This study was a single-centre randomized controlled trial that took place at the Institute for Nutritional Medicine at the School of Medicine & Health of the Technical University of Munich (TUM). The study protocol has been approved by the local ethics committee (vote number: 45/22 S-NP) and was registered at the German Clinical Trials Register (Registration number: DRKS00025291). All participants had given written informed consent before participation.

Participants were recruited through social media (Facebook, Instagram), an advertising banner in the Munich subways, and flyers distributed to doctor’s offices in the Munich region between March 17th and August 9th 2022. Participants who met the following eligibility criteria were included into the study: adults (women, men), age between 18 and 70 years, BMI between 30.0 and 40.0 kg/m2, no severe diseases (e.g. diagnosed diabetes mellitus, cardiovascular disease, and cancer), ownership of a smartphone. Inclusion criteria were checked through a screening phone call and confirmed during the first study visit. Included participants were randomized to two study groups (ADHOC and EXPECT) with an allocation ratio of 1:1 using the Stat Trek Random Number Generator [9]. The trained study team which enrolled participants did not know the randomization list. Allocation was performed by the database after entering the participants’ data. Participants of both groups attended three study visits in total (baseline: V1, after 12 weeks: V2, after 24 weeks: V3) and received allowance of 25 € for V2 and V3 each.

The DiHA “Oviva Direkt für Adipositas” (Oviva AG, Potsdam, Germany) is available for iOS and Android and delivers a 12-week multimodal weight loss intervention program according to the German guidelines for the prevention and treatment of obesity [10]. In the beginning of the weight loss intervention (for ADHOC at V1, for EXPECT at V2 after 12 weeks of “waiting” period), participants were guided through the app installation by a member of the study team. In the first week of app use, participants received a phone call by a qualified coach employed by the app provider. This call aimed to ensure patients’ safety and the appropriate use of the medical device. Furthermore, the app included a private chatroom to ask questions if needed.

The mode of action of the app-based program included three main elements: self-management, self-monitoring, and education (Fig. 1). As self-management approach, participants could set their daily/weekly goals by choosing ones suggested by the app or by choosing self-appointed ones. For self-monitoring purposes, participants could enter various data on e.g. nutrition, physical activity, and body weight. According to the data entered by the users, automated feedback was generated in form of weight trajectory curves, reminders (e.g. to enter the current weight at least once a week), motivating notifications, and interpretations (e.g. which food categories are under- or over-represented in the diet). For education, learning content (e.g. aetiology of obesity, lifestyle recommendations for weight loss, dealing with relapse) was provided on a weekly basis by text, audio, or video format. Regular learning success controls were offered. In summary, the weight loss program of the app uses different behavioural change techniques clustered by Michie et al. [11].

Screenshots of the study app.

To mimic a real-world setting, participants in the ADHOC group did not get any further advices besides of using the app for 12 weeks. The ADHOC group received the app intervention in the first 12 weeks and was allowed to use the app for further 12 weeks of follow-up. The app was not uninstalled by the study team.

Sociodemographic data was collected at baseline through a standardized questionnaire. Anthropometrics (e.g. body weight, height, and composition) were objectively measured at all three study visits by the study team. Body height was measured to the nearest cm by using a stadiometer (SECA 214, Seca GmbH & Co., KG, Germany) and body weight and body composition was measured in light clothing, with an empty bladder, and without shoes by a bioimpedance analysis scale (BC-418MA, Tanita Europe B. V., Netherlands). The BMI has been calculated as body weight (in kg) divided by the square of body height (in m).

Health-related quality of life was assessed at all three study visits through the validated EuroQol (EQ-5D-5L) questionnaire [12] with five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and five response levels (no problems, slight problems, moderate problems, unable to/extreme problems) for each dimension. Responses were converted to an EQ-5D index value by using a standard EQ-5D-5L value set for Germany. The conversion was done in RStudio with the package “eq5d” by Morton Fraser [13]. In addition, the EQ-5D-5L includes a visual analogue scale from 0 (“The worst health you can imagine.”) to 100 (“The best health you can image.”) at which participants rated their current, subjective perception of health ( = EQ VAS score).

Data on app usability and user acceptance were collected after 12 weeks of app intervention. The two dimensions “perceived usefulness” and “perceived ease of use” (each represented by four items) of the validated questionnaire Technology Acceptance Model 3 (TAM 3) [14] were used to calculate a total score. A higher total score means a more positive judgement of app usefulness and ease of use. By creating tertiles of the total score, completers were divided into three groups (“low” = assessment tends to be negative, “middle” = assessment is neither negative nor positive, “high” = assessment tends to be positive). To assess the app system usability, the validated questionnaire System Usability Scale (SUS) [15] was used. A higher SUS score stands for a more positive judgement of the app system usability. By creating tertiles of the SUS score, completers were divided into three groups (“low”, “middle”, and “high”).

Data on minutes spent on the app per week tracked by the app was used as a proxy for the intensity of app usage.

The primary analysis population was all study participants providing weight data after 12 weeks of intervention ( = completers analysis). Integrity and plausibility checks were performed. Absolute and relative frequency, means, and standard deviations were calculated. For comparison of baseline characteristics between the groups, a Two sample t-test (for normally distributed outcomes) or a Mann-Whitney-U test (for non-normally distributed outcomes) was used. Normality was tested using the Shapiro-Wilk test and by graphical inspection of the distribution in each group. Variance homogeneity was checked by using Levene’s test. For categorical outcomes, Pearson’s chi-squared test or Fisher’s exact test was used. Multiple linear regression analysis adjusted for gender, age, and baseline body weight was conducted for comparison of changes after 12 weeks between the groups and for examining the association between time spent on the app and weight reduction. The 24 weeks data analysis was focussing on changes within the groups and not on group comparisons. For comparisons of the TAM and SUS scores, the Kruskal-Wallis rank sum test was conducted. P-values < 0.05 were considered as statistically significant. All statistical analyses were performed using RStudio (4.1.0). Assuming a 20% dropout, 156 participants are enough to show with a statistical power of 80%, a significance level of 0.05, and a standard deviation of 6%, a weight loss effect of 3% in the intervention group (ADHOC) which is statistically significant compared to the control group (EXPECT). The post-hoc power analysis with G*power [16] resulted in 99.9%.

For missing data on the primary outcome, last observation-carried forward (LOCF) imputation method was conducted. For the LOCF analysis after 12 weeks, the last weight tracked by the app was used as the last observation if measured weight was missing after 12 weeks. For the LOCF analysis after 24 weeks, the last measured body weight at the study centre and/or the last tracked weight by the app (self-reported) was used.