Suicide Risk and the SSRIs

The selective serotonin reuptake inhibitors (SSRIs) were a welcome innovation in the treatment of depression. Although these drugs are no more efficacious than their predecessors, they are safer in overdose and more acceptable in terms of adverse effects.¹ The popularity of the SSRIs in general, and of fluoxetine (Prozac) in particular, created its own problems. For instance, Prozac achieved a recognition normally given to household items in which the brand name has become identified with the generic product.

Selective serotonin reuptake inhibitors can cause agitation and activation, particularly at the start of treatment (eg, such sensations are not pleasant and resemble the akathisia associated with the neuroleptics^{2 - 3} ). In the setting of lowered mood, developing new aversive symptoms might further depress mood and increase the risk of suicide. Also, paradoxical reactions (ie, very occasionally causing agitation) are potential adverse effects of the benzodiazepines, and it is possible that something similar might be observed with SSRIs

The current debate about the SSRIs and either self or externally directed aggression initially concerned adult patients, but recently the controversy has extended to children receiving the same drugs, with similar claims being made concerning both increased suicidality and psychological dependence resulting from SSRI use.⁴ These concerns have found a receptive media audience. For instance, 2 investigative programs by the British Broadcasting Corp Panorama team generated the largest television audience response they have ever known, receiving several hundred e-mails from the public, nearly all describing adverse effects due to the SSRIs.^{5 - 6}

Unfortunately, evidence on the role of SSRIs in treating adolescent depression, which is indisputably an important problem, is far from perfect. Selective serotonin reuptake inhibitors are certainly frequently prescribed in this age group, few formal studies have been performed, and not all of them are published.⁷ Only in the last few weeks has one company, GlaxoSmithKline, agreed to publish all the data it has on paroxetine, possibly stimulated by the lawsuit filed by New York State's attorney general.⁸ Sensitivities about research involving children mean that both academics and industry are increasingly wary about carrying out randomized trials in this age group. The paradox, however, is that it is in these vulnerable populations that high-quality evidence is most needed, yet most difficult to obtain. Relying on evidence from studies of adult populations is far from ideal.

The regulatory authorities are faced with a dilemma. The distant specter of thalidomide and the enormous impact of other perceived failures of regulation, such as the bovine spongiform encephalopathy crisis in the United Kingdom, linked to the general precautionary climate has increased the pressures on many regulatory institutions. Fear of being blamed if subsequent evidence of harm emerges means that the regulatory bodies have to operate in an increasingly nervous environment. Hence both the US Food and Drug Administration and the Medicines and Healthcare Products Regulatory Agency (MHRA) in the United Kingdom have recently issued warnings concerning the use of the SSRIs in children. This is not a "ban" as some media have reported, since SSRI use in children has always been off label, except for fluoxetine. Perusal of any of the summaries of product characteristics for the SSRIs confirms that use in children is not recommended by the manufacturers (again with the exception of fluoxetine)^{9 - 10} but remains widespread in those younger than 18 years, presumably because of the continuing need to do something to help the serious problem of depression in adolescents. The only drug for which the UK MHRA Committee on Safety of Medicines concluded that there is adequate evidence to support use in those younger than 18 years is fluoxetine.⁹

Although it is reasonable to ask questions about the SSRIs and suicide, it is more difficult to answer them. The problem is that depression is unequivocally and substantially associated with suicide and deliberate self-harm. Depression is also unequivocally and substantially associated with the prescription of antidepressants. Thus looking for an association between antidepressants and suicide and/or deliberate self-harm is going to be difficult. However, the UK General Practice Research Database (GPRD) provides perhaps the best setting to explore the issue. It is the world's largest computerized database of patient records from primary care, with more than 35 million patient-years of information. Because all prescriptions are recorded electronically, it is particularly useful for pharmacoepidemiological studies and has been used extensively for this purpose, most often by Jick and colleagues¹¹ from the Boston Collaborative Drug Surveillance Program, whose study on antidepressant prescribing and attempted and completed suicide is reported in this issue of JAMA.

In epidemiology size does matter, and there is no faulting the UK GPRD on that basis. But size alone is no guarantee of accuracy, unless the data are untainted by bias. This may affect either the recording of exposure or of outcome, and in either case size will not overcome the problem. In the study by Jick et al, recording of exposure, ie, antidepressant prescription, is likely to be complete and bias free because the UK GPRD exists largely for that purpose. Recording of the outcomes may be more problematic. Death certification of suicide is imperfect in all countries, and recording of deliberate self-harm is very variable. Some probably go unnoticed. The majority are seen acutely in the hospital emergency department setting. In theory, attendances should be communicated to the general practitioner and hence find their way onto the UK GPRD. In practice many episodes go unrecorded. This will however only affect this analysis if misclassification is associated with prescribing of one particular class of antidepressant, and this seems unlikely. Self-harm itself is a broad term—in the study reported in this issue the authors have chosen to restrict the analysis to the UK GPRD codes for "attempted suicide" and a variety of codes used for unintentional drug overdose. This means that they have excluded all the codes for deliberate cutting or self-mutilation, behaviors that are certainly self-harming but in general not considered to be motivated by suicidal thoughts.

The authors have chosen the strategy of comparing 3 designated antidepressants: namely amitriptyline, fluoxetine, or paroxetine, with a reference antidepressant, dothiepin. The 2 SSRIs are those about which there have been most concerns. Amitriptyline is the reference tricyclic antidepressant. The choice of dothiepin needs some justification. This drug has never been licensed in the United States and is also one of the most dangerous antidepressants in overdose.^{12 - 13} Nevertheless, it was the most frequently prescribed antidepressant in the United Kingdom during the period of analysis.

In any naturalistic study like this, the challenge is comparing like with like. The danger comes from confounding, for which the only certain remedy is randomization. Confounding might arise, for example, if treatment was preferentially initiated with the tricyclics, and then switched to the SSRIs if first-line treatment was unsuccessful. Selective serotonin reuptake inhibitors would then be associated with more treatment-resistant depression and a higher risk of suicide.

The authors have gone to considerable trouble to reduce this risk. Their strategy was to treat confounding by restriction—analyzing only those patients receiving their first-ever antidepressant prescription. The overall analytic strategy is to compare cases of deliberate self-harm, and then suicide, in a case-control design in which exposure is receipt of any of the 4 antidepressants.

The results confirm that antidepressant prescription is indeed associated with suicidal behavior, and strongly so. This simply means that antidepressants are being prescribed for the right indication, and that they do not immediately eliminate suicide risk. That we knew. The article by Jick et al also shows a 4-fold increased risk for suicidal behavior for patients first prescribed an antidepressant drug within 1 to 9 days before their index date. This too is not unexpected—patients see their general practitioner when their distress is most intense. But the hypothesis being tested is that over and above the known association of antidepressant prescribing and suicidal behavior (in which the confounder is the presence of depressive disorder), there is also a specific link in which one class of antidepressants, the SSRIs, increases that risk still further.

The results do not offer much support for the hypothesis. The key results are that the adjusted relative risk estimates comparing amitriptyline, fluoxetine, and paroxetine each to dothiepin as the reference group were 0.83, 1.16, and 1.29, respectively. None of these were statistically significant although the risk estimates for paroxetine approached significance. However, the authors appropriately caution against overinterpreting this borderline result. Most UK general practitioners are now aware that the older tricyclic drugs are more dangerous in overdose, and it remains plausible that there is a tendency to prescribe the newer SSRIs for patients about whom the physician has more concerns about suicidal risk. Only a small such bias could cause the observed results. Moreover, there was no evidence for the alleged withdrawal phenomenon, which is another of the concerns that have been raised about the SSRIs.¹⁴ Stopping medication did not lead to an increase in risk, as postulated by some.

The findings from Jick et al¹¹ provide useful data in what is still a somewhat messy situation. Public anxiety fueled by media reports has transferred itself to the already nervous regulatory authorities, and it is unlikely that this study alone will restore confidence. Instead, the general crisis of confidence in the regulatory bodies in both the United States and United Kingdom is likely to accelerate as society becomes increasingly obsessed with questions of risk and then blame. Yet, irrespective of the decisions of regulatory bodies, general practitioners and psychiatrists will continue to see depressed patients in general, and adolescents in particular, at risk of suicide and will confront difficult questions of clinical management. Fortunately, suicide is uncommon in adolescents , and no suicides were noted in this cohort. The UK GPRD database documents 15 adolescent suicides during the period 1993 to 1999. None of these patients received an antidepressant drug. It is unknown whether their outcome would have been different if they had been prescribed antidepressants. However, it is probable that clinicians who do wish to prescribe antidepressants to depressed adolescents may be increasingly likely to return to the older tricyclics rather than to the SSRIs because of the recent warnings. The irony is that the tricyclics seem to be ineffective in this age group and certainly are more dangerous in overdose.¹⁵ Whatever decision clinicians reach, careful monitoring of adolescents (for activation, agitation, and suicidal ideation) prescribed any antidepressant remains essential.