Poster B32, Sunday, March 25, 8:00-10:00 am, Exhibit Hall C
Computational modeling as a tool for detecting medication response in ADHD
Mads Pedersen1,2, Michael J. Frank1, Sigurd Ziegler2, Mats Fredriksen3, Guido Biele4; 1Brown University, 2University of Oslo, 3Vestfold Hospital Trust, 4Norwegian Institute of Public Health
Stimulant medication reduces symptoms in ADHD, but response rates are only around 60-70%, and non-responders are typically identified only after several months. Recent studies have examined the influence of patients' neurobiological, clinical, and social characteristics on medication response, but no clinically useful predictors have been identified. Cognitive processes captured by computational modeling are likely closer to adaptive functioning than neurobiological variables and can thus potentially provide a bridge between less specific global measures of functioning and fine-grained neurobiological characteristics. To test the potential benefit of using computational modeling for predicting medication response we applied a modified drift diffusion model to data on Conners’ Continuous Performance Test (CPT) II collected in a prospective open-label study on adult ADHD-patients. 250 patients followed during their first year of medication treatment were tested on the CPT before starting medication treatment, and at multiple time points during treatment. At the end of the study, 160 patients (70%) continued on medication. To measure potential differential effects of medication we retrospectively grouped patients based on medication status at the one-year endpoint. After 6 weeks of treatment, patients who at end-of treatment responded to medication showed substantially stronger positive effects of medication on performance in the CPT compared to non-responders. Specifically, performance improved more for responders through an increased rate of evidence accumulation (Bayes Factor: 32.99) and extended decision threshold (Bayes Factor: 25.49). The results reported here show the potential benefit of using computational modeling as a tool for predicting medication response in ADHD.
Topic Area: EXECUTIVE PROCESSES: Monitoring & inhibitory control