Effects of Face-to-Face and eHealth Blended Interventions on Physical Activity, Diet, and Weight-Related Outcomes among Adults: A Systematic Review and Meta-Analysis

An increasing number of studies are blending face-to-face interventions and electronic health (eHealth) interventions to jointly promote physical activity (PA) and diet among people. However, a comprehensive summary of these studies is lacking. This study aimed to synthesize the characteristics of blended interventions and meta-analyze the effectiveness of blended interventions in promoting PA, diet, and weight-related outcomes among adults. Following the PRISMA guidelines, PubMed, SPORTDiscus, PsycINFO, Embase, and Web of Science were systematically searched to identify eligible articles according to a series of inclusion criteria. The search was limited to English language literature and publication dates between January 2002 and July 2022. Effect sizes were calculated as standardized mean difference (SMD) for three intervention outcomes (physical activity, healthy diet, and weight-related). Random effect models were used to calculate the effect sizes. A sensitivity analysis and publication bias tests were conducted. Of the 1561 identified studies, 17 were eligible for the systematic review. Studies varied in participants, intervention characteristics, and outcome measures. A total of 14 studies were included in the meta-analyses. There was evidence of no significant publication bias. The meta-analyses indicated that the blended intervention could lead to a significant increase in walking steps (p < 0.001), total PA level (p = 0.01), and diet quality (p = 0.044), a significant decrease in energy intake (p = 0.004), weight (p < 0.001), BMI (p < 0.001), and waist circumferences (p = 0.008), but had no influence on more moderate-to-vigorous physical activity (MVPA) or fruit and vegetable intake among adults, compared with a control group. The study findings showed that blended interventions achieve preliminary success in promoting PA, diet, and weight-related outcomes among adults. Future studies could improve the blended intervention design to achieve better intervention effectiveness.


Introduction
In today's society, millions of adults live unhealthy lifestyles. Insufficient physical activity (PA) and unhealthy diets are regarded as the dominant form of lifestyle leading to many health problems [1]. Insufficient PA (e.g., at least and accumulated 150 min of moderate PA per week) and unhealthy dietary behaviors (e.g., less than five portions of fruit and vegetable intake per day, high in saturated fats, trans fatty acids) are the leading risk factors for death and life-years lost, while 1.6 million deaths annually can be attributed to insufficient PA [2]. However, risk behaviors commonly co-occur. A previous study reported approximately 50% of co-occurrences of unhealthy diets with physical inactivity among adults [3]. Engaging in multiple risk behaviors can lead to negative effects on health, such as chronic diseases and mortality [4]. An overwhelming body of studies suggest that

Study Inclusion Criteria
The eligibility criteria were defined based on population, type of intervention, comparisons, type of outcomes, and study type (PICOs).

Type of Population
This study targeted adults aged 18 years and above. Both clinical and non-clinical subjects were included. Exclusion criteria included studies with participants who were <18 years old, or under special situations that seriously affected their feeding ability and physical mobility (e.g., physical disability).

Type of Interventions
The blended intervention with the aim of promoting physical activity and dietary behaviors simultaneously was included. There is no consistent definition of blended intervention, and it was applied in various formats [25,26,29,30]. In this study, it was defined as intervention programs that include both face-to-face sessions and elements of eHealth interventions, such as a short message service (SMS), smartphone applications (APP), email, telephone counselling, a website, and a social medium. Within blended interventions, face-to-face physical contacts should be added to eHealth interventions or, vice versa, the eHealth intervention may be applied as an implement to existing faceto-face contacts. These 2 interventions can be performed simultaneously or sequentially. Non-blended interventions are those that only comprise of a stand-alone, face-to-face intervention or an eHealth intervention.

Type of Comparisons
The comparators in this study were defined as control groups (e.g., face-to-face intervention, usual care intervention, no intervention, or eHealth intervention). Eligible studies should compare a blended intervention group to at least 1 control group.

Type of Outcomes
Eligible studies should include both PA outcomes (e.g., energy expenditure, walking steps, time spent on moderate to vigorous PA (MVPA), and diet outcomes (e.g., energy intake, fruit and vegetable intake, diet quality). Both objective and self-report measurements were acceptable. All units were acceptable (e.g., minutes, steps, servings, calories, kilograms, score). In addition to PA and diet outcomes, weight-related outcomes (e.g., body weight, BMI, waist circumference, body fat, waist-to-hip ratio) which are closely related to PA and diet were also included. The studies that did not include PA or diet outcomes simultaneously were deemed ineligible.

Study Type
Both pilot studies and main studies of randomized controlled trials (RCTs) or cluster RCTs were included. All other study designs, such as pure qualitative study and reviews, were not eligible. Publications that were not written in English were excluded.

Search Strategies
Systematic searching was conducted in the following 5 electronic databases: PubMed, SPORTDiscus, PsycINFO, Embase, and Web of Science. The date of publication covered the last 20 years (between 1 January 2002 and 31 July 2022). The search keywords were based on a combination of a thesaurus and a previous study [17], focusing on face-to-face and eHealth blended interventions regarding weight-related behaviors (PA and diet). The specific search terms used in each database are presented in the Multimedia Appendix A.

Study Selection
All identified articles were exported into Endnote 20 for duplicate checking and further screening. After deleting duplications (MY), screening was conducted following 2 steps. First, the titles and abstracts were independently screened by 2 authors (MY and DP). In cases where the 2 authors disagreed over the eligibility of an article, the disagreement was resolved by a 3rd author (YD). Then, full texts were obtained and screened by MY. The procedures guiding article inclusion are presented in the flow chart in Figure 1. The reference lists of eligible articles were further reviewed by MY to identify related studies via hand-searching. Grey literature (e.g., working papers, unpublished studies, conference proceedings or abstracts, dissertations) was not considered eligible in this study.
All identified articles were exported into Endnote 20 for duplicate ther screening. After deleting duplications (MY), screening was cond steps. First, the titles and abstracts were independently screened by 2 DP). In cases where the 2 authors disagreed over the eligibility of an art ment was resolved by a 3rd author (YD). Then, full texts were obtained MY. The procedures guiding article inclusion are presented in the flow The reference lists of eligible articles were further reviewed by MY t studies via hand-searching. Grey literature (e.g., working papers, unp conference proceedings or abstracts, dissertations) was not considere study.

Data Extraction
One author MY extracted data from the included studies. Data ex ducted using the specific framework for this review (Table 1), which author, year of publication, nation, participants characteristics (sample ratio of female, mean age/age range), intervention characteristics (study tion duration, theoretical underpinning, intervention condition, contro come measures, and main findings.

Data Extraction
One author MY extracted data from the included studies. Data extraction was conducted using the specific framework for this review (Table 1), which includes the first author, year of publication, nation, participants characteristics (sample size, type, gender ratio of female, mean age/age range), intervention characteristics (study design, intervention duration, theoretical underpinning, intervention condition, control condition), outcome measures, and main findings. Statistically signficant between-group differences in PA, calorie, carbohydrate, sugar, total fat, and saturated fat intake, weight, and BMI, p < 0.05

Risk of Bias Assessment
The methodological quality of RCTs was assessed using CRIBSHEET (Cochrane risk-ofbias tool for randomized trials (RoB2) SHORT VERSION) [31]. RoB2 evaluated 5 domains, including the randomization process, the effect of assignment to intervention, missing outcome data, measurement of the outcome, and selection of the reported result. Overall quality was considered as low risk ("low risk" in all domains), some concerns (at least 1 domain with "some concern"), and high risk (at least 1 domain with "high risk" or several domains with "some concerns"). The assessment was independently assessed by 2 reviewers (MY and DP). Inconsistencies between the reviewers were resolved through mutual discussion until consensus [17].

Strategy for Meta-Analysis
When at least 3 studies reported the same exposure and sufficient information was available from the studies, the quantitative data were pooled into RevMan 5.4 (The Cochrane Collaboration, 2020) for meta-analysis to identify the between-group effects [32]. All variables quantifying the amount of physical activity, diet, and weight-related outcomes were extracted. Most of the included studies reported the mean change from baseline in each group without the primary data of mean and SD. Hence, the mean change in outcomes from baseline to the post-intervention was extracted from the intervention group and control group separately for further meta-analysis following previous studies [33,34]. The values of the outcome variables (i.e., mean change, standard deviation of mean change (SDs), and sample size in each group) were also extracted.
When the mean change and SDs were not available, the following equations were used to calculate them: Mean change = Mean post-test − Mean pre-test . SDs = √ (SD pre-test ) 2 + (SD post-test ) 2 − (2 × Corr × (SD pre-test ) × (SD post-test )). The correlation coefficient 0.80 was used for imputation of the SDs between both sets of time points [34,35]. Ten meta-analyses were completed for walking, MVPA, total PA, energy intake, fruit intake, vegetable intake, diet quality scores, weight, BMI, and waist circumference.
The random-effect models which allow generalization of inferences beyond the studies included in a particular meta-analysis were applied to calculate pooled mean changes with 95% confidence intervals using the inverse variance approach [36]. The Cochran's Q test and I 2 statistics were used to test the heterogeneity of the included studies [37,38]. An I 2 statistic of 25% is considered low heterogeneity, 50% moderate heterogeneity, and 75% high heterogeneity [39]. For the Q statistic, statistical heterogeneity was set at p < 0.1 [40,41]. The results of the meta-analysis are presented through Forest plots. Effect sizes were calculated as standardized mean difference (SMD) because of the variability in outcome measures [42]. SMD of 0.2, 0.5, and 0.8 are suggested corresponding to small, medium, and large effects [43].
Sensitivity analyses were applied to assess the robustness of included studies when the included studies in a meta-analysis indicated high heterogeneity [44]. The Egger's regression test was conducted for detecting publication bias by using STATA 16.0 (College Station, Texas, USA), the statistical publication bias was set at p < 0.1 [40]. The subgroup analysis was not performed due to the limited number of included studies, i.e., each sub-category was required to contain at least 4 studies [34,45].

The Intervention Effects on PA
To identify the between-group effects of interventions on physical activity, three meta-analyses were conducted, including walking, MVPA, and total PA (see Figure 3). There was no evidence showing significant publication bias (Egger's test, p > 0.1).

The Intervention Effects on PA
To identify the between-group effects of interventions on physical activity, three metaanalyses were conducted, including walking, MVPA, and total PA (see Figure 3). There was no evidence showing significant publication bias (Egger's test, p > 0.1).   Five studies involving 438 participants measured walking counts. The overall effect demonstrated that the blended intervention led to significant increases in step counts (SMD = 0.45, Z = 4.38, 95% CI 0.25 to 0.66, p < 0.0001) when compared with the results of the control group with low heterogeneity (I 2 = 8%).
The synthesized effect size of intervention on MVPA was analyzed by meta-analysis on 6 studies that included 767 participants. The result showed that the blended intervention led to no significant promotion in MVPA (SMD = 0.92, Z = 1.11, 95% CI −0.70 to 2.55, p = 0.27) compared with control groups. The heterogeneity test showed significance among MVPA (I 2 = 99% > 50%). The sensitivity analyses indicated no significant modification in magnitude when individual study data were removed from the analysis one at a time.
A total of 3 studies involving 352 participants measured total PA level. For the analysis of blended interventions versus control, the SMD on total PA level was 0.62 (Z = 2.56, 95% CI 0.15 to 1.10, p = 0.01) with high heterogeneity (I 2 = 76%, p < 0.10). The results showed that the blended intervention led to significant promotion in total PA level. The pooled effect was significantly modified in magnitude (SMD 0.38 Z = 3.05, 95% CI 0.14 to 0.62, p = 0.002) when one study with high heterogeneity was removed from the analysis [55].

The Intervention Effects on Diet
To identify the between-group effect of interventions on dietary behaviors, four metaanalyses were conducted, including energy intake, fruit intake, vegetable intake, and diet quality scores (see Figure 4). There was no evidence showing significant publication bias (Egger's test, p > 0.1).
The synthesized effect size of intervention on energy intake was analyzed by metaanalysis on 10 studies that included 1630 participants. The overall effect demonstrated that the blended intervention led to a significant decrease in energy intake (SMD −0.39, Z = 2.91, 95% CI −0.65 to −0.13, p = 0.004) when compared with the results of a control group. The heterogeneity test showed significance among the included studies (I 2 = 80%, p < 0.1), which was high to elaborate a reliable result. The pooled effect was significantly modified in magnitude (SMD −0.47 Z = 5.64 95% CI −0.64 to 0.31, p < 0.00001) when one study with high heterogeneity was removed from the analysis [60].
Three studies involving 452 participants measured fruit intake. The result showed that the blended intervention led to no significant increase in fruit intake (SMD = 0.45, Z = 1.51, 95% CI −0.14 to 1.04, p = 0.13) compared with the control group. The heterogeneity test showed significance among these three studies (I 2 = 85%, p < 0.01). Sensitivity analyses indicated no significant modification in magnitude when individual study data were removed from the analysis one at a time.
Three studies involving 452 participants measured vegetable intake. The result showed that the blended intervention led to no significant increase in vegetable intake (SMD = 0.59, Z = 1.34, 95% CI −0.27 to 1.44, p = 0.18) compared with the control group. The heterogeneity test showed significance among these three studies (I 2 = 93%, p < 0.01). The pooled effect was significantly modified in magnitude (SMD 1.07, Z = 3.22, 95% CI 0.38 to 1.55, p = 0.003) when one study with high heterogeneity was removed from the analysis [58].
The synthesized effect size of intervention on total diet score (higher score means healthier diet) was analyzed by meta-analysis on five studies that included 262 participants. There was a negligible heterogeneity among the included five studies (I 2 = 0%, p = 0.86 > 0.01). The result demonstrated that the blended intervention led to healthier dietary behavior when compared with the control group (SMD = 0.36, Z = 2.85, 95% CI 0.11 to 0.60, p = 0.004), which suggested a small effect size (SMD > 0.2).

The Intervention Effects on Diet
To identify the between-group effect of interventions on dietary behaviors, four meta-analyses were conducted, including energy intake, fruit intake, vegetable intake, and diet quality scores (see Figure 4). There was no evidence showing significant publication bias (Egger's test, p > 0.1). The synthesized effect size of intervention on energy intake was analyzed by metaanalysis on 10 studies that included 1630 participants. The overall effect demonstrated that the blended intervention led to a significant decrease in energy intake (SMD −0.39, Z

Intervention Effects on Weight-Related Outcomes
To identify the between-group effect of interventions on weight-related outcomes, three meta-analyses were conducted, including weight, BMI, and waist circumference (see Figure 5). There was no evidence showing significant publication bias (Egger's test, p > 0. Eight studies including 913 participants measuring weight were included in metaanalysis. For the analysis of blended interventions versus control, the SMD on weight was −0.42 (Z = 6.26, 95% CI −0.55 to −0.29, p < 0.001) with negligible heterogeneity (I 2 = 0%, p = 0.60). The result suggested that blended intervention can lead to significant weight loss.
The synthesized effect size of intervention on BMI was analyzed on five studies that included 424 participants. The overall effect demonstrated that the blended intervention led to a significant decrease in BMI (SMD −0.68, Z = 6.11, 95% CI −0.90 to −0.46, p < 0.001) when compared with the results of the control group. The heterogeneity test showed no significance among the included studies (I 2 = 13%, p = 0.33).
Five studies involving 620 participants measured waist circumference. The overall effect demonstrated that the blended intervention led to a significant decrease in waist circumference (SMD = −0.47, Z = 2.66, 95% CI −0.82 to −0.12, p = 0.008) when compared with the control group. The result showed high heterogeneity (I 2 = 70%). Eight studies including 913 participants measuring weight were included in metaanalysis. For the analysis of blended interventions versus control, the SMD on weight was −0.42 (Z = 6.26, 95% CI −0.55 to −0.29, p < 0.001) with negligible heterogeneity (I 2 = 0%, p = 0.60). The result suggested that blended intervention can lead to significant weight loss.
The synthesized effect size of intervention on BMI was analyzed on five studies that included 424 participants. The overall effect demonstrated that the blended intervention led to a significant decrease in BMI (SMD −0.68, Z = 6.11, 95% CI −0.90 to −0.46, p < 0.001) when compared with the results of the control group. The heterogeneity test showed no significance among the included studies (I 2 = 13%, p = 0.33).
Five studies involving 620 participants measured waist circumference. The overall effect demonstrated that the blended intervention led to a significant decrease in waist circumference (SMD = −0.47, Z = 2.66, 95% CI −0.82 to −0.12, p = 0.008) when compared with the control group. The result showed high heterogeneity (I 2 = 70%).

Principle Findings
To the best of our knowledge, this is the first study to systematically review and metaanalyze the evidence of face-to-face and eHealth blended lifestyle interventions targeted at promoting PA, diet, and weight-related outcomes among adults. Compared with the research of stand-alone, face-to-face and eHealth interventions, the blended intervention is still under development. However, this review has indicated that the number of faceto-face and eHealth blended interventions showed an increasing trend over the past five years for health problems, while the majority (14/17, 82.4%) of the included studies were conducted within the last five years. Current findings showed that the blended intervention could lead to a significant improvement in walking steps, total PA level, diet quality, and a significant decrease in energy intake, weight, BMI, and waist circumference when compared with a control group. Though all the included studies adopted a blended intervention approach, there was still a high variability in participants (e.g., cultural background, age, and percentage of sex), intervention characteristics (e.g., the frequency of face-toface and eHealth intervention, the channel of eHealth intervention, content and duration of intervention), and measurements. This indicates that the blended intervention, as a promising intervention paradigm, is underexplored and does not have relatively wellacknowledged guidelines or standards such as CONSORT [63] and AGREE [64].
Despite the high variability in eligible studies, several notable trends can still be found. First, current blended intervention studies paid more attention to specific populations, such as people who were overweight and obese, people with chronic diseases, and pregnant/parturient women. In addition, the included studies recruited higher proportions of female participants. It remains unclear who would benefit most from the blended intervention based on the current studies targeting non-representative populations. Therefore, the effectiveness of blended interventions in promoting PA and diet among the general population should be further explored in future research. This should include old adults, as insufficient PA and an unhealthy diet are prevalent among this population [65,66].
Second, most of the included studies (15/17, 88.2%) had only two study arms. One study with four study arms reported that the blended group showed higher adherence rates compared with two stand-alone intervention groups and significantly better results on BMI compared with the eHealth group [27]. However, it was found that the adherence to blended interventions was equal to or worse than control groups in all studies except for two. The possible reason for this might be that the blended intervention included both face-to-face and eHealth sessions, which may lead to a heavy burden of time and energy engagement for participants in the blended intervention group compared with that of the control group. In addition, as the drop-out issue is a key challenge in eHealth interventions, the design of the eHealth session might also affect the adherence rate of participants in blended intervention groups. Previous studies assumed that blended interventions may be more effective and cost-effective compared with stand-alone, faceto-face, and eHealth interventions [29,30,[67][68][69]. Comparing the effectiveness of blended interventions with stand-alone interventions would help to find the most cost-effective and effective intervention strategies to help people improve their healthy behaviors. The most used behavior change theories are Social Cognitive theory (SCT) and Self-Determination Theory (SDT). Nevertheless, over half of the included studies were not based on a behavior change theory or model, while previous systematic reviews indicated that theory-based interventions are more effective in health behavioral changes than those theoretically unaware [70].
Third, although all the included studies combined the eHealth and face-to-face intervention session, the optimal and cost-effective dose of face-to-face sessions and eHealth sessions which can help increase the effectiveness of the blended intervention is still unknown. The face-to-face dose was small in most of the included studies, with only one face-to-face session before or after the eHealth intervention. However, those who have difficulty using electronic equipment or expressing their thoughts and feelings in writing through electronic devices might need more face-to-face sessions, such as older adults [18].
Regarding the outcome measurements, over half of the included studies measured PA data using objective methods, while three studies collected PA data via both objective methods and self-reported questionnaires. It seems that objective approaches of data collection regarding PA (e.g., pedometers, accelerometer; wearable fitness tracker, activity monitor, smart phone application,) and weight (e.g., portable automatic BMI stadiometer) are recommended to improve accuracy during data collection. In terms of diet, all data were collected by subjective approaches. The Food Frequency Questionnaire was the most used approach for data collection on diet. It was suggested that computer-assisted recall could be applied to increase the accuracy of diet data in future research [17]. In addition, previous studies indicated that the data from subjective methods may be more heterogeneous than data from objective methods [16,71]. To identify the effects of blended interventions more precisely, future studies should adopt objective measurements to increase the accuracy of outcomes. Furthermore, due to the limited number of eligible studies included in dataanalysis, no subgroup analysis or sensitivity analysis were conducted to evaluate the effect of measurements. In addition, as few studies conducted follow-up data collection, the long-term effect of blended lifestyle intervention on promoting PA, diet, and weight-related outcomes is unknown.

Intervention Effectiveness
Regarding the effectiveness of blended interventions, the results indicated significant increases in step counts and total PA but not in MVPA. One possible reason could be that among the six eligible studies included in the meta-analysis, participants in five studies include populations with special physical conditions, including patients with familial hypercholesterolemia [58], pregnant/parturient women [48,49], and overweight/obese populations [27,51]. Only one study targeted young men [57]. The characteristics of movement behaviors among most participants may contribute more to the low-intensity exercise (walking) and total PA but not to the MVPA. Therefore, more studies targeting the healthy population in normal situations are warranted to explore the effects of blended interventions on MVPA.
Current findings suggested the superiority of blended interventions in promoting decreased energy intake and increased diet quality compared with stand-alone, face-to-face usual care or the wait-list control group. The possible reason might be (i) the majority of included studies provided personalized feedback both in face-to-face intervention sessions and in eHealth intervention sessions based on the daily/weekly data of PA or diet behavior input or reported by the participants. It has been also well established in other studies [72] that personalized feedback can significantly affect health behavior changes. (ii) All included studies for meta-analysis showed high adherence rates to the lifestyle intervention at over 85%, which can efficiently increase the intervention effects [73]. In addition, most included studies targeted the overweight or obese population, which implies that the blended intervention might be effective to help those participants who need special dietary needs with the purpose of weight control. Today there is a high demand in NCD patients (e.g., obesity, diabetes) for medical support, but the traditional face-to-face intervention or treatment struggles to meet their needs [74]. The blended intervention mode can facilitate timely individualized feedback not only through face-to-face intervention at the clinic or healthcare center but also through eHealth intervention at home (e.g., web-based program, Apps). As a result, such an intervention mode can promote patients' adherence to the entire treatment, and they may eventually obtain the health benefits of the intervention [56]. In terms of the effect of blended intervention on fruit and vegetable intake, there was no significant result. The possible reason for this might be that only three studies were included in the meta-analysis, while two of the three studies reported significant increases in fruit and vegetable intake, while one study weighted over 35% reported an insignificant result. The effectiveness of blended interventions on promoting fruit and vegetable intake should be further explored in future studies.
Referring to the effects of the blended intervention on weight-related outcomes, it was found that weight, BMI, and waist circumference of participants improved significantly. Regular PA and diet courses and feedback were applied in all included studies, which implies that the blended intervention can efficiently help with weight management by targeting weight-related behaviors including PA and a healthy diet. Such a conclusion is in line with the findings of a blended intervention on unhealthy lifestyle change among employees at risk of chronic diseases [75].

Limitations
There are several limitations in this review. First, the omission of appropriate topics or relevant studies may have occurred by not including key terms or studies outside the search time frame and other databases. Second, only 5 out of 17 studies were considered as low risk, while 12 studies did not provide detailed information regarding intervention deviation, outcome measurements, or appropriate measurements. Third, the included studies showed a high degree of heterogeneity in participants, study design, and outcome measures. In the meta-analysis, the comparison groups were also non-uniform, including the waiting-list group and usual care group. Finally, because of the limited eligible studies, a subgroup analysis was not conducted. Therefore, interpretation of the results should be undertaken with caution.

Conclusions
Our study demonstrated that face-to-face and eHealth blended interventions achieve preliminary success in increasing walking and time spent in physical activity, promoting diet quality, and decreasing energy intake and weight-related outcomes, including weight, BMI, and waist circumference among adults. The finding highlights the need for future trials that aim to explore the theoretical foundation, intervention deviation, and outcome measurements for blended intervention studies, which would help improve effectiveness. In addition, more comparable studies on blended intervention, stand-alone face-to-face, and eHealth intervention on weight-related healthy lifestyle behaviors are warranted in the future to identify the most effective approach for health promotion.