Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors present a retrospective analysis of weight-loss effects due to the eHealth app zanadio in two large cohorts.

The overall rationale is clear.

Introduction: Fine.

Methods: Generally fine.

Please specify the frequency of data collection for body weight (monthly?). This is highly relevant for the LOCF analysis.

Results: Generally fine.

Please emphasize the massive drop-out already in the first 3-month term; please report attrition in plain numbers and in percentage of total cohort. (not just in the discussion)

Table 2, second half: The "n" should be identical for all four terms of the risk factor cohort (4314) and the CVD cohort (1418), respectively. LOCF is continued throughout the entire observation period, not just until the first missing visit.

Please shorten decimals in accordance to plausible raw data precision, e.g. body weight with one decimal, only.

Please report, how many individuals in which cohort achieved clinically relevant weight loss (>5%) and how many maintained that success until the end of the observation period.

Discussion: Can only be evaluated after major revision.

Author Response

Reviewer 1

Comment 1: Please specify the frequency of data collection for body weight (monthly?). This is highly relevant for the LOCF analysis.

Response 1: Users are encouraged to self-monitor their body weight on a weekly basis as part of their therapy but can also do so at any time. This enabled us to analyse weight changes at months 3, 6, 9, and 12 at the end of each prescription phase. We have clarified this in Materials and Methods section 2.2 “The zanadio Intervention”. Analysis of weight tracking engagement shows that patients on average logged weight >1 time per week indicating that patients followed the recommendation. These details have been added to section “3. 1. Treatment Persistence and Weight Tracking”.

Comment 2: Please emphasize the massive drop-out already in the first 3-month term; please report attrition in plain numbers and in percentage of total cohort. (not just in the discussion)

Response 2: We have added a new paragraph to the results section which explicitly details treatment persistence across the prescription phases. Figures 2 and 3 now additionally contain the plain numbers of patients included in the analysis.

Comment 3: Table 2, second half: The "n" should be identical for all four terms of the risk factor cohort (4314) and the CVD cohort (1418), respectively. LOCF is continued throughout the entire observation period, not just until the first missing visit.

Response 3: We thank the reviewer and apologise for the ambiguity. We have clarified both the analysis and the table legend. For the LOCF analysis, each patient's last recorded weight is carried forward to month 12, and the linear mixed model is fitted on the full cohort (N=4,314 for the risk-factor sample and N=1,419 for the CVD sample); the analysis n is therefore constant across the observation period, as the reviewer requests. 'Number of prescription codes' enters the model as a between-patient stratification factor representing programme exposure, not as a within-patient visit. Accordingly, the figures in the 'Number of patients' column of Table 2 are not a diminishing per-visit n but the sizes of these mutually exclusive exposure subgroups, which necessarily differ and together comprise the full cohort. We have revised the Table 2 legend to state this explicitly and to clarify that the last observation is carried forward to month 12 within every subgroup. We retained the exposure-stratified presentation because collapsing it would remove the dose-response relationship between programme exposure and weight change that is central to this analysis.

Comment 4: Please shorten decimals in accordance to plausible raw data precision, e.g. body weight with one decimal, only.

Response 4: This has been amended across the results section

Comment 5: Please report, how many individuals in which cohort achieved clinically relevant weight loss (>5%) and how many maintained that success until the end of the observation period.

Response 5: We thank the reviewer for this valuable suggestion. We have added a dedicated responder analysis to the results section (Section 3.4, 'Responder Rates,' Figures 4 and 5), reporting the proportion of patients achieving 3%, 5% and 10% weight loss, stratified by number of prescriptions and reported separately at months 3, 6, 9 and 12. Responder rates were derived from the LOCF values, so that a patient classified as a responder at a given timepoint is one whose last recorded weight met the respective threshold, providing a conservative estimate. In the established CVD cohort, the proportion reaching the clinically relevant threshold of ≥5% weight loss increased with program exposure and was sustained in patients continuing treatment: among patients with at least 4 prescriptions, ≥5% responder rates plateaued at approximately 55–57% from month 6 onward (55.2% at month 6, 56.7% at both months 9 and 12), while the proportion reaching ≥10% continued to rise to 35.8% at 12 months. A comparable pattern was observed in the risk-factor cohort, where ≥5% responder rates plateaued between approximately 53% and 60%, and ≥10% responses reached 30.4% at 12 months in the ≥4-prescription group. Regarding maintenance specifically: because these rates are based on last-observation-carried-forward values extended to month 12, the stability of the ≥5% responder rate between months 6/9 and 12 in patients with continued prescriptions indicates that, for those remaining in the program, the clinically relevant response was largely sustained to the end of the observation period. We present these as cross-sectional responder proportions under the LOCF assumption rather than as an individual-level maintenance analysis, and we note in the limitations that LOCF assumes no change after the last recorded weight.

Reviewer 2 Report

Comments and Suggestions for Authors

The reviewed article addresses the important issue of modern obesity therapy using IT tools, monitoring applications, and lifestyle support – non-pharmacological treatment.

The article comprehensively presents the research methodology, study group, study results, and discussion. The literature is carefully selected. The authors emphasize the existing limitations of the study and the need for periodic clinical examination of participants rather than self-reporting of data, which may be subject to error.

Furthermore, the average age of the participants may indicate that not all participants were sufficiently proficient in using the application. This should be noted in the limitations, which could also have influenced the study results.

However, the reported average weight loss of over 5% with non-pharmacological treatment of obesity lasting 6-12 months is a good result and, as the authors noted, may translate into modification of cardiovascular risk factors.

The study requires further long-term follow-up to determine the percentage of obesity relapse and study adherence.

An interesting topic that could contribute to improving the care of patients with obesity, adhering to recommendations, and motivating them to change their lifestyle. In this context, it is worth citing the TERMINOS study: "Pilot Clinical Trial of Time-Restricted Eating in Patients with Metabolic Syndrome": Świątkiewicz I., Mila-Kierzenkowska C., Woźniak A., Szewczyk-Golec K., Nauszkiewicz J., Wróblewska J., Rajewski P., Eussen J.P.M. S., Faerch K., Manoogian N.C. E., Panda S., Taub R. P.: Nutrients 2021, 13,346. https://doi.org/10.3390/nu13020346
Świątkiewicz, Iwona, Jarosław Nuszkiewicz, Joanna Wróblewska, Małgorzata Nartowicz, Kamil Sokołowski, Paweł Sutkowy, Paweł Rajewski, Krzysztof Buczkowski, Małgorzata Chudzińska, Emily N. C. Manoogian, and et al. Feasibility and Cardiometabolic Effects of Time-Restricted Eating in Patients with Metabolic Syndrome. Nutrients (2024);16, no. 12: 1802. https://doi.org/10.3390/nu16121802

Author Response

Reviewer 2: 

Comment 1: The average age of the participants may indicate that not all participants were sufficiently proficient in using the application. This should be noted in the limitations, which could also have influenced the study results.

Response 1: This is an important observation. We now describe this in the limitations section. 

Comment 2: The study requires further long-term follow-up to determine the percentage of obesity relapse and study adherence.

Response 2: We thank the reviewer for this observation. We agree that long-term follow-up data are needed to assess sustained weight loss and obesity relapse rates beyond the 12-month observation window. This is acknowledged in the limitations and future directions sections of the manuscript, where we note that further studies linking app-based data with routine clinical data and longer follow-up are required to assess long-term cardiometabolic outcomes.

Comment 3: An interesting topic that could contribute to improving the care of patients with obesity, adhering to recommendations, and motivating them to change their lifestyle. In this context, it is worth citing the TERMINOS study: "Pilot Clinical Trial of Time-Restricted Eating in Patients with Metabolic Syndrome": Świątkiewicz I., Mila-Kierzenkowska C., Woźniak A., Szewczyk-Golec K., Nauszkiewicz J., Wróblewska J., Rajewski P., Eussen J.P.M. S., Faerch K., Manoogian N.C. E., Panda S., Taub R. P.: Nutrients 2021, 13,346. https://doi.org/10.3390/nu13020346
Świątkiewicz, Iwona, Jarosław Nuszkiewicz, Joanna Wróblewska, Małgorzata Nartowicz, Kamil Sokołowski, Paweł Sutkowy, Paweł Rajewski, Krzysztof Buczkowski, Małgorzata Chudzińska, Emily N. C. Manoogian, and et al. Feasibility and Cardiometabolic Effects of Time-Restricted Eating in Patients with Metabolic Syndrome. Nutrients (2024);16, no. 12: 1802. https://doi.org/10.3390/nu16121802

Response 3: We thank the reviewer for highlighting this excellent pilot study. We have included it in the discussion. 

Reviewer 3 Report

Comments and Suggestions for Authors

The authors presented a manuscript entitled “Real World Self-Reported Weight Loss during the Use of a Lifestyle Therapy Prescription Digital Therapeutic in Adults with Obesity and Self-Reported Cardiovascular Risk Factors or Cardiovascular Disease”, which addresses a highly relevant, timely, and clinically useful topic: the real-world effectiveness of a prescription digital therapeutic (zanadio), for weight management in adults with obesity who also present with self-reported cardiovascular disease (CVD) or cardiovascular risk factors.

One of the strengths is the large retrospective sample included for risk factors and for established CVD), the authors explore percentage weight change over a 12-month timeline. They employ appropriate statistical methods, comparing a Missing-at-Random approach via linear mixed-models with a conservative Last Observation Carried Forward (LOCF) sensitivity analysis. The manuscript is well-structured, the introduction sets a strong background and the discussion properly contextualizes the clinical utility of lifestyle modifications.

However, before publication several methodological gaps, reporting limitations, and risks of bias must be addressed to elevate the rigor of the paper.

The abstract states that attrition is typical for weight loss programs, but this volume of missing data introduces a massive healthy-user or selection bias (only highly motivated responders continue tracking weight). The authors must explicitly include a comprehensive analysis or a flowchart in the main body (not just as a supplementary file). They must compare the baseline characteristics (age, initial BMI, gender, comorbidities) of those who dropped out versus those who completed 12 months to determine if data truly meets MAR assumptions or if it is Missing Not At Random.

 

Also, the authors note that missing weight logs do not inherently mean non-compliance with the app modules. To make this study robust, the authors should provide data on engagement metrics (e.g., number of app logins per week, number of completed therapeutic modules). Correlating weight loss with actual module completion rather than just the administrative metric of "prescription codes" would vastly improve the paper's scientific impact.

On the other hand, given the massive adoption of GLP-1 receptor agonists globally, a patient simultaneously taking semaglutide could completely confound the weight-loss results attributed to the "zanadio" app. The authors must acknowledge this as a major limitation. If possible, a sub-analysis excluding all known medication users should be emphasized.

In general, correct some typographical errors through the text: Ex: Line 195, AVOVA.

Author Response

Reviewer 3

 

Comment 1: The abstract states that attrition is typical for weight loss programs, but this volume of missing data introduces a massive healthy-user or selection bias (only highly motivated responders continue tracking weight). 

Response 1: We thank the reviewer for this important observation. We agree that the volume of attrition introduces a potential responder bias that warrants explicit acknowledgement beyond the general statement in the abstract. We have addressed this in several places in the revised manuscript. First, we added a responder analysis stratified by “number of prescriptions” to the manuscript (see Methods and additional Results Section “3.4 Responder Rates”). In the limitations section, we have added that the attrition pattern may reflect a responder bias whereby individuals who continued tracking weight and renewing prescriptions likely represent those with a favourable biological response to treatment who are also a more motivated and engaged subset of the broader population prescribed zanadio. In the discussion, we have added a sentence following the reporting of persistence rates noting that those contributing weight data at later timepoints are potentially a cohort of individuals with a favourable biological response who are also highly motivated, and that the reported weight loss outcomes may not be generalised to all patients prescribed zanadio.

Comment 2: The authors must explicitly include a comprehensive analysis or a flowchart in the main body (not just as a supplementary file). 

Response 2: We have added an additional paragraph to the results section which describes the attrition in more detail across each prescription phase. Figures 2 and 3 now contain the number of patients with a weight recording at months 3, 6, 9 and 12 alongside the weight changes.

Comment 3: They must compare the baseline characteristics (age, initial BMI, gender, comorbidities) of those who dropped out versus those who completed 12 months to determine if data truly meets MAR assumptions or if it is Missing Not At Random.

Response 3: The patient characteristics at first, second, third and fourth follow-up have been added to Table 1. 

Comment 4: The authors note that missing weight logs do not inherently mean non-compliance with the app modules. To make this study robust, the authors should provide data on engagement metrics (e.g., number of app logins per week, number of completed therapeutic modules). Correlating weight loss with actual module completion rather than just the administrative metric of "prescription codes" would vastly improve the paper's scientific impact.

Response 4: We thank the reviewer for this constructive suggestion. We agree that engagement metrics such as module completion rates and login frequency would provide valuable context and would strengthen the mechanistic interpretation of the findings. While a full correlational analysis of engagement with weight outcomes falls outside the scope of the current retrospective study, we have acted on this comment by explicitly acknowledging in the limitations section that granular engagement data were not incorporated into this analysis, that prescription codes as a proxy for program exposure do not capture depth of individual therapeutic engagement, and that future analyses linking engagement metrics with weight outcomes are needed. In addition, weight tracking behavior was analysed, which is described in the Methods section and reported in the Results section “Weight Tracking Persistence”

Comment 5: On the other hand, given the massive adoption of GLP-1 receptor agonists globally, a patient simultaneously taking semaglutide could completely confound the weight-loss results attributed to the "zanadio" app. The authors must acknowledge this as a major limitation. If possible, a sub-analysis excluding all known medication users should be emphasized.

Response 5: We thank the reviewer for raising this important point. We agree that concomitant GLP-1 receptor agonist use is the most material potential confounder, and we address it directly here and in an expanded limitations section. Two considerations bound its plausible magnitude in the present cohort.

First, drug availability. Semaglutide at the obesity-licensed dose (Wegovy, 2.4 mg) did not become available in Germany until July 2023, whereas all patients received and activated their first zanadio prescription between 15 November 2020 and 14 August 2022. The only GLP-1 therapy licensed for obesity during the study period was liraglutide 3.0 mg (Saxenda), self-reported by only n=2 patients in the entire cohort. Semaglutide was available as Ozempic, but was licensed and reimbursed solely for type 2 diabetes; any use for weight management would have been off-label.

Second, reimbursement structure. zanadio is a DiGA reimbursed within the statutory health insurance (GKV) system and prescribed by a panel physician (Kassenarzt). Under §34 SGB V, GLP-1 receptor agonists prescribed for obesity are classified as lifestyle agents (Abnehmmittel) and excluded from GKV reimbursement, rendering them self-pay only; off-label prescription of a diabetes-licensed GLP-1 such as Ozempic for weight loss additionally exposes the prescriber to recourse (Regress) within the statutory system. A clinician operating within the GKV framework to prescribe a reimbursed DiGA is therefore structurally unlikely to issue a concurrent off-label GLP-1 script for the same indication. This does not exclude privately funded use, or legitimate on-label use in patients with comorbid type 2 diabetes — neither of which we can detect, and which we now acknowledge explicitly — but it makes widespread concomitant semaglutide use for weight loss, the scenario the reviewer rightly flags,implausible at a magnitude that would "completely confound" the observed effect during this window.

We have expanded the limitations section accordingly, both to acknowledge that undetected concomitant medication use remains a limitation and to provide this contextual clarification on drug availability.

Regarding the proposed sub-analysis, excluding the small number of known self-reporters (n=2, Saxenda) would not resolve the uncertainty around unreported use, and would risk creating a misleading impression of having controlled for a confound that remains, by its nature, unquantifiable. We therefore prefer to treat it transparently as a limitation rather than imply false precision.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The authors took the comments into account and made changes that significantly improved the manuscript. 
The presentation of the tables could still be improved to make them easier to read. 

Author Response

Comment: The presentation of the tables could still be improved to make them easier to read.
Response: We thank the reviewer for this observation. All tables have been reformatted for improved readability, including landscape orientation where needed, revised column headers for clarity and conciseness, and improved spacing and layout throughout.