Mikel Gomes1, Kristine Færch1, Maja Bramming1, Mads Reinholdt Sørensen1, Anders Strathe1
1Novo Nordisk A/S
Introduction: It is hypothesized that an exposure-response weight predictor algorithm developed from randomized controlled trial data can predict long-term individual body weight changes in response to subcutaneous semaglutide treatment for weight management using self-reported, real-world data, collected through an app-based patient support solution. Moreover, it is well documented that the response to GLP-1 treatment differs between sexes and therefore it is relevant to investigate the prediction error for each sex, when predicting the response to treatment. Methods: Data from one 52-week, phase 2 trial and two 68-week phase 3 trials with semaglutide in people with overweight or obesity have previously been used to develop an exposure-response model [1]. In the current study, that model was applied to a real-world data set from patients prescribed semaglutide (Wegovy®) by their treating healthcare professional and enrolled in a digital patient support program (WegovyCare®) from January 2023 to May 2024. In addition to baseline data (sex and body weight), the app allowed users to track semaglutide doses and body weight, serving as model inputs. The predictive ability of the model was evaluated in 2 scenarios: 1) a half-year treatment scenario (26 +/- 4 weeks), where model forecasts were updated at weeks 4, 8 and 16; and 2) a full-year treatment scenario (52 +/- 4 weeks), where model forecasts were updated at weeks 8, 16 and 28. Model bias was estimated as the mean predicted body weight minus the mean observed (i.e., self-reported) body weight. Median absolute weight loss and inter-quartile range (IQR) was also compared by sex. Results: A total of 1,797 WegovyCare® app users from Denmark (87.6%), Switzerland (6.9%), and Germany (5.5%) were included in the study at baseline. The population was 81% women, with mean (SD) self-reported baseline bodyweight of 102 (17.8) kg in women and 119 (20.1) kg in men. The mean (SD) age was 47.1 (11.7) years for women and 51.8 (11.5) years for men. The dataset had a gradual decline in the number of individuals represented over time as users enrolled and discontinued app use. In the half-year scenario, 596 persistent WegovyCare® users (479 female, 117 male) had data eligible for analysis. The mean final body weight in this subset was 86.9 kg for women and 103.2 kg for men, corresponding to 16% and 14% reductions, respectively. The median absolute weight loss and IQR was 15.5 [12.4, 19.7] kg for women and 15.9 [13.0, 20.9] kg for men. Model bias was low for both women and men (ranging from 1.0 to 1.6 kg in women and from –0.6 to 0.5 kg in men) and improved with time after successive model updates. The mean (SD) final therapeutic dose was 1.4 (0.6) mg. In the full-year scenario, the mean final body weight reported by 206 persistent WegovyCare® users (174 female, 32 male) was 81.1 kg for women and 105.7 kg for men, corresponding to a 22% and 15% weight loss, respectively. Median absolute weight loss and IQR was 22.6 [16.9, 27.9] kg for women and 18.5 [14.8, 24.1] kg for men. Again, model bias was low for both women and men (ranging from –0.9 to 1.0 kg in women and –1.4 to 0.8 kg in men). The mean (SD) final therapeutic dose was 1.6 (0.7) mg. Conclusion: We demonstrated that long-term weight loss can be predicted with adequate precision and low bias from self-reported information on sex, dose, and body weight during initial treatment with semaglutide. Even though females lost more weight than males, the prediction error appeared to be similar between sexes. Patients treated with weight management medications as well as health care professionals may benefit from having a weight predictor integrated into a patient-support solution to better guide decision-making and dose adjustments towards individual treatment targets. Conflict of Interest: All authors are employed by Novo Nordisk A/S. KF, MG, MB, MRS, and AS hold shares in Novo Nordisk A/S. Funding: No external funding was obtained for this project.
[1] Strathe et al. DOM 2023; 25 (11): 3171-3180.
Reference: PAGE 33 (2025) Abstr 11449 [www.page-meeting.org/?abstract=11449]
Poster: Drug/Disease Modelling - Other Topics