Impulsive aggression in adults with attention-deficit/hyperactivity disorder

Abstract

Dowson JH, Blackwell AD. Impulsive aggression in adults with attention-deficit/hyperactivity disorder.

Objective:  DSM-IV criteria for attention-deficit/hyperactivity disorder (ADHD) include examples of ‘impulsivity’. This term can refer to various dysfunctional behaviours, including some examples of aggressive behaviour. However, impulsive aggression is not included in the DSM-IV criteria for ADHD. The associations of impulsive aggression with ADHD were investigated.

Method:  Seventy-three male adults with DSM-IV ADHD, and their informants, completed questionnaires. Impulsive aggression was assessed by ratings of two criteria for borderline personality disorder (BPD), involving hot temper and/or self-harm.

Results:  Logistic regression indicated that features of DSM-IV ADHD were predictors of comorbid impulsive aggression. However, compared with ADHD features, verbal IQ and comorbid psychopathology were more strongly associated with impulsive aggression.

Conclusion:  The findings support the inclusion of features of impulsive aggression, such as hot temper/short fuse, in the ADHD syndrome in adults. These overlap with features of BPD. The findings inform the selection of research samples.

Significant outcomes

• • 
  In male adults with ADHD, the main domains of DSM-IV ADHD were predictors of comorbid impulsive aggression.
• • 
  Compared with ADHD features, verbal IQ and comorbid psychopathology were more strongly associated with impulsive aggression.
• • 
  In adults, impulsive aggression can be considered as a part of the ADHD syndrome, which overlaps with features of borderline personality disorder. The findings are relevant for the selection of research samples.

Limitations

• • 
  There is limited reliability for current assessments of ADHD.
• • 
  Heterogeneity of subjects with ADHD can involve ADHD features, IQ, comorbidity and previous treatment.
• • 
  The present study did not include women.

Introduction

There is uncertainty about the status of impulsive aggression as a feature of the syndrome of attention-deficit/hyperactivity disorder (ADHD), because aggression is found in the context of disorders which are commonly comorbid with ADHD, such as conduct disorder (CD) and, in adults, antisocial and borderline personality disorders (1). The syndromal context of impulsive aggression in subjects with ADHD may have important implications for the selection of research samples and the efficacy of drug treatments and other interventions which have a specific role for the treatment of ADHD.

The DSM-IV ADHD criteria (2) include examples of ‘impulsivity’. This term can refer to several variably co-occurring behaviours with different neurobiological substrates, which are features of several psychiatric syndromes (3–6). Impulsivity has been considered to be reflected by some examples of impaired focusing on a task, rapid poorly planned responses to stimuli, lack of perseverance and disorganization.

Impulsivity has also been associated with some forms of aggression (7) and one definition of impulsivity is ‘a lowered threshold for motoric actions, particularly aggressive behaviour, in response to environmental stimuli’ (8). Aggression can be reflected by verbal aggression (i.e. expressions of anger and/or threats of violence to self or others), aggression against property, autoaggression (i.e. various forms of self-harm) and physical aggression to others (9). Although impulsive aggression is not included in the DSM-IV criteria for ADHD, aggression has often been reported in subjects with ADHD (10–13) and is included in the ‘Utah’ criteria for adult ADHD (14). Also, aggression in subjects with ADHD has been reported to be modified by drug treatments for ADHD (15, 16).

Various associations with aggression have been reported in subjects with ADHD. In children with ADHD, impulsivity and emotional lability predicted adult antisocial behaviour (17) and, in children and young adults, the DSM-IV ‘combined’ ADHD type predicted more self-harm and antisocial behaviour compared with the ‘inattentive’ type (18, 19). In addition, ADHD has been associated with increased risk of suicidal ideation in female adolescents (20) and of attempted and completed suicide in male adolescents and young adults (19, 21, 22). In another study in children, ADHD comorbid with CD predicted more physical aggression compared with ADHD alone, while comorbid oppositional defiant disorder (ODD) predicted more verbal aggression (23, 24). However, there are conflicting claims about the associations of ADHD with aggression in the absence of comorbid disorders which involve aggression, in particular, of CD and, in adults, of DSM-IV ‘cluster B’ personality disorders (PDs) (25). The latter include borderline personality disorder (BPD), some features of which can be considered as examples of impulsive aggression, i.e. temper outbursts, hitting people, throwing things and some examples of repeated self-harm. Although a follow-up study to adulthood of clinic-referred boys found that CD, but not ADHD, significantly predicted antisocial PD (26), another follow-up study involving official arrest data found that even ADHD boys with ‘low defiance ratings’ had significantly higher offender rates than normal controls (27).

Aims of the study

The aim was to investigate associations of impulsive aggression in adults with ADHD, to evaluate the validity of impulsive aggression as a part of the ADHD syndrome in adults and to inform the selection of subjects for future research into treatments for ADHD.

Material and methods

Subjects

Male patients were sequentially recruited from referrals to a psychiatric outpatient clinic for the assessment and management of ADHD in adults aged 18–65. Patients were referred from primary care services for the local catchment area as well as from other psychiatric services in the region. Patients received a DSM-IV diagnosis of ADHD, based on a clinical interview with a psychiatrist, informants’ ratings of past childhood and subjects’ and/or informants’ ratings of adult behaviour. Only a minority of potential subjects who attended (<10%) did not provide informants and were excluded as potential subjects. About 25% of referrals did not meet the present criteria for ADHD.

Exclusion criteria were i) a verbal intelligence quotient (IQ) of ≤90 as estimated by the National Adult Reading Test (NART) (28), ii) significant visual or motor impairment, iii) history of (or current) pervasive developmental disorder, neurological disorder (including brain injury and tic disorders), bipolar disorder, schizophrenia or other psychotic disorders, and iv) current major depressive disorder. Criteria ii–iv were evaluated within a semi-structured routine assessment by a psychiatrist.

Clinical assessments

In addition to an interview by a psychiatrist, which was part of the process of diagnosis (see below), questionnaires based on DSM-IV ADHD, described and evaluated by Barkley and Murphy (29), were completed by each patient, by an informant who had known the patient since childhood (usually a parent, but in a minority of cases another family member) and by an informant who had known the patient since the previous 6 months (usually a partner, parent, friend or sibling). The questionnaires included observer- ratings and self-ratings for the three sets of DSM-IV ADHD criteria, i.e. inattention (nine items), hyperactivity (six items) and impulsivity (three items), for past behaviour when aged between 5 and 12 years (by the informants regarding childhood) and for adult behaviour in the previous 6 months (by the patients and informants regarding adulthood).

All subjects in the study received a DSM-IV diagnosis of ADHD, according to the following criteria which have been validated by previous studies (30–32): i) DSM-IV ratings from an informant in relation to childhood features endorsed a specified number (see below) of the nine criteria for the ‘predominantly inattentive type’ (PIT) and/or of the nine criteria for the ‘predominantly hyperactive-impulsive type’(PHIT), ii) DSM-IV ratings by the patient and/or informant in relation to behaviour in the previous 6 months endorsed a specified number of the nine criteria for the PIT and/or for the PHIT, which corresponded to endorsements in relation to childhood (see below) and iii) a judgement made by a psychiatrist, based on all information, that the patient’s symptoms of ADHD often interfered with ability to function and were not explained by the presence of another disorder.

For the diagnosis of the PIT, PHIT or the ‘combined’ type of ADHD, a minimum number of six endorsed criteria were required for each of the relevant sets of nine DSM-IV criteria in both childhood and adulthood. For a diagnosis of ‘ADHD in partial remission’, the minimum number of endorsed DSM-IV criteria in relation to recent behaviour (but not related to childhood behaviour, where it remained at six) was reduced to three in each set of nine criteria. (The DSM-IV concept of ‘ADHD in partial remission’ relates to an unspecified number of endorsed criteria, which is less than that required for diagnosis of the full syndrome. The present definition, involving at least three endorsed criteria, was arbitrarily chosen as 50% or more of the minimum number of endorsed criteria for diagnosis of the full syndrome.) For the diagnosis of ‘ADHD not otherwise specified’, the minimum number of endorsed criteria for both childhood and adult behaviour was reduced to three in each of the sets of nine criteria. Diagnosis of adult ADHD, for all diagnoses, was contingent on endorsement of the ADHD type (or specified minimum number of criteria) in relation to childhood behaviour which corresponded to the ADHD type related to adult behaviour, e.g. an ‘inattentive’ type diagnosis required at least six endorsed criteria for the informant’s ‘inattentive’ ratings for childhood and at least six endorsed criteria for the subject’s and/or informant’s ratings for recent behaviour.

The informants’ ratings for childhood also included ratings for the 15 DSM-IV CD criteria between the ages of 5 and 18 and the eight DSM-IV criteria for ODD between the ages of 5 and 12. For all these ratings (except for CD), each item was rated ‘never or rarely’, ‘sometimes’, ‘often’ or ‘very often’. ‘Often’ or ‘very often’ were considered as endorsements. For CD, the informants’ ratings for childhood were ‘present/absent’.

Patients were also assessed by the ‘Attention-deficit Scales for Adults’ (ADSA) (33). The ADSA is a self-report questionnaire relating to symptoms of ADHD over the patient’s recent adult life. Fifty four statements are each rated as never/seldom/sometimes/often/always and then scored between 1 (never) and 5 (always).

Scores for nine scales are obtained, for a range of ADHD-related characteristics which are reflected by the names of the scales, i.e. attention-focus/concentration; interpersonal; behaviour-disorganized activity; co-ordination; academic theme; emotive; consistency/long-term; childhood; and negative-social. A total score of ≥172 is considered ‘much above average’ in relation to ADHD characteristics. Patients also completed the self-report ‘Brief Symptom Inventory’ (34), which gave a ‘global severity index’ (GSI). Clinically significant symptoms, related to a range of psychopathology, are indicated by a value of 0.6 and above for males.

Features of DSM-IV PDs (but not PD diagnoses) were assessed by the self-report screening test for the Structured Clinical Interview for DSM-IV Personality Disorder (SCID II) (35, 36). This provided present/absent ratings for all the DSM-IV PD criteria, except for the antisocial PD criteria.

In the present study, DSM-IV BPD criteria 5 and 8 were selected as examples of impulsive aggression. As ratings for these criteria are associated with aggression, as are the ratings for the CD criteria included in the SCID II screening test, these were excluded from the total score of ‘PD features’ for the analyses of associations between impulsive aggression and ‘PD features’.

Impulsive autoaggression was identified if one or both of the following SCID II questions for BPD criterion 5 were endorsed: ‘Have you tried to hurt or kill yourself or threatened to do so?’, ‘Have you ever cut, burned, or scratched yourself on purpose?’

Impulsive externally directed aggression was identified if one or both of the following SCID II questions for BPD criterion 8 were endorsed: ‘Do you often have temper outbursts or get so angry that you lose control?’, ‘Do you hit people or throw things when you get angry?’

The study’s assessment procedures were approved by the Cambridge Local Ethics Committee. All subjects who were approached gave written informed consent.

Data analysis

P ≤ 0.05, two-tailed, was selected to indicate significance in the analyses of rating scale data. As this was an exploratory study, P-values were not adjusted for multiple comparisons. Backward stepwise logistic regression analysis (Likelihood Ratio method) was used to identify variables significantly associated with impulsive aggression (spss Version 12; SPSS Inc., Chicago, IL, USA).

Results

Seventy-three male subjects were recruited, whose characteristics are shown in Table 1. Impulsive externally directed aggression was endorsed in 29 of the 73 subjects and impulsive autoaggression in 34 subjects. Eighteen subjects endorsed both categories of impulsive aggression.

Table 1.   Characteristics (means, SDs) of adult males with DSM-IV attention-deficit/hyperactivity disorder (ADHD)

NART, National Adult Reading Test; ADSA, Attention-Deficit Scales for Adults – Total score, GSI, Global Severity Index of the Brief Symptom Inventory.
Number	73
Age (years)	29.0 (9.0)
NART verbal IQ	110.5 (8.4)
ADHD
Combined	29 (40%)
Inattentive	26 (35%)
Hyperactive-impulsive	2 (3%)
ADHD ‘In partial remission’
Combined	2 (3%)
Inattentive	5 (7%)
Hyperactive-impulsive	0
ADHD ‘Not otherwise specified’	9 (12%)
ADSA Total score	204 (20)
GSI	1.5 (0.7)

Table 2 shows the effects of ADHD features and selected covariates as predictors of impulsive aggression. Significant predictors are shown, controlling for the effects of covariates in each analysis. The predictive effects of a combination of the following three ratings of DSM-IV ADHD features were examined: ratings of inattention, hyperactivity and impulsivity. A separate logistic regression was performed on each such combination of data from the informants for childhood, from the informants for adulthood and from self-reports for adulthood. For each data set, the predictive effects of a combination of the above three ratings without additional variables were examined, as well as their effects relative to, and in combination with, additional covariates previously selected to be of potential relevance. For the informants’ ratings for past childhood ADHD, the selected additional covariates were ratings for past ODD and CD, while for ratings of adult ADHD, the selected covariates were IQ, GSI, age and ‘PD features’.

Table 2.   Significant variables (P ≤ 0.05) remaining in backward stepwise logistic regressions of predictive variables for impulsive aggression in 73 adult males with ADHD

Variables and origins of ADHD data	Predictor variables	Coefficient	Standard error	Wald	P-value	exp(B)
NART, National Adult Reading Test; ADSA, Attention-Deficit Scales for Adults – Total score, GSI, Global Severity Index of the Brief Symptom Inventory; ODD, past oppositional defiant disorder; CD, past conduct disorder; PD features, features of DSM-IV personality disorders excluding antisocial personality disorder and borderline personality disorder criteria 5 and 8; B, estimated regression coefficient; SE, standard error of the coefficient estimate; Wald, the ratio B to SE^2; exp(B), the predicted change in odds for a unit increase in the parameter. Covariates: ODD and CD for informants’ data regarding childhood; IQ, GSI, age and PD features for other data.
DSM-IV ADHD features
Predictors of externally directed aggression
Informant regarding adulthood	Hyperactivity	0.351	0.148	5.61	0.018	1.42
Self-report regarding adulthood	Impulsivity	0.535	0.212	6.34	0.012	1.71
Predictors of autoaggression
Self-report regarding adulthood	Inattention	0.373	0.133	7.93	0.005	1.45
DSM-IV ADHD features and covariates
Predictors of externally directed aggression
Informant regarding childhood	CD	0.350	0.122	8.28	0.004	1.42
	ODD	0.276	0.125	4.85	0.028	1.32
Informant regarding adulthood	IQ	−0.161	0.045	12.83	<0.001	0.85
	GSI	1.600	0.544	8.86	0.003	5.05
	Hyperactivity	0.502	0.211	5.63	0.018	1.65
Self-report regarding adulthood	IQ	−0.149	0.043	12.20	<0.001	0.86
	PD features	0.093	0.028	11.07	0.001	1.10
	Inattention	0.318	0.157	4.13	0.042	1.38
Predictors of autoaggression
Informant regarding adulthood	GSI	1.423	0.435	10.69	0.001	4.15
Self-report regarding adulthood	GSI	1.167	0.448	6.77	0.009	3.21
	Inattention	0.270	0.138	3.86	0.049	1.31
ADSA scales 1–9
Predictors of externally directed aggression
	Scale 6	0.167	0.065	6.55	0.010	1.18
Predictors of autoaggression
	Scale 6	0.209	0.068	9.54	0.002	1.23
	Scale 2	0.098	0.047	4.39	0.036	1.10
	Scale 8	0.300	0.150	3.10	0.046	1.35
ADSA scales 2, 6 and 8 and covariates
Predictors of externally directed aggression
	IQ	−0.150	0.044	11.80	0.001	0.86
	Scale 6	0.187	0.075	6.30	0.012	1.21
Predictors of autoaggression
	Scale 6	0.205	0.067	9.36	0.002	1.23
	GSI	1.181	0.494	5.73	0.017	3.26

The nine ADSA scales were also evaluated as predictors of impulsive aggression. ADSA scales 2, 6 and 8 were significant predictors; these scales were also evaluated in combination with the covariates IQ, GSI, age and ‘PD features’. ADSA scale 2 (‘Interpersonal’) relates to dysfunctional and short-term personal relationships, scale 6 (‘Emotive’) relates to a low threshold for rapid mood change, involving ‘over-reaction’, ‘moods easily triggered’, ‘getting agitated quickly’ and ‘easily excitable’, while scale 8 (‘Childhood’) relates to childhood clumsiness and poor school performance.

In summary, there were more predictors of externally directed aggression than of autoaggression. All the three main domains of ADHD (inattention, hyperactivity and impulsivity) were predictors of externally directed aggression, in particular, hyperactivity and impulsivity; other predictors were past ODD and CD, IQ, GSI, PD features and ADSA scale 6 (‘emotive’). Autoaggression was predicted by inattention (but not by the other two domains), GSI and ADSA scales 6, 2 and 8.

Discussion

In view of the reported gender differences for the features, prevalence and comorbidity of ADHD (1), potential female subjects were excluded from the study, as the number of potential female subjects (i.e. 21) was insufficient for parallel analyses. The mean GSI rating was above the threshold for ‘caseness’, which is consistent with reports of a range of psychiatric comorbidity in adults with ADHD (1). The mean ADSA total score indicated ‘very considerable problems’ related to ADHD (33). Various aspects of the present methodology have been discussed previously (30–32).

The features of BPD which were evaluated as examples of impulsive aggression were predicted by DSM-IV ADHD features and also by ADSA scales 2, 6 and 8 (Table 2); however, unlike the DSM-IV ADHD features, ADSA scales 2 and 6 reflect behaviours which usually incorporate impulsive aggression. The BPD criteria 5 and 8 reflect several characteristics which have been linked to impulsivity, such as poorly planned responses to stimuli and a lowered threshold for action, in particular, aggression (see Introduction). Although the questions related to the BPD criterion involving self-harm do not specify aspects of impulsivity other than aggression (i.e. autoaggression), self-harm in the context of BPD often involves several aspects of impulsivity (37). Table 2 indicates that the three domains of DSM-IV ADHD features (inattention, hyperactivity and impulsivity) can predict aspects of impulsive aggression.

When analysed without covariates, adult ADHD-related impulsivity and hyperactivity predicted temper outbursts/hitting people/throwing, while self-reported adult ADHD-related inattention predicted threats/actual self-harm. Further analyses examined the predictive effects of ADHD in relation to, and compared with, other variables which were selected as potential additional predictors of impulsive aggression. The non-ADHD potential predictors of impulsive aggression were informants’ ratings of past ODD and CD for analyses of predictive effects of past childhood ADHD and, for data on adult ADHD, the selected covariates were IQ, age, ‘PD features’ (but omitting those which incorporated aggression – see Material and methods) and the range of psychopathology evaluated by the GSI. Several studies have found associations of these covariables with aggression and/or ADHD. A history of past ODD, and to a lesser extent past CD, both of which incorporate aspects of aggression, have been reported to be associated with features of adult ADHD (32), verbal IQ has been negatively correlated with aggression in various male samples (38, 39) and features of ADHD have been reported to diminish with increasing age (1). Also, as increased aggression has been associated with comorbidity in ADHD samples (23, 24), the range of comorbidity reflected by ‘PD features’ and GSI were also selected as covariates. It was found that, compared with the features of DSM-IV ADHD, some of these covariates were more significant as predictors of impulsive aggression, i.e. IQ, ‘PD features’, GSI and past CD. Also, for the ADSA scales, IQ was a more significant predictor of impulsive aggression than ADSA scales 2, 6 or 8. However, age was not a significant covariable in any analysis.

These findings suggest that, in adult males, DSM-IV ADHD is associated with features of impulsive aggression. This is consistent with the ‘Utah’ criteria for adult ADHD, which include hot temper with a ‘short-fuse’, explosive short-lived outbursts, a rapid calming-down and transient ‘loss of control’ (14). The present findings are also consistent with studies which have reported overlap between ADHD and BPD involving affect regulation and impulse control (40). Further, the results show that, in future studies of the efficacy of treatment regimens for ADHD, the associations of ADHD-related impulsive aggression with IQ and comorbidity are potentially important variables. Therefore, the non-ADHD significant predictor variables in the present study should be matched or controlled for in future studies of treatments for ADHD.

Various factors may confound generalization of the findings, such as the heterogeneity of ADHD, variations in referral practices to specialist ADHD clinics, limitations of current assessment methods and variations between studies involving recruitment, diagnostic procedures, gender, selection criteria, IQ, previous treatments and comorbidity. It should be noted that there have been several suggested definitions of ‘ADHD in partial remission’, for example, involving level of functioning or a threshold of four endorsed criteria out of each set of nine DSM-IV criteria. The present sample was heterogeneous in respect of concurrent psychotropic medication and previous treatments for ADHD. Other possible confounding factors include an uncertain relationship between some examples of the DSM-IV ‘predominantly inattentive’ type to other ADHD presentations and a recognition that DSM-IV ADHD criteria are less applicable to adults than to children.

The evidence of overlap between the clinical presentations of ADHD and BPD, involving examples of impulsive aggression, indicates the need for further research into the effects of psychostimulants, as well as other drugs with efficacy for ADHD, on impulsivity in the context of other disorders. However, although methylphenidate (MPH) has been reported to be effective in reducing impulsive aggression in subjects with ADHD when this is comorbid with BPD or CD and/or ODD, this effect could just reflect a reduction in a nonspecific enhancement, by ADHD, of the impulsive aggression in the comorbid disorder (12, 15, 41, 42). Moreover, treatments for ADHD do not have an agreed place in the management of BPD, as shown by reviews of drug treatments for BPD which have recommended the first-choice use of non-stimulant drugs, with only relatively modest effects (43, 44). Also, parenteral MPH has been reported to induce severe dysphoria in subjects with BPD (45). Nevertheless, a study of 84 children with CD, with and without ADHD, reported a beneficial effect of MPH on antisocial behaviour related to CD alone (46), while psychostimulants have been claimed to reduce aggression in brain-injured patients (47), developmental disorders (48, 49), autism (50) and in two patients with bulimia nervosa and ‘cluster B’ PD traits (51).These claims may reflect the variety of underlying neurobiological substrates for any example of impulsive aggression. However, subgroups of subjects with impulsive aggression in the absence of ADHD may have specific profiles of neurobiological substrates which are also associated with ADHD and respond to psychostimulants or other ADHD-related treatments.

Acknowledgements

The authors thank the subjects and Lorraine Allen who provided secretarial assistance.