The statistics for the epidemiology and course of major depressive disorder (MDD) remain impressive. As of 2000, MDD ranked as the fourth leading cause of disability for women and seventh for men, and it is projected to be second only to cardiovascular diseases by the year 2020 (+1, +2). In the United States, 16.2% of the population have had MDD in their lifetime and 6.6% have experienced an episode of major depression within the past year (+3).

Of those with MDD, about 90% are classified as at least moderately depressed. The odds ratios for women compared with men are 1.7 for lifetime MDD and 1.4 for 12-month prevalence. Fifty percent of persons who have had one episode of major depression will have at least one more episode, as will 70% of those who have had two episodes and 90% of those who have had three episodes (+4). About 20%—25% of persons who have major depression will develop chronic depression, with the episode lasting at least 2 years (+5). It has been estimated that as many as 30% of patients who present for primary care have MDD (+6, +7).

These epidemiological data indicate that major depressive disorder continues to be widespread, is frequently chronic and recurrent, and presents physicians of all disciplines with a substantial clinical challenge.

According to DSM-IV-TR (+8) criteria, a diagnosis of MDD requires that at least five of nine symptoms be present (Table 1+), although many other symptoms have been associated with the disorder (Table 2+). However, in a key genetic epidemiological study of twins, Kendler and Gardner found that the presence of as few as four symptoms of MDD with less than severe impairment and with symptoms lasting less than 14 days in one twin was associated with a substantial risk of MDD in the co-twin (+9). Dysfunction can occur with two of the nine symptoms, and functioning is unequivocally worse for patients with severe depression, defined as meeting threshold criteria for three or more of the nine symptoms (+10—+12). With as few as two of the nine symptoms of depression associated with dysfunction, cause and effect cannot be determined (do the symptoms cause the dysfunction or does the dysfunction cause the symptoms?), leaving the threshold for treatment unclear. A full spectrum of depressive disorders has been proposed (+8) (Table 3+).

Minor and subsyndromal depression may precede an episode of MDD as prodromal symptoms, occur after an episode of MDD as residual symptoms, or persist without ever fully meeting the diagnostic criteria for MDD (+13—+16). Whether or not minor and subsyndromal depression should be treated with antidepressants is still an open question, with some studies finding no difference between active drug and placebo. For example, in a placebo-controlled study comparing paroxetine and problem-solving psychotherapy for elderly primary care patients with minor depression (+17), these treatments were modestly better than placebo, but functioning improved only among participants whose functioning was the worst at baseline. Moreover, remission rates were similar for all three treatment groups: paroxetine, 53.1%; psychotherapy, 44.0%; and placebo, 49.1%. A similar study of younger primary care patients with minor depression found no differences among treatment groups receiving paroxetine, psychotherapy, or placebo (+18). However, Judd and colleagues (+19) found active drug to be superior to placebo in a study of fluoxetine, which was significantly more effective than placebo in improving symptoms of minor depression and improving psychosocial functioning.

When minor or subsyndromal depression is the outcome of an episode of major depression, in contrast to an outcome of full remission, patients are at higher risk of relapse and recurrence (+13, +20). Remission, therefore, has become the goal of treatment (+21, +22). Unfortunately, only a minority of patients achieve remission with available treatments, and multimodal interventions or thoughtful polypharmacy may be necessary (+6, +12). Residual symptoms have increasingly attracted the attention of researchers, and targeted interventions are under development for the management of persistent fatigue, insomnia, and amotivational states (+21—+23).

Although minor and subsyndromal depression are associated with distress and dysfunction, it is not clear when subsyndromal and minor depression should be treated: What is the threshold for treatment? In major depressive disorder, residual symptoms after a course of treatment are also associated with distress and dysfunction. In both cases, if distress and dysfunction are present, clinicians should treat their patients to remission whenever possible. One problem in practice, however, is that it can be difficult to determine when and whether remission occurs if depressive symptoms are not measured periodically. Patient rating forms provide a simple means of measuring symptoms. The Quick Inventory of Depressive Symptomatology—Self-Report version (QIDS-SR) (+24) or the Beck Depression Inventory II (BDI-II) (+25) can easily be incorporated into clinical practice. Patients can complete such instruments in a few minutes and help clinicians track progress (or lack thereof). Reviewing the self-report with the patient allows the clinician to have a frank and focused discussion with the patient about residual symptoms, negotiate the next steps, and evaluate the effectiveness of any interventions.

Nearly three-quarters of persons with MDD have at least one other comorbid psychiatric disorder (+3). Most common are anxiety disorders (59%), followed by impulse control disorders (30%) and substance use disorders (24%). Those with comorbid disorders seek help more often and, as would be expected, do not function as well as those without any comorbid disorders (+3). One of the great paradoxes of modern clinical trials of treatments for MDD is that many studies exclude most patients with comorbid disorders, which limits the generalizability of results. Zimmerman and colleagues found that among patients treated for depression in a large outpatient clinic, only 14% would meet criteria for inclusion in clinical trials (+26). Clinicians should actively probe for comorbid anxiety, impulse, and substance use disorders and treat them accordingly.

Medical comorbidity and MDD are reciprocal: Patients with medical comorbidity have higher rates of depression, and those with MDD have higher rates of some medical disorders, such as cardiovascular disease (+27—+29). Cardiovascular outcomes have been shown to be adversely affected by MDD. For patients who have depression after myocardial infarction (MI), the risk of death is three to four times higher than for those without depression, even when the analysis is covaried by left ventricular function (+30—+32). Research suggests that while social support may not directly affect survival, high levels of social support may act as a buffer against the high risk of death in patients with post-MI depression (+33).

Major depression, but not minor depression, has been found to be an independent risk factor for the development of MI—and it is as strong a risk factor as impaired left ventricular function (+34). The SADHART trial, although not designed to address the question of whether antidepressant treatment improves cardiovascular outcomes for patients with post-MI depression, suggested that sertraline is a safe treatment for depression in these patients and that it might improve cardiovascular outcomes (+35). Ongoing studies are addressing this question. When consulting with medical colleagues, it is important to avoid the mind-set that some depression is normal and temporary after MI and therefore that it should not be treated. Whenever depressive symptoms are present, they should be assessed (using a patient-rated measure), treatment should be initiated, and progress should be tracked.

Stress has long been implicated in the precipitation of episodes of major depression (+36), but the pathophysiology of the relationship between stress and depression is quite complex. Stress has been shown in animal models to decrease gene expression of BcL-2 and BDNF (+37), proteins necessary to maintain robust and healthy neurons (+38). Antidepressants have been found to increase gene expression of BDNF and block the decrease precipitated by stress (+39). A longitudinal study of more than 1,000 persons between ages 11 and 26 years found a gene-environment interaction that suggests that individuals with the short arm allele of the serotonin transporter gene develop depression under conditions of stressful life events or traumatic parenting, whereas those with the homozygous long arm of the gene appear protected from developing major depression (+40). While stress may be a causal factor, alone it may not be sufficient to precipitate major depressive episodes.

The gene-environment relationship becomes more complex when theories of neurogenesis come into play. BDNF is necessary for the generation of new neurons in the hippocampus and dentate gyrus in adults (+39, +41). Because of animal models showing a decrease in BDNF with stress and an increase with antidepressants, some researchers have postulated that in human beings impaired neurogenesis may be at the heart of the pathophysiology of depression (+39). Although early studies did not find a significant difference in hippocampal volume between depressed individuals and healthy control subjects (+42, +43), a decrease in BDNF is consistent with more recent neuroimaging data that show decreased hippocampal volume in patients with chronic or recurrent MDD (+44, +45). No actual evidence for decreased neurogenesis in adult humans has been produced, however, nor any fundamental data on the function of new neurons and the clinical correlates of decreased neurogenesis (+46). The pros and cons of these two arguments are elegantly discussed by Sapolsky (+47).

The large pharmaceutical companies all advertise their antidepressants as first-line agents; but "first-line" in this context is a marketing tool, not an evidence-based concept. Instead of conceptualizing one antidepressant or another as first-line, it makes sense to tailor the treatment choices to the individual patient and, as much as clinically appropriate, involve the patient in the decision of which medication to try and which side effects to risk. In this model of shared decision making, the physician and the patient reach concordance about the best course of action (+48, +49). When concordance is reached, patients are more apt to take their medication regularly than when they are merely told what to do by their physician. This model of concordance and shared decision making is new to psychiatry and still evolving in other areas of medicine (+48, +50). Those interested in more detail on these issues might want to read Jay Katz’s classic, The Silent World of Doctor and Patient (+51), and Carl Schneider’s counterargument to Katz, The Practice of Autonomy (+52). Those interested in better communication of risk and negotiating around treatment choices might want to read Gerd Gigerenzer’s excellent book Calculated Risks (+53) and Barry Schwartz’s The Paradox of Choice (+54) as well as Ury and Fisher’s classic on negotiation, Getting to Yes (+55).

In the past few years, the hypothesis that dual-action antidepressants—agents that inhibit reuptake of both serotonin and norepinephrine—are better than single-action antidepressants has been at the center of the efficacy wars. A fundamental part of the terrain of this battleground is the emphasis on remission (becoming completely free of depression) as opposed to response (showing 50% improvement) (+22). In a widely cited analysis of pooled data, Thase and colleagues found a remission rate of 45% among patients who received venlafaxine, a reuptake inhibitor of both norepinephrine and serotonin, compared with 35% for various selective serotonin reuptake inhibitors (SSRIs) and 25% for placebo (+56). In contrast, Montgomery and colleagues reported that escitalopram was superior to venlafaxine (+57). Why the discrepancy? A careful look at the Thase study reveals that the SSRI in six of the eight studies analyzed used fluoxetine, an agent that has a long half-life and possibly a slower onset of action than other SSRIs, so the results of the meta-analysis may not be generalizable to other SSRIs. Also, one of the studies lasted only 6 weeks, and some of the studies may have included SSRI-resistant patients.

It is essential that the treatment choices be tailored to the individual patient, typically on the basis of estimates of tolerability, possible symptom clusters, and prior treatment, and for the clinician and the patient to negotiate about a limited number of antidepressants and to share decision making.

The past two decades have seen the emergence of evidence-based psychotherapies, notably cognitive behavior therapy (CBT) and its variants (+58) and interpersonal therapy (IPT) for depression (+59). While studies with adult patients have generally found that the efficacy of these therapies can equal that of antidepressants, it is important to note that the subjects who agree to participate in clinical trials are those who are willing and able to participate in psychotherapy. Those who are too depressed to do the assignments associated with the psychotherapy obviously would not participate in the trials. Behavioral activation therapies can sometimes be helpful in these instances. Converging evidence supports the practice of combined therapy with antidepressants and targeted psychotherapies. In one key study, the combination of nefazodone and the cognitive behavioral analysis system of psychotherapy (CBASP) was superior to either treatment alone (+60). Similarly, the combination of CBT and fluoxetine was better than either one alone for depression in adolescents (+61). Results of a meta-analysis further support the combination of medication and targeted psychotherapy (+62).

Since the seminal studies by Prien et al. (+63), Keller et al. (+64), and the Pittsburgh group (+65, +66), clinical practice has shifted from time-limited treatment for each depressive episode to long-term maintenance treatment for many patients. Guidelines suggest that all patients receive continuation treatment for 4 to 6 months after response to prevent a relapse. Patients who have had three or more lifetime episodes, those whose index episode lasted 2 years or longer, those with double depression, and those whose first onset of depression occurred at 50 years of age or later should receive indefinite maintenance treatment to prevent a recurrence of depression.

Although indefinite treatment is beneficial for many patients, most stop taking their antidepressants within 6 months (+67, +68). To address this reality, several sequential studies of cognitive behavior therapy have been developed that would help maintain euthymia after discontinuation of antidepressants. Paykel and colleagues found that CBT was effective for residual symptoms and prevented relapse (+69). Giovanni Fava found that well-being therapy, a modification of CBT, also prevented recurrence (+70). Segal and Teasdale developed a group CBT called mindfulness-based cognitive therapy (MBCT) that shows promise in this area and is being assessed for efficacy in a randomized clinical trial (+71, +72).

The Food and Drug Administration (FDA) held extensive hearings in September 2004 on the risk-benefit ratio of antidepressants for children and adolescents (+73). An FDA advisory panel examined data from randomized clinical trials with some 5,000 pediatric subjects and found little evidence of efficacy for antidepressants in this age group and some evidence of increased suicidal thoughts. In response to the FDA hearings, the American Psychiatric Association issued a statement that access to antidepressants should be maintained and that close monitoring is needed for adolescents and children taking antidepressants (+74). The American College of Neuropsychopharmacology formed its own task force to assess pediatric antidepressant use. After reviewing clinical trials, epidemiological studies, and autopsy studies, the task force concluded that SSRIs and other newer antidepressant drugs "do not increase the risk of suicidal thinking or suicide attempts" in youths (+75). There are other possible explanations for increased suicidal thoughts in youths taking antidepressants, such as initial activation of psychomotorically retarded depressives (giving them the energy to act on latent impulses), precipitation of mania or panic, and induction of akathisia. In examining this controversial issue, both efficacy and the side effect of emergent suicidal ideation and behavior need to be addressed.

First, randomized clinical trials of antidepressants for children and adolescents have a high placebo response rate, making it difficult to detect any therapeutic signal of the antidepressants. If one assumes that the antidepressants are ineffective for these patients, then the high placebo response rate confirms that the antidepressants do not work. On the other hand, if one assumes the opposite, that the antidepressants are effective, then the criticism is leveled at the methods used in the trials, which inflate the placebo rate. Generally, however, symptoms improve for most youths in clinical trials whether they are taking the active antidepressant or placebo. Much has been written about the placebo effect in antidepressant trials (+76—+82). The issue is inherently complex, and it may be further complicated in the case of pediatric trials. Rater bias and competency are among the most difficult factors to assess. Baseline depression scale scores may be inflated at the start of the trial to meet inclusion criteria, and scores may be deflated as the trial progresses with the expectation of improvement; as a result, both participants taking antidepressants and those taking placebo show improvement. Furthermore, because of safety concerns and because of volunteer bias, youths who participate in trials are moderately depressed at most, not severely depressed. With these factors combined, the results of randomized clinical trials show a minimal level of effectiveness for antidepressants that may not be generalizable to the population that will be treated (+83). Parker has noted the discrepancy between patients in clinical care and subjects in clinical trials and questioned the relevance of modern clinical trials for actual practice (+84). Parker’s discussion refers to trials in general; the problem is compounded in pediatric studies, because parents decide whether to enroll their children in clinical trials—and parents of severely depressed youths will very likely be reluctant to have their children participate in research.

Second, no suicides occurred during any of the trials, and although the relative risk ratio for active drugs compared with placebo approached 2 (indicating that about twice as many taking active drug had suicidal thoughts or behaviors), the absolute difference was about 1.5% (+73). Suicidal thoughts and suicide attempts were among the events counted, but given inconsistencies in the rules and definitions, it is unclear what was actually assessed. Another consideration is that the randomized clinical trials specifically excluded subjects considered to be at risk of suicide (those who had recently attempted suicide or currently had suicidal ideation). Thus, the trials are not designed to show the efficacy of the antidepressants in reducing suicidal ideation or behavior in those who have these problems at baseline.

It is likely that the analytical challenges of a high placebo response and the lack of any completed suicides in the studies led the FDA advisory panel to recommend that antidepressants carry a warning label indicating that pediatric patients be followed carefully, especially in the early stages of treatment, rather than an outright ban on prescribing antidepressants to youths.

Thus, when using antidepressants to treat a depressed child or adolescent, monitor carefully for the emergence or exacerbation of suicidal thoughts (perhaps with additional measures such as the QIDS-SR or another patient-rated instrument appropriate for youths), see the patient frequently enough to detect any exacerbation, and evaluate carefully for symptoms of a bipolar disorder.

Because only a minority of patients achieve remission with antidepressant monotherapy, many clinicians turn to either augmentation or combination treatments for patients who have a partial response or no response to antidepressants (+85). The sections below provide a brief review of what is new with these strategies.

Technically, augmentation in this context is the addition of a non—antidepressant medication to an antidepressant medication regimen. The oldest and most studied of such agents is lithium (+86—+88). In a meta-analysis, Bauer and colleagues found that the average response rate with the addition of lithium was 45%, compared with 18% with placebo (+86). Several other reports, however, found no significant difference between lithium and placebo augmentation (+88). The only study published to date that has addressed the issue of duration of treatment with lithium augmentation found that patients who responded to and continued taking lithium in addition to their antidepressant did substantially better than those who were given placebo in addition to the antidepressant (+89). Few studies have been published that examine the efficacy of lithium augmentation for SSRIs and other more recently introduced antidepressants.

Thyroid hormone augmentation, specifically with T₃, is the second most widely studied augmentation, but this agent, too, has not been adequately studied for augmentation of the newer antidepressants. The largest of recent studies compared lithium and thyroid augmentation of tricyclic antidepressants and found them about equally effective and better than placebo (+90).

One practice favored by many clinicians and slowly gaining supporting evidence from randomized controlled trials is augmentation of antidepressants with atypical antipsychotic medications for patients with nonpsychotic, treatment-resistant unipolar depression. Olanzapine (+91), risperidone (+92, +93), ziprasidone (+94, +95), aripiprazole (+96), and quetiapine (+97) have been studied and found to be superior to placebo as augmentation of antidepressants for nonpsychotic depression. The long-term risks and benefits of using these medications will become clearer over time. One concern is that atypical antipsychotics have been associated with the metabolic syndrome (+98, +99).

Another agent favored in clinical practice that has garnered some supportive data is the addition of the antinarcolepsy agent modafinil. An initial case series supported the use of modafinil, and subsequent reports have confirmed that it might be useful for patients with residual fatigue and sleepiness (+100—+102). Some data support the use of la-motrigine (+103—+106), pramipexole (+107), folate (+108, +109), and s-adenosylmethionine (+110—+113), but more studies are needed to establish efficacy for these agents.

Combination treatment refers to the use of two different antidepressants together. Although combination treatment is commonly used, few studies have been published. Initial case series have suggested that the combination of SSRIs and bupropion has potential (+114, +115). A controlled trial of the combination of SSRIs and mirtazapine found that mirtazapine was substantially better than placebo (+116).

Switching within the same antidepressant class may be useful because mechanisms of action may differ across agents in the same class (+117). Overall, the probability of response to a second SSRI after nonresponse to a first SSRI is about 50% (+118). Reported response rates include 76% for patients who were switched from fluoxetine to citalopram (+119), 63% for patients switched from sertraline to fluoxetine (+120), and 50% for patients switched from one SSRI to another (+121).

Switching classes of antidepressant is an alternative strategy (+122—+124). In a study of chronically depressed patients treated with imipramine or sertraline in which nonresponders were switched from one agent to the other, Thase and colleagues reported response rates of 44%—60% after the switch (+124). Switching from SSRIs to venlafaxine yielded response rates from 30% to 69% (+123, +125—+127). Switching to mirtazapine, bupropion, tricyclics, and monoamine oxidase inhibitors has also been studied (+128—+135).

None of the augmentation, combination, or switching options cited above, with the exception of lithium and thyroid hormone, have been evaluated in published comparison studies. To address this gap in the literature, the National Institute of Mental Health sponsored the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study (+136, +137), the largest clinical trial ever conducted for depression (www.star-d.org). Patients with MDD who present in either psychiatric or primary care settings are invited to participate and are treated with citalopram (open-label) at maximally tolerated doses for up to 12 weeks. Those who do not remit are randomly assigned to a second treatment through equipoise-stratified randomization (+138). Equipoise-stratified randomization essentially allows the patient and the clinician to decide which group of options makes clinical sense. For example, patients who find that they are unable to tolerate citalopram would not want to opt to be randomized to an augmentation strategy, whereas those who are doing reasonably well with citalopram but fall short of remission would not want to stop citalopram treatment and start something else. Some patients may choose all options or may opt out of having cognitive behavior therapy as either an augmentation or a switch strategy.

The augmentation/combination options if citalopram fails to bring a patient to remission include buspirone, bupropion, or CBT; the switch options include sertraline, bupropion, venlafaxine, or CBT. If these options fail to bring about remission, subsequent levels include augmentation with either lithium or thyroid hormone and switching to either mirtazapine or nortriptyline. Finally, if these options fail, patients are randomized to either tranylcypromine or the combination of venlafaxine and mirtazapine. Publication of results is expected to begin this year.

Depressive disorders, including major, minor, and subsyndromal depression, are widespread. Our understanding of the pathophysiology of these disorders is growing but far from complete. A stress-diathesis model has empirical support, but the story of the relationship between stress, genes, and neurogenesis is just beginning. Clinicians use widely different thresholds for prescribing an antidepressant, but when one is prescribed, the patient needs to be monitored carefully and systematically. Clinicians should consider using depression rating scales such as the QIDS-SR or the BDI-II. Taking the medication regularly is more important than which medication the patient takes. Comorbid conditions, especially anxiety disorders, need to be diagnosed and treated. New paradigms of shared decision making and negotiation can lead to greater concordance and improved outcomes. New psychotherapies are gaining ground, and combined treatment has increasing empirical support. Finally, treating depressed youths has become more complicated because of the controversy surrounding SSRIs and suicidality, and treatment must be delivered with great care.