IntroductionBackground

Metabolic syndrome (MetS) is a highly prevalent condition that affects up to approximately 30% of adults aged >65 years worldwide []. It consists of multiple symptoms, namely abdominal obesity, glucose intolerance, hypertension, and high cholesterol as well as low high-density lipoprotein []. It is associated with a substantially increased risk of premature morbidity and mortality from diabetes and cardiovascular disease (CVD) []. Low levels of physical activity (PA) are strongly associated with MetS, including obesity and overweight [], high blood pressure [], and insulin intolerance []. Furthermore, low levels of activity are significantly associated with increased risk of complications of MetS, including diabetes and CVD [,]. In addition, research has found that losing weight by approximately 5% to 10% results in significantly reduced MetS-associated markers [] in patients with existing disease, highlighting that MetS may be modifiable through lifestyle-related weight interventions. Dietary modifications, including reduced sodium, sugar, and fat intake, are also highly beneficial for reducing the risk of the syndrome and its complications [].

In recent years, mobile health (mHealth) has increasingly been used to support behavior changes related to weight loss, including improving dietary intake and physical activity []. Research on the use of mHealth interventions has found support for a moderate effect size for assisting with weight loss []. This includes the use of SMS text messaging for behavior change and mHealth apps that target weight loss using a range of behavior change techniques (BCTs) [], including self-monitoring, feedback, goal setting, education, tips, personal tailoring, reminders, encouragement, and social and professional support []. mHealth is a form of health care that enables timely accessibility, portability, and personalized medicine tailored to the needs of the user [,]. It includes smartphones, PDAs, MP3 players, iPads (Apple Inc), smart clothing, and smart watches [,].

Emerging research in the mHealth field has focused on developing conversational agents that can simulate human professional interactions for managing different health problems [], including weight issues []. Furthermore, avatars have also been developed to display a conversational coach in addition to written conversational text, simulating real-life interactions with a professional, such as a live fitness coach [,]. Having a conversational coach complement or replace metabolic-related health advice from professionals may increase accessibility and enable more timely health monitoring and diagnosis of health conditions [] such as MetS if physicians also gain access to patient data. Given that technology in the field is advancing, it is time to determine whether these conversational agents are effective for assisting with MetS-associated risk factors, including overweight, obesity, physical inactivity, and unhealthy dietary intake. There is also a need to better understand what types of weight-related and MetS-related studies have been undertaken using conversational agents and to identify challenges with the technology and future areas of research.

Aims

This review aimed to better understand the evidence surrounding the use of conversational coaches for metabolic-related risk factors and biomarkers. Furthermore, this review aimed to determine whether conversational coaches are effective for improving weight-related behaviors and metabolic indicators and whether conversational agents are acceptable for consumers as agents of behavior change.

Research QuestionsResearch question (RQ) 1: How effective are conversational agents (chatbots and avatars) for weight-related behaviors, including diet and exercise?RQ 2: How effective are conversational agents for improving metabolic risk factors, including blood pressure, cholesterol, abdominal obesity, and glucose (diabetes management)?RQ 3: What are consumers’ perspectives on the use of chatbots?
Methods

A systematic review of PubMed and MEDLINE was conducted in December 2021 for all relevant studies on conversational coaches for metabolic risk factors published over the last 10 years. Google Scholar was also searched for any additional papers along with manual hand searching.

Inclusion and Exclusion Criteria

This review included studies on chatbots or avatar conservational agents that acted as coaches for improving metabolic health behaviors, including dietary intake (sodium and sugar intake), PA, and weight (including abdominal obesity). Studies that evaluated one or more physiological indicators of metabolic health or risk factors for MetS, such as diabetes, glucose intolerance, hypertension, cholesterol, and serum triglycerides, were also included. Studies must have been published in the English language to be included. Chatbots that were used for survey reasons but not primarily for targeting weight-related or metabolic risk factors were excluded. Studies whose primary focus was not on conversational coaches were excluded (including those that had an avatar element but did not primarily focus on evaluating it). Studies on wearables that did not include avatars or chatbots were excluded. Studies in pregnant women were excluded.

Search

The keywords included word variations for “chatbot,” “virtual assistant,” “virtual coach,” or “avatar”; weight-related behaviors, including “diet,” “exercise,” or “weight”; and metabolic risk factors, including “hypertension,” “cholesterol,” or “diabetes.” The search strategy is shown in .

PubMed search strategy example.

1.Cardiometabolic risk factors

“obesity”[MeSH Terms] OR “obese”[tiab] OR “obesity”[tiab] OR “overweight”[tiab] OR “overweight”[tiab] OR “BMI”[tiab] OR “Body mass index”[tiab] OR “Body mass index”[MeSH Terms] OR “physical activity”[Tiab] OR adiposity [tiab] OR weight gain[tiab] OR body weight[tiab] OR “abdominal visceral fat”[Tiab] OR “adipose tissue”[MeSH Terms] “weight loss”[Mesh] OR “weight loss”[tiab] or “metabolic syndrome”

Diet and physical activity

diets[tiab] OR “diet”[mesh] OR diet[tiab] OR “energy intake”[tiab] OR nutrition[tiab] OR “diet, food, and nutrition”[MeSH Terms] OR diets[tiab] OR Caloric restriction[tiab]OR “physical activity”[tiab]

hypertension[tiab] OR “Blood Pressure”[tiab] OR Prehypertension[tiab] OR BP[tiab] OR “Systolic blood pressure”[tiab] OR SBP[tiab] OR “Diastolic blood pressure”[tiab] OR DBP[tiab] OR cardiovascular[tiab] OR hypotensive[tiab] OR “Hypertension”[MeSH] OR “Blood Pressure”[MeSH] OR “Prehypertension”[MeSH]

“cholesterol”[MeSH Terms] OR cholesterol[tiab]

“Diabetes Mellitus”[MeSH] or diabetes[tiab] or diabetic[tiab] or prediabetes[tiab] or pre-diabetes[tiab] OR “glucose”[MeSH Terms] OR “glucose”[tiab]

AND

2. Technology

chatbot*[tiab] OR chat bot[tiab] OR chat-bot[tiab] OR chatter bot[tiab] OR chat bots[tiab] OR chat-bots[tiab] OR chatter bots[tiab] OR chatterbot*[tiab] OR smart bot[tiab] OR smartbot[tiab] OR smart bots[tiab] OR smartbots[tiab] OR smart-bot*[tiab] OR virtual agent*[tiab] OR virtual character*[tiab] OR virtual coach*[tiab] OR virtual human[tiab] OR avatar*[tiab] OR embodied agent*[tiab] OR relational agent*[tiab] OR animated character*[tiab])

1 AND 2

Textbox 1. PubMed search strategy example.Screening and Data Extraction

Titles were screened for relevance to the RQs, followed by abstract and full-text retrieval of eligible studies that met the inclusion criteria. A second reviewer (LL) screened the abstracts and full texts against the inclusion and exclusion criteria to ensure agreement. Quantitative and qualitative data were extracted and summarized in a tabular format, including study characteristics, measures, outcomes, and intervention details.

ResultsGeneral Description

LL and ME screened the final selected papers individually. A total of 52 full texts were selected [,,-]; however, after double peer screening, 1 protocol and 1 dated technology were removed. The final number included 50 papers [,,-,-,]. Details of the search process and reasons for exclusion are illustrated in [].

Most of the studies were feasibility and usability studies. A few studies were qualitative and explored consumer perspectives on conversational agents for weight-related behaviors [,]. The countries where the studies were conducted included Australia, the United States, Italy, Spain, and Taiwan [,,-]. Most of the studies explored virtual agents for diet and exercise, with only 2 (4%) exploring chatbots for hypertension management [,]. The majority were conducted among adults, but 3 (6%) were conducted among teenagers and preteens [,,]. The study characteristics and results are summarized in .

‎Figure 1. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart of the search and screening process. MetS: metabolic syndrome. Table 1. Study characteristics.Study and yearLocation, N, and designSex (%)Age (years)Health targets and measuresTechnology and
proceduresOutcomesEcheazarra et al [], 2021Location: Spain
‎ N=112
‎ Design: 2-year RCTa
‎ Female: 42Mean 52.1BPbTensiobot (telegram app)
‎ Reminders to check BP
‎ Education on how to properly check BP using videos
‎ Warnings and graphic feedback on BP
‎ GPsc can connect with the app to access patient data
‎ Advice offered 24/7
‎ No significant differences in adherence between groups
‎ Bot group had higher levels of knowledge on good practice skills for BP (t=2.11; df=82.3; 95% CI 0.39-12.6; P<.05)
‎ Measurements (P<.05)
‎ Bot found to be acceptable/likable
‎ Adherence after intervention: 85%
‎ Griffin et al [], 2021Location: United States
‎ N=15
‎ Design: mixed methods questionnaires with semistructured interviews qualitative
‎ Female: 53Mean 59 (SD 11)BPTheoretical discussion around chatbots for hypertension medication management
‎ Most patients were interested in and open to trying a chatbot for hypertension medication management and reminders
‎ Privacy concerns and usability with mobile phones
‎ Larbi et al [], 2021Location: Switzerland
‎ N=30
‎ Design: feasibility study
‎ Female: 50Range 18-69PAdMYA social media chatbot
‎ Perceptions of usefulness and informativeness: 53%
‎ User friendly: 83%
‎ Failed to understand user input: 63.3%
‎ Potential confusion with using the technology 43.3%
‎ Lin et al [], 2021Location: Taiwan
‎ N=96
‎ Design: factorial experimental study with 4 arms
‎ Female: 53Mean 21.5; range 18-42PA (core muscle exercise)Increase in PA (vector movement) of 986.7 (SD 1.03) points in in normal realistic avatar relative to muscular avatar
‎ Higher self-efficacy for core muscle exercise in normal avatars vs muscular avatars in female participants (+0.66, SD 0.1 points) and higher levels than in male participants (+0.9, SD 0.2 points)
‎ P<.05
‎ Dol et al [], 2021Location: The Netherlands
‎ N=71
‎ Design: qualitative study
‎ Female: 100Mean 44.4 (SD 12.86); range 19-70Emotional eatingConversational coach for emotional eating
‎ The design of the conversational coach should integrated dialectal behavioral coaching strategies, as preferred by participants with emotional eating behavior
‎ Lin et al [], 2021Location: Taiwan
‎ N=104
‎ Design: experimental design study
‎ Female: 50Mean 70.39 (SD 6.51); range 60-88PA perceived exertion
Self-efficacyAssigned to either age-matched or young avatars for PA Theory: Proteus effect of avatar embodiment
‎ Watched videos in a digital gym where they exercised
‎ Wore a head-mounted display
‎ Older male participants assigned to young avatars had higher perceived exertion than counterparts assigned to older ones (+1.56, SD 0.31 points; male participants only)
‎ Female participants assigned to young avatars had higher self-efficacy for future exercise than counterparts (+0.45 points) and male participants
‎ P<.05
‎ Maher et al [], 2021Location: Australia
‎ N=31
‎ Design: proof-of-concept study
‎ Female: 67Range 45-75PA, Mediterranean diet, and weightAIf Paola chatbot teaches users about exercise and uses BCTsg, including goal setting, self-monitoring, and feedback
‎ Mean increase in diet score: 5.7 (95% CI 4.2-7.3)
‎ Mean PA increase: 109.8 min (95% CI 1.9-217.9; P<.01)
‎ Mean weight loss: 1.3 kg (95% CI −2.5 to −0.7; P<.05)
‎ P<.01
‎ No significant changes in blood pressure
‎ Hickman et al [], 2021Location: United States
‎ N=109
‎ Design: 2-arm RCT
‎ Female: 59Mean 52 (SD 11)Hypertension, quality of the physician-patient interactionAvatar intervention or video on hypertension management
‎ Scores for the quality of the patient-provider interaction were better over time (F3=5.25; P<.01) in the within-subjects analysis along with a time by experimental condition interaction (F3=2.91; P<.05)
‎ Between-subject effects per treatment were insignificant
‎ No significant changes in blood pressure
‎ Napolitano et al [], 2021Location: United States
‎ N=136
‎ Design: feasibility study (12 weeks)
‎ Female: 100Mean 27.8 (SD 5.4)Weight, diet, and PA; exercise self-efficacyConversational coach gave lessons on health behaviors
‎ No significant results were found for differences in weight, PA, or consumption of fast food between the intervention arm and control groups
‎ High attrition 44%
‎ Goal achievement for nutrition <10%
‎ Santini et al [], 2021Location: Austria, Italy, and Netherlands
‎ N=60 (2 waves)
‎ Design: qualitative study with focus groups and phone interviews
‎ Female: 53.3% wave 1; 51.6% wave 2Mean 61.9Health behaviors, diet, and PAEmbodied coach for diet and PA
‎ Desire for the avatar to motivate older adults to exercise
‎ Supportive tone and language that is not authoritarian or patronizing
‎ Krishnakumar et al [], 2021Location: India
‎ N=102
‎ Design: pre-post intervention (1 arm)
‎ 16 weeks
‎ Female: 31.4Mean 50.8Diabetes (blood sugar), diet, PA, and weight (logged)Wellthy CARE mobile app
‎ The use of the Wellthy CARE digital therapeutic for patients with T2Dh showed a significant reduction in the mean levels of HbA1ci −1.16% (95% CI −1.40 to −0.92; P<.01); FBGj (−11 mg/dL), and PPBGk (−22 mg/dL); P<.05
‎ Weight decreased by 1.32 kg (95% CI −0.63 to −2.01 kg) after 16 weeks
‎ Dhinagaran et al [], 2021Location: Singapore
‎ N=60
‎ Design: one arm web-based feasibility stud
‎ Female: 62Mean 33.7Diet, PA, sleep, and stressChatbot for diabetes prevention, diet, exercise delivered over Facebook Messenger (Meta Platforms Inc)
‎ Engagement: 50%
‎ Retention: 93%
‎ Satisfaction: high at 92%
‎ 50% agreed that the chatbot was acceptable and usable
‎ No significant changes in health behaviors including PA
‎ Minimal improvement in diet: increase in fruit intake (3 portions) by 4% and vegetables once per day by 2%
‎ To et al [], 2021Location: Australia
‎ N=116
‎ Design: quasi-experimental study (6 weeks)
‎ Female: 81.9Mean 49.1 (SD 9.3)PAFitbit plus a chatbot in the Messenger app
‎ Usability score: 89.4%
‎ Desire to continue using: 35.4%
‎ Helped them: 53%
‎ Mean PA increase: 154.2 min/week (95% CI 2.28-5.63)
‎ OR for meeting PA guidelines: 6.37 (95% CI 3.31 to 12.27)
‎ Mean steps/day increase: 627 (95% CI 219 to 1035)
‎ Mitchell et al [], 2021Location: United States
‎ N=158
‎ Design: mixed methods survey with qualitative interviews study
‎ Female: 100Mean 56 (SD 11) intervention; 57 (SD 11) controlDiabetesAvatar for diabetes self-management
‎ Avatars provide support for diabetes self-management via 3 areas (self, social, and physical) that are linked with engagement
‎ Strombotne et al [], 2021Location: United States
‎ N=590
‎ Design: quasi-experimental study
‎ Female: 11Mean treatment=58.1; control=57.7Diabetes and risk factorsConversational coach and ketogenic diet
‎ BP decrease (systolic): 1.4 mm Hg (95% CI −2.72 to 0.14)
‎ Diastolic BP levels decreased: −1.43 (95% CI −2.72 to −0.14) mm Hg
‎ HbA1c decreased: −0.69 (95% CI −1.02 to 0.36)
‎ Diabetes medication fills: −0.38 (95% CI −0.49 to −0.26)
‎ BMI: −1.07 (95% CI −1.95 to −0.19) kg/m2
‎ Alves Da Cruz [], 2020Location: Brazil
‎ N=27
‎ Design: cluster randomized crossover trial
‎ Female 48.1Mean 63.4 (SD 12.7)HRl, BP, and RRmAvatars with exergames for PA in patients undergoing cardiovascular rehabilitation
‎ Increase in HR (z=82.8; P<.01) and RR (z=12.9; P<.01) during and (5 min) after exergame
‎ Changes in systolic BP but not diastolic with differences within moments z=11.26 (P<.01)
‎ With no statistical significance between groups
‎ Kowalska et al [], 2020Location: Poland
‎ N=249
‎ Design: cross-sectional study
‎ Female: 36.5Mean 65.3 (SD 13.8)CVDnTelehealth voice technology with health professionals and voice conversational agent
‎ High desirability for telehealth consultations with a cardiologist combined with a conversational agent
‎ Desirability for telemonitoring of vitals: 67.5%
‎ 70.7% wanted a consultation with a cardiologist remotely
‎ Piao et al [], 2020Location: South Korea
‎ N=106
‎ Design: Exploratory Randomized Controlled Trial
‎ 12 weeks
‎ Female: 56 intervention; 57 controlRange 20-59Health behaviors (diet and exercise); SRHIoLifestyle coaching chatbot
‎ Informed by habit formation
‎ Cues and goals
‎ Significant improvement in health behavior
‎ The intervention group had higher scores on the SRHI of 7.12 (SD 5.57) with P<.05 at 4 weeks; no significant differences between groups at 12 weeks, PA remained higher after 12 weeks (P<.05)
‎ Naylor et al [], 2020Location: United States
‎ N=20
‎ Design: pilot study
‎ N/ApMean 8.4 (SD 1.3)VO2 (mL × kg–1 × min–1) using indirect calorimetry questionnaire on liking and motivationChildren played tennis with their friend and an avatar
‎ Increased VO2 during game play in both cooperative (3.8 + 1.8 mL × kg-1 × min-1) and competitive play (4.4 +1.8 mL × kg-1 × min-1) compared with resting condition (P<.01)
‎ Children liked exercising more in cooperative games than in competitive games (P<.01)
‎ No differences between game styles in motivation for PA (P>.05)
‎ Hahn et al [], 2020Location: United States
‎ N=42 (child and parent dyads [n=40 completed baseline and follow-up measures])
‎ Design: pilot intervention
‎ Female (children): 55.2Treatment: mean 8.06 (SD 1.10); control: mean 7.5 (SD 1.38)PA using Fitbit and self-report on motivation for PAChildren wore Fitbit with a personalized dog avatar for socializing and support (digital fitness kiosk); theory informed (social cognitive theory)Completion rate: 81.63%
‎ Mean number of PA goals reached: 3.28
‎ Mean time playing with pets: 20.35 min
‎ Mean number of active min: 66 min (no statistical significance was found)
‎ Navarro et al [], 2020Location: United States
‎ N=305
‎ Design: 3-arm RCT
‎ Female: N/AMean 20.0 (SD 2.2); range 18-37Cardiac frequency, step counts, accelerometer, and HR monitorRandomly assigned to avatars embodying them (same face) or different from them (strangers)
‎ Avatars wore normal clothes or gym clothes
‎ Higher cardiac output (frequency) from 6 to 12 min in users of avatars that had a similar appearance (face)
‎ Higher output in users with avatars that additionally wore sports clothing at 6-7 and 10-minute periods
‎ Support for the Proteus effect hypothesis
‎ No changes in step count
‎ Davis et al [], 2020Location: Australia
‎ N=28
‎ Design: pilot single-arm study
‎ Female: 68Mean 56.2 (SD 8); range 45-75Diet: Mediterranean diet adherence tool. Weekly log for diet and step count; activity tracked using a wrist worn tracker (Garmin) that syncs with Paola. Minutes of moderate to vigorous PA assessed with Active Australia SurveyConversational assistant Paola for diet and PA consisted of educational modules, weekly check-ins, and 24/7 availability for PA and diet questions
‎ 12-week pilot
‎ Assisted with increasing PA (step goal achieved 59% of the time)
‎ Adherence to diet: 91%
‎ Navarro
et al [], 2020Location: Spain
‎ N=42
‎ Design: 3 arms—2 avatars vs control
‎ Female: 100Mean 31.9 (SD 11.7); range 19-61PA, IPAQq, self-efficacy to regulate exercise, and PA enjoyment scale (PACESr)Avatar: ideal (perfect body) or normal (matching the participant) and web-based intervention without the avatar
‎ Increased PA in all groups (F1,39=15.8; P<.01; web-based intervention effects)
‎ No effects of time by avatar assignment, ie, interaction
‎ Balsa et al [], 2020Location: Portugal
‎ N=20
‎ Design: usability study with qualitative interviews
‎ Female: experts 88.9%; end users 27.3%Mean 62.62; mean end users 70.9; mean experts 54.3Usability of the app for diabetes medication adherence and improving lifestyle behaviors, diet, and PAThe conversational coach resembles a human
‎ Integrated BCTs: goal setting, self-monitoring, feedback, and social support/counseling
‎ Usability score: 73.75 (SD 13.31) (indicates high usability of the coach)
‎ Chin et al [], 2020Location: United States
‎ N=15
‎ Design: feasibility study
‎ Female: 60%Mean 67 (SD 5.84)PAHealth coach for PA
‎ As part of a PA program using a Google Home device (Google LLC)
‎ Usability was high
‎ 80% of the participants did not experience challenges when interacting with the conversational coach
‎ Fadhil et al [], 2019Location: Italy
‎ N=19
‎ Design: validation study (4 weeks)
‎ Female: 42Mean 28.5; range 19-53Diet and PA questionnaires via chatbot and motivation (HAPAs)CoachAI text based conversational agent
‎ Tailored coaching for habits
‎ Participants were satisfied with the agent
‎ High trust to share personal information to the coach
‎ Ahn et al [], 2019Location: United States
‎ N=67
‎ Design: field study (3 days)
‎ Female: 61.19Mean 11.24 (SD 0.85); range 9-13PA and basic psychological needsUse of a digital dog, with and without a points-based reward system
‎ Higher levels of PA in the rewards points group briefly versus control (F1,58=5.32; P<.05)
‎ No significant effects on PA over time
‎ Stephens et al [], 2019Location: United States
‎ N=23
‎ Design: feasibility study
‎ Female: 57Mean 15.2; range 9.7-18.5Weight management; pre-diabetesTess text-based chatbot counsellor for healthy behavior change usability assessed with progress toward goals and engagement
‎ Usefulness rate: 96%
‎ Progress toward goals frequency: 81%
‎ Srivastana et al [], 2019Location: United States
‎ N=10
‎ Design: usability study
‎ Female: 70Range 44-67PrediabetesThe web-based module used to support diabetes prevention education and a mobile app that is an electronic diary and a coach
‎ Success of modules 60% as they meet weight loss of 5%
‎ Compliance with dietary recommendations: 59%-87%
‎ Compliance with PA: 52%-93%
‎ Thompson et al [], 2019Location: United States
‎ N=27
‎ Design: pilot feasibility study
‎ Female: 73 (teens)Range 10-15DiabetesConversational agent with human features
‎ Conversations around diabetes education
‎ Attrition: low (<10%)
‎ High satisfaction: >80%
‎ Technical issues<10%
‎ Teens and families had a positive experience
‎ Thompson et al [], 2018Location: United States
‎ N=48
‎ Design: laboratory-based study
‎ Female: 50Range 12-14PAPA exergame with an avatar coach
‎ Completion: 87.5%; teens enjoyed the game (mean enjoyment score 68%)
‎ Vigorous PA during 74.9% of the game
‎ Duncan-Carnesciali et al [], 2018Location: United States
‎ N=198
‎ Design: cross-sectional, survey-based design using quantitative and qualitative paradigms
‎ Female: 97.5Range 26-76DiabetesAvatar for diabetes management
‎ Ethnicity including Arab or Middle Eastern, Asian, and White or European descents as well as age were significantly associated with an excellent rating of the video with P<.05
‎ Klaassen et al [], 2018Location: N/A
‎ N=21
‎ Design: usability study
‎ Female: 52Mean 13.9DiabetesConversational coach game with feedback
‎ Integrates BCTs including information on consequences
‎ Usability index of 44.18 (SD 21.18; low)
‎ Sinoo et al [], 2018Location: Netherlands
‎ N=21
‎ Design: experimental study
‎ Female: 37Mean 9.2 (SD 1.1)Diabetes self-managementAvatar for gameplay and diabetes self-management vs robot
‎ Preference for the robot (mean friendship score 4.0, SD 0.6) over the avatar (mean friendship score 2.9, SD 0.7) as a companion
‎ Usability moderate: 58.7 (SD 24.5)
‎ Similarity of avatar to robot led to greater friendship (P<.01)
‎ Tongpeth et al [], 2018Location: Australia
‎ N=22 (development of the application)
‎ N=10 (feasibility testing)
‎ Design: pilot feasibility
‎ Female: 10Mean 52.2 (SD 10.4)Cardiovascular: acute coronary syndrome managementAn interactive, avatar-based education application for improving patients’ knowledge of, and response to, acute coronary syndrome symptoms
‎ Symptom recognition increased: 24%
‎ Satisfaction: 87.3%
‎ Knowledge increase: 15.7%
‎ Friedrichs et al [], 2014Location: Netherlands
‎ N=958
‎ Design: 3-arm RCT
‎ Female: 60.4Mean 42.9 (SD 14.5)PA; Dutch Short questionnaireAvatar with a web intervention or a digital web-based text condition versus control
‎ Significant increases in PA in the intervention arms versus control with B=0.39 in the avatar arm and B=0.44 in the text arms (P<.05)
‎ No differences between the text arm or the avatar arm for PA
‎ Stein et al [], 2017Location: United States
‎ N=70
‎ Design: longitudinal observational study
‎ Female: 74.5Mean 47 (SD 1.8); range 18-76Weight and dietary intakeLark Weight Loss Health Coach (participants were a part of a diabetes prevention weight loss program)
‎ Advice on dietary intake and PA
‎ BCTs used include motivation, encouragement, reminders, and education
‎ 31% increase in healthy eating
‎ Mean weight change: −2.4 kg (SE 0.82; 95% CI −4.03 to −0.77)
‎ Thompson [], 2016Location: United States
‎ N=43 (round 2)
‎ Design: mixed methods survey with qualitative interviews
‎ Female: 50Range 12-14Preferences for a PA interventionExergame with a self-representation avatar
‎ Desired gameplay with the avatar that could be controlled by eliciting the desired action: 62.5% male and 58.3% female
‎ Personalized avatar: 41.7%
‎ Most common avatar features to be customized:
‎ Body: 95.8%
‎ Clothing: 93.8%
‎ Hair color: 87.5%
‎ Behm-Morawitz et al [], 2016Location: United States
‎ N=90, female=100% (the 2 male participants were excluded)
‎ Design: qualitative research and RCT
‎ Female: 100Range 18-61Weight and PA self-efficacyAvatar (embodied) and video game to promote PA
‎ Findings support the use of the avatar for weight management t18=2.15 (P<.05) with the intervention losing 1.75 lbs versus 0.91 lbs in the control
‎ No effects on dietary self-efficacy
‎ Strong correlation with avatar sense of self-presence and confidence in meeting health goals (r=0.95; P<.01)
‎ Themes: avatar benefits include motivation and assisting with self-efficacy for PA
‎ Barrier: games are not for everyone
‎ Kuo et al [], 2016Location: Taiwan
‎ N=76
‎ Design: 2-arm intervention in laboratory
‎ Female: 63.15Mean 21.2Eating behavior observed in laboratoryAvatar that embodied the participants or was a weight-reduced (thinner) version of them
‎ Avatars that embodied a thinner version of the participants shaped eating behaviors more compared with identical self-avatars; including selecting less ice cream (Cohen d=0.35; F1,73=7.8; P<.01) and opted for sugar free drinks (Cohen d=0.29; F1,73=6.0; P<.01)
‎ Ruiz et al [], 2016Location: United States
‎ N=41
‎ Design: laboratory study
‎ Female: 0Mean 64 (SD 7)Cardiovascular behavioral risk factors (diet and exercise)Avatar vs a voice (nonanimated) for behavior change linked with CVD
‎ Avatar increased intentions (+1.56 points) to improve lifestyle behaviors relative to controls (Cohen d=0.77 P<.01; t36=2.48)
‎ Differences in confidence to change risk of heart disease was nonsignificant
‎ LeRouge et al [], 2015Location: United States
‎ N=41
‎ Design: user-centered design, 3 phases with focus group and interviews
‎ N/ATeenagers: 12-17Perceptions of the avatar for diet and exerciseInteractive avatar coach
‎ Desire for a fun human-like interaction
‎ Desire for a lifestyle coach and personal embodiment avatar and an authoritarian one
‎ Desire for customization of the avatar
‎ Advice on activity on the go and meals when eating at home
‎ Goal setting
‎ Technical issues could be a barrier including the internet
‎ Thomas et al [], 2015Location: United States
‎ N=37
‎ Design: feasibility and usability study with pre-post test
‎ Female: 100Mean 55.0 (SD 8.2)Weight-related eating behaviorsConversational coach for weight (focuses on dietary intake and managing eating behaviors)
‎ The coach assisted with perceptions of increased self-control over eating (confidence to control eating: +1 point (SD 0.2; P<.01) and skills for controlling eating +0.7 points (SD 0.1; P<.01)
‎ Ruiz et al [], 2014Location: United States
‎ N=150
‎ Design: RCT
‎ Male: 100Mean 62 (SD 7.9)Diabetes (knowledge)Computer program with an avatar to increase diabetes knowledge and medication (adherence)
‎ There were no significant differences between the intervention group and control group in terms of knowledge, with P=.95
‎ Satisfaction levels were higher in the digital intervention group (F4=3.11; P<.01)
‎ Li et al [], 2014Location: Singapore
‎ N=140
‎ Design: factorial design experiment
‎ Female: 41Range 9-12PA attitudes, motivation, and game performanceAssigned to varying avatars (normal and overweight)
‎ Healthy weight avatars linked with greater scores in motivation for Nintendo exercise (F1,134=5.49; P<.05 [boys]) attitude, and performance (F1,134=2.27; P<.05 [girls])
‎ Napolitano et al [], 2013Location: United States
‎ N=128 (phase I) N=8 (phase 2)
‎ Design: mixed methods (pilot usability testing) study with interviews
‎ Female: 100Mean 34.1 (SD 13.0); range 18-60 (phase 1)Weight, PA [], and weight self-efficacy; satisfaction; preferences survey and interviewsAvatar for diet and exercise
‎ Informed by social cognitive theory
‎ Behavioral modeling
‎ Targeted self-efficacy
‎ 4 weeks
‎ The avatar helpful: 87.5%
‎ Mean weight loss after 4 weeks: 1.6 (SD 1.7) kg
‎ All women found that it helped with their diet and exercise
‎ Most were interested in the avatar
‎ Bickmore et al [], 2013Location: United States
‎ N=122
‎ Design: 4-arm RCT (2 months)
‎ Female: 61Mean 33.0 (SD 12.6); range 21-69Diet (NIHt/NCIu fruit and vegetable scan) and PA (IPAQ)Animated counselor
‎ for diet and PA (separate and combined)
‎ No significant differences between groups in PA after adjustment
‎ Fruit and vegetable servings significantly increased in the diet arm (F3,103=4.52; P<.01)
‎ No significant differences in weight or PA between groups
‎ Likability: Karen was perceived as nice by 35% of the participants
‎ 50% of the participants found Karen helpful
‎ Johnoson-Glenberg et al [], 2013Location: United States
‎ N=19
‎ Design: pilot feasibility study (pre-post study)
‎ N/AGrades 4-12 (ages 9-18)Diet (nutrition and food choice test and knowledge)Diet and exercise game (exergame) with an alien interactive coach
‎ Differences in dietary knowledge of nutrition pre and post intervention (t=4.13; P<.01) and knowledge of the My Plate content in the study (t=3.29; P<.01)
‎ Ruiz et al [], 2012Location: United States
‎ N=30
‎ Design: comparative pilot with three arms (with randomization) intervention
‎ N/AN/APA3D avatar-based VR intervention
‎ Participants completing a 3D VR intervention mediated by avatars resembling the participants showed significant improvement in PA (P<.05)
‎ No significant effects of the intervention on obese or overweight participants
‎ Mestre et al [], 2011Location: France
‎ N=6
‎ Design: laboratory experimental study
‎ N/ARange 19-25PA enjoymentDigital coach paced participants in a VR bicycling setting
‎ The VR coach and VR cycling were associated with higher levels of PA enjoyment (F2,10=13.24; P<.001) in the feedback group
‎

aRCT: randomized controlled trial.

bBP: blood pressure.

cGP: general practitioner.

dPA: physical activity.

eVR: virtual reality.

fAI: artificial intelligence.

gBCT: behavior change technique.

hT2D: type 2 diabetes.

iHbA1c: hemoglobin A1c.

jFBG: fasting blood glucose.

kPPBG: postprandial blood glucose.

lHR: heart rate.

mRR: respiratory rate.

nCVD: cardiovascular disease.

oSRHI: Self-Report Habit Index.

pN/A: not applicable.

qIPAQ: International Physical Activity Questionnaire.

rPACES: physical activity enjoyment scale.

sHAPA: Health Action Process Approach.

tNIH: National Institutes of Health.

uNCI: National Cancer Institute.

Weight

A few studies evaluated the effects of conversational assistants for weight loss [,-,]. The study by Maher et al [] in Australia found that the conversational assistant (chatbot) Paola assisted with a weight loss of 1.3 kg at 12 weeks follow-up (95% CI –0.1 to –2.5). In addition, there was a mean waist circumference reduction of 2.5 cm at follow-up compared with baseline (95% CI –3.5 to –0.7). The chatbot used a range of BCTs, including goal setting, self-monitoring, education, social support, and feedback to users on PA and the Mediterranean diet []. A study in the United States found that the Lark Weight Loss Coach, an artificial intelligence–powered bot, assisted participants with a weight loss of 2.38% (95% CI –3.75 to 1.0) with a mean use of 15 weeks []. The conversational agent was informed by cognitive behavioral therapy and used a range of BCTs, including education, encouragement, and reminders surrounding dietary and PA targets []. The determinants of weight loss included the duration of using the artificial intelligence program and engaging with it, logging meals, and the number of counseling sessions completed []. A large study in the United States examining the use of an avatar coach that targeted self-efficacy and modelled vicarious experiences for diet and PA (4 weeks) found that women lost an average of 1.6 (SD 1.7) kg at follow-up []. A study in India found that an avatar coaching app with calls from health professionals assisted with a weight loss of 1.39 kg (95% CI –0.63 to –2.01; P<.01) at 16 weeks []. A randomized controlled trial (RCT) with a qualitative component found that avatars increase motivation and PA self-efficacy linked with weight loss []. However, some studies did not report any significant weight loss [,].

Diet

A few studies evaluated the effects of conversational coaches (chatbots and avatars) on dietary intake and found that overall, the coaches assisted with ameliorating dietary habits and goals [,,,,,,]. A study in the United States found that healthy dietary intake improved in 30% of participants who were using a conversational weight loss coach []. Another study found that eating behaviors improved in users of a conversational eating coach, which included increases in the mean scores for the perceptions of skills to eat healthily and self-control over their eating habits (0.7 increase in points) as well as confidence to control food consumption in social situations (1.0 increase in points; P<.01) []. The Paola chatbot study found a mean increase in the Mediterranean diet score [] of 5.8 points at 12 weeks follow-up []. Similarly, a study of Karen, an animated counselor, found significant increases (F3,103=4.5; P<.01) in fruit and vegetable intake in the diet intervention arm relative to the control group []. A further study found that eating behaviors were shaped by the appearance of the avatar, with healthier eating behavioral patterns in participants who had thinner avatars including reduced portions of ice cream and opting for healthier sugar-free drink alternatives [].

Physical Activity

A few conversational assistant PA coaches, including chatbots and avatars, were evaluated, and overall, they assisted with increasing PA [,,,,,,,]. Most of them involved exergames with the avatar. However, one of the studies did not find any improvements in PA among the 2 avatars, attributing improvements only to the web-based part of the intervention [], and another study did not find a difference between the web-based intervention and the chatbot (only when considering a standard control) []. A preliminary usability study in Australia found that step count goals increased 59% of the time in users of the chatbot that targeted PA and that participants had a preference for personalization and greater knowledge-based content []. Another pilot study of Paola, the chatbot in Australia, found that it assisted with increasing mean step count by 109 minutes per week at 12 weeks follow-up (95% CI 1.9-217.7) []. A study involving an exergame that used a PA avatar coach in teens found that 75% of the time, participants engaged in 15.88 (SD 5.8) minutes of vigorous PA throughout the game []. Participants also wanted the avatar to have a supportive and nonpatronizing or nondisparaging tone in interactions regarding PA and found that it could motivate older adults when adequately personalized []. Similarly, a study in children also found that they desired the option to personalize the avatar, including controlling and customizing its physical appearance during game play when exercising [].

Proteus Effect

The Proteus effect is a phenomenon wherein individuals embody and emulate the behaviors of their virtual characters such as avatars [,]. A few studies demonstrated support for the Proteus effect when it came to PA behaviors, although the type of avatar varied. A study in Taiwan found that younger looking avatars were associated with higher levels of PA than older looking avatars but only in women. Male participants had higher levels of PA than female participants who used an older looking avatar, highlighting differences between sexes []. A further study found a higher cardiac output resulting from increased intensity of PA in adult users of an avatar that resembled them and wore gym clothes when compared with avatars that appeared unfamiliar like strangers in regular clothing, which reduced heart rate []. Similarly, a study in Taiwan found increases in physical activity assessed in movements (986.7 points higher) in users of a “normal avatar”, more closely resembling them than the most muscular avatar []. They also found that self-efficacy was higher (0.66 points) for core muscle exercises in female participants assigned to normal avatars relative to their muscular counterparts and male participants assigned to the same standard avatar (0.9 points higher), with P<.05 []. Similarly, dietary behavior was also shaped by thinner embodied avatars in another study [].

Diabetes

Most diabetes studies were feasibility studies. The results of diabetes conversational coaches were mixed. A few studies did not have positive findings concerning the applications with avatars for diabetes [,]. However, one study reported a usability score of 73, which is relatively high. Notably, the study integrated a range of BCTs, including goal setting, feedback, self-monitoring, social support, and counseling []. Low usability scores were reported in a few studies, including one that reported an overall score of 44.58 (SD 21.18) []. Similarly, an RCT of a diabetes coaching avatar did not find that knowledge increased relative to controls, but intervention participants