Post hoc analysis: early changes in ADHD-RS items predict longer term response to atomoxetine in pediatric patients

Data from 5 atomoxetine trials in pediatric outpatients with attention-deficit/hyperactivity disorder (ADHD) were divided into training and validation data sets to develop models predicting atomoxetine treatment response, using changes in individual ADHD Rating Scale (ADHD-RS) items early in treatment. Treatment response was predicted after 1 week by a > or =1-point score decrease in ADHD-RS item 15 ("easily distracted;" positive predictive values [PPVs]: 84.9%, 74.3%, and 73.3%; negative predictive values [NPVs]: 52.6%, 50.5%, and 46.3%; training and 2 validation data sets, respectively); after 2 to 3 weeks, by a > or =1-point score decrease in ADHD-RS item 1 ("fails to give close attention or makes careless mistakes;" PPV = 77.7% and 77.9%) and by the absence of a > or =1-point score decrease on ADHD-RS items 1 and 10 ("on the go;" NPV = 72.2% and 77.5%), or by the combination of items 1 and 10 (PPVs: 75.1% and 75.4%; NPVs: 72.2% and 77.5%; training and validation data sets, respectively).