The concept of violent suicide, its underlying trait and neurobiology: A critical Perspective

Abstract

Suicide is one of the leading causes of death and represents a significant public health problem world-wide. Individuals who attempt or die by suicide represent a highly heterogeneous population. Recently, efforts have been made to identify sub-populations and variables to categorize them. A popular dichotomy in suicide research of the past years is violent versus non-violent suicide - based on the method. This dichotomy is important given that there is an association between method of attempted suicide and risk of subsequent death by suicide. The differentiation concerning suicide methods is also critical regarding preventive efforts. In this review, we have tried to approach the concept of violent suicide from different perspectives, including a discussion about its definition and overlapping categories. In addition, we have critically discussed aggression as underlying trait, the question of intent to die, and sociodemographic, environmental, neuropsychological, and neurobiological factors potentially associated with violent suicide.

1. Introduction

Suicide is one of the leading causes of death and represents a significant public health problem in the United States and around the world. The general demographic groups impacted by a high relative rate of death by suicide in the US are middle-aged, White men, and young Native American men (Nock et al., 2008). Cross-national data from the World Health Organization shows that the burden of suicide shifted over the past fifty years from Western to Eastern Europe and now seems to have the highest impact in Asia (Varnik, 2012).

Violent suicide is in most cases defined by the method. Several publications use the Asberg’s criteria (Giner et al., 2014; Penas-Lledo et al., 2015), which defines hanging attempts, the use of firearms, jumping from heights, several deep cuts, car crash, burning, gas poisoning, drowning, electrocution, and jumping under a train as violent attempts; whereas drug overdoses are considered to be nonviolent suicide attempts (Asberg et al., 1976). Others define violent suicide as all methods but poisoning (Stenbacka and Jokinen, 2015) and again other publications define violent suicide as all methods but drowning, gas poisoning and substance poisoning (Dumais et al., 2005). As for violent suicide, few findings are published on demographic characteristics. It has been well established that men use violent methods more often than women, both in suicide and suicide attempts. A large European study, involving four countries looked at these gender-specific differences in greater detail. Suicidal acts were 3-4 times more lethal in men than in women; men were more likely to choose more lethal suicide methods and the lethality was even higher in men when using the same method (Mergl et al., 2015). A switch from non-violent attempts to violent death by suicide was significantly more common in men compared to women (Bradvik, 2007). Individuals who attempt or die by suicide represent a highly heterogeneous population. Thus, efforts have been made to identify sub-populations and variables to categorize them. Popular dichotomies in suicide research of the past years are violent versus non-violent suicide attempters, high-lethality versus low-lethality attempters, impulsive versus non-impulsive suicide attempters and low intent to die versus strong intent to die. These dichotomous variables have been examined and operationalized by various scales (e.g., Beck’s Suicide Intent Scale). Definitions and subpopulations are overlapping; hence in this review, we will focus on violent suicide but equally address related terms and definitions.

2. Aggression as the personality trait underlying violent suicide

One of the prime suspects to consider, when searching for the underlying causes of violent suicide, is aggression as a trait. The hypothesis, although not explicitly formulated, suggests that aggression as a trait is associated with both self-directed and other-directed violence. Aggression as a personality trait is not included in the Big-Five model (Goldberg, 1990), but plays an important role in theory of psychoanalysis and is hence operational in the Structured Interview of Personality Organization (Horz-Sagstetter et al., 2017). This was used by Baus et al. in a study of 120 Borderline Personality Disorder patients (Baus et al., 2014). They found that patients with a history of suicide attempt had higher aggression scores and a worse overall level of personality organization. However, when differentiated between other-directed and self-directed aggression, it became clear that the determining factor was self-directed aggression (p <0.001) and not other-directed aggression (p= 0.297) (Baus et al., 2014). A similar study compared personality traits (by means of proxy-based interviews) of borderline patients, who died by suicide, to those who did not die by suicide. Impulsivity and aggression traits could predict suicide, but when controlling for Cluster B personality disorder, the in-between group variance was fully explained by this factor (McGirr et al., 2007). Giegling et al. (2009) examined a sample of suicide attempters in regards to their method (violent vs. non-violent) among other variables. Also assessing anger, aggression and temperament via several interview inventories, they tried to associate these traits with violent/non-violent suicide attempt, but no significant correlation was found (Giegling et al., 2009). A study comparing data from suicide research centers in New York City and Madrid, suggested that higher lethality of suicide in New York City and concomitant higher aggression scores (assessed by the Brown-Goodwin Scale) might be derived from the same diathesis (Baca-Garcia et al., 2006).

To further elucidate the association between suicide and violence, Jokinen et al. (2010) constructed and validated the Karolinska Interpersonal Violence Scale. It measures both exposure to violence and expressed violent behaviour in both childhood and during adult life (Jokinen et al., 2010). Contrary to previous findings, the association between violent behavior in adulthood and completed suicide was significant (Jokinen et al., 2010; Stefansson et al., 2015). These investigators also found an association between violent adult behavior and non-suicidal self-injury in a sample of suicide attempters (Sahlin et al., 2015). A recent multi-center study, using the same scale, found that a high overall score was associated with repeated suicide attempts and violent suicide attempts (Haglund et al., 2016).

Another approach was to regroup individuals who had a history of self-directed violence and those who had a history of violence against others and compare their personality traits (Stalenheim, 2001) or temperament (Engstrom et al., 1999) within the groups or with a control group. Engstrom et al. (1999) matched a sample of criminal offenders and a sample of suicide attempters for age and psychiatric diagnosis and found that most of their personality traits (based on the Karolinska Scales of Personality) did not significantly differ, displaying high trait of anxiety and low socialization. Interestingly, impulsiveness and inhibition of aggression were in the normal range (Engstrom et al., 1999), again not supporting the hypothesis of a common trait of aggression in these subpopulations.

The increased risk of suicide among criminal offenders has been widely reported (Webb et al., 2011) but the nature of this correlation has not been solved yet. Stålenheim (2001) for instance, found no correlation between suicide attempts and violent criminality in a sample of criminal offenders during forensic psychiatric examinations. The subgroup of suicide attempters within this sample of forensic patients showed significantly higher scores in impulsive aggression and had a lower socioeconomic upbringing than the other examined offenders (Stalenheim, 2001).

The above listed publications aim to correlate aggression as a trait or temperament with self-directed violence including violent and non-violent suicide attempts. Other studies listed hereafter aim to find correlations between violent behavior and suicide attempts, not explicitly defining aggression as a trait that can be measured but rather implicitly suggesting a common ground between them. Webb et al. (2013) found a positive correlation between violent offenses and suicide, although when controlling for the method (violent vs non-violent), it became evident that it was actually the rate of non-violent suicide that was elevated in violent offenders (Webb et al., 2013). Interestingly, several recent studies have revealed that social factors and psychiatric comorbidities play a crucial role in conveying the suicide effect in samples of criminal offenders. Stenbacka et al. (2014) followed a sample of 48,834 men over the course of 35 years and found a 5-fold elevated suicide rate in the subsample that had a record of violent offenses. Additionally, they found the subsample of non-violent criminal offenders to have a two times higher suicide rate compared to the other men being followed by the researchers. After controlling for social factors, the correlation decreased and when additionally controlled for psychiatric factors such as inpatient treatment for drugs and alcohol, the higher risk only stayed significant for the non-violent sample (Stenbacka et al., 2014). A similar outcome was presented in a retrospective case-control study; in the crude model sexual and violent offenders seemed to have an elevated suicide risk compared to non-violent offenders. After correction for psychiatric and social risk factors, the relative risk was almost equivalent to the risk among non-violent offenders (Webb et al., 2012). These two studies are strong advocates for the case that both violent suicide and violent offenses are not to be separated from social factors and that the environment plays a crucial role in the formation of a “violent trait”.

A study, that successfully established an association between violent behavior and suicide, focused on violent behavior within the last year of suicide victims. Even after controlling for alcohol misuse, which is a common confounder in that association, they found violence in the last year to be a predictor of suicide. Interestingly, this association was strongest in women and younger individuals (Conner et al., 2001), which comprise the demographic significantly underrepresented in completed suicide. A very well-known paradigm in this context is the cycle of violence (Widom, 1989), claiming that exposure to violence as a child predicts aggressive behavior as an adult. Moberg et al. (2014) revisited this hypothesis in a sample of suicide attempters and healthy volunteers as a control group. It was shown that aggressive behavior in childhood was predictor of aggressive behavior in adulthood and also exposure to violence predicted interpersonal violence in adult patients in this sample of suicide attempters but the multivariate logistic regression analysis was only significant when including substance abuse diagnosis and age as covariates (Moberg et al., 2014). A different approach to a similar question was undertaken by Dumais et al. (2005); life-time aggression was investigated in individuals who died by suicide. The method of suicide (violent vs. non-violent) was significantly associated with life-time aggression assessed via psychological autopsy, although the authors controlled for age, sex, substance disorders, and other major psychopathology as confounder variables (Dumais et al., 2005).

3. Question of intent and violent suicide

The questions of intent and the level of violence of a suicidal act are closely connected concepts. Suicide attempters with high intent – according to a study including 203 patients – are less likely to be married, more likely to be male, have a diagnosis of major depressive disorder (MDD) and experience a higher number of recent life events in the month preceding the attempt (Kumar et al., 2006). According to Banwari et al. (2013) suicide attempts in MDD and schizophrenia patients share common features but schizophrenia patients show a significantly higher intent to die, compared with MDD patients (Banwari et al., 2013). According to a study by Stefansson et al. (2015), intent to die and violence both contribute to the risk of suicide but are not correlated with each other and therefore confer two different components of the suicide risk. Contrary to these findings Jordan and Samuelson (2016) showed that repeated acts of committing violence were associated with high suicide intent in individuals with a prior history of suicide attempts (Jordan and Samuelson, 2016).

In a retrospective study, Giner et al. (2014) compared violent suicide attempters, defined as by Asberg’s criteria, with suicide attempters who did not meet these criteria. They also differentiated between serious suicide attempts, i.e., overdoses requiring hospitalization from non-serious attempts such as overdoses that did not require hospitalization. They concluded that violent and serious suicide attempters were significantly more likely to repeat suicide attempts and showed relatively higher lethality of their attempts. Violent suicide attempters defined by Asberg’s criteria were also more likely to be men and of older age than those who did not meet the criteria (Giner et al., 2014).

In the discussion of violent versus non-violent methods and intent to die, physical accessibility is an important factor. Anestis et al. (2015) found that suicide victims who died by the means of firearms had significantly fewer prior attempts than those who used other methods (Anestis, 2016). First of all this discrepancy could of course be explained by a different level of intent to die in-between groups, but it is also an important reminder of how closely death by suicide is associated with the physical accessibility of the method (Anestis and Anestis, 2015). On a side note it seems important to state that suicide prevention in the United States is impaired by the great percentage of suicidal individuals who already have access to fire arms, hence actions and policies are needed to limit the access of vulnerable populations to fire arms that they have purchased in previous years in order to make a difference in statistical numbers (Mann and Michel, 2016). A somewhat contradicting study shows that individuals, who used a non-violent method in the past, tend to die from non-violent methods rather than re-attempting suicide with a violent method (Jamison and Bol, 2016).

4. Sociodemographic and environmental factors

Little is known about early risk factors in violent suicide. A Finish study suggests that the risk factors might be similar to the general risk factors for suicide, however conduct problems at school might be more prevalent in violent suicide attempters (Stenbacka and Jokinen, 2015). An interesting association between birth length and violent suicide attempts was established by Swedish researchers by the help of birth records. Low birthweight and short birth length for gestational age were both associated with violent suicide (Mittendorfer-Rutz et al., 2008).

Vyssoki et al. (2014) observed potentially sunshine-induced changes in suicide rates over the course of 40 years. They found a significant positive correlation between number of sunshine hours per day and violent but not non-violent suicide (Vyssoki et al., 2014). Several preceding publications also suggest that there might be an association between hours of sunshine and violent suicide in Belgium (Linkowski et al., 1992; Maes et al., 1994), Italy (Preti and Miotto, 1998), and Australia (Lambert et al., 2003). Data from the subtropical region in Taiwan suggests that seasonality might also contribute to the rate of violent suicide. In their model of seasonality, they also included climatic data such as temperature, relative humidity, rainfall, hours of sunshine, and pressure. This model revealed that ambient temperature was significantly correlated with high rates of violent but not non-violent suicide. However, contrary to previous findings, they did not replicate the association of violent suicide with hours of sunshine (Lin et al., 2008).

With regard to the diagnosis and comorbidities, there are very few studies that focus on the clinical and epidemiological associations of violent suicide. A correlation between substance abuse disorder and non-violent (as opposed to violent) suicide attempts was established in a sample of Veterans Affairs (VA) patients (Ilgen et al., 2010). This finding is supported by another study linking non-violent suicide attempts to a higher blood alcohol concentration compared to suicide attempts by violent methods (Zupanc et al., 2013). Sheehan et al. (2015) examined the relationship between drug abuse and violent methods of suicide, aiming to establish correlations between a specific drug and a specific method of suicide. They found a correlation between the use of opiates and firearm suicide (Sheehan et al., 2015). Multivariate analyses in a sample of mood disorder patients revealed higher impulsivity and aggression levels as well as higher lethality of attempt in bipolar patients compared with MDD patients; this effect was particularly associated with male gender (Zalsman et al., 2006). Medico-legal investigation of 18,894 deaths revealed an association between the acute influence of tetrahydrocannabinol and the violent suicide method jumping from a height, whereas alcohol was not associated with any violent suicide method at all (Lundholm et al., 2014).

When it comes to prescription drugs such as antidepressants and its association with attempted suicide/suicide, many studies have established, that neither antidepressants as a group, nor SSRIs specifically increase of the risk of suicide compared to non-treated individuals (Isacsson et al., 2005; Tiihonen et al., 2006). The association between SSRIs and suicide by violent means has not been fully explored yet; the data seems to vary through age groups and gender (Fazel et al., 2007).

5. Neuroscience of violent suicide

5.1. Neuropsychology

Several publications include violent suicide attempters as a subsample in their design, aiming to find deficits in neuropsychological domains. Keilp et al. (2001) focused their tests on executive function in a sample of patients having major depressive episodes. Within this sample, suicide attempters were rated based on Beck’s medical damage scale and divided into high-lethality and low-lethality suicide attempters. Violent suicide attempts occurred almost exclusively in the high-lethality group. Individuals with high-lethality attempts showed the most impairment in memory and executive functioning and also had the lowest scores of all groups in general intellectual functioning and attention. Only motor functioning showed homogeneous results for all depressed subjects as well as healthy controls. Since impaired motor functioning, being a sensitive tool to detect or confirm diffuse brain damage such as caused by hypoxia, was not impaired in the high-lethality sample, the authors concluded that the impairments registered/assessed in this study might not be a result of the high-lethality attempts but a risk factor for committing high-lethal attempts (Keilp et al., 2001). A follow-up study, 12 years later (Keilp et al., 2013) included a higher sample number and an extended number of assessments. Interestingly, they did not replicate their previous findings when comparing high-lethality and low-lethality suicide attempters. No significant difference was noted in their performance in domains such as attention, memory, abstract/contingent learning and impulse control among others. High lethality attempters even outperformed low-lethality attempters in object alternation, which measures response inhibition and which is also particularly sensitive to ventral prefrontal cortex lesions. Additionally, it was also assessed if the last suicide attempt was violent or non-violent. They noticed that in this follow-up study the violent attempters were predominantly represented in the low-lethality group, whereas in the previous study they were primarily found in the high-lethality group. When focusing on the group of violent suicide attempters, it became clear that their performance was significantly poorer in abstract/contingent learning and impulse control and a trend into the same direction could be observed for object alternation (Keilp et al., 2013).

In another study following up on these previous findings, the same group focused on ventral prefrontal cortex (VPFC) dysfunction in suicide attempters and replicated the finding that VPFC dysfunction might not be characteristic for all suicide attempters. They suggest that carefully-planned, non-violent, highly-lethal suicide attempts might occur in a specific subgroup of suicide attempters who do not share the same characteristics such as VPFC dysfunction with the rest of the population (Keilp et al., 2014).

The Iowa Gambling Task is a very common tool to measure impaired decision-making. A study testing four groups (healthy control subjects, affective disorder patients, violent suicide attempters and non-violent suicide attempters) revealed significantly higher disadvantageous choices in the group of suicide attempters compared to all other groups even when controlling for confounders such as substance abuse and previous experiences of trauma or coma (Jollant et al., 2005). A similar study including healthy subjects, affective disorder patients without previous suicide attempts, non-violent suicide attempters and violent suicide attempters found no significant difference but a trend of violent suicide attempters having the poorest performance in the Iowa Gambling Task (Gorlyn et al., 2013). Another aspect of suicide and decision-making is delayed gratification versus impulsivity. Delay discounting measures the preference between immediate and delayed gratification. When comparing high-lethality suicide attempters to four other subgroups (non-suicidal depressed patients, low-lethality suicide attempters, suicidal ideators and healthy controls) within a sample of elderly individuals, Dombrovski et al. (2011) showed that low-lethality suicide attempters had the strongest preference for immediate rewards compared to the other groups, whereas high-lethality suicide attempters were more willing to wait for larger rewards.

5.2. Neurobiology

Early studies, examining the neurobiology of violent suicide, found benzodiazepine receptor to be upregulated in PFC postmortem tissue of suicide victims who died by violent means compared to other subgroups of suicide and controls who died for natural causes (Pandey et al., 1997; Rochet et al., 1992).

The implication of cytokines in the pathogenesis of suicide is becoming more and more prominent. Interleukin-6, measured in cerebrospinal fluid (CSF), peripheral blood, and postmortem brain tissue, has shown robust associations with violent suicide and suicide completion (Gananca et al., 2016). In addition, high CSF-insulin in a violent sub-sample of recent suicide attempters was independent of MDD diagnosis (Westling et al., 2004).

The ABCB1 gene codes for an energy-dependent efflux pump that has been implicated in drug-resistance. A genotypic analysis of survivors of a recent suicide attempt revealed a significant association between the ABCB1 haplotype TTT in three SNPs and violent method (according to Asberg’s criteria) (Penas-Lledo et al., 2015). Previous findings from forensic autopsies had already suggested an association between completed suicide and ABCB1 polymorphisms (Boiso Moreno et al., 2013).

Perroud et al. (2008) investigated a sample of violent suicide attempters in regards to a gene-environment effect of the functional polymorphism Val66Met (BDNF) and sexual abuse. They found that violent suicide attempters with the Val/Val polymorphism were more likely to have suffered abuse compared to Val/Met and Met/Met polymorphism carriers (Perroud et al., 2008). Interestingly, a similar study found a negative correlation between the Val/Val polymorphism, impulsive aggression and serious life events in borderline personality disorder patients (Wagner et al., 2010). In order to identify new candidate genes, a transcriptomic expression profile of orbitofrontal cortex tissue of violent suicide victims and non-psychiatric controls was performed and the results included potentially biologically relevant agents of the ‘central nervous system development’, ‘homophilic cell adhesion’, ‘regulation of cell proliferation’ and ‘transmission of nerve impulse’ (Thalmeier et al., 2008).

The relationships between cholesterol, suicide and violence have been extensively studied in the past 20 years. The first study to demonstrate the association between violent suicide attempts and cholesterol levels was by Tanskanen et al. (2000). Using five independent population surveys as well the National Mortality Register, it was shown that the serum total cholesterol concentration was positively related to violent but not non-violent suicide. The same trend could be shown for the concentration of high density lipoprotein cholesterol, but not reaching statistical significance (Tanskanen et al., 2000). In the same year, another study examined this relationship comparing the serum cholesterol levels of drug-free violent suicide attempters and non-violent suicide attempters following a drug-overdose. Interestingly, they found the cholesterol levels to be 30% lower in the violent suicide attempter samples, even when controlling for sex and age (Alvarez et al., 2000). Two studies by the same group investigated a sample of schizophrenic suicide attempters and found significantly lower cholesterol serum levels in violent suicide attempters compared with suicide attempters using non-violent methods (Marcinko et al., 2005). They also found significantly lower serum cholesterol levels in the violent suicide attempter group compared to a control group of schizophrenic patients (Marcinko et al., 2004). A similar case-control study looked at women with a history of suicide attempts and compared the serum cholesterol concentration between violent suicide attempters and non-violent suicide attempters. Again, a lower concentration was predictive for violent methods within this sample of patients (Vevera et al., 2003). Interpersonal violence in suicide attempters, as previously mentioned, has been explained by the cycle of violence (Moberg et al., 2014). Asellus et al. (2014) tried to find biological mediators of this association. Again, assessing exposure to violence in childhood as well as adult interpersonal violence in a sample of patients with a history of suicide attempts, they additionally measured levels of serum cholesterol and found it to be a predictive marker. When dividing patients into two groups based on their serum cholesterol levels, the association between exposure to violence as a child and reported interpersonal violence in adulthood was only significant for the sample of patients having serum cholesterol levels below the median (Asellus et al., 2014). Two postmortem studies, relevant to the question of fatty acid composition and violent suicide, reported the following findings: significantly lower cholesterol levels in the PFC of violent suicide attempters, significantly higher cholesteryl ester hydrolase (LIPA) expression in violent suicide attempters, a significantly lower cholesterol/phospholipid level ratio in violent suicide attempters and increased levels of phospholipids (mediated by LIPA) in violent suicide attempters when compared with non-violent suicide attempters (Freemantle et al., 2013; Lalovic et al., 2007).

It seems rather natural also to suspect an involvement of HMG-CoA inhibitors as a potential culprit since there are reports of psychiatric adverse drug reactions. But the evidence remains unclear due to the limited amounts of trials including psychiatric patients (Cham et al., 2016). Other studies suggest adjunct antidepressant effects of statins when prescribed with SSRIs (Kohler-Forsberg et al., 2017). Given that chronic cholesterol depletion (by statins) cause serotonin receptor 1A (5HT1A) dysfunction in vitro (Shrivastava et al., 2010), there might be a balancing effect in the theory behind adjunct therapy.

Studies related to violent suicide and the serotonin axis include candidate gene studies, gene expression studies, postmortem brains studies, plasma serotonin levels as well as studies involving the cerebrospinal fluid. In a candidate gene study with a sample of schizophrenic patients, who were genotyped for the promoter of the serotonin-transporter gene (5HTTLPR) polymorphism, significant differences in the genotype frequencies were found when comparing a subgroup of patients who had a history of violent attempts with a subgroup of patients, who had a history of non-violent attempt and to a subgroup who had no history of suicide attempt at all. A low-activity short allele was found more frequently in the sample of violent attempters (Bayle et al., 2003). Another study comparing suicide attempters to a healthy control sample found no significant differences in the genotype frequency of the 5HTTLPR polymorphism. The suicide attempters consisted of predominantly (98%) of suicide attempts by violent means (Hranilovic et al., 2003). Tryptophan hydroxylase (TPH) is the rate-limiting enzyme in the synthesis of serotonin. Stefulj et al. (2005) examined promoter polymorphisms of the TPH1 gene and found no significant differences between a sample of violent suicide attempters and healthy controls, although they yielded for a very homogeneous sample (only men of Croatian descent) (Stefulj et al., 2005). In another study, the same group found age-dependent differences in the genotype frequency in polymorphism A218C of the same gene; elderly victims of violent suicide were more likely to have the CC genotype compared with healthy controls (Stefulj et al., 2006). A postmortem study on the locus coerelus in suicide attempters and controls revealed an increased number of tyrosine hydroxylase immunoreactive neurons in violent suicide victims compared with non-violent suicide victims and controls (Gos et al., 2008). A genetic study of adolescents with suicide attempts by Zalsman et al.(2011) found the homozygosity for the T allele of the HTR2A 102T/C polymorphism to be associated with lower impulsivity (p=0.03) and aggression (p=0.01) compared to TC carriers. Monoamino oxidase A (MAOA) was also genotyped, the gene codes for an enzyme responsible for the degradation of serotonin as well as other neurotransmitters; hence it might be implicated in the pathogenesis of psychiatric disorders. They found low activity MAOA genotypes to be associated with suicidality (p=0.04) (Zalsman et al., 2011).

A candidate gene study with the MAOA-uVNTR polymorphism in a sample of violent and non-violent suicide attempters revealed no significant differences in allele frequency. The authors also performed a meta-analysis of the same polymorphism pooling 7 case-control studies and focusing on suicidal behavior in general. No significant associations were found (Hung et al., 2012). Alvarez et al. (2000) found lower serotonin (5-HT) content in platelets of violent suicide attempters compared to controls. 5-Hydroxyindoleacetic acid (5-HIAA) is a serotonin metabolite, which can be measured in the CSF. 5-HIAA as well as the dopamine metabolite homovanillic acid (HVA) were assessed in the CSF of recent suicide attempters who used violent methods. CSF 5-HIAA concentrations in the violent suicide group were significantly lower than in control subjects and when comparing impulsive violent attempters to non-impulsive attempters, impulsive suicide attempters again had significantly lower levels of HIAA. HVA levels showed no significant difference (Cremniter et al., 1999). A similar study studied plasma serotonin (5-HT), 5-HIAA, HVA, and platelet 5-HT levels in recent violent suicide attempters. Plasma 5-HIAA and platelet 5-HT concentrations were significantly lower in violent suicide attempters than in controls and plasma HVA was again was not associated with suicide behavior (Spreux-Varoquaux et al., 2001). Moberg et al. (2011) hypothesized that exposure to violence in the childhood might have an effect on the serotonin axis in adulthood. They measured CSF 5-HIAA levels in a sample of suicide attempters and found significantly lower levels in suicide attempters with higher scores in adult interpersonal violence. In women, the level of CSF 5-HIAA was inversely correlated with the exposure to violence as a child (Moberg et al., 2011).

Wallner and Machatschke (2009) proposed an evolutionary hypothesis for the associations between violence, cholesterol and serotonin. Central cholesterol plays a key role in the exocytosis of serotonin vesicles and hence it was suggested that there might be regulatory mechanisms in place that diminish the threshold for aggressive behavior in times of starvation (Wallner and Machatschke, 2009). Since most of the central cholesterol is synthesized in situ, and only a small fraction of it crosses the blood-brain barrier (Dietschy and Turley, 2001), this theory has its own limitations.

5.3. Neuroimaging

So far – to the best of our knowledge- no studies have been published approaching violent suicide attempts by neuroimaging techniques. One study worth discussing included subsample patients with a history of recent violent suicide attempt. They measured the binding potential of the serotonin transporter (5HTT) and dopamine transporter (DAT) by single photon emission computerized tomography (SPECT) in patients with a recent serious suicide attempt. There was no significant difference in DAT or 5HTT binding between the violent suicide attempters, the non-violent suicide attempters and healthy controls (Lindstrom et al., 2004). Also, it has been proposed that increased serotonin_2A receptor binding (5-HT_2A) might be most prevalent in suicide victims who died by violent means (Meyer, 2012).

6. Conclusions

In this review, we have tried to approach the concept of violent suicide from different perspectives, including a discussion about its definition and overlapping categories, the role of the supposedly underlying trait “aggression” and general findings in neurobiology, neuropsychology and neuroimaging.

The question of aggression remains a matter of debate: several findings (Baca-Garcia et al., 2006; Conner et al., 2001; Dumais et al., 2005; Haglund et al., 2016; Jokinen et al., 2010; Stefansson et al., 2015; Webb et al., 2011) point to aggression as an underlying trait in both self-directed and other-directed violence, whereas as many other publications (Baus et al., 2014; Engstrom et al., 1999; Giegling et al., 2009; McGirr et al., 2007; Stalenheim, 2001; Stenbacka et al., 2014; Webb et al., 2012) found this not be true. We also discussed the question of intent and how it is different from the concept of violent suicide, as well as the questions of physical accessibility of violent means. Environmental factors taken into account include hours of sunshine, climate and the influence of legal or illegal drugs. Neuroscience findings that have been addressed here include neuropsychological studies, neurobiological studies and neuroimaging studies. Most studies related to violent suicide are using a neurobiological approach, in which the triangle of suicide-violence-serotonin features as most prominent. Neuroimaging might not be the most adequate method to explore the neuroscientific background of violent suicide, since violent suicide attempts are more likely to be lethal and the sample of patients that qualify for neuroimaging only constitutes a subsample of patients – the individuals who survived a violent suicide attempt. The same of course is true for studies including peripheral blood markers and candidate gene studies. Hence, post mortem brain studies seem to be the best approach to further elucidate the neurobiological basis of violent suicide.