Selective Serotonin Reuptake Inhibitors in the Acute Treatment of Depression

Stuart A. Montgomery

The impetus for the development of the selective serotonin reuptake inhibitors (SSRI) was the perceived need for antidepressants with an improved therapeutic profile compared with the traditional tricyclic antidepressants (TCA). The TCAs made a great contribution to giving depressed patients a better quality of life when they were introduced more than a generation ago, but they have many limitations (see Mood Disorders Linked to the Reproductive Cycle in Women).

The TCAs produce a wide range of pharmacological actions affecting a variety of neurotransmitters; many of these pharmacological effects produce unwanted and sometimes dangerous side effects without being of apparent therapeutic benefit for the depression. The need for drugs with an improved safety profile compared with the older TCAs was recognized early. The anticholinergic effects of the TCAs, for example, lead to worsening in concentration and memory and it has been suggested may worsen dementia. The TCAs have marked effects on cardiac function and can be cardiotoxic in therapeutic dosage as well as overdose (28). This limits their usefulness particularly in the elderly, who are at increased risk of undetected impaired cardiac function. They are toxic in overdose (12, 49), and, in prescribing them a doctor may be providing patients with the means to harm themselves.

The older TCAs are poorly tolerated by patients, which contributes considerably to the therapeutic failure of these antidepressants due to poor compliance (33). Their well-known anticholinergic effects, such as dry mouth and blurred vision, are often cited by patients as the reason for withdrawing from treatment. The difficulty experienced in tolerating the unwanted side effects also prejudices the use of therapeutic doses. Even if treatment is initiated at low doses and increased slowly, it is sometimes difficult to reach the full therapeutic dose.

The need for better-tolerated, safer antidepressants than the TCAs is clear. It was hoped that the SSRIs, which act selectively on the serotonergic system and avoid activity on receptor systems that appear unrelated to antidepressant effect, would not be associated with the unwelcome side effects characteristic of the TCAs. This selectivity of pharmacological action was also expected to result in safer drugs particularly in cardiotoxicity and overdose toxicity.

Another reason for developing the SSRIs was the possibility that improved efficacy could be achieved by focusing the pharmacological activity on a specific neurotransmitter system. There is a considerable body of evidence to indicate the importance of serotonin in the pathophysiology of depression and to suggest the serotonergic pathway as the most closely involved with mood regulation. An antidepressant acting primarily on serotonin might therefore lead to an improved therapeutic effect. A second possibility might be that the SSRIs would be of particular benefit in a particular subgroup of depression (64). Asberg et al. (5) suggested that there was a bimodal distribution of 5-hydroxyindoleacetic acid (5-HIAA) in the cerebrospinal fluid (CSF) of depressed patients, which was a marker of a subgroup. Therefore, SSRIs might offer treatment for this subgroup of patients characterized by a dysfunction in serotonin metabolism. It appears, however, that SSRIs have similar efficacy to reference antidepressants and that the evidence for serotonin-specific depressions is weak.

The SSRIs were developed in response to the need for drugs that were less "pharmacologically dirty" than the TCAs. The selectively acting antidepressants that are the fruit of that development have provided a pharmacological tool for investigating the clinical effects of activity on specific neuronal systems as a means of unraveling the underlying mechanisms of antidepressant action.

The list of SSRIs that have reached the market or are in development is considerable. Several have been in clinical use for some years. This chapter discusses those SSRIs that have currently been licensed as antidepressants in the United States, Canada, or Europe, as well as their efficacy, side effects, and safety in the acute treatment of depression. The SSRIs currently available are citalopram (available in Europe since 1988), fluoxetine, fluvoxamine (available in Europe since 1983), paroxetine, and sertraline. Standard doses of these SSRIs are shown in Table 1.

The standards for assessing the efficacy of new antidepressants are more stringent than was formerly the case and as a result there is evidence from placebo-controlled studies on which to base a judgment. There are also studies that included both placebo and a reference comparator antidepressant, and these provide reassurance about the population of patients investigated and the size of effect observed.

Table 2, Table 3, Table 4, and Table 5, summarize the efficacy studies of the different SSRIs. In some cases individual centers participating in multicenter studies have published separately. To avoid citing the same data twice these have been excluded.

Citalopram was developed initially using reference-controlled studies only. These have not all been published, but the antidepressant efficacy was suggested by Bech and Cialdella (9, see also refs. 3 and 59). A small, as yet unpublished, placebo-controlled study also suggested efficacy compared with placebo. This was later confirmed in two large multicenter placebo-controlled studies (see Table 2). In the 142 patient study of Mendels et al. (39), which has not yet been fully published, citalopram in a dose of 20 mg to 80 mg was better than placebo. In the study of Montgomery et al. (50), which included 199 patients, 40 mg of citalopram was clearly effective compared with placebo but 20 mg was not.

Fluoxetine was the first SSRI to appear in the United States, although it was a latecomer in Europe. Its antidepressant efficacy is shown in its superiority compared with placebo demonstrated in three large studies and in three out of four smaller studies (25, 29, 52, 56, 60, 67, 68). The efficacy assessment was based on 834 patients treated with fluoxetine compared with 397 treated with placebo and 227 treated with TCAs (see Table 3). The number of patients included in the large multicenter studies was sufficiently large to justify comparison between response to fluoxetine and to the comparator, imipramine. The results suggest that fluoxetine has a very similar level of efficacy to the TCAs (41).

Fluvoxamine is the longest available SSRI in Europe. The efficacy of fluvoxamine has been investigated in small placebo-controlled studies, some of which were positive and others negative (16, 26, 31, 36, 37, 58). Confirmation of efficacy was provided by a collective analysis of the eight centers participating in two large multicenter placebo and reference-controlled studies (2). Four of the centers published their results separately (20, 34, 53). In the main analysis and in other reference-controlled studies, fluvoxamine was found to have very similar levels of efficacy to reference TCAs (see Table 4).

Two of the studies and two of the centers in the multicenter studies did not find a difference from placebo for either fluvoxamine or the active comparator (31, 34, 36, 53).

A number of small studies conducted in Europe indicated the probable efficacy of paroxetine, which was later confirmed in two large studies, one placebo controlled and the other placebo and reference antidepressant controlled (see Table 5). Four studies following identical protocols compared paroxetine with placebo in 273 patients and three were positive (13). Six studies compared paroxetine with placebo and an active control, imipramine, in 726 patients and five of these were positive (22).

In the direct comparison with placebo in the merged four-study analysis a significantly better response was found in the group treated with paroxetine at 2 weeks. In the larger six-study comparison this apparently early onset of effect was also seen at 1 week. The response in the imipramine and paroxetine groups was seen to be very similar, and a quantitative difference large enough to be considered clinically important did not emerge until 3 to 4 weeks.

Sertraline has been shown to be effective as an antidepressant in two large multicenter placebo- or placebo- and reference-controlled studies (24, 55) (Table 6). In the large placebo-controlled study (24), which included 369 patients, sertraline showed unequivocal efficacy. This study compared three fixed doses of sertraline between 50 mg and 150 mg. All these doses were seen to be effective, although there was some indication of a better response with the higher dose. In the reference- and placebo-controlled study, which included 448 patients, sertraline and amitriptyline were both shown to be significantly better than placebo with rather similar antidepressant response on the active drugs.

The SSRIs have been extensively compared with reference antidepressants, both in the placebo-controlled studies that included a reference comparator and in stand-alone active comparator studies of varying size. Comparator antidepressants have included amitriptyline, imipramine, clomipramine, maprotiline, oxaprotiline, mianserin, and dothiepin; these studies have generally found no significant differences between treatment groups.

Relatively small active control studies can only provide support for the concept of equivalent efficacy because a very large number of patients is needed to indicate similar efficacy. In general, these studies have reported similar efficacy with occasional studies reporting either increased efficacy with the SSRI (26, 52) or less efficacy (3). However the multicenter studies that have been conducted with the SSRIs during the clinical development program have been large enough to demonstrate that the SSRIs are of the same order of efficacy as the reference TCAs.

Another way of assessing the relative efficacy of antidepressants is to carry out metaanalyses of the databases. This has been done for some of the SSRIs; for example, fluoxetine (54) and paroxetine (22) appear to have very similar efficacy to the TCAs. This judgment is supported by the metaanalysis of the published literature on all of the SSRIs taken together (47). In this analysis, the SSRIs and TCAs were associated with the same rate of drop outs because of lack of efficacy, but there were significantly fewer dropouts because of side effects with the SSRIs.

Direct comparisons between SSRIs are generally lacking, so that we cannot state categorically that one is more effective than another. The results of the efficacy studies and the similar size of therapeutic effect indicate that there are unlikely to be clinically significant differences between them in efficacy.

The selection of the correct dose for antidepressants has often been based on a combination of guess work, following observation of the effects of drugs in animals, and dose titration studies. Neither provide an adequate basis for selecting the optimum therapeutic dose. In dose titration studies, it is customary to increase the dose to the maximum tolerated at which a response is observed. Because of the delay in antidepressant response, this approach can lead to the dose being raised beyond necessary; the response to a low dose may be attributed to the later higher dose. Patients may thus be exposed unnecessarily to high doses of antidepressants, which are often accompanied by increased side effects, without any advantage in therapeutic efficacy. The more reliable way of establishing the optimum dose is via fixed dose studies that directly compare response to a range of doses given in a fixed dosage regimen. Without this type of study to establish the minimum effective dose, a drug may be introduced at too high a dose.

Fixed-dose studies with several of the SSRIs have shown that the earlier studies in the development program were carried out using too high a dose. The initial studies of fluoxetine, for example, were carried out using an 80-mg/day maximum dose. Later fixed-dose studies showed that 20 mg and 40 mg were significantly better than placebo and that the 60 mg dose although effective was associated with increased side effects and an increased number of withdrawals because of side effects (68). In a later fixed-dose study, 5 mg, 20 mg, and 40 mg of fluoxetine were found to be significantly better than placebo although a larger treatment effect was seen with the 20 mg and 40 mg dose (67). The minimum effective dose of fluoxetine has not therefore been established, and the dose of 20 mg may be more than is necessary in some patients.

It is possible that fluvoxamine was introduced at too high a dose and that the side effect of nausea could have been reduced without a loss of efficacy if a lower dose had been recommended. It has become more routine in recent years to carry out fixed-dose studies of new antidepressants, and as a result data are available for most of the SSRIs to indicate the minimum effective dose. Thus the minimum effective dose of paroxetine was established in fixed dose studies as 20 mg (32). There is some suggestion that a higher dose may be more effective in patients with melancholia. The minimum effective dose of citalopram is thought to lie between 20 mg and 40 mg. The 20-mg dose has been found to be effective in a general practice population, but in a hospital-based population 40 mg was found to be more effective (50).

Some patients may need higher doses than the generally optimum therapeutic dose, and here the SSRIs, which are safer and more tolerable than the TCAs, offer an advantage. A study in which the dose was raised from 20 mg to 60 mg of fluoxetine in nonresponders at 3 weeks showed no advantage compared to patients continued on 20 mg. The evidence with paroxetine and citalopram, however, does suggest that raising the dose may be helpful for nonresponders, particularly those with melancholia symptoms or severe depression. Higher doses may also benefit patients with both depression and concomitant obsessive–compulsive disorder.

The SSRIs have been studied in both acute and long-term treatment; the long-term treatment studies have found that full antidepressant doses are needed for prophylaxis or maintenance treatment. Fluoxetine has been found to be effective in reducing the risk of new episodes of depression in doses of 20 mg (57) and 40 mg (45). The dose shown to be effective in both the continuation and prophylactic phase of treatment was approximately 100 mg of sertraline (21) and 20 to 30 mg of paroxetine (46). Citalopram was found to be effective at doses of both 20 and 40 mg in a 6-month continuation treatment study (51).

The overview of the efficacy of SSRIs is that they are as effective as the TCAs with which they have been compared. Their speed of onset of action of general antidepressant effect also appears for the most part to be similar. There are however certain aspects of efficacy that seem to differentiate the SSRIs from the TCAs.

It is a difficult task to establish efficacy in subgroups of depression, as the clinical trials are often too small in size to permit a legitimate analysis. However the databases with the new SSRIs are sufficiently large to enable metaanalyses to be carried out to investigate differential effects in different subgroups.

Efficacy in severe depression is an important aspect of treatment, and when new drugs are introduced their efficacy in this respect is often challenged. Because severe depression is not usually studied separately, a metaanalysis of the database can provide answers. Several of these have been carried out to see if the SSRIs are as effective as the reference TCAs in severe depression. The approach has been to stratify patients included in studies into moderate and severe depression, defined by a cutoff score on a severity rating scale at the start of treatment. The answer provided by this type of metaanalysis has been that the SSRIs are as effective as the reference comparator in both moderate and severe depression; for example, such a metaanalysis has been carried out for fluoxetine (41).

There is also evidence to suggest that in some cases the SSRIs are more effective in severe depression than the comparator TCA. An example is the analysis of the placebo- and reference-TCA-controlled studies with paroxetine, which divided patients into severe [baseline Hamilton Depression (HAMD) score {ewc MVIMG, MVIMAGE,!greateq.bmp} 28] and moderate depression (baseline HAMD score < 28). In the moderate depression group, there was little difference between the responses on paroxetine and imipramine, both of which appeared to have rather similar levels of efficacy in moderate depression. In the severe depression group, however, patients responded significantly better to paroxetine than imipramine, which was not significantly different from placebo (42).

Although the results of metaanalyses must be viewed with caution, this was not an isolated result. A similar metaanalysis of the fluvoxamine database also found an advantage for the SSRI compared with imipramine in severe depression (40).

A characteristic side effect of the SSRIs as a class is a transitory increase in nervousness or anxiety early in treatment in some patients. This quality might be expected to be prejudicial in treating depressed patients with prominent anxiety symptoms. Furthermore, it has often been thought that a sedative antidepressant was necessary for the treatment of anxiety in depression, and SSRIs, which lack the sedative effects of the TCAs, might therefore appear to be at a disadvantage. The metaanalyses that have been carried out indicate that, on the contrary, the SSRIs have a differential advantage in treating anxiety within depression.

An advantage compared with a reference TCA in treating the anxiety component of depression was reported in the early small comparator study of zimelidine (48), an early SSRI briefly available in Europe. This could have been a chance finding because of the small size of the study. However metaanalyses of the databases of three SSRIs (fluvoxamine, fluoxetine, and paroxetine) have all reported an advantage for the SSRI. The advantage is reflected in an earlier response of the anxiety symptoms measured on the Hamilton Rating Scale for depression (23, 41, 43, 66) and the Covi anxiety scale (43).

The response to SSRIs has also been investigated in metaanalyses of the databases in the subgroup of patients with more severe anxiety symptoms at the start of treatment. For example, a significant advantage was reported for fluoxetine compared with the comparator TCA in 786 patients identified as suffering from moderate or severe levels of agitation at entry to the study (41).

These results are all from retrospective analyses and have to be treated with caution. Nevertheless efficacy on anxiety symptoms accords with the positive results seen in the treatment of anxiety disorders such as panic disorder with SSRIs. The differential effects are interesting because they suggest that sedative antidepressants are not the automatic treatment of choice for the amelioration of anxiety symptoms in depression.

Suicidal thoughts are present in a very high proportion of depressed patients. In studies of clinical efficacy the frequency of suicidal thoughts is high in spite of the attempts that are made to exclude these patients. It has been hypothesized that serotonin reuptake inhibitors might have a beneficial effect on suicidality and small early studies reported a differential advantage for SSRIs compared with comparator antidepressants early in treatment (48, 52). Both paroxetine and fluvoxamine were found to reduce the number of suicidal thoughts more than comparator antidepressants in the metaanalyses of the databases (42, 43). The response of patients with frequent suicidal thoughts at the start of treatment was also reported to be less frequent during treatment with fluvoxamine than with imipramine (66).

There have been a small number of anecdotal reports of a provocation or worsening of suicidal thoughts attributed by the authors to treatment with the SSRI fluoxetine (18, 38, 63 ). These were open observations in patients in whom suicidal thoughts did not occur de novo, many of whom were receiving concomitant medication; therefore, attribution to a particular treatment is hazardous. Suicidal thoughts are integral to depressive illness and fluctuate during the course of an episode. There is a natural increase associated with the illness, so that, if the depression worsens, so may the suicidal thoughts. This increase has been reported during treatment with placebo in the metaanalysis of the database of fluvoxamine and of paroxetine (42, 43).

The metaanalyses of the databases of the SSRIs have been scrutinized for the possibility that these drugs might provoke suicidal thoughts and the evidence clearly suggests that, rather than provoking, the SSRIs protect against the emergence of suicidal thoughts. The number of patients who at the start of the study had zero or very low scores on suicidal thoughts items of rating scales and who subsequently developed high suicidal thoughts scores is seen to be significantly higher on placebo than on the active antidepressants in these metaanalyses (8, 42, 43).

Clinically the SSRIs discussed here behave in very similar ways and are effective in the same range of patients. They do, however, vary to some extent both in their selectivity for 5-HT receptors and in their potency. There are also some differences in the pharmacokinetics and pharmacodynamics between compounds, and these may have a bearing on clinical practice. In any discussion of relative selectivity, potency, or pharmacokinetics the metabolites of the drugs have also to be taken into account. For example clomipramine, which is a potent 5-HT reuptake inhibitor, is not included with the SSRIs because it has an active desmethyl metabolite that is a potent norepinephrine reuptake inhibitor.

Citalopram, fluoxetine, and sertraline are converted to their active metabolites which are potent and selective for serotonin. Paroxetine and fluvoxamine are transformed to inactive metabolites. The activity of the active metabolites of SSRIs on neurotransmitter reuptake varies from little in therapeutic doses, and probably not clinically relevant in the case of sertraline, to being more selective and potent than the parent drug, for example fluoxetine. Norfluoxetine is three times more selective than fluoxetine and is probably more important than fluoxetine for clinical efficacy.

The drugs discussed here are rapidly absorbed and distributed with a time-to-peak plasma concentration of 2 to 8 hr. Clearance times vary widely. The plasma half-life of citalopram is approximately 36 hr, fluvoxamine 15 hr, paroxetine 20 hr, and sertraline and norsertraline 25 hr and 66 hr, respectively. The longest is with fluoxetine, which, although the parent drug has a half-life of approximately 1 to 3 days, remains in the blood for very long periods because of the long 7- to 15-day half-life of the active metabolite, norfluoxetine. The long half-life of fluoxetine and its metabolite imposes cautious management when treatment is to be changed, because the drug lingers in the blood for some time after the treatment has been discontinued, thereby increasing the risk of drug interactions. The persistence of the drug has some advantages for patients with irregular compliance.

There is a wide interindividual variation in steady-state plasma levels achieved with SSRIs. These drugs do not appear to be associated with a narrow therapeutic window, although there have been reports of poorer response associated with high plasma concentrations with some drugs (1, 48).

Combinations of antidepressants are sometimes used, although the benefit to be derived from polypharmacy has not been demonstrated in blinded studies and is in any case likely to be outweighed by the increased risks of the combined side effects. Drug interactions have been reported for a variety of drugs. For example, a large variety of commonly used antiarrhythmics, beta blockers, neuroleptics, and TCAs, such as amitriptyline, imipramine, clomipramine, desipramine, and nortriptyline, are metabolized by the same cytochrome P450-P2-D6 system. Interactions are therefore expected and caution is needed with concomitant therapy. The SSRIs fluoxetine, citalopram, paroxetine, and sertraline are also metabolized by the same system and the same need for caution applies in varying degrees. It is not surprising therefore that interactions have been reported between TCAs and SSRIs leading to increased drug plasma levels (4, 7, 65). This matters less with the SSRIs, which have a wider safety margin, but may be important for the TCAs, which are often used at doses close to their toxic limit. The situation is complicated by the genetic polymorphism that exists for these enzyme systems and allows some 7% of Caucasians, for example, to be poor metabolizers of these drugs. Inhibition of their own enzyme system by some neuroleptics, antiarrythmics, and SSRIs suggests that clinicians should become more cautious in the concomitant use of a wide variety of drugs and more circumspect in the use of drugs with a narrow therapeutic range or a poor safety margin. The potency of the SSRIs as inhibitors of the metabolism of the P450-P2-D6 varies and is reported in descending order of potency as paroxetine, fluoxetine, sertraline, citalopram, and fluvoxamine (17).

The combination of SSRIs with MAO inhibitors can lead to the appearance of the serotonin syndrome, which is potentially lethal. Rapid death from hyperthermia following combination of SSRI and irreversible MAO inhibitors have been reported (14) and the newer reversible MAO inhibitors may not be entirely free of risk (61). There should be an interval between treatments with an MAO inhibitor and SSRIs, which should be of a length to take the pharmacokinetics of the SSRI into account.

The SSRIs have been shown to have significantly less anticholinergic side effects compared with the TCAs and this is considered to be an important explanation for their improved tolerability. They are associated with characteristic side effects expected from their pharmacological actions.

The main side effects, which are generally mild and short-lived, are gastrointestinal symptoms and male sexual difficulties. The gastrointestinal symptoms are generally found to be mild and to reduce with continued treatment. It is important to base any assessment of the relative level of side effects on blinded comparisons with placebo in order to obtain an unbiased estimate. For example, 4% of those given paroxetine and 3% given a placebo withdrew because of nausea, which is the most common side effect (32). The gastrointestinal symptoms appear to be dose related, so that too high a dose is likely to cause nausea. Significant weight loss, mainly in those who are overweight, is seen, especially with higher doses of SSRIs. Sexual difficulties in men seems to be a consistent effect, with some two to three times more men reporting abnormal ejaculation than with active controls.

Although the SSRIs as a group are seen to improve the anxiety symptoms of depression somewhat more than comparator TCAs, a few patients experience an increase in anxiety symptoms or agitation early in treatment. These appear to have been reported more frequently with fluoxetine, which may relate to the 5-HT1C agonist properties of fluoxetine (69).

There are also reports of movement disorders with the SSRIs, which is perhaps not surprising in view of the close role that serotonin plays in dopamine autoregulation. Extrapyramidal symptoms are a class effect, rather than relating to one or other SSRI in particular. There have been a number of reports of movement disorders with fluoxetine, although a causal relationship has not been established (6, 10, 11, 35, 62). The number of spontaneous reports to the Committee on Safety of Medicines in the United Kingdom of this type of reaction appeared to be higher with paroxetine (15), but prescription-event monitoring studies suggest that the rate is low and that paroxetine does not cause a greater frequency of extrapyramidal reactions than other SSRIs (30).

The level of convulsions reported with the TCAs is high; a figure of 0.5% is frequently cited for the doses of TCAs used in the comparator trials. At higher doses, convulsion rates increase substantially and rates between 1% and 2% have been reported (19, 27). The convulsion rate seen at doses over 200 mg of TCAs is of concern. The SSRIs as a group appear to have fewer reports of convulsions. Paroxetine, fluvoxamine, and sertraline have been reported to have significantly fewer reports of convulsions than the TCAs.

The potential for antidepressants to precipitate manic episodes in bipolar patients is widely accepted, and the use of TCAs, where the risk is elevated, is contraindicated. It has been postulated that manic switches may be associated with altered levels of 3-methoxy-4-hydroxyphenylglycol (MHPG), the metabolite of norepinephrine, it seems possible that SSRIs might have a safer profile. There is a suggestion from the controlled studies that some of the SSRIs may be associated with fewer cases of switches to mania in bipolar patients. For example, the database on paroxetine reveals that the rate of switches on paroxetine was 2.2% compared with 11.6% with comparator antidepressants (44). However this type of comparison has to be based on relatively small numbers, as bipolar patients are generally excluded from the early stages of a clinical trial program. This type of effect needs closer examination and needs to be confirmed in prospective studies.

The TCAs are associated with a high level of completed suicides attributed to overdose with the drug per million prescriptions (12). This has been attributed to the cardiotoxicity due to the ionotrophic effect of the TCAs on the myocardium. By contrast, there have been very few reports of suicide with SSRIs. In the United Kingdom, for example, the rates of suicides per million prescriptions is 34 with the TCAs compared with 1.7 with the SSRIs (Henry, personal communication). The SSRIs appear therefore to be relatively safe in overdose.

The level of side effects reported in placebo-controlled studies is consistently low with the SSRIs at the lower effective dose with few effects occurring with greater frequency than placebo. This contributes to the tolerability of these antidepressants and results in fewer side-effect–attributed discontinuations than is usual with the TCAs (47).

The SSRIs have now established themselves as a welcome new class of antidepressants. They are clearly effective with a somewhat wider spectrum of action and selective efficacy in obsessive–compulsive disorder and its concomitant depression, where the traditional TCAs are apparently ineffective. The major clinical advantage of the SSRIs lies in their improved side-effect profile and better tolerance, which is reflected in the better compliance seen even in the controlled studies. They appear safer than the TCAs on cardiotoxic measures and are associated with fewer deaths from overdosage. Increasingly clinicians are recognizing the importance of using these safer well-tolerated antidepressants as first line treatment.

published 2000