Abstract
Clozapine treatment may have beneficial effects on behavioral outcomes in psychotic disorders, including violent offending. Although clozapine and other antipsychotics have been linked to lower levels of violent behavior, these have been primarily in small selected samples, and population-based estimates have been limited and imprecise. We aimed to assess the effect of clozapine treatment on the rate of violent and nonviolent offending. We carried out a within-person mirror-image study of the Swedish population with linked prescription, hospitalization, and sociodemographic registers. Outcomes were violent, nonviolent, and overall offences occurring before and after clozapine, or olanzapine, initiation. Comparison of effects of clozapine and olanzapine on key variables was modeled with interaction terms. We found periods of mirror-image observation time with clozapine treatment were associated with a much lower rate of violent offending compared to periods before treatment (rate ratio [RR]: 0.13 (95% CI: 0.05, 0.34). Reductions in nonviolent offences were smaller in magnitude (RR: 0.37, 95% CI: 0.17, 0.80). There was a statistically greater rate reduction effect on violent offences for clozapine than olanzapine (RR for interaction: 4.84, 95% CI: 1.56, 14.86, P = .002). In patients with psychotic disorders, clozapine treatment is associated with a lower rate of violent offending compared to olanzapine.
Keywords: violence, clozapine, antipsychotic effectiveness, olanzapine, within-subject design, mirror image study
Introduction
Clinical management of psychotic disorders typically involves a combination of psychological and pharmacological therapy, with the aim of eliminating or limiting symptoms and optimizing functioning.1 However, violent offending is also an important adverse outcome in psychotic disorders2 and is more common in patients diagnosed with psychotic disorders compared to the general population.3 Patients with psychotic disorders are often intermittently treated,4 and studies suggest violence is higher in untreated patients.5
Some studies have reported lower levels of violence in people treated with antipsychotics, particularly second-generation drugs,6,7 and especially clozapine.8,9 Such observations are complicated by the strong possibility of confounding by indication.10 First, a violent episode may trigger a psychiatric evaluation, and the initiation of treatment. Second, given that clozapine requires a commitment by the patient to accept oral medication and frequent blood tests, it may be that people who are prescribed clozapine are systematically different from those prescribed other treatments in ways that mean that the simple comparison of violence occurrence between groups may not be valid.
Accumulating evidence continues to support the clinical effectiveness of clozapine on symptoms and hospital use in treatment refractory schizophrenia.11,12 National registers linked to prescribing information have clarified the real-world effectiveness of antipsychotic drugs for a range of outcomes.13,14 However, identifying convictions for violent behavior from national registers is not straightforward. A single conviction may refer to a mixture of separate constituent offences, some of which may be violent, and others nonviolent; eg, a person may be convicted for a combination of theft, assault, and a drug-related offence. Fazel et al15 reported that antipsychotic treatment reduced violent convictions (ie, convictions for offences where at least one offence was violent) in a Swedish population cohort, comparing periods of time on treatment with time off treatment, over a 3-year period. Effect estimates for clozapine were underpowered in the Fazel study to investigate clozapine, and focused on a conviction outcome, rather than on individual offences which comprised convictions.
In this article, we address these issues by (a) considering occurrence of violent offending in people treated for psychotic disorders with clozapine, comparing equal time periods before initiation with periods of time after, (b) test whether any effect of clozapine on violent offending is greater than that expected of a general antipsychotic effect, by comparing the effect of clozapine to that of olanzapine, the most commonly prescribed antipsychotic drug in Sweden, and (c) assessing violent and nonviolent offences separately. We draw upon registry data on clozapine and olanzapine prescriptions in Sweden, linking it with national data on convictions to identify violence-related outcomes. Our analysis is within-subject, ie, all comparisons made are of offending before initiation compared to after initiation within patients.
Methods
Using a within-subject design, also known as a mirror-image model, we compared the rate of offences during treatment with clozapine or olanzapine with periods of time of equal duration prior to the initiation of that treatment.
Data Sources
The unique Swedish personal identity number16 was used to link information from the following population-based registers:
The Causes of Death Register, comprising information on all deaths of Swedish residents since 1952 with causes of death coded according to the International Classification of Diseases (ICD).17
The National Patient Register, including all individuals admitted to psychiatric or general hospitals, with complete coverage for all in-patient care since 1987, and for hospital-based (as opposed to primary care-based) outpatient care since 2006.18
The Total Population Register, containing comprehensive information on age, sex, place of residence, and other relevant demographic characteristics19 on Swedish people.
The Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA), which integrates existing data from the labor market, educational and social sectors.20
The Register of Court Conviction, containing information on all court convictions and offences in Sweden for individuals 15 years of age or older since 1973.21
Finally, the Prescribed Drug Register,22 which contains patient identities for all dispensed prescribed drugs to the entire Swedish population since July 1, 2005, classified using the 5-level anatomical therapeutic and chemical classification system (ATC).
Derivation of Study Population
All prescriptions for clozapine and olanzapine (ATC-code N05AH02 and ATC-code N05AH03, respectively) registered from July 2005 until June 2012 were retrieved, excluding individuals who were prescribed both medicines, either concurrently or at different points in this period. Those that had a start date during 2005 were excluded, as a conservative measure to ensure the only new initiations of clozapine treatment were included. Of these, all Swedish people born 1955–1988 who had a first prescription of clozapine or olanzapine between January 1, 2006 and December 31, 2010 were kept. Those without a psychotic disorder or schizoaffective disorder (ICD-10 F20-F29) were further excluded (257 people).
In order to be confident that individuals included in the analysis were exposed to sustained periods of treatment, we limited our study to individuals treated with each drug for a minimum of 8 weeks. We identified 1176 people living in Sweden who were initiated on clozapine during the study period, and 4527 who were initiated on olanzapine. Among those prescribed clozapine, 1126 received more than 1 prescription of clozapine, of which 1086 had complete information on observation time (40 had missing data on the end of observation time, as defined later), of which 1004 were treated with clozapine for longer than 8 weeks and were included in the analysis. Among those prescribed olanzapine, 3967 had more than 1 olanzapine prescription, of which 3238 had complete information on observation times (729 had missing data on the end of observation time, as defined later), of which 2258 were continuously treated for 8 weeks or longer. Thus, our analysis was based on 1004 subjects treated with clozapine and 2258 subjects treated with olanzapine. To evaluate any influence of the 8 weeks criterion on our results, we inspected data on individuals prescribed each drug for less than 8 weeks.
Definition of Observation Time in Subjects
Data on convictions were collected for individuals in the study population described earlier, for (a) as long as possible following initiation of the drug, and (b) a period of time of equal duration prior to the initiation of the drug.
First, the “forward” observation time at risk for these outcomes was defined, using the Total Population Register, as the elapsed number of days from the date of initiation of the drug to either:
the discontinuation date for the drug (defined as the last date of prescribed medication where this occurred prior to a period of 6 months without a prescription for the drug, or without an inpatient psychiatric admission during this period), or
date of emigration, or
date of death, or
date of the end of the study period, which was December 31, 2011.
Second, having identified the forward observation time at risk, a backward observation time was defined for each subject of the same length. In the event that the backward observation time extended to a point before the start of the prescription register, the forward observation time was shortened to match the backward observation time. Data on offences were then gathered, classified by whether they occurred in the “before” period (prior to initiation of the drug) or the “after” period (during drug treatment), within the mirror-image observation time.
Measurement of Outcomes
Dates of all offences for which there were convictions during the mirror-image observation time (January 1, 2006–December 31, 2011) were collected for all study participants. We classified offences into violent offences and nonviolent offences. Violent offences included manslaughter, homicide, assault, gross assault, assault on a public official, arson, murder, unlawful threat, sexual crimes, crimes involving a weapon, cruelty to an animal, and infanticide. A full list of offences, and their classification into “violent” and “non-violent,” is displayed in table 1. Counts for violent and nonviolent offences, and overall offences were generated based on this information.
Table 1.
Coding of Offences Used in This Study
| Violent offences | Manslaughter |
| Homicide | |
| Gross assault | |
| Arson | |
| Infanticide | |
| Assault | |
| Cruelty to animals | |
| Sexual crimes | |
| Murder by carelessness | |
| Unlawful threat | |
| Weapon-related crime | |
| Assault on a public official | |
| Robbery | |
| Nonviolent offences | Acquisitive offence |
| Vehicle offence | |
| Theft | |
| Disorderly conduct | |
| Contact ban | |
| False alarm | |
| Drug offence |
Measurement of Covariates
Owing to the within-subject design, account was taken of characteristics that did not change over time: data on these were available for gender, age, highest educational attainment (categorized into compulsory education (≤9 years), “high school” education (10–12 years), and university or higher (≥13 years)), born in Sweden, the age and year of psychotic disorders diagnosis, and the age and date of drug initiation. We also considered characteristics that changed over time: employment status, presence of salary (as a binary variable yes/no), social salary (salary derived from social benefits, in quintiles), living in 1 of top 3 biggest cities (Stockholm, Gothenburg, or Malmö), and the presence of unemployment benefit (a state benefit specifically for unemployment). This information was available 2 years before drug start, and at the time of drug start, and was used to measure characteristics in the before and during observation periods within the mirror-image observation time, respectively.
Analysis
All analyses were performed in Stata 14.23 Offence rates were expressed per 100 persons. Owing to large numbers of zeroes (ie, observation periods where no offences occurred), Poisson, zero-inflated Poisson, and zero-inflated binomial regression models were compared on fit, assessed by both the Akaike Information Criteria and the Bayes Information Criteria. Zero-inflated negative binomial regression models gave best fit, and this model framework, incorporating robust standard errors, was retained. Counts of violent, nonviolent, and overall offences were compared by estimating the main effect (rate ratio [RR]) of treatment status with clozapine/olanzapine, comparing offence rate before treatment with during treatment. Difference in violence-reducing effects between clozapine and olanzapine were estimated by including an interaction term for drug (clozapine vs olanzapine). All covariates were entered into zero-inflated negative binomial regression models in order to arrive at an adjusted estimate. Given the within-subject design, only time-changing covariates, namely employment status, income, residing in 1 of Sweden’s 3 biggest cities, and unemployment benefit receipt were evaluated as potential confounders, by deriving and adjusting for categorical indicators for the before and the during observation period within the mirror-image observation time. Age and calendar year at drug initiation/psychotic disorders diagnosis, gender, highest educational attainment, and whether the person was born in Sweden were not included as covariates because they did not vary within subjects.
Zero-inflated negative binomial models are estimated in 2 parts,24 consisting of a negative binomial model, in this study estimating counts of violent offences in patients who offend, which was the focus of our analysis. Zero-inflated negative binomial models also estimated a logit model, predicting excess zeros, in this study, zeros refer to periods of observation within the mirror-image observation time without offences, and we included inflation coefficients for age and gender, reflecting that these were the main influences on zero-offending.25 The negative binomial model also estimates a dispersion parameter, quantifying the extent to which variance exceeds that expected under a Poisson model. Supplementary analyses tested crude and fully adjusted associations stratified by gender (supplementary tables S4 and S5), and by mirror-image observation time, dichotomized at 3 years (supplementary tables S6 and S7).
Results
Table 1 describes the coding of offences into violent and nonviolent categories used for this study. Table 2 summarizes sample characteristics. A total of 2258 people treated with olanzapine met criteria for the study, of which 1385 (61.3%) were male, compared to a slightly greater proportion in clozapine-treated patients (66.0%, n = 1004). More than three-quarters of the olanzapine patients were born in Sweden (76.2%), compared to nearly 80% of the clozapine group. Treatment for 2 years or more was more common among clozapine subjects than olanzapine (51.4% compared to 31.9%). Any admission for mood disorder was also commoner in the olanzapine group (36.5%, compared to 32.0% in the clozapine group). Between treatment groups, there was difference in duration of observation time, with a higher proportion of olanzapine patients treated for less than a year, and a higher proportion of clozapine patients (about a third) treated for more than 3 years compared to the olanzapine group (around a fifth). There was statistical evidence for differences between clozapine-and olanzapine-treated groups for all covariates included in this study.
Table 2.
Description of Clozapine (n = 1004) and Olanzapine (n = 2256) and Samples by Non-Time Changing Characteristics, Based on Prescription Registers for the Whole of Sweden, Reflecting First Withdrawal of Each Drug Between January 1, 2006 to December 31, 2010
| Clozapine | Olanzapine | |
|---|---|---|
| Count (%) | Count (%) | |
| Year of treatment start | ||
| 2006 | 204 (20.32) | 664 (29.41) |
| 2007 | 194 (19.32) | 490 (21.7) |
| 2008 | 201 (20.02) | 360 (15.94) |
| 2009 | 213 (21.22) | 372 (16.47) |
| 2010 | 192 (19.12) | 372 (16.47) |
| Gender | ||
| Male | 663 (66.04) | 1385 (61.34) |
| Female | 341 (33.96) | 873 (38.66) |
| Born in Sweden | 797 (79.38) | 1721 (76.22) |
| Duration of treatment | ||
| 8 weeks–1 year | 224 (22.31) | 1001 (44.33) |
| 1–3 years | 446 (44.42) | 788 (34.90) |
| 3 years or more | 334 (33.27) | 469 (20.77) |
| Educational attainment at treatment start | ||
| ≤9 years | 343 (34.16) | 702 (31.09) |
| 10–12 years | 576 (57.37) | 1317 (58.33) |
| University or higher | 53 (5.28) | 172 (7.62) |
| Missing | 32 (3.19) | 67 (2.97) |
| Age at drug start (years) | ||
| 18 < 28 | 206 (20.52) | 295 (13.06) |
| 28 < 38 | 331 (32.97) | 578 (25.6) |
| 38 < 48 | 319 (31.77) | 920 (40.74) |
| 48 < 58 | 148 (14.74) | 465 (20.59) |
| Start periods | ||
| Before 2007 | 398 (39.64) | 1154 (51.11) |
| After 2008 | 606 (60.36) | 1104 (48.89) |
| Period of psychotic disorders diagnosis | ||
| 1970–1982 | 34 (3.39) | 102 (4.52) |
| 1983–1991 | 137 (13.65) | 295 (13.06) |
| 1992–2001 | 291 (28.98) | 585 (25.91) |
| 2002– | 542 (53.98) | 1276 (56.51) |
| Age at psychotic disorders diagnosis | ||
| 18 < 28 | 536 (53.39) | 861 (38.13) |
| 28 < 38 | 310 (30.88) | 763 (33.79) |
| 38 < 48 | 141 (14.04) | 527 (23.34) |
| 48 < 58 | 17 (1.69) | 107 (4.74) |
Table 3 summarizes data on offences. On the basis of 369 offences in the clozapine group and 960 offences in the olanzapine group, we estimated a rate reduction of around 75% in the clozapine group and 50% in the olanzapine group with statistical evidence of difference between the 2 drugs (P value for interaction between drug and period of observation = .015). The rate reduction for nonviolent offences, comparing before treatment to during treatment for clozapine, was 63% after adjustments, compared to 39% for olanzapine. For violent offences, the fully adjusted rate reduction for treatment compared to before treatment was 87% for clozapine, and 8% for olanzapine (RR for clozapine: 0.13, 95% CI: 0.05, 0.34, RR for olanzapine: 0.82, 95% CI: 0.47, 1.43, P value for the interaction between drug and period of observation = .002). In the final adjusted model for overall offences, female gender (compared to male gender) predicted zero offences during periods of observation (P = .027), but age did not, and neither gender nor age were statistically evidenced predictors of 0 counts for nonviolent, or overall offences.
Table 3.
Descriptive Data (Absolute Counts and Rate of Offences Per 1000 Person Years of Observation) for Overall, Violent, and Nonviolent Offences for Olanzapine (n = 2258) and Clozapine Treatment (n = 1004)
| Clozapine | Olanzapine | Interaction Term | P Value for Interaction | |||
|---|---|---|---|---|---|---|
| Total number of violent offences | 103 | 506 | ||||
| Individuals with any violent offence (percentage of the overall treatment group) | 63 (6.27) | 144 (6.38) | ||||
| Number of violent offences before treatment (rate) | 96 (74.41) | 376 (95.71) | ||||
| Number of violent offences during treatment (rate) | 7 (5.43) | 130 (33.09) | ||||
| Effect of drug on violent offence rate | RR | 95% CI | RR | 95% CI | ||
| Crude | 0.07 | 0.03, 0.18 | 0.35 | 0.20, 0.61 | ||
| Fully adjusteda | 0.13 | 0.05, 0.34 | 0.82 | 0.47, 1.43 | 4.82 (1.56,14.86) | P = .002 |
| Total number of nonviolent offences | 222 | 507 | ||||
| Individuals with any nonviolent offence (percentage of the overall treatment group) | 69 (6.87) | 193 (8.55) | ||||
| Number of nonviolent offences before treatment (rate) | 161 (124.79) | 326 (82.99) | ||||
| Number of nonviolent offences during treatment (rate) | 61 (47.28) | 181 (46.07) | ||||
| Effect of drug on nonviolent offence rate | RR | 95% CI | RR | 95% CI | ||
| Crude | 0.38 | 0.17, 0.83 | 0.56 | 0.40, 0.76 | ||
| Fully adjusteda | 0.37 | 0.17, 0.80 | 0.61 | 0.44, 0.86 | 1.66 (0.68,4.04) | P = .263 |
| Total number of overall offences | 369 | 960 | ||||
| Individuals with any overall offence (percentage of the overall treatment group) | 128 (12.75) | 304 (13.46) | ||||
| Number of overall offences before treatment (rate) | 295 (122.69) | 635 (161.64) | ||||
| Number of overall offences during treatment (rate) | 74 (30.78) | 325 (82.73) | ||||
| Effect of drug on overall offence rate | RR | 95% CI | RR | 95% CI | ||
| Crude | 0.25 | 0.12, 0.51 | 0.51 | 0.12, 0.51 | ||
| Fully adjusteda | 0.24 | 0.12, 0.48 | 0.62 | 0.45, 0.85 | 2.55 (1.20,5.44) | P = .015 |
Also shown are crude and adjusted offence (overall, nonviolent, and violent) rate ratios with 95% confidence intervals for during vs before treatment with clozapine and olanzapine. Rate ratios were estimated from zero-inflated negative binomial regression models with offence rate as the dependent variable and period of observation (dichotomized into before treatment and during treatment), as the main independent variable of interest. Models took account of clustered before and during treatment data within individuals. Model estimates for zeroes are presented in supplementary table S1.
aFully adjusted models are adjusted for urban residence, salary presence, employment status, and unemployment benefit receipt.
Time-varying characteristics for olanzapine- and clozapine-treated subjects are shown in supplementary table S1. The proportion of people working fell slightly for subjects treated with both drugs. The presence of salary fell among both the olanzapine group (22.7%–17.8%) and the clozapine group (15.8%–9.7%). Model estimates from the 0 prediction part of final models for violent, nonviolent, and overall offences are displayed in supplementary table S2. In this article, final estimates for the effects of clozapine and olanzapine on offending, interaction terms and interaction P values are presented in table 3. Zero-prediction coefficients and dispersion parameters for violent, nonviolent, and overall offences are presented in supplementary table S2. Estimates restricted by gender gave similar results, however among women, statistical evidence was insufficient at the 5% alpha level. The fully adjusted rate reduction for clozapine treatment on violent offences among men was 0.04 (95% CI 0.02, 0.12), and for olanzapine it was 0.38 (95% CI 0.21, 0.69, P value for interaction <.001). In women the adjusted rate reduction for clozapine on violent offences was 0.34 (95% 0.08, 1.39), and for olanzapine it was 0.55 (95% CI 0.09, 3.30, P value for interaction 0.681, supplementary table S4). Among 492 individuals prescribed clozapine for less than 8 weeks, there were no violent offences during the mirror-image observation time before initiation, and 4 offences in the mirror-image observation time following initiation.
Discussion
Summary of Findings
In the population of Sweden, clozapine treatment was associated with greater reductions in overall and violent offending, but not nonviolent offending, compared to olanzapine, the most commonly prescribed antipsychotic drug in Sweden. This was not accounted for by confounding by time-changing socioeconomic characteristics. Effects of clozapine on nonviolent offending were statistically similar to olanzapine, suggesting that clozapine may offer specific benefits on the risk of violent offending in people with psychotic disorders.
Limitations and Strengths
There may be local factors that determine clozapine prescription, and our sample may not be representative of clozapine users outside Sweden. Our results, in particular the estimates of pretreatment rate of offences, could have been affected by an underlying trend toward less offending in patients with psychotic disorders as they get older, irrespective of how they are treated. Initiation of treatment in some individuals could have occurred as a result of violent behavior, eg, triggering arrest and subsequent psychiatric evaluation. We did not assess the effect of concurrent medicines in this study, including the effect of concurrent treatment with clozapine and olanzapine. Although data on psychiatric diagnoses were comprehensive, misclassification is a possibility. A validation study involving record review of admission diagnoses in Sweden suggested good correspondence, with kappa values of between 0.74 and 0.76.26 Around 85% of patients with an admission diagnosis of schizophrenia in Sweden had Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) schizophrenia diagnoses assigned by clinical raters, in another study.27 Although adjustment was made for time-varying covariates, factors such as age, gender, and calendar time could not be examined directly, due to the chosen design. Mirror-image studies cannot take account of the possible effects of health policy changes on the background rates of the outcome. Adjusted estimates should, therefore, be interpreted with caution.28 However, we were primarily interested in comparison between clozapine and olanzapine. We think the influence of these factors is likely to have been similar between the 2 treatment groups, and therefore unlikely to fully explain differences between olanzapine and clozapine treatment observed in this study. We included prescribing data from 2006; having such data prior to 2006 would have afforded a longer study period, and increased the ability to assess the impact on our results of change in underlying offending patterns over time. It is possible that predictors of offending that were not accounted for in this study, such as personality disorders or cognitive impairment, could have affected our results. We included patients only treated for longer than 8 weeks in our analysis—limiting the generalizability of our findings to people receiving treatment for at least this length of time. No violent offences occurred in mirror-image observation times among individuals prescribed clozapine for less than 8 weeks. Prior to the availability of inpatient prescription data in 2006, patients may have initiated clozapine/olanzapine during an inpatient admission, but received the first-recorded prescription only after discharge, resulting in a start time for the mirror-image observation period which was later than the true start time for the drug, and misclassification of time on treatment as time off treatment. Any bias introduced by this would likely be toward underestimating the effectiveness of both drugs.
On the other hand, our analysis was based on dates for offences rather than convictions in contrast with previous population-based studies, although we did not distinguish among offences comprising a conviction, beyond the classification of offences into violent and nonviolent. Offences data were taken from a whole population-based register of convictions with effectively total coverage; bias introduced by missing data on convictions is very unlikely. We had information on women, in contrast to one previous study on this topic in Sweden15 that was restricted to men, and had access to enough data on clozapine to arrive at a precise estimate of rate reduction of violent offences attributable to clozapine treatment. Information on convictions and their aligned offences was from a national register of court proceedings, not based on self-report. We studied both overall and violent offences as the outcome. Our data were based on dispensed prescriptions for these drugs, and made the assumption that dispensing of the drug was equivalent to full adherence, making it analogous to an intention to treat analysis.
Explanations
The observed reduction in violent offending rate for both antipsychotic drugs indicates that this could be a class effect of antipsychotic drugs as a whole, with clozapine being particularly effective, consistent with the superiority of clozapine over other antipsychotics in other areas.29,30 Clozapine could improve engagement with health care staff, social cognition, reduce irritability, improve social and occupational functioning, or effects on psychotic symptoms could mediate the effect. Clozapine-treated patients are typically affected by more severe illness and more treatment resistant symptoms, and clozapine treatment requires greater contact with the mental health system. In this regard, the current study was not able to distinguish among possible active components of clozapine treatment in relation to offending, including the role of increased contact with the health care system. One future approach to examining this could be to compare clozapine with another treatment that also involves increased contact with the health care system, such as long-acting injections.
Previous Literature
There is a consistent observational association between psychotic disorders and violence.3 Violence risk in psychotic disorders may be related to clinical status,7 concurrent substance use disorders, or to nonadherence with antipsychotic medication.5 Although Swanson et al6 found that the violence-reducing effects of atypical antipsychotics was greater than for typicals, an analysis of data from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE), by the same investigators, indicated that violence reduction in newer antipsychotics was not significantly greater than for perphenazine, a traditional typical antipsychotic drug.31 Convictions may happen a significant period of time after the offences themselves, leading to bias in effect estimates; this analyzed date information on offences within convictions, and therefore benefited from greater statistical power. In contrast to previous work, this study also adjusted for employment, salary presence, unemployment benefits, and place of residence as time-changing covariates that may have had an influence both on prescription of each drug and the offending outcome. Typically, measurement of violence in pharmacological studies has been done by independent observers using rating scales; few studies have used legal/administrative outcomes such as criminal conviction.5 Stevens et al32 report a randomized controlled trial showing no effect of assertive specialized treatment on offending in first episode psychosis patients, suggesting the need for specific, rather than universal interventions for violence reduction.
How Our Results Fit In
As far as we are aware, this is the first report of violence-reducing effects of clozapine in a population-based sample of both men and women and in a within-subject observational design. We also found an (albeit weaker) effect for olanzapine, in accordance with some previous work.6 Our results suggest that the effects are independent of socioeconomic factors that might also have an influence on offending rates.
Conclusions
We found strong statistical evidence for a violence-reducing effect of clozapine in whole population data from Sweden that was larger in magnitude than olanzapine. Clozapine may be more effective than olanzapine at reducing violent offending behavior.
Funding
Stockholm County Council and the Swedish Research Council (2012–2264). The data linkages and staff costs have also been supported by grants from the Stockholm County Council (2007008); Swedish Council for Working Life and Social Research (2007–2064); Swedish Research Council (523-2010-1052); and Swedish Regional agreement on medical training and clinical research.
Supplementary Material
Acknowledgments
Dr Bhavsar was supported by a Wellcome Trust Clinical Research Training Fellowship (101681/Z/13/Z). This article represents independent research part funded by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health. The authors have declared that there are no conflicts of interest in relation to the subject of this study.
References
- 1. Gelder M, Gath D, Mayou R.. Oxford Textbook of Psychiatry. Oxford: Oxford University Press; 1989. [Google Scholar]
- 2. Walsh E, Buchanan A, Fahy T. Violence and schizophrenia: examining the evidence. Br J Psychiatry. 2002;180:490–495. [DOI] [PubMed] [Google Scholar]
- 3. Fazel S, Gulati G, Linsell L, Geddes JR, Grann M. Schizophrenia and violence: systematic review and meta-analysis. PLoS Med. 2009;6(8):e1000120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Davis JM, Schaffer CB, Killian GA, Kinard C, Chan C. Important issues in the drug treatment of schizophrenia. Schizophr Bull. 1980;6(1):70–87. [DOI] [PubMed] [Google Scholar]
- 5. Swartz MS, Swanson JW, Hiday VA, Borum R, Wagner HR, Burns BJ. Violence and severe mental illness: the effects of substance abuse and nonadherence to medication. Am J Psychiatry. 1998;155:226–231. [DOI] [PubMed] [Google Scholar]
- 6. Swanson JW, Swartz MS, Elbogen EB. Effectiveness of atypical antipsychotic medications in reducing violent behavior among persons with schizophrenia in community-based treatment. Schizophr Bull. 2004;30(1):3–20. [DOI] [PubMed] [Google Scholar]
- 7. Swanson JW, Swartz MS, Van Dorn RA, et al. . A national study of violent behavior in persons with schizophrenia. Arch Gen Psychiatry. 2006;63(5):490–499. [DOI] [PubMed] [Google Scholar]
- 8. Krakowski MI, Czobor P, Citrome L, Bark N, Cooper TB. Atypical antipsychotic agents in the treatment of violent patients with schizophrenia and schizoaffective disorder. Arch Gen Psychiatry. 2006;63(6):622–629. [DOI] [PubMed] [Google Scholar]
- 9. Buckley P, Bartell J, Donenwirth K, Lee S, Torigoe F, Schulz SC. Violence and schizophrenia: clozapine as a specific antiaggressive agent. Bull Am Acad Psychiatry Law. 1995;23(4):607–611. [PubMed] [Google Scholar]
- 10. Walker AM. Confounding by indication. Epidemiology. 1996;7(4):335–336. [PubMed] [Google Scholar]
- 11. Siskind D, McCartney L, Goldschlager R, Kisely S. Clozapine v. first- and second-generation antipsychotics in treatment-refractory schizophrenia: systematic review and meta-analysis. Br J Psychiatry. 2016;209(5):385–392. [DOI] [PubMed] [Google Scholar]
- 12. Land R, Siskind D, McArdle P, Kisely S, Winckel K, Hollingworth SA. The impact of clozapine on hospital use: a systematic review and meta-analysis. Acta Psychiatr Scand. 2017;135(4):296–309. [DOI] [PubMed] [Google Scholar]
- 13. Taipale H, Tanskanen A, Tiihonen J, Mehtälä J. Comparative effectiveness of antipsychotic drugs for rehospitalization in schizophrenia—A nationwide study with 20-year follow-up. Schizophr Bull. 2017;44(6):1381–1387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Tiihonen J, Mittendorfer-Rutz E, Torniainen M, Alexanderson K, Tanskanen A. Mortality and cumulative exposure to antipsychotics, antidepressants, and benzodiazepines in patients with schizophrenia: an observational follow-up study. Am J Psychiatry. 2016;173(6):600–606. [DOI] [PubMed] [Google Scholar]
- 15. Fazel S, Zetterqvist J, Larsson H, Långström N, Lichtenstein P. Antipsychotics, mood stabilisers, and risk of violent crime. Lancet. 2014;384(9949):1206–1214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. The swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol. 2009;24(11):659–667. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. World Health Organization. The ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines. Geneva: WHO; 1992. [Google Scholar]
- 18. Ludvigsson JF, Andersson E, Ekbom A, et al. . External review and validation of the Swedish national inpatient register. BMC Public Health. 2011;11:450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Ludvigsson JF, Almqvist C, Bonamy AK, et al. . Registers of the Swedish total population and their use in medical research. Eur J Epidemiol. 2016;31(2):125–136. [DOI] [PubMed] [Google Scholar]
- 20. Longitudinal Integration Database for Health Insurance and Labour Market Studies(LISA by Swedish acronym) English. Sweden: Federal Ministry of Education and Research; 2012. https://www.scb.se/LISA [Google Scholar]
- 21. Gauffin K, Vinnerljung B, Fridell M, Hesse M, Hjern A. Childhood socio-economic status, school failure and drug abuse: a Swedish national cohort study. Addiction. 2013;108(8):1441–1449. [DOI] [PubMed] [Google Scholar]
- 22. Wettermark B, Hammar N, Fored CM, et al. . The new Swedish Prescribed Drug Register–opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf. 2007;16(7):726–735. [DOI] [PubMed] [Google Scholar]
- 23. StataCorp. Stata Statistical Software: Release 14. Special Edition 2014. [Google Scholar]
- 24. Greene WH. Accounting for Excess Zeros and Sample selection in Poisson and Negative Binomial Regression Models. New York: New York University, Leonard N. Stern School of Business, Department of Economics; 1994. [Google Scholar]
- 25. Witt K, van Dorn R, Fazel S. Risk factors for violence in psychosis: systematic review and meta-regression analysis of 110 studies. PLoS One. 2013;8(2):e55942. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ekholm B, Ekholm A, Adolfsson R, et al. . Evaluation of diagnostic procedures in Swedish patients with schizophrenia and related psychoses. Nord J Psychiatry. 2005;59(6): 457–464. [DOI] [PubMed] [Google Scholar]
- 27. Dalman Ch, Broms J, Cullberg J, Allebeck P. Young cases of schizophrenia identified in a national inpatient register–are the diagnoses valid? Soc Psychiatry Psychiatr Epidemiol. 2002;37(11):527–531. [DOI] [PubMed] [Google Scholar]
- 28. Kishimoto T, Nitta M, Borenstein M, Kane JM, Correll CU. Long-acting injectable versus oral antipsychotics in schizophrenia: a systematic review and meta-analysis of mirror-image studies. J Clin Psychiatry. 2013;74(10):957–965. [DOI] [PubMed] [Google Scholar]
- 29. Stroup TS, Gerhard T, Crystal S, Huang C, Olfson M. Comparative effectiveness of clozapine and standard antipsychotic treatment in adults with schizophrenia. Am J Psychiatry. 2016;173(2):166–173. [DOI] [PubMed] [Google Scholar]
- 30. Wimberley T, MacCabe JH, Laursen TM, et al. . Mortality and self-harm in association with clozapine in treatment-resistant schizophrenia. Am J Psychiatry. 2017;174(10): 990–998. [DOI] [PubMed] [Google Scholar]
- 31. Swanson JW, Swartz MS, Van Dorn RA, et al. ; CATIE Investigators Comparison of antipsychotic medication effects on reducing violence in people with schizophrenia. Br J Psychiatry. 2008;193(1):37–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Stevens H, Agerbo E, Dean K, Mortensen PB, Nordentoft M. Reduction of crime in first-onset psychosis: a secondary analysis of the OPUS randomized trial. J Clin Psychiatry. 2013;74(5):e439–e444. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.