The rational drug discovery used to produce the selective serotonin reuptake inhibitors (SSRIs) explains why they are similar as a class in so many ways (Table 5.1) as well as why they differ as a class from tricyclic antidepressants (TCAs). As in the preceding chapter, a summary of the similarities among the SSRIs using the STEPS approach will be given rather than an exhaustive review.

Ideally, this chapter would be based on data from studies in which patients were randomly assigned to one of several treatment arms: one for each SSRI at their respective, comparable antidepressant doses and one for a placebo. Ideally, these studies would contain an adequate number of patients randomly assigned to each of these discrete treatment arms to determine whether there are meaningful differences with regard to safety, tolerability and efficacy. Since there are 5 SSRIs, such studies would have at a minimum 6 treatment groups (ie, each SSRI and the placebo arm). Most researchers would probably not want to gamble on doing such a study with only 1 dose per drug because of the criticism that the dose chosen for a given SSRI was not the truly comparable dose for that drug and thus biased the study outcome relative to that drug. Instead, they would have more than 1 dose arm for each SSRI. However, each dose added per drug would multiply the number of groups needed for the study and thus its size and complexity. To adequately power these studies to test for expected differences between these different drugs would require hundreds of patients per group. Given these considerations, it should not be surprising that such studies do not exist. Moreover, such studies will almost undoubtably never be done for quite practical reasons of cost and logistics.

Since the ideal data does not exist for the purposes of this book, the next best approach is taken by relying on summary data from comparable studies of rigorous design. To be included, the data has to come from double-blind, placebo-controlled, adequately powered studies. The presence of a placebo group permits comparisons to be drawn across the studies by comparing the drugs on the basis of placebo-adjusted differences in such outcome variables as adverse effects and efficacy. In some instances, comparable data has not been published for all of the SSRIs and, therefore, some SSRIs cannot be included in the analysis. Those instances are pointed out in the text.

Safety

There are multiple facets to this broad heading, including:

• Therapeutic index
• Long-term safety
• Risk of pharmacodynamically and pharmacokinetically mediated drug-drug interactions

The available data indicate that the SSRIs as a group are remarkably similar in all of these ways with the exception of pharmacokinetically mediated drug-drug interactions. Given the complexity of that topic and the fact that it is a major distinguishing characteristic of these otherwise quite similar drugs, it is discussed at length in Sections 7 and 8.

Since all SSRIs have been designed to avoid affecting fast sodium channels in contrast to TCAs, they all have a wide therapeutic index (ie, the gap between the effective dose and a potentially toxic dose). They do not affect intracardiac conduction.^36,100,165 Patients have survived overdoses of each of the SSRIs that were many times their usually effective antidepressant doses without serious toxicity including:

• No arrhythmias
• No disturbance of blood pressure
• No seizures
• No coma
• No respiratory depression

All of these adverse effects do occur with overdose of TCAs as little as 5 times their therapeutic doses.²²⁵ In most instances of an overdose of only an SSRI, there is no need for medical intervention beyond observation and addressing the reasons for the overdose.^36,100,165

Drug-drug interactions, whether pharmacodynamically or pharmacokinetically mediated, are a safety issue with any drug since polypharmacy is a common clinical practice particularly in the most fragile patients (ie, the elderly and those with multiple medical illnesses). As mentioned above, pharmacokinetically mediated drug-drug interactions with SSRIs will be the subject of a latter section; however, this phenomenon will be briefly mentioned here since there is overdose risk with these drugs when they are taken in combination with other drugs.

Overdoses in the form of a suicide attempt are typically done with more than one drug.^36,100,165 SRI-induced inhibition of a specific cytochrome P450 (CYP) enzyme can affect the toxicity resulting from such an overdose in two ways due to inhibition of the CYP enzyme which mediates the metabolism of the concomitant drug. If the concomitantly ingested drug normally has extensive first pass metabolism dependent on that enzyme, then SSRI-induced inhibition of that enzyme should increase the bioavailability of the other drug and thus increase the toxicity of the overdose. The inhibition of the enzyme should also delay the clearance of the other drugs and thus increase the duration of their toxicity. An increased duration will involve longer care and potentially an increased risk of time-dependent sequelae such as intercurrent infection in an overdose patient with compromised ventilation due to respiratory depression. One example would is an overdose involving a TCA and an SSRI that is capable of causing substantial inhibition of the metabolism of TCAs (eg, fluoxetine- or paroxetine-induced inhibition of CYP 2D6) (refer to Section 8 for details).

There can be many other examples involving overdoses with a wide range of drugs from benzodiazepines to cardiovascular drugs to narcotics. This scenario is based on the known pharmacology of these drugs and pharmacokinetic principles, but to date there are no studies which have tested whether this scenario meaningfully affects clinical outcome in such multiple drug overdoses.

A reduced risk of pharmacodynamically mediated drug-drug interactions is a class advantage of SSRIs over TCAs as discussed in Section 4. Nonetheless, SSRIs can have pharmacodynamically mediated adverse drug-drug interactions, primarily with drugs that also affect serotonin mechanisms of actions (MOAs). Such interactions are a class issue common to all of the SSRIs and a direct result of the MOA they were designed to share (ie, the inhibition of serotonin uptake). The most serious of these adverse interactions is the central serotonin syndrome that can occur when monoamine oxidase inhibitors (MAOIs) are combined with SSRIs.²⁶⁶ Minor variants of this syndrome in terms of the number of symptoms, their severity, and their duration can occur when a variety of serotonin active drugs (eg, lithium, busiprone) are added to SSRIs or MAOIs.

This pharmacodynamically mediated drug interaction results from the combined indirect serotonin agonism caused by both the inhibition of serotonin degradation produced by the MAOI and the inhibition of serotonin uptake produced by the SSRI. Together, these two actions can create a potentially catastrophic dysregulation of a variety of basic brainstem mechanisms, regulated by the central serotonin neural system and produce a syndrome consisting of:

• Hyperthermia
• Diaphoresis
• Gastrointestinal distress
• Mental status changes
• Myoclonus²⁶⁶

The central serotonin syndrome produced by the combined use of SSRIs and MAOIs is particularly dangerous because of its severity and duration due to the persistent action of MAOIs. The duration may also be increased the longer the half-life of the SSRI. In severe cases, this interaction can be fatal.

To avoid this interaction, SSRIs should not be started until 2 weeks after the discontinuation of MAOIs to allow for replenishment of MAO activity. Similarly, MAOIs should not be started until there has been virtual full washout of the SSRIs, which takes 2 weeks for all of the SSRIs except fluoxetine. A minimum of 5 weeks is recommended for fluoxetine if the daily dose was 20 mg/day, and longer if the daily dose was higher, due to the nonlinear pharmacokinetics of fluoxetine (see Section 6).

Tolerability

While the SSRIs as a class differ from the TCAs in terms of avoiding a number of adverse effects that are mediated by the blockade of histamine, acetylcholine, and a₁-adrenergic receptors, as well as fast sodium channels, all SSRIs as a class produce adverse effects that are the result of the MOA they are designed to share (ie, indirect serotonin agonism by inhibiting the serotonin uptake pump).

The adverse effects produced by a drug can be determined in several ways.²¹¹ The most common is by determining what adverse complaints or physiological effects are seen to a statistically significant greater degree in the drug-treated group versus a parallel placebo-treated group. Table 5.2 presents the data on this issue for 4 of the SSRIs: fluvoxamine, fluoxetine, paroxetine and sertraline. The data in this table is from the double-blind, placebo-controlled clinical trial databases with these 4 SSRIs. The presence of the placebo control allowed for a comparison of the placebo-adjusted rates of specific adverse effects for these 4 SSRIs (ie, the incidence rate of each specific SSRI minus the incidence rate that occurred on its parallel, double-blind, placebo control). The reader who is interested in a further discussion of this approach is referred to a recent paper on this subject.²¹¹ Unfortunately, comparable data for citalopram is not available for this analysis. Hence, comparative statements about the rate of specific adverse effects cannot be made about citalopram. Nevertheless, citalopram has been reported to produce the same type of adverse effects now known to be caused by serotonin uptake inhibition.¹⁷³

As can be readily seen in Table 5.2, there are only modest differences in the incidence rates for various adverse effects for these different SSRIs consistent with how these drugs are developed. Since data in this table are from the clinical trials databases with these drugs, it represents the average incidence on the various doses of that SSRI used in its clinical trials. The incidence and severity of many of these adverse effects are dose-dependent as seen in Figure 5.1, which shows the discontinuation rate due to adverse effects as a function of dose for the 3 SSRIs which have published fixed-dose studies.

TABLE 5.2 — Comparison of the Placebo-adjusted Incidence Rate (%) of Frequent Adverse Effects for SSRIs* Item Fluoxetine (n=1730, n=799)1 Fluvoxamine (n=222, n=192)1 Paroxetine (n=421, n=421)1 Sertraline (n=861, n=853)1 Headache 4.8 2.9 0.3 1.3 Nervousness2 10.3 7.6 4.9 4.4 Tremors 5.5 6.1 6.4 8.0 Insomnia 6.7 4.0 7.1 7.6 Drowsiness3 5.9 17.2 14.3 7.5 Fatigue4 5.6 6.2 10.3 2.5 Dizziness/lightheadedness 4.0 1.3 7.8 5.0 Vision disturbances 1.0 0 2.2 2.1 Nausea 11.0 25.6 16.4 14.3 Diarrhea 5.3 – 0.4 4.0 8.4 Dry mouth 3.5 1.8 6.0 7.0 Anorexia 7.2 8.6 4.5 1.2 Dyspepsia 2.1 3.2 0.9 3.2 Frequent micturation 1.6 0.6 2.4 0.8 Constipation 1.2 11.2 5.2 2.1 Sweating 4.6 – 1.3 8.8 5.5 Respiratory5 5.8 – 1.3 0.8 0.8 Palpitations6 – 0.1 NA 1.5 1.9 Urinary retention7 — NA 2.7 0.9 * Data for fluoxetine, paroxetine and sertraline is from reference 211; data for fluvoxamine is from reference 89. Incidence of each respective adverse effect for patients taking each drug minus the incidence for each drug’s parallel placebo condition. 1The first value is the number of patients on that medication, while the second represents those treated in the parallel, placebo group. 2Nervousness is a composite of the following terms: nervousness, anxiety, agitation. 3Includes somnolence, sedation. 4Includes asthenia, myasthenia, hypokinesia. 5Includes respiratory disorder, upper respiratory infection, flu, dyspnea, pharyngitis, sinus congestion, oropharynx disorder. 6Includes tachycardia. 7Includes micturition disorder, difficulty with micturition, and urinary hesitancy. NA = Not available

FIGURE 5.1 — Discontinuation Rate Due to Adverse Events as a Function of Dose for Three SSRIs* References: 1283, 284; 280; 385

FIGURE 5.2 — Antidepressant Efficacy as a Function of Dose for Three SSRIs* References: 1283, 284; 280; 385

TABLE 5.3 — Adverse Events for Each SSRI that Occurred ³ 1% More Often Than With Other SSRIs* Fluoxetine Fluvoxamine Paroxetine Sertraline Nervousness/agitation/anxiety† Respiratory complaints Headache Nausea Drowsiness Constipation Anorexia† Anorexia† Frequent micturition Asthenia/fatigue† Dizziness Sweating Loose stools Tremors Dry mouth * Placebo-adjusted rates using the results of placebo-controlled, double-blind studies for each SSRI respectively. Based on analysis of data in Table 5.2. † Dose-dependent adverse effects of SSRIs which can mimic symptoms of major depression. NOTE: Insomnia is another dose-dependent, adverse effect of SSRIs and also can be a symptom of major depression. Insomnia is not shown above because its incidence as an adverse effect is virtually identical for fluoxetine, paroxetine and sertraline, but is lower for fluvoxamine (Table 5.2).

Given the dose-dependent nature of many of SSRI-mediated adverse effects, a specific SSRI can be at a disadvantage in Figures 5.1 and 5.2 and Table 5.3 if the doses used predominantly in its clinical trials development were higher than its usually effective, therapeutic dose while the other SSRIs were dosed closer to their usually effective dose. In fact, the use of higher than necessary doses was the case for virtually all of the SSRIs, particularly in the early phases of their clinical trials. The reason is that the relatively benign adverse effect profile of the SSRIs in comparison to the TCAs allowed the investigators to titrate to the highest doses permitted in the ascending dose design studies that were typically used in the early studies with these drugs. This dose issue is one of the limitations of this approach in contrast to having data from the ideal study described at the beginning of this chapter. Nonetheless, the data in Table 5.2 is state-of-the-art and fortunately the dosing issue is somewhat mitigated by the fact that overdosage during the clinical trials was virtually universal with these drugs. With this caveat in mind, Table 5.3 lists specific adverse effects that have at least a 1% higher incidence rate on a specific SSRI in comparison to the other 3.

The SSRIs as a class also produce a variety of sexual dysfunction adverse effects, including anorgasmia and decreased libido (Table 5.4). Although an analysis of the clinical trial database for each SSRI suggests that fluvoxamine and fluoxetine are less likely to produce these effects than paroxetine and sertraline, clinical experience suggests that all SSRIs produce a comparable rate at their usually effective, minimum antidepressant dose.^130,280,296 One possible reason for the lower rates in the clinical trial databases for fluvoxamine and fluoxetine is that these two SSRIs were the first to be extensively studied and that unfamiliarity with this adverse effect may have contributed to an under-reporting of its occurrence. Again, the ideal study described at the beginning of this chapter would more convincingly answer this question; in the interim, clinicians will have to assess this matter using the available data and their clinical experience. Clearly, sexual dysfunction is an unintended effect that can apparently be produced by serotonin uptake inhibition.

Of importance, a number of the dose-dependent adverse effects produced by the SSRIs can mimic clinical depression (Tables 5.3 and 5.4). These include:

• Nervousness/agitation/anxiety
• Drowsiness or daytime tiredness
• Anorexia
• Fatigue
• Sexual dysfunction, such as decreased libido

This fact has several implications. First, the emergence of such adverse effects at higher doses may in part account for the fact that the antidepressant response rate to the SSRIs tends to be lower on average at doses higher than the usually effective, minimum dose based on the result of fixed-dose studies (Figure 5.2). Second, physicians may misinterpret the emergence of such adverse effects as the need for higher doses and thus increase the dose unwittingly, worsening the situation. The late emergence of these effects as a result of gradual drug accumulation may be interpreted as the drug losing its effectiveness. This phenomenon is one reason the physician should ensure that the patient has had an adequate trial on the usually effective, minimum dose of each SSRI. For the same reason, the physician may also wish to consider a dose reduction rather than increase if the patient initially appeared to respond and then has a recurrence of symptoms. This phenomenon also is relevant to the potential role of therapeutic drug monitoring (TDM) with SSRIs which is discussed later in this section.

Efficacy - Acute

The major indication for the use of SSRIs is the treatment of major depression. There are 3 types of efficacy that are important in this condition:

• Acute amelioration of the depressive syndrome
• Maintenance of that improvement during the vulnerable period for a relapse
• Prophylactic treatment to prevent the occurrence of a new episode^{92,150,151,233}

There is a variable amount of data for the various SSRIs on the first two uses and no systematic data for any of them on their ability to prevent recurrent episodes for a period greater than 1 year. The comparable data existing for these 2 uses in major depression does not reveal any difference among the SSRIs in terms of either the induction of acute response or maintenance of that response for a period up to 1 year (Table 5.5).

Before reviewing that data, it is important to note that the efficacy of SSRIs is not limited to major depression but extends to several other conditions including obsessive-compulsive disorder, panic disorder, and even conditions such as premature ejaculation.^156,199 The efficacy of these drugs in these other conditions appears to be due to their indirect serotonin agonism via serotonin uptake inhibition and hence is probably a class phenomenon. Different SSRIs have approval for several of these other indications in different countries, others have applications pending for such approval, and still others are in clinical testing in hopes of obtaining data that will support application for formal labeling for these conditions. Fluvoxamine is an example of vagaries of such approvals. It is formally labeled in the US for the treatment of obsessive-compulsive disorder but not for major depression even though it has that indication in many other countries. Since major depression is the principal, clinical use of these drugs and has the largest amount of comparative data, it will be the focus of this discussion of their comparative efficacy.

Two approaches were taken to compare the efficacy of the different SSRIs in terms of the acute relief of depressive symptoms. The first was a meta-analysis of the double-blind, placebo-controlled studies and the second was a comparison of the results of placebo controlled, fixed-dose studies. The latter is critical to determining whether there is a difference among the drugs in terms of their efficacy at the usually effective, minimum dose. Of necessity, these comparisons had to be limited to the SSRIs for which published data existed from such studies.

It was possible to do the meta-analysis with 4 of the 5 SSRIs: fluvoxamine, fluoxetine, paroxetine and sertraline.²²⁶ The results of such studies with citalopram were not published in sufficient detail at the time of this meta-analysis to permit that SSRI to be included. Based on the results of this meta-analysis, each SSRI produces approximately a 60% overall response rate (ie, at least a 50% reduction in symptoms as a result of treatment) and a 30% higher response rate than a parallel, placebo control. This meta-analysis suggests that approximately the same percentage of patients with major depression respond to approximately the same degree to each of these 4 SSRIs. The data that has been published on citalopram suggests its efficacy is comparable.^174,177

Double-blind, placebo-controlled, fixed-dose studies are the only way to determine the optimal dose of a drug and what is a comparably effective dose for different drugs in the same class (ie, having the same MOA mediating that outcome). Such studies have been done with all of the SSRIs, but the results have not been presented yet for the citalopram or fluvoxamine studies. Therefore, this analysis is of necessity limited to the studies with fluoxetine, paroxetine and sertraline (Figure 5.2).

Based on available studies, each of these 3 SSRIs have a flat-dose antidepressant-response curve meaning that they produce approximately the same average response rate at each dose above their usually effective, minimum dose over their clinically relevant dosing range (Figure 5.2). Based on these studies, the usually effective, minimum dose for fluoxetine and paroxetine is 20 mg/day and for sertraline 50 mg/day. Paroxetine was the only 1 of the 3 to include an ineffective dose, 10 mg/day.⁸⁰ In 1 study, fluoxetine (5 mg/day) was as effective as 20 mg/day on the Hamilton Depression Rating Scale but was not effective on other measures. This lack of robust effectiveness and the limited studies with fluoxetine 5 mg/day has been the basis for concluding that 20 mg/day is the optimal dose.⁶ The conclusion that 50 mg/day of sertraline is its optimal dose based on its fixed-dose study is further supported by a review of its other efficacy studies.²²⁹

If anything, the response rate in the fixed-dose clinical trials tends to be lower at higher doses of fluoxetine and paroxetine. As mentioned above, one reason for this result may be the emergence of dose-dependent, adverse effects which mimic symptoms of major depression. Consistent with this explanation, the discontinuation rate due to adverse effects increases at higher doses for all 3 of these SSRIs (Figure 5.1).

Given the delayed onset of antidepressant response seen with SSRIs and other antidepressants, an increase in dropouts at higher doses will also bias against the response rate at higher doses since patients will be prone to discontinuing the trial before they may reasonably be expected to respond. Regardless of the reason, the decrease in the average magnitude of the antidepressant effect for fluoxetine and paroxetine reinforces the recommendation to ensure that the patient has had an optimal trial on the usually effective dose of the SSRIs before attempting a dose increase unless the physician uses a TDM approach to judge the adequacy of the dose in the patient as discussed later in this chapter.

The usually effective dose of fluvoxamine and citalopram have not been convincingly established due to the absence of published data from appropriately designed and executed fixed-dose studies. Nonetheless, a meta-analysis of the placebo-controlled studies with citalopram has been published in 2 separate publications.^174,177 The results of this meta-analysis suggest that citalopram also has a flat-dose antidepressant-response curve in terms of antidepressant efficacy.

Nevertheless, this meta-analysis is not an adequate substitute for publication of the fixed-dose studies. Its interpretation is compromised by the fact that the studies involve different criteria for subject selection, different designs, and different scales. Additionally, it is based on the data from only approximately 60% of the patients who were assigned to the citalopram treatment arms (ie, a selected group of patients who stayed on treatment for at least 4 weeks). A case was made using this meta-analysis that 20 mg/day of citalopram is the usually effective dose in patients who do not have severe and/or melancholic major depression. However, the published results of the 1 fixed-dose study with citalopram showed that 40 mg/day was statistically superior to placebo at weeks 3, 4 and 6 and superior to 20 mg/day at weeks 4 and 6 (p < 0.05), while 20 mg/day was not statistically superior to placebo at any time.¹⁷⁸ In contrast to the meta-analysis, the analysis of this fixed-dose study used the more conservative and more generally accepted "last observation carried forward" approach. Until the results of the other fixed-dose studies with citalopram are adequately presented, the results from the published, fixed-dose study serves as the basis for concluding that 40 mg/day is the usually effective, minimum dose for citalopram as shown in Table 3.7.

Returning to the fixed-dose studies with fluoxetine, paroxetine and sertraline, the magnitude of the antidepressant effect (ie, the absolute reduction in depressive symptoms as quantitated using the Hamilton Depression Rating Scale) was also virtually identical for these 3 SSRIs at their usually effective, minimum dose (Figure 5.2). These findings suggest that each of these drugs treats approximately the same percentage of patients to approximately the same extent at their respective minimally effective doses. Consistent with this finding is the fact that each of these SSRIs at these doses produces approximately a 70% to 80% inhibition of serotonin uptake using the platelet as a surrogate marker for the effect on the central serotonin neurons (Table 3.7). The plasma concentration of each SSRI at comparable antidepressant doses are consistent with the concentrations predicted to be needed to produce this magnitude of uptake inhibition based on the in vitro IC50 data.

Figure 5.3 further illustrates that the major effect on serotonin uptake inhibition with SSRIs occurs at the lower end of their clinically relevant dosing range using data from studies with sertraline as a representative SSRI.²²³ In this study, normal volunteers were randomly assigned to 1 of 4 fixed-doses of sertraline. The plasma levels that they achieved and the degree of serotonin uptake inhibition that occurred on each of the 4 doses was measured. The platelet was used as a surrogate in lieu of measuring the effect of the drug on central serotonin neurons. As is apparent in Figure 5.3, 80% of serotonin uptake inhibition was achieved at plasma levels produced by the 50 mg/day dose. While both the plasma levels of the drug and the degree of serotonin uptake inhibition increased as expected with the higher fixed-doses, the higher levels at 200 mg/day were associated with only an additional 8% increase in serotonin uptake inhibition further reflecting the fact that majority of the effect had already occurred and the plateau portion of the dose-response curve had been reached. The fact that this dose-response curve is so similar to the dose-response curve for the antidepressant efficacy of the drug (ie, that is relatively flat above 50 mg/day) provides an heuristic explanation consistent with the presumed MOA for both effects being the result of the effect of sertraline on the serotonin uptake pump. The relationship illustrated in Figure 5.3 is not unique for sertraline but is simply illustrative of curves that can be drawn for any of the SSRIs which have been so studied. This phenomenon is again consistent with how these drugs were rationally developed.

FIGURE 5.3 — Relationship Between Daily Dose of Sertraline, Mean Plasma Levels of Sertraline, and Mean Reduction in Serotonin Uptake by Platelets After 14 Days of Drug Administration at One of Four Fixed Doses

Reference: 223

All of the above observations raise serious questions about the widespread clinical practice of routinely using higher doses of SSRIs before there has been an adequate trial on their usually effective, minumum dose. The only double-blind, adequately powered study to test whether early nonresponders to an SSRI would benefit from a higher dose randomly assigned patients who had not responded after 3 weeks of treatment with fluoxetine 20 mg/day to either stay on 20 mg/day or be treated with 60 mg/day.⁷⁷ At the end of an additional 5 weeks of treatment, an equal number of patients had responded in both groups and the time course for response was also the same. This study is fully consistent with the flat-dose antidepressant-response curve and indicates that 3 weeks is not a sufficient trial to determine that the usually effective dose of this SSRI is inadequate for a patient. Parenthetically, fluoxetine is not the ideal SSRI for this type of study since the long half-lives of the parent drug and active metabolite may make a delayed response more likely than with the other more intermediate-lived SSRIs. Nonetheless, the data exists for this SSRI and not the others.

The reason to adequately test the antidepressant response to the usually effective, minumum dose is to optimize the tolerability of the drug and to avoid unintended effects on CYP enzymes (see Sections 7 and 8). While the effect on antidepressant response and serotonin uptake inhibition of the SSRIs has on average reached the plateau portion of their dose-response curve, that is not true for their discontinuation rate due to adverse effects (Figure 5.1) nor for the effects of specific SSRIs on specific CYP enzymes (eg, fluoxetine on CYP 2D6). Thus, the trade-off for using higher doses than necessary is to increase the degree of CYP enzyme inhibition produced, possibly the number of CYP enzymes inhibited and number and severity of dose-dependent adverse effects which can mimic major depression. All of these adverse consequences can occur with higher than the usually effective dose without improving antidepressant efficacy.

These observations also raise serious question about the widespread clinical practice of switching from one SSRI to another if the patient fails to respond to the first SSRI after a reasonable therapeutic trial. Conceivably, there is some as yet unidentified difference in the spectrum of antidepressant activity of these drugs, but the available data is not suggestive that a meaningful difference exists. There is certainly no compelling double-blind, placebo-controlled data to support this practice. It is clearly an important deficiency in our knowledge when faced with a patient who has not responded to a member of this class of drugs.

In addition to acute efficacy, fluoxetine, paroxetine and sertraline have been studied in terms of their ability to prevent a recurrence of a depressive episode in the 1-year interval following the induction of an acute response (Table 5.5). The design of these studies is similar so that the results can be reasonably compared: patients were brought into remission on the SSRI and after a period of several months of stabilization were randomly assigned in a double-blind fashion to either remain on the SSRI or be switched to placebo. As with acute efficacy, the results of these studies are amazingly consistent: continued treatment with each SSRI produced an approximately 30% reduction in the risk of relapse compared to the parallel, placebo control in the 1-year period of follow-up observation. A similar study was done with citalopram, but for 6 months rather than 1 year. Like the other 3 SSRIs, citalopram was superior to placebo having a 20.5% lower risk of relapse at that point.¹⁷⁹ That result is comparable to the result with the other 3 SSRIs at the same time point in the longer studies with those drugs. A relapse prevention study has not been published with fluvoxamine. The expectation is that it will be similarly effective.

Therapeutic Drug Monitoring

The discussion about the flat-dose antidepressant-response curve is not meant to imply that there are no patients who will benefit from a dose other than the usually effective, antidepressant dose. The phrase "on average" is key to that discussion. Like the effect on serotonin uptake inhibition (Figure 5.3) and specific CYP enzyme inhibition (Figure 8.3 and 8.4), the antidepressant effect and adverse effects of SSRIs must be concentration-dependent although the "signal-to-noise" problems in such research makes it difficult to show a strong correlation between these clinical effects and plasma drug levels.

Since the plasma levels achieved on the same dose of the same SSRI can differ among patients, there are undoubtably some patients who need a lower dose to achieve the concentration that usually occurs on the usually effective, antidepressant dose and, conversely, some patients need a higher dose to achieve the same levels. The problem is that it is obviously difficult to detect this fact using dose alone.

One approach to this problem is simply careful dose titration based on clinical assessment of response. However, there are several problems with this approach. First, there is no compelling data about what is an adequate duration to know that the usually effective dose is wrong for that specific patient. The only data on this issue indicates that 3 weeks is an inadequate trial for fluoxetine, but not what is an adequate trial for this or any other SSRI. Second, the clinician would have to be able to distinguish between dose-dependent, adverse effects that mimic depressive symptoms and the depressive symptoms themselves. The placebo response rate in clinical trials does not suggest that even experienced clinical investigators can consistently determine when a response is specifically treatment related. For these reasons, TDM can have a role in deciding whether a specific patient needs a dose other than the usually effective, minimum dose.

A few general comments may be helpful to put this discussion in perspective. Unlike TCAs, TDM with SSRIs will almost undoubtably never be a standard-of-care issue. The avoidance of serious toxicity is the overriding rationale for TDM. That is not an issue with the SSRIs in contrast to the TCAs. Due to their narrow therapeutic index, their potentially life-threatening toxicity, and wide interindividual variability in clearance, some patients on conventional antidepressant doses of TCAs can experience serious adverse consequences (eg, delirium, seizures, arrhythmias).²²¹ The avoidance of such toxicity primarily in patients who are deficient in CYP 2D6 function either due to genetics or concomitantly prescribed drugs (eg, fluoxetine, paroxetine) is the reason for obtaining a plasma level once early in the course of treatment with a TCA to permit a rational dose adjustment if needed.

A second and less pressing reason to use TDM is to increase efficacy and tolerability by adjusting the dose upward in patients who have rapid clearance to increase efficacy and downward in those with slow clearance to improve tolerability and thus efficacy.²²¹ This reason would be applicable to drugs like the SSRIs which have a sufficiently wide therapeutic index such that serious toxicity is not a concern, but nevertheless can produce a greater increase in adverse effects than an increase in efficacy on average at higher doses (Figures 5.1 and 5.2), especially since some of those adverse effects can be confused with lack or loss of efficacy (Table 5.3).

The problem is establishing the plasma levels of the various SSRIs that will produce the optimal balance of antidepressant efficacy and tolerability. One approach that has been used is to try to correlate plasma levels of the various SSRIs with antidepressant response in clinical trials. Such studies have been published with all of the SSRIs and have all failed to show a relationship between drug concentration and antidepressant response: citalopram,^31,79 fluoxetine,^232,142 fluvoxamine,^37,140 paroxetine,^158,269 and sertraline.²²³ That is not surprising for many reasons including the "signal-to-noise" problem created by placebo response and the fact that these studies used doses at or above the usually effective, minimum dose. In other words, they examined the plateau portion of the dose-response curve where one would predict that there would be no relationship between concentration and response.

The concentration-dependent nature of antidepressant response with the SSRIs is apparent from the flat-dose antidepressant-response relationship and the fact that it parallels the dose- and concentration-dependent effect on serotonin uptake inhibition (Figure 5.3). This relationship is consistent with a minimum threshold concentration that produces approximately 70% serotonin uptake inhibition and exerts a step-function effect on antidepressant response. However, a study demonstrating such a relationship will have to focus on doses which produce concentrations both below this level and somewhat above, in contrast to the studies that have been done. Such a study will also have to involve hundreds of patients to overcome the "noise" in current antidepressant clinical trials research. The interested reader is referred to a previous paper that discusses the technical problems of doing such research in psychiatry.²³⁰

The ideal TDM study with the SSRIs described above will most likely never be done for cost and logistical reasons. There is no compelling incentive on the part of any government or private research sponsor to demonstrate the concentration-dependent nature of antidepressant response to SSRIs. In fact, the manufacturer may actually have a disincentive out of concern that the demonstration of such a relationship might be used to suggest that TDM is necessary with their medication and thus create a perception that may adversely impact the clinical acceptance of the drug.

FIGURE 5.4 — Estimated Minimum Effective Drug Concentration: Average Plasma Drug Concentration Achieved in the Group Treated With the Minimum, Effective Dose

In the absence of data from such an ideal TDM study, the fixed-dose studies can be used to estimate the optimal plasma level range as illustrated in Figure 5.4. Fixed-dose studies take advantage of group difference to establish the superiority of the drug over placebo. As illustrated in this figure, some patients will do as well on placebo as the best patient does on the drug at any dose, and some patients on the drug will do as poorly as any patient on placebo. In other words, there is complete overlap in the range of the response between the two groups. That is what is meant by the term, "noise," in these studies. The ability to show a difference between drug and placebo is based on the fact that the curve for the drug-treated group is skewed to the responder side, whereas the curve of the placebo-treated group is skewed to the nonresponder side to a statistically significant extent The usually effective, minimum dose is the lowest dose that separates the drug from placebo. In the case of the SSRIs, it is also the best dose for most patients (Figure 5.2).

The usually effective dose defines an expected concentration range (Figure 5.4) based on the variability of the clearance of the drug in different patients. Since this dose on average separates drug treatment from placebo, the concentration that on average is achieved by this dose must also separate the drug from placebo. This concentration then is an estimate of the minimum desired concentration. Since higher doses on average increase adverse effects more than efficacy, the average plasma levels produced by those doses define the upper threshold where generally nuisance adverse effects outweigh additional therapeutic benefit. It would be ideal to have a low dose in the fixed-dose study that does not separate the drug from placebo because the concentration achieved by that dose will more convincingly establish the minimum threshold concentration. Such data exists for paroxetine (10 mg/day)⁸⁰ and possibly for citalopram (£ 20 mg/day).¹⁷⁸ If that has not been established, then one does not know whether a lower threshold will actually be better on average in terms of the balance between efficacy and tolerability.

This information can help the clinician in the decision-making process. A patient may have failed on the usually effective, minimum dose because s/he is not responsive to this MOA or because plasma concentration of the drug is below the usually effective threshold due to an unusually rapid clearance of the drug in that patient. TDM can be used to provide information relevant to the latter possibility. If TDM reveals substantially lower plasma drug levels than would normally be expected for patients on this dose, the problem may be noncompliance or rapid clearance. In the former case, the physician could visit with the patient and determine what is leading to compromised compliance. In the event of rapid clearance, a trial of a higher dose will be more rational than switching to another SSRI or to another class of antidepressants. The goal will be to ensure that each patient receives a trial of a dose which will result in that patient achieving plasma drug levels that are usually therapeutic for that drug. If the patient has a level considerably above the range produced by the usually effective, minimum dose due to unusually slow clearance, then a dose reduction may be warranted to determine whether the less than optimal response is due to adverse effects that are mimicking depressive symptomatology.

The goal of this TDM approach is to aid rather than to dictate the dose adjustment. Such TDM data would have to be considered by the physician along with other factors specific to that patient such as clinical assessment of efficacy and tolerability. As stated above, TDM is clearly not a necessity with SSRIs. Their wide therapeutic index means that clinicians can titrate the dose over large ranges without concern about serious toxicity. Some physicians will, therefore, never use TDM with these drugs. Others may use this approach in nonresponders. A few might even wonder whether this TDM approach might have sufficient clinical and cost-effectiveness advantages to use it early in treatment as a routine aspect of treatment. The goal in this latter approach would be to reduce the time needed to determine whether the usually effective, minimum dose in that patient is producing the usually expected plasma levels in the patient. The potential advantages will be reducing the duration of the illness by being able to more rapidly make dose adjustments and by not erroneously concluding that the patient is not responsive to the drug when the problem is unusually rapid or slow clearance of the drug.