Secondary Analysis of Treatment Effect Heterogeneity of priovi in Borderline Personality Disorder

The digital therapeutic priovi has small average effects for borderline personality disorder (BPD). In this preregistered secondary analysis, we aimed to explore the heterogeneity and potential moderators of these effects. Adults received priovi plus treatment as usual or treatment as usual alone. At the 3-month primary endpoint (N = 528), model-based recursive partitioning showed heterogeneous effects on Borderline Symptom List-23 (BSL-23) change (p = 0.022) and on clinically relevant response (p = 0.007). Benefit was larger at higher baseline symptom load: a representative tree split on Patient Health Questionnaire-9 at a score of 23, and the higher-score subgroup had roughly twice the BSL-23 reduction (conditional average treatment effects −0.40 versus −0.19). Baseline BSL-23, depressive symptoms, and recent life-threatening events ranked as the leading moderators. Quartile calibration of predicted CATEs was weak and non-monotonic. The moderators are hypothesis-generating and should not guide patient selection without external validation.