This review aims to summarize recent eHealth research on using smartphones and Web 2.0 technologies for weight management, comparing them with other interventions, and provide recommendations for future research and practice.

The objectives of this systematic review and meta-analysis are to evaluate the overall effectiveness of smartphone and Web 2.0 interventions in weight management. It seeks to identify the key components of these interventions and analyze the characteristics of the target populations, including demographic and socioeconomic factors. It also targets to determine user adherence and engagement levels, and how these factors influence the success of the interventions. Finally, it aims to highlight gaps in the current literature and propose directions for future research.

The protocol for this study was registered at PROSPERO and the registration no. is CRD42024556096. A systemic literature search of two databases was conducted through May, 2024 to pick out studies related to the efficacy of smartphone and Web 2.0, intervention compared to a control intervention in achieving body measurement and physical activity changes; Medicine (via PubMed; National Library of Medicine, Bethesda, MD; started in 1966) and ScienceDirect (Elsevier; started in 1997). Search strategy used for data search included the words (mobile phone OR smartphone OR Web 2.0, technologies) AND (weight management). In addition, reference lists from relevant original research and reviews were also reviewed.

Studies generated from the databases were screened for inclusion criteria. The study selection process is summarized in Figure 1.

All studies with open access that assessed a smartphone or app intervention compared to a control group in randomized control trials, with weight-related body measures (i.e., body weight, BMI, waist circumference) and physical activity changes (steps/day) expressed in terms of mean and standard deviation performed in population of adults and published between the years of 2015–20124 were included.

The exclusion criteria was as follow 1)

no original research i.e., reviews, editorials, or non-research letters

Ethical approval was not required as only published freely accessed data was analyzed.

Data was extracted from articles that met the selection criteria and compiled using a table developed in Microsoft Word. Data such as author, year of publication, sample size, intervention, study outcomes etc. was collected. Study outcomes (change in body weight, BMI, waist circumference and physical activity) were recorded in terms of mean and standard deviation. Two studies (16, 17) provided data in terms of percentage which was then converted into gross values using reverse percentage formulas. Four studies (16, 18–20) were three arm RCTs. In such cases the interventional group involving smartphone or web 2.0, technology exclusively was considered for comparison with the control. The risk of bias was assessed using the Cochrane risk of bias tool- robvis, considering random sequence generation, allocation concealment, blinding of participants and personnel, blinding of output assessment, incomplete outcome data and selective reporting (18). Each was categorized as clearly yes, not sure or clearly no.

For each study, the net effect size was calculated as the change in body weight measures and physical activity outcomes resulting from intervention from baseline till the end of intervention in the intervention group minus the change in the control group identical measures during the same. For studies (16, 17) providing data in percentages following formulas using baseline values were used to calculate gross value in standard measuring units;

Convert the mean percentage change to a decimal:Δ=Δ%÷100Calculate the mean difference in standard units:ΔW=Δ×WIf there was a need to calculate the standard deviation of the mean percentage difference, it is proportional to the baseline standard deviation:ΔSD=Δ×SDIf the mean and SD for change between baseline and the end of the intervention were not reported (19), it was calculated using the following equation (20).

For mean;ΔW=Wpost–WpreFor SD;SD2diff=SD2pre+SD2post−2×ρ×SDpre×SDpostWhere SDpre corresponds to the SD at baseline, SDpost corresponds to the SD at the end of interventions, and ρ is the correlation coefficient for correlations between measurements taken at baseline and at the end of the intervention assuming it to be perfect i.e., 1.

For body weight, BMI and waist circumference, weighted mean differences were estimated using the random effects models. For physical activity outcomes standardized mean differences were estimated using random effects model. Heterogeneity was quantified with 12 statistics, which describes the proportion of total variation in study estimates as a result of heterogeneity (21). The publication bias was assessed by using Egger's test and funnel plots. Statistical analyses were performed using Review Manager software, version 5.4 (The Nordic Cochrane Centre, The Cochrane Collaboration).

The search strategy generated total of 2,629 articles from two different sources. Out of these, 10 articles were included in this meta-analysis (16–20, 24–28). Studies published from 2015 to 2024 were included. All studies included were randomized control trials. The sample size ranges from 30 (19, 22) to 750 (23). The rest of the characteristics of included clinical trials are given in Table 1. Only two studies (19, 22) provided data regarding the change in physical activity in terms of steps/day. For one study the mean change of step/day in intervention group was 7,054.6 whereas in the control group it was 5,002.4. For the second study the intervention group showed a mean change of 736.71 as compared to a control group with mean change of 218.79 in steps/day.

All 10 researches provided data regarding weight change in the experimental and control groups before and after the intervention. The overall mean difference of −2.12 suggests that, on average, the experimental group performs better than the control group. The high I² value (100%) and significant Chi² test (P < 0.00001) indicate substantial heterogeneity among the studies. The overall effect is statistically significant (Z = 3.65, P = 0.0003), indicating a significant difference between the experimental and control groups. The funnel plot indicates a low likelihood of publication bias and suggests that the meta-analysis results are robust. The studies are fairly symmetrical around the overall mean difference, with most points close to the vertical line, reinforcing consistency in the effect estimate. Forest plot and funnel plot are given in Figure 2.

Out of 10, 7 studies provided data regarding waist circumference. The forest plot indicates that the experimental treatment consistently outperforms the control across the included studies, with a combined mean difference of −2.81 95% CI ranging from −4.06 to −1.57. The Z-test for the overall effect is 4.43 (P < 0.00001), showing a statistically significant overall effect favoring the experimental treatment. The results are statistically significant, despite high heterogeneity, suggesting that the experimental treatment is generally more effective, but the effect size varies between studies. The funnel plot indicates a low likelihood of publication bias. Forest and funnel plots are given in Figure 3.

Out of 10, 5 studies provided data regarding BMI. In the forest plot the overall mean difference suggests a small reduction in body weight in the experimental group compared to the control group, but the confidence interval includes zero, indicating that this result is not statistically significant at the 95% confidence level.

High heterogeneity (I² = 98%) suggests substantial variability among the study results, which should be explored further to understand the sources of this heterogeneity. The funnel plot indicates a low likelihood of publication bias. Forest and funnel plots are given in Figure 4.

To find out the source of high heterogeneity, we performed sensitivity analysis by excluding one study at a time. Sensitivity analysis done for change in body weight produced a variety of values for pooled WMDs ranging from −160 to −2. 55 with the exclusion of (24) producing maximum deviation from original pooled WMDs whereas no significant change in heterogeneity was observed in any case. For change in waist circumference, the range for pooled MDs was −1.56 to −3.35 with heterogeneity not varying significantly. For change in BMI, the range was −0.31 to −0.96 with no significant change in heterogeneity. Sensitivity analysis shows that excluding individual studies does not significantly change the overall heterogeneity. This reinforces that the observed variability is systematic and not due to outliers.

We also performed subgroup analysis by dividing studies on the basis of study design, sample size, study duration and type of intervention to find out the source of heterogeneity. Details are given in Table 2.

Interventions generally show a significant reduction in body weight and waist circumference, especially in studies with larger sample sizes and those conducted for about 6 months.

For BMI, significant reductions are observed mainly in studies with moderate sample sizes and 6-month duration.

These findings suggest that digital interventions, especially those using more direct communication methods like text messages and emails, can effectively reduce body weight, waist circumference, and BMI over certain periods and sample sizes.

The study design (whether the study is an RCT or a three-arm RCT) and the duration of the study (especially around 6 months) seem to be the primary factors contributing to the variation in the meta-analysis results. These factors likely influence the effectiveness and consistency of the interventions being analyzed.

Different study designs might involve variations in control conditions or additional intervention groups. This can lead to differences in observed effect sizes because the comparison conditions are not uniform across studies. The effectiveness of the intervention might be context-dependent. Differences in how interventions are implemented can cause variation in results.

The length of the study can influence how long participants adhere to the intervention, as well as how sustainable the intervention's effects are. Shorter studies might show initial benefits that don't persist, while longer studies might reveal whether these benefits are maintained or diminish over time.

Future research should consider these sources of heterogeneity, potentially focusing on more standardized study designs and appropriate durations to reduce variability and increase the reliability of findings.

The risk of bias assessment for each Cochrane item and each included study in shown in Figure 5. The table indicates that while many studies have strong methodologies in several areas, allocation concealment is a recurring issue. This could affect the overall reliability and validity of the findings in the meta-analysis. Addressing these biases in future research would improve the quality and robustness of evidence. Details are given in Figure 5.

This study's limitations include high heterogeneity among the included trials, which indicates variability in study designs and intervention protocols that could affect the consistency of the results. Additionally, the reliance on self-reported data for physical activity and body measurements may introduce reporting biases. The exclusion of non-English studies and the limited number of trials available for certain outcomes may also affect the generalizability of the findings. Lastly, potential technological issues and variability in user engagement with digital interventions were not thoroughly explored.