Predicting Methylphenidate Response in Children With ADHD: Theoretical, Empirical, and Conceptual Models COLIN B. DENNEY, PH.D, AND MARK D. RAPPORT, PH.D. ABSTRACT ObJective: To evaluate the theoretical merit and empirical validity of models designed to predict response to methylpheni- date (MPH) among children with attention-deficit/hyperactivity disorder (ADHD). Method: Seventy-siX children with ADHD received each of 4 counterbalanced doses of MPH (5, 10, 15, and 20 mg) in the context of a double-blind, placebo-con- trolled, within-subject (crossover) experimental design. Logical and conceptual foundations of 3 models of MPH response were subjected to critical scrutiny. and patterns of relationship anticipated on the basis of these models were subjected to empirical analysis. Results: The conceptual foundations of all reviewed models were found to be sUbstantially flawed, and none provided an adequate empirical basis for predicting response to MPH among children with ADHD. Conclusions: The observed pattern of relationships suggests that magnitudes of response to MPH in domains of classroom attention and behavioral disinhibition are correlated and differentially predictive of response on measures of academic performance and teacher-rated behavior. J. Am. Acad. Child Ado/esc. Psychiatry, 1999,38(4):393--401. Key Words: methylphenidate, attention-deficit/hyperactivity disorder, titration. Methylphenidate (MPH) has an enviable record as a safe and effective treatment for children with attention-deficit/ hyperactivity disorder (ADHD) (for reviews, see Barkley, 1998; Gadow, 1992). Not all children respond to MPH, and in many cases, benefits are observed in some domains but not others (Pelham and Milich, 1992). The concep- tual and clinical implications of this frequently reported observation have fueled continuing interest in identify- ing predictors of MPH response. For example, Loney (1986) and others (e.g., Ross and Ross, 1982) have argued that establishing predictive models of MPH response may facilitate identification of distinct, etiologically homo- geneous subgroups of children with ADHD and fore- stall the time and expense of controlled clinical drug trails. Studies designed to derive and comprehensively eval- uate models predicting drug response are necessarily complex and must include several methodological fea- tures. First, guiding theoretical or conceptual founda- Accepted September 23. 1998. Dr. Rapport is Proftssorof Clinical Child Psychology and Chair of Graduate Studies, and Dr. Denney is a postdoctoralftllow, Department of Psychology, University of Hawaii, Honolulu. Reprint to Dr. Rapport, Department of Psychology, Gartley Hall, 2430 Campus Road, University of Hawaii, Honolulu. HI 96822. 0890-8567/99/3804-0393©1999 by the American Academy of Child and Adolescent Psychiatry. ]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 38:4, APRIL 1999 tions that embrace multiple levels of abstraction and analysis are of great benefit, as they facilitate interpreta- tion of either positive or null findings. Second, data should be collected that enable comparison of models differing in rationale. Third, multiple doses should be administered in the context of a repeated-measures cross- over design to evaluate the generality of predictors across a wide dose range. Finally, to evaluate predictive power rigorously at the clinical level, it is necessary to use statis- tically robust methods for determining responder status and optimal doses rather than relying solely on clinical judgment or arbitrary cutoff scores. The scope of these criteria helps to explain the pau- city of research attention given to prediction of MPH response in children with ADHD and should temper criticisms of the literature that have been advanced in this area. Previous studies concerning prediction of MPH response in children are reviewed below. Empirical Models Taylor and colleagues (1987) adopted an atheoretical approach emphasizing the empirical identification of variables capable of distinguishing drug responders from nonresponders prior to drug administration. Thirty- eight boys were categorized into responder and non- responder groups on the basis of clinical judgment after 393