The global prevalence of obesity has tripled since 1975, and the economic burden of obesity and overweight is projected to reach 3.3% of global gross domestic product by 2060 [1]. Beyond economic considerations, obesity substantially increases risks for numerous chronic conditions including type 2 diabetes, cardiovascular disease, certain cancers, as well as overall mortality. Despite recognition of obesity as a complex chronic disease requiring multifaceted interventions, historically available treatments have shown limited long-term efficacy [2].

Recent advancements in pharmacotherapy have transformed the obesity treatment paradigm. Tirzepatide (Mounjaro) is a novel dual glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) receptor agonist (RA) that has demonstrated remarkable efficacy for weight management in randomized controlled trials (RCTs) [3]. The SURMOUNT-1 trial reported mean weight reductions of 15% to 20.9% at 72 weeks, significantly surpassing previous pharmacotherapies [4]. These results have generated substantial clinical interest; yet, questions remain regarding their generalization to real-world settings and the scope for adjuncts to support or optimize outcomes.

Digital weight loss services (DWLSs) combine pharmacotherapy with app-based tools, regular weight tracking, and professional health education and tailor nutrition-driven coaching. These services may address critical barriers to obesity treatment, including limited health care provider time, accessibility challenges, and inadequate behavioral support [5-7]. Understanding how digital engagement impacts pharmacotherapy outcomes could critically inform clinical recommendations and health service design. Furthermore, digital health interventions represent a promising complement to pharmacotherapy. Recent systematic reviews indicate that technology-based approaches can enhance weight loss outcomes by improving self-monitoring, accountability, personalized feedback, and remote support [8-10]. Studies of remotely delivered weight loss services have shown the efficacy of such digital intervention [11]. However, apart from a few recent studies [5], there is limited evidence regarding how digital engagement influences outcomes specifically in patients using dual GLP-1 and GIP RA for weight management.

This study conducted a retrospective evaluation of the Voy DWLS, which is available throughout the United Kingdom. Our objectives were to characterize the real-world effectiveness of tirzepatide on weight loss, explore the associations between digital engagement and outcomes, and identify factors that correlate with engagement.

This retrospective study used an open cohort approach. The study period spanned early February 2024 to early August 2025, with patients contributing follow-up from the date of first prescription to the earliest date of last weight measurement logged or reached 12 months of follow-up. The Voy digital health platform, a commercial telehealth DWLS for obesity management, was developed by a multidisciplinary team of clinicians, behavioral scientists, and software developers to provide remote behavioral support through live group video coaching sessions, text-based in-app support, dynamic educational content, and the direct supply of tirzepatide for weight management. Drawing upon established frameworks in behavior change and self-management support, the platform combined core digital tools (eg, real-time weight monitoring, medication adherence tracking, and personalized coaching sessions) into a single interface accessible via smartphone or web browser.

Participants became aware of the Voy program through multiple channels: targeted social media campaigns, clinician referrals, word-of-mouth recommendations, and general web searches. They self-enrolled and self-paid via the Voy website, where they completed an online screening questionnaire covering medical history, BMI, and lifestyle factors. Thereafter, the participants interacted with the Voy digital health platform’s weight management program, which integrates dual GLP-1 and GIP RA pharmacotherapies with digital behavioral support to enhance weight loss outcomes. Subsequently, participants self-enrolled via the Voy website, where they completed an online screening questionnaire covering medical history, BMI, and a range of lifestyle factors. Upon enrollment, participants underwent an initial assessment to confirm eligibility, including verification of age, BMI, and the absence of exclusion criteria (Figure 1). All participants underwent comprehensive clinical screening before treatment initiation. Structured safeguards ensured accuracy and safe prescribing: photo identification and full-body photographs for identity and BMI verification; detailed clinical questionnaires capturing medical history, contraindications, and current medications; and individual review by qualified licensed clinical prescribers with cross-checking for clinical red flags. The service operates under Care Quality Commission registration with internal clinical audit processes, escalation protocols for safety concerns, and regular safety reviews. This approach adheres to General Medical Council and General Pharmaceutical Council standards for remote prescribing in the United Kingdom, with clinician-led prescribing decisions and independent verification of eligibility criteria to ensure patient safety.

The initial monthly cost of enrollment for tirzepatide was US $273.72. Participants receive tirzepatide (Mounjaro) via multidose prefilled KwikPen containing 4 weekly doses. Administration is a subcutaneous injection once weekly, with doses ranging from 2.5 to 15 mg. The starting dose is 2.5 mg, increased every 4 weeks in 2.5 mg increments as tolerability permits. New tirzepatide pens are prescribed every 28 days. Monthly costs cover medication, clinical support from qualified prescribing clinicians, personalized coaching from registered dietitians and nutritionists, and access to digital tracking tools and educational content (see Multimedia Appendix 1 for detailed program description). Although online weight tracking was mandatory for continued prescriptions, every other component of the Voy platform was optional.

Participants attended group onboarding sessions and were offered fortnightly coaching to enhance engagement and adherence. Coaches were trained based on principles from social cognitive theory, self-determination theory, the transtheoretical model, and the theory of planned behavior [12,13]. These techniques focused on fostering intrinsic motivation, goal setting, and problem-solving to promote sustainable lifestyle changes tailored to participants’ individual progress.

Participants were adult residents of the United Kingdom, aged between 18 and 75 years, with a BMI of 30 kg/m² or higher or >27.5 kg/m² with obesity-related comorbidities who were enrolled into the Voy digital weight loss service and were prescribed tirzepatide.

Eligibility required access to a smartphone or tablet. Exclusion criteria included history of self-reported eating disorders (eg, anorexia nervosa and bulimia nervosa), pregnancy or patients seeking fertility, known allergies or hypersensitivity to any components of tirzepatide, and severe medical conditions with established contraindications for tirzepatide as per Multimedia Appendix 2.

Engagement metrics and weight loss outcomes were available for the entire sample of participants who used the digital platform during the study period at baseline. Complete demographic information (eg, age, sex, and comorbidities) was available for the entire sample participants who enrolled during the initial phase of the service.

Self-reported weight measurements could introduce reporting bias. To mitigate this, participants were encouraged to provide accurate measurements through regular reminders and had the option to upload progress photographs, enhancing data validity. Additionally, data validation checks were performed internally and by statisticians to identify and address implausible values. Selection bias was minimized by including all eligible participants who initiated the program within the study period, ensuring the sample was representative of the population using the DWLS.

Missing data were addressed through the MMRM approach, which uses all available observations without requiring complete cases. This method operates under the missing at random assumption, whereby missingness may depend on observed covariates and previous measurements but not on unobserved values conditional on observed data. MMRM provides unbiased parameter estimates and maintains statistical efficiency in the presence of intermittent missing data, consistent with statistical methodology for landmark obesity trials [4]. For time-to-event analyses, participants who remained on treatment were censored at their last prescription date or study end, with censoring appropriately incorporated into Kaplan-Meier and Cox regression models.

This retrospective open cohort study was an analysis of deidentified data collected during the routine clinical care of adults treated by Voy. The study was approved by the University College London Research Ethics Committee (Project ID 2025-0906-775). The study adhered to the principles outlined in the Declaration of Helsinki. Participants provided informed consent for their anonymized data to be used for ethically approved research and service improvement purposes upon enrollment in the program. To protect participant privacy, all data were deidentified prior to analysis, with direct identifiers (including names, contact details, and unique patient identifiers) removed and replaced with study-specific codes. Data were stored on secure servers with access restricted to authorized research personnel only. Participants did not receive any compensation for their participation in this study, as the analysis was conducted retrospectively using data collected during routine clinical care.

This study adhered to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [14] (Multimedia Appendix 4).

Among 126,553 participants who initiated tirzepatide, 6746 (5.3%) met criteria for maximally digital engagement. Baseline characteristics stratified by engagement status are presented in Table 1. The overall cohort had a mean age of 42.3 (SD 12.4) years, was predominantly female (99,905/126,553, 78.9%), and had a mean baseline BMI of 35.3 (SD 6.2) kg/m². Compared to nonengaged participants, digitally engaged individuals were significantly older (engaged: mean 44.9, SD 12.0 years vs nonengaged: mean 42.2, SD 12.5 years; P<.001), more likely to be female (engaged: 5486/6746, 81.3% vs nonengaged: 94,419/119,807, 78.8%; P<.001), and had higher baseline BMI (engaged: mean 36.4, SD 6.7 kg/m² vs nonengaged: mean 35.2, SD 6.2 kg/m²; P<.001). Engaged participants also reported higher prevalence of comorbidities, including type 2 diabetes (engaged: 250/6746, 3.7% vs nonengaged: 3681/119,807, 3.1%; P<.001), hypertension (engaged: 978/6746, 14.5% vs nonengaged: 11,592/119,807, 9.7%; P<.001), hypercholesterolemia (engaged: 707/6746, 10.5% vs nonengaged: 7738/119,807, 6.5%; P<.001), and PCOS (engaged: 591/6746, 8.8% vs nonengaged: 7285/119,807, 6.1% of female participants; P<.001).

Among engaged participants (n=6746), all met the 3 required criteria by definition: coaching session attendance (mean 1.00, SD 0.00), app use (mean 1.00, SD 0.00), and weekly weight tracking (mean 1.00, SD 0.00). Among nonengaged participants (n=119,807), partial engagement was common: 86,881 (72.5%) used the app (mean 0.73, SD 0.45), 18,864 (15.7%) attended coaching (mean 0.16, SD 0.36), and tracking adherence (mean 0.09, SD 0.27) was lower. These patterns indicate that while most participants used some platform features, especially 93,627 of 126,553 (74%) of the overall cohort using the digital app, simultaneous engagement across all 3 modalities was achieved by 6,746 of 126,553 (5.3%) of users.

This cohort study demonstrated the notable real-world effectiveness of tirzepatide in a UK-based DWLS and highlighted an association between higher digital engagement and greater weight loss. After controlling for demographic factors and comorbidities, digitally engaged participants achieved an absolute 5.3 percentage points greater weight loss at month 12 compared to nonengaged users (P<.001). This engagement effect emerged gradually, with trajectories beginning to diverge around month 2 and differences progressively widening thereafter. By month 12, digitally engaged participants achieved a mean weight loss of 22.9% compared to 17.6% for nonengaged participants.

Nearly all tirzepatide users eventually achieved ≥5% weight loss regardless of engagement status, but more substantial differences were observed for the engaged group for higher weight loss thresholds. Notably, engaged participants demonstrated substantially higher cumulative probabilities of achieving ≥15% weight loss (2091/6746, 31% vs 14,017/119,807, 11.7%; HR 1.67, 95% CI 1.60-1.75) and ≥20% weight loss (1079/6746, 16% vs 6710/119,807, 5.6%; HR 1.79, 95% CI 1.68-1.91) by study end. The temporal pattern of weight loss differences is meaningful. While statistically significant differences emerged by month 2, the magnitude of difference progressively increased throughout the observation period. In other studies, a 21% weight loss end point was typically reached at week 72 as opposed to at week 52 (month 12) as evidenced in our findings [15].

The temporal differences in weight loss associated with engagement status were further highlighted in our analysis, which revealed significantly accelerated achievement of clinically meaningful weight loss thresholds. While engagement conferred substantial advantages for ≥5% weight loss achievement (5445/6746, 80.7% vs 44,329/119,807, 37%; RR 2.18; HR 1.49, 95% CI 1.45-1.53), more substantial differences emerged for more ambitious thresholds. Notably, engaged participants demonstrated substantially higher cumulative probabilities of achieving ≥10% weight loss (3623/6746, 53.7% vs 26,240/119,807, 21.9%; RR 2.45; HR 1.55, 95% CI 1.50-1.60), ≥15% weight loss (2091/6746, 31% vs 14,017/119,807, 11.7%; RR 2.65; HR 1.67, 95% CI 1.60-1.75), and ≥20% weight loss (1079/6746, 16% vs 6710/119,807, 5.6%; RR 2.88; HR 1.79, 95% CI 1.68-1.91).

Our analysis of baseline characteristics identified that female sex, older age, and higher BMI were associated with increased engagement likelihood, with male participants having 14% lower odds (OR 0.86, 95% CI 0.81-0.92), each decade of age conferring 18% higher odds (OR 1.18, 95% CI 1.15-1.20), and every 5 kg/m² BMI increase associated with 14% higher odds (OR 1.14, 95% CI 1.12-1.16; all P<.001). Among comorbidities, PCOS (59% higher odds; OR 1.59, 95% CI 1.45-1.74), fatty liver disease (52% higher odds; OR 1.52, 95% CI 1.32-1.76), and hypercholesterolemia (34% higher odds; OR 1.34, 95% CI 1.23-1.47) were positively associated with engagement, while diabetes was associated with 17% lower engagement odds (OR 0.83, 95% CI 0.73-0.95;; P=.009). These findings highlight important demographic factors influencing digital health use that may inform future program design and implementation.

This study has several strengths, including its large sample size, real-world setting, and comprehensive assessment of weight trajectories over time. The 126,000-participant cohort provides robust power at early time points (n=40,582 at 2 months); while 12-month completers (n=2653) match other digital studies, the real-world implementation enhances generalizability and ecological validity. Additionally, the operational definition of digital engagement captured meaningful interaction with service components while maintaining practical relevance.

While baseline differences between engaged and nonengaged participants reached statistical significance due to the large sample size (n=126,553), the absolute magnitude of these differences was modest and unlikely to be clinically meaningful. The mean age difference was less than 1 year (mean 42.3, SD 12.5 vs mean 44.9, SD 12.0 years), and the BMI difference was approximately 1 kg/m² (mean 35.2, SD 6.2 vs mean 36.4, SD 6.7 kg/m²). These small baseline differences are insufficient to account for the substantial treatment effects observed, particularly the 7.4% absolute difference in weight loss at 2 months and the near 3-fold difference in treatment persistence. Furthermore, the consistency association between engagement and outcomes across all time points, combined with our multivariable adjustment for baseline characteristics, supports the conclusion that digital engagement genuinely enhances pharmacological efficacy rather than these findings being attributable to baseline confounding.

Several limitations warrant consideration. First, as a retrospective service evaluation, participants were not randomly assigned to engagement conditions, introducing potential selection bias, and we cannot ascertain the causality of this association. While we adjusted for observed confounders, unmeasured factors (eg, motivation and socioeconomic status) may influence both engagement and outcomes. Second, weight measurements were self-reported via home scales, potentially introducing measurement and reporting error. However, this limitation applied to both engaged and nonengaged groups, likely minimizing differential bias. We did not systematically capture concurrent participation in external nonpharmacological weight management programs; we note this as a limitation and as a priority for future data collection. Additionally, ethnicity data were not systematically collected during the study period, limiting our ability to examine potential disparities in engagement or outcomes across ethnic groups. Future implementations of DWLS should prioritize collecting comprehensive demographic data, including ethnicity and socioeconomic indicators, to ensure equity in access and outcomes and to identify populations that may benefit from targeted engagement strategies. The monthly cost for the service in this study service represents a significant financial consideration that may limit access primarily to individuals with higher socioeconomic status. This raises important considerations regarding health inequalities, as individuals who might benefit most from weight management interventions may face financial barriers to accessing digital health services that integrate pharmacotherapy with behavioral support.

Moreover, the rolling enrollment design means that participants were at different stages of their treatment journey during the evaluation period. While this reflects real-world implementation, it complicates the interpretation of longitudinal patterns. Our approach of examining both absolute outcomes at each time point and conducting subgroup analyses of participants with sufficient follow-up time helps address this limitation. Finally, our operational definition of digital engagement, while evidence-informed, represents one approach among many possible definitions. Future research should explore alternative engagement metrics and potential dose-response relationships between engagement intensity and outcomes.

The magnitude of additional weight loss associated with digital engagement was substantial in the context of standard obesity treatment. Previous research indicates that each 5% reduction in body weight confers significant cardiometabolic benefits, including improvements in glycemic control, blood pressure, and lipid profiles [16,17]. Our findings suggest that digital engagement may help individuals reach more substantial weight loss thresholds, potentially amplifying the health benefits of tirzepatide treatment. This suggests that digital engagement may have cumulative benefits, potentially by supporting medication adherence, dietary modifications, and behavioral changes that enhance long-term outcomes [18,19]. Similar patterns have been observed in other behavioral interventions for chronic disease management, where ongoing support provides incremental benefits over time [20].

The identified correlates of digital engagement offer insights for service optimization. The higher likelihood of engagement among female participants, older adults, and those with higher baseline BMI suggests that these demographic groups may find digital support particularly valuable. Conversely, younger male participants with lower BMI appear less likely to engage, potentially benefiting from targeted engagement strategies; yet, digital health engagement literature has shown that male participants benefit more than female participants [21]. Interestingly, studies have proposed that people with diabetes have a slower velocity of weight loss in both the GLP-1 [22] and GLP and GIP-1 [23] agonist exposure groups (albeit, faster velocity in tirzepatide vs semaglutide).

Furthermore, our study confirmed that diabetes status was significantly associated with a lower likelihood of digital engagement (adjusted OR 0.83, 95% CI 0.73-0.95; P=.009). This may reflect complex physiological and psychological mechanisms influencing weight loss velocity in people with diabetes. To address this, DWLS should consider implementing more intensive engagement and coaching interventions tailored for patients with diabetes. We propose that digital services pilot such enhanced approaches to test whether outcomes can be improved. By contrast, the positive association between comorbidities, particularly PCOS [24] and engagement, suggests that individuals with obesity-related health concerns may perceive greater value in comprehensive support services.

Several mechanisms may explain the enhanced outcomes associated with digital engagement. Health coaching provides accountability, problem-solving support, and personalized guidance that can address common barriers to weight management [25]. Regular weight tracking enhances self-monitoring, a behavioral strategy consistently associated with improved weight outcomes [8]. Additionally, digital platforms may facilitate greater medication adherence through reminders, side effect management, and addressing various concerns, particularly important for injectable medications like tirzepatide that require consistent administration [26].

Our findings align with previous digital weight management studies while revealing important distinctions. Xu et al [27] (n=153) defined engagement as any daily food tracking, identifying thresholds of 28.5%-39.4% of days for ≥3%-5% weight loss. W8Buddy used a liberal definition (any platform activity within 14 days), achieving 83.1% engagement with 0.74 kg per month additional weight loss. Our study used the strictest definition (coaching, weekly weight tracking, and app login combined), resulting in only 5.3% meeting criteria, yet demonstrating the largest effect (5.3 percentage points at 12 months) [28]. This contrast, where stricter criteria with fewer engaged participants yielding greater outcomes, suggests that multimodal engagement combining behavioral support, self-monitoring, and clinical oversight produces beneficial effects beyond single-modality interventions. Notably, partial engagement remained common (app use: 93,627/126,553, 74% and coaching: 37,224/126,553, 29.4%), and even nonengaged participants achieved substantial weight loss (17.6%), indicating that platform features provided benefit across engagement levels. These findings demonstrate that while engagement definitions substantially influence observed rates, the consistent association between digital engagement and enhanced outcomes persists across platforms and intervention designs.

The integration of digital services with pharmacotherapy represents an evolving care model addressing limitations of traditional obesity treatment approaches. By providing remote support, behavioral tools, and clinical monitoring, DWLSs may bridge critical gaps in obesity care while reducing barriers related to geography, provider availability, and stigma [29,30]. Our findings suggest that such integrated approaches may optimize the effectiveness of novel obesity pharmacotherapies.

Our findings have important implications for clinical practice, health service design, and research. Clinically, practitioners should consider recommending digital support services alongside tirzepatide prescriptions, particularly for patients who may benefit from additional accountability and behavioral guidance. From a service design perspective, our results support investment in research and development for digital infrastructure that facilitates engagement, particularly during early treatment phases when engagement patterns appear to be established.

Some self-funded DWLS can exacerbate health inequalities; therefore, future research should examine the cost-effectiveness of such integrated DWLS compared to medication-only approaches and explore alternative funding models, including potential National Health Service commissioning pathways, that could improve accessibility across socioeconomic groups while maintaining service quality and clinical outcomes [31]. Additionally, qualitative studies exploring patient experiences and clinician perspectives could identify specific digital components that contribute most to engagement and outcomes. Finally, RCTs comparing different digital support intensities would address causality questions and optimize resource allocation.

This study demonstrates that digital engagement was significantly associated with enhanced weight loss outcomes among individuals using tirzepatide for obesity management. By month 12, engaged participants achieved –22.9% weight loss and an absolute difference of –5.3% compared to nonengaged participants, with differences emerging early and increasing over time. These findings highlight the potential value of integrated care models combining pharmacotherapy with digital support services. As novel obesity medications continue to transform treatment possibilities, optimizing their implementation through complementary digital strategies represents a promising approach to addressing the global obesity epidemic.