ADHD: New Approaches to Subtyping and Nosology

A perennial and seemingly intractable nosological problem for ADHD has been whether to consider this condition as a single, unitary disorder or as comprising important subtypes—or even sub-disorders. Up until 1980, there was only one disorder in the nosology (variously named minimal brain damage; hyperkinetic reaction of childhood, and other terms). However, in 1980 DSM-III (American Psychiatric Association, 1980) introduced ADHD with and without hyperactivity. ADHD without hyperactivity was not operationally defined, but the implication was that those children could be impulsive, but not hyperactive. This was conceivable at that time because under DSM-III, ADHD had three behavioral dimensions. However, in 1987, after extensive factor analyses, DSM-III-R (American Psychiatric Association, 1987) abandoned the three-dimensional structure and any mention of subtypes. It was left to DSM-IV (American Psychiatric Association, 1994) to propose a revised picture with three subtypes based on a rational division of a two-factor structure (inattention and hyperactivityimpulsivity). The predominantly inattentive type was similar to the DSM-III ADHD without hyperactivity, except that these children were expected to be below threshold on a single hyperactivity-impulsivity dimension, rather than on hyperactivity. Crucially, they were not defined as being “without” hyperactivity (or hyperactivity-impulsivity) but only as being below the threshold for the combined subtype—meaning they could have up to 5 symptoms of hyperactivity-impulsivity. This led to substantial dissatisfaction among critics who noted that normative hyperactivity-impulsivity in children was well below 5 symptoms, so that some children in this group were still more hyperactive-impulsive than normal. Further problems arose with the recognition that the subtypes were not temporally stable, and their biological distinctions were faint (reviewed in detail by Willcutt et al., 2012). However, in the absence of a compelling body of data supporting an alternative structure, and in view of the need to convey heterogeneity in some fashion, DSM5 (American Psychiatric Association, 2013) opted only to soften the subtype definitions by repositioning them as presentations, leaving it to future editions to replace this nosological structure with a superior description of ADHD’s heterogeneity. An additional problem, alluded to by Willcutt and colleagues (Willcutt et al., 2012), is that the pattern of results in many cognitive, neuropsychological, and biological studies of the DSM-IV ADHD subtypes has been one consistent with a severity model. That is, if we assume that ADHD is a continuous dimension or two continuous dimensions, then arbitrarily cutting these into “types” will simply create a mild and a severe group. One of my objections to many findings about ADHD subtypes was that on measures of neuropsychological functioning, a consistent picture was that the ADHD combined type performed significantly worse than the ADHD inattentive type, which in turn performed worse than controls. This typical picture is seen in our data in a recent report (Nikolas & Nigg, 2013). Figure 1 illustrates the problem schematically. Panel A (Severity) shows a typical finding: the ADHD inattentive type scores in between the controls and the ADHD combined type. Contrary to what many publications have concluded, this is not evidence of valid subtypes. Rather, it is evidence of a continuous dimension of severity that has been arbitrarily divided into subtypes. This is because the ADHD combined type has more symptoms than the ADHD inattentive type. To conclude that a group with significantly more symptoms has significantly more neuropsychological problems (putting it cynically, significantly more symptoms) verges on the tautological or else the trivial. Needed in my view has been evidence that a group with fewer symptoms has worse performance on a validator—this would truly be evidence of the configural variation that characterizes true subtypes, rather than simply recapturing the communication convenience provided by the arbitrary cut on a true dimension. Panel B (Configural) illustrates this hypothetical situation— on some biological probes, the putative subtype with fewer ADHD symptoms nonetheless shows worse performance. The same success would be achieved if two ADHD types had similar numbers of symptoms, but different profiles of weakness on a panel of probes. Our group has approached this problem from two directions at once. In the first approach, we use neurophysiological measurements to attempt to evaluate the clinical proposals—such as a “pure inattentive” group or a callousunemotional group. In the second approach, we use empirical clustering methods and then evaluate their validity with cross-validation using physiological and clinical measures. In this report, I summarize key recent findings that may stimulate clinical thinking, research, and discussion.