Developing a Risk Score to Guide Individualized Treatment Selection in Attention Deficit/Hyperactivity Disorder

Objective: To develop a risk score for treatment failure that could potentially be used to individualize treatment selection between lisdexamfetamine dimesylate (LDX) and osmotic-release oral system methylphenidate (OROS-MPH) in children and adolescents with attention deficit/hyperactivity disorder (ADHD).

Methods: The study used data from patients with ADHD receiving LDX (N = 104) or OROS-MPH (N = 107) in a phase III randomized clinical trial. A prediction model was developed to estimate risk scores for failing OROS-MPH, where treatment failure was defined as less than 25% improvement in the ADHD Rating Scale IV total score from baseline. Patients were ranked by their predicted risks of OROS-MPH failure to define high-risk subpopulations. Outcomes of LDX and OROS-MPH were compared within subpopulations.

Results: The prediction model for OROS-MPH failure selected seven predictors (age, disease duration, and five ADHD Rating Scale IV item scores) and had an in-sample C statistic of 0.860. Among all patients, LDX had a 17% (95% confidence interval 7.1%-27.8%) lower treatment failure rate than that of OROS-MPH; differences in failure rates ranged from 17% to 43% across subpopulations, increasingly enriched for high-risk patients. Similar heterogeneity across subgroups was observed for other efficacy measures.

Conclusions: In the overall trial population, LDX was associated with a lower rate of treatment failure compared with OROS-MPH in patients with ADHD. A more pronounced benefit of LDX over OROS-MPH was observed among subpopulations with a higher predicted risk of failing OROS-MPH. The present study showed the feasibility of individualizing treatment selection. Future research is needed to prospectively verify these results.