Severe depression is associated with an increased "lifetime risk" of suicide. It is estimated that 40% of those affected attempt suicide and 4% to 10% die by suicide (1). In contrast to the use of lithium (QUILONUM etc.) in bipolar disease (a-t 2004; 35: 2-4) there is so far no convincing evidence from controlled studies or epidemiological surveys that taking antidepressants can in the short or long term prevent suicide (2,3). The correlation of increasingly frequent prescriptions with a fall of the suicide rate in the total population is not sufficient to prove a benefit as numerous other factors including the reductions in carbon monoxide in car exhausts and access to firearms have an influence on the incidence of completed suicide. Possible effects of antidepressants are lost in the overall statistics (2).

Since the early 1990s, there has been increasing evidence from studies in healthy volunteers, cases series and randomised clinical trials that selective serotonin reuptake inhibitors (SSRI) such as paroxetine (SEROXAT etc.) can increase suicidal tendencies (cf. a-t 2003; 34: 70) (4,5). SSRIs double the incidence of anxiety and aggressiveness compared to placebo - characteristics that can promote suicidal tendencies (6). The American licensing authority FDA reacted now after more than ten years. It assessed the data in October 2004 as so worrying that it ordered the strongest warning ("black box") on the use of antidepressants in children and adolescents, for whom the evidence is clearest (7). At the same time, an expert group of the British Committee on Safety of Medicines made a negative assessment of the benefit-harm relation of citalopram (SEPRAM etc.), escitalopram (CIPRALEX), mirtazapine (REMERGIL etc.), paroxetine, sertraline (ZOLOFT, GLADEM) and venlafaxine (TREVILOR) in children and adolescents on the basis of a lack of proof of efficacy and a suspected increase in suicidal tendencies. A positive recommendation was given for fluoxetine (FLUCTIN etc.) on the basis of data from two randomised controlled trials despite a trend toward an increased suicidality compared to placebo (odds ratio [OR] 1.6; 95% confidence interval [CI] 0.9-3.1) (8). In both investigations, however, a benefit could be calculated only after redefining the primary endpoint (cf. a-t 2004; 35: 45-6). According to the current recommendation of the EMEA, SSRIs should no longer be used in children and adolescents except in licensed indications - in Germany, that means only fluvoxamine (FEVARIN etc.) and only in compulsive disorders (9).

The suicide risk for all age groups is estimated in three recent studies - two meta-analyses and one case control study - (10-12). In a meta-analysis of all the published randomised placebo- or drug-controlled studies on SSRIs in various indications, suicide attempts were recorded in only half of the studies (345 of 702). Taking SSRIs increases the risk of a suicide attempt twofold in placebo-controlled studies (OR 2.28; 95% CI 1.14-4.55). However, the suicide rate does not differ between the two groups (OR 0.95; 95% CI 0.24-3.78). Suicide attempts are equally frequent with tricyclic antidepressants and SSRIs (OR 0.88; 95% CI 0.54-1.42) (10).

Another systematic review analyses the results of 477 published and unpublished randomised controlled studies that were submitted to the British licensing authority by pharmaceutical companies for drawing up safety reports (8). The authors did not find a statistically significant increase either for suicide (OR 0.85; 95% CI 0.2-3.4) or for self-injury (OR 1.57; 95% CI 0.99-2.55; no data on paroxetine) or suicidal thoughts (OR 0.77; 95% CI 0.37-1.55; no data on paroxetine), but the confidence intervals are wide. Therefore, even a markedly increased risk cannot be excluded (11). In both analyses it is suspected that there is a considerable number of unreported cases because of underreporting of events, so that the data are of doubtful validity. Even few unreported events can shift the results markedly. In addition, only published data are considered in the first meta-analysis. However, negative studies often remain unpublished, which was repeatedly demonstrated particularly for SSRIs (a-t 2003; 34; 62-3) (13,14).

The currently available incomplete data from randomised studies and the meta-analyses based on these lack validity for other reasons: suicidal thoughts and attempts are partially misclassified (e.g. events during "wash-out periods" assigned to placebo groups (5)) or miscoded (e.g. "emotional lability" instead of suicidality; cf. a-t 2005; 36: 1-2). Extensive selection of the study patients - exclusion of illegal drug use and acute suicidality or restriction to patients known to respond to SSRIs and tolerate them - prevents the study results from being applicable to practice and euphemising the extent of adverse effects. Furthermore, the studies are not designed to measure suicide events, they record potential suicidality inadequately using only one item in a depression scale and are too short to enable the risk of long-term use to be assessed realistically. The oblique presentation of data by the manufacturers are worrying and nurture distrust even in the licensing authorities. Thus Lilly apparently deliberately concealed safety data (a-t 2005; 36: 1-2). Researchers who have addressed the risk potential of SSRIs critically have been placed under pressure personally or professionally (15,16).

According to epidemiological data, the risks of tricyclic antidepressants and SSRIs are possibly similar with regard to suicidality; in a case control study of 146,000 patients from Great Britain, who received an antidepressant for the first time because of depression, the rate of suicide and self-injury was correlated with the prescription of SSRIs or tricyclic antidepressants. The authors did not calculate any difference for use of SSRIs compared to tricyclic antidepressants for completed suicide (OR 0.57; 95% CI 0.26-1.25) or self-injury (OR 0.99; 95% CI 0.86-1.14). However, the subgroup analysis for children and adolescents from 10 to 18 years showed increased autoaggressive behaviour with SSRIs (OR 1.59; 95% CI 1.01-2.50). A comparison with non-use or placebo is lacking. It thus remains uncertain how both antidepressant classes influence the suicide risk (12). Epidemiological studies offer the advantage of a broader spectrum of patients, but involve the risk of bias by partially unknown influences. The results must therefore be interpreted with caution.

Besides the discussion of the suicidality induced by antidepressants, the extent of the efficacy of these agents is also increasingly questional; according to an analysis of 47 studies, which were submitted to the FDA between 1987 and 1999 for the licensing of citalopram, fluoxetine, nefazodone (NEFADAR, off market), paroxetine, sertraline and venlafaxine, the placebo effect accounts for 68% to 89% of the drug effect. The rate of clinical success is probably even less as nine studies which showed no significant difference were not considered in the calculation. When the analysis is limited to three antidepressants with complete data, the result is an improvement of only 2 points (out of a total of 50 or 62 points respectively) in the HAMILTON depression scale (HAMD*), which is used in many studies, compared to placebo (17). Even the small benefit compared to placebo might be simulated by "unblinding" the patients through typical side effects. Doubts about a relevant antidepressant efficacy also apply to tricyclic antidepressants (18). Thus, a systematic analysis of antidepressant studies with "active" placebo (atropine to produce anticholinergic side effects without antidepressant effect) showed only minimal superiority for the drug (19).

Proof of a dose-effect relation is lacking so far. No data support increase in dosage, if the effect is inadequate. The very low benefit of the antidepressants is known to the manufacturers; apparently eight placebo-controlled studies are planned to obtain the two positive studies, which are required for licensing in US (20). More than 50% of the placebo-controlled studies of antidepressants have a negative result, but they do not count for the authorities when deciding on licensing (20). For reboxetine (EDRONAX, SOLVEX), which is licensed in Germany (but not in the USA), only one of the eight studies submitted was positive (21).

	Evidence of a preventive effect on suicide is lacking for all antidepressants.
	Summaries of data from randomised controlled studies indicate an increase in suicidal tendencies through the use of serotonin reuptake inhibitors such as paroxetine (SEROXAT etc.). Tricyclic antidepressants may involve a similar risk.
	The risk cannot be determined reliably with the available studies because of the considerable method-related problems. Prospective long-term studies with adequate recording of safety data are therefore urgently required.
	Measured by the usually employed HAMILTON depression scale the benefit of antidepressants is only slightly above that of placebo. Even this slight effect can be simulated by typical interference effects because of unblinding.
	Despite decades of use, neither the efficacy nor the safety of the available antidepressants has been adequately proven.

(M = meta-analysis)

1

FINLEY, P.R.: Ann. Pharmacother. 2004; 38: 1739-42

2

GUNNELL, D., ASHBY, D.: BMJ 2004; 329: 34-8

3

KHAN, A. et al.: Arch. Gen. Psychiatry 2000; 57: 311-7

4

BREGGIN, P.R.: Int. J. Risk Safety Med. 2004; 16: 31-49;
im Internet unter: http://www.breggin.com

5

HEALY, D.: Psychother. Psychosom. 2003; 72: 71-9

6

BRENT, D.A.: N. Engl. J. Med. 2004; 351: 1598-601

7

FDA Public Health Advisory of 15. Oct. 2004:
http://www.fda.gov/cder/drug/antidepressants/SSRIPHA200410.htm

8

CSM expert working group:
http://medicines.mhra.gov.uk/ourwork/monitorsafequalmed/safetymessages/SSRIfinal.pdf

9

EMEA: Press declaration of 25. Apr. 2005

10

FERGUSSON, D. et al.: BMJ 2005; 330: 396ff

11

GUNNELL, D. et al.: BMJ 2003; 330: 385ff

12

MARTINEZ, C. et al.: BMJ 2005; 330: 389ff

13

MELANDER, H. et al.: BMJ 2003; 326: 1171-3

14

WHITTINGTON, C.J. et al.: Lancet 2004; 363: 1341-5

15

ANTONUCCIO, D.O.: Prevention and Treatment 2002; 5: Article 25; can be found at
http://journals.apa.org/prevention/volume5/pre0050025c.html

16

HEALY, D.: Perspect. Biol. Med. 2002; 45: 250-63

17

KIRSCH, J. et al.: Prevention and Treatment 2002; 5:
http://journals.apa.org/prevention/volume5/pre0050023a.html

18

MONCRIEFF, J.: BMJ 2005; 330: 420

19

MONCRIEFF, J. et al.: Active placebos versus antidepressants for depression. The Cochrane Database of Systematic Reviews 2004, Issue 1

20

THASE, M.E.: Efficacy of Newer Antidepressants; on the Internet at:
http://www.medscape.com/viewprogram/2546_index

21

MEDAWAR, C. et al.: Lancet 2004; 363: 2087

The HAMILTON depression rating scale (HAMD) is a questionnaire often used in clinical studies containing 17 items (50 points) or 21 items (62 points) regarding depressive symptoms.