Standard Antidepressant Pharmacotherapy for the Acute Treatment of Mood Disorders

Michael J. Burke, M.D., Ph.D. and Sheldon H. Preskorn, M.D.

This chapter surveys and synthesizes our current knowledge of the standard pharmacological treatment for the acute treatment of depression and points out areas requiring further research. The information presented is based on a literature review of clinical trials data, postmarketing studies, and meta-analyses since the last volume of this text. We have restricted our scope to antidepressant drugs marketed or soon to be marketed in the United States.

The treatment of mood disorders has been and continues to be in a state of dynamic evolution. In the last volume of this text, Psychopharmacology: The Third Generation of Progress (1987), tricyclic antidepressants (TCAs) and irreversible monoamine oxidase inhibitors (MAOIs) were considered "the standard" antidepressant pharmacotherapy, and all other antidepressant drugs were designated as "new". Since that time, multiple antidepressant drugs have been developed and approved in the United States at a rate that is unparalleled in the history of antidepressant pharmacotherapy, with eight new antidepressant agents becoming available for clinical use in just the last eight years. As rapidly as these new drugs have been marketed, they have enjoyed widespread use and quickly become part of the standard pharmacopoeia for the treatment of depression.

The goal of this chapter is to conceptualize the initial phase of antidepressant pharmacotherapy in a way that will: 1) provide a framework to assist the clinician in the organization, differentiation, and use of the currently available antidepressant medications and new agents as they become available; and 2) generate issues and questions germane to the research-oriented psychiatrist. Our focus will be issues relevant to the initiation phase of antidepressant pharmacotherapy and the successful induction of symptom remission. Because depressive illness is characterized by a high risk of relapse and recurrence, important treatment issues remain even after successful short term therapy. These issues relevant to the continuation and maintenance phases of antidepressant pharmacotherapy will be covered separately in a subsequent chapter (Long-Term Treatment of Mood Disorders) by Prien and Kocsis.

Perhaps the most popular class of antidepressant drugs worldwide is the selective serotonin reuptake inhibitors (SSRIs). As such, Selective Serotonin Reuptake Inhibitors in the Acute Treatment of Depression by Montgomery is devoted exclusively to this antidepressant class; only general principles regarding the SSRIs and specific issues related to the early phase of antidepressant treatment will be referred to in this chapter. At present, data suggest possible antidepressant efficacy of some benzodiazepines (i.e., alprazolam), psychostimulants, sex steroids (i.e., estrogens), and azapirones (i.e., buspirone). However, drugs from these classes cannot be considered standard antidepressants at this time, and hence they are not included in this chapter. A number of antidepressant agents, some with novel mechanisms of action, that are not marketed in the United States or other English-speaking countries. These agents will not be addressed per se. The interested reader is referred to a recent report (52).

It is relevant for both the research psychiatrist and clinician to understand the basis for the recent expansion in the number and types of new antidepressant medications. Antidepressant pharmacotherapy has evolved over the past three decades, with a trend toward increasing pharmacodynamic selectivity (Fig. 1). Over the last decade, drug discovery and development in psychiatry has gone from a process based almost exclusively on serendipity to a process of rational drug design by molecular targeting. Antidepressant pharmacotherapy is the first area in psychopharmacology to have benefited significantly from such targeted development.

The history of antidepressant pharmacotherapy begins with the tricyclic antidepressants (TCAs) and monoamine oxidase inhibitors (MAOIs). Both of these classes of antidepressant agents were successful, were characterized by having diffuse effects on the central nervous system (CNS), and were essentially discovered by chance. TCAs were originally developed to be better neuroleptics. The antidepressant properties of MAOIs were found while searching for antitubercular drugs.

Such chance discoveries proved to be of seminal importance. First, they demonstrated that major depression was amenable to medical intervention, just like other medical conditions such as hypertension and diabetes. Second, they served as road maps to improve our understanding of the mechanisms of action mediating both their desired antidepressant effects and their undesired effects. The latter was critical to the era of rational drug development that followed.

The problem with drugs discovered by chance is that they frequently have multiple mechanisms of action, as in the case of TCAs, or they have such a basic mechanism of action that they impact a wide variety of systems, such as MAOIs which affect serotonin, norepinephrine, and dopamine metabolism. The reason is that chance discovery is usually dependent on the drug having a large signal-to-noise ratio. The signal, or effect, must be larger than the background or random noise in order to be recognized. Unfortunately, whereas a drug with multiple mechanisms of action may produce a large "signal" (i.e., have a broad clinical activity in a heterogenous population), a complex pharmacodynamic profile also means that the drug will produce a number of undesired as well as desired effects and have a narrower therapeutic index than a drug rationally developed to affect only the mechanism(s) of action mediating the desired effect.

This issue can be better understood by examining the pharmacology of TCAs, which served as the cornerstone of antidepressant pharmacotherapy for almost 30 years. TCAs have multiple mechanisms of action that occur over a relatively narrow concentration range. Essentially, the TCAs exhibit multiple biological activities with roughly the same potency, such that patients taking these medications are likely to experience multiple effects in addition to those attributed to the antidepressant mechanisms of action. Figure 2 illustrates the pharmacodynamic profiles for a number of antidepressant agents and the relative potency of the various mechanisms of action exhibited by these agents.

Some of the mechanisms of action of TCAs, such as the inhibition of sodium conductance through excitable membranes (gNa+), cause potentially serious effects on cardiac conduction and occur at a concentration less than one order of magnitude above that needed to inhibit the neuronal uptake of norepinephrine (NE) and serotonin (SE), the mechanisms of action believed to mediate their antidepressant effects (72). This is the reason that an overdose of these drugs of only 10 times the therapeutic dose can have serious cardiotoxic effects (8).

One of the obvious goals of rational drug development has been to produce agents with improved tolerability and safety, and this has pointed the direction to development of increasingly selective agents. A potential downside to this strategy, given our limited understanding of the pathophysiology underlying major depression, is that a drug or class of drugs having a highly specific and hence limited range of actions might have a narrower spectrum of clinical activity in a syndromic illness like major depression. The history of medicine repeatedly shows that syndromic illnesses are usually composed of groups of heterogeneously distinct conditions with regard to their underlying pathophysiology and pathogenesis.

The dilemma then, from a drug development standpoint, is to develop new medications with the optimal balance between the number of mechanisms of action needed for the widest spectrum of antidepressant activity and, at the same time, maximize safety and tolerability. Narrowing the pharmacologic profile of new antidepressant agents offers potential advantages in terms of safety and tolerability. However, when coupled with the heterogenous nature of syndromal major depression, this may likely reduce our ability to detect efficacy (e.g., reduced signal to noise ratio) and lead to false negative results. With the current goal of developing antidepressant agents with increasingly precise mechanisms of action, an important question is at what point will efficacy results shift from representing the agents being studied to primarily reflecting the heterogenous population of depressed patients? How to modify the design of clinical trials to avoid this situation is a critical research issue overlapping the need to refine diagnostic practices and advance understanding of pathophysiology.

Of note with regard to the trend toward developing increasingly selective antidepressant agents, the most recent additions to the antidepressant armamentarium are agents that combine more than one biological activity presumed to mediate their antidepressant effect. Just a few years ago, one prominent direction for antidepressant development was creation of specific biogenic amine receptor subtype agonists and antagonists. For these agents, the problems of demonstrating efficacy and identifying the subpopulation of patients likely to have a positive response was and is germane. However, the recently marketed antidepressant agents appear to have achieved a broad spectrum of clinical efficacy by combining multiples of these desired mechanisms of action. In the case of nefazodone, it is first a serotonin-2-A receptor antagonist but also acts to inhibit serotonin reuptake. In the case of the newest antidepressant, mirtazapine, the agent is a selective antagonist for several serotonin and one norepinephrine receptor subtype (see later discussion).

Historically, the classification of antidepressant agents was quite straightforward (i.e., two classes—the TCAs and the MAOIs). In actuality, this system or lack of a system, combined the use of structure (e.g., TCA) and function (e.g., MAOI). Over time, minor structural modifications of the TCAs were made to produce structural analogs with the goal of improving the adverse effect burden. This expanded the use of the classification system based on structure (i.e., tertiary versus secondary amine TCAs). As new antidepressants have been made available the structural basis for classifying these drugs has increasingly lacked utility and has been complicated by incorrect usage of terms such as heterocyclic, tricyclic, and tetracyclic. With the rapid expansion of the antidepressant armamentarium, an acceptable classification system is critical to assist the practicing physician with what might be the confusing process of selecting agents from among the many treatment options.

A classification system based on the mechanism of action presumed to be responsible for the antidepressant effects eliminates the confusion around significance of structure. A pharmacodynamic system of classification has heuristic advantages, in that it incorporates the current theories of disease pathophysiology, can easily accommodate new agents as they become available, and provides a rational basis for sequential treatment selection in clinical practice. Although the connection between mechanisms of action and antidepressant response is hypothetical, this type of classification is based on the established pharmacology of the drugs. Such pharmacology clearly mediates some of the actions of the drugs, even if it is eventually proven that these actions are not responsible for the antidepressant effects. A classification based on presumed mechanism(s) of action also is more consistent with how antidepressant medications are being developed.

Using a functional classification system exclusively (i.e., eliminating reference to molecular structure), the currently available antidepressant agents can be organized into eight classes, based on their known pharmacology and putative mechanism of action (Table 1). This approach has the advantage of "splitting" antidepressant drugs into categories that emphasize their differences. For many clinicians, this approach may be most useful for treatment selection, particularly in those cases when a patient has failed to respond to an initial antidepressant trial, and the plan is to select an alternative agent. Despite the wide variety of safe antidepressant choices, in a recent poll of practicing physicians, 50% stated their first choice in a patient who failed treatment with a selective serotonin reuptake inhibitor (SSRI) would be another SSRI. Of those polled, 50% also state that if the patient failed a treatment trial with a second SSRI their next choice would be another SSRI. These results certainly speak to the popularity of the SSRI class of antidepressant agents but also suggest that the diversity of antidepressant agents is not being used to full advantage.

An alternative function-based classification scheme takes advantage of the fact that antidepressants can be conceptualized as falling into three primary categories, defined by the biological mechanism presumed to mediate antidepressant efficacy. The three primary categories include: inhibitors of neuronal uptake of neurotransmitter; inhibitors of the enzymatic metabolism of neurotransmitter; and, biogenic amine receptor antagonists (Table 2). Agents within a class can be further subdivided based on their degree of selectivity (see below). The advantage of this classification approach is that similarities as well as differences among antidepressant agents are highlighted. For the present, either approach to function-based classification provides an organized framework for the clinician and researcher to accommodate new classes of antidepressant agents as they are developed with similar or novel molecular targets not currently represented in the pharmacopoeia (e.g., second messenger systems, ion channels of excitable membranes).

Within the three major categories of antidepressants, specific agents are further classified with regard to their selectivity for a specific neurotransmitter, selectivity for the primary antidepressant mechanism of action over other non-antidepressant biological actions, and the combined presence of more than one putative antidepressant mechanism that is likely to be active at clinically relevant concentrations of the drug (Table 2). Hence, among those antidepressants in the broad class of amine reuptake inhibitors, drugs are further classified by the targeted neurotransmitter (e.g., selective serotonin or norepinephrine reuptake inhibitors). Drugs that potently target more than one neurotransmitter are referred to as "combined" reuptake inhibitors (e.g., combined serotonin-norepinephrine or dopamine-norepinephrine reuptake inhibitors). Reuptake inhibitors with relatively potent biological activity on presumed non-antidepressant mechanisms are classified as non-selective (e.g., TCAs). The same system is also used to subdivide antidepressants within the class designated as receptor antagonists (Table 2).

The clinically relevant pharmacodynamic profile for each antidepressant agent is derived from in vitro determination of the relative potency of a drug for different mechanisms of action and the likelihood that, under routine use, the drug achieves sufficient concentration in vivo to affect a site of action given the in vitro potency. If a drug affects a site of action more potently than its presumptive antidepressant site of action, the non-antidepressant mechanism of action is clearly part of the clinically relevant pharmacodynamic profile. For example, amitriptyline is more potent at blocking histamine and muscarinic acetylcholine receptors than inhibiting neuronal reuptake of norepinephrine (see Fig. 2). Therefore, the antidepressant effect of norepinephrine reuptake inhibition can only occur simultaneously with antihistamine and anticholinergic effects of the drug.

A specific biological activity also becomes part of the clinically relevant pharmacodynamic profile of an antidepressant drug if its potency is within one order of magnitude of affecting the site of action believed to be mediating antidepressant efficacy. The rationale is the pharmacologic rule of thumb that there must be more than one order of magnitude difference in the in-vitro potencies of two biological activities to state confidently that one can achieve concentrations of the drug capable of affecting one site without affecting the other. With this rule, each antidepressant is used as its own reference point to determine which biological mechanisms are likely to be active under clinically relevant dosing conditions.

The clinically relevant pharmacodynamic profile for each antidepressant agent can be summarized using a binary system which incorporates the preclinical in vitro data on potency combined with the in vivo data on concentration of drug achieved under conditions of routine practice (Table 3). Characterizing antidepressants in this way has heuristic value, in that it anticipates the clinical pharmacology of these drugs and provides direction for clinical use.

At one level, the approach to pharmacotherapy for depression has remained unchanged over the last several decades; after diagnosis the clinician initiates his choice of antidepressant agent and then titrates the drug dose over time until symptoms resolve. However, with expansion of the antidepressant armamentarium and a refined understanding of the clinical pharmacology of antidepressant drugs, the conceptualization of short-term treatment for depression has changed dramatically in recent years. The days of the "French cooking" approach to antidepressant pharmacotherapy are giving way to an era of rational drug therapy.

Rational antidepressant pharmacotherapy utilizes the pharmacologic diversity of medication options to advantage in treatment selection by tailoring treatment to the individual patient based on the specific patient profile and the known pharmacology of the antidepressants. Patient features such as age, comorbid physical illness, and concomitant medications can be used to anticipate vulnerability to particular mechanisms of action exhibited by some but not all antidepressants and the likelihood of significant drug interactions.

Attention to the pharmacodynamic and pharmacokinetic profile of antidepressants, of use in the initial treatment selection, also provides a rational basis for sequential use of antidepressant agents when the initial therapy is unsuccessful. Hence, if a patient fails an antidepressant that targets one neurotransmitter, the clinician can next select an agent that: affects the same neurotransmitter via a different mechanism of action; or, targets a different neurotransmitter; or, has combined effects on more than one neurotransmitter. This approach to sequential drug use is supported by considerable data demonstrating that failure to respond to a drug from one antidepressant class does not predict a failed response to a drug from another class (38, 68).

Antidepressant treatment has been conceptualized as occurring in three phases (45). The initiation of antidepressant pharmacotherapy and successful remission induction is referred to as the "acute phase" of treatment. This is followed by the "continuation phase" of treatment, with the goal of maintaining symptom remission from the depressive episode and preventing relapse. In those cases of recurrent depressive disorders, a "maintenance phase" of treatment is instituted to maintain euthymia and prevent a new episode of depression from occurring (i.e., relapse).

The duration of the acute phase of antidepressant treatment may be 6–12 weeks or longer, depending on the individual patient response and the number of medication changes necessary to achieve the desired clinical outcome. In the last decade, considerable attention has been focused on the necessary duration of a "therapeutic" treatment trial, and a general consensus has been that a minimum of 3–6 weeks of continuous therapy at optimal dose is necessary to determine that a patient is refractory to a particular antidepressant agent (20, 75). Data suggest that some patients who fail to respond to an antidepressant agent during this period will go on to become a treatment responder if the treatment trial duration is increased (21). However, in a recent analysis of 590 cases, patients who failed to show at least minimal improvement by the fourth week of antidepressant treatment were unlikely to have an optimal antidepressant response without a change in pharmacotherapy (77). The authors suggest that, in those patients without improvement by week 4 and those patients who show early minimal improvement but no benefit by week 5, the treatment should be changed to an alternative antidepressant agent.

The guidelines for the duration of an adequate antidepressant treatment trial, supported by clinical trial and post-marketing study data, are being incorporated into routine clinical practice. However, it is not uncommon that patients may be prematurely judged treatment non-responders and have antidepressant drug changes after brief treatment trials of even a few days. It is clearly possible that within a matter of days a patient will be identified as a treatment non-responder because they are unable to tolerate a particular antidepressant agent. In contrast, it is unlikely that antidepressant efficacy can be accurately assessed after a brief trial of less than 7–14 days (20, 38).

Antidepressant drug dosing is a prominent issue during the acute phase of therapy. For the first antidepressants (e.g., TCAs and MAOIs), dose titration was an expectation, with the clinician advancing the drug dose in increments to optimize tolerability and efficacy. With the advent of the SSRIs, clinicians encountered the first antidepressant drugs where dose titration was not always necessary to induce remission. This was a difficult lesson to learn, and early on, patients treated with SSRIs routinely experienced a rapid titration to high doses during the first 1–2 weeks of therapy. There is little data to support this practice, particularly for an agent like fluoxetine with an active metabolite that typically requires several weeks to accumulate to a steady state plasma concentration after a dosage change (2, 67, 73).

Renewed interest in the pharmacokinetics of antidepressant agents has focused attention on the broad interpatient variability in plasma concentrations achieved at a given drug dose due to variability among physically healthy people in absorption, metabolism and elimination of the drug. This degree of variability is compounded by additional factors, such as age and co-morbid physical illness (11, 67, 68). The upshot is that patients may develop inappropriately high or low plasma concentrations of an antidepressant under routine dosing conditions. In the case of inappropriately high plasma concentrations, the patient is likely to experience additional adverse effects of the medication, which may be misinterpreted as a worsening of the depression (e.g., insomnia, restlessness, anorexia, etc.), leading to a further increase in drug dose.

In recent years, the concept of a "therapeutic window" for the SSRIs has been discussed in the literature, with the suggestion that in nonresponders it may be as appropriate to decrease as to increase the drug dose (13). Data suggest that the minimum formulations of the SSRIs are effective in a large percentage of patients. Conceivably there are patients who will have inadequate plasma concentrations of drug and hence require titration to higher doses. However, rather than the rule, the selected population of patients requiring high doses should be considered an exception and should be titrated upwards only after a reasonable trial on the minimum effective drug doses (66).

Emphasizing the use of the minimally effective antidepressant dose in routine practice is particularly relevant to the SSRIs, where a flat dose-response curve for efficacy has been established (67). In the case of other antidepressant reuptake inhibitors (e.g., venlafaxine) and receptor antagonists (e.g., nefazodone) ascending dose-response data for efficacy provide a rationale for the use of dose titration during the acute phase of treatment (63). Nonetheless, to optimize treatment response (i.e., minimize adverse effects) titration to higher dose should occur in those patients who have failed or only partially responded to the recommended initial drug dosing (67).

The issue of antidepressant dosing becomes increasingly complex with the considerations of interpatient variability in pharmacokinetics and the differences in dose-response relationships among antidepressant agents for both efficacy and toxicity. Clearly, one research goal is identification of concentration-response relationships, where they exist, for newer antidepressants similar to those established for TCAs.

Therapeutic drug monitoring (TDM) can serve several purposes during the acute, continuation, and maintenance phases of antidepressant treatment, including checking compliance and increasing the safe and efficacious use of drugs with well defined concentration-response relationships. The underlying principles and goals of TDM of psychotropic drugs have recently been reviewed (75). General features of psychotropic drugs that make TDM useful include: a small therapeutic index, large interindividual variability in the dose-plasma level relationship, difficult detection of early toxicity, long delay in onset of action, and well-defined concentration-response relationships.

To date, among all classes of antidepressants, the use of TDM has been well established only for TCAs (73). The application of TDM is particularly suited to the TCAs because, as a class: 1) there can be as much as a 30-fold variability in plasma TCA levels despite administration of the same dose of the same TCA to physically healthy individuals (60, 62, 74); 2) the drugs exhibit a narrow therapeutic index; and 3) plasma concentration-therapeutic response relationships have been determined (73).

The strongest of these plasma concentration-response relationships have been demonstrated for nortriptyline, desipramine, amitriptyline and imipramine (Table 4) [73]. In terms of likelihood of response, TCAs, when used based on clinically determined dose titration alone, will generally produce a response rate (i.e., at least a 50% drop in depressive symptom severity) of 60–70% and a remission rate of 30–40% (73). When the dose is adjusted based on plasma drug level monitoring, TCA remission rates increase ranging from 42% for imipramine to 70% for nortriptyline (73, 57).

Unlike the TCAs, the SSRIs do not fit the profile of drugs that require TDM for their safe and effective use. This is primarily due to their wide therapeutic index and flat dose-antidepressant response curves (21, 75, 66). As would be predicted based on a flat dose-antidepressant response relationship, attempts to determine an SSRI plasma concentration-antidepressant response relationship have in general been unsuccessful (43, 71). The literature on attempts to establish plasma concentration-antidepressant response relationships for SSRIs has recently been reviewed (32). The utility of TDM to check compliance may have application for use with SSRIs at some point in the future with the exception of fluoxetine. Due to the long half-lives of the parent compound and its metabolite, there would have to be gross noncompliance for TDM to detect it (66).

The flat dose-response curves for SSRIs are compatible with their presumed mechanism of action (serotonin reuptake inhibition), since the minimum effective doses of all SSRIs produce approximately 80% inhibition of platelet serotonin uptake in patients (48, 46). Because of the flat dose-response relationship for antidepressant efficacy and the broad therapeutic index of the SSRIs, it is unlikely that TDM will ever become a routine standard with the use of these agents. However, in treatment non-responders, TDM may help to determine whether the patient is developing plasma drug levels substantially lower than expected such that a dose increase should be attempted before concluding that the patient is not responsive to the SSRI (see Treatment-Resistant Depression on treatment resistant depression by Drs. Thase and Rush). The average plasma concentrations of SSRIs in treatment responders is available for reference (32, 67).

There is good reason to believe that TDM might be advantageous in guiding the dose adjustment of bupropion (32, 61, 70). A number of concentration-antidepressant response studies have had surprisingly consistent findings (30, 33, 36, 61). These studies suggested that the highest antidepressant response rate occurs at plasma levels of bupropion in a range of 20–75 ng/ml. Only one of these studies measured the metabolite levels in addition to the parent compound (30). This is unfortunate, because bupropion has a complex metabolism, with three active metabolites which circulate in concentrations in excess of the parent compound (61). This issue is of more than academic interest, because in the study by Golden, et al. (30), high plasma levels of metabolites, in particular hydroxybupropion, were associated with poorer antidepressant response and may be in part responsible for the seizure risk of this drug (31, 93). A slow-release form of bupropion has recently been marketed, but there are no reports as yet relating plasma concentration to response.

There is no substantial database to suggest utility of TDM with the use of trazodone. Despite over ten years of broad clinical usage, little is known about trazodone plasma concentrations and clinical response (44, 32). The metabolite of trazodone, mCPP, can achieve higher plasma levels than the parent compound, but plasma level-response studies of trazodone and the metabolite mCPP have not been reported (63). The metabolite is not believed to contribute significantly to the antidepressant properties of the parent compound trazodone. Specific studies of mCPP have identified this agent as a potent agonist of the 5-HT 2C receptor and when administered alone it can be anxiogenic (39, 12, 104).

Among the other recent additions to the antidepressant armamentarium, there are no published data on the relationship between plasma concentration and therapeutic response. Venlafaxine has linear pharmacokinetics and one active metabolite contributing to the pharmacologic effects of the parent compound (63). These features, combined with the ascending dose-response curve, suggest a potential for relating plasma drug concentration to efficacy. Although venlafaxine appears to be without mortal toxicity, considering the broad recommended dosing range (i.e., from 75–375 mg/day), TDM may provide some utility in dosing to achieve maximum benefit.

Nefazodone exhibits nonlinear pharmacokinetics and is biotransformed into three pharmacologically active metabolites: hydroxynefazodone; triazolodione; and mCPP. Both hydroxynefazodone and triazolodione appear to contribute to the overall efficacy of nefazodone. mCPP is believed to contribute minimally to the clinical pharmacology of nefazodone (63, 32). There are no available studies relating plasma concentrations of nefazodone and its metabolites to clinical response. Similar to nefazodone, mirtazapine is also a potent 5-HT 2A antagonist without published data that supports a relationship between plasma levels achieved during treatment and clinical response.

Conventional TDM is not applicable to MAOIs. However, the degree of platelet MAO inhibition appears to be a useful bioassay to guide dose adjustment. Studies support maximum antidepressant response to the three irreversible MAOIs requires over 80% inhibition of platelet MAO (19, 79, 10).

Despite the widened therapeutic index that characterizes many of the recently marketed antidepressants, the potential applications of TDM to antidepressant pharmacotherapy are significant. Hence, this will remain an active area of research. Important factors continuing to drive the interest in TDM are the broad interpatient variability in antidepressant drug clearance, the need for further refinement of dosing guidelines beyond the dose ranges generated from clinical trials, and the need for objective measures to apply to patients judged to be treatment nonresponders during the acute phase of pharmacotherapy and in those not uncommon cases where patients experience relapse or recurrence of illness, despite maintenance of the drug doses that successfully induced remission.

Related to therapeutic drug monitoring (TDM) but a topic unto itself is the issue of antidepressant metabolites. Most antidepressants undergo extensive biotransformation prior to their elimination from the body. The ratio of metabolites to parent compound can be highly variable based on interindividual differences in rates of biotransformation and elimination. Clinical response may at times be primarily a function of the metabolite. Therefore, we need to know the pharmacology of antidepressant metabolites in order to properly interpret the clinical response observed in patients.

We know surprisingly little about antidepressant drug metabolites. Even for drug classes like the TCAs and MAOIs, after 30 years of marketing, the significance of their metabolites is unclear. The hydroxylated metabolites of TCAs inhibit biogenic amine uptake, but their potency and significance relative to the tertiary and secondary amine parent compounds are unknown. We do know that, in elderly persons and patients with decreased renal function, the polar metabolites can accumulate several times in excess of the parent compound (68). However, the clinical significance of increased levels of circulating hydroxylated metabolites, both in terms of antidepressant response and adverse effects is obscure.

A case in point for the clinical relevance of antidepressant metabolites is clomipramine. This agent is the most potent TCA in terms of serotonin uptake inhibition, but its metabolite desmethylclomipramine—like other secondary amine TCAs—is actually a more potent norepinephrine uptake inhibitor. In some patients receiving clomipramine, more than 70% of their circulating drug concentration will be the demethylated metabolite (66). Without knowledge of the metabolite's activity and relative plasma concentration, the clinician might conclude incorrectly that a patient who fails to respond to clomipramine is not responsive to serotonin uptake inhibition and hence not use an SSRI.

Among SSRIs, there are substantial differences with regard to active metabolites. Norfluoxetine, the active metabolite of fluoxetine, is equipotent to the parent compound both at blocking the serotonin reuptake site and inhibiting the P-450 isoenzyme CYP 2D6. This metabolite, once steady state is achieved, circulates at plasma levels 2–3 times higher than that of the parent compound fluoxetine and exhibits a half-life 4–8 times that of the parent compound. This profile is in contrast to other SSRIs, like sertraline and paroxetine, where the metabolites do not appear to contribute to the clinical effects of the parent compounds in a meaningful way (66, 67).

Multiple active metabolites have been identified for several of the more recently marketed antidepressants. While there is data on the activity of some of these metabolites, the data comes from in vitro studies of the isolated metabolite. The limitation of this approach is that the actual contribution of the metabolite to the clinical activity of the antidepressant remains obscure. As discussed in the preceding section on therapeutic drug monitoring, mCPP is an active metabolite of both trazodone and nefazodone. In studies of the isolated metabolite mCPP, the compound is a potent agonist of the 5-HT2C receptor, and when administered alone to patients, is anxiogenic, presumably due to this mechanism (39, 12, 104). It is presumed that mCPP does not accumulate to a concentration that would permit significant contribution to the clinical activity of either trazodone or nefazodone, but this has not been studied per se. As a substrate for the P-450 isoenzyme 2D6 (cyp 2D6), it is not unreasonable to expect that mCPP plasma concentrations would increase when the parent antidepressants are co-administered with inhibitors of CYP 2D6 (63). In such cases, the clinical manifestations of this metabolite may be significant.

Mirtazapine, venlafaxine, and bupropion are also biotransformed into several metabolites. Only one metabolite of venlafaxine, 0-desmethylvenlafaxine, has prominent activity and appears to contribute equally to the pharmacology of the parent drug (H). Bupropion has three active metabolites which have received some attention (30, 70). During antidepressant treatment, these metabolites accumulate in concentrations several times higher than that of the parent compound (61). There is some case material suggesting that high plasma levels of these metabolites, particularly hydroxybupropion, are associated with an increased incidence of serious adverse effects as well as poorer antidepressant response (31, 61). However, systematic investigation is still needed, and these provocative preliminary findings have not influenced clinical practices as yet.

Further systematic study of antidepressant drug metabolites is critical to the refinement of our understanding of the clinical activity of antidepressant drugs and will play a permissive role in the development of plasma concentration-response relationships for both efficacy and toxicity. The interested reader is referred to more detailed discussions of this pertinent topic (66, 73, 32).

A product of the 1990s and the era of managed care medicine is the practice guideline to optimize treatment outcomes and maximize the efficient use off health care resources. Ideally, a practice guideline is supported by the relevant medical literature and represents a consensus of opinion among "experts" and practicing clinicians. In recent years, a number of practical how-to publications for the treatment of depression have become available (38, 75, 20). In large part these guidelines present algorithms for the selection and usage of antidepressant pharmacotherapy during the acute and continuation phases of treatment. The how-to approach to the treatment of depression is timely in light of the recent expansion and diversity of antidepressant treatment options. It offers several advantages, not the least of which is consistency of treatment practices and improved quality of clinical data.

Among the credible algorithms providing direction for the treatment of depression there are a number of shared features that have served to advance the rational approach to antidepressant pharmacotherapy and deserve mention. The first of these common features is the concept of tailoring treatment selection to the individual patient. Efficacy, tolerability, and toxicity issues related to antidepressants can be discussed in the abstract, but the clinical relevance of a pharmacologic profile for a given agent depends in large part on the patient. Characteristics such as age, co-morbid medical illness and concomitant medications will have a major impact on the treatment risk profile and likelihood of a successful outcome. Hence, any reasonable treatment algorithm should guide the clinician through a matrix of patient variables and drug features including the pertinent pharmacodynamics and pharmacokinetics.

Other common features in antidepressant practice guidelines that have served to advance rational pharmacotherapy are an emphasis on careful dose titration, the importance of treatment duration in defining a therapeutic medication trial, and an approach to partial and non-responders that is focused on the clinical pharmacology of antidepressant options (see Treatment-Resistant Depression). An important addition to the next edition of this chapter will be a careful analysis of the clinical benefit and scientific support of proposed antidepressant algorithms.

Our ability to determine the merits of one drug versus another is quite limited. To understand those limitations, one must understand the clinical trials research done as part of the drug development process. Perhaps the first thing to understand is that it is an expensive process. The cost is estimated at over 250 million dollars to bring a new antidepressant to market. For that reason, the focus of this development process is to successfully support the new drug application which is what the FDA evaluates to determine whether the drug is sufficiently safe and effective to warrant approval for marketing.

Typically, the total clinical trials experience with a drug involves treating approximately 2500 people with the new investigational agent. Of these, only a few hundred are treated for more than a few months, and only 10–20 patients are exposed to medications for over one year (Table 5). Hence, in typical investigational drug development, the cumulative human exposure to the active agent is relatively small. As such, their are real limits of clinical trial data in terms of drawing conclusions about long-term adverse effects and the cumulative safety risk of a new compound. The vast majority of data come from trials in which the goal was to establish an optimal dose. As a result, it is not uncommon that over 90% of patients are treated with doses other than those that will be commonly used in clinical practice. Also, the databases are frequently limited to at most a couple of comparator agents. Almost invariably, most of the comparisons are done with a tertiary amine TCA (i.e., amitriptyline, imipramine).

The entry criteria for clinical trials in drug development is such that the population studied is a narrow subset of the population that the clinician treats. In fact, the population studied is more representative of the patients treated by primary care physicians rather than those treated by psychiatrists. The reason for limiting the scope of the patients treated goes back to the concern that if widely divergent groups are included in the trial, it will compromise the ability to determine what was due to the drug versus what was due to the natural variability in the sample being studied.

The typical clinical trial participant suffers from mild to moderate depression. All participants are screened and generally excluded if the have concomitant psychiatric illness such as psychosis, specific types of personality disorders (e.g., antisocial personality disorder), substance abuse, suicide risk, unstable medical conditions and concomitant medications. There are also age (i.e., less than 18 and greater than 65 years old) and frequently gender (i.e., females of childbearing potential) exclusion criteria. The matter of restricting women of childbearing potential from the early stages of clinical trials is currently undergoing review, given recent attempts to encourage research on the treatment of illnesses affecting women. An interesting area for future research would be to characterize the subpopulation of depressed patients who participate in clinical trials research. Why these patients answer advertisements for clinical trials, enter cumbersome treatment programs that disrupt their daily schedules and agree to accept a chance of receiving placebo is not known. Readers are referred to a specific review for further discussion of clinical trials (92).

With these limitations of clinical trials research in drug development, much of the information that clinicians would like to know will not be available for years after the medication has been released. For a new drug, the extent, duration and diversity of human exposure expands exponentially in the first several years after marketing. However, from a research standpoint, the studies that follow marketing often suffer from design limitations. The most frequent limitation is the absence of a placebo control group and inadequate power to separate the efficacy of one drug from another except in the most obvious ways. Properly controlled studies require funding by agencies such as the National Institute of Mental Health (NIMH). Such funding is quite limited. Hence, there are still many unanswered questions about TCAs, despite the fact that these drugs were the gold standard of treatment for major depression for nearly 30 years.

To summarize the previous section, clinical trials research on antidepressants is limited with regard to number of patients studied, the rarefied population of patients selected to participate in the trials, and the study design that often differs considerably from the conditions of routine clinical practice. Despite these limitations, the clinical trial data is more relevant to the acute treatment phase than to any other phase of antidepressant pharmacotherapy because the vast majority of study participants are treated for 90 days or less (Table 5).

Clinical trial data represent the edge of knowledge about a newly released medication. This data establishes the drug efficacy. Once marketed, experience in the real world goes beyond efficacy to establish the actual effectiveness of a drug and ultimately determines its acceptance. Effectiveness of a medication can be conceptualized as the algebraic sum of its efficacy, tolerability, safety, and ease of use. The remainder of this paper will review these components of effectiveness as they relate to the currently marketed antidepressants and to the extent that existing data permit.

The rate of onset of antidepressant action is important. The faster the illness can be brought into remission, the less the patients' suffering and the less the likelihood of death due to an otherwise treatable illness. Many, if not all antidepressants, have aspired to the throne of faster onset-of-action; as of yet, none have convincingly claimed it (5).

In the case of TCAs and mirtazapine, their antihistamine-mediated effects may offer the possibility of a faster drop in the total HAM-D score due to sedation. Trazodone has been reported to provide early anxiolysis, but this claim has not been demonstrated under controlled conditions (83). Nefazodone has been shown to produce a rapid (i.e., by 1 week) and statistically significant improvement in sleep and anxiety symptoms associated with depression (25). However, the full antidepressant response appears to take an interval of time comparable to the SSRIs. SSRIs show both early activation and anxiolysis (4). Bupropion and MAOIs frequently have activating effects on anergic patients. Although these reported features may contribute to a psychiatrist's antidepressant selection for a given patient, there is no data to support a superior response rate of the overall affective illness.

In the venlafaxine clinical trials, patients on 375 mg/day demonstrated a response rate five times that of placebo after one week of treatment (63). These data are consistent with the observation from basic studies that venlafaxine, at doses that inhibit both serotonin and norepinephrine reuptake, produces a more rapid development of beta-adrenergic receptor down regulation than agents that only potentiate one of the biogenic amines (3). Taken together, this onset-of-action data is provocative and may be applicable in cases of a severely ill patients for whom shortening the time of antidepressant action is particularly important. Under routine conditions, it is more likely that patients will be on lower doses of venlafaxine that have been titrated to optimize tolerability. In such cases, response rate is likely to be similar to that of other antidepressant agents.

The manner in which onset of antidepressant action has been measured is problematic with regard to determinations of absolute response rate and comparison of different agents. Onset of action is usually defined as the time when the difference in depression severity between a group of patients treated with the antidepressant can be separated from a group of patients treated with placebo, or, in rarer instances, from a group treated with a comparator agent. This approach obscures and confuses the issue. It is based on the ability to detect an average difference between drug and placebo efficacy, as opposed to determining the true onset of antidepressant action. The problem is further complicated by the fact that the most reproducible phenomenon in psychiatric research is the first two-week drop in the average depression severity score in patients with major depression entered into a double-blind, placebo controlled study.

There is also the question of what is meant by the phrase, "onset of antidepressant action"? Some have suggested that the phrase can be defined as a drop of at least a specified minimum in the severity of the patient's depressive syndrome as assessed by a standardized rating scale (e.g., Montgomery-Asberg Depression Rating Scale, Hamilton Depression Rating Scale). This suggestion raises many questions. What size drop would be clinically significant? Some have proposed a 4-point change on a rating scale as significant. However, these scales are not ratio scales, meaning that 4 points at one place in the scale are not equal to 4 points somewhere else. For example, a drop from 8 to 4 or a drop from 34 to 30 on the Hamilton Depression Rating Scale (HAM-D) may not have the same clinical impact as a drop from 20 to 16. A score of 8 on the HAM-D is generally considered remission, whereas the patient with a score of 30 is still severely ill. In contrast, a score of 20 is often the minimum severity score needed to enter a clinical trial treatment program, while a patient with a score of 16 would not be accepted. A further question is: Does it matter which 4 points change? If the patient has prominent insomnia, which can account for 6 points on the HAM-D, a tertiary amine TCA can produce a 4 point drop quickly due to anti-histamine mediated sedation. Is this equivalent to onset of antidepressant response?

Despite these problems, the issue of rapid onset of antidepressant action is important and will continue to demand attention. It has been suggested that patients who begin to exhibit some improvement within the first week of treatment will go on to be treatment responders (18). Other researchers have suggested that the patient with an early, robust response is likely to be a placebo responder. At the other end of the spectrum, a subpopulation of "slow responders" has been identified (19, 14). These patients required in excess of eight weeks before significant improvement was observed. Because of the variance in rate of response, a likely drug responder should begin to demonstrate some significant symptom reduction within the first three weeks of treatment (25). Based on this belief, clinical guidelines suggest a minimum 4–6 week treatment trial duration (75, 38, 20).

Efficacy is a fundamental criteria in the selection of an antidepressant agent. It would seem likely that antidepressants with different mechanisms of action would have differences in either their spectrum of activity (i.e., subtypes of the disorder in which they work preferentially) or in terms of their overall efficacy (i.e., antidepressants with certain mechanisms of action would work in a larger percentage of patients). Despite this logic, there is no compelling evidence that any one antidepressant has greater efficacy than another in the acute phase of antidepressant treatment. The reader is referred to Long-Term Treatment of Mood Disorders by Drs. Prien and Kocsis for discussion of efficacy and effectiveness issues related to the continuation and maintenance phases of antidepressant treatment.

There are now hundreds of antidepressant efficacy studies comparing a single drug to placebo or, in the case of the newer antidepressants, the test drug has been compared to a tricyclic antidepressant. Many of these studies are well controlled, while others are partially controlled or open-label trials. Despite the size of the clinical trial literature, the ability to compare efficacy among the different classes of antidepressants has been limited, in part due to the fact that the average efficacy study compares only one or two agents to placebo. More recently, the meta-analytic approach has been used to integrate the results of these separate studies into an evaluation of the relative efficacy of multiple antidepressant drugs. Meta-analysis is a well established research method for making quantitative comparisons using the data from multiple sources (96).

Restricting the data sources to double-blind, placebo controlled studies, the general finding from meta-analyses by independent research groups has been that: 1) marketed drugs representing the major classes of antidepressants are effective when compared to placebo; 2) the overall response rate for these antidepressant drugs in nonpsychotic depressed patients was approximately 65%, with an average placebo response of 35%; and 3) there was no evidence that the efficacy of a specific drug or drug from a specific antidepressant class was superior (5, 38, 102).

The most common comparator agents used in antidepressant clinical trials is a tertiary amine TCA, usually amitriptyline or imipramine, as the gold standard of antidepressants. An underlying and unresolved question, in this regard, is what defines a "good" study when using TCAs? Comparison studies using tertiary amine TCAs are technically difficult to do, due to the narrow therapeutic and tolerability ranges of these drugs. The investigator almost invariably finds himself in a "no win" situation. If the study permits gradual titration of the dose, most patients on TCAs will finish on doses which many psychiatrists consider inadequate to test the efficacy of these drugs. The reason is that many patients simply can not or will not tolerate such doses. If the study calls for aggressive dosing of the tertiary amine TCA, then there will be a large drop out rate. In that instance, the study can again be faulted for being an inadequate test of the efficacy of the TCA. In fact, both of these outcomes correctly reflects the difficulty with using tertiary amine TCAs in clinical practice.

With a 30-fold interindividual variability in plasma drug concentration at a given dose, what is an adequate TCA dose to use for efficacy comparisons? There is compelling data to suggest that response rates to TCAs can be markedly increased by adjusting the drug dose based on plasma level determinations (73, 72). Does this mean that TCA dosing in clinical trials should be based on therapeutic drug monitoring? If so, how should results be interpreted from studies failing to use TDM? The vast majority of clinical trials use tertiary amine TCAs. Is this the most appropriate comparison or should a secondary amine, like desipramine or nortriptyline, which is better tolerated be used? There are no ready answers to these questions. However, secondary amine TCAs are more potent as antidepressants than tertiary amine TCAs and as a result, lower concentrations of these drugs are needed for treatment. Hence, the secondary amine TCAs are better tolerated and there is an increased likelihood that the patient will complete a therapeutic treatment trial. These features of secondary amine TCAs suggest they would be better comparators in efficacy studies of new agents.

For these and other reasons, we are often left to conclude that there is no difference in the overall efficacy of different types of antidepressants. The response rate (i.e., percentage of patients who experience at least a 50% reduction in the severity of their depressive episode) on average will be between 55–70%. The remission rates (i.e., the percentage of patients who experience a complete resolution of their depressive episode), although usually not reported, on average will be about 10–20% less than the response rates. Neither measure convincingly separates the existing classes of drugs when taken as a whole. However, by examining the database for individual classes of antidepressants, one can identify features of specific agents that will likely be desirable for some patients and features that will be undesirable for others.

With regard to the "spectra of activity" for the various classes of antidepressant agents, it is reasonable to comment on the identified subtypes of major depressive disorder. These subtypes, defined by clinical symptomatology, have different natural courses and respond to different forms of somatic therapy. The depressive subtypes include: melancholic, also referred to as "endogenous" depression; atypical, also referred to as "mood reactive" depression; and psychotic depression. One could add to this list an additional subtype, bipolar depression, which although it may present with melancholic, atypical or psychotic symptoms, has its own particular course and treatment risks.

In the early days of antidepressant pharmacotherapy, when treatment options were limited primarily to TCAs and MAOIs, considerable effort was made to identify which patients would preferentially respond to these antidepressant agents. Data from those and subsequent studies suggested that melancholic depression responds well to TCAs and that patients with atypical depression had a preferential response to MAOIs (47, 76). As the antidepressant armamentarium has expended in the last decade, so have the studies examining treatment efficacy of the new antidepressant agents in the various depressive subtypes. There is no ample evidence that both melancholic and atypical depression respond well to selective serotonin reuptake inhibitors (38, 63). The reader is referred to Selective Serotonin Reuptake Inhibitors in the Acute Treatment of Depression on the selective serotonin reuptake inhibitors for further discussion.

For the psychotic and bipolar subtypes of depressive disorders, there are relevant clinical issues that apply in general to all antidepressant pharmacotherapy and deserve mention here. There is consensus that patients with psychotic depression require either electroconvulsive treatment (ECT) or combination therapy with an antidepressant plus an antipsychotic agent (56). Antidepressant medication alone is unlikely to produce an optimal response in psychotically depressed patients. In combination with an antipsychotic agent, there is no convincing data that any one class of antidepressant is superior for treating psychotic depression.

The bipolar depression subtype (i.e., an episode of depression in a patient with a previous history of mania) has received considerable attention over the years. This attention is because antidepressant treatment of the bipolar subtype has the particular risk of inducing a manic episode. The literature regarding the ideal antidepressant agent for bipolar depression is not clear at this time. Historically, MAOIs were believed to carry a particular risk of inducing mania in vulnerable patients, but now there are data that suggest MAOIs may be beneficial for certain types of bipolar patients (see section on MAOIs). In recent years, there has been discussion regarding the use of bupropion as the antidepressant of choice for bipolar depressed patients, but this claim is premature (see section on combined dopamine-norepinephrine reuptake inhibitors).

At the present time, the literature suggests that any antidepressant agent may induce mania in a bipolar patient. Hence, for depression treatment in this population the recommendation of combination therapy with an antidepressant agent plus a "mood stabilizer" is a reasonable approach (38). Acceptable mood stabilizing agents include lithium, carbamazepine, and valproic acid. Of interest, among these mood-stabilizing agents, only lithium has been shown to have distinct (albeit marginal) antidepressant properties (89).

There is a substantial body of data which indicates that the selective serotonin reuptake inhibitors (e.g., fluoxetine, sertraline, paroxetine, fluvoxamine) are superior to placebo and equal in efficacy to the TCAs (6, 22, 24, 80, 67). Like the TCAs, the SSRIs are a broad-acting class of antidepressants. Of note, patients who fail to respond to TCAs can still respond to SSRIs and vice versa. In double-blind crossover studies (1, 24, 54), 60–65% of patients who fail to respond to monotherapy with an SSRI responded when switched (without breaking the blind) to a secondary amine TCA such as nortriptyline or desipramine. The same percentage held true when switching TCA nonresponders to an SSRI. From a research perspective, these data are provocative. Does it lend support to the theory of different populations of depressed patients based on pathophysiology (e.g., NE-mediated versus SE-mediated) or is it more a reflection of the difference in tolerability and toxicity profiles for TCAs and SSRIs? Characterizing these populations of drug-specific responders is an important direction for further study. The reader is referred to Selective Serotonin Reuptake Inhibitors in the Acute Treatment of Depression for a further discussion of this class of antidepressants by Dr. Montgomery.

Antidepressants characterized by the ability to inhibit the neuronal reuptake of both serotonin (SE) and norepinephrine (NE) can be further subdivided into selective and non-selective, based on the clinical relevance off their non-antidepressant pharmacodynamic profile (Table 2). The non-selective subgroup of agents is dominated by the tertiary amine tricyclic antidepressants (TCAs). Venlafaxine is the only compound currently in the selective subgroup.

Among the TCAs, amitriptyline and imipramine have been the most frequently studied, either directly or as a control for another active agent. In the vast majority of these studies, the TCA has been more effective or equally effective to placebo (25). In general, there is a 30% drug-placebo difference. Studies failing to demonstrate a definite superiority of TCA over placebo typically suffered from methodological inadequacies (e.g., insufficient dose, inappropriate rating scales, and small nonhomogeneous patient populations). In studies comparing amitriptyline with another tertiary amine TCA, imipramine, the two drugs were found to be equally effective. Likewise, studies comparing the secondary amine TCAs, desipramine and nortriptyline, to tertiary amine TCAs found no difference in efficacy although secondary amines were more potent (e.g., lower doses and lower TCA plasma levels necessary for antidepressant response) [72].

TCAs are broad-acting and have been shown to be effective in treating all depressive subtypes. There has developed a belief among some psychiatrists that TCAs are the most effective treatment for the severe melancholic subtype of major depressive disorder, also referred to as endogenous depression and mood-nonreactive depression. The belief that TCAs are a more effective treatment for severely depressed inpatients is supported by the Danish University Antidepressant Group (DUAG) studies (16, 17). In these controlled, multicenter trials, severely depressed inpatients had a superior response to clomipramine, compared with treatment with SSRIs (either citalopram or paroxetine).

In fact, the most striking response to TCAs (e.g., change in HAM-D rating) has been seen in severely depressed patients. This observation is consistent with work showing that higher HAM-D scores and a full neurovegetative syndrome predict the greatest drug-placebo response difference (41, 57). Comparative interpretation is somewhat complicated by the fact that efficacy trials for newer antidepressant agents have in large part enrolled only patients with depression of moderate severity.

One can speculate as to why TCAs may be more effective than SSRIs in severely depressed patients. Is there an advantage to targeting more than one biogenic amine system (e.g., SE and NE)? Do the other, presumed "non-antidepressant" effects of TCAs (e.g., antihistaminic sedation) give them an advantage in treating the severely depressed patient population? The former explanation is supported by a recent study with venlafaxine, also a mixed SE and NE reuptake inhibitor. In a double-blind, randomized assignment study, venlafaxine was found to have superior efficacy over fluoxetine in treating severely depressed inpatients (63). In contrast to TCAs, venlafaxine is devoid of "non-antidepressant" effects on histamine, acetylcholine and adrenergic receptors, which suggests that these mechanisms do not contribute substantially to the antidepressant efficacy of TCAs.

The most recent addition to the antidepressant armamentarium, mirtazapine, is similar to TCAs and venlafaxine, in that it has direct effects on both norepinephrine and serotonin. This should provide another test of the benefit of a dual-acting antidepressant agent. As yet studies suggesting particular efficacy of mirtazapine in severely depressed patients have not been published.

From the standpoint of future research design examining antidepressant efficacy relative to illness severity, a question is how depressed inpatients differ from depressed outpatients. Does a difference exist and is it only a function of severity of illness? Is it a function of co-morbidity? What are the variables leading to a patient's hospital admission for depression? Suicidality? Psychosis? Substance abuse? Once in the hospital, what factors may affect treatment response? Does inpatient status allow for more rapid titration to higher drug doses because the patient can go to their room and lie down? Since the severely depressed patients requiring hospitalization and those with concomitant illness have been excluded from clinical trials, the differential treatment response of these populations has yet to be clearly determined. Nonetheless, the issues raised are relevant to both data interpretation and antidepressant treatment selection in clinical practice.

Bupropion, the only marketed aminoketone antidepressant, has been found to be an effective antidepressant in a number of double-blind studies (51, 106). It is unique in both its apparent mechanism of action and its spectrum of antidepressant efficacy. It has no appreciable effect on the uptake of serotonin, and its most potent action is blockade of dopamine reuptake. Hence, its antidepressant mechanism is unclear. Interestingly, in vitro studies using rat brain synaptosomes, found bupropion to be an order of magnitude less potent than sertraline (an SSRI) at blocking dopamine reuptake (Fig. 2) [9].

There is evidence that bupropion is effective in patients who are TCA nonresponders. In double-blind, placebo-controlled trials of this agent, the largest difference between bupropion and placebo occurred in patients who had historically failed to respond to TCAs (90). In a parallel study, bupropion was compared to amitriptyline in patients with either primary or secondary major depressive disorder (MDD). In this study, secondary referred to an MDD which became apparent after the onset of another Axis I diagnosis, almost invariably an anxiety disorder. While amitriptyline was equally effective in both groups of patients, bupropion was substantially more effective in primary as opposed to secondary MDD (55). Consistent with this finding, bupropion (in contrast to TCAs) was found to be ineffective in treating panic disorder patients (87). Taken together, these data support a distinction between the clinical psychopharmacologic activity of bupropion and that of TCAs.

There are limited data from uncontrolled, open trials which suggest that bupropion may have particular application in the treatment of the bipolar subtype of depression (i.e., depressed phase of bipolar disorder). In these studies, bupropion was effective in treating the depressed phase of bipolar disorder and (unlike TCAs) did not appear to induce rapid cycling or an affective shift to mania (36, 103). Based on these data and an increasing number of anecdotal case reports, psychiatrists perceive this agent as more effective or at least less risky to use in the bipolar population. This conclusion is premature. There is clearly a need for controlled studies examining antidepressant therapy in patients with this complex affective disorder. The reader is referred to a recent review for a more detailed discussion of the somatic treatment of depression in bipolar disorder (105).

Monoamine oxidase inhibitors (MAOIs) are the only group of agents in the antidepressant class characterized as inhibitors of the enzymatic metabolism of neurotransmitters. In a recent meta-analysis of controlled efficacy studies, the response rate to MAOIs was comparable to the tricyclic antidepressants (38). At one time, the TCAs were considered to be more effective than MAOIs. A review of earlier studies showing a poorer response to MAOIs as compared to TCAs suggests this was the result of subtherapeutic doses of MAOIs administered to treatment-resistant populations (e.g., patients with psychotic depression). Several large, double-blind studies that established the efficacy of phenelzine and tranylcypromine as equal to TCAs and superior to placebo have recently been reviewed (91).

MAOIs have a spectrum of activity that differs somewhat from TCAs (50). Whereas phenomenological predictors of response have not been established per se for the other classes of antidepressants, data support some clinical features that predict preferential response to MAOIs. These features include mood reactivity, irritability, hypersensitivity to rejection, hypersomnia, hyperphagia, and psychomotor agitation. They are collectively referred to as reverse vegetative symptoms and are thought to define an "atypical" subtype of depressive disorder. This subtype of depression has also been referred to as hysteroid dysphoria or mood-reactive depression. Repeated studies have found a preferential response to MAOIs in this selected population (91). More recently, patients who meet the profile of the atypical depressive subtype have also been shown to respond well to SSRIs.

Uncontrolled studies and anecdotal reports have led to a belief among many psychiatrists that there is significant risk of inducing mania in depressed bipolar patients treated with an either MAOIs or TCAs. However, in a double-blind comparison with imipramine, data suggested that anergic, bipolar patients responded particularly well to MAOIs (97). As is the case for bupropion, further controlled studies are needed to determine the best treatment of depression in bipolar disorder. These gaps in our knowledge of treatment response in "complicated" depression (e.g., bipolar disorder, hospitalized patients, pregnancy) reflect the practice of excluding several populations of patients important to psychiatrists from clinical trials research.

The MAOIs currently marketed in the United States are irreversible and nonselective inhibitors of monoamine oxidase. Phenelzine, tranylcypromine and isocarboxazid inhibit both forms of monoamine oxidase (i.e., Type A and B). Moclobemide, marketed in 1993 in Europe, is the first selective and reversible inhibitor of monoamine oxidase-A (RIMA). Clinical trials of moclobemide in Europe have found it equal in efficacy to TCAs and superior to placebo (7). A unique spectrum of activity has not been defined for this compound. One concern has been raised that in routine practice most patients require higher than recommended doses of moclobemide for an optimal antidepressant outcome. At these higher doses of moclobemide, the degree of enzyme selectivity is diminished.

Selegiline (l-deprenyl), an irreversible but selective inhibitor of MAO—type B, is marketed in the United States for the treatment of Parkinson's disease. Approximately 70% of the monoamine oxidase in the human brain is MAO-B. Currently, there are no published data on the antidepressant efficacy of selegiline, but this is clearly a reasonable avenue of investigation.

There are now three antidepressant agents that fall into the category of receptor antagonists. Trazodone, the first drug to be marketed in this class, was at one time the most popular antidepressant by brand name (Desyrel) in the United States. Most psychiatrists speculated that this popularity had more to do with the broad therapeutic index of this compound than with its efficacy. Review of the randomized, controlled trials with trazodone showed that the average drop in depressive symptom severity achieved was comparable to that seen with either TCAs or SSRIs and superior to placebo (44, 26). Similar to TCAs and SSRIs, trazodone is broadly effective with no substantial evidence to support a unique spectrum of activity. Despite these data, trazodone is considered by many psychiatrists to be a "weaker" antidepressant and has been criticized as being less effective in severely depressed patients.

One possible explanation is the dose-limiting problems of sedation and cognitive slowing or a "drugged feeling" on trazodone. These adverse effects may limit the number of patients who can reach therapeutic doses for antidepressant response. The problem is compounded by the short half-life of trazodone (i.e., 3–9 hours), which requires dosing to be divided into equal amounts given at least three times per day (66). These features of the drug may contribute to noncompliance and subtherapeutic treatment and in that way adversely affect remission rates in clinical practice. A review of trazodone efficacy studies found that those studies reporting a relatively poor response rate were characterized by use of a rapid dose escalation paradigm and a higher drop out rate (85). The conclusion was that a slower dose titration and overall lower doses (150–300 mg/day) maximized response.

Nefazodone bears some structural similarity to trazodone, but nonetheless there are important differences between these two molecules (9, 15, 23, 82). In terms of clinical pharmacology, nefazodone is an antagonist of 5-HT-2A receptors. However, at clinically relevant plasma concentrations, nefazodone also inhibits neuronal reuptake of serotonin.

The majority of clinical efficacy trials with nefazodone used a targeted-dose-range design which facilitates generation of a dose-response curve, and indeed an ascending dose-response curve was identified over the clinically relevant dosing range. Using this design in a series of double-blind controlled trials, nefazodone at doses of 300–400 mg/day was superior to placebo and equivalent in efficacy to imipramine and fluoxetine (63). Interestingly, at 600 mg/day, the response to nefazodone was virtually the same as to placebo. This result suggests that there is generally no advantage to a nefazodone dose exceeding 500 mg/day. At present there are no published data to show that nefazodone has other than a broad spectrum of activity. With the finding of rapid anxiolysis during the first week of nefazodone treatment, it has been suggested that this agent may be uniquely suited for use in the anxious/depressed patient, but no conclusive statement can be made in this regard (25).

Mirtazapine, similar to trazodone and nefazodone, has a prominent action as an antagonist of 5-HT-2A receptors. In contrast to these other agents, mirtazapine has potent effects on other serotonin receptor subtypes (e.g., 5-HT-2C and 5-HT-3) and alters noradrenergic transmission as an antagonist of presynaptic alpha-2 receptors (68). In conventional double-blind, randomized trials of efficacy, mirtazapine was superior to placebo and comparable to amitriptyline. In a follow-up protocol, patients who had not been effectively treated with either placebo or amitriptyline were treated with mirtazapine in a double-blind design. Interestingly, mirtazapine was as effective in treating both the placebo and imipramine non-responders as it was in the first phase of the study (68). This data suggest that, although TCAs and mirtazapine both have dual actions on serotonin and norepinephrine, there may be significant differences in their spectra of clinical efficacy.

Antidepressant tolerability is a critical issue in the short-term treatment of affective disorders and as such an importance consideration in antidepressant treatment selection. Patients are already experiencing considerable morbidity secondary to their illness. Additional burden associated with the adverse effects of antidepressant treatment can lead to poor compliance and treatment failure.

Tolerability is assessed in several ways during the clinical trials development program for a new drug. One way is to note any adverse event that occurs to the patient during the treatment interval, whether or not it is believed to be related to the treatment. At the conclusion of the study, all reports of treatment emergent adverse effects are tallied and summarized by major categories (e.g., complaints of "lightheadedness", vertigo, "woozy" sensation or "unsteadiness" may be collapsed in the category, "lightheadedness"). The incidence of a given adverse event category can then be compared across treatments in the same double-blind controlled studies (e.g., the incidence for the investigational drug vs. the incidence for placebo vs. the incidence for another marketed antidepressant).

This approach is limited by the fact that all reports of adverse effects are given the same weight, whether the adverse effect occurred once during treatment or was persistent throughout the treatment period. Nonetheless, the same is also true for the placebo condition, so that the difference should still be meaningful. Of clinical importance is the rate-limiting quality of a given adverse effect. This can be determined by whether the patient discontinues treatment as a result of the adverse effect. Such information is recorded during the treatment trial.

It is worth emphasizing that after a drug is marketed there is never again the large scale studies needed to address the important questions concerning tolerability and safety. Despite this importance, clinical trials have a number of limitations in determining the tolerability and safety profiles of antidepressant drugs. In a 6–8 week clinical trial, the development of tolerance to acute adverse effects and the occurrence of late-emergent adverse effects will likely be missed. The case of rare adverse effects occurring at a frequency of less than 1% is particularly problematic. There may only be a chance to study four or five patients who develop the effect out of the 2–3000 trial participants, which limits efforts to characterize rare adverse effects and determine their significance. Once the drug is marketed, the detection of adverse effects is generally dependent on spontaneous reporting by clinicians.

In terms of study design, a confounding issue is how to best assess the participants for adverse effects. This debate varies from fear of overendorsement of adverse effects by using a laundry checklist approach to fear of missing important drug effects by relying completely on spontaneous reports by the trial participant. The significance of the assessment approach becomes particularly important with regard to adverse effects the patient may be hesitant to report (e.g., sexual dysfunction).

An equally important limitation in clinical trials is the focus on pharmacodynamically mediated adverse effects at the near exclusion of those that are pharmacokinetically mediated. This is due in large part to the fact that participants are screened for medical illness and the use of other medications. As a result, significant pharmacokinetic drug interactions may not be determined until years after marketing (e.g., inhibition of P450 hepatic enzymes by fluoxetine) [66, 34, 35].

Among antidepressant agents, our understanding of tolerability issues is most advanced for the TCAs, by virtue of their years on the market and the absolute number of patient exposures. The underlying problem with the tolerability of TCAs is that a number of the mechanisms of action responsible for producing adverse effects (e.g., those mediated by histamine or muscarinic receptor blockade, etc.) occur at concentrations lower than or the same as those needed to block the neuronal uptake for NE and SE (Fig. 2) [9, 15, 82]. As a result, even rational dose adjustment based on therapeutic drug monitoring may not substantially improve the tolerability of the TCAs. The patient has to experience a large number of effects to receive the benefit of the mechanism that mediates antidepressant response.

In repeated studies, a number of adverse effects of TCAs occur in excess of placebo controls. These include the anticholinergic mediated effects (e.g., dry mouth, constipation, urinary retention, blurred vision), the antihistaminic effects (e.g., sedation, weight gain), and the anti-alpha adrenergic effects (e.g., orthostatic hypotension, lightheadedness). This adverse effect profile becomes increasingly problematic as a function of the age of the patient. The reasons for this include: 1) elderly patients develop higher TCA plasma levels on a given dose (11); 2) older patients are more sensitive to the adverse effects (e.g., they develop TCA-induced delirium at lower plasma drug levels) [68]; and, 3) the multiple pharmacodynamic actions of TCAs increase the potential for additive or synergistic interaction with other drugs the elderly are likely to be taking.

There is a belief among psychiatrists that as a group the newer antidepressants have substantially fewer adverse effects. In fact, as a group the newer antidepressants have less anticholinergic and antihistaminic effects than tertiary amine tricyclics and are considered to be more selective than TCAs in terms of their presumed antidepressant mechanism of action (81, 82). Not surprisingly, in comparison studies with tertiary amine TCAs, the average adverse effect burden and discontinuation rate secondary to adverse effects has been generally lower for newer agents (93, 64). These data are consistent with recent outcome studies which show a lower level of medication compliance with TCAs than for SSRIs and bupropion when antidepressant treatment occurs under routine practice conditions (42, 40).

However, non-TCA antidepressants are not without adverse effects. SSRIs potentiate the effect of serotonin at all synaptic terminals and affect a number of different regional systems in the CNS. Their adverse effect profile is consistent with this broad serotonin agonism and includes gastrointestinal disturbance, agitation, somnolence or insomnia, tremor, anorexia, sexual dysfunction and dizziness (38, 64). The incidence of nausea, diarrhea and sexual dysfunction occur in excess of that seen in active control patients receiving tertiary amine TCAs (81). The mechanism for gastrointestinal disturbance is considered likely due to stimulation of 5-HT-3 receptors and the nervousness/restlessness to stimulation of 5-HT-2C and 5-HT-1D receptors (37, 58, 59, 68).

Venlafaxine appears to have an adverse effect profile somewhat similar to the SSRIs, which is not surprising since it is a potent inhibitor of serotonin reuptake. Nervousness, tremor, sweating, dizziness and somnolence are dose-dependent (64). An adverse effect of increasing blood pressure is likely related to the other antidepressant mechanism of venlafaxine, namely potentiation of norepinephrine by blocking reuptake of this neurotransmitter. The incidence of an elevation in blood pressure is reported to be rare below doses of 225 mg/day (63, 64). It is not known whether unstable, elevated blood pressure is a serious risk factor for this adverse effect, since such patients were excluded from the clinical trials.

Similar to the SSRIs, trazodone, nefazodone, and bupropion are essentially devoid of anticholinergic and antihistaminic effects but not without significant adverse effects. In the case of bupropion, the principal adverse effects include restlessness, activation, tremors, insomnia, and nausea (70). While these adverse effects rarely require discontinuation, aggravation of psychosis, which has been reported, and occurrence of seizures do (18, 31).

In the case of trazodone sedation, cognitive slowing and orthostasis are consistent dose-limiting adverse effects (4). Results from the clinical trial development program for nefazodone suggest that these effects may be less of an issue than for trazodone (64). The most common dose-dependent, treatment emergent adverse effects for nefazodone were nausea, dizziness and somnolence. In a recent comparative analysis with a number of other antidepressants (e.g., paroxetine, fluoxetine, venlafaxine, imipramine), dizziness was most common with nefazodone (63, 64). The mechanism for the dizziness is likely to be in part due to alpha-1 receptor antagonism by nefazodone but may relate to 5-HT-2A receptor antagonism also. Of note, the incidence of sexual dysfunction, particularly ejaculation/orgasm disturbance was considerably less than for other serotonin active agents (63, 64).

Knowledge of the relevant adverse effect profile of mirtazapine is limited at the time this chapter is being written. In the conventional clinical trials, mirtazapine was better tolerated than the comparator agent amitriptyline and exhibited an adverse effect profile consistent with its pharmacodynamic profile (68). Mirtazapine has low affinity for muscarinic cholinergic receptors and alpha-1 adrenergic receptors and similar to SSRIs produced minimal anticholinergic side effects or orthostatic hypotension. The agent produced minimal nausea, which is consistent with mirtazapine's antagonism for 5-HT-3 receptors. There was minimal reporting of anxiety/restlessness, consistent with the drug acting as an antagonist of 5-HT-2C receptors (94). The most prominent adverse effect of mirtazapine was sedation related to the potent anti-histamine properties of the drug. Weight gain, mediated by the anti-histamine mechanism, may be an issue. The clinical community awaits the real-world test of experience with this agent to clarify the tolerability burden relative to other antidepressant options.

Finally a comment about MAOIs. These agents are generally well tolerated if the patient observes the tyramine-restricted diet and avoids medications containing sympathomimetic amines. The most common and treatment-limiting adverse effect is hypotension (38, 78). While changes in blood pressure are most often orthostatic in nature, a general reduction in blood pressure can be seen in some patients. This adverse effect may present as fatigue or decreased motivation which emphasizes the importance of routine blood pressure monitoring in patients treated with a drug from this class.

There is perhaps no other area where rational drug development has proven more successful than in widening the therapeutic index of new antidepressant drugs and increasing their safety. This need has long been recognized since patients with depressive episodes are at risk to take overdoses of their medication either intentionally or through inattentiveness. Some patients who are slow metabolizers of a drug may also achieve toxic plasma concentrations on conventional doses. The latter is the case with TCAs whether the patient is a slow metabolizer due to a genetic deficiency or drug-induced (e.g., fluoxetine) deficiency in the hepatic isoenzyme 2D6 (34, 35).

There are now multiple antidepressant options characterized by having broad therapeutic indices that reduce the risk of serious sequelae from overingestion. For many of the newer antidepressant agents, plasma concentrations an order of magnitude above the therapeutic range are only associated with minor adverse effects. Exceptions to this rule are the TCAs and bupropion. The difference between a therapeutic dose of TCA and a dose with the likelihood of causing serious toxicity can be as little as 5-fold (e.g., 200 mg/day vs. 1000 mg) [8]. The toxicity of TCAs affects the brain (producing delirium and seizures) and the heart (producing various types of conduction disturbances and even sudden death) [69].

There is an even narrower therapeutic index with the immediate release formulation of bupropion than with TCAs, although the consequences are generally much less serious. Doses of bupropion above 450 mg/day are associated with a substantial increase in the risk of grand mal seizures (12). Seizures are themselves typically nonlethal, and death is a rare possibility in overdose alone, due to the absence of adverse effects on the cardiovascular or respiratory systems (70).

SSRIs, venlafaxine, and drugs within the antidepressant class of receptor antagonists (e.g., trazodone, nefazodone, mirtazapine) have wide therapeutic indices and are nonlethal in overdose (63, 66, 68). MAOIs are somewhere between TCAs and SSRIs in terms of safety in acute over-ingestion.

The CNS toxicity of TCAs is plasma concentration-dependent, often exhibiting the sequential clinical stages of agitation, delirium, seizures, coma and death. The relative risk for delirium on a tertiary amine TCA increases 14 and 38 fold above plasma levels of 300 ng/ml and 450 ng/ml, respectively (72, 73).

The CNS toxicity of bupropion leading to seizures is dose-dependent. The seizure incidence is >0.4% at doses of 450 mg/day or less; approximately 1% at doses of 451–600 mg/day; and over 2% at doses of 601–900 mg/day (18). The pathogenesis of bupropion-induced seizures is likely to be in part pharmacokinetically mediated (66). This conclusion is based upon several observations: 1) the incidence of seizures is dose-dependent; 2) seizures generally occur within days of a dose increase; 3) seizures generally occur within the first few hours after the dose; and 4) bulimic patients with increased lean body mass may have a higher risk for seizures than do other patients.

As a class, the fundamental feature of TCAs responsible for their toxicity, in contrast to the other major classes of antidepressant drugs, is the ability to stabilize electrically excitable membranes through the inhibition of sodium conductance (i.e., blockade of sodium fast channels). This action is responsible for their cardiotoxicity and lethality in overdose (69). The most characteristic and serious sequelae of this mechanism of action involves slowing of cardiac conduction, manifested as prolongation of the QT interval, intraventricular conduction defects, and AV block leading to malignant arrhythmias (29).

The cardiotoxicity of TCAs is directly related to plasma concentration of the drug (66, 73). The slowing of cardiac conduction by TCAs occurs even at therapeutic plasma concentrations. As TCA levels increase above the recommended therapeutic range, the incidence and severity of the cardiotoxicity increases. Above plasma levels of 350 ng/ml, 70% of physically healthy young depressed patients will develop a first degree heart block versus less than 3% below 350 ng/ml (72).

The cardiac effects of TCAs are similar to the electrophysiologic properties of the Type 1A antiarrhythmic compounds (e.g., quinidine, procainamide, disopyramide). This slowing of cardiac conduction by TCAs was believed to be of possible benefit in some patients with pre-existing ventricular arrhythmias, causing reduction in premature ventricular contractions (29). Recently, this belief has been upset by findings from the Cardiac Arrhythmia Suppression Trial (CAST), which demonstrated an increased mortality in patients treated with antiarrhythmic drugs. In this long-term follow up study, patients with cardiac arrhythmias and those suffering an acute myocardial infarction were more likely to have a fatal outcome if they were being treated with an antiarrhythmic drug of the Type 1A class (98, 99).

Since TCAs are indistinguishable from Type 1A antiarrhythmic drugs, in terms of their effects on intracardiac conduction, this CAST data must now be included in the risk/benefit consideration when prescribing TCAs. Although the full interpretation is yet unclear, TCAs may become a second-line treatment for any patients suffering from cardiovascular disease or at risk for an acute myocardial infarction (e.g., age over 45, positive family history) [84].

Among the newer antidepressant drugs, SSRIs (6), bupropion (70), venlafaxine (63), nefazodone (63), and mirtazapine (68) have no direct effects on cardiac conduction. Trazodone had little direct action on the heart in preclinical and clinical trials. However, since its release there have been reports that trazodone aggravated arrhythmias in patients with preexisting ventricular conduction disease (100).

An important point is that TCAs were available for over thirty years before we have fully understood some of their risks. What do we actually know about the safety of newer antidepressant drugs? As discussed in the earlier section on clinical trials, results from clinical trial development programs have real limitations in terms of number of patients exposed to a new drug, the duration of that exposure and the doses used (Table 5). Given the widespread use of antidepressants and the increasing emphasis on long-term treatment of depression, there is a pressing need for ongoing controlled studies of cumulative drug safety.

In recent years, there has been a renewed interest in the pharmacokinetics of drug metabolism and the issue of pharmacokinetic drug interactions. This interest has been accompanied by major advancements in our understanding of drug metabolism, including the identification and cloning of important P-450 isoenzymes involved in routine Phase I drug metabolism and the international acceptance of a single nomenclature system for these enzymes that has facilitated systematic study and communication.

With the increased understanding of antidepressant metabolism has come an increased awareness of the contribution of pharmacokinetic drug interactions to the toxicity profile of antidepressant drugs. At least four antidepressant agents have been identified to have potent inhibitory effects on hepatic P-450 isoenzymes, These agents (e.g., fluoxetine, paroxetine, fluvoxamine, nefazodone) have the potential to alter the metabolism of co-administered prescription and over-the-counter drugs which may be a source of considerable toxicity. The reader is referred to recent reviews on this important topic (34, 35).

An axiom of neuropsychopharmacology is that the brain adapts to drugs. In fact, adaptation to the acute synaptic effects of antidepressants has been proposed to be what truly mediates relief of the depressive syndrome and account for the delay between initiating an antidepressant and the onset of antidepressant response. Such adaptation includes alteration in receptor density and/or sensitivity to their neurotransmitters. Such adaptation also leads to the potential for withdrawal symptoms.

The likelihood and nature of withdrawal symptoms following antidepressant discontinuation varies substantially among and within the different classes of antidepressants discussed in this and the other chapters on mood disorders. As a general rule, the likelihood and nature of the withdrawal is a function of the acute synaptic effects of the drug (i.e., its neuropharmacology), whereas the likelihood and the severity of withdrawal within a class is a function of the half-life of drugs in that class.

Antidepressants do not cause medically serious withdrawal symptoms, in contrast to sedative-hypnotics. Nor are the symptoms generally as unpleasant as those due to opiate withdrawal. In fact, withdrawal symptoms due to antidepressants are primarily limited to tertiary amine tricyclic antidepressants and to serotonin uptake inhibitors (i.e., serotonin selective reuptake inhibitors and venlafaxine).

In case of tertiary amine tricyclic antidepressants, withdrawal symptoms due to apparent cholinergic rebound can occur when patients are abruptly discontinued from antidepressant doses (i.e., 100–300 mg/day). These symptoms include gastrointestinal disturbance, autonomic symptoms, anxiety, agitation and disrupted sleep. These symptoms are often so mild that there is no need to reinstitute the antidepressant and then taper it more slowly.

Withdrawal symptoms due to serotonin uptake inhibitors are more likely to present as clinically significant problems. The likelihood and severity of these symptoms is inversely correlated with the half-life of the serotonin uptake inhibitor, being more common with fluvoxamine, paroxetine or venlafaxine than with citalopram, fluoxetine or sertraline. Symptoms include dizziness, sweating, nausea, insomnia, tremor, confusion, vertigo, fatigue, rhinorrhea, agitation, headaches, incoordination, irritability, memory problems, abdominal cramps, increased dreaming, gait disturbance, paraesthesia, and a flu-like syndrome. Knowledge of this withdrawal syndrome is important because it may be misdiagnosed as worsening of the underlying psychiatric syndrome including the development of a manic episode. Such a misdiagnosis can result in the patient being consigned to indefinite treatment with mood stabilizers.

Women are at increased risk for depressive disorders (see Mood Disorders Linked to the Reproductive Cycle in Women by Dr. Parry). As such, the risks associated with antidepressant drug use during pregnancy has been a long-standing concern. For ethical reasons, double-blind randomized studies of antidepressant drug effects on the fetus and mother are unavailable. Hence, most of what we know comes from case reports, uncontrolled retrospective studies of clinical outcomes, and basic research.

From basic research comes support for the clinicians' concern. In rats, prenatal exposure to D-amphetamine is associated with decreased neuronal density in the prefrontal cortex and decreased serotonin levels in the medial prefrontal cortex (95). In mouse embryo, serotonin uptake sites appear to mediate critical morphogenetic events in craniofacial development (88). Mouse embryos exposed to antidepressant agents that block the serotonin uptake site had craniofacial defects associated with extensive cell death in the mesenchyme (88).

In contrast to the straight forward nature of the basic science data, the clinical literature is difficult to interpret. There are both reports of toxicity as well as case reports and retrospective studies where live birth rate and the rate of fetal anomalies associated with antidepressant exposure do not appear different than in the general population (49, 86). In addition to the obvious limitations of uncontrolled, retrospective studies of ambulatory patients, additional limitations of the clinical data on fetal toxicity of antidepressants is a focus on "major" malformations and absence of follow-up to assess developmental problems (i.e., so called "soft-signs" of neurological impairment).

A recent report on the results of fetal exposure to antidepressants prospectively examined 228 pregnant women taking fluoxetine (14). Compared to 254 controls, there was no difference in the rate of pregnancy loss or occurrence of major structural anomalies in the drug-exposed newborns. The incidence of three or more minor anomalies was significantly higher in the drug exposed infants. Those infants exposed to fluoxetine in the third trimester had higher rates of premature delivery, admission to special care nurseries, and poor neonatal adaptation. This last finding is of particular interest in light of the clinical myth that fetal exposure to antidepressants in the third trimester is less toxic than first trimester exposure.

Another recent prospective study looked specifically at the effects of in utero exposure to antidepressant medication on subsequent cognitive and behavioral development (53). The study population consisted of 80 mothers who received a tricyclic antidepressant during pregnancy, 55 mothers who received fluoxetine during pregnancy, and 84 control mothers who received no antidepressant medication during gestation. The children from these pregnancies were assessed by a psychometrician for cognitive ability, language skill, temperament, and behavior, which were similar for children in all study groups. This study suggested that in utero exposure to the tricyclic antidepressants and fluoxetine does not have significant detrimental effects on subsequent early childhood neurodevelopment. A caveat to interpretation of this data is that all children examined were prepubescent. In animal models of in utero exposure to neurotoxin, the offspring are most likely to show abnormal behavior after the onset of puberty, when hormonal changes presumably begin to exert significant effects on brain function.

Recent prospective studies of antidepressant use during pregnancy are encouraging in that they suggest minimal effects on fetal development, risk of spontaneous abortion, and subsequent early childhood neurodevelopment. Given the gravity of potential adverse effects, it is clear that the subject of antidepressant use in pregnancy requires continued research efforts and clinical caution. Whatever the hazards associated with antidepressant use during pregnancy, they must be weighed in a risk:benefit ratio with the risks accompanying the untreated depression (e.g., suicide, impaired self- and neonatal care, etc.). A goal for the next edition of this chapter is a detailed analysis of the literature on this clinically related topic.

A separate but related topic to the use of antidepressant medication during pregnancy is the use of antidepressant medication during lactation. Although it is always desirable to minimize exposure of newborns to any drug, considering the benefits of breast-feeding and the frequent necessity of antidepressant pharmacotherapy post-partum, the issue of antidepressant pharmacotherapy during lactation is a dilemma. A recent review of the literature suggests that among antidepressant agents there may be differences with regard to risk to the nurslings (101). There are a number of published reports on nursing mothers who were taking antidepressant medication where serum antidepressant levels were measured in the nurslings. Among these reports, amitriptyline, nortriptyline, desipramine, clomipramine, dothiepin, and sertraline were not found in quantifiable amounts in the nurslings, and no adverse effects were reported. Other reports identified significant accumulation of doxepin and fluoxetine in nursling serum associated with adverse effects. The author of the review suggests that those antidepressants not identified in nursling serum are the treatments of choice for depression in the lactating female. However, it is important to acknowledge that these data are primarily case-based and uncontrolled, there is a significant margin of error in the serum drug levels, and the issue of drug metabolites was not thoroughly addressed. Hence, the issue of safety and antidepressant of choice for lactating mothers is not yet well defined. Nonetheless in those cases where the risk:benefit decision favors treatment, the case-based data on nursling drug levels is a relevant consideration in treatment selection.

Great strides have been made in the treatment of depression over the past decade due to the ability to rationally develop medications with targeted mechanisms of action. The major benefits of the new generation antidepressants have been in the areas of drug tolerability and safety, and ease of administration. The major disappointment is that remission rates have not improved. As a result, a substantial portion of patients fail to respond to an initial trial of monodrug therapy. This means that either there are forms of the illness that do not respond to pharmacotherapy or that there are different forms of the illness requiring different pharmacologic mechanisms of action for treatment. In the latter case, the challenge is to identify subpopulations of patients responding to specific pharmacologic interventions or to continue the trial and error approach with the sequential use of different agents. An alternative possibility is to develop drugs with multiple mechanisms of action that will treat a broad spectrum of patients but yet retain the tolerability and safety of the new generation antidepressants.

published 2000