Eur Arch Psychiatry Clin Neurosci (2001) 251: Suppl. 2, II/72–II/75 © Steinkopff Verlag 2001

� Abstract Affective disorders are characterized by firsta high recurrence risk, second a 30–50 times increasedsuicide risk and third a 2- to 3 times increased overallmortality. In contrast to a populistic belief no scientificevidence exists that antidepressant treatment, particu-larly long-term treatment, could reduce the the risk ofsuicidal acts in depressive patients with a history of sui-cide attempts. Data, however, coming from interna-tional, systematic, retrospective analyses of well-docu-mented long-term courses of illness in reliably diagnosedpatients, and from a large national, prospective long-term trial on the prophylactic efficacy of lithium versuscarbamazepine and amitriptyline has accumulated in thelast 10 – 15 years strongly supporting a (possibly specific)antisuicidal effect of lithium. The large collaborativeIGSLI study (International Group for the Study ofLithium-treated Patients) covering 5,616 patient yearsclearly showed that adequate long-term lithium treat-ment significantly reduces and even normalizes theexcess mortality of patients with affective disorders. Ametaanalysis on 17,000 patients pooled from 28 studiesdemonstrated that the rate of suicidal acts is 8.6 foldhigher in patients without lithium as compared to thosewith regular lithium treatment.

A post-hoc analysis of a large multicenter, controlledlong-term trial found no suicidal acts in 146 patientsrandomized to lithium compared to 9 suicidal acts in 139patients randomized to carbamazepine.

Reanalysis of the data from the IGSLI study supportsthe concept of the specificity of lithium, i.e., evidencecould be provided that lithium also reduces suicidalbehavior in patients who do not benefit from the lithiumtreatment in terms of episode reduction.

Conclusion Lithium has to be considered as a first linemood stabilizer in affective disorders, particularly inpatients with a history of suicide attempts. Extreme cau-tion is required when lithium is discontinued or a patientis switched to another mood stabilizer, because such apatient might have been protected against suicidalimpulses in spite of an incomplete response as to thenumber and quality of depressive/manic episodes.

� Key words Suicide · lithium · specificity · carba-mazepine · affective disorders

Dedicated to Prof. Per Bech in appreciation and admiration for his con-tinuous support and promotion of a rational and optimized pharma-cotherapy in psychiatry.

Introduction

Affective disorders are characterized by 1) the high riskof recurrences and 2) the 2- to 3-fold increased mortalityrate which is mainly caused by the 30 – 50 times higherrisk of suicide. For many years it was a silent, apparentlyself-evident assumption that vigorous treatment of acutedepressive episodes and antidepressive long-term treat-ment would also reduce the suicide risk and the excessmortality. However, in spite of the worldwide use of anti-depressants, epidemiological studies hardly supportedsuch hope. The question, thus, arises: are there any phar-macological strategies to prevent suicidal acts and toreduce the overall excess mortality? In fact, a number offindings from the past 15 years, coming from varioussources, various groups and various countries have accu-mulated sound evidence that appropriate lithium long-term medication, administered and monitored accord-ing to the state of the art, does decrease the suicide riskand thus reduces or even normalizes the excess mortal-ity in bipolar and schizoaffective as well as unipolarpatients.

Prof. Dr. med. B. Müller-Oerlinghausen (�)Jebensstr. 3, 10623 Berlin, GermanyTel.: +49-30-31001-361Fax: +49-30-31001-366E-Mail: bmoe@zedat.fu-berlin.de

B. Müller-Oerlinghausen

Arguments for the specificity of the antisuicidal effect of lithium

II/73

Historical aspects of the antisuicidal effect of lithium

How did the detection of this fascinating antisuicidaleffect come about?

Based on observations made in our lithium clinic in1985 we began a systematic follow-up of a group ofpatients with a high suicide risk, i.e., those with at leastone suicide attempt in their history before onset oflithium prophylaxis and with lithium medication main-tained for at least one year. We observed a very markedreduction of suicide attempts which intriguinglyoccurred not only in responders towards lithium pro-phylaxis – in terms of episode reduction – but also innon-responders. In addition, many more suicidal actswere committed by patients having discontinued lithiumthan by those under regular lithium treatment (Table 1).

At that time in the early 1980s only scattered infor-mation was available in the literature on a possible asso-ciation of lithium prophylaxis and change of suicidality.In addition, Barraclough (1972) had observed that sui-cide victims sufferinged from mental illness often hadreceived inadequate drug treatment. He estimated that,in his sample, 20 of 100 suicides could have prevented byappropriate lithium treatment. The best documentedinformation came from Hanus and Zapletálek (1984).They reported that whereas 25 out of 99 patients withaffective disorders tried to commit suicide in a period of5 years preceding the start of lithium treatment, only 4suicide attempts and no suicides occurred within a 5-years course of lithium treatment.

After the fall of the wall between West and East Ger-many we became aware of the findings of Felber et al.(1994) from Dresden, who had observed practically iden-tical changes as we had in the former West Berlin, butduring the time of socialism he had not been allowed topublish his findings. The Germany findings were alsosupported by a publication from the U.K. (Coppen et al.1991) referring to mostly unipolar patients (Table 2).

These preliminary findings were the beginning of avery fruitful international scientific collaboration over

the past 15 years. In 1986, Mogens Schou had started theInternational Group for the Study of Lithium-treatedPatients (IGSLI) which included 6 specialized lithiumcenters in different countries. The outcome criterion ofthe first IGSLI mortality study was the standardized mor-tality ratio (SMR), i.e., the death rate of a patient sampleis compared with the odds for dying in an exactly indi-vidually matched sample of the general population(Müller-Oerlinghausen et al. 1992). The IGSLI groupfound an SMR of 0.89, i.e., it did not differ from the mor-tality of the normal population. We could also show insuccessive studies that the mortality increased after dis-continuation of lithium to the expected values inuntreated patients with affective disorders (Müller-Oerlinghausen et al. 1992). Nilsson (1995), although shedid not observe a full normalization of the SMR, alsoshowed a significant increase after discontinuation oflithium up to the expected figure of around 3 (Table 3).

Tab. 1 Main findings of the first Berlin study on the antisuicidal properties of lithium(Causemann and Müller-Oerlinghausen 1988; Müller-Oerlinghausen et al. 1992)

Lithium treatment > 1 year N = 68

No. of suicide attempts before lithium treatment 2.1 � 0.2 55 patients on regular lithium treatment 2 suicides

4 suicide attempts 13 discontinuers of lithium 4 suicides

7 suicide attempts

No. of suicide attempts/patient Before lithium During lithium

Responders n = 42 1.3 � 0.2 0.00** Non-responders n = 26 1.5 � 0.2 0.23 � 0.13**

** = P < 0.01

Tab. 2 Completed suicides in patients on and off lithium

On lithium Off lithium

Müller-Oerlinghausen et al. (1992) 1 of 55 4 of 13 Felber and Kyber (1994) 1 of 36 3 of 36 Coppen et al. (1991) 1 of 103 13 of 103*

* nontreatment group

Tab. 3 Standardized mortality rate (SMR) of patients with affective disorders duringappropriate lithium treatment and after its discontinuation

SMR

During lithium After treatment discontinuation

Müller-Oerlinghausen et al. (1992) 0.89 2.54** Lenz et al. (1994) 0.86 1.8* Nilsson (1995) 1.8*** 3.1***

(SMR 1.0 indicates that mortality does not differ from that of the general popula-tion) significantly different from 1.0: *p < 0.05, **P < 0.01, ***p < 0.001

Tab. 4 Standardized mortality rate (SMR) of 5616 patient years (average lithiumtreatment time 81 months) (Ahrens et al. 1995)

SMR 95 % conf. interv. Expected

All deaths 1.14 0.74 – 1.69 ca. 2 – 3 Suicides 5.22 1.70 – 12.18 ca. 30 – 78 Cardiovasc deaths 0.93 0.45 – 1.71 ca. 3 – 8 Other 1.04 0.64 – 1.61

Reanalyzing the data from the IGSLI main studyallowed us to differentiate between the suicide mortalityand other causes of mortality. As Table 4 shows, the spe-cific suicide-induced mortality – in contrast to the car-diovascular mortality – is not fully normalized, but it is

about one tenth of what one would expect in untreatedpatients (Ahrens et al. 1995).

In the meantime, other groups have replicated andsupported these original observations, and it should beemphasized that at present no study exists which wouldhave falsified the concept of a suicide preventive effect oflithium long-term treatment (Schou 1998). Tondo et al.(1997) reviewed the existing evidence and reported thatthere is an 8.6-fold increased risk of suicidal acts inpatients off lithium as compared to those on lithium.These data are extracted from 28 studies covering 17,000patients. Further support comes from a recent and care-ful Swedish study by Kallner et al. (2000) which refers tomore than 6,000 patient years and which supports theIGSLI analysis of 5,600 patient years, since according tothese authors the suicide- induced mortality is stillincreased (SMR = 14.0), at least in bipolar patients, but isdoubled (SMR = 33) in patients off lithium. It also indi-cates that patients being treated in a specialized lithiumclinic have a lower SMR than those being taken care byGPs, psychiatrists in private practices, etc.

Specificity of lithium?

In view of these rather robust and consistent empiricalfindings, the intriguing question arises whether the sui-cide- preventive effect of lithium should be considered aspecific effect and what could be the underlying mecha-nism?

For the sake of clarity we may subdivide the issue ofpotential specificity into two questions:1. Is this effect specific for lithium salts, i.e., is it not

shared by other drugs, such as other mood stabilizersor antidepressants, etc.?

2. Is this effect strictly related to the episode-preventiveeffect of lithium prophylaxis or might it act inde-pendently?

The question whether the antisuicidal effect is shared byother psychotropic agents was addressed in the GermanMAP study, a prospective, randomized, controlled trialwith a treatment duration of 2.5 years. 146 bipolar andschizoaffective patients were randomized to lithium, 139to carbamazepine. No suicidal act was observed in thelithium group, 4 suicides and 5 suicide attempts occurredin the carbamazepine group – a statistically significantdifference (Thies-Flechtner et al. 1996; Ghaemi andGoodwin 1999).

Another argument could be taken from the study byModestin and Schwarzenbach (1992) who were runninga follow-up of 64 former psychiatric inpatients who hadcommitted suicide within one year after discharge andcompared them with a carefully matched control groupof patients who had not committed suicide. A signifi-cantly higher proportion of the controls had been receiv-ing various kinds of psychopharmacotherapy includingseven patients who had been treated with lithium. How-

ever, none of the 64 patients who committed suicide hadbeen receiving lithium at the time of his or her death – theonly statistically significant difference between the twogroups.

What could be the mechanism of this suicide-preven-tive effect? Our hypothesis from the very beginning hasalways been that lithium differs from other mood stabi-lizers and also from most antidepressants by its verymarked serotonin-agonistic effects which are predomi-nantly related to its pre-synaptic functions. It appears atleast an attractive speculation that this serotoninergicaction, possibly in connection with other effects, isrelated to its very well-established antiaggressive effectsin animals as well as humans, but also to its antisuicidaleffects.

The second and decisive question is, whether the anti-suicidal effect of lithium would also occur in patients notresponding optimally in terms of episode prevention.

This question again implies two different issues: a) Are there neurobiological concepts and epidemiolog-

ical data to support the idea of suicidal behavior as anindependent nosological entity?

b) Does lithium effectively reduce suicidal behavior alsoin patients who do not benefit from the lithium treat-ment in terms of episode reduction?

In fact, some concepts and findings suggest that suicidalbehavior might be seen as a particular, possibly anger-related form of affective dysregulation, also associatedwith a disturbance of the 5HAT system, and, thus, as anindependent nosological syndrome.

Data from the recent large WHO study shows that theprevalence of suicidal behavior is not fully related to theexistence of ICD psychiatric diagnoses but that it occursfrequently in symptomatic individuals and in subjectswith subthreshold disorders. Many such people, accord-ing to WHO data from Germany, are characterized bysymptoms of overt or suppressed anger which van Praag(1992) considers as one of the core constituents of thestress syndrome, together with anxiety. Van Praag pos-tulated that in certain types of depression – characterizedby a “5HT-deficit” – anxiety and aggression regulation isprimarily disturbed while mood lowering is a derivativesymptom. Consequently, he expected that certain drugssuch as l-tryptophan, the azapirones or lithium mightameliorate anxiety and/or aggression regulation via reg-ulation of the 5HT system to exert in addition an overalltherapeutic effect in depression.

As to the second aspect, namely whether the antisui-cidal effect of lithium would also occur in patients notresponding optimally in terms of episode prevention,part of the IGSLI data does support such a concept.

For this subanalysis we selected only patients with atleast one suicide attempt in the past before onset oflithium medication (n = 176). We divided the sampleinto three subgroups according to their response tolithium long-term treatment in terms of reduction ofdepressive in-patient episodes. Despite the clearly differ-ent overall efficacy of lithium prophylaxis a statistically

II/74

II/75

significant reduction of suicide attempts occurred in allthree groups, even in the poor responders who did notshow a significant decrease of the depressive in-patientepisodes (Table 5). In other words, in 50% of the clear-cut non-responders no other further suicide attempt wasobserved during lithium treatment.

The standardized suicide mortality in the poorresponders was 17.0 compared to an expected figure ofapprox. 100. Certainly, these findings can neither provethe suicide-preventive effect of lithium nor its potentialspecificity. However, the accumulated evidence stronglysupports such a hypothesis.

Which consequences should be drawn from our pres-ent state of knowledge? We feel that, first, the potentialantisuicidal effect should be considered in setting upmodern guidelines for maintenance treatment in affec-tive disorders (Müller-Oerlinghausen et al. 2001). Fur-thermore, doctors should be extremely cautious in dis-continuing lithium because of alleged non-response.Despite insufficient episode reduction the patient mayhave been protected against suicidal impulses. There-fore, it could be wiser to add a second mood stabilizerinstead of switching the patient to a monotherapy withe.g. carbamazepine.

References

Ahrens B, Müller-Oerlinghausen B (2001) Does lithium exert an inde-pendent antisuicidal effect? Pharmacopsychiatry 34: 132–136

Ahrens B et al. (1995) Excess cardiovascular and suicide mortality ofaffective disorders may be reduced by lithium prophylaxis. J AffectDisord 33: 67–75

Barraclough B (1972) Suicide prevention, recurrent affective disorderand lithium. Br J Psychiatry 121: 391–392

Causemann B, Müller-Oerlinghausen B (1988) Does lithium preventsuicides and suicidal attempts? In: Birch NJ (Ed) Lithium: InorganicPharmacology and Psychiatric Use. IRL Press Limited, Oxford, pp23–24

Coppen A et al. (1991) Does lithium reduce the mortality of recurrentmood disorders? J Affect Disord 23: 1–7

Felber W, Kyber A (1994) Suizide and Parasuizide während und außer-halb einer Lithumprophylaxe. In: Mueller-Oerlinghausen B, Berg-hoefer A (Eds) Ziele und Ergebnisse der medikamentösen Prophy-laxe affektiver Psychosen. Thieme, Stuttgart, pp 53–59

Ghaemi SN, Goodwin FK (1999) Use of atypical antipsychotic agents inbipolar and schizoaffective disorders: review of the empirical liter-ature. J Clin Psychopharmacol 19: 354–361

Hanus K, Zapletálek M (1984) Sebrevrazedná aktivita nemocnychaktivnimi poruuchmi v prübehu lithioprophylaxe. CeskoslovenskáPsychiatrie 80: 97–100

Kallner G et al. (2000) Mortality in 497 patients with affective disordersattending a lithium clinic or after having left it. Pharmacopsychia-try 33: 8–12

Lenz G et al. (1994) Mortalität nach Ausscheiden aus der Lithiumam-bulanz. In: Mueller-Oerlinghausen B, Berghoefer A (Eds) Ziele undErgebnisse der medikamentösen Prophylaxe affektiver Psychosen.Thieme, Stuttgart, pp 49–52

Modestin J, Schwarzenbach F (1992) Effects of psychopharmacotherapyon suicide risk in discharged psychiatric inpatients. Acta PsychScand 85: 173–175

Müller-Oerlinghausen B et al. (1992) Suicides and parasuicides in ahigh-risk patient group on and off lithium long-term medication. JAffect Disord 25: 261–270

Müller-Oerlinghausen B et al. (1992) The effect of long-term lithiumtreatment on the mortality of patients with manic-depressive andschizoaffective illness. Acta Psych Scand 86: 218–222

Müller-Oerlinghausen B et al. (2001) Manic-depressive disorder. TheLancet (in press)

Nilsson A (1995) Mortality in recurrent mood disorders during periodson and off lithium: a complete population study in 362 patients.Pharmacopsychiatry 28: 8–13

Schou M (1998) The effect of prophylactic lithium treatment on mor-tality and suicidal behaviour: a review for clinicians. J Affect Disord50: 253–259

Thies-Flechtner K et al. (1996) Effect of prophylactic treatment on sui-cide risk patients with major affective disorders. Pharmacopsychi-atry 29: 103–107

Tondo L et al. (1997) Effect of lithium maintenance on suicidal behav-iour in major mood disorders. In: Stoff JM, Mann JJ (Eds) The Neu-robiology of Suicide: From the Bench to the Clinic. Ann N Y AcadSci 836: 339–351

van Praag HM (1992) Depression, aggression, and anxiety: a biologicalhypothesis about their interrelation. Eur Neuropsychopharmacol 2:393–402

Tab. 5 Specificity of lithium. Reanalysis of IGSLI data from 176 high risk patients(Ahrens and Müller-Oerlinghausen 2001)

No. of suicide No. of suicide attempts per year attempts per year before lithium during lithiumtreatment treatment

Poor response (n=41) 0.33 0.10 p < 0.007Questionable response (n=81) 0.27 0.06 p < 0.000 Excellent response(n=45) 0.26 0.02 p < 0.000