AIM: To evaluate the 12-month effectiveness and patient persistence of a semaglutide-supported digital weight-loss service (DWLS) in a real-world German cohort.
METHODS: This retrospective study analysed 4535 patients who initiated the Juniper Germany DWLS. Primary endpoints included 12-month medication and data submission adherence (receiving ≥ 10 orders) and mean weight-loss percentage. Effectiveness was assessed via a per-protocol (PP) analysis of an adherent sub-cohort (n = 546) and an intention-to-treat (ITT) analysis of the full cohort using Last Observation Carried Forward (LOCF) and a Worst-Case Scenario (WCS) sensitivity model. Binary logistic and multiple linear regressions identified predictors of attrition (churn) and efficacy.
RESULTS: Of the total cohort (n = 4535), mean 12-month weight loss was 9.47% (LOCF). However, sensitivity analysis using a WCS model-assuming 0% weight loss for all dropouts-yielded a mean of 2.40%. Among the 12% of patients (n = 546) who remained adherent to both medication and data submission protocols, mean weight loss was 15.13% (±7.1). Program persistence was significantly bolstered by achieving a healthy BMI milestone (OR = 0.29, p < 0.001), encountering a weight-loss plateau (OR = 0.41, p < 0.001), and higher baseline obesity severity (Obesity Class I-III: p < 0.023). Conversely, high clinical complexity (≥ 3 comorbidities) more than doubled the risk of churn (OR = 2.12, p = 0.002), while previous GLP-1 RA use also significantly increased churn risk (OR = 1.57, p = 0.001). While moderate early engagement (2-10 logs) was a strong predictor of persistence (p < 0.001), intensive tracking (> 10 logs) did not significantly reduce churn risk (p = 0.393).
CONCLUSION: The population-level impact of this DWLS is fundamentally limited by high attrition (88%) in an unsubsidised environment. While significant weight loss is possible for a non-representative adherent minority, the mean effectiveness for the total initiating population (2.40% WCS) highlights that real-world success is strictly conditional upon long-term persistence. The study found that both clinical success and setbacks were associated with program adherence. It also observed that patients of high clinical complexity were at greater risk of attrition. Finally, it found that high-friction manual weight logging serves as a marker of commitment rather than anxiety as had been reported in app-based DWLSs where tracking has significantly lower friction. These findings suggest that the predictive value of digital engagement is fundamentally shaped by DWLS delivery mode and the behavioural cost of engagement.
