Lessons from SSRIStories #5: What Does Research Tell Us About the Connection between SSRIs and Violence?

by Julie Wood

Summary of Post #4: Medication-induced violence is different from regular violence. It can manifest as bizarre and random actions without apparent motive. Sometimes, it is based on a sudden impulse without any warning. Other times, thought distortion leads to elaborate plans, but the violence still makes no rational sense. In many cases, such as school or other mass shootings, perpetrators plan their own suicide as part of the violence. Sometimes SSRI violence involves overreactions so extreme as to defy rational explanation. SSRI suicides are often the result of sudden impulses, and bereaved loved ones agonize because they missed warning signs that never existed.

It is intriguing that everyone accepts that alcohol and street drugs can cause undesirable behaviour, while being reluctant to believe that prescription medications, which are often chemically similar to or the same as banned substances, might have similar effects. Prescription medications and street drugs alike can trigger violent impulses. We have no problem understanding that mind-altering drugs might trigger violence, but we seem not to understand that many prescription drugs are mind-altering, just like street drugs.

In 1998, Steadman et al noted that:

“FOR 75 YEARS, studies have attempted to estimate the prevalence of violence committed by people discharged from psychiatric facilities in the United States and to compare that rate with the prevalence of violence by others in their communities. These studies have been invoked in legal and policy debates… Four methodological problems consistently have compromised this work… [and in addition] inclusion criteria limit the generalizability of reported findings.” [1]

In other words, many researchers went looking for a connection, which they strongly believed in, between mental illness and crime. The result was an historical body of research that claims to show that higher rates of violence are associated with mental illness, but the methodologies were flawed and the results cannot be extrapolated.

A recent article in Huffington Post [2] included the following paragraph:

“More studies, more conflicting results – The link between SSRI use and violence is controversial turf, and previous studies have yielded conflicting results. A 2010 PLOS One study that used data from the U.S. Food and Drug Administration found that SSRI use was associated with increased violence, for example, while a study published the same year in the Journal of Policy Analysis and Management reported exactly the opposite.”

The first study [3] referenced was conducted for the Institute for Safe Medication Practices, using the US Food and Drug Administration (FDA) Adverse Event Reporting System (AERS) database. The researchers, Moore, Glenmullen and Furberg (MGF) identified the top 31 prescription drugs associated with violence toward others, and calculated a PRR [4], showing how prevalent the side effect is for the subject drug relative to other side effects and other drugs. Two of the most common SSRI antidepressants – fluoxetine (Prozac) and paroxetine (Paxil, Seroxat) were in the top 3. Fluvoxamine (Luvox), venlafaxine (Effexor) and sertraline (Zoloft) were among the top 12. Escitalopram (Lexapro), Citalopram (Celexa) and Bupropion were in the top 20. Duloxetine (Cymbalta) was 25th. The probability that these drugs are associated with more violence than other drugs by coincidence is essentially zero – there is no realistic chance that this is the case. In other words, no matter what other research shows, the FDA data reveal that some drugs definitely are associated with increased violence.

The other study [5], which supposedly “reported exactly the opposite” actually did no such thing. In the second study, the authors:

“consider possible links between the diffusion of new pharmaceuticals used for treating mental illness and crime rates. We describe recent trends in crime and review the evidence showing that mental illness is a clear risk factor both for criminal behavior and victimization. We summarize the development of a number of new pharmaceutical therapies for the treatment of mental illness that came into wide use during the “great American crime decline.”… we find some evidence that the expansion of psychiatric drugs is associated with decreased violent crime rates, but not property crime rates. We find no robust impacts on homicide rates and no effects on arrest rates. Further, the magnitudes of the estimated effects of expanded drug treatment on violent crime are small. Our estimates imply that about 5 percent of the decline in crime during the period of our study was due to expanded mental health treatment.”

This study examined violent crime rates, supposedly against the most likely causal factors. It is full of unsubstantiated information, making claims like the last few decades were “a period of dramatic technological advances in the treatment of mental illness”. It provides no evidentiary basis for its statements, or its conclusion. This study purports to explain the drop in crime rates of the G7 countries, and notes a correlation between reduced crime and increased use of psychiatric medications. They imply that the correlations they found could indicate a causal link, while ignoring the two main factors to which the drop in crime rates is most commonly attributed.

These two factors are demographic changes; specifically, a decline in the percent of young males in the population, and dramatically reduced illegal drug use. As another analysis of the phenomenon noted: “epidemics of crack cocaine and heroin appear to have burnt out.” [6] Other analyses have entirely accounted for the reduction in crime without any reference to mental health treatments. In fact, remembering that illegal and legal drugs can both cause violent behaviours, it would be odd if reduced crime were attributable both to the decreased use of illegal drugs and the increased use of legal drugs.

The Huffington Post took the title of the second article at face value and assumed that its findings were reliable, and incompatible with those of the MGF FDA data study. Journalists quite reasonably accept that the conclusions of all published studies are valid. They have no reason to suspect that some research is actually propaganda for a particular viewpoint. Thus, the average person reading the Huffington Post article, and others like it, is given the impression that research on the relationship between medication and violence is inconclusive.

Problems in methodology and interpretation

In the same 1998 study cited above regarding the problematic history of research in this area, Steadman et al [7] found that unless drugs or alcohol are involved, people with mental disorders do not pose any more threat to the community than anyone else. Steadman’s research team was referring to illegal drugs, but, as the study based on FDA data reveals, prescription medications can significantly increase propensity to violence. Prescription medications and street drugs alike can lead to violent acts, such as assaults, suicides, homicides, and physical abuse.

It is sometimes argued that it is not the medications, but the underlying conditions, that cause problems. In other words, violent and suicidal people are given drugs to help them, but the drugs do not completely eliminate these tendencies. However, other research has demonstrated that drugs can induce thoughts and behaviours in ordinary people who have never experienced them previously.[8]

A 2015 Swedish study [9] found that taking SSRIs increased the risk of being involved in a violent crime across the board for all ages but most markedly for Swedes 15 to 24 years old. The authors caution that: “the analytical approach used does not fully account for time-varying risk factors such as symptom severity or alcohol misuse that might affect an individual’s risk of committing a violent crime”.

There are a number of reasons to think that the link between violence and treatment may be even stronger than this study showed. One factor is the low incidence of identifiable drug-related violence as a rate. Homicides and assaults resulting from medication may occur in less than 1% of the population taking (or withdrawing from) particular medications. However, if ½ % of the 13% of the US adult population taking antidepressants [10] became involved in violence, this would be about 172,000 people. If 1% of that violence was deadly, 1,720 people would be affected.

One thousand and seven hundred people may be a very tiny percent of the US population, but they are still a lot of people, who all want to live and be healthy. Fewer people are killed in air crashes each year, yet nobody thinks that ensuring aviation safety is unnecessary.

Because population studies do not deal with individual situations, they can only look at overall trends, which are affected by many variables, many of which cannot be identified and none of which are controllable. Population studies search for correlating rates of increasing (or decreasing) violence with increasing prescriptions, and then measure the statistical significance of the correlations, and then try to analyze the likelihood of causation. This approach reflects an implicit assumption that populations are homogenous with respect to their reaction to medications. But we know that this is not the case.

Research has shown that up to 15% of the population are unable to properly metabolize SSRIs effectively,[11] and these poor metabolizers are the people most likely to have extreme negative reactions. When drugs are not metabolized, continuing to take them has the same effect as increasing the dose. This being the case, and considering that some people may not be able to metabolize the drugs at all, it is possible that two separate and offsetting effects are hidden in the data. Perhaps most of the population taking SSRIs has blunted emotional reactions, and commits fewer violent acts as a result. By contrast, the poor metabolizers may experience a dramatic increase in risk of violence. In a population study, this offsetting effect might cancel out any overall increase in violence. So, the overall population trend identified might be accurate, but it would not be appropriate to conclude that antidepressants lower violence, when an identifiable segment of the population, however small in percentage terms, is put in serious danger by the medication.

Note that reducing violent acts is not the same as reducing crime. Most crimes are property crimes, and remembering that SSRIs work by dulling peoples’ consciences, their impact on non-violent crime might be different. This is not a fact, it is simply a possible hypothesis, which is neither unambiguously supported nor refuted by available research.

The studies can suffer other methodological problems, as well. The Swedish researchers defined the end of a treatment period as the date of the last prescription dispensed. In the authors’ own words: “Another possible source of underestimation is that we used a conservative approach to measure the end of a treatment period (we defined this as the date the last SSRI prescription in a treatment period was dispensed”. Thus they did not take into account that the period after stopping a drug may actually be higher risk than the period on treatment, and they were including times when individuals might have been in withdrawal as non-treatment periods to contrast with when those same people were on the drug. Since the withdrawal period is known to be a higher risk time for some, this approach would have caused greater underestimation of the drug impact than the authors thought.

They examined the potential role of alcohol as a confounding factor (“The possible role of alcohol as a time-varying confounder was tested by using hospitalisations for alcohol intoxication as an outcome, showing an increased risk during times of medication”) when in fact what those data may have been showing was that alcohol was part of a causation chain. The authors were conservative in their analysis and discounted the impact of the SSRI when alcohol was involved, when the medications (fluoxetine, citalopram, paroxetine, sertraline, fluvoxamine and escitalopram) might have caused the drinking in some people, and a more pronounced reaction to the drinking, which may have increased the risk of violence.

The same study excluded people who might have been of interest: “we wanted to restrict the sample to those adherent with SSRIs, individuals with a single SSRI prescription within a six-month-period were excluded from all further analyses as no assumptions could be made about their medication adherence.” However, those who are not reacting well to SSRIs may be the ones who become “non-adherent”, perhaps after a bad experience. In this way the researchers may have inadvertently limited their study subjects to those best able to tolerate SSRI medication, and least likely to become violent.

The Swedish study is interesting because of its design that had subjects acting as their own “control”. Given that, despite its extremely conservative approach, a positive association between violence and SSRIs was identified, one can conclude that the true association was potentially much stronger.

Another challenge for studies that look for medication links to violence is definitions. The Swedish study included: “attempted, completed and aggravated forms of: homicide, manslaughter, unlawful threats, harassment, robbery, arson, assault, assault on an official, kidnapping, stalking, coercion, and all sexual offences.” The MGF study used the MedDRA codes for homicides, physical assault, physical abuse, homicidal ideation and “violence-related symptom” (aggression) reported to the FDA. Clearly, while there is overlap, these definitions are different.

A clear connection

One very obvious difference is that the MGF study, while clearly showing that certain medications have a higher relationship to violence, has included thoughts along with actions in their definition. MGF ranked the violent acts according to MedDRA codes, from most serious (homicides) to violence-related symptom (least serious) and selected the most serious side effect reported in each report. This is not to say that MGF were wrong to include ideation and feelings, but this definition reduces comparability to studies based on crime stats.

Updated data was extracted from RxISK.org (120 months starting Jan 1, 2004, US & Canadian data), which include the data used by MGF (69 months, starting Jan 1, 2004, US only) show that the number of thought-related symptoms greatly outnumber cases of realized violence.

Click here to see the data extracted from RxISK.org [12].

The RxISK update was not able obtain individual reports, so that some violence events will duplicated in the codes; that is, there will be more violence events than reports. However, by using definitions that include only one or two categories that can be assumed to be mutually exclusive, or have minimal duplication, this is minimized. Also, as long as the events are not added or subjected to quantitative analysis, this duplication will not change the ranking of medications in terms of their association with violence. TABLE 1 shows the rankings, based on PRRs, of different violence definitions.

Three additional drugs qualified for inclusion. These were Trazodone (Desyrel), Amitriptyline (Elavil), and Nortriptyline (Pamelor). The authors had mentioned the first two but at Q3, 2009, they did not meet the criteria for inclusion. In addition, Imipramine (Tofranil) was included, even though it did not meet the criteria, because the RxISK update covered a number of definitions of violence (different combinations of MedDRA codes). One of these definitions included suicide. It is generally accepted that drug-induced suicide is the same violent impulse that causes people to harm others, but turned inward. Imipramine was included because it has a PRR of 15.3 for suicide, and for one definition, “Deadly Drugs” (homicides plus suicides) it ranks higher than Chantix, the drug most associated with violence under the MGF definition.

The MGF study is important because it confirms what observation /anecdotal data show: there is a definite relationship between certain drugs, and violence.

Table 1

Table 1 illustrates how changing the definition of violence can significantly change the ranking of a drug on its relationship to violence.

The updated data, despite the slight difference in PRR calculation, basically confirm the original findings. The one exception is Intron, which did not show PRRs> 2 in the update. PRR> 2 is a fairly reliable indicator that the violent incidents reported did not happen during the prescription by chance. PRRs in the ranges indicated for the SSRIs are clearly not coincidence. In the time between Sept 30, 2009 and the end of 2014, many additional reports have been sent to the FDA.

There are a few interesting things that can be seen from both the Moore, Glenmullen and Furberg study and the RxSK update, including:

• Of these 31 (35) drugs associated with violence, all but 8 are psychiatric drugs;

• 11 of the top 31 examined by MGF were antidepressants, including all 10 of the SSRIs. All these same medications plus 3 older antidepressants qualified for the update. (Imipramine was also included because of its PRR of 15.3 for suicide, but it did not qualify under the original criteria);

• When violence that did not involve a physical act of violence (e.g. homicidal ideation) was removed, different drugs joined the top group. Varenicline (Chantix) which showed up as the top drug for violence in the original study, and was second in the update, fell to 19th in the ranking.

• In the update, when only physical violence was counted, Diazepam (Valium) and Zolpidem (Ambien) moved from 12th and 5th, to 5th and 2nd

The main thing to remember is that even when research is accurate, what it shows is not always as easy and clear as we would like and it is important to interpret it carefully.

[1] Violence by people discharged from acute psychiatric inpatient facilities and by others in the same neighborhoods, by Steadman HJ1, Mulvey EP, Monahan J, Robbins PC, Appelbaum PS, Grisso T, Roth LH, Silver – Archives of General Psychiatry E.1998 May;55(5):393-401.

[2] How To Interpret That Study Linking Violence And Antidepressants; It doesn’t necessarily mean what you think. by Erin Schumaker – Huffington Post, Sept 17, 2015,

[3] Prescription Drugs Associated with Reports of Violence Towards Others, by Thomas J. Moore, Joseph Glenmullen, Curt D. Furberg – PLOS, Published: December 15, 2010 DOI: 10.1371/journal.pone.0015337

[4] Proportional Reporting Ratio. The PRR is defined as the ratio between the frequency with which a specific adverse event is reported for the drug of interest, relative to all adverse events reported for that drug, relative to the frequency with which the same adverse event is reported for all drugs

[5] A cure for crime? Psycho-pharmaceuticals and crime trends, by Dave E. Marcotte and Sara Markowitz

© 2010 by the Association for Public Policy Analysis and Management, DOI: 10.1002/pam.20544, 29 OCT 2010

[6] The curious case of the fall in crime, The Economist, July 20, 2013

[7] Violence by people discharged from acute psychiatric inpatient facilities and by others in the same neighborhoods, by Steadman HJ1, Mulvey EP, Monahan J, Robbins PC, Appelbaum PS, Grisso T, Roth LH, Silver – Archives of General Psychiatry E.1998 May;55(5):393-401.

[8] Healy D (2000), Emergence of antidepressant-induced suicidality. Primary Care Psychiatry

Read, Cartwright, and Gibson, 2014, Adverse emotional and interpersonal effects reported by 1829 New Zealanders while taking antidepressants, Psychiatry Research published April, 2014

[9] Selective Serotonin Reuptake Inhibitors and Violent Crime: A Cohort Study, by Molero, Lichtenstein, Zetterqvist, Gumpert, and Fazel, PLOS Published: September 15, 2015, DOI: 10.1371/journal.pmed.1001875

[10] Nearly 7 in 10 Americans Take Prescription Drugs, Mayo Clinic, Olmsted Medical Center Find, by Dr. Jennifer St. Sauver, Mayo Clinic News Releases, Jun 19, 2013

[11] Antidepressant-induced akathisia-related homicides associated with diminishing mutations in metabolizing genes of the CYP450 family, Lucire Y, Crotty C, DovePress Open Access to Scientific and Medical Research, August 2011 Volume 2011:4 Pages 65—81 DOI http://dx.doi.org/10.2147/PGPM.S17445

[12] To get zone data, go to RxISK.org, click on the zone of interest (suicide, hair, violence, etc). Then at the bottom of the page, enter the drug of interest, and click on CONTINUE