The separate successes of psychotherapy and psychopharmacology have led to the hope that these treatment modalities can be combined for optimal treatment outcome. However, examination of the treatment-outcome literature indicates that, depending on the disorder, the severity or chronicity of symptoms, and the phase of treatment, combination treatment may or may not provide clear benefits over that provided by either modality alone and may substantially increase the costs of care.

Discussions of the efficacy of combined pharmacologic and psychosocial treatments are aided by delineation of the nature of each treatment modality. For pharmacotherapy, the quality of the therapeutic relationship, instruction, and support offered in conjunction with medication treatment may differ widely among clinicians in any one setting and may differ further depending on the setting and the type of provider (e.g., primary care physicians vs. the psychiatric specialist). For example, variable results between trials of the same agent (Greenberg, Bornstein, Greenberg, & Fisher, 1992; Hooper & Amsterdam, 1998) indicate the dependence of pharmacotherapy outcomes on the context of medication delivery. Indeed, for issues of medication adherence and adequacy of treatment alone, there is evidence for greater adequacy of treatment when antidepressant medication is prescribed by a psychiatrist as compared to a primary care physician (Weilburg, O’Leary, Meigs, Hennen, & Stafford, 2003).

Pharmacotherapy that involves a clear model of treatment, expectations of success, supportive interventions, and close follow-up presumably provides a valid comparison condition by which to judge whether an independent psychosocial intervention has additive value. Fortunately, many of these conditions are met in standard pharmacotherapy trials. Randomized, double-blind treatment trials help ensure that patients are given a model of treatment, close follow-up, and ample opportunity to contact the treatment team. Although clinician and patient expectations of success may be hampered by the knowledge that up to half of the patients will receive a placebo (depending on the trial), it can be assumed that pharmacologic treatment trials provide an index for the value of medications delivered in the context of at least minimal psychotherapeutic information and support. Whether this level of support is greater or less than that found in clinical practice is dependent on the setting and clinician factors noted here.

For psychosocial therapy, there is wide variability in the type and targets of treatment, which may range from simple supportive or medication-compliance interventions to fully independent and intensive treatment protocols. Accordingly, any empirical review of combination treatments must take into account the type and intensity of treatments offered. This task is simplified by the predominance of cognitive-behavioral therapy (CBT) in the outcome literature. As such, the perspective on combination treatments provided here is largely specific to short-term CBT protocols (commonly 12 to 20 sessions).

As evaluated by well-controlled clinical trials, CBT and pharmacologic treatments have repeatedly demonstrated efficacy for mood and anxiety disorders and in most cases have been shown to offer approximately equal acute outcome (e.g., Dobson, 1989; Gould, Buckminster, Pollack, Otto, & Yap, 1997; Gould, Otto, & Pollack, 1995). Across diagnostic domains, there is some suggestion that CBT is a more tolerable treatment relative to medications and is especially more cost-effective, particularly when the long-term outcomes provided by CBT are considered (e.g., Antonuccio, Thomas, & Danton, 1997; Hofmann et al., 1998; Otto, Pollack, & Maki, 2000). Indeed, one of the consistent features of the effects of CBT in mood and anxiety disorders is that it has strong relapse-prevention effects. This is a particularly important feature when cost efficacy is assessed relative to medications; short-term CBT is associated with strong maintenance of treatment gains, whereas pharmacotherapy often requires ongoing treatment to prevent relapse (e.g., see Antonuccio et al., 1997; Otto et al., 2000).

Given the availability and efficacy of CBT and pharmacotherapy, there is hope that they can be combined to strikingly increase the efficacy of treatment. In most cases, this hope has not been realized, although there are specific situations and applications where it appears that combined treatments have advantages over either modality alone. On the other hand, use of both modalities of treatment may double costs of acute care, with continued escalation of pharmacotherapy costs over time. Whether there are reliable benefits in terms of additional symptom improvement or cost offsets is dependent on the nature of the disorder under treatment and the degree of additive benefit conferred by the combination treatment strategy. In the following sections, we consider the advantages and disadvantages of combined treatment for the treatment of unipolar depression, bipolar disorder, and anxiety disorders.

Perhaps most is known about combined psychosocial and pharmacologic treatment in the depression-treatment literature. Interpersonal psychotherapy (IPT) and CBT have been investigated for their efficacy alone and in combination with medications for the treatment of unipolar depression. For depressed outpatients, studies have provided only limited support for the hypothesis that psychopharmacology and these psychosocial treatments can be combined for additive efficacy. Although rates of treatment response tend to be numerically higher for combined treatments relative to either modality alone, these differences tend to be small and often do not achieve statistical significance in acute outcome studies (Beck, Hollon, Young, Bedrosian, & Budenz, 1985; Browne et al., 2002; Friedman et al., 2004; Hollon et al., 1992; Murphy, Simons, Wetzel, & Lutzman, 1984; Simons, Murphy, Levine, & Wetzel, 1986). For example, Murphy and associates (1984) compared the efficacy of the tricyclic antidepressant nortriptyline to CBT delivered alone, CBT in combination with placebo, or CBT in combination with nortriptyline in 70 patients with major depression. Differences between treatments failed to reach significance, but patients who received pharmacotherapy in combination with CBT appeared to have a slightly higher response rate (72%; defined as a Hamilton Depression Rating Scale score of less than 7) than those who received the two monotherapies (CBT alone, 63%; pharmacotherapy alone, 50%), though it was in the same range as that of the CBT-plus-placebo group (76%).

There are some indications, however, that additive effects can be observed in chronic patients. In an early study, Blackburn, Bishop, Glen, Whalley and Christie (1981) found an advantage for combined treatments (CBT with antidepressants—most typically, amitriptyline or clomipramine) in a cohort of patients characterized by severity or chronicity (e.g., a longer duration or frequency of episodes or a greater severity of psychopathology scores, inpatient treatment). In that trial, combination treatment was associated with an advantage over medication alone in the range of a very large effect size (d=1.33) according to Cohen’s (1977) standards (see Friedman et al., 2004). Very large effect sizes for combined treatment over medication alone were also obtained in the study of an inpatient cohort by Bowers (1990).

More recently, Keller et al. (2000) examined the outcome for 662 patients who received the antidepressant nefazodone alone, psychotherapy alone, or a combination of these treatment strategies. The psychotherapy under study was a 16- to 20-session treatment protocol developed specifically for chronic forms of depression that focused on the modification of interpersonal issues using cognitive and behavioral strategies. In the intent-to-treat analyses, 48% of patients treated with nefazodone or psychotherapy met responder criteria; in the combined treatment condition, 73% achieved a response, reflecting a medium-to-large effect size (d=.67). Remission criteria were achieved by 29% of the nefazodone group, 33% of the psychotherapy group, and 48% of the combined therapy group. As evaluated by endpoint depression-severity scores, the combined treatment group had an advantage in the range of a moderate effect size over either of the individual treatments.

The Keller et al. (2000) trial provides evidence for different profiles of response for the two modalities of treatment. It appears that the antidepressant nefazodone treatment led to more rapid response early in the acute trial, whereas psychotherapy led to a greater trajectory of response in the second half of the 12-week study. Accordingly, the combined treatment group may have achieved early benefits from medication and then additive benefits from psychotherapy. The brevity of the trial, however, prevents firm conclusions about the trajectory of improvements from individual or combination treatment strategies over a longer treatment interval.

In terms of the complementary action of pharmacotherapy and CBT, there is emerging evidence that CBT and pharmacotherapy may share some of the same brain targets in terms of limbic and cortical changes with treatment but that each treatment may offer a modality-specific sequence of change. Specifically, Goldapple et al. (2004) provided preliminary evidence that CBT and pharmacotherapy target different primary sites, with "top-down" effects (medial frontal and cingulated cortices) provided by CBT and "bottom-up" effects (brainstem, insula, and subgenual cingulate) provided by pharmacotherapy with paroxetine. Modulation of both these circuits results in changes in prefrontal hippocampal pathways, providing a rationale for combination treatment at the level of analysis of brain changes. However, there is evidence that the modulation of hippocampal and frontal activity may occur in reverse patterns, raising questions about the conditions by which the unique pathways to prefrontal hippocampal modulation will be synergistic instead of interfering. These intriguing findings await further study in randomized controlled trials, but imaging studies do provide a unique perspective on the mechanisms behind the separate and combined effects of CBT and pharmacotherapy.

Imaging results are consistent with the notion that, should one mechanism of change fail (i.e., top-down or bottom-up), the alternative mechanism may provide benefit. As witnessed by clinical studies, there is evidence that depressed patients failing one modality of treatment (CBT or pharmacotherapy) can achieve adequate response when switched to the other modality. Specifically, Fava et al. (1997a) found a 63% remission rate for CBT for 19 patients who had failed to respond to at least two adequate trials of pharmacotherapy, and Stewart, Mercier, Agosti, Guardin, and Quitkin (1993) found a strong response rate among CBT nonresponders who were switched to pharmacotherapy. Given these apparent complementary modes of action among patients who otherwise may fail one modality of treatment, advantages for combination treatments may be more evident when evaluated in cohorts with greater likelihood of failure to respond to one strategy or the other. The more advantageous treatment effects observed in chronic or severe patients (as noted earlier) suggests that selection of patients on these characteristics may lead to an improved cost-benefit ratio for combination treatment relative to the application of combination strategies for patients at every level of severity. However, this principle may be apt only for the acute treatment phase. Short-term CBT also appears to be particularly effective as a relapse-prevention strategy, and hence its use in medication-treated patients may lead to considerable cost offsets over the long term if patients are successfully protected from recurrent episodes. This application of CBT is considered in the following section.

Relapse prevention in the treatment of unipolar depression

There is consistent evidence that adding psychotherapy (CBT or IPT) to the acute phase of pharmacological treatment offers comparable efficacy to maintenance pharmacotherapy in helping protect against relapse. For example, Simons et al. (1986) examined 1-year outcome for patients who received CBT, pharmacotherapy, or a combination of these strategies during acute treatment. Pharmacotherapy was discontinued at Month 3, and by the end of the 1-year follow-up period, 66% of those patients who had received nortriptyline alone relapsed, as compared to 28% of the patients who had received CBT alone or in combination with nortriptyline (reflecting a large effect size, d=1.0). In a similarly designed study, Evans and colleagues (1992) found a 20% relapse and a 15% relapse (over 2 years) in patients treated with CBT alone or in combination with antidepressants, respectively, as compared to a 32% relapse rate among those who maintained pharmacotherapy treatment and a 50% relapse rate among patients who were treated with short-term antidepressant treatment alone. In these studies, CBT was added during the acute treatment phase, and it appeared to have long-term relapse-prevention effects, although sample sizes were limited.

CBT also appears to be efficacious in preventing relapse when implemented following the acute phase of pharmacotherapy. In a series of studies, Fava and associates (Fava, Grandi, Zielezny, Rafanelli, & Canastrari, 1996; Fava, Rafanelli, Grandi, Canastrari, & Morphy, 1998) examined the long-term protective effects of CBT added to the care of depressed patients who had already responded to antidepressant treatment. Following initial response, patients were randomized to CBT or a clinical management control condition, and the antidepressant treatment was tapered. Over the next 4 years of follow-up, 70% of patients in clinical management relapsed, compared to 35% in CBT, reflecting a large effect size (d=.91). By the 6-year follow-up assessment, absolute rates of relapse in these two conditions were no longer significantly different. Nonetheless, there was still strong evidence for a differential course for the two groups of patients. Even though relapse was now common in the CBT-treated patients, these patients tended to have a single relapse, as compared to the multiple relapses presented by the clinical management group over the 6-year follow-up period (Fava et al., 1998). Similar findings have been reported for IPT when added as a replacement strategy for antidepressant treatment (Frank, Kupfer, Wagner, McEachran, & Cornes, 1991).

The protection against relapse conferred by CBT appears to extend to patients with recurrent depression, defined as a history of three or more episodes of depression. Fava, Rafanelli, Grandi, Conti, and Belluardo (1998) examined the 2-year outcome of recovered outpatients with a history of multiple episodes provided with either CBT or clinical management. Antidepressant treatment was tapered in the context of this treatment. By 2 years, 80% of patients treated with clinical management had relapsed, compared to only 25% in the short-term CBT group.

In summary, if not selected as an initial monotherapy for treating major depression, CBT (and in many cases IPT) can complement pharmacotherapy in a combination treatment strategy. Evidence suggests that during acute treatment, combination treatment can offer an increased likelihood of response to patients characterized by a chronic course or pharmacotherapy nonresponse. In addition, CBT appears effective when applied as a relapse-prevention/medication-discontinuation strategy for patients who have responded to medication. Indeed, the maintenance of treatment effects over time is one feature of CBT that contributes to its favorable cost-effectiveness relative to pharmacotherapy (Antonuccio et al., 1997). Also, CBT may emerge as a preferred long-term maintenance strategy as research evaluates the possible deleterious effects of long-term antidepressant use (see Fava, 2003).

Combination therapy for bipolar disorder

The role of CBT in protecting against relapse and in treating symptoms that are not managed by pharmacology alone extends to the treatment of bipolar disorder. Pharmacotherapy has been particularly emphasized in the management of bipolar disorder; but even with the broad array of mood-stabilizing, antipsychotic, antidepressant, and antianxiety medications available, the most common course of bipolar disorder continues to be one of regular relapses to depression or mania/hypomania (e.g., Gitlin, Swendsen, Heller, & Hammen, 1995; see also, O’Connell, Mayo, Flatlow, Cuthbertson, & O’Brien, 1991). In the face of these limitations of pharmacotherapy alone, there is a growing body of literature documenting the efficacy of CBT or family therapy to improve medication adherence, reduce symptoms, and protect against relapse in medicated patients with bipolar disorder. It is important to note that this treatment is considered an adjunct to the use of mood stabilizers; hence, the assessment of combined psychosocial and pharmacologic treatment of bipolar disorder is evaluated only relative to monotherapy with medication (primarily mood stabilizers).

There are several targets of treatment for psychosocial interventions for bipolar disorder, each of which appears to be valuable in the adjunctive management of bipolar disorder: (a) improving medication adherence; (b) reducing the impact of psychosocial stressors such as negative life events, anxiety comorbidity, and negative communication patterns (e.g., expressed emotion); (c) trying to prevent or limit the severity of episodes during their prodromal state; and (d) directly intervening with bipolar depression (for review, see Miklowitz & Goldstein, 1997; Newman, Leahy, Beck, Reilly-Harrington, & Gyulai, 2002; Otto, Reilly-Harrington, & Sachs, 2003).

Concerning medication adherence, it appears that brief CBT interventions directed at improving medication adherence have significant impact on the course of bipolar disorder. Lifelong medication use is frequently recommended for bipolar patients, but poor compliance is evident in the majority of patients within the first several months of treatment. For example, by the 2nd month of treatment, it is not uncommon for half of treatment samples to have failed to maintain adequate treatment with mood stabilizers (Johnson & McFarland, 1996). Evidence for the specific value of psychosocial interventions for improving medication adherence in bipolar patients is provided by Cochran (1984). Using a brief (six-session) CBT intervention targeting adherence, Cochran obtained significantly better adherence and fewer hospitalizations over a 6-month follow-up period as compared to a treatment-as-usual comparison condition.

Broader treatment protocols emphasizing education about bipolar disorder and its management—as well as cognitive restructuring, problem solving, and sleep- and routine-management interventions (e.g., to prevent sleeplessness from triggering an episode)—have also met with success. For example, in an initial small-scale study, Lam et al. (2000) examined the outcome of CBT in a sample of bipolar outpatients who at pretreatment had an average history of seven manic episodes and nine depressive episodes. These patients, all taking mood stabilizers, were randomized to either routine care or routine care combined with 12 to 20 sessions of CBT delivered over 6 months. CBT was associated with significantly fewer manic, hypomanic, and depressed episodes than the comparison condition. Similar results have been reported by other small-scale studies (Perry, Tarrier, Morriss, McCarthy, & Limb, 1999; Scott, Garland, & Moorhead, 2001; Zaretsky, Segal, & Gemar, 1999).

In an important large-scale study, Lam et al. (2003) examined the effects of 12 to 18 sessions of CBT (M=14 sessions) delivered to bipolar patients over the course of 6 months, with 2 additional booster sessions delivered over the next 6 months. This treatment was compared to a treatment-as-usual control condition. Patients in both groups were taking similar levels of mood stabilizers and concomitant additional pharmacologic agents (e.g., antidepressant and antipsychotic medications). Outcome was examined over the course of 12 months. Although the CBT program resulted in improved medication adherence, analyses indicated that the significant beneficial effects of CBT remained when medication adherence levels were controlled. CBT, as compared to the treatment-as-usual control condition, was associated with significantly fewer days (approximately one-third less) spent in a mood episode, with evidence of similar beneficial effects on depressed, hypomanic, and manic mood episodes. Significant effects were also seen on reductions in days hospitalized due to a mood episode; CBT (combination treatment) was associated with a mean reduction of 8 days of hospitalization over the course of a year. This reduction in days hospitalized suggests potentially huge cost-offset gains for combination treatment. Accordingly, for bipolar disorder, combination treatment with CBT has the potential to enhance outcomes while reducing costs of care, taking into account both inpatient and outpatient costs.

In addition to CBT, family therapy has also been shown to be an effective combination treatment strategy for bipolar disorder. Specifically, Miklowitz and colleagues have developed and tested a family treatment emphasizing psychoeducation, problem solving, and improvement of communication patterns (Craighead & Miklowitz, 2000; Miklowitz & Goldstein, 1997). Evaluations of this treatment indicate that this approach is effective in reducing depressive symptoms, rates of relapse, and hospitalizations in patients with bipolar disorder as compared to pharmacologic treatment and case management alone (Miklowitz et al., 2000; Miklowitz et al., 2003; Miklowitz, George, Richards, Simoneau, & Suddath, 2003; Rea et al., 2003). Structured group treatment incorporating some of these same treatment elements, but with only minimal family participation, also appears to offer efficacy for bipolar disorder (Colom et al., 2003).

In summary, there is consistent evidence supporting a role for specific psychosocial treatments offered as an adjunct to the pharmacologic management of bipolar disorder. Bipolar patients, in general, are characterized by multiple relapses, and they receive only partial protection from mood-stabilizing medications. CBT and family-focused therapy appear to offer extra protection against relapse while treating current symptoms.

In many ways, outcome studies of combined treatment for anxiety disorders have provided findings similar to those for the treatment of major depression. There is some evidence for advantages of combined treatment effects over monotherapy for panic disorder, but these effects frequently do not reach significance (for review, see Foa, Franklin, & Moser, 2002). Additionally, there is evidence that patients failing to respond to one modality of treatment can benefit from a switch to the other modality of treatment. CBT can also be used as a replacement strategy for patients wishing to discontinue their medications. However, unlike studies of major depression, studies of combined treatment have presented evidence of long-term, deleterious effects of combined treatment, especially when medication is discontinued. Each of these findings is considered in the following paragraphs.

For pharmacotherapy, there is good evidence that anxious patients can improve from monoamine oxidase inhibitors, serotonin selective reuptake inhibitors, and tricyclic antidepressants, as well as benzodiazepine treatment (Lydiard, Brawman-Mintzer, & Ballenger, 1996; Pollack & Smoller, 1996). Presumably, these medications have their mechanisms of action from blocking—via different biochemical pathways—the severity of anxiety elicited by the cues specific to each disorder (e.g., social interactions for patients with social anxiety disorder; see Otto, Safren, Nicolaou, & Pollack, 2003). For the average patient, this attenuation of the anxiety response appears to work as long as medication treatment is continued. However, when treatment is discontinued, relapse is common (Noyes, Garvey, Cook, & Samuelson, 1989; Noyes, Garvey, Cook, & Suelzer, 1991; Pollack & Smoller, 1996), although studies of panic disorder indicate that relapse rates attenuate the longer a patient is kept on medication (Mavissikalian & Perel, 1993).

In contrast to pharmacotherapy, CBT is less directed at attenuating the anxiety response as it is at eliminating exaggerated fears and the avoidance responses that help maintain anxiety disorders. In exposure-based procedures, patients are repeatedly exposed to feared stimuli under controlled conditions, allowing fears to dissipate (extinguish) as patients acquire a sense of safety in the presence of these stimuli. The exact stimulus used depends on the disorder under treatment. For example, for panic disorder, core exposure procedures include exposure to feared somatic sensations (e.g., breathlessness, dizziness) as well as feared situations (situations associated with panic attacks). Exposure procedures are characteristically combined with informational and cognitive interventions aimed at maximizing the disconfirmation of exaggerated fears via careful evaluation of ongoing experiences.

CBT and pharmacotherapy appear to offer clear efficacy at short-term outcome points (e.g., 12—15 sessions of CBT, 3 months of treatment); but unlike pharmacotherapy, CBT appears to offer long-standing benefits after the cessation of formal treatment (for meta-analytic reviews, see Christensen, Hadzi-Pavlovic, Andrews, & Mattick, 1987; Gould, Buckminster, et al., 1997; Gould et al., 1995; Gould, Otto, Pollack, & Yap, 1997; Otto, Penava, Pollock, & Smoller, 1996). Indeed, for at least one anxiety disorder (social anxiety disorder), the trajectory of outcome over follow-up periods suggests that during acute treatment patients learn a method for approaching their social fears and avoidance such that patients continue to increase the magnitude of their treatment gains over time (Heimberg, Salzman, Holt, & Blendell, 1993).

An issue for combination treatment is whether the anxiety-attenuating properties of medication successfully enhance or interfere with the long-term outcome of exposure-based CBT. Because encouragement of reentry into avoided situations or events is considered a strategy that extends medication treatment effects (Sutherland & Davidson, 1995; Telch & Lucas, 1994), the combined treatment strategies considered here refer to fuller CBT protocols of treatment rather than simply exposure instruction added to psychopharmacology.

There is some evidence for advantages of combined treatment effects over monotherapy for panic disorder (Barlow, Gorman, Shear, & Woods, 2000; Marks et al., 1993), social phobia (Blomhoff et al., 2001), obsessive-compulsive disorder (Cottraux et al. 1990; Hohagen et al., 1998), and generalized anxiety disorder (Power, Simpson, Swanson, & Wallace, 1990); but there is also evidence for no advantage (e.g., Franklin, Abramowitz, Bux, Zoellner, & Feeny, 2002; van Balkom et al., 1998; for meta-analytic review, see Gould et al., 1995) or for the claim that some of the advantages of combined treatment are accounted for by pill taking alone (Barlow et al., 2000; Power et al., 1990). Of particular concern is a group of studies that indicate how combined treatment may sap some of the stronger effects of CBT over time. Two large multicenter trials provided evidence of these deleterious effects in the treatment of panic disorder; in each study, the authors suggested that some of the benefits of CBT provided during medication treatment are lost when medication is discontinued such that the long-term effects of combined treatment appear to be inferior to CBT alone (Barlow et al., 2000; Marks et al., 1993). Similar evidence has recently been documented for the combined treatment of social phobia; as with panic disorder, there is evidence that combined treatment does not have an advantage over CBT alone over long-term intervals (Haug et al., 2003).

The apparent cost of combined treatment to the long-term maintenance of treatment gains is fully consistent with data from the animal laboratory. Procedurally, exposure-based CBT is similar to animal models of extinction of conditioned fears. Research on the extinction of fear learning indicates that, rather than representing a weakening of fear associations, extinction represents the active learning of an alternative meaning ("relative safety") of the original fear cue. After extinction training, memories of the original fear learning and the extinction learning are in competition, with the dominant memory being determined by context (for review, see Bouton, 2002). For example, a fear learned in Context A and extinguished in Context B is likely to reappear if the animal is retested in Context A or in a brand new context (i.e., Context C). Contexts include, for example, the physical environment, the time of day, or internal cues such as drug or emotional state. It is changes in internal cues that may play havoc with the maintenance of treatment gains in combined treatment.

Animal research indicates that a change in internal state (such as anxiety reduction from a benzodiazepine) is a powerful-enough context such that adequate safety learning from exposure (extinction) is achieved only in that context (Bouton, Kenney, & Rosengard, 1990). When the drug state is withdrawn, so is the learned safety. Accordingly, patients acquire "safety" with CBT offered alone or in conjunction with medication; but later, when the internal context is changed via discontinuation of medication, safety learning appears to be compromised in those who had received the combined treatment. In contrast, patients who have not taken—and, logically, not stopped—medication do not undergo a shift in context, and their maintenance of treatment gains are correspondingly strong (cf., Barlow et al., 2000; Haug et al., 2003; Marks et al., 1993).

But what about patients who do not discontinue their medication treatment? Is it likely that combined treatment offers optimal outcome under conditions of continued pharmacotherapy? It is noteworthy that in the Barlow et al. (2000) trial, the data were suggestive of advantages of combined treatment for patients who continued this regimen to the 6-month evaluation point; it was only after medication discontinuation that the outcome for these patients decreased to levels below that for CBT alone. Accordingly, there has been some suggestion that, among the subset of patients who can remain adherent to ongoing pharmacotherapy over time, there is additive benefit for combined treatment. However, there are reasons to expect long-term interference effects for combined treatment (see Otto, 2002). Animal research suggests that inclusion of occasional aversive outcomes during extinction trials may make extinction learning more resistant to loss, due to inclusion of the aversive events in the context of safety learning (see Bouton, 2002). Likewise, during CBT, breakthrough episodes of strong anxiety (e.g., occasional panic attacks) during the course of extinction learning may have beneficial effects on the maintenance of treatment gains because exposure-based (extinction) learning is acquired in the context of these events. Particularly strong and early suppression of anxiety with drug treatment may aid the quick acquisition of clinical benefit but may deprive patients of robust safety learning so that when anxiety or panic occurs in the future, there is a stronger return of fear ("I can’t handle this; I am back to square one"). Although this source of relapse may be particularly important if anxiety returns quickly upon medication discontinuation, it may also explain greater relapse rates observed in a naturalistic study of CBT-treated patients who remained on medication as compared to patients who completed CBT medication-free (Otto, Pollack, & Sabatino, 1996).

In summary, some of the strongest evidence for deleterious effects of combined treatment are derived from studies that have a context shift brought about by medication discontinuation. There is only limited evidence for deleterious effects of combined treatment in patients who maintain their pharmacotherapy. Nonetheless, for some of these patients, the pharmacologic blockade of anxiety during CBT may provide a context for learning that is too narrow for broad generalization; should episodic anxiety occur in the future, memories of the original fear learning may predominate.

The internal context may be particularly salient for patients with panic disorder; hence, context shifts due to medication may be particularly difficult for these patients. Evidence for similar effects in other anxiety disorders is evident to only a limited degree in one study providing relevant data. For obsessive-compulsive disorder, there was an absence of interference effects in at least two long-term studies, but in these studies, continued medication treatment over the outcome interval was common (Cottraux et al., 1990; Hembree, Riggs, Kozak, Franklin, & Foa, 2003) and may have attenuated context effects. Nonetheless, given the clinical reality of poor medication adherence over the long run in clinical samples of anxiety patients (e.g., Cowley, Ha, & Roy-Byrne, 1997) and in clinical samples of patients taking antidepressant medications (Sirey et al., 1999; Weilburg et al., 2003), clinicians should be aware that for patients who receive combination treatment, context shifts are likely over time, if for no other reason than poor medication adherence.

There is evidence that some of the beneficial effects of CBT on long-term course can be maintained if CBT is reinstated or is ongoing at the time of medication discontinuation. This evidence comes primarily from studies of the application of CBT to benzodiazepine discontinuation difficulties. In panic disorder, discontinuation of benzodiazepine treatment can lead to symptoms as bad as or worse than those of pretreatment; furthermore, discontinuation failure is common (Fyer et al., 1987; Noyes et al. 1991). Brief CBT conducted during and after the course of taper appears to be effective in aiding benzodiazepine taper, reducing residual panic symptoms and leading to long-term maintenance of treatment gains (Hegel, Ravaris, & Ahles, 1994; Otto et al., 1993; Spiegel, Bruce, Gregg, & Nuzzarello, 1994). In terms of context effects, ensuring that CBT extends to the posttaper period appears to ensure long-term maintenance of "safety" learning in the medication-free context. There is also initial evidence that such strategies apply equally well to discontinuation of antidepressant medication (Schmidt, Wollaway-Bickel, Trakowski, Santiago, & Vasey, 2002; Whittal, Otto, & Hong, 2001). For example, Schmidt and associates randomly assigned patients with panic disorder to either continue or discontinue their serotonin selective reuptake inhibitors treatment in the context of group CBT. No adverse effects of antidepressant discontinuation were evident, and patients in both treatment conditions improved dramatically, with no significant differences between groups and with over 75% of patients meeting criteria for high-end state functioning.

CBT can also be used successfully for patients who have failed to respond to pharmacotherapy. In two studies of patients with panic disorder (Otto, Pollack, Penava, & Zucker, 1999; Pollack, Otto, Kaspi, Hammerness, & Rosenbaum, 1994) and in an initial small study of patients with posttraumatic stress disorder (Otto, Hinton, et al., 2003), CBT was found to offer significant benefit to patients who did not respond to an adequate trial of pharmacotherapy. There is also evidence that patients with panic disorder who fail to respond to CBT can achieve benefit from medications (Kampman, Keijsers, Hoogduin, & Hendriks, 2002; cf. Fava et al., 1997b).

The sequential addition of CBT to patients who have been treated with medications may be a particularly realistic strategy given limitations in the number of providers who offer CBT relative to the wide availability of medication treatment from primary care as well as specialty medical providers. There is evidence that the addition of exposure interventions alone can improve outcome for medication-treated patients. That is, while there is evidence that medications may add little to CBT alone when long-term outcomes are considered, there is evidence that combination treatment with CBT does improve outcome relative to medications alone (Barlow et al., 2000; Gould et al., 1995; Telch & Lucas, 1994). Indeed, the evidence indicates that the beneficial effects of imipramine treatment can be greatly attenuated by instructions from clinicians who discourage situational exposure (Telch, Agras, Taylor, Roth, & Gallen, 1985). Without exposure, patients do not have a mechanism to translate anxiety suppression into the fuller reduction of fears and impairment that is brought by reentry into feared situations. The value of exposure instructions to pharmacotherapy have become clear enough that, for the treatment of social phobia, clinical researchers in pharmacotherapy have recommended that exposure instructions be considered a standard part of pharmacotherapy and that the prescription given by the physician include exposure instructions as well as medication.

It is important to note that the available information on combination treatments for anxiety disorders is in the greatest abundance for panic disorder. Here the evidence suggests that combination treatments probably should not be adopted as a standard, first-line treatment. Although patients may enjoy a strong treatment response acutely, these gains may come with limitations in the longevity of treatment gains; furthermore, combined treatment adds significantly to the costs of treatment as compared to CBT alone (see Gould et al., 1995; Otto et al., 2000). Interestingly, the results from ongoing studies of combination treatments in obsessive-compulsive disorder and social phobia support the notion that intensive CBT can offer treatment outcome similar to that of combination treatment. It remains to be seen for these disorders whether medication use in combination with CBT saps the effects of CBT over time.

Novel combination treatment strategies for anxiety disorders

The combination treatment strategies reviewed here represent the application of two independent strategies, each designed to treat affective disorders on its own. But are there alternative ways to conceptualize the combination of psychosocial and pharmacologic strategies? Rather than focus on the anxiolytic, antidepressant, or mood-stabilizing effects of pharmacotherapy, can medication be used to directly enhance the change processes specific to psychotherapy? One such strategy relies not on anxiolysis but on the enhancement of the extinction learning that occurs in CBT.

This approach grew out of advances in animal research that mapped some of the core pathways and neurotransmitters involved in fear extinction (e.g., Davis, Falls, & Gewirtz, 2000; Davis & Myers, 2002). Specifically, the glutamatergic N-methyl-d-aspartate (NMDA) receptor has been shown to be critically involved in learning and memory, and this learning appears to be augmented by NMDA agonists such as D-cycloserine (Schwartz, Hashtroudi, Herting, Schwartz, & Deutsch, 1996; Tsai, Falk, Gunther, & Coyle, 1999). Interestingly, the extinction of conditioned fear appears to be facilitated by D-cycloserine given in individual doses prior to extinction (exposure) trials in an animal model (Walker, Ressler, Lu, & Davis, 2002).

These exciting findings from the animal laboratory have recently been reproduced in a sample of adults with anxiety disorders. Specifically, Ressler et al. (2004) randomized 30 patients with a fear of heights (acrophobia) to one of three treatment conditions: virtual-reality exposure (VRE) therapy plus D-cycloserine (500 mg), VRE plus D-cycloserine (50 mg), or VRE plus placebo pill. The medication and placebo were administered in a double-blind fashion in individual doses only before each of two weekly VRE sessions. Differential exposure effects were evident by the second session and at the posttreatment assessment. Patients who had received D-cycloserine reported significantly less fear to all height stimuli (perceived floors of elevation) than those treated with placebo, with maintenance of these treatment benefits at a 3-month follow-up assessment.

These dramatic findings await replication and extension to other anxiety disorders, but they do suggest that a new approach to combined treatment—where the pharmacotherapy element is targeted to memory enhancement rather than affect management—may offer a new perspective on the value of combined treatment.

For mood and anxiety disorders, it is important to remember that the investigation of combination treatments is a relatively young science. The future may bring additional information and new ways of combining these treatments that may change the now-cautious recommendations. For unipolar depression, the available evidence suggests that for the average patient, improvement with monotherapy with either pharmacotherapy or structured psychotherapy (i.e., CBT or IPT) is likely and that combination treatment may add only subtle benefit during the acute phase of treatment. However, for chronic or severe cases or in cohorts more likely to fail to respond to either treatment alone, combination treatment may offer a strong advantage. Also, for patients who initiated treatment with pharmacotherapy alone, CBT or IPT should be considered a treatment for residual symptoms or to help protect against relapse. Exciting new work on changes in brain function across successful treatment (e.g., Goldapple et al., 2004) has provided a new tool not only for understanding the central nervous system correlates of CBT and pharmacotherapy but also for distinguishing when the metabolic changes brought by these treatments have synergistic, as compared to competing, effects on brain function.

For bipolar disorder, the case for combination treatments is much easier to support as a matter of course. Medication treatment is considered to be a necessary first step for controlling the disorder; but given problems of noncompliance, breakthrough and residual symptoms, and high rates of relapse, there is ample need for additional treatment strategies. A number of psychosocial treatments have shown promise for helping fill this clinical need.

For anxiety disorders, issues of combination treatments are best informed by research on panic disorder. However, this research provides a somewhat convoluted answer. In the short and moderate term, there does appear to be some additive value to combination treatment. Yet, there are indications that combined treatment does not have the staying power of CBT alone such that treatment gains in the short term need to be balanced against the potential deleterious effects that medications may have for the maintenance of treatment gains from CBT. On the other hand, if medication treatment is selected as the initial treatment strategy, there appears to be every reason to add exposure instructions as part of pharmacotherapy to extend treatment gains. Moreover, CBT can be considered at a later stage to bolster treatment gains or to maintain and extend treatment gains during medication discontinuation.

The march of empirical research will undoubtedly bring new treatments and new combination treatment strategies to the field. At present, a new potential for combination treatment is suggested by emerging research on the use of D-cycloserine in combination with exposure-based treatment of anxiety disorders. Pending further development of this and other novel strategies, clinicians should be cautious about assuming two treatments are always better than one. Instead, the evidence supports judicious application of combined treatments specific to the characteristics of the disorder and the needs of the patient.