sample size, age distribution and type of population); (5) baseline means of adiposity parameters (weight, body mass index (BMI) and waist circumference (WC)); (6) type of intervention (PDA, smartphone or web-based); (7) comparison groups; (8) length of intervention; and (9) percentage of dropouts.
The Cochrane Collaboration's tool for assessing the risk of bias (RoB2) [15] was used to assess the risk of bias of the included RCTs. The evaluation of six domains is included in this tool: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome and selection of the reported result. Each domain could be assessed as having a low risk of bias, some concerns or a high risk of bias.
For non-RCTs, the ROBINS-I tool was used [16]. This tool evaluates the risk of bias according to seven domains: bias due to confounding, bias in the selection of participants for the study, bias in the measurement of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in the measurement of outcomes and bias in the selection of the reported result. Overall bias could be considered as "low risk of bias" if all domains were classified as "low risk", "moderate risk of bias" if all domains were classified as "low risk" or "moderate risk", "serious risk of bias" if there was at least one domain rated as "serious risk", "critical risk of bias" if there was at least one domain rated as "critical risk" and "no information" if there was no clear indication that the study had a serious or critical risk of bias and there was a lack of information in one or more domains.
The literature search, data extraction and quality assessment were conducted by two independent reviewers (IC-R and RF-R), and a third reviewer (CA-B) was included when inconsistencies remained after discussion. Kappa statistics was calculated to assess the agreement rate between reviewers.

Statistical Analysis and Data Synthesis
To compute the pooled estimate of the ES and its 95% confidence intervals (CIs) for weight change, we used the DerSimonian and Laird method [17]. A standardized mean difference score was calculated, using Cohen's d index as the ES statistic, in which negative ES values indicate a weight loss in favor of behavioral weight management interventions using lifestyle mHealth self-monitoring. Cohen's d values represented the following: (i) weak effects when values were around 0.2, (ii) moderate effects when values were around 0.5, (iii) strong effects when values were around 0.8 and (iv) very strong effects when values were greater than 1.0 [18]. Additionally, a pooled estimate of the mean weight change difference in kg was calculated.
Adherence to behavioral weight management interventions using lifestyle mHealth self-monitoring was calculated as the risk of dropping out of the lifestyle mHealth self-monitoring group versus other interventions or the control group. Relative risk (RR) was used as the risk estimate.
Subgroup analyses were performed based on the type of mHealth intervention (PDA, smartphone or web-based), the type of comparison group (usual care, paper record or wait-list) and the length of the intervention (≤3 months, six months and ≥12 months). Sensitivity analyses were conducted to assess the robustness of the summary estimates and to detect whether any particular study accounted for a large proportion of heterogeneity. Random-effects meta-regressions were used to investigate whether the results were associated with the age of participants and the baseline means of weight, BMI or WC, since these variables may explain the observed heterogeneity.
Finally, the Egger test [19] (p < 0.10 considered as statistically significant [20]) and a visual inspection of the funnel plots were used to assess publication bias. STATA SE software, version 15 (StataCorp, College Station, TX, USA), was used for the statistical analyses.

Risk of Bias
For RCTs, as evaluated by the RoB2 tool, 47.4% of studies showed some concerns regarding the risk of bias and 52.6% showed a high risk for overall bias (mainly as a consequence of a high risk of bias in the measurement of the outcome domain) (see Appendix A Figure A1). Among non-RCTs, as evaluated by the ROBINS-I tool, the risk of bias was scored as moderate in 33.3% of studies and serious in 66.7% (mainly as a consequence of a serious risk of bias in the missing data domain) (see Appendix A Figure A2).

Risk of Bias
For RCTs, as evaluated by the RoB2 tool, 47.4% of studies showed some concerns regarding the risk of bias and 52.6% showed a high risk for overall bias (mainly as a consequence of a high risk of bias in the measurement of the outcome domain) (see Appendix Figure A1). Among non-RCTs, as evaluated by the ROBINS-I tool, the risk of bias was scored as moderate in 33.3% of studies and serious in 66.7% (mainly as a consequence of a serious risk of bias in the missing data domain) (see Appendix Figure A2).

Subgroup Analyses
Subgroup analyses considering the type of mHealth intervention and the type of comparison group showed that a greater effect was observed when the mHealth intervention used a smartphone  Table 3).

Subgroup Analyses
Subgroup analyses considering the type of mHealth intervention and the type of comparison group showed that a greater effect was observed when the mHealth intervention used a smartphone  Table 3).

Sensitivity Analyses
The pooled ES estimate was not significantly different when data from each individual study were removed from the analyses one at a time.

Meta-Regressions
The random-effects meta-regression models for the effects on weight loss and adherence showed that age (p = 0.365 and 0.462) and baseline means of weight (p = 0.724 and 0.593), BMI (p = 0.440 and 0.979) and WC (p = 0.677 and 0.428) were not related to heterogeneity across studies (see Appendix A Table A2).

Publication Bias
Evidence of publication bias was found in both the funnel plot asymmetry and Egger's test (p = 0.016 for effect on weight loss and p = 0.008 for adherence).

Discussion
This systematic review and meta-analysis provides an overview of the evidence supporting lifestyle (diet and physical activity) mHealth self-monitoring, as part of a behavioral weight management approach, as a suitable intervention for weight management in adults with overweight or obesity, resulting in a moderate decrease in weight and higher adherence to intervention, greater than with other interventions. Additionally, this meta-analysis shows that interventions delivered through smartphones are the most effective mHealth approach to achieve weight management in adult populations with overweight or obesity. The effect of behavioral weight management interventions using lifestyle mHealth self-monitoring interventions was more pronounced when they were compared to usual care and in the short-term (less than six months). Furthermore, behavioral weight management interventions using lifestyle mHealth self-monitoring interventions showed a higher adherence than: (i) paper records at any time and (ii) any other intervention at six months and twelve months.
A previous systematic review on the effect of mHealth applications on weight loss highlighted that the use of these interventions is widely accepted, easy to use and helpful in achieving weight loss goals [12]. Additionally, there is evidence for the consistent and significant positive relationship between lifestyle (diet and physical activity) and behavior (self-monitoring) strategies and successful weight management [5]. Likewise, our systematic review and meta-analysis supports the notion that mHealth self-monitoring interventions have a moderate effect on reducing weight, which may represent a mean weight loss of 1.78 kg greater than with other intervention types.
Our results not only show a positive effect on weight loss but also fewer dropouts of subjects included in mHealth interventions in the short and long-term. The mechanisms through which behavioral weight management interventions using lifestyle mHealth self-monitoring are effective may be explained from two perspectives: the user and the clinical setting [41]. For users, mHealth interventions enhance patients' self-efficacy and empowerment and improve daily life autonomy and adherence to treatment [42]. Moreover, the mHealth approach reduces contact with the clinical setting and, as a consequence, decreases the workload for health care workers (physicians, nurses and nutritionists, especially in primary care) [43,44].
There is a variety of devices that could be used for mHealth. Our subgroup analyses support the notion that smartphones, a technology available to a high proportion of the population worldwide [45], are the most effective mHealth devices for weight management. This, along with the high prevalence rates of both physical inactivity and obesity, has triggered a growing interest in the development of smartphone applications for health, fitness and diet, which have increased exponentially in the last few years [46]. More than half of smartphone users may have downloaded a health application [47]; however, the use of these applications for clinical outcomes is still very limited and even non-existent in most contexts [48].
The limitations of this study are as follows: (1) the risk of bias assessment showed that a few studies presented some concerns or moderate risk of bias, while most showed a high or serious risk of bias. It should be noted that the main reason behind the high risk of bias in the included studies was the impracticality of blinding interventions, but this limitation is difficult to overcome in this type of intervention. (2) The lack of studies using devices that did not allow for comparison between them, with reliable results only being obtained for smartphones. (3) The intervention groups could be very different considering that they used different types of applications; however, they have the common characteristic that they were carried out through an mHealth device. (4) Many of the studies did not control for the effect of other covariates which could affect the results, such as educational or socioeconomic level. (5) Regarding the analysis of adherence, it cannot be assumed that lifestyle mHealth self-monitoring is the only reason for participants to dropout, since these interventions are usually part of a broader behavioral weight management approach. (6) Finally, this meta-analysis showed publication bias, mainly due to the lack of studies with small sample sizes.
Even with the risk of being branded as opportunists, we are not reluctant to emphasize the importance of mHealth interventions in times when face-to-face contact must be limited, such as those we are living in currently, particularly in clinical settings, in which the transmission of infectious diseases, namely Influenza or Sars-CoV-2, may be greater. Additionally, everything appears to indicate that the current concept of treatment will lead to an increase in the use of mHealth in daily clinical practice [49].

Conclusions
In summary, our study demonstrates that lifestyle mHealth self-monitoring interventions, as part of a behavioral weight management approach, are suitable interventions for short-term weight management in adults with overweight/obesity. Considering our results and the population's accessibility to smartphones, this type of device could be a useful and largely scalable tool for weight management. Thus, future well designed RCTs and controlled clinical trials with higher statistical power are essential in order to reinforce the evidence, which is still weak, to demonstrate that effective mHealth interventions could eventually change the current paradigm of lifestyle prescription, increasing patients' self-management of disease, developing new clinical practice guidelines and facilitating workflow in everyday clinical consultations. Funding: This study was funded by FEDER funds.

Conflicts of Interest:
The authors declare no conflict of interest.   Low risk Some concerns High risk Figure A1. Risk of bias for randomized controlled trials (RCTs) using the Cochrane Collaboration's tool for assessing the risk of bias (RoB2).