Does methamphetamine use increase violent behaviour? Evidence from a prospective longitudinal study

Aims: To determine whether violent behaviour increases during periods of methamphetamine use and whether this is due to methamphetamine‐induced psychotic symptoms. Design: A fixed‐effects (within‐subject) analysis of four non‐contiguous 1‐month observation periods from a longitudinal prospective cohort study. Setting: Sydney and Brisbane, Australia. Participants: A total of 278 participants aged 16 years or older who met DSM‐IV criteria for methamphetamine dependence on entry to the study but who did not meet DSM‐IV criteria for life‐time schizophrenia or mania. Measurements: Violent behaviour was defined as severe hostility in the past month on the Brief Psychiatric Rating Scale (BPRS) (corresponding to assault/damage to property). Days of methamphetamine and other substance use in the past month were assessed using the Opiate Treatment Index. Positive psychotic symptoms in the past month were identified using the BPRS. Findings: There was a dose‐related increase in violent behaviour when an individual was using methamphetamine compared with when they were not after adjusting for other substance use and socio‐demographics [cf. no use in the past month: 1–15 days of use odds ratio (OR) = 2.8, 95% confidence interval (CI) =1.6–4.9; 16+ days of use OR = 9.5, 95% CI = 4.8–19.1]. The odds of violent behaviour were further increased by psychotic symptoms (OR = 2.0, 95% CI = 1.1–3.6), which accounted for 22–30% of violent behaviour related to methamphetamine use. Heavy alcohol consumption also increased the risk of violent behaviour (OR = 3.1, 95% CI = 1.4–7.0) and accounted for 12–18% of the violence risk related to methamphetamine use. Conclusions: There is a dose‐related increase in violent behaviour during periods of methamphetamine use that is largely independent of the violence risk associated with psychotic symptoms.

comorbidity; methamphetamine; psychosis; schizophrenia; substance use; Amphetamine

Methamphetamine (also known as ‘ice’ or ‘crystal meth’) is notorious for its association with violent behaviour [1] , [2] . Epidemics of use have been marked by rises in assaults and violent crime and case reports have implicated the drug in homicides [2] , [3] , [4] . Violence associated with methamphetamine use is characterized by its capricious and often bizarre nature, this seeming to be fuelled by methamphetamine‐induced paranoia [4] . With between 14 and 53 million users of the drug world‐wide [5] , the potential capacity of methamphetamine to incite violence is a significant public health concern.

Despite its reputation, evidence falls short of showing a causal link between methamphetamine use and violence [1] . Epidemiological evidence is based on case reports of methamphetamine users behaving violently [2] , [3] , [4] , high rates of methamphetamine use among violent offenders [6] and cross‐sectional studies that show higher rates of violence among people who use the drug compared to those who do not [7] , [8] , [9] . It is not clear from these studies whether methamphetamine use causes violent behaviour or whether people who use the drug are otherwise predisposed to violence (Fig. [NaN] ). Violent behaviour among people who use methamphetamine could be accounted for by pre‐morbid risk factors for violence that are common in illicit drug‐using populations [10] , [11] , [12] . In particular, the childhood antecedent to antisocial personality disorder, conduct disorder, is predictive of both later substance use and later violent offending [13] , [14] .

One way to overcome confounding by pre‐morbid risk factors is to use a within‐subject, or ‘fixed‐effects’, analysis of longitudinal panel data. This type of analysis examines the likelihood of an event (in this case, violent behaviour) during periods when an individual is exposed to a risk factor (e.g. methamphetamine use) relative to when they are not exposed to that risk factor. Examining changes within individuals over time eliminates confounding by pre‐existing individual characteristics and other time‐invariant factors (e.g. heritable traits, personality, gender, previous adverse life events) in so far as these factors do not change over the course of the study. Factors that vary during this time (e.g. changes in other drug use) need to be adjusted for, as in any conventional regression analysis. Fixed‐effects analysis is commonly applied within the econometrics literature [15] , and to a lesser extent within public health research [16] , [17] , [18] , to eliminate the confound of pre‐morbid factors.

A potentially important mediating factor for methamphetamine‐related violence is the paranoid psychotic state that the drug can induce [4] , [19] (Fig. [NaN] ). There is a dose–response increase in psychotic symptoms during periods of methamphetamine use [20] and these symptoms often co‐occur with high levels of hostility [21] . Psychosis is known to increase the risk of violence, this being related to the capacity of delusions to incite anger, increase a person's perception of threat in their environment and their propensity to misinterpret events [4] , [22] , [23] . Understanding whether psychotic symptoms are responsible for methamphetamine‐related violent behaviour is important, because it will determine whether the risk of violence is confined to people who have methamphetamine psychosis or whether the risk extends to the broader population of methamphetamine users.

The aim of the current study was to determine the risk of violent behaviour attributable to methamphetamine use net of pre‐morbid risk factors. We achieve this by applying a within‐subject (fixed‐effects) analysis to longitudinal panel data from a prospective cohort of methamphetamine users [24] . As noted above, in this design each participant acts as their own control, avoiding confounding by pre‐morbid factors that do not change during the study. The relationship between methamphetamine use and violent behaviour was assessed over four discrete non‐contiguous 1‐month periods, while adjusting for concurrent changes in other drug use and socio‐demographics. We also examined whether psychotic symptoms exacerbated the risk of violent behaviour, and whether this could account for the relationship between methamphetamine use and violent behaviour.

A total of 278 participants met DSM‐IV criteria for methamphetamine dependence on entry to the study and none met DSM‐IV criteria for life‐time schizophrenia or mania. DSM‐IV diagnoses were assessed using the Composite International Diagnostic Interview (CIDI) [25] . Participants were selected from the Methamphetamine Treatment Evaluation Study (MATES) cohort [24] . The MATES cohort included 400 people entering community‐based drug treatment services in Sydney and Brisbane, Australia, for methamphetamine use, and 101 methamphetamine users from Sydney who were not in treatment (i.e. recruited through needle and syringe programmes and outreach services) and who screened positive for dependence on methamphetamine. Other inclusion criteria for MATES were being aged at least 16 years, comprehension of English, being willing to participate in follow‐up interviews and not having been in methamphetamine treatment, other in‐patient drug treatment, or in prison, in the month prior to entering the study.

From the MATES cohort, 17 participants were excluded because they did not meet DSM‐IV criteria for methamphetamine dependence on recruitment. A further 59 participants were excluded because they met DSM‐IV criteria for either life‐time schizophrenia or a life‐time manic episode, and 138 participants were excluded because this diagnostic information was not available (i.e. these participants did not partake in the follow‐up interviews when these diagnoses were made). Finally, nine further participants were excluded because they had not used methamphetamine during any of the 1‐month periods analysed in the current study.

A structured interview schedule was administered at baseline and at each follow‐up (3 months, 1 year and 3 years after the baseline interview). Recruitment of the cohort took place in 2006 and 2007, while follow‐up interviews spanned the period from 2006 to 2010. Interviews were conducted face‐to‐face or by telephone. All participants were volunteers who provided informed consent, and who were reimbursed for their time and travel expenses (up to $AU40 per interview). All the participants (n = 278) were re‐interviewed at 3 months and 12 months after entry to the cohort and 82% (n = 230) were interviewed at 3 years. Participants who dropped out at 3 years did not differ significantly from other participants on methamphetamine use (1–15 days: 44 versus 40%; 16+ days 22 versus 16%) or violent behaviour (24 versus 19%). The current study used data on past‐month outcomes measured at each of these four time‐points, totaling 1064 months of data for all the participants combined.

The following measures were taken for the past month at each of the four time‐points in the study.

Violent behaviour was defined as a score of 6 or 7 on the BPRS hostility item in the past month. The BPRS is a semi‐structured psychiatric interview that is used to rate the severity of various psychiatric symptoms on a scale of 1–7, with a score of 1 reflecting no symptom present, scores of 2–3 representing mild/subclinical symptoms and scores of 4+ indicating clinically significant/pathological symptoms [26] —these being rated as either moderate (scores of 4 or 5) or severe (scores of 6 or 7). The probes and anchor point ratings for the BPRS hostility item can be found in the Supporting information, Appendix S1. The anchor point rating for a score of 6 is: ‘has assaulted others, but with no harm likely (e.g. slapped or pushed) or destroyed property (e.g. knocked over furniture, broken windows)’; and for a score of 7: ‘has attacked others with the definite possibility of harming them, or with actual harm (e.g. assault with a hammer or weapon)’. The BPRS is designed specifically to monitor psychiatric symptom severity over time and is used widely in clinical trials [27] . Published quality assurance procedures, which we adopted in our study, ensure high levels of inter‐rater reliability and can be maintained when the scale is used by trained interviewers with pre‐doctoral degrees [27] . Ratings in our study were made by trained interviewers (honours‐level psychology graduates or equivalent) and weekly meetings were held to review BPRS ratings in order to maintain inter‐rater agreement and avoid rater drift [27] . A selection of interviews from the MATES cohort [24] (n = 64) were audio‐recorded and rated by a second interviewer for inter‐rater reliability. Inter‐rater agreement for our definition of violent behaviour was 94%, yielding a kappa of 0.86.

Days of methamphetamine use in the past 4 weeks was measured using the Opiate Treatment Index [28] . Self‐reported abstinence from methamphetamine use was confirmed in a subsample of the entire MATES cohort (n = 83) using hair toxicology, with false reporting of abstinence occurring in only 6% of cases (detailed elsewhere [24] ). The main route of methamphetamine administration (oral, intranasal, smoked, intravenous) during the past 4 weeks was also recorded.

Days of use in the past 4 weeks was measured for other substances, including cannabis, heroin, cocaine, ecstasy, hallucinogens, alcohol and tobacco.

Psychotic symptoms were defined as a score of four or greater on any of the BPRS items of suspiciousness, unusual thought content or hallucinations in the past month [26] . This definition of psychotic symptoms has been used previously to assess psychotic symptoms in methamphetamine‐using samples [29] and yielded inter‐rater agreement of 93% and a kappa of 0.86 in this sample [20] .

These included unemployment, living alone and unstable accommodation (no fixed address, living in a boarding house/shelter, refuge or other temporary accommodation).

Time‐invariant measures were age, sex, duration of methamphetamine use and primary route of methamphetamine administration on recruitment, and a diagnosis of childhood conduct disorder (assessed using a modified version of the Diagnostic Interview Schedule [30] ).

Data were analysed using Stata SE version 11.2 [31] . All tests were two‐sided with significance set at P < 0.05. A random‐effects logit model (xtlogit) was used to assess the relationship between methamphetamine use and violent behaviour over time. The fixed‐effects model option was used to determine the within‐subject effect for methamphetamine use on violent behaviour. The influence of time‐invariant factors on violent behaviour was assessed using a random‐effects model in which the between‐subject effect (the mean score for each participant over their observation period) and the within‐subject effect (the difference between the participant's score at each time‐point and their mean score for the observation period) were modelled separately. Modification of the effect of methamphetamine use on violent behaviour was tested using an interaction contrast between the time‐invariant factor and the within‐subject effect contrast for methamphetamine use.

The main outcome measure was violent behaviour in the past month. The main predictor variable was days of methamphetamine use in the past month (no use, 1–15 days, 16+ days). Time‐varying covariates were psychotic symptoms, other substance use and socio‐demographic factors.

To assess the extent to which the relationship between methamphetamine and hostility was mediated by psychotic symptoms, the ‘explained fraction’ approach, as described by Whitehead et al. [32] , was used: [odds ratio (OR)a – 1] – (ORb – 1)]/(ORa – 1), where ORa represented the OR for the unadjusted relationship between methamphetamine use and hostility, and ORb the relationship between methamphetamine use and violent behaviour after adjusting for psychotic symptoms.

Participants had a mean age of 31.7 years [standard deviation (SD) = 8.1 years]. The majority were male (72%), single (72%) and unemployed (78%). Most were Australian‐born (89%) and nominated English as their preferred language (96%). They had a median of 10 years of schooling (range 6–12), 44% had completed a tertiary technical or trade qualification and 6% had completed a university degree. The majority (76%) had a history of conduct disorder.

All participants met DSM‐IV criteria for methamphetamine dependence in the year prior to entering the study; they had used the drug for a mean of 13.1 years (SD = 7.9 years) and 72% usually injected it. Methamphetamine use occurred during 58% of all observed months. During months of methamphetamine use, participants used the drug on a median of 8 days (range 1–28 days) and injection was typically their main route of administration (79% cf. 14% smoking and 6% snorting or swallowing). Other substance use consisted primarily of tobacco (89% of months, median of 28 days of use), cannabis (57% of months, median of 20 days use) and alcohol (62% of months, median of 6 days use), with other substance use being less common.

Half (51%) of the sample reported violent behaviour during at least one of the observed months, and hostility was present during 20% of the observed months in total (49% of these involved a BPRS score of 6 and 51% a score of 7). Psychotic symptoms occurred among 60% of the sample during at least one of the observed months, and psychotic symptoms were present for 25% of the observed months. The relationship between methamphetamine use and psychotic symptoms is documented elsewhere [24] .

Violent behaviour was 6.2 times more likely to occur when a participant was using methamphetamine relative to when they were not using the drug [95% confidence interval (CI) = 3.8–10.2, P < 0.001]. This relationship was dose‐dependent, with low use (< 16 days in the past month) increasing the odds of violent behaviour fourfold and heavier use (16+ days in the past month) producing a 15‐fold increase (Table [NaN] ).

The relationship between violent behaviour and methamphetamine use, other substance use, socio‐demographics and psychotic symptoms

1 Empty cells indicate variables not included in the model. aSimultaneous fixed‐effects regression including all substance use and socio‐demographic variables. bSimultaneous fixed‐effects regression including only substance use and socio‐demographic variables that showed evidence of a relationship with hostility in Model 1. cSimultaneous fixed‐effects regression including variables in model 2 together with psychotic symptoms. dRelative to fewer than 16 days of use. CI = confidence interval; OR = odds ratio.

The relationship between methamphetamine use and violent behaviour persisted after adjustment for other drug use and socio‐demographic variables (model 1, Table [NaN] ). Heavy alcohol use (16+ days use in the past month) was the only other factor that remained significantly predictive of violent behaviour in this model. Model 2 (Table [NaN] ) shows the relationship between methamphetamine use and violent behaviour adjusting only for heavy alcohol consumption; lower levels of methamphetamine use (1–15 days) tripled the odds of violent behaviour and heavy use (16+ days) predicted a 13‐fold increase. Using the explained fraction approach described by Whitehead et al. [32] , we compared the OR for methamphetamine use and violent behaviour in model 2 to the unadjusted odds ratio and found that heavy alcohol use accounted for 18% of the relationship between low methamphetamine use (1–15 days) and violent behaviour {[(3.8–1) – (3.3–1)]/(3.8–1)} and 12% of the relationship between high methamphetamine use (16+ days) and violent behaviour {[(14.7–1) – (13.1–1)]/(14.7–1)}.

Adding psychotic symptoms to this model showed that they further increased the odds of violent behaviour (model 3, Table [NaN] ). The relationship between methamphetamine use and violent behaviour was only slightly attenuated. Comparing the OR of methamphetamine use and violent behaviour in models 2 and 3 [32] showed that psychotic symptoms accounted for 22% of the relationship between low methamphetamine use levels (1–15 days) {[(3.3–1) – (2.8‐1)]/(3.3–1)} and violent behaviour, and 30% of the of the relationship between heavy methamphetamine use levels (16+ days) and violent behaviour {[(13.1–1) – (9.5‐1)]/(13.1–1)}.

The predicted probability of violent behaviour derived from model 3 is shown in Fig. [NaN] . Without psychotic symptoms, the probability of violent behaviour was 9% when participants were not using methamphetamine, rising to 24% during periods of low use (< 16 days) and 52% during periods of heavy use (16+ days). Psychotic symptoms increased the probability of violent behaviour across all levels of methamphetamine use. There was no significant interaction effect between methamphetamine use and psychotic symptoms (OR = 1.0, 95% CI = 0.9–1.1, P = 0.482).

Conduct disorder increased the overall risk of violent behaviour (OR = 1.9, 95% CI = 1.1–3.6, P = 0.037), but did not modify the relationship between methamphetamine use and violent behaviour. The risk of violent behaviour declined with age, but heavy methamphetamine use (16+ days) produced a slightly greater increase in the odds of violent behaviour among older participants (age by 16+ days use interaction: OR = 1.1, 95% CI = 1.0–1.2, P = 0.033). Other time‐invariant measures were not related to violent behaviour (P > 0.05, analyses not shown).

There was a clear dose–response increase in violent behaviour when participants were using methamphetamine compared to when they were not using the drug. This effect was especially large for frequent methamphetamine use (i.e. 16+ days of use in the past month), which increased the odds of violent behaviour 10‐fold (cf. threefold with less frequent use) after adjusting for shifts in other drug use, socio‐demographics and psychotic symptoms. Although psychotic symptoms significantly exacerbated the risk of violent behaviour, the relationship between methamphetamine use and violent behaviour was largely independent of psychotic symptoms, suggesting a direct relationship between the drug and violent behaviour. Heavy alcohol consumption also increased the risk of violent behaviour, but accounted for only 12–18% of the relationship between methamphetamine use and violence.

These findings indicate that violence is a key harm associated with methamphetamine use, with the probability of violent behaviour increasing from 10% during periods of abstinence to 60% during periods of heavy methamphetamine use. Violent behaviour was characterized mainly by interpersonal violence, ranging from altercations that led to fights to seemingly unprovoked physical attacks. Efforts are needed to quantify the impact of this interpersonal violence on assault rates. Our findings also fortify reports of violent behaviour associated with methamphetamine emergency presentations, and highlight the need to resource such services adequately in order to manage methamphetamine‐related violence [33] , [34] . Generic protocols for psychiatric emergencies (e.g. Castle et al. [35] ) may not be adequate or appropriate for use in this context [36] . Although the evidence base for treating methamphetamine dependence is limited [37] , [38] , existing effective interventions (e.g. structured psychological and behavioural interventions [37] ) need to be promoted and made accessible.

These results support the existence of a causal relationship between methamphetamine use and violent behaviour by showing that this relationship cannot be accounted for by pre‐morbid risk factors for violence among people who use the drug. We also adjusted for concurrent shifts in other drug use and several indicators of social adversity (unstable accommodation, unemployment, living alone) and found that these could not explain more effectively the relationship between methamphetamine use and violence. However, we cannot confirm the direction of the relationship between methamphetamine use and violence (i.e. whether methamphetamine use increased violence or violence increased methamphetamine use); nor can we confirm the mechanism underpinning this relationship. The pharmacological mechanisms posited to underpin methamphetamine‐related violence are monoamine dysregulation within the fronto‐limbic pathway with chronic use [39] and the sympathetic arousal (fight‐or‐flight response) during acute intoxication [40] , [41] . However, we cannot eliminate the potential contribution of non‐pharmacological factors (e.g. economically motivated violent crime) or factors closely linked to methamphetamine use (e.g. sleep deprivation, hypoglycaemia, cognitive impairment or acquired brain injury) to violent behaviour.

One question raised by our research is whether pre‐morbid risk factors for violence, such as conduct disorder, are necessary for methamphetamine use to precipitate violent behaviour. That is, whether the drug is having a disinhibiting effect on people predisposed to violence, or whether it is capable of inciting violent behaviour in people who are not predisposed to such behaviour. We failed to find evidence that methamphetamine's effect on violence was dependent upon people having conduct disorder, but we were limited in our capacity to detect such an effect because the vast majority of our sample had a history of conduct disorder. As is the case with alcohol [42] , we found some evidence that the relationship between methamphetamine and violent behaviour was dependent on an individual's characteristics, in that older users of the drug, who were generally less likely to engage in violent behaviour than their younger peers, were comparatively more vulnerable to the methamphetamine's effect on violent behaviour.

Although heavy alcohol consumption increased the risk of violence, it accounted for only a small component of the relationship between methamphetamine use and violence (12–18%) and had a much smaller effect on violent behaviour than did methamphetamine use. This raises an important question about the contribution of methamphetamine use to violence among young adults who ‘binge drink’, who often co‐consume stimulants with alcohol [43] . While the relationship between alcohol use and violence is well established [18] , [42] , the potential combined effect of alcohol and stimulant use on violent behaviour has not been well researched.

The current findings are specific to chronic methamphetamine users and should not be generalized to other populations, including recreational stimulant users, whose drug use is less frequent and who have fewer pre‐morbid risk factors for violence. In order to examine whether the relationship between methamphetamine use and violence was due to psychotic symptoms produced by the drug, we also needed to exclude participants who had a diagnosis of schizophrenia or mania from our sample (and a further 138 participants for whom this diagnostic information was not available). We therefore cannot make inferences about the effect of methamphetamine on violence in people with primary psychotic disorders.

Our definition of violent behaviour corresponded broadly to self‐reported assaults and/or damage to property, this being gleaned from a semi‐structured psychiatric interview (the BPRS). Definitions and measures of violent behaviour vary considerably across research studies [44] , but the BPRS was ideal for our study because it is a clinically valid instrument that assesses changes in psychiatric symptom severity across time. However, it relies upon accurate event recollection [27] , which may be impaired with methamphetamine and other heavy substance use. Note that our definition of violent behaviour would not necessarily correspond to violent offending.

In conclusion, violent behaviour is a key harm associated with the use of methamphetamine. A better understanding of how methamphetamine use contributes to the level of violence within the community is needed. Resources that identify and manage methamphetamine‐related violence are essential within clinical and frontline settings that have regular contact with methamphetamine‐using populations, particularly those services that manage individuals with methamphetamine psychosis. Further research is needed to elucidate the mechanism underpinning the relationship between methamphetamine use and violent behaviour in chronic users of the drug.

Dan Lubman has provided consultancy advice to Lundbeck and has received travel support and speaker honoraria from Astra Zeneca, Bristol Myers Squibb, Janssen and Lundbeck. Rebecca McKetin, Jake Najman, Sharon Dawe, Peter Butterworth and Amanda Baker report no financial relationships with commercial interests.

This research was funded by the National Health and Medical Research Council (Project Grant no. 350974) and the Australian Government Department of Health and Ageing. Data were collected through the Methamphetamine Treatment Evaluation Study (MATES), conducted by the National Drug and Alcohol Research Centre, University of New South Wales. The authors acknowledge the contribution of other project investigators: Robert Ali, Richard Mattick, Joanne Ross, Nicole Lee and Abdullah Mamun. Thanks also go to the research officers who assisted with data collection (Shelley Cogger, Erin Kelly, Kate Hetherington, Grace Ho, Cathie Sammut, Sagari Sarkar, Rachel Sutherland and Miriam Wyzenbeek), participating treatment agencies and the research participants.

Graph: The posited relationship between pre‐morbid factors, methamphetamine use, methamphetamine‐induced psychotic symptoms and violent behaviour

Graph: Probability of violent behaviour in the past month predicted from model 3 by days of methamphetamine use and psychotic symptoms

Graph: Appendix S1 Brief Psychiatric Rating Scale hostility item.