Obesity remains a global health crisis, affecting 890 million adults worldwide and contributing to diseases such as type 2 diabetes, cardiovascular disease, and certain cancers [1]. In the United Kingdom, projections suggest that by 2050, the annual cost of obesity to the National Health Service (NHS) could reach £9.7 billion (US $12.1 billion), with wider societal expenses approaching £49.9 billion (US $62.4 billion) [2]. Although lifestyle interventions form the foundation of obesity treatment, their long-term effectiveness varies considerably, with more than half of lost weight typically regained within 2 years of weight loss [3].

Clinical trials of glucagon-like peptide-1 receptor agonists (GLP-1RAs), including semaglutide and tirzepatide, have demonstrated significant weight reductions, ranging from 14.9% to 20.9% over 68 to 72 weeks, respectively [4-6]. However, significant weight regain occurs following treatment discontinuation, with STEP-1 trial participants regaining approximately two-thirds of lost weight within 1 year of withdrawing from semaglutide (2.4 mg) and lifestyle intervention [7]. The National Institute for Health and Care Excellence (NICE) guidance recommending tirzepatide for obesity management alongside lifestyle modification [8] has no maximum treatment duration, suggesting potential long-term use. This necessitates the development of sustainable, cost-effective approaches to implement prolonged GLP-1RA treatment and lifestyle modifications.

To date, eligible people living with obesity in the United Kingdom have primarily accessed GLP-1RAs and lifestyle support through NHS specialist weight management services (SWMSs). SWMSs typically offer multidisciplinary team (MDT) support including dietitians, physicians, psychologists, and physiotherapists, often through a tiered system requiring general practitioner (GP) referrals [9]. However, because of underfunding, these services face significant challenges including long waiting lists, understaffing, and limited treatment flexibility, hindering timely access and potentially impacting outcomes [8,9]. In response to these issues and the anticipated demand for newer medications such as tirzepatide, NICE guidance suggests prescribing, monitoring, and multidisciplinary behavioral support could potentially be implemented in primary care [8].

However, concerns exist regarding the capacity of primary care services, particularly GPs, to absorb this additional workload. Despite manufacturer surveys indicating GP interest [8], recent data highlight increasing patient caseloads (17% more patients per GP since 2015) and appointment demands (5.46 million more appointments in February 2024 vs 2019), alongside a 3% decrease in full-time equivalent GPs [10,11]. Significant service transformation appears necessary to meet this demand within the NHS.

The NICE Health Technology Evaluation 14 (HTE14) guidance specifically highlights the potential role of remote interventions in increasing access to GLP-1RA medication and associated support [9]. Emerging evidence suggests that remotely delivered SWMSs could offer a scalable approach to address service provision gaps [9,10,12]. For example, a preliminary 12-week service evaluation of a remotely delivered program combining GLP-1RA (semaglutide) with app-based behavioral support (the same program evaluated in this study) showed that participants lost an average of 6.4% of body weight [12]. This earlier evaluation also suggested feasibility, with relatively low withdrawal rates over 12 weeks, manageable side effects, and generally positive or neutral participant experiences [12]. Furthermore, early economic modeling within the NICE HTE14 process indicated potential cost-effectiveness for remote SWMSs compared with traditional in-person services [9]. However, longer-term (≥12 months) real-world evidence for remotely delivered GLP-1RA–supported weight management interventions is crucial to substantiate their potential role in expanding access to NICE-recommended obesity treatments.

This retrospective service evaluation of a remotely delivered, semaglutide-supported weight management program by Second Nature was conducted to evaluate its effectiveness, feasibility, acceptability, and potential cost-effectiveness over a 12-month period in a real-world commercial setting.

This retrospective analysis used routinely collected service data from participants of a remotely delivered, semaglutide-supported commercial weight management program provided by Second Nature Healthy Habits Ltd, a UK-based digital health company. The study focused on participants who initiated the program between September and December 2023. As a service evaluation of existing data, the analysis primarily reflects outcomes among those who remained in the program and provided data at 12 months (ie, a completer analysis), particularly for the primary effectiveness outcome.

The core intervention, delivery methods, and data collection procedures were consistent with our previously reported 12-week evaluation [12]. Program adaptations were implemented following NICE guidance for tirzepatide, which, unlike semaglutide at the time, had no recommended treatment duration limit [8]. Consequently, the intervention was modified from a fixed duration to an open-ended format supporting individualized treatment periods. Program phases were redesignated from “transition” and “maintain” to “progress” and “momentum” to reflect observed participant behavior patterns, as significant weight loss often continued during these later phases. While this cohort used semaglutide exclusively and was followed for 12 months, these structural modifications enabled extended support for ongoing GLP-1RA–assisted weight management where clinically appropriate. No participants in this cohort received tirzepatide.

This evaluation study met Health Research Authority and Medical Research Council criteria for service evaluation [13]: it assessed current care using existing, anonymized data from an established intervention, did not involve any change to standard practice, and did not involve randomization or a control group. The Health Research Authority explicitly states that “because both audit and service evaluation are considered part of usual professional practice, they are exempt from oversight processes that govern research—that is, they do not require review by a research ethics committee” [14]. The study analyzed anonymized, non-NHS data that could not identify individuals, and thus did not require formal ethics approval (Multimedia Appendix 1). Participants were not compensated for taking part in the study. Participants consented to the research use of anonymized data through acceptance of the program’s privacy policy at registration. The evaluation adhered to General Data Protection Regulations, maintaining participants’ right to request data deletion.

Retrospective data were extracted from the Second Nature database in December 2024 for all participants (N=341) who initiated the intervention between September and December 2023. This timeframe allowed for a full 12-month potential follow-up period. All participants meeting the eligibility criteria during this period were included in the baseline analysis.

Eligible participants were adults aged 18 to 75 years living with obesity (BMI ≥30 kg/m²) who had access to and could use a smartphone or tablet. Exclusion criteria included current or a history of eating disorders; pregnancy, breastfeeding, or actively trying to conceive; specific allergies to medication ingredients; concurrent use of certain medications (eg, other GLP-1RAs and insulin); and specific health conditions (eg, current or historical thyroid cancer, active cancer, inflammatory bowel disease, pancreatitis, severe liver or kidney disease, heart failure, multiple endocrine neoplasia type 2, gallbladder issues, or diabetic retinopathy). People living with type 2 diabetes were eligible if they were not taking any of the excluded medications (eg, metformin was allowed).

Participants were allowed to transition to a version of the digital program that focused only on behavioral support without pharmacotherapy; they were transferred to a separate intervention and were not included in this analysis.

Participants self-referred to the program, joining through digital marketing channels such as Google or Facebook, through word-of-mouth referrals from friends or family, or through guides written by Second Nature that were discovered through search engine results. The cost to the participants varied between £179 (US $224) and £299 (US $374) per month, depending on the dose of semaglutide prescribed.

Analyses were performed using R statistical software (version 4.3.3). For effectiveness, we calculated mean weight loss and SD. We evaluated statistical significance using paired 2-tailed t tests (P<.05) and also calculated the proportions of participants achieving ≥10% and ≥15% weight loss thresholds. Statistical analysis included descriptive statistics for baseline characteristics and primary weight change outcomes. We also descriptively analyzed weight change by subscription status (“active” vs “inactive” paid subscriptions). Program retention was analyzed by calculating active subscription rates and categorizing withdrawal reasons. Monthly behavioral changes, side effects, and participant experiences were analyzed as proportions of total survey respondents. Program engagement was assessed through mean monthly Home screen views. Cost comparisons used the NICE IAT to model costs per 100,000 population, assuming 9.50064% eligibility, 30% digital service uptake, and 75% initiation rate. We standardized annual per-patient costs and calculated percentage differences between the intervention and NHS standard care models [18].

Content analysis was conducted to analyze the free-text responses to the monthly survey question, “How have you found the last few weeks of the program?” For each question, at each time point, the responses were coded and organized into higher-level categories by MW. These frameworks were then refined based on the protocol outlined in a previous study to create 9 higher-level categories that captured key themes in participants’ experiences [12]. Each response could be categorized into multiple themes to reflect the complexity of participants’ experiences. For example, when a participant submitted a free-text response of “ok,” “fine,” “not bad,” “nothing to report,” or “same as before” to the “How have you found the last few weeks on the program?” question, it would be categorized as a “neutral experience.”

A total of 341 participants initiated the program between September and December 2023 and were included in the baseline analysis (Figure 1). Baseline characteristics are presented in Table 1. The majority were women (282/341, 82.7%), with a mean age of 48.2 (SD 11.7) years and a mean baseline BMI of 37.7 (SD 6.9) kg/m². There were no statistically significant differences in baseline characteristics between those who remained active and those who became inactive at 12 months (data not shown; P>.05 for all comparisons).

At the 12-month time point, of the 341 participants, 135 (39.6%) maintained an active paid subscription. The remaining 206 (60.4%) cancelled their subscriptions and became inactive. The mean duration of program participation for the entire cohort was 8.2 (SD 4.2) months. For active participants, the mean duration was 12.5 (SD 0.6) months (reflecting ongoing participation at data extraction), while for inactive participants, the mean duration before cancellation was 5.4 (SD 3.0) months. Weight data at 12 months were available for 179 (N=341, 52.5%) participants of the baseline cohort. This included all 135 (100%) active participants and 44 (n=206, 21.4%) inactive participants who continued to record weight data after cancellation.

When asked about fruit and vegetable consumption, most participants (151/280, 54%) at month 1 and (62/119, 52.1%) at month 12 reported consuming 3 to 4 different items per day across all time points (Tables 2 and 3). The proportion consuming ≥5 items remained relatively stable from 63 (n=280, 22.4%) at month 1 to 28 (n=119, 23.5%) at month 12.

The proportion exercising 3 to 4 times per week or every day increased, whereas the proportion exercising 1 to 2 times per week stayed relatively consistent. In addition, the number of participants who reported no exercise per week decreased from 60 (n=280, 21.5%) participants in month 1 to 9 (n=119, 7.6%) in month 12. An increase in cooking frequency was observed over 12 months, except for those who reported cooking “every day” at baseline, which decreased from 92 (n=280, 33%) participants to 22 (n=119, 18.6%) participants, and those who “never” cooked, which decreased from 4 (n=280, 1.4%) participants to 1 (n=119, 0.8%) participant.

The participant responses were categorized into 9 distinct areas: change in appetite, illness, intervention benefits, negative experience, neutral experience, other priorities, positive experience, and slow progress. Initially, most participants reported positive experiences (211/320, 65.9% at month 1), but this steadily declined to just over a third (41/120, 34.2%) by month 12. Conversely, neutral experiences increased from 15.9% (51/320) at month 1 to 56.7% (68/120) at month 12 (Table 5). Reported intervention benefits were reduced throughout the program, decreasing from 17.2% (55/320) at month 1 to 3.3% (4/120) at month 12. Similarly, reports of appetite changes declined from 8.4% (27/320) to <1%. Negative experiences and reduced effectiveness were consistently reported by 1% (1/120) of participants.

This 12-month retrospective service evaluation provides real-world insights into a remotely delivered, semaglutide-supported commercial weight management program. The primary finding is that participants who completed 12 months of the program (179/341, 52.5%) of the initial cohort achieved substantial and clinically significant weight loss, with a mean reduction of 17.3% (18.1 kg). Those who maintained an active subscription throughout achieved even greater mean weight loss (19.1%, 20 kg). However, these findings must be interpreted cautiously because of the high withdrawal rate, with 47.5% (162/341) of participants providing no 12-month weight data, and 60.4% (206/341), cancelling their subscriptions, meaning the results primarily reflect outcomes in a self-selected group of completers.

The program demonstrated feasibility regarding medication administration and side-effect management among those who remained. Side effects were most common early in the program but decreased significantly over time, with the vast majority of participants reporting no side effects after 12 months. Self-injection was not a significant barrier. Acceptability, as judged by participant experience surveys among respondents, remained largely positive or neutral at 12 months; although positive sentiment decreased over time, neutral responses increased. Engagement with the digital platform, measured by Home screen views, showed a typical pattern of decline over the 12 months. Encouragingly, positive lifestyle changes related to exercise and cooking frequency appeared to be sustained or improved over the 12 months among survey respondents.

The intervention’s estimated cost was substantially lower than traditional NHS SWMS models, suggesting the potential for cost savings if implemented within a public health care system. However, this requires confirmation through formal health economic analysis in that context.

The mean weight reduction of 19.1% (SD 8.0%) for participants with an active subscription observed in this study are comparable with outcomes reported in both controlled trials and real-world evaluations of GLP-1RA medications, respectively [19-21]. For example, reviews of the semaglutide STEP randomized controlled trials showed mean weight reductions of 14% to 17% over 68 weeks [19], while tirzepatide trials demonstrated weight loss of 14.7% to 20.9% across populations with and without diabetes [20]. However, direct comparisons are limited by differences in study design (randomized controlled trials vs real-world evaluation), participant populations (trial eligibility vs commercial users), follow-up duration, the intensity of lifestyle support, and analytic approach (intent-to-treat [ITT] vs completer analysis).

Comparison with other real-world, remotely delivered, GLP-1RA–supported behavioral change programs with completer analyses are more relevant. Two real-world studies in the United States reported 12-month weight reductions with semaglutide when combined with remote support: one found a reduction of 13.8% to 15.3% and another reported a reduction of 13.9% [21,22]. This study demonstrated higher rates of substantial weight loss, with 77.7% (139/179) of participants achieving ≥10% and 61.5% (110/179) achieving ≥15% at 12 months, compared with 38.7% and 18.4%, respectively, in comparable remotely delivered program in the United States [22].

An important finding relates to the inactive participants who provided 12-month weight data (44/206, 21% of those who cancelled). These individuals, who stopped the paid program (including medication and coaching) after an average of 5.4 months, still maintained a mean weight loss of 11.9% at the 12-month mark. This observation requires cautious interpretation because of the small sample size and potential for selection bias. However, it aligns with recent evidence suggesting less weight regain after stopping semaglutide when combined with behavioral support compared with medication withdrawal alone [7,23]. The STEP 1 trial extension showed a regain of about two-thirds (approximately 11.6% absolute weight) of the weight 1 year after stopping semaglutide and intensive behavioral therapy [7], whereas another study reported only 3% regain 12 months after semaglutide cessation in participants who continued a nutritional program [23]. Our finding, while preliminary and based on limited data, suggests the potential role of the initial behavioral intervention component in mitigating weight regain after program cessation.

Program retention rates in remotely delivered weight management services vary substantially. The 12-month retention rate of 39.6% (135/341) observed in this study falls between previously reported rates ranging from 22.7% at 5 months [24] to 58% at 9 months [25] for other GLP-1RA–supported remotely delivered interventions, although structural and financial differences limit direct comparisons. The observed decline in app engagement (Home screen views) is typical for digital health interventions [22,26] and does not necessarily equate to a lack of benefit but highlights the need for strategies to maintain long-term engagement, if required. The improvement in side-effect tolerability over time is consistent with clinical trial data [19] and is reassuring for long-term use. However, the fact that 26% (49/206) of participants who withdrew cited side effects suggests that early intolerance remains a significant barrier for a subset of users, contributing to the lower prevalence of side effects observed in completers (a form of survivorship bias).

It is possible that the behavioral support provided in this study increased and supported the maintenance of exercise, dietary quality, and home cooking to support and enhance the benefits of the GLP-1RA medication. For example, exercise and maintaining dietary quality have been recommended during the use of GLP-1RAs to retain muscle mass during weight loss [27-31]. However, comparison with existing literature is constrained because of the lack of comprehensive behavioral data in GLP-1RA research. Typically, studies of GLP-1RA medication have focused on caloric intake and satiety measures rather than broader lifestyle modifications [32]. Therefore, further research on the impact of behavioral support on GLP-1RA outcomes is needed.

On the basis of NICE modeling data [9,18], this remotely delivered intervention appears to be substantially less costly than traditional secondary care SWMSs. Calculations using the NICE IAT suggest potential direct cost reductions of 40% to 50% per 100,000 population (£1.5-1.8 million; US $1.9-2.25 million) compared with the HTE14 standard care estimate [18]. While these are preliminary estimates based on modeling assumptions and confidential commercial costs, they support the hypothesis that remote delivery could offer a more resource-efficient way to provide the necessary wraparound support for GLP-1RA therapy, potentially freeing up NHS resources or enabling wider access. A formal health economic evaluation within an NHS setting is required to confirm these potential savings.

The strengths of this study include its relatively large initial sample size, 12-month follow-up duration, and real-world setting, providing insights beyond those from controlled trials. The use of connected scales offers objective weight data, and the collection of data on retention, engagement, side effects, behaviors, and experience provides a multifaceted evaluation.

However, the findings of this study have several limitations. The observational design without a comparator group inherently limits the ability to definitively attribute observed weight loss solely to the intervention, as secular trends or unmeasured participant factors could play a role. This design issue is compounded by the substantial participant withdrawal rate (206/341, 60.4%) by 12 months. Consequently, the primary effectiveness analysis relies on data from completers (179/341, 52.5% of the initial cohort), a group likely to differ systematically from those who dropped out (eg, in motivation, medication tolerability, or initial success). This reliance on a completer analysis significantly limits the generalizability of the effectiveness findings. An ITT analysis with appropriate imputation for missing data, while beyond the scope of this service evaluation, would be necessary to estimate effectiveness across all individuals who initiated the program.

Generalizability is also influenced by the nature of the participant cohort. As it is a commercial program requiring self-payment, participants primarily enrolled via digital marketing, search engines, or word-of-mouth referrals. This self-selection process likely yielded a group with sufficient financial resources and motivation, potentially differing from the broader NHS patient population. The high proportion citing cost as a reason for withdrawal underscores this potential selection bias and suggests that retention patterns and possibly outcomes might differ substantially in a publicly funded model. Relatedly, while initial health conditions were reviewed during onboarding (partly via NHS Summary Care Record access), detailed baseline data on comorbidities (including type 2 diabetes status), ethnicity, or concurrent medications were not accessible for this specific analysis, restricting a full understanding of the cohort’s characteristics and factors influencing outcomes.

Methodological limitations also extend to data collection and measurement. Behavioral changes were assessed using self-reported data from nonvalidated questionnaires, introducing potential recall and social desirability bias. Furthermore, the absence of baseline behavioral data precludes a direct assessment of change from preintervention habits. Program engagement was measured by Home screen views, which is an unvalidated metric of engagement. More granular data on the use of specific features such as coaching interactions or self-monitoring tool entries were not included in this analysis but may offer avenues for future investigation. Although connected scales provided objective weight data, the real-world setting meant weighing frequency and conditions were not strictly standardized.

The evaluation also lacked data on broader clinical outcomes (eg, metabolic parameters) or patient-reported outcomes, such as quality of life. In addition, participants could transition to a separate nonmedication program, and these individuals were excluded; thus, the analysis does not capture outcomes for those who may have stopped medication but continued behavioral support under a different structure. Moreover, while the last available weight was used for inactive participants with 12-month data, the specific weight trajectory immediately preceding withdrawal for the majority who became inactive earlier was not analyzed.

Finally, although the 12-month duration exceeds that of many digital health evaluations, it may be insufficient to assess the long-term sustainability of weight loss and behavioral changes, particularly concerning potential medication cessation effects. Future research should aim to address these limitations, for instance, through designs incorporating matched controls or randomization, standardized measurements using validated instruments, comprehensive baseline and outcome data collection, ITT analyses, and extended follow-up periods.

This real-world service evaluation suggests that a remotely delivered, GLP-1RA–supported weight management program is an effective, feasible, acceptable, and cost-effective approach for achieving weight loss and improving health behaviors in people living with obesity. These programs could address the current service provision issues with specialist in-person services, but further research is needed to evaluate their use within a public health care setting, assess their impact on other clinical and psychological measures, and evaluate sustained effects with continuation and discontinuation of the medication.