The Efficacy of Short-term Psychodynamic Psychotherapy in SpecificPsychiatric Disorders:  A Meta-analysis FREE

The place of psychoanalytic and psychodynamic treatments within psychiatryis controversial.^1- 4 Forlong-term psychoanalytic psychotherapy and psychoanalysis, convincing outcomeresearch is urgently needed.^1- 2 Inthe field of short-term psychodynamic psychotherapy (STPP), more evidenceof efficacy is available. Various meta-analyses have addressed the efficacyof STPP.^5- 7 Svartbergand Stiles⁶ found STPP to be superior to ano-treatment control condition, but inferior to alternative psychotherapies.According to Crits-Christoph,⁵ STPP achievedlarge effect sizes compared with untreated waiting-list control patients andwas found to be equally effective as other forms of treatment such as cognitive-behavioraltherapy (CBT) or psychopharmacological treatment. Although the effect sizesof STPP assessed by Anderson and Lambert⁷ werea bit lower than those reported by Crits-Christoph, their results corroboratedthe findings of Crits-Christoph.⁵ One factorthat heavily influences the outcome of a meta-analysis is the selection andthe quality of the studies included.⁸ UnlikeSvartberg and Stiles,⁶ Crits-Christoph⁵ and Anderson and Lambert⁷ includedstudies of interpersonal therapy (IPT) as representative of STPP. However,the relation of IPT to STPP is controversial.⁹ Accordingto empirical results, IPT seems to be very close to CBT.¹⁰

Furthermore, the methodological quality of the studies included in ameta-analysis plays an important role. As Messer and Warren^11- 12 havepointed out, many studies included in the meta-analysis of Svartberg and Stiles⁶ showed severe conceptual and methodological flaws.This is also true for the meta-analysis of Grawe et al,^13- 14 whichcompared STPP and CBT. To avoid these flaws, Crits-Christoph⁵ includedin his meta-analysis only studies that fulfilled rigorous inclusion criteria(eg, use of therapy manuals, experienced therapists, minimum number of sessions).Wampold et al¹⁵ assessed the efficacy of bonafide treatments and did not find differences between different methods ofpsychotherapy. Up to the present, 2 disorder-specific meta-analyses of STPPhave been published. The meta-analysis of Leichsenring,¹⁶ whichused criteria similar to those of Crits-Christoph,⁵ foundSTPP and CBT to be equally effective in the treatment of depression. For thetreatment of personality disorders, significant and large effect sizes forSTPP and CBT were found in the meta-analysis of Leichsenring and Leibing.¹⁷

Randomized controlled studies (RCTs) are regarded as the gold standardfor the demonstration that a treatment is effective. This applies not onlyto psychotherapy research but to the realm of evidence-based medicine in general.^18- 22 Thisassumption is reflected in the criteria proposed by the Task Force on Promotionand Dissemination of Psychological Procedures of Division 12 (Clinical Psychology)of the American Psychological Association for the definition of empiricallysupported treatments.²³ According to thesecriteria, empirical support for psychotherapeutic methods can only be providedby RCTs in which a therapy group is compared with a control condition (waitinglist or placebo group) or with an already established therapy.^8,23- 24 Furthermore,the use of therapy manuals and the treatment of a specific disorder are required.

Since the publication of the meta-analyses of Svartberg and Stiles,⁶ Crits-Christoph,⁵ Graweet al,¹³ and Anderson and Lambert⁷ andthe report of the Task Force,²³ several studiesof STPP have been published that have not yet been included in meta-analysesof STPP. The most recent study included by Anderson and Lambert was publishedin 1993, ie, more than 10 years ago.²⁵ Forthis reason, we conducted a new meta-analysis that included the more recentstudies of STPP. As the results of a meta-analysis depend on the quality ofthe studies included, we applied rigorous inclusion criteria.

This meta-analysis addressed the following questions: (1) What is theevidence of improvement in target problems, general psychiatric symptoms,and social functioning after STPP? (2) Do the effects of STPP exceed the effectsof untreated control groups and treatment-as-usual (TAU) groups? (3) Are theredifferences in efficacy between STPP and other forms of psychotherapy? (4)How stable are the effects of STPP? and (5) Is there a correlation betweenoutcome and duration of treatment?

Apart from conceptual and technical differences, there are some therapeuticelements that the different models of STPP have in common.^11- 12 Withregard to formal characteristics, they are time limited (usually 16-30 sessionswith a range of 7-40 sessions) and performed in a face-to-face setting, with1 or 2 sessions a week.^12,26- 28 Therapiesincluding 6 or fewer sessions are regarded as ultrabrief.²⁸ Thetherapeutic techniques of STPP are elaborations and modifications of generalprinciples of psychodynamic psychotherapy. Short-term psychodynamic psychotherapyis characterized by the following features.¹² First,therapists are usually relatively active and foster the development of a therapeuticalliance and a positive transference. Second, STPP focuses on specific conflictsor themes that are formulated early in therapy. Third, attention is paid toadherence to the focus, the setting of achievable goals, and termination issues.The focus is on the patients’ experiences here and now, including theirsymptoms. With regard to transference, the emphasis is more on the here-and-nowdimension, ie, on the present relationship between the patient and the psychotherapist,which is not necessarily traced back to the past. In a review of empiricalstudies, Blagys and Hilsenroth²⁹ identified7 features that were observed significantly more frequently in STPP comparedwith CBT, including focus on affect, resistance, identification of consistentpatterns (of relationships, feelings, and behaviors), past experiences, interpersonalexperiences, the therapeutic relationship, and wishes, dreams, or fantasies.

We collected studies of STPP that were published between January 1,1970, and September 30, 2004, by performing a computerized search using MEDLINE,PsycINFO, and Current Contents. The following key words were used: psychodynamic therapy, psychoanalytically oriented therapy, psychoanalytictherapy, and randomized controlled trial.A total of 141 partly overlapping journal articles were found. The only studiesincluded in the meta-analysis fulfilled the following methodological requirements:(1) a randomized controlled design was applied; (2) a specific form of STPPas represented in a treatment manual or manual-like guide was applied, andtreatment integrity was ensured; (3) therapists were specifically trainedand/or experienced in STPP techniques; (4) patients with specific psychiatricdisorders were treated (no mixed samples or analogue studies); (5) the patientsample was clearly described; (6) diagnostic procedures and outcome measuresfor which reliability and validity have been demonstrated were used; and (7)data were reported that are necessary to calculate pretreatment-posttreatmenteffect sizes.

Contrary to the meta-analyses of Crits-Christoph⁵ andAnderson and Lambert,⁷ studies of IPT werenot included (eg, Elkin et al³⁰ and Wilfleyet al³¹), because the relation of IPT to STPPis controversial, and empirical results suggest that IPT is very close toCBT.⁹ Thus, this review includes only studiesfor which there is a general agreement that they represent models of STPP.As it is questionable to aggregate the results of very different outcome measuresthat refer to different areas of psychological functioning, we assessed theefficacy of STPP separately for target symptoms, general psychiatric symptoms(ie, comorbid symptoms), and social functioning.³² Thisprocedure is analogous to the meta-analysis of Crits-Christoph.⁵ Asoutcome measures of target problems, we included patient ratings of targetproblems and measures referring to the symptoms that are specific to the patientgroup under study, eg, measures of anxiety for studies investigating treatmentsof anxiety disorders.³³ For the efficacy ofSTPP in general psychiatric symptoms, broad measures of psychiatric symptomssuch as the Symptom Checklist-90 and specific measures that do not refer specificallyto the disorder under study were included; eg, the Beck Depression Inventoryapplied in patients with personality disorders.^34- 35 Forthe assessment of social functioning, the Social Adjustment Scale and similarmeasures were included.³⁶

We calculated within-group effect sizes for all studies using the Cohen d statistic. For each measure, we subtracted the posttreatmentmean from the pretreatment mean and divided the difference by the pretreatmentstandard deviation of the measure.³⁷ If therewas more than 1 patient group, we calculated a pooled baseline standard deviation,as suggested by Rosenthal.³⁸ If necessary,signs were reversed so that a positive effect size always indicated improvement.We assessed effect sizes separately for measures of target problems, generalpsychiatric symptoms, and social functioning. Whenever multiple measures wereapplied in a study, we assessed the effect size for each measure separatelyand calculated the mean effect size to assess the overall outcome of the studyin the respective area of functioning. We computed effect sizes that wereunweighted and those weighted by the sample size to yield unbiased estimatorsof effect sizes.^38- 39 To examinethe stability of psychotherapeutic effects, we assessed effect sizes separatelyfor assessments at termination of therapy and follow-up.

Furthermore, between-group effect sizes were assessed according to themethod described by Cohen³⁷ by calculatingthe difference between preoutcome and postoutcome (follow-up) measures ofthe STPP group minus the difference between preassessment and postassessment(follow-up) measures of the comparison group, divided by the pooled standarddeviation before therapy. These between-group effect sizes give the differencein the magnitude of change between STPP and the comparison group in unitsof standard deviation. An effect size of zero indicates that STPP and comparisongroups are equal in therapeutic effect. All calculations were made so thata positive effect size indicates superiority of STPP.

The first rater extracted the following information from the articles:(1) name of the authors, (2) year of publication,(3) psychiatric disordertreated with STPP, (4) model of STPP, (5) duration of STPP, (6) comparisongroup, (7) sample size in each group, and (8) means and standard deviationsfor each outcome measure. This information was checked by the second rater,and disagreements were resolved by consensus, including the ratings of thethird rater.

Seventeen studies of STPP met the inclusion criteria.^40- 57 Thestudies are described in Table 1 (datanecessary for the assessment of effect sizes were provided by Susan Bögels,PhD [written communication, January 23, 2004]; Paul Crits-Christoph, PhD [writtencommunication, November 11, 2003]; and P. J. Cooper, PhD, and Helena Romaniuk,PhD [written communication, November 24, 2003]).

Of the 17 studies that were identified, 13 (76%) have not yet been includedin other meta-analyses of STPP. Only 2 of the 19 studies that Svartberg andStiles⁶ included in their meta-analysis metour inclusion criteria.^45,54 Ofthe 26 studies in the meta-analysis of Anderson and Lambert,⁷ 4could be included.^41,54- 56 Ofthe 11 studies included by Crits-Christoph,⁵ weincluded 4.^41,54- 56 Crits-Christoph⁵ and Anderson and Lambert⁷ includedstudies of IPT, which we excluded. Only 2 studies of the meta-analysis ofLeichsenring¹⁶ were included,^52,54 andonly 1 study⁵⁰ of the meta-analysis of Leichsenringand Leibing.¹⁷ Accordingly, only a limitedportion of the studies that were included in previous meta-analyses of STPPwas included in our meta-analysis. Thus, most of the studies included in ourmeta-analysis have not been included in the previous meta-analyses.

In the 17 studies included in our meta-analysis, different models ofSTPP were applied (Table 1). The mostfrequently applied concepts of STPP were the methods developed by Luborsky,⁶⁵ Horowitz,⁷³ Shapiroand Firth,⁶⁶ and Davanloo.²⁷

In the 17 studies of STPP, the number of sessions conducted ranged from7 to 40. Short therapies of 7 sessions were performed in the study of Hamiltonet al,⁴⁸ whereas longer therapies were performedin the studies of Winston et al⁵⁵ and Svartberget al,⁵³ who studied the treatment of personalitydisorders (Table 1). The mean numberof sessions of STPP was 20.97 (SD, 10.90).

Mean length of follow-up was about 1 year (mean, 61.42 weeks; SD, 71.26weeks).

In the studies selected, the following psychiatric disorders were treatedwith STPP: major depression (2 studies), maternal depression (1), posttraumaticstress disorder (1), bulimia nervosa (2), anorexia nervosa (1), opiate dependence(2), cocaine dependence (1), cluster C personality disorders (3), borderlinepersonality disorder (1), somatoform pain disorder (1), chronic functionaldyspepsia (1), and social phobia (1).^40- 57 Itis desirable to perform meta-analyses of STPP separately for specific psychiatricdisorders, eg, for somatoform disorders or anxiety disorders. However, thenumber of RCTs of STPP is not yet large enough for that purpose. Nevertheless,it is still useful at the present state of research to evaluate the efficacyof STPP in psychiatric disoders in general, ie, across different psychiatricdisorders.

Within-group effect sizes of STPP, other forms of psychotherapy, TAU,and waiting list are presented in Table 2.

Short-term psychodynamic psychotherapy yielded an effect size of 1.39after therapy and of 1.57 at follow-up (Table2). To determine whether the effect sizes were different from zero,we performed unpaired, 2-tailed t tests. Both effectsizes differed significantly from zero (t₁₆ = 6.94 [P<.001] and t₁₅ = 7.10 [P<.001],respectively). The limits of a 95% confidence interval (CI) were 0.97 and1.82 at the posttherapy assessment and 1.10 and 2.04 at the follow-up assessment.Adjustment for sample size^38- 39 yieldednearly identical results (1.37, t₁₆ = 7.00[P<.001] and 1.54, t₁₅ = 7.15 [P<.001], respectively).

Short-term psychodynamic psychotherapy yielded an effect size of 0.90after therapy and of 0.95 at follow-up. Both effect sizes differed significantlyfrom zero (t₁₄ = 7.25 [P<.001] and t₁₂ = 6.90[P<.001], respectively). The limits of a 95% CIwere 0.64 and 1.17 at the postassessment measurement and 0.65 and 1.25 atthe follow-up assessment. Adjustment for sample size again yielded nearlyidentical effect sizes of 0.89 (t₁₄ = 7.29[P<.001]) and 0.93 (t₁₂ = 6.95 [P<.001]).

Short-term psychodynamic psychotherapy yielded an effect size of 0.80after therapy and of 1.19 at follow-up. Both effect sizes differed significantlyfrom zero (t₁₀ = 7.29[P<.001] and t₇ = 4.65[P = .002], respectively). The limits ofa 95% CI were 0.56 and 1.05 at the posttherapy assessment and 0.58 and 1.79at the follow-up assessment. Again, adjustment for sample size yielded nearlyidentical effect sizes of 0.79 (t₁₀ = 7.28;[P<.001]) and 1.16 (t₇ = 4.64 [P = .002]).

To assess the stability of effects, we compared the pretherapy and posttherapyeffect sizes with the pre–follow-up effect sizes for only those studiesthat included follow-up assessments. For STPP, the pre–follow-up effectsizes were stable (target problems, 1.44 vs 1.57; general psychiatric symptoms,0.91 vs 0.95; social functioning, 0.89 vs 1.19). This was also true for CBT(target problems, 1.37 vs 1.33; general psychiatric symptoms, 1.01 vs 0.97;social functioning, 0.97 vs 1.05).

Four studies provided data for waiting-list control patients (Table 1). No treatment yielded small effect sizesthat ranged from 0.12 to 0.27 (Table 2).Because of the small number of studies (n = 4), we did not performtests of significance for the between-group effect sizes of STPP and no treatment.

Three studies provided data for TAU (Table1). We decided to regard the control condition of the study of Gowerset al⁴⁷ as TAU rather than as no treatment,because some of these patients had “received extensive treatment elsewhere.”Treatment as usual yielded effect sizes that ranged from 0.22 to 0.95 (Table 2). Again, we did not perform tests ofsignificance for the between-group effect sizes of STPP and TAU because ofthe small number of studies (n = 3).

To compare STPP with no treatment and TAU, we tested whether the between-groupeffect sizes between STPP on the one hand and no treatment and TAU on theother significantly differed from zero (t tests forindependent samples). According to the results, STPP was significantly superiorto no treatment and TAU with regard to target problems, general psychiatricsymptoms, and social functioning (Table 3).Because of the small number of studies, these tests were performed only forthe posttherapy effect sizes. The between-group effect sizes were large fortarget problems and medium to large for general psychiatric symptoms and socialfunctioning (Table 3).

Fifteen studies included a comparison of STPP with other forms of psychotherapy(Table 1). For these comparisons wecalculated between-group effect sizes (Table 3). They ranged from 0.02 to 0.23, thus corresponding to small effectsizes according to Cohen.³⁷ To test for differencesbetween STPP and other forms of psychotherapy in these 15 studies, we performedmultiple analyses of variance (MANOVA).⁷⁴ Ina first MANOVA, we compared the pretherapy-posttherapy effect sizes betweenSTPP and other forms of psychotherapy in target symptoms, general psychiatricsymptoms, and social functioning. In a second MANOVA, we compared the pre–follow-upeffect sizes between STPP and other forms of psychotherapy on the 3 outcomevariables. Both MANOVAs yielded an insignificant result at the posttherapyassessment (Wilks Λ = 0.92; F_3,15 = 0.49[P = .69])and at follow-up (Wilks Λ = 0.89;F_3,13 = 0.53 [P = .67]).According to these results, STPP and other forms of psychotherapy did notdiffer significantly concerning their effect sizes at the end of therapy orat follow-up.

We tested whether the effect sizes of STPP correlate with the durationof therapy (number of sessions). For STPP, all Spearman rank correlationswith any outcome measure were insignificant (for all, r_s ≤ 0.21 [P = .64]).

The place of psychoanalytic and psychodynamic treatments within psychiatryis controversial. This meta-analysis addressed the efficacy of STPP in specificpsychiatric disorders. At present, the number of RCTs is not yet large enoughto perform meta-analyses of STPP for specific psychiatric disorders separately.For this reason, our meta-analysis addressed the question of how effectiveSTPP is in general in the treatment of psychiatric disorders. We studied theeffects in target problems, general psychiatric problems, and social functioning.

The data presented herein corroborate the results reported by Crits-Christoph,⁵ Anderson and Lambert,⁷ andLeichsenring.¹⁶ However, this meta-analysisadds to the literature for several reasons. First, more rigorous inclusioncriteria were applied. As a consequence, only a minority of studies includedin previous meta-analyses were regarded as adequate for inclusion in thismeta-analysis. Second, studies of IPT were not included because the relationof IPT to STPP is controversial. Empirical results seem to support the exclusionof IPT from a meta-analysis of STPP.¹⁰ Thus,the database of this meta-analysis is different from those of previous meta-analyses.Third, 76% of more recently published studies included in this meta-analysishad not yet been included in other meta-analyses of STPP. Fourth, the stabilityof effects of STPP was assessed by studying the pretreatment and posttreatmenteffects and the effects yielded in follow-up studies.

According to the results, STPP yielded significant and large effectsizes. With regard to between-group effect sizes, which allow for a strictertest of efficacy and are more appropriate to the fact that only RCTs wereincluded, STPP was significantly superior to no treatment and TAU. No differencewas found between STPP on the one hand and CBT or other psychotherapies onthe other concerning changes in target problems, general psychiatric problems,and social functioning. These results differ in part from those reported bySvartberg and Stiles,⁶ who found STPP to besuperior to no-treatment controls but inferior to alternative psychotherapies.⁶ The discrepant results can be explained by the selectionof studies. In our meta-analysis, only studies were included that fulfilledrigorous inclusion criteria that, among other things, ensured that STPP wasdelivered competently. The results of this meta-analysis are consistent withthose of the meta-analysis of Wampold et al,¹⁵ whodid not find differences between bona fide methods of psychotherapy. The insignificantdifferences in effect size between STPP and other forms of psychotherapy foundin this meta-analysis correspond in magnitude quite well to the data reportedby Grissom,⁷⁵ who found a mean difference ineffect size of 0.23 between different forms of psychotherapy.

Effect sizes can be transformed into percentages of nonoverlap.^37,76- 77 According to theseresults, patients treated with STPP are better off with regard to their targetproblems than 92% of the patients before therapy. At follow-up, which wason average 13 months after termination of therapy, they were better off than95% of the patients. As an effect size of 1.00 corresponds to a success rateof 72%, clearly more than 72% of the patients were successfully treated withSTPP.⁷⁷ These results are highly relevant forclinical practice. This is true for another result as well. Comparing theeffect sizes yielded at the end of therapy and follow-up, the effects of STPPproved to be stable and even tended to increase. This result is consistentwith that reported by Anderson and Lambert,⁷ whofound a slight superiority of STPP at follow-up assessment. The effects ofCBT also were maintained to a high degree at follow-up.

For a valid evaluation of the efficacy of STPP, follow-up assessmentsseem to be necessary, as can be illustrated by the following example. In 2RCTs, significant improvements in bulimia nervosa after STPP were found.^45- 46,78 In the central measuresspecific to bulimia nervosa (ie, bulimic episodes and self-induced vomiting),STPP was as effective as CBT.^45- 46,78 Apartfrom this, CBT was superior to STPP on some other measures of psychopathology.^45- 46 However, a follow-up study of thesample of Fairburn et al⁴⁵ that used a longerfollow-up period found that both forms of therapy (STPP and CBT) proved tobe equally effective and partly superior to a behavioral form of therapy.⁷⁸ Results like this can only be detected by long-termfollow-up studies.

In this meta-analysis, TAU yielded effect sizes between those of psychotherapyand no treatment. These results are consistent with the overall findings accordingto which “the ranking for therapeutic success is generally therapy,placebo, and control (do nothing or wait).”⁷⁵ Treatmentas usual generally can be expected to be superior to placebo and inferiorto psychotherapy. The relatively large effect sizes of TAU found in this meta-analysiscan be attributed to the study of Gowers et al,⁴⁷ whichreported large effect sizes in target problems (1-year follow-up, 1.20; 2-yearfollow-up, 1.71) and measures of social functioning (2-year follow-up, 0.95).These large effect sizes achieved by TAU may be explained by the fact thatsome of the control patients of this study had “received extensive treatmentelsewhere.”^47(p165) In thisstudy, patients with anorexia nervosa were treated, and outcome was assessedby weight gain and physical, nutritional, social, and sexual adjustment andmental state (Morgan and Russel scores). In this case, an “extensive”TAU, as described by the authors, can be expected to be relatively effective.

The studies reviewed herein refer to RCTs of STPP in specific psychiatricdisorders (Table 1). However, for somespecific psychiatric disorders, there are no RCTs of STPP at all. This istrue for dissociative disorders, and for some specific forms of personalitydisorders (eg, compulsive, avoidant, and narcissistic). This is also truefor CBT; however, to our knowledge, there are no studies of CBT in the treatmentof dissociative and conversion disorder. With regard to personality disorders,only studies of CBT in avoidant and borderline personality disorder exist.^79- 80 Furthermore, for some forms of anxietydisorders, RCTs of STPP are needed. For panic disorder and agoraphobia, onlyRCTs in which STPP was combined with pharmacological treatment exist.^81- 83 This is also surprising,as anxiety is one of the central concepts of psychoanalytic and psychodynamictheory and therapy.⁸⁴ With regard to generalizedanxiety disorder, the study of Durham et al⁸⁵ comparingSTPP and CBT did not fulfill the inclusion criteria of this meta-analysis(no manual for STPP, no specific training of therapists, no checks of adherenceto and competence for STPP). In addition, STPP and CBT were not equally carefullyperformed. In that study, STPP served as a kind of control group or a “strawman.” As Smith et al⁸⁶ put it:

A comparison therapy might be set up as a kind of straw man overwhich the favored therapy would prevail. The comparison therapy (often an‘insight therapy’) would be treated with fairly obvious disdainand would be given not much opportunity for success.

This is known as the investigator-allegiance effect.^86- 87 Furthermore,studies of STPP in children and adolescents are urgently needed.⁸⁸ Studiesof STPP can be one aspect of what Kazdin⁸⁸ calleda research agenda for child and adolescent psychotherapy.

In the studies reviewed herein, different models of STPP were applied.It is an interesting question, if and how they empirically differ, as “brandnames of therapy can be misleading.”¹⁰ Studiesaddressing this problem are relevant for consideration if some of the differentmodels of STPP are close enough to be lumped together. Ablon and Jones¹⁰ recently compared CBT and IPT as they were performedin the National Institute of Mental Health Treatment of Depression Study.According to the results, both forms of therapy adhered most strongly to theideal prototype of CBT. In addition, adherence to the CBT prototype yieldedmore positive correlations with outcome measures across both types of treatment.However, STPP was not included in this comparison. Thus, it is not clear howSTPP, CBT, and IPT empirically differ with regard to therapist behavior. Comparingprototypic sessions of different variants of STPP empirically would be a veryinteresting and promising project of research. Other forms of therapy (eg,CBT) should be included.

The studies evaluated in this meta-analysis did not include comparisonswith pharmacological treatments. There are only a few randomized controlledstudies that examined the effects of a combination of STPP and pharmacologicaltreatments. In the RCT of Zitrin et al⁸² andKlein et al,⁸¹ STPP and CBT were combined withimipramine hydrochloride in the treatment of phobias. According to the results,STPP and CBT were equally effective in the treatment of agoraphobia, mixedphobia, and simple phobia.⁸¹ In this study,however, no manual of STPP was used. In the study of Wiborg and Dahl,⁸³ STPP combined with clomipramine hydrochloride wasmore successful in the treatment of panic disorder at follow-up than an exclusivetreatment with clomipramine. In this study, 80% of the sample had panic disorderand agoraphobia. In 2 studies, the combined treatment of STPP and antidepressiveswas more successful than the pharmacological treatment alone.^89- 90 However,in the study reported by de Jonghe et al,⁸⁹ andKool et al,⁹¹ this was true only for patientswith a comorbid personality disorder. In the mega-analysis of DeRubeis etal,⁹² CBT was as effective as pharmacologicaltherapy in the treatment of severe depression. According to the mega-analysisof Thase et al,⁹³ the combined treatment ofpsychotherapy (CBT or IPT) and antidepressives was more successful in severedepression than was psychotherapy alone. In milder depression, psychotherapyand the combined treatment were equally effective. With regard to STPP, furtherstudies of STPP are necessary that examine the combined treatment comparedwith STPP alone and pharmacological treatment alone.

According to the results of this and other meta-analyses, STPP is aneffective treatment of psychiatric disorders and yields stable results. Furtherresearch should study not only the effects of STPP in specific psychiatricdisorders but also the active ingredients of STPP. In addition, the activeingredients of STPP should be compared with other forms of therapy, eg, CBT.Furthermore, effectiveness studies should address whether the methods of therapythat have proved to work in RCTs are effective in the field. Data on healtheconomics also should be included.

Correspondence: Falk Leichsenring, DSc,Department of Psychosomatics and Psychotherapy, University of Goettingen,von Siebold Str 5, 37075 Goettingen, Germany (fleichs@gwdg.de).

Submitted for Publication: March 9, 2004; finalrevision received May 25, 2004; accepted June 9, 2004.