45 Biological predictors of suicidal behaviour in mood disorders

Predicting suicide is difficult due to the low base rate, even in high-risk groups, and the multi-causal nature of suicidal behaviour. Retrospective and cross-sectional studies have identified a number of biologic abnormalities associated with suicide and suicide attempt. Prospective studies provide estimates of the predictive utility of biologic measures. Here, we review prospective studies of suicidal behaviour and serotonergic, noradrenergic, dopaminergic systems and the hypothalamic–pituitary–adrenocortical (HPA) axis function in mood disorders. The most promising biologic predictors are low CSF (cerebrospinal fluid) 5-hydroxyindoleacetic acid (5-HIAA) and HPA axis dysfunction as demonstrated by dexamethasone non-suppression that are each associated with about 4.5-fold greater risk of suicide.

Identifying individuals at imminent risk for suicidal behaviour is a major challenge for clinicians. However, prediction of suicidal behaviour is difficult due to the relative rarity of the event as well as the multidetermined causes of such behaviour.

This chapter reviews prospective studies of the serotonergic, dopaminergic and noradrenergic systems and hypothalamic–pituitary–adrenal axis, and systems in relation to suicidal behaviour in mood disorders, to assess the potential of biologic measures for improving the prediction of suicidal behaviour.

Prospective studies of suicide completion and the serotonergic system consistently report that, in mood disorder individuals, low baseline cerebrospinal fluid levels of 5-HIAA (the serotonin metabolite) levels and a history of attempting suicide predict those who go on to complete suicide (Åsberg et al. 1976a, 1976b; Roy et al. 1989; Träskman-Bendz et al. 1992a; Samuelsson et al. 2006). The association between CSF 5-HIAA and suicide attempts is less clear (see Mann et al. 1996 for a review). Cross-sectional and retrospective studies find little evidence for a relationship between violent suicide attempt method and CSF 5-HIAA levels. Rather, it seems that the choice of a violent method is more dependent on what methods are available or commonly used in a society. However, greater planning and medical lethality of suicide attempt correlate with lower CSF 5-HIAA (Mann and Malone 1997; Placidi et al. 2001; Oquendo et al. 2003b). Insofar as the time between sampling CSF 5-HIAA and suicide attempt did not affect this relationship to lethality (Mann and Malone 1997) it appears that CSF 5-HIAA is a stable biochemical trait that may prospectively predict suicide and the lethality of non-fatal suicidal behaviour.

Suicidal acts are associated with aggressive and impulsive traits that, in turn, are also associated with serotonergic dysfunction (Oquendo and Mann 2000; Baca-Garcia et al. 2001; Placidi et al. 2001; Oquendo et al. 2003b). In prospective studies of alcoholic fire setters and violent offenders, lower levels of CSF 5-HIAA predict future aggression against property or homicide (Virkkunen et al. 1989, 1996). A 15-year follow-up study of army veterans (Faustman et al. 1993), supports the association between suicide, aggression and lower CSF 5-HIAA insofar as it found that those under 40 years of age who died by suicide, accident, or homicide, had significantly lower CSF 5-HIAA and homovanillic acid (HVA) compared with living controls.

Lower CSF 5-HIAA has also been associated with severity of lifetime aggressivity and a history of higher lethality suicide attempt (Träskman-Bendz et al. 1992a; Mann et al. 1996; Placidi et al. 2001). It is likely that definitions of impulsiveness and aggressive traits need to be refined. Perhaps greater suicide intent correlates with greater aggression, but inversely with impulsiveness, in which case serotonin system deficiency would more likely be related to intent rather than impulsiveness. Alternatively, different parts of the serotonin system may modulate intent and impulsiveness, as we have reported in depressed suicide attempters using positron emission tomography (PET) scanning (Oquendo et al. 2003b). The causal relationship between lower serotonin function and aggressive and/or impulsive behaviours has been demonstrated by the increase in aggressiveness and impulsiveness following lowering serotonin function transiently by acute tryptophan depletion in healthy male volunteers (Salomon et al. 1994; Cleare and Bond 1995; Moeller et al. 1996). Clearly the same kind of study would be unethical in those at risk for suicidal behaviour.

Clinical prospective studies have suggested that risk for suicidal acts might be associated with phase of depressive episode or subtype of depressive disorder (Berglund and Nilsson 1987; Leon et al. 1999; Brinkman-Sull et al. 2000; Coryell et al. 2002). Serotonergic abnormality has been observed in euthymic patients with a history of major depression, via a depressive response to tryptophan depletion (Price et al. 1991; Heninger et al. 1992; Delgado et al. 1994; Moreno et al. 1999; Smith et al. 1999) and a blunted prolactin response to serotonin release by fenfluramine during remission (Flory et al. 1998; Smith et al. 1999), suggesting that an abnormality in serotonergic function may underlie a predisposition to recurrent episodes of major depression. A prospective study of CSF 5-HIAA across the course of depressive illness found that, while CSF 5-HIAA increased on recovery, levels remained lower in individuals who had the lowest levels during a depressive episode compared to those with higher levels during a depressive episode and who achieved recovery (Träskman-Bendz et al. 1984). It is crucial to note that the abnormalities in serotonergic function associated with suicidal behaviour can be distinguished from those of a major depressive episode/disorder by their anatomical distribution in the brain (Mann et al. 2000). Thus, individuals that have mood disorders have a widespread abnormality in serotonin function affecting most of the prefrontal cortex and many other cortical and subcortical areas (Milak et al. 2005). In contrast, in post-mortem brain tissue from suicides, or in depressed suicide attempters identified by PET, is the abnormality is localized to parts of the prefrontal cortex, perhaps reflecting those brain regions involved in suicide intent, decision-making, and impulse regulation (Mann et al. 2000; Oquendo et al. 2003b). The generalizability of this localized serotonin system deficit and suicidal behaviour is supported by findings of low CSF 5-HIAA in suicide attempters with schizophrenia or personality disorders compared to psychiatric controls (Brown et al. 1982; Ninan et al. 1984; Gardner et al. 1990; Cooper et al. 1992).

Genetic studies may shed light on the origins of serotonergic system dysfunction in suicidal behaviour. In a longitudinal genetic study, Caspi et al. (2003) found a functional polymorphism in the promoter region of the serotonin transporter (5-HTTLPR) was associated with the likelihood of developing major depression and suicidality in relation to stressful life events in adulthood.

Abnormality in the dopaminergic system has been documented in major depression (for reviews see Mann and Kapur 1995; Dailly et al. 2004), however, with regard to suicidal behaviour, post-mortem and retrospective studies of dopaminergic function are few and inconclusive (see Mann 2003). Prospective studies of dopaminergic system function and suicidal behaviour have also produced divergent findings. (Roy et al. 1986; Roy 1992).

The ratio between monoamine metabolites has shown a stronger association with suicidal behaviour, although the predictive value of these ratios is unclear. Such ratios factor out common variance due to characteristics such as the shared CSF transport system and effects on monoamine metabolites levels related to CSF gradient due to variation in length of the spinal canal.

Engstrom et al. (1999) found lower HVA/5-HIAA and HVA/MHPG (3-methoxy-4-hydroxyphenylglycol) ratios in individuals with a history of suicide attempt compared with surgical controls at baseline, although there was no difference during follow-up between past attempters and those who completed suicide. Roy et al. (1986) reported a lower baseline CSF HVA/ 5-HIAA ratio in depressed subjects compared to controls. Moreover, among depressed subjects who had a history of suicide attempts, dexamethasone suppression test (DST) non-suppressors had a significantly lower mean CSF HVA/ 5-HIAA ratio than suppressors. This may indicate that depressed individuals who attempt suicide have a more marked imbalance between the turnover of dopamine and serotonin in terms of relatively lower dopaminergic activity or turnover.

There is evidence that abnormal functioning in the noradrenergic system is associated with both major depression and suicidal behaviour. In post-mortem studies fewer noradrenergic neurons in the locus coeruleus in the brainstem have been observed in depressed suicide victims (Arango et al. 1996), along with indications of cortical noradrenergic overactivity such as decreased alpha and high affinity beta₁-adrenergic receptor binding (Arango et al. 1993). While data are limited on the role of the noradrenergic system in suicidal attempt and suicide (see Mann 2003), severe anxiety and/or agitation increase suicide risk and are associated with noradrenergic overactivity (Fawcett et al. 1997).

Additional support for a role for increased noradrenergic activity as a predictor of suicidal behaviour is provided by a treatment study of a norepinephrine reuptake inhibitor (NRI), maprotiline, that found that individuals maintained on the NRI after remission of depressive episode have higher rates of suicidal behaviour than those on placebo despite a lower likelihood of depression relapse (Rouillon et al. 1989).

HPA axis abnormalities, most commonly dexamethasone resistance, have been observed in suicidal patients in diagnostically heterogeneous populations (Bunney et al. 1969; Meltzer et al. 1984; Nemeroff et al. 1988; Roy 1992; Träskman-Bendz et al. 1992b; Inder et al. 1997; van Heeringen et al. 2000; Brunner et al. 2001; Coryell and Schlesser 2001). Prospective biological studies of suicidal behaviour have investigated HPA function using both urinary measures of cortisol and the dexamethasone suppression test (DST).

In prospective studies of mood disorder samples that include both individuals with and without prior suicide attempts, seven of nine studies reported that the majority of subjects who completed suicide over the course follow-up were DST non-suppressors (Boza et al. 1988; Carroll et al. 1981b; Coryell 1990; Coryell and Schlesser 1981; Norman et al. 1990; Roy et al. 1986; Yerevanian et al. 1983) while two studies found no relation (Black et al. 2002; Träskman-Bendz et al. 1992a). Coryell and Schlesser (2001) estimated that, over a 15-year follow-up period, DST non-suppressors had a fourteenfold higher risk of suicide compared to suppressors. In that study, the next most powerful predictor, a prior serious suicide attempt, indicated only a threefold increase in risk.

For suicide attempt, three of seven studies found an association between DST non-suppression and seriousness of suicide attempts at baseline (Targum et al. 1983; Norman et al. 1990), and over the follow-up period (Coryell 1990). Serious suicide attempts, including attempts prior to baseline resulting in high medical damage (Norman et al. 1990) and necessitating hospitalization (Targum et al. 1983) were associated with DST non-suppression. In the third study, DST non-suppressors were more likely to make a psychologically, rather than medically, serious attempt in follow-up (Coryell 1990). Three further studies (Carroll et al. 1981b; Roy 1992; Träskman-Bendz et al. 1992a) reported baseline associations between violent method of suicide attempt and abnormal HPA axis function, however, only one study achieved statistical significance. In that study Roy (1992) found that at baseline individuals who had a previous violent attempt had significantly higher maximum post-DST plasma cortisol levels than those who had a previous non-violent attempt, however, no significant differences were observed between attempters, violent or not, and non-attempters during a 5-year follow-up.

Non-suppression on the DST may be associated with suicide because it predicts non-response to antidepressant treatment or a tendency for early relapse. Prospective studies provide evidence for both possibilities. Two studies (Yerevanian et al. 1983; Targum 1984) noted that DST non-suppressors, particularly those who fail to normalize over the course of inpatient treatment, had worse outcomes in terms of depression remission and relapse, circumstances which clinical follow-up studies have suggested increase risk for future suicidal acts (Oquendo et al. 2002). Targum et al. (1983) found that MDD subjects who normalized during treatment had the same relapse rate as non-suppressor psychiatric controls, while those who did not normalize had a higher incidence of relapse and poorer treatment response compared to normalizers.

Developing sensitive prediction models for suicide and suicide attempt is crucial for prevention but is difficult due to the multiplicity of contributory risk factors and the low base rate of suicidal behaviour. Anomalies in several biological systems have been associated with suicidal behaviour in mood disorders and prospective biological studies, while not yet conclusive, suggest some potential for prediction based on biological measures. Meta-analysis of prospective studies of completed suicide show that CSF 5- HIAA levels and dexamethasone non-suppression yielded odds ratios for prediction of suicide of 4.48 and 4.65 respectively (Mann et al. 2006). Given the multi-determined nature of suicidal behaviour no one biological index will be adequate to predict suicidal behaviour, however including multiple biological markers in a model, for example CSF 5-HIAA and dexamethasone response, in order to assess both trait and state related risks (Mann et al. 2006) may increase predictive power. Including more than one tests results in some trade-off of sensitivity (requiring a positive result on any single test) versus specificity (requiring a positive result on more than one test). Therefore, also integrating other biological tests reviewed elsewhere, into multivariate predictive models alongside clinical and genetic risk factors to develop still more sensitive and specific predictive models, is a major challenge for this field of research (Mann and Currier 2007).