Abstract
The conceptualization of ADHD has changed over the years, from motoric disinhibition in DSM-II (hyperkinetic reaction), inattention in DSM-III (attention-deficit disorder, ADD), and both inattention and hyperactivity in DSM-IIIR and DSM-IV [1] (attention-deficit/hyper-activity disorder, ADHD). However, while DSM-IIIR considers ADHD to be a unitary condition, DSM-IV conceptualizes ADHD as having three subtypes: inattentive (ADHD-I), hyperactive-impulsive (ADHD-HI) and combined (ADHD-C). Subtype classification is based on the presence of six or more symptoms of hyperactive impulsive behaviour, inattentive behaviour or both. ICD-10 [2], on the other hand, does not include these subtypes. Instead, it defines a category of hyperkinetic disorder, characterized by severe, pervasive inattentiveness, impulsivity and overactivity. ICD-10 and DSM-IV are very similar in the behaviours that are considered to be the basis for the diagnosis but vary in the rules for weighting them. Impulsiveness and hyperactivity are pooled in DSM-IV but are considered separately in ICD-10. However, in ICD-10, impulsive/hyperactive behaviours alone do not justify the diagnosis; inattentive restlessness must also be present. Additionally, criteria for inattention and hyperactivity/impulsivity, as well as impairment, have to be met in more than one situation (e.g. school, home) for ICD-10, whereas for DSM-IV, impairment from the symptoms has to be present in two or more settings. Hence, hyperkinetic disorder in ICD-10 resembles ADHD-C in DSM-IV, while the DSM-IV definition picks up a more common, less distinctive pattern.
Modifications to the diagnostic system should not be seen in isolation because they can have wide implications. For example, changes in diagnostic criteria seem to have had a major impact on the number of children diagnosed with ADHD [3], [4], the use of medication [5], educational practices and social perceptions [6]. This is particularly the case for disorders like ADHD, which rely on cut-offs for symptoms and impairment, rather than on pathology findings for its recognition. In this line, Canino et al. [7] showed that requiring a rating of <69 on the Children's Global Assessment Scale, as well as Diagnostic Interview Schedule for Children (DISC) DSM-IV criteria for diagnosis, halved the prevalence of DSM-IV ADHD from 8% to 3.7%.
The aims were to examine the validity of the DSM-IV subtypes of ADHD, especially because 10 years have elapsed since the publication of DSM-IV. While symptoms may persist into adulthood, adult data was not considered since it is scarce and would make the review unnecessarily complex. Likewise, the review focuses on studies which provide information for the various subtypes; data referring only to ADHD-C or to general issues about the diagnosis of ADHD are not considered unless relevant to subtype validity.
Method
Studies published in English relating to the different subtypes of ADHD from 1994 to 2004 were reviewed after searches of Medline, Pre-Medline, PsycINFO, EMBASE, Web of Science and manual perusal of articles' reference lists.
Reliability of diagnosis
There is limited data on the reliability of ADHD diagnosis using structured interviews. The data available suggests that reliability is fair to good (kappa = 0.48–0.91) [8]. Problems assessing reliability are compounded by the choice of informant. Mitsis et al. [9] studied 74 clinically referred children using the ADHD module of the DISC and found that agreement between parents and teachers was poor for individual ADHD subtypes. Diagnosis based on either parent or teacher report frequently yielded a diagnosis of either ADHD-I or ADHD-HI. However, when cross-informant data were used to form diagnoses, these subtypes became relatively rare, with most cases meeting criteria for ADHD-C, suggesting that parent reports of ADHD behaviours at school are not an adequate substitute for direct teacher input. This could have resulted from genuine cross-situational differences in the child's behaviour in some cases; in other cases, this may have been due to the categorical cut-off of six symptoms for the diagnosis of ADHD subtypes, so that reports that differed by one or two symptoms might have appeared discrepant, although they do not necessarily reflect serious disagreement between the informants.
There is more information on the psychometric properties of scales to measure ADHD symptoms, which has been reviewed recently [10]. Overall, test-retest reliability is reasonably good when the same type of informant (e.g. parent) is considered, but decreases substantially when agreement between different informants (e.g. parent, teacher) is examined. For example, Wolraich et al. [11] studied 243 children with ADHD and found low agreement between parent and teacher reports of ADHD symptoms according to DSM-IV-based questionnaires (inattentive: r = 0.34, hyperactive-impulsive: r = 0.27). When the two-setting requirement was strictly enforced, disagreement between informants decreased diagnostic rates for all three subtypes: ADHD-I from 15% to 5%, ADHD-HI from 11% to 3% and ADHD-C from 23% to 7%.
In summary, overall reliability of ADHD diagnosis when using structured interviews or questionnaires is acceptable, but it can decrease considerably when data from different informants is included, depending on how data are treated. For example, if both informants have to agree to consider a symptom present, reliability and prevalence drops. However, there is practically no data on the reliability of ADHD-I and ADHD-HI.
Phenomenology
Studies using factor analysis to evaluate the structure of DSM-IV ADHD symptoms have found that a twofactor model: (i) inattention; and (ii) hyperactivity/impulsivity, showed the best fit [12], [13]. Latent class approaches towards ADHD classification have been investigated recently in the context of genetic studies [12],[14–16]. While DSM-IV subtypes require positive endorsements of six or more symptoms of inattention and/or hyperactivity-impulsivity, latent class assignment has no analogous threshold cut-off. Rather, individuals are grouped into classes based on similar profiles of endorsement for the 18 ADHD symptoms [15] and take into account subsyndromal states. The validity of categorical versus dimensional approaches to classification has been discussed elsewhere [17] and is not addressed here. Overall, findings of factor analytical and latent class analyses are broadly consistent with the DSM-IV subtypes and support their face validity.
Comorbidity
Internalizing disorders
Earlier studies, mostly DSM-III based, suggested that there were differences in the patterns of comorbidity with internalizing disorders between attention deficit disorder with and without hyperactivity, with the latter showing a greater prevalence of comorbid anxiety and depression than the former [18], [19]. However, recent DSM-IV-based research has called this into question [20–22]. Power et al. [22] for example, reported similar levels of anxiety and depression among children with ADHD-C and ADHD-I when the severity of externalizing disorders was controlled for. Consistent with this, studies including all three subtypes report higher rates of anxiety and depression in those with ADHD-C and ADHD-I compared to those with ADHD-HI [4], [23], [24] suggesting that if differences in comorbidity exist, these refer to ADHD-HI. Faraone et al. [23] also found that those with ADHD-C had higher rates of bipolar disorder and tic disorder.
Externalizing disorders
There is considerable agreement that externalizing disorders like ODD and CD occur most commonly in children with ADHD-C, followed by ADHD-HI and ADHD-I [4],[23–25]. It is unclear, however, which of the various externalizing behaviours is most salient in distinguishing between the three subtypes. Researchers comparing children with ADHD-C and ADHD-I have reported that ODD and CD are more common in those with ADHD-C compared to ADHD-I [20], [21]. However, hyperactivity was not always the most salient distinction. Rather, it was the level of aggression or social deficits that best discriminated between the subtypes [26], [27].
Learning disability
The association between ADHD and learning disorders is well established [28]. Much less clear is whether associations vary according to ADHD subtype, although academic problems seem to be more common in those with ADHD-I and ADHD-C compared with ADHD-HI [4], [25], [29]. Faraone et al. [23] also reported that ADHD-C children had higher lifetime rates of language disorders, while Morgan et al. [21] found a higher proportion of children with a mathematical learning disability among those with ADHD-I compared to ADHD-C.
Substance use
People attending substance dependence treatment facilities have a high prevalence of ADHD [30], but comorbidity between conduct disorder and ADHD is high, particularly in clinical populations. The majority of studies conclude that ADHD in the absence of conduct disorder does not increase the risk of drug use [31], [32]. Nevertheless, young people may start using substances earlier when conduct disorder coexists with ADHD [33]. This may not be true for the inattentive subtype of ADHD [34].
Some studies show that adolescents with ADHD are more likely to smoke cigarettes [35], [36] but there is much less information about the association between smoking and specific ADHD subtypes. Tercyak et al. [37] studied a community sample of 1066 10th-grade students and found that a history of smoking and current smoking were associated with clinically significant ADHD-I symptoms.
In summary, there is much data showing that children with all three DSM-IV subtypes, particularly ADHD-C, have increased rates of other disorders. However, evidence about differential associations with specific subtypes is scarce and often conflicting. On the whole, reports suggest that ADHD-HI and ADHD-C are more likely to be associated with aggression and externalizing disorders, while ADHD-I and ADHD-C with internalizing disorders, learning disabilities and smoking.
Epidemiology
Surveys that used teacher or parent rating scales usually resulted in higher prevalence rates of ADHD than those using structured interviews. For example, in studies using rating scales, the average prevalence of DSM-IIIR ADHD was 9.1% [4], [25] while the average prevalence of DSM-IV ADHD was 15.4% [4], [25],[38–42]. In the latter studies, ADHD-I was the most frequent (8.0%), followed by ADHD-C (4.0%) and ADHD-HI (3.4%). Among studies using interviews or other clinical judgements, the average prevalence of DSM-IIIR ADHD was 4.3% [43], [44], while the average prevalence of DSM-IV ADHD was 6.3% [7],[45–47]. ADHD-I was the most common subtype (45% of all those with ADHD), followed by ADHD-C (34%) and ADHD-HI (21%) [46], [47]. That is, the increase in prevalence of DSM-IV ADHD is similar in size to the prevalence of the newly introduced ADHD-I. How impairment is treated also has a very large impact on prevalence estimates. For example, requiring a global functioning rating of < 69 and DSM-IV criteria reduced prevalence rates from 8.0% to 3.7% in one study [7].
These surveys vary considerably in their methodology and it is not valid to average the results. Nevertheless, these data as a whole illustrate that using DSM-IV compared with DSM-IIIR diagnostic criteria increases the estimated point prevalence of ADHD, whether rating scales or interviews are used. As a whole, studies also show that ADHD-I is the most common subtype (about half of all ADHD cases), while ADHD-HI is the least common. Rates of ADHD-C are similar to those of ADHD-I.
Age and gender
Overall, DSM-IV ADHD is more prevalent in males than in females [7],[46–48]. Although prevalence of ADHD as a whole decreases with increasing age [7], [47], [48], prevalence of ADHD-I may increase [46]. The DSM-IV field trials, which included 380 clinic-referred youth aged 4–17 years showed that: (i) ADHD-C participants were more likely to be younger and male; (ii) ADHD-I participants were more likely to be female and older than ADHD-C participants; and (iii) those with ADHD-HI were the youngest [3]. McBurnett et al. [29] and Nolan et al. [41] also found that children with ADHD-HI were more frequently in the younger age group compared to those with ADHD-I, and Biederman et al. [49] reported that ADHD-I was more common among females.
These findings suggest that the three subtypes have slightly different age and gender distributions. This may be due to ADHD manifesting itself differently at various developmental stages: impulsivity and hyperactivity would predominate in younger children, while inattention would be more common in adolescents. Alternatively, since ADHD symptoms may decrease with increasing age, it may be that symptoms extinguish at different rates: impulsivity and hyperactivity declines first, while inattention lessens later. In this line, Biederman et al. [50] studied 128 boys with ADHD over 4 years and found that age was significantly associated with decline in total ADHD symptoms and symptoms of hyperactivity, impulsivity, and inattention. Symptoms of inattention remitted for fewer subjects than did symptoms of hyperactivity or impulsivity. Another possibility is that children with ADHD may move through the subtypes as they grow older, manifesting hyperactive-impulsive symptoms mostly from ages 4–5, both hyperactive-impulsive and inattentive symptoms from ages 6–12, and inattentive symptoms thereafter. More prospective studies of children with the various subtypes of ADHD are required to clarify these issues.
Environmental factors
Psychosocial risk factors
Graetz et al. [47] studied a nationally representative sample of Australian young people and found that all three subtypes were socially disadvantaged compared to controls but ADHD-C was the most clearly linked to social adversity; they were the most likely to live in households where there was a single parent, household income was lower, parents had left school earlier and the percentage of parents in employment was lower. Parent education was higher for ADHD-I and ADHD-HI than for ADHD-C, and parental employment was lower for ADHD-HI and ADHD-C than for ADHD-I.
Children reared in institutions have increased rates of inattention and overactivity suggestive of ADHD [51]. These effects seem to be specific (e.g. not accounted for by low birth weight, malnutrition or cognitive impairment) and persist over time. It is possible that institutional deprivation may cause a syndrome similar to ADHD. However, there are no data linking this to a specific subtype.
ADHD and the family
Podolski and Nigg [52] examined role-distress and coping in mothers and fathers of 66 children aged 7–11 with ADHD and found that parents of children with ADHD-C and ADHD-I expressed more role dissatisfaction than parents of controls. For mothers, both children's inattention and oppositional-conduct problems contributed to role distress, while for fathers, role distress was only associated with children's oppositional or aggressive behaviour. Peris and Hinshaw [53] studied 131 girls aged 6–12 with ADHD and found that high parental expressed emotion, particularly criticism, was associated with both ADHD and aggression in ADHD-C and ADHD-I.
Genetic factors
Behavioural genetic studies
Faraone et al. [54] studied 140 boys with ADHD, 120 controls and their first-degree relatives, and found that rates of ADHD among relatives of each subtype were greater than among relatives of controls. ADHD-HI was found almost exclusively among relatives of ADHD-HI probands but this was not the case for ADHD-C and ADHD-I. However, Smalley et al. [55] did not find evidence of family specificity for individual ADHD subtypes in a cohort of 132 affected sibling pairs and their relatives.
Hudziak et al. [12] investigated ADHD symptoms in 1549 female adolescent twin pairs. They reported that an eight-latent class model best accounted for the symptom variance and latent classes. These classes roughly correspond to ADHD-I, ADHD-HI and ADHD-C. Neuman et al. [14] compared latent class findings from the sample of female adolescent twins used by Hudziak et al. [12] to 430 male and 430 female children and adolescents. They found that among the female twins, approximately 80% of MZ and 52% of DZ twins were concordant for class membership, while among the male and female children and adolescents, 51% of female and 41% of male siblings were in the same latent class, a rate approximately equal to that detected among the DZ female twins, suggesting that ADHD class membership is familial to a large extent. Todd et al. [16] and Rasmussen et al. [15] also found that ADHD subtypes defined by DSM-IV and latent class criteria exhibited significant same subtype clustering among twin and twin/non-twin sibling pairs in Missouri and Australia, respectively.
Molecular genetic studies
ADHD appears to be associated with abnormalities in the dopaminergic system, with DAT1, DRD4 and DRD5 being implicated as major susceptibility genes. Waldman et al. [56] studied 122 children and families and found that levels of hyperactive-impulsive symptoms were related to the number of DAT1 high-risk alleles. Siblings discordant for the number of high-risk alleles differed markedly in their levels of both hyper-active-impulsive and inattentive symptoms, suggesting linkage of ADHD with DAT1, especially for the ADHDC subtype. Lowe et al. [57] studied 14 probands and their parents and found an association between ADHDC, ADHD-I and the DRD5 locus. Todd and Lobos [58] also found a significant association between the DRD4 3′exon 3-repeat allele and a novel talkative-impulsive latent-class-defined subtype of ADHD. However, studies examining the association of ADHD with the DAT1 3′VNTR polymorphism [59], DRD4 5′120-base-pair polymorphism [60] and DRD2 [58] have yielded negative results. Qian et al. [61] studied 202 nuclear ADHD families, 340 ADHD cases and 226 controls in China and found an association between the low enzyme-activity catechol-O-methyl-transferase (COMT) Met allele and boys with ADHD-I, and an association between the Val allele and females with ADHD. Manor et al. [62] initially found an association between COMT polymorphism and ADHD-C but subsequently failed to replicate this finding. In a recent study, Todd et al. [63] also found a significant association between the 5′intron 2-single-nucleotide polymorphism of the nicotinic acetylcholine receptor alpha-4-subunit gene (CHRNA4) and severe inattention problems.
The rapidly growing body of research on the genetics of ADHD shows that this condition is to a large extent inherited. Although it is still unclear whether high heritability applies to the various subtypes, latent class studies suggest that subtypes tend to breed true, and that they may be associated with specific chromosomal abnormalities
Biological factors
Neuropsychological deficits
Earlier studies of attention deficit disorder without hyperactivity (DSM-III) suggested that children with this condition had deficits in information retrieval and processing, were less alert, and had more memory problems than those with attention deficit disorder with hyperactivity. Factor analysis showed that these features clustered together in a factor which was subsequently called ‘sluggish cognitive tempo’ [64]. However, recent research suggests that symptoms of sluggish cognitive tempo do not discriminate between DSM-IV subtypes [65].
Children with ADHD have been found to consistently exhibit poorer performance on measures of executive function, vigilance and perceptual speed but usually perform within normal limits on a variety of verbal or spatial measures [66]. Children with ADHD-C and ADHD-I have deficits in inhibition, processing speed, vigilance and planning. In addition, children with ADHD-C also have deficits in conceptual reasoning. Children with ADHD-HI, on the other hand, do not seem to exhibit significant cognitive deficits [67–70].
Motor ability
Piek et al. [71] found that boys with ADHD had significantly poorer motor ability than controls and a higher rate of movement difficulties consistent with developmental coordination disorder. Those with ADHD-I had poorer fine motor skills, while those with ADHD-C had poorer gross motor skills. Severity of inattentive symptomatology was found to be a significant predictor of motor coordination difficulties. Other studies [72], [73] suggest that boys with inattentive symptomatology (ADHD-C or ADHD-I) have poorer motor skills.
Neurophysiological studies
Quantitative electroencephalographic (EEG) techniques show by and large that children with ADHD have increased alpha and beta band frequency power and decreased P300 amplitude than controls, although these findings have low sensitivity and specificity [74]. Later studies have also found neurophysiologic differences between the various subtypes. Clarke et al. [75] reported that EEG power changed faster in children with ADHDC compared to those with ADHD-I, with power levels becoming similar with age, supporting a two-component model of ADHD, with the hyperactive-impulsive component maturing with age and the inattentive component remaining stable. Johnstone et al. [76] found differences in residual event-related potentials between children with ADHD-C, ADHD-I and controls, suggesting differences in stimulus processing and regional activation or inhibition.
Neuroimaging studies
Neuroimaging studies indicate dysfunction of the frontostriatal circuits in patients with ADHD, implying dysregulation of the dopaminergic and noradrenergic systems [77]. However, evidence about specific subtypes is lacking.
Response to treatment
Most of the controlled trials include only children with DSM-IIIR ADHD or DSM-IV ADHD-C, the best example being the NIMH Collaborative Multi-site Multi-modal Treatment Study of Children with ADHD [19]. Several clinical trials examining the efficacy and tolerability of various types of stimulant and non-stimulant medication included participants with either two [78], [79] or three [80], [81] of the DSM-IV subtypes of ADHD. However, treatment response was reported collectively and not according to subtype. In the studies that analysed results according to subtype, Stein et al. [82] found that in children and adolescents with ADHD-C, higher doses of long-acting methylphenidate were associated with increased reduction of inattention and hyperactivity symptoms. However, in those with ADHD-I, symptom improvement occurred at lower doses, with less benefit derived from higher doses.
Psychosocial treatments for ADHD abound [83]. However, interventions usually focus on symptoms such as aggression and off-task behaviour in heterogeneous diagnostic groups and the results cannot be generalized to the specific ADHD subtypes. A recent study including 103 stimulant-responsive 7–9 years-old children with DSM-IIIR ADHD without learning and conduct disorders treated for 2 years with stimulants and multimodal psychosocial treatment showed that multimodal psychosocial treatment did not enhance academic achievement, emotional adjustment [84], social functioning [85] or decrease symptoms [86]. Multimodal psychosocial treatment included parent training or family therapy, academic organizational skills training, individualized academic assistance, academic remediation, social skills training and individual psychotherapy. However, this does not necessarily mean that psychosocial treatment is ineffective in children with the other two subtypes. Antshel and Remer [87] found that social skills training led to improved assertiveness in children with ADHD, but did not affect other domains of social competence. Children with ADHD-I improved more than those with ADHD-C.
Impairment
Psychosocial functioning
Research on clinic-referred samples has shown that ADHD subtypes may exhibit different types of impairment [88]. In the DSM-IV field trials, Lahey et al. [3] found that ADHD-C and ADHD-HI children were more globally impaired than ADHD-I children, while ADHD-C and ADHD-I children had more academic problems than those with ADHD-HI and controls. Several studies have also shown that ADHD-C children were the most impaired in multiple domains. Children with ADHD-I had more academic problems, while those with ADHD-HI had more social and behavioural problems [4], [25], [29], [39], [40], [47]. Faraone et al. [23] found that children with ADHD-C were more impaired than those with ADHD-I and ADHD-HI, had higher rates of counseling and multimodal treatment, and were more likely to be placed in special classes.
Academic achievement
Prospective studies of New Zealand and Australian birth cohorts showed an association between earlier ratings of hyperactivity and later adverse educational outcomes, like continuing school difficulties, problems with attention and poor reading in adolescence, but provided no information about the various subtypes [89], [90]. In a community sample of 1154 child and adolescent twins, Todd et al. [91] found that those with ADHD-C and ADHD-I showed significant deficits in cognitive and achievement testing, worse grades and increased use of special education resources compared to those with ADHD-HI and controls.
Onset of impairment
Applegate et al. [92] found that nearly all youth who met symptom criteria for ADHD-HI displayed impairment before the age of 7, whereas 18% of youth with ADHD-C and 43% of youth with ADHD-I did not manifest impairment before 7 years. Parents also reported that the first symptom of ADHD emerged 2.5 years earlier than the onset of impairment. This may suggest (i) that symptoms of ADHD typically do not create impairment until several have emerged, or (ii) that some children manifest impairment later than others. The latter is likely to be due to variations in situational demands (e.g. at school) rather than variations in the ontogeny of the syndrome.
In summary, each of the three ADHD subtypes shows unique patterns of impairment. Children with ADHD-C are impaired in most domains, those with ADHD-I tend to have more academic problems and manifest impairment later, while children with ADHD-HI seem to have more social and behavioural problems and show impairment earlier.
Conclusion
A decade has elapsed since the publication of DSM-IV and much research has been published since then. Yet, evidence supporting the validity of ADHD-I and ADHD-HI is still scarce. It is apparent that the new classification has resulted in an increase in the prevalence of ADHD, which seems to largely reflect the addition of ADHD-I. There are indications that diagnosis of ADHD-C is more reliable than the other two subtypes, although specific data on the reliability of ADHD-I and ADHD-HI are lacking. Factor analytical studies of symptom patterns and some genetic research support the validity of the distinction between the three subtypes, although this research also questions the DSM-IV rules for diagnosis (e.g. number of symptoms). Studies examining comorbidity, correlates and neuropsychological deficits show complex and often contradicting results. Overall, they suggest that ADHD-HI in particular may be different from the other two subtypes. For example, learning problems and neuropsychological deficits are less common in this group than in the other two. Symptoms of ‘sluggish cognitive tempo’, believed to be specific to ADHD-I, do not seem to discriminate between the subtypes. Given that inattention is the hypothesized ‘core’ symptom of ADHD, it remains to be demonstrated that hyperactive-impulsive children are part of the same condition. ADHD-HI might be more closely related to the oppositional-conduct disorder construct than to the inattentive-hyperactive one. Long-term prospective studies of young children with the three subtypes of ADHD are needed to answer some of these questions, for example, whether the subtypes largely reflect a developmental progression, whether they persist unchanged as children grow up, or whether they have similar outcomes in adulthood.
One of the main research deficits refers to the treatment of ADHD-I and ADHD-HI. Almost all treatment trials are based on participants with ADHD-C. While other subtypes are included in recent studies, results are consolidated for reporting. Although some trials describe a significant reduction in symptoms of inattention and hyperactivity-impulsivity in children with ADHD-C, it cannot be assumed that children with ADHD-I or ADHD-HI would obtain the same benefit until this is shown to be the case. For example, it is possible that children with ADHD-HI may not benefit as much from medication, or that the optimal dosage required may vary according to subtype, or that they may respond better than ADHD-C to psychosocial treatments.