Table 2.
Individual study characteristics (n=35).
| Author; Year | Country; publication type | Study design; duration; n | Intervention used | Compliance | Key outcomes |
| Goldstein, 2016 [32] | United States; thesis dissertation | Pilot trial; 6 weeks; n=12 | DietAlert and WWa | EMAb survey response: 94.6%; retention: 100% (with compensation) | Model accuracy: 0.67-0.72; specificity: 0.68-0.72; sensitivity: 0.45-0.70; higher data volume ↑ model outcomes, especially sensitivity. |
| Goldstein et al, 2017 [10] | United States; journal paper | Developmental; N/Ac; N/A | DietAlert and WW | N/A | N/A |
| Forman et al, 2019a [33] | United States; journal paper | Pilot; 8 weeks; n=43 | OnTrack (formerly DietAlert) and WW | EMA survey response: 85.1%; retention: 97.7% | Negative predictive value 80%; 70.15% alerts opened; 3.13% weight loss; app easy to use, had minimal issues, moderately useful and enjoyable; unplanned lapses ↓ over time. |
| Forman et al, 2019b [34] | United States; journal paper | RCTd; 10 weeks; n=181 | OnTrack (formerly DietAlert) and WW | EMA survey response: 62.9%; retention: 88.4% | Specificity 83.8%; Sensitivity 69.2%; 46.9% alerts opened; 2.1% weight loss when moderated by diet type; high satisfaction reported; 72.8% risk alerts received as helpful or accurate; dietary lapses ↓ over time. |
| Goldstein et al, 2020 (Primary) [35] | United States; journal paper | Randomized trial; 10 weeks; n=121 | OnTrack (formerly DietAlert) and WW | EMA survey response: 65.4% (8 questions) and 60.5% (17 questions); retention: 84.3% | Accuracy 79.7% (8 questions) vs 79.9% (17 questions); specificity 84.4% (8 questions) vs 81.7% (17 questions); sensitivity 71.6% (8 questions) vs 77.7% (17 questions); 46.9% alerts opened; 3.4% weight loss; 72.84% risk alerts received as helpful or accurate. |
| Goldstein et al, 2021a (Secondary) [36] | United States; journal paper | Secondary analysis of randomized trial; 10 weeks; n=121 | OnTrack (formerly DietAlert) and WW | EMA survey response: 63.1% | 0.49% weight loss; 50.3% accessed intervention library per week; lapse frequency not statistically significantly associated with percentage weight loss. |
| Goldstein et al, 2021b (Secondary) [37] | United States; journal paper | Secondary analysis of randomized trial; 10 weeks; n=121 | OnTrack (formerly DietAlert) and WW | EMA survey response: 62.9% | Lapses occurrence: multiple 28.5% (LASSO accuracy 69.8%), planned 22.9% (67.2%), off-plan 16.4% (67.9%), larger portion 12.5% (67.7%), unknown points 10.8% (70.6%), unintended time 8.1% (69%); ↓unplanned lapses (larger portion, unintended time, multiple) ↑ weight loss; 3.7% weight loss observation. |
| Burke et al, 2017 [29] | United States; journal paper | Pilot randomized trial; 12 weeks; n=39 | SMARTER and LoseIt! | SMe: 53.5%; SM and FBf: 55.9%; SM, FB, and F2Fg: 65.3% adherent to self-monitoring; retention was 74% (with compensation). | All groups lost weight and ↓SBPh; no between-group differences for weight loss, SBP, DBPi, and self-efficacy; DBP ↓ in SM and SM+FB+F2F; ↑ self-efficacy only in SM. |
| Burke et al, 2020 [27] | United States; journal paper | Developmental; N/Ac; N/A | SMARTER and Fitbit | N/A | N/A |
| Burke et al, 2022 [28] | United States; journal paper | RCT; 6 months; n=502 | SMARTER, Fitbit, and 1-to-1 90-minute dietary counselling | Around 54.8% of feedback messages were opened; retention was 86%. | Both groups lost weight, ↓ BMI, and WCj; no between-group differences. |
| Burke et al, 2022 (Primary) [26] | United States; journal paper | RCT; 12 months; n=502 | SMARTER, Fitbit, and 1-to-1 90-minute dietary counselling | Around 42.2% of feedback messages were opened; retention was 78.5%. | Both groups lost significant weight (–2.16 kg); no between-group weight loss difference; % days adherent to the calorie goal was higher and declined more slowly in SM and FB; ↑ feedback messages opened→↑ adherence to calorie goal and weight loss. |
| Cheng et al, 2023 (Secondary) [30] | United States; journal paper | Secondary analysis of RCT; 12 months; n=502 | SMARTER, Fitbit, and 1-to-1 90-minute dietary counselling | Not stated | Minimal diet quality improvement overall; weight loss ≥5% linked to higher HEI-2015k [38] scores at 6 months, but not sustained at 12 months. |
| Bizhanova et al, 2022 (Secondary) [39] | United States; journal paper | Secondary analysis of RCT; 12 months; n=502 | SMARTER, Fitbit, and 1-to-1 90-minute dietary counselling | Around 66.5% met the PAl goal at week 1; median adherence was higher in SM and FB (165.2%) vs SM (106.3%) at 12 months, but no within-group differences; adherence was nonlinear and nonsignificant in both groups. | Greater PA adherence was linked to male sex, more feedback message engagement, higher baseline self-efficacy, week-1 adherence, greater weight loss at week 4 and 12 months, and fewer mental health issues; MLm models (random forest regression, regression tree model, and LASSOn model) identified week-1 PA attainment as the strongest predictor. |
| Kariuki et al, 2024 [40] | United States; journal paper | Qualitative study; 6 focus groups; n=23 | N/A | N/A | Successful weight loss group: SMARTER helpful, organized, effective with consistent use; Unsuccessful weight loss group: weight gain or no weight loss experienced, feedback discouraging, harsh or out of context; Overall: both diet and PA are key in weight loss; most messages are inaccessible due to a 1-hour window, timing, and personalization needed for such applications. |
| Bond et al, 2014 (Primary) [41] | United States; journal paper | Quasi study; 4 weeks; n=35 | B-MOBILE | Not stated; 85.7% retention (with compensation). | ↓ sedentary time, ↑ light, and MVPAo in all conditions; 3-minute breaks gave the largest effects (–5.9% sedentary time, +3.9% light PA); 90% found real-time feedback motivating and found smartphone prompts helpful in reducing sedentary time; 6-minute PA breaks were most preferred. |
| Thomas and Bond, 2015 (Secondary) [42] | United States; journal paper | Secondary analysis of quasi-study; 4 weeks; n=35 | B-MOBILE | Adherence: 3 minutes (89.4%), 6 minutes (86.7%), and 12 minutes (77%). | High engagement across all conditions; 3-minute prompts produced the most prompts/day (7), walking breaks/day (6.5), and shortest latency (23 minutes); walking minutes significantly different in 3-minutes (37.2) and 6-minutes (38.7) condition vs 12-minutes (32.5 minutes); 3-minute condition yielded similar total walking time as 6-minute as participants often exceeded required duration of prompt. |
| Westenenk, 2023 [25] | The Netherlands; thesis dissertation | Microrandomized trial; 14 days; n=13 | Ancora Health | 92.9% retention | Sending 1 prompt ↑ steps vs no prompt; evening prompts ↓ steps vs morning; 2 prompts/day ineffective. |
| Shapiro et al, 2012 [31] | United States; journal paper | RCT; 12 months; n=170 | Text4Diet | Increased adherence to knowledge testing, decreased adherence to first and follow-up weight and step queries over time, and 76% retention (with compensation). | Modest weight loss, no group differences; higher SMS text messaging adherence → greater weight loss; ↑ steps linked with ↑ weight loss at 12 months; moderately strong satisfaction with program; ↑ satisfaction with pedometer component → ↑ weight loss at 6 months; 85% would pay US $4.99/month. |
| Haggerty et al, 2016 [43] | United States; journal paper | Randomized trial; 6 months; n=20 | Text4Diet | 100% retention | 90% lost weight; telemedicine > texting (–7.6% vs –4.1%); ↓ interleukin-2 (IL-2) postintervention. |
| Gupta and Sood, 2015 [44] | India; conference paper | RCT; 4 weeks; n=33 | Let’s Exercise | Not stated | App rated effective and useful; 84.6% satisfied; 60% would continue using the app. |
| Patrick et al, 2009 (Primary) [45] | United States; journal paper | RCT; 4 months; n=65 | mDIET | Two-thirds of messages were responded to by participants at the end of 4 months, and retention was 83.3%. | Intervention lost –3.16% weight vs –1.01% control; satisfaction high (92% would recommend). |
| Norman et al, 2012 (Secondary) [46] | United States; journal paper | Secondary analysis of RCT; 4 months; n=65 | mDIET | Not stated |
Intervention improved weight change, fruit and vegetable intake, and eating behavior; fruit and vegetable intake and eating behavior were inversely associated with weight loss and were mediators of weight loss. |
| Valle, et al, 2020 (Primary) [47] | United States; journal paper | Microrandomized trial; 12 weeks; n=53 | Nudge | Two-thirds of messages were viewed by participants, and retention was 98.1%. | Message viewing declined weekly (–0.15/day) and with weight gain (+1 lb = –0.08) or longer lapses in weighing (–0.063/day). Viewing also fell with activity tracking lapses (–0.03/day). Likelihood of message viewing increased with more previous messages viewed (+0.07 per 1%) and more days meeting diet goals (+0.14/day). |
| Hurley et al., 2024 (Secondary) [48] | United States; Journal Paper | Secondary analysis of microrandomized trial; 12 weeks; n=53 | Nudge | Not stated | Each unmet goal decreased the odds of viewing messages by 34.8%. Odds of viewing also declined daily (OR=0.977). Baseline depressive symptoms did not moderate these effects. |
| Rajanna et al, 2014 [49] | United States; conference paper | Developmental; not stated; user study (n=4), formative evaluation (n=2), summative evaluation (n=2) | Step Up Life | Not stated | Around 80% of users across the 3 evaluation phases (user study, formative evaluation, and summative evaluation) showed strong interest in the app. |
| Van Beurden et al, 2021 [50] | United Kingdom; journal paper | Developmental; N/A; N/A | ImpulsePal | N/A | N/A |
| Chen et al, 2024 [51] | United States; conference paper |
|
GatorTrack and FatSecret | Completed 57.14% of daily logs, 15.09% of notification clicks, and 18% of notifications; 100% retention (with compensation). | Context-based notifications led to shorter click response time (12.33 vs 18.42 minutes), higher click rate (19.05% vs 13.96%), and higher completion rate (21.77% vs 17.32%). The average overall log rate was higher in the context-based condition (58.87% vs 55.54%) but was not significant. |
| Everett et al, 2018 [52] | United States, journal paper | Quasi-experimental (1-group pretest-posttest); 3 months; n=55 | Sweetch | 86% retention |
↑ PA (2.8 metabolic equivalent task h/wk), ↓ weight (–1.6 kg, 2%), ↓ BMI (–0.6 kg/m2), ↓ WC (–1.4 cm), and ↓ A1c (–0.1%); high acceptability (78%). |
| Purpura et al, 2011 [53] | United States; conference paper |
Randomized trial (conceptual); 9 weeks; n=26 | Fit4Life (fictional) | N/A | Focuses on ethical reflection rather than efficacy. |
| Spanakis et al, 2017 [54] | The Netherlands; journal paper | Mixed method (study I: developmental and 2 weeks; study II: RCT and 8 weeks); n=100 | ThinkSlim | Study I: 80%-81% of assessments done, study II: 70.5% of assessments done, and ~9.9 hours of application usage. | ML clustered 6 eater types; study II showed the feasibility of adaptive feedback based on these 6 eater types. |
| Finkelstein et al, 2015 [55] | United States; conference paper | Pilot, randomized crossover study; 8 weeks; n=30 | Fitbit One and Android smartphone app | 90% retention | Inactivity decreased during “message-on” periods (24.6% vs 30.4%), and step count increased (6199 vs 5615 steps). Most participants expressed high acceptance and willingness to use the app in the future. |
| Hermsen et al, 2019 [56] | The Netherlands; journal paper | RCT; 15 weeks; n=141 | 10sFork | 86.5% retention (with compensation). In the group using the online dashboard, only 55.3% used the dashboard. |
Eating rate slowed (−1.8 bites/min), the success ratio of >10 seconds between bites increased by 22.5%, and BMI decreased by 0.5-0.8; the dashboard added no benefit. Vibrotactile feedback had a small to moderate effect on bite rate and a moderate to large effect on the success ratio, and both effects remained significant for 8 weeks. |
| Mendi et al, 2013 [57] | Turkey; conference paper | Developmental; N/A; N/A | Android app and wrist-worn sensor | N/A | N/A |
| Moses et al, 2023 [58] | United States; conference paper | Mixed method (pilot quantitative study and qualitative study); not stated; n=29 | SMS JITAIp | Not stated | Participants preferred non-numeric messages (nonsignificant), although numeric messages were still positively received. There was a statistical significance between message type and comprehension error rates. |
| Gao, 2021 [59] | United States; thesis dissertation | Quasi (1-group posttest); not stated; lab testing (n=28) and field testing (n=4) | ADM-IPAq | Not stated | More than 43% of segments were detected as eating activities, with the best accuracy of 74.3%. |
aWW: Weight Watchers.
bEMA: ecological momentary assessment.
cN/A: Not applicable.
dRCT: randomized controlled trial.
eSM: self-monitoring.
fFB: feedback.
gF2F: face-to-face.
hSBP: systolic blood pressure.
iDBP: diastolic blood pressure.
jWC: waist circumference.
kHEI-2015: Healthy Eating Index 2015.
lPA: physical activity.
mML: machine learning.
nLASSO: least absolute shrinkage and selection operator.
oMVPA: moderate to vigorous physical activity.
pJITAI: just-in-time adaptive intervention.
qADM-IPA: Automated Diet Monitoring Intelligent Personal Assistant.