Abstract
Differential response to the Systems Training for Emotional Predictability and Problem Solving (STEPPS) program was compared in subgroups identified through latent class analysis (LCA). STEPPS is an evidence-based group treatment program for patients with borderline personality disorder (BPD). A reanalysis of data was conducted using data from a 20-week randomized controlled trial and 1-year follow-up. Subjects (n = 164) with DSM-IV BPD were assessed for comorbid Axis I and II disorders and selected clinical variables. Severity was assessed using the Zanarini Rating Scale for BPD (ZAN-BPD) and the Borderline Evaluation of Severity Over Time (BEST). Three- and four-class models were identified with the four-class model having the better fit. The latter included a high severity (HS) class (26%), an affective instability/substance abuse (AISA) class (16%), an empty/dissociation/identity disturbance (EDID) class (27%), and a low severity (LS) class (30%). High impulsiveness predicted membership in the HS class. Improvement was determined using a linear mixed-effects model. Those most likely to benefit were those in the HS group characterized by high symptom severity, Axis I and II comorbidity, problem relationships, abandonment fears, and intense anger. This work should help further efforts to match patients with treatments based on sociodemographic, diagnostic, and other illness characteristics.