maintenance of gains following treatment discontinuation (e.g., Barrett, Rapee, et al., 1996; Kendall et al., 1997) and SAD diagnosis has not been a predictor of poor outcome in these studies. Thus it can be surmised that CBT provides lasting benefits for these children. As mentioned above, SAD may be more common in young children, and its treatment and maintenance of gains have not been studied in adolescent samples. The treatment of selective mutism has also received little attention in the acute treatment literature, thus even less is known about maintenance of gains, especially in adolescents. In general, the available data indicate partial response to acute treatment, which suggests continued impairment. What is unknown is whether the posttreatment gains are maintained over time.

Relapse rates following discontinuation of acute benzodiazepine treatment in adults are as high as 80%, indicating the absence of maintenance of medication-produced gains (Rickels, Case, & Diamond, 1980; Rickels, Case, Downing, & Fridman, 1986). A few studies demonstrate lasting benefit for GAD patients treated for 6 months or more with benzodiazepines and buspirone. However, concern over physiologic dependence and potential for abuse make long-term treatment with benzodiazepines controversial. A recent placebo-controlled trial of venlaflaxine XR showed that benefits from this medication continued following 6 months of treatment and that it was well tolerated over the long term. Several CBT studies involving GAD in youth have indicated maintenance of gains up to 12 months after CBT discontinuation. In addition, there are supportive data for 3.5-year (Kendall & Southam-Gerow, 1996), 6-year (Barrett, 1998), and 7.4-year (Kendall et al., 2004) follow-ups. Although not all of these treated cases had GAD as the principal diagnosis and not all evidenced complete initial benefit from CBT, those individuals who do benefit seem to be able to maintain their gains for several years, even without continued treatment.

The adult literature on maintenance of gains following treatment of specific phobias generally suggests that benefits associated with CBT are lasting even when the treatment is delivered in a single 3-hour long session (Öst, 1989; Öst et al., 1991, 1997). Studies of children and adolescents have indicated that CBT produces significant gains immediately after treatment up to and at 1-year follow-up (Öst et al., 2001; Silverman et al., 1999a). The pharmacotherapy literature on this topic is underdeveloped and thus no conclusions can be drawn about maintenance of gain over time using this approach.

Mavissakalian and Perel (1992) compared patients who responded to 6 months of acute treatment with adult patients who had a similar response to a 6-month acute treatment period and then additionally received 1 year of half-dose maintenance imipramine treatment. Both groups entered a double-blind discontinuation study for 3 months, followed by a 3-month drug-free period. The group who had acute and maintenance treatment did significantly better during this 6-month period than the group that had only acute treatment. Maintenance treatment with antidepressants may have a significant prophylactic effect and reduce the rate of relapse after treatment discontinuation (Mavissakalian & Perel, 1992). A follow-up study found prophylactic efficacy for maintenance treatment continued beyond 1 year of maintenance and into a second year. However, the persistence of difficult side effects, particularly weight gain, was noted with imipramine over long-term treatment (Mavissakalian & Perel, 2001). A recent naturalistic study (Simon et al., 2002) involving 78 patients who achieved remission in an 8-week acute phase (with benzodiazepine alone, antidepressant alone, or both) and then received maintenance pharmacotherapy for 2 years found that 46% of these patients relapsed over the maintenance phase and that combination therapy did not confer any added advantage.

As briefly noted earlier, a multicenter trial studied the efficacy of medication therapy and CBT and explored whether combined therapy (medication plus CBT) is more effective than either therapy alone in a randomized, double-blind placebo study (Barlow et al., 2000). In terms of acute response, combined treatment did not differ from medication or CBT alone, but was better than placebo. The 6-month maintenance response rate of the combination therapy was high (approximately 57% for the Panic Disorder Severity Scale [PDSS] and 60% for the CGI) and significantly different from medication alone or CBT alone but not significantly different from CBT plus placebo. After the 6-month maintenance phase, patients had treatment discontinued and were followed for 6 months. Improvement in CGI was highest for CBT plus placebo (41%), 32% for CBT alone, 20% for imipramine alone, and 26% for combined treatment. This study and another large-scale RCT using alprazolam (Marks et al., 1993) indicate that combined treatment for panic disorder may be advantageous at posttreatment and during maintenance but may attenuate the benefits of CBT following medication discontinuation (for a review see Foa, Franklin, & Moser, 2002). Ollendick's (1995) small multiple-baseline study included follow-up over a 6-month interval, and these results affirmed the lasting effects of a similar treatment for a pediatric sample. Much more research employing controlled designs needs to be done using pediatric samples.

Little is known about how long SRI medication should be continued in OCD. In practice, many adult patients continue taking their medication for at least one year; some seem to require indefinite treatment. There are only three published double-blind SRI discontinuation studies (Koran, Hackett, Rubin, Wokow & Robinson, 2002; Pato, Zohar-Kadouch, Zohar, & Murphy, 1988; Romano, Goodman, Tamura, Gonzales, & the Collaborative Research Group, 2001) in adults with OCD. Each uses a different SRI (clomipramine, fluoxetine, sertraline) and comes to a different conclusion regarding the effects of dis continuing the SRI (substantial [89%] recurrence over 12 weeks for clomipramine, moderate relapse [32%] after one year and not different from staying on SRI [21%] for fluoxetine, low rate of recurrence [24%] after 28 weeks but significantly more than staying on SRI [9%] for sertraline). Given the paucity of blind studies and the methodological differences between them (e.g., relapse definition, length of follow-up, procedure for placebo substitution), the posttreatment effects of SRIs in OCD remain unclear.

A recent comparative trial of clomipramine and EX/RP in adults suggested that the relapse rate with clomipramine may not be as high (i.e., 45%) as previously thought. However, even with this lower relapse rate, EX/RP showed superior maintenance of gains 12 weeks after treatment discontinuation (Simpson et al., 2004). Optimal lengths of SRI treatment to minimize relapse, and predictors of who can discontinue without relapsing are lacking. Moreover, there is no agreement on the criteria to define relapse. The few open studies of CBT in pediatric OCD that have included follow-up data suggest good maintenance of gains following treatment discontinuation (Franklin et al., 1998; March et al., 1994; Wever & Rey, 1997), and the one study of pharmacotherapy discontinuation in pediatric OCD suggests that relapse is common (e.g., Leonard et al., 1991). A systematic follow-up study (done when clomipramine was available only investigationally and access to CBT was limited) of 54 children and adolescents with OCD treated acutely with clomipramine was completed (Leonard et al., 1993). Of the 54 subjects, 38 (70%) were taking psychoactive medication at follow-up; 23 of 54 (43%) patients still met diagnostic criteria for OCD, and 43 (80%) were improved from baseline. Of the six patients (11%) who were totally asymptomatic, three were taking medication at the time of reevaluation; thus, only three could be considered to be in true remission. This finding suggested that patients were improved at follow-up, but most were on medication maintenance, and only 3 of 54 were asymptomatic and not on medication. Although not every patient had attempted discontinuation of medication, this study for patients treated with an SRI indicates that they may require long-term maintenance therapy.

Deblinger and colleagues' (Deblinger and Heflin, 1996; Deblinger et al., 1999) work with sexually abused children provided evidence for the long-term effectiveness of a 12-week CBT treatment. March, Amaya-Jackson, et al. (1998) reported a robust beneficial effect and maintenance of gains up to 6 months in a small study of CBT for school-age children and adolescents who experienced a single incident traumatic experience. Pharmacological studies in adolescents need to take into account the two adult randomized studies of SSRIs in the treatment of PTSD that indicate relapse with discontinuation of medication (Martenyi, et al., 2002a). These adult studies raise the same issue as that in the treatment of other anxiety disorders—whether a combination of medication and trauma-focused CBT would provide greater resistance to relapse.

It is important to note that there has been no research specifically addressing maintenance treatment and relapse prevention for anxiety disorders in youth and adolescents. In general, the ultimate goal of most psychosocial treatment programs for anxiety disorders is to equip children with skills that will help them manage anxious distress after treatment discontinues; a “cure” for anxiety is not the goal (see Kendall, 1989). Because some degree of anxious arousal is likely to persist after treatment, modifying dysfunctional expectations and distorted processing styles can help to enable more adaptive functioning. The use of in vivo exposure tasks in treatment provides performance-based experiences of coping that bolster a child's confidence for future situations. Therapeutic intervention may be only a first step, but it is a step that helps to alter the maladaptive developmental trajectories of these children so that they are better able to address the inevitable challenges emerging in their lives. Upon completion of a treatment program, the guiding principle (and hope) is for the child to continue to practice the skills learned.

There are several clinical strategies to help guide youth toward consolidation of treatment-produced gains. First, the therapist should shape and encourage “effort” attributions regarding the management of anxiety. Youth should be encouraged to reward their hard work and coping efforts, even if the successes are only partial. A second principle for continued posttreatment functioning includes introducing children to the concept of “lapses in efforts,” rather than “relapses” (see also Brownell, Marlatt, Lichtenstein, & Wilson, 1986; Marlatt & Gordon, 1985). Mistakes and partial successes are not viewed as incompetence or inability; rather, they can be constructively framed as vital and inextricably linked to the learning process. Within this framework, children can label and accept inevitable setbacks as temporary and then proceed to work on forward-looking problem solving. Mistakes are viewed as an acceptable part of the learning process and not as excuses for giving up or confirming anxious cognition.

Maintenance treatment may require combinations of CBT and medications, booster sessions, or a return and reexperience of the initial therapy. Alternate approaches, or even as-yet undeveloped treatments may be needed, especially for cases that are refractory to the otherwise reasonably successful program. To date, the field is lacking information about how best to maintain treatments, prevent relapses, and integrate psychological and pharmacological approaches to maximize long-term gains.

There are no controlled studies in any age group on how to address partial responders and nonresponders with social anxiety disorder. Cognitive-behavioral therapy is a logical candidate to augment partial drug response, and Liebowitz and colleagues have recently found CBT to be helpful in augmenting partial SSRI responders in an open trial (unpublished data). Gabapentin and clonazepam are also considered possible augmenting agents for partial SSRI re

sponders. Complicating this research is the fact that there is no accepted definition of partial response.

Most adult GAD patients who improve with acute treatment do not reach full remission of symptoms. In early trials, full remission occurred in as few as one-third of GAD patients. Unfortunately, there are no systematic studies of treatment options for patients with partial medication response in GAD. Partial response and a nonresponder rate also characterize CBT outcomes for GAD. Unfortunately, there are no systematic trials assessing nonresponders to initial treatment.

There are no controlled data on how to address or augment partial responders and there are no controlled data on how to treat nonresponders to an initial drug or CBT trial.

A recent double-blind, placebo-controlled trial with adults found that pindolol had a beneficial augmentation effect on fluoxetine-treated panic patients (Hirschmann et al., 2000). Patients with panic disorder (N = 25) who had not responded to 8 weeks of fluoxetine (and prior to that, two other trials of antidepressants) were randomly assigned to pindolol or placebo for 4 weeks. The pindolol group achieved significant improvement on ratings compared to the placebo group. However, since the augmentation period studied was so brief, it is difficult to determine if these effects would be sustained and whether they would actually lead to remission. Three studies have examined the effects of benzodiazepine treatment combined with antidepressant treatment. Imipramine plus alprazolam was compared to imipramine plus placebo. The combi nation group responded with therapeutic effect more quickly. Clonazepam augmentation of paroxetine was shown to be superior to paroxetine alone (Mathew, Coplan, & Gorman, 2001). Goddard et al. (2001) studied 50 panic patients and randomized them to either .5 mg of clonazepam three times per day plus sertraline or placebo clonazepam plus sertraline for the first 4 weeks. At week 1, 41% of the combination group evidenced improvement compared to 4% of the placebo group. At 3 weeks, there was a significantly higher (63%) response rate in the combination group than that in the placebo group (32%), but this difference did not emerge at any other point in the trial. This study suggests that benzodiazepine augmentation of SSRIs at the beginning of treatment can lead to earlier improvement. Recently, Kampman, Keijsers, Hoogduin, and Hendriks (2002) found that paroxetine augmentation of 43 nonresponders to an acute phase of CBT significantly reduced overall anxiety and agoraphobic behaviors. Nonresponders pose a special challenge, yet there is a paucity of systematic investigation in this area.

In considering medication augmentation, a benzodiazepine, such as clonazepam, is occasionally added, but disinhibition, dependence, and tolerance to the medication have limited the enthusiasm for this choice in the long run (Leonard et al., 1994). Adult studies support a trial of neuroleptic augmentation in SRI non-or partial responders (McDougle et al., 1995). In a Type 1 controlled risperidone versus placebo SRI augmentation study, risperidone addition was superior in reducing OCD symptoms (McDougle et al., 2000). Neuroleptic augmentation has not been systematically studied in children; a case series of children who were refractory to SRI therapy improved significantly after risperidone was added (Fitzgerald, 1999). T. Owley, S. Owley, Leventhal, and Cook (2002) reported four cases of 8-to 25-year-olds who were partial responders to an SSRI and subsequently responded after mixed salts of dextroamphetamine (Adderall) was added. The authors speculated that the “del

icate” balance of serotonergic and dopaminergic systems may be affected by both neuroleptic or Adderall augmentation, resulting in increased serotonergic transmission.

There is no accepted definition of a nonresponder. Sometimes nonresponders have meant patients who have had a suboptimal response to one therapy, but such a definition would include both partial responders and nonresponders. If a nonresponder is defined as someone who has had absolutely no improvement from an initial treatment, there are no published controlled data on how to treat such patients.

There is only one placebo-controlled study on augmentation for partial responders: combat veterans with PTSD who were minimally responsive on an SSRI (sertraline) were given augmentation with the atypical antipsychotic medication olanzepine (Stein, Westenberg, & Liebowitz, 2002). Olanzapine augmentation was associated with statistically significantly greater reduction than that with placebo in specific measures of posttraumatic stress, depressive, and sleep disorder symptoms (Petty et al., 2001). One recently completed study found an augmentation effect for PE compared to sertraline continuation in patients who evidenced at least a 20% symptom reduction following 10 weeks of open label sertraline (Foa, Franklin, & Moser, 2002). In a subgroup analysis of these data, patients who experienced only a marginal response to 10 weeks of open label sertraline benefited substantially when PE was added, compared to those who continued on sertraline for an additional 5 weeks.

There are no controlled data on how to treat nonresponders to pharmacologic or CBT treatment for PTSD in adults. Given the high rates of comorbidity with PTSD, it is recommended that patients be given evidence-based treatments for their comorbid conditions and that thereafter the patient's PTSD symptoms be reassessed.

There is little to no research on the management of partial response and nonresponse in the treatment of adolescent anxiety disorders. This area is one that sorely needs to be addressed in future research. Although the literature has yet to provide the data needed to guide evidence-based treatments for partial responders and nonresponders, several issues still warrant our consideration.

The evidence accrued to date informs us about the choice of an initial treatment to be undertaken for anxious children and, at least to some extent, anxious adolescents. There is a noted absence of studies specifically about adolescents and what to do for them at the various other stages of treatment. Studies reviewed above have evaluated the efficacy of various treatments for children identified with anxiety disorders and these data guide our treatment choices. However, it is likely that some of the cases treated in these studies had prior experience with one or another treatment, and may or may not have been partially refractory to those earlier treatments. More detailed analyses of initial treatment response may reveal useful information about the moderating role of prior treatment experiences in the efficacy of the treatment being evaluated. The field has nevertheless made meaningful progress in identifying at least a few initial treatments that are quite promising for anxious youth.

We know little about what to do when the youthful patient's response to treatment is not favorable. Even when approximately two-thirds of patients respond favorably to CBT, for example, there are still one-third of the patients who did not respond well and may need something additional. One might speculate that within the CBT approach, a combination of more practice, increased exposure tasks, and help with the use of new skills in new, challenging situations would be worthwhile. One might also speculate that a combination of approaches may be valuable. Treatment of a nonresponsive client can be complicated by the fact that a nonresponder to