|
 We
believe that a valid classification is an essential step in science. In
medicine, and hence in psychiatry, classification is diagnosis.
—Robins and Guze, 1970, p. 107
As
conceptualized in DSM-IV, attention-deficit/hyperactivity disorder
(ADHD) is a common syndrome of childhood characterized by problems of
inattention and/or hyperactive-impulsive behavior that is frequently associated
with comorbid psychiatric conditions and educational or vocational failure.
The prevalence of ADHD is estimated to be between 3% and 9% in children
and several-fold higher in boys. These prevalence estimates and recent
prescription patterns for stimulant medications have led to controversies
in the lay press regarding possible overdiagnosis and overtreatment of
ADHD.
The
results of modern twin studies suggest that much of the controversy surrounding
the supposed epidemic of ADHD is secondary to fundamental misconceptualizations
about its nature. Notwithstanding the tremendous advances in treatment
and research that have been made possible by the adoption of a standard
diagnostic nomenclature, the DSM approach is categorical. In contrast,
the results of several population-based twin studies from around the world
are most compatible with ADHD representing the extreme of one or more
continua of problems with inattention and hyperactivity-impulsivity (for
example, Levy et al., 1997). If correct, ADHD is akin to common medical
conditions such as hypertension or diabetes in that the criteria for defining
who is ill represent an arbitrary cutoff along a continuum of blood pressure
or serum glucose level, respectively. The definition of a case, then,
becomes entwined with the efficacy of our treatment approaches, whether
pharmacological or behavioral. The conceptualization of ADHD as a quantitative
trait is in keeping with the noncategorical tradition such as represented
by the Attention Problems (AP) scale of Achenbach (1977).
Whether
viewed as a discrete disorder or as a quantitative trait, estimates of
the heritability of ADHD are uniformly high (reviewed by Faraone and Biederman,
1998; Todd, 2000). It is unclear whether the same basic heritable elements
predispose to the presence of problems with attention, hyperactivity,
and impulsivity or whether unique genetic elements contribute to each
problem area. Similarly, it is unclear whether the presence of comorbid
psychiatric problems or learning difficulties is due to pleotrophic genes
that may result in different disorders, to coheritability of genetic elements
predisposing to different problem areas, or to other mechanisms such as
environmental effects or assortative mating among parents.
Which
children are seen in our clinics is determined by complex referral patterns.
Hence, clinically based studies of ADHD have potential biases, particularly
regarding comorbidity. School-based studies overcome some of these difficulties,
but they exclude individuals who are not in school. The most severely
affected cases, and perhaps the most interesting from a genetic point
of view, are thereby undersampled.
A
complementary approach that may overcome some of these difficulties is
to sample individuals from birth records. Though not without possible
shortcomings, birth records–based twin studies have the advantage
of sampling the entire spectrum of possibly affected individuals using
genetically informative designs. Coupled with analytic approaches that
are agnostic regarding the nature of the disorder or the mechanisms of
comorbidity, such population-based twin studies offer a powerful approach
to resolving these issues. We are in the process of collecting data from
such a birth records–based twin sample from the state of Missouri
(Hudziak et al., 1998, 2000). Our initial results from this study suggest
that ADHD is underdiagnosed, has complex mechanisms of comorbidity, and
is genetically heterogeneous.
Our
basic approach has been to collect full DSM-IV symptom information
on all twins irrespective of whether or not they qualify for a diagnosis.
Using a categorical version of traditional factor analysis called latent
class analysis (McCutcheon, 1987), we subdivide the population sample
of twins into mutually exclusive classes based on twin or parent report
of DSM symptoms. Defining topologies of phenotypes in this manner
has the advantage that information about the occurrence of symptom clusters
from different diagnostic domains is extracted without previous conceptualizations
of diagnostic boundaries or patterns of comorbidity. A latent class approach
also naturally tests whether the derived classes represent distinct entities
or points along a continuum of severity. It should be emphasized that
as with all statistical approaches, the results of any latent class analysis
are dependent on the quality of the data analyzed, the sample size, and
the underlying reality of nature.
Our
currently published work is compatible with the presence of 2 latent classes
which identify 2 of the 3 DSM-IV ADHD subtypes: primarily inattentive
and a combined inattentive/hyperactive-impulsive subtype (Hudziak et al.,
1998; Neuman et al., 1999). The majority of individuals in these subclasses
qualify for a diagnosis of ADHD. Classes also identify many individuals
who do not meet DSM-IV diagnostic criteria. These classes generalize
across age and sex of the children as well as across different methods
of ascertainment and data collection and diagnostic nomenclatures (Neuman
et al., 1999). In our larger analysis of female twins, we also identified
a relatively uncommon latent class consistent with the DSM-IV hyperactive-impulsive
subtype of ADHD (Hudziak et al., 1998). This class also contains a number
of individuals who did not meet DSM-IV diagnostic criteria. In
contrast to most clinic-based studies of ADHD, we find that the inattentive
ADHD latent class is not associated with an increased frequency of symptoms
or diagnoses of disruptive behavior disorders, anxiety disorders, or depression.
Although
the presence of these 3 ADHD latent classes is superficially compatible
with the categorical definition of DSM-IV ADHD, the patterns of
symptoms in these classes closely follows patterns of symptoms in other,
less severe, latent classes. We have interpreted this to mean that there
are 2 or more continua or domains (inattentive, hyperactive-impulsive,
or combined) of ADHD problems in the general population and that the latent
classes we observe are points along these continua. More recently, however,
we have found that the individual latent classes appear to “breed
true” in that we do not find the expected increased frequency of
twin pairs having a mild and severe form of latent class membership as
would be predicted by a continuum model. That is, when both twins have
ADHD symptoms, they tend to be in the same latent class. A general model
illustrating the difference between such multiple heterogeneity and continuum
views is shown in Figure
1.
In
summary, our current studies are most compatible with a number of genetically
independent forms of ADHD that differ in severity of the symptom domains
defined in DSM-IV and vary in comorbidity pattern. This suggests
that the use of DSM-IV diagnostic criteria for genetic studies
may be problematic. Perhaps the use of latent class categories or other
phenotype definitions will allow the identification of more genetically
homogeneous subtypes for study. Furthermore, these results may offer an
explanation of why children differentially respond to different treatment
paradigms. Different genetic forms of illness might be expected to respond
to different interventions.
top of page
|
