Anti-depressants and suicide

Abstract

Suicide takes the lives of around a million people each year, most of whom suffer from depression. In recent years there has been growing controversy about whether one of the best-selling anti-depressants – selective serotonin reuptake inhibitors (SSRIs) – increases or decreases the risk of completed suicide. Randomized clinical trials are not informative in this application because of small samples and other problems. We present what we believe are the most scientifically credible estimates to date on how SSRI sales affect suicide mortality using data from 26 countries for up to 25 years. We exploit just the variation in SSRI sales that can be explained by institutional differences in how drugs are regulated, priced, and distributed, as reflected by the sales growth of new drugs more generally. We find an increase in SSRI sales of 1 pill per capita (12% of 2000 sales levels) reduces suicide by 5%.

Introduction

Suicide claims the lives of about a million people around the world each year (Goldsmith et al., 2002) and currently ranks 11th among leading causes of death in the United States (National Vital Statistics Reports, 2007), yet economists have devoted surprisingly little attention to the topic. There is a very small theoretical literature that seeks to understand the nature of suicidal behavior (e.g., Hamermesh and Soss, 1974; Cutler et al., 2001; Becker and Posner, 2004). Even less attention has been devoted to the problem of suicide prevention, which has the potential to improve social welfare either by changing suicidal people's desire for self harm, or by preventing them from acting on their desires.²

This paper examines the effects on suicide from one of the most important, but increasingly controversial, tools for preventing suicide—modern anti-depressant drug treatment. Specifically, we provide what we believe to be the most scientifically credible estimate to date for the causal effects on suicide mortality from selective serotonin reuptake inhibitors (SSRIs). The SSRIs were introduced in the 1980s, and by 2000 were the most commonly prescribed drug class in the U.S. and the third best-selling drug class in the world (IMS Health, 2006). Yet SSRIs have been the subject of recent government safety warnings in the U.S. and U.K., which have led to large, widespread reductions in their use (Gibbons et al., 2007a) as well as sharp divergence in professional opinion about the safety of SSRIs. One researcher involved in the FDA's reviews of SSRIs told the New York Times, “Sitting up there and having the public yell that you’re killing their children is no fun.” A medical historian told the Times “It's like a religious war,” with a level of argument not seen since “the 1960s and 1970s, when scientists were challenging psychoanalysis” (Carey, 2006).

The expected net effect of the introduction and growing use of SSRIs on suicide mortality is ambiguous, a priori. Anti-depressants may help people persevere through difficult but transitory periods of their lives. On the other hand, most anti-depressants appear to improve patient energy levels before they improve mood, which may contribute to an increase in the risk of suicide during the early stages of treatment (FDA, 2006). There could also be what Viscusi, 1984, Viscusi, 1985 terms a “lulling effect,” for example if medical practitioners react to the improved safety and reduced side effects of SSRIs relative to the older tri-cyclic anti-depressants (TCAs) by reducing the vigilance of supervision of potentially suicidal patients.

Concern about the safety of SSRIs has been motivated by several meta-analyses of randomized clinical trials (RCTs), which suggest that SSRI treatment elevates the risk for suicidal thoughts and non-lethal self-injurious behavior among pediatric patients and perhaps even adults as well. But the sample sizes used in these RCTS are too small to detect policy-relevant effects on suicide mortality. As a result, most RCTs rely on measures of non-lethal suicidality, which for reasons discussed further below are not very informative about drug impacts on mortality. Several non-experimental studies have compared trends in suicide and SSRI use across jurisdictions over time, a design that provides better statistical power than RCTs to detect impacts on suicide mortality, but raises concerns about the endogeneity of SSRI utilization rates. The magnitude and even the sign of the bias that may result are difficult to predict. Adverse changes in mental health may increase both SSRI sales and suicide, which would lead standard panel data analyses to understate any protective effects of SSRIs on suicide. But countries might also encourage SSRI use as part of a broader effort to improve mental health. Given the limitations of the existing empirical evidence, there remains great uncertainty about the public health effects of one of the world's most widely used pharmaceutical products.

In this paper we present what we believe to be the first estimates for the effects of SSRIs on suicide using both a plausibly exogenous source of identifying variation and adequate statistical power to detect effects on mortality that are much smaller than anything that could be detected from randomized trials. We construct a panel dataset with suicide rates and SSRI sales per capita for 26 countries for up to 25 years. Since SSRI sales may be endogenous, we exploit institutional differences across countries that affect how they regulate, price, distribute and use prescription drugs in general (Berndt et al., 2007). Since we do not have direct measures for these institutional characteristics for all countries, we use data on drug diffusion rates as a proxy. We show that sales growth for SSRIs is strongly related to the rate of sales growth of the other major new drugs that were introduced in the 1980s for the treatment of non-psychiatric health conditions. This source of variation in SSRI sales helps overcome the problem of reverse causation and many of the most obvious omitted-variables concerns with past studies. Our research design may also have broader applications for the study of how other drug classes affect different health outcomes.

Our instrumental variables (IV) estimates suggest that an increase in SSRI sales of 1 pill per capita (around a 12% increase over 2000 sales levels) would reduce suicide mortality rates by around 5%. This relationship holds up even after conditioning on country and year fixed effects, country-specific linear trends, and many of the risk factors that previous research identifies for suicide (Goldsmith et al., 2002). Our estimates imply around 1 suicide is averted for every 200,000 pills sold.³ Our IV estimates are about twice as large as those from OLS, which is important because the magnitudes of impacts – not just their signs – matter for benefit–cost or cost-effectiveness analyses of health interventions. Commonly used SSRIs can be obtained in the U.S. for around $0.11 per pill,⁴ which suggests a cost per statistical life saved from increasing SSRI use of around $22,000—far below most other government regulations or policies.

One drawback of country-level data is that SSRI sales information is not available for different population sub-groups, such as by age or gender. This limits our ability to identify heterogeneity in treatment effects, which could in principle be valuable for helping health policymakers target SSRI use within the population. However, the degree to which regulators can in fact influence SSRI use in targeted ways remains unclear. For example, SSRIs were widely used for “off-label” treatment of depression among adolescents and children prior to FDA approval (Olfsen et al., 2002a, Olfsen et al., 2002b, and Zito et al., 2003).⁵ After the FDA issued a warning in 2004 about SSRI use in pediatric patients, SSRI sales declined among almost all adult age groups as well (Gibbons et al., 2007a). Given this broad-based response to age-targeted warnings, the question we address here seems relevant to a broad range of policy decisions.

The most important concern with our estimates comes from the fact that our instruments are not randomly assigned across countries, and so there is necessarily the question of whether they are orthogonal to other determinants of suicide. A variety of specification tests provide some support for our research design.

One specific concern is that prescription drugs may diffuse more rapidly in higher-income countries (Slade and Anderson, 2001). However, we show that our results are not affected when we control for economic conditions. A related concern is that new drugs may diffuse more quickly in countries that are more intensive users of medical care overall, so that our IV estimates might be picking up effects that more or better health care may have in reducing the prevalence of (or pain associated with) chronic health problems that lead some people to contemplate suicide. But we find that the rate at which new drugs diffuse is unrelated to the trend in health spending in our sample. Countries that have made similar policy decisions about how to operate their general health care systems have made different choices about how to regulate, price or distribute prescription drugs. We also show there is no estimated SSRI “effect” on accident mortality, and that predicted SSRI sales are most strongly associated with declines in suicide mortality among teenagers and young adults, rather than among older age groups for which chronic health problems or pain are most common. As a more general specification test, we compare countries that our design predicts to have faster versus slower rates of growth in SSRI sales, and show they have similar “pre-treatment” trends in suicide mortality in the period before SSRIs are introduced. One might still worry that the use of other mental health treatments may have increased over the 1990s during the time when SSRI sales were increasing rapidly, but among the set of countries for which we can obtain data we find no increases in psychotherapy or use of older TCA anti-depressants during the 1990s.

The next section discusses the pathways through which anti-depressant drugs could affect suicide, while Section 3 reviews available evidence on this question. We discuss our data in Section 4 and empirical methods in Section 5. The main findings are in Section 6, while implications are discussed in Section 7.