Journal of Consulting and Clinical Psychology1984, Vol. 52, No. 2, 268-279

Copyright 1984 by theAmerican Psychological Association, Inc.

A Comparative Study of Four Forms of Psychotherapy

William E. PiperMcGill University and Allan Memorial Institute

Montreal, Quebec, Canada

Elie G. Debbane and J. P. BienvenuAllan Memorial Institute, Montreal, Quebec,

Canada

Jacques GarantHopital Pierre Boucher, Longeuil, Quebec, Canada

Four forms of psychoanalytically oriented psychotherapy involving experiencedtherapists and psychiatric outpatients were compared in an outcome investigation.Patients received either individual therapy or group therapy that lasted either 6months or 24 months. A comprehensive set of outcome scores provided by patients,therapists, and an independent assessor was monitored each 6 months including a6-month follow-up assessment. Therapy outcome, therapy process as viewed bythe therapists, and cost-effectiveness were examined. What emerged as importantwas the particular form of therapy received, not the general type of therapy or thegeneral duration of therapy when considered as independent dimensions. The resultsfavored long-term group therapy and short-term individual therapy over long-termindividual therapy and short-term group therapy. Advantages and difficulties as-sociated with each form of therapy are discussed.

After providing individual and group psy-chotherapy for outpatients in our unit for sev-eral years, we decided to conduct a study thatcompared four forms of psychotherapy. Therewere two types of therapy, individual andgroup, and two time durations, 6 and 24months. These two main dimensions (type andduration) were combined to produce fourforms of therapy: short-term individual (STI),short-term group (STG), long-term individual(LTI), and long-term group (LTG). We wereinterested in studying the work of experiencedtherapists rather than those who were inex-perienced or still in training. Each therapistin our study treated one set of patients witheach of the four forms of therapy. We werealso interested in assessing patients after a rea-sonable follow-up period. Therefore, patientswere assessed 6 months after treatment endedin addition to before, during, and just after

This research was funded in part by research grant 6605-1456-44 from the Health Services and Promotion Branch,Health and Welfare, Canada.

Maria K. Ramsay served as the independent assessorof the patients.

Requests for reprints should be sent to William E. Piper,Department of Psychology, Stewart Biological SciencesBuilding, McGill University, 1205 Docteur Penfield Av-enue, Montreal, Quebec, Canada H3A 1B1.

treatment ended. Initial assessments began inJuly, 1977, and follow-up assessments endedin February, 1982.

In planning the project we were aware ofseveral issues that have characterized the psy-chotherapy field. First, reviews of comparativestudies (Luborsky, Singer, & Luborsky, 1975;Smith, Glass, & Miller, 1980) have made itclear that not many comparative studies con-trasting individual versus group therapy orshort-term versus long-term therapy have beenconducted, particularly those that are strongmethodologically. From existing studies re-viewers have concluded that different forms oftherapy have produced rather similar results.In the terminology of Luborsky and his col-leagues, "tie results" characterize the box scorecomparisons between individual and grouptherapy and between time-limited and time-unlimited therapy. In the terminology of Smithand her colleagues, there is no significant re-lationship between type of modality and effectsize or between duration of therapy and effectsize. Nonsignificant results such as these usu-ally raise more questions than they answer. Inthis case they leave readers wondering whethertype of modality or duration makes a differ-ence or whether other factors were responsiblefor the lack of significant findings.

268

COMPARATIVE STUDY OF PSYCHOTHERAPY 269

Second, we were aware of growing interestin the development of "brief" forms of therapy.Interest has spread from the areas of behaviortherapy and crisis intervention to more tra-ditional forms of psychotherapy. A number ofrecent books concerning short-term therapyhave emerged (Budman, 1981; Davanloo,1978; Malan, 1976; Sifneos, 1979; Small,1979). In support of brief therapy it has beenargued that a predetermined, short-term du-ration accelerates the pace of work, that themost significant gains come early in the courseof therapy, and that short-term therapy canlessen overdependency on the therapist. Al-though many of the arguments are compellingand deserve attention, the data base on whichthey rest resembles more a collection of con-vincing case presentations than a series of sys-tematic studies.

Third, we were aware of a growing interestin developing more economical forms of treat-ment. In this context choices between indi-vidual and group therapy and between short-term and long-term therapy take on economicimportance. The concept of cost-effectivenesshas become a familiar one. The concept in-dicates that both cost and quality of care areimportant in evaluating treatment. As moreeconomical treatments are attempted, it isreasonable to expect that at times minimalstandards for quality will not be attained. Notall treatments can be expected to maintaintheir effectiveness in abbreviated forms. Qual-ity of care needs to be considered as carefullyas cost in studies that compare various formsof treatment.

Interest in studying our own clinical workmore carefully and interest in obtaining datarelevant to such issues as the relative efficacyof therapies that differ in modality and du-ration, the advantages and disadvantages withbrief therapies, and the concept of cost-effec-tiveness, resulted in the current research proj-ect. The present article focuses primarily onthe outcome differences among the four formsof therapy.

Method

Patients

Patients were referred for outpatient therapy to the In-dividual and Group Psychotherapy Unit of the Allan Me-morial Institute in Montreal. The patient composition

consisted primarily of patients suffering from neurotic ormild-to-moderate characterological problems. They typ-ically presented outpatient complaints concerning anxiety,depression, low self-esteem, and difficulties with interper-sonal relationships. In terms of the Diagnostic and Sta-tistical Manual of Mental Disorders (DSM-II; AmericanPsychiatric Association, 1968), 66% of the patients receiveda diagnosis of neurosis and 30% received a diagnosis ofpersonality disorder. Patients manifesting problems ofpsychosis, addiction, sexual deviation, or sociopathic be-havior were excluded. The patients were discouraged fromtaking psychotropic medication and none was prescribedby the therapists. One hundred and six patients were ac-cepted and began treatment with one of the four formsof therapy. Their average age was 32 years (SD = 8.5;range = 18 to 56). Sixty-two percent were women. Notquite half the population was single (45%), about a thirdwas divorced or separated (32%), and about a quarter wasmarried (23%). Eighty-five percent of the patients wereemployed. Their average formal education was 13.1 years.

Therapists

The therapists were three male psychiatrists who had12, 6, and 2 years of postboard (Fellow, Royal College ofPhysicians, Canada) experience treating patients with in-dividual and group psychotherapy. Each had experiencewith the four forms of therapy being studied. All threewere psychodynamically oriented, two having receivedformal training from a psychoanalytic institute. For eachtherapist the practice of outpatient psychotherapy repre-sented a major portion of his professional work. Eachtherapist served as his own control across the four formsof therapy.

Therapy

Psychoanalytically oriented psychotherapy was thetreatment orientation. The techniques of clarification, in-terpretation, and exploration were central. The therapistattempted to clarify the patient's underlying conflicts asdifferentiated from his or her stated conflicts, which wereoften expressed in complaints about relationships andthrough symptoms. The therapist interpreted the patient'sunderlying conflicts in terms of dynamic components, thatis, in terms of conflict between a wish (libidinal or ag-gressive), that was considered unacceptable, and mobilizedanxiety that set up the operation of defenses against thewish. The therapist also attempted to explore the patient'sconflicts as traced through his or her current relationshipsoutside of therapy, his or her immediate relationship withthe therapist, and the patient's past (early, parental) re-lationships. Short-term individual therapy, a relatively re-cent form of analytic therapy, followed the technique asdescribed by Malan (1976). Individual therapy consistedof one 55-min session per week and group therapy of one90-min session per week. Although technical manuals werenot constructed, a considerable amount of time was spentprior to the project discussing the technical applicationof the four forms of therapy. Remainers in the long-termforms of therapy attended an average of 76 sessions andremainers in the short-term forms attended an average of22 sessions for a ratio of 3.5 to 1.

270 PIPER, DEBBANE, BIENVENU, AND GARANT

Procedure

Patients were interviewed by a staff psychiatrist andaccepted or rejected for psychotherapy. The psychiatristrated each patient on several selection criteria includingverbal skill, chronicity of problems, psychological mind-edness, motivation for individual therapy, and motivationfor group therapy. Patients had to be willing to accepteither individual or group therapy. During the first 2 yearsof the project patients were told that they would be offeredeither LTI or LTG therapy and that a final decision wouldawait additional assessment. During the third and fourthyears of the project patients were told that they would beoffered either STI or STG therapy. Patients were contactedby an independent assessor (psychologist) who obtainedtheir informed consent and conducted a standard assess-ment interview and questionnaire battery concerning theoutcome variables of the study. Assessor blindness to thepatient's therapy assignment (individual or group) wasguaranteed because the decision was made after the as-sessment. Patients were matched in pairs on the basis ofage, sex, and the composite sum of the five selection criteriaratings into two sets of 7 or 8 patients. Patients were thenrandomly assigned to either the individual or group formof therapy. Thus, assignment to therapy was not completelyrandom across the four forms, but only between LTI andLTG therapy and between STI and STG therapy. Thetherapists met with each patient individually for one ortwo sessions as the final preliminary step before treatment.Patients were re-evaluated by the independent assessor each6 months. The last evaluation occurred 6 months aftertreatment ended. Dropouts (unilateral premature termi-nators) were replaced if they occurred within the firstquarter of therapy. All therapy sessions were audiotaped.

Measures

Outcome measures were selected to cover interpersonalfunctioning, traditional psychiatric symptomatology, andpersonal target objectives. The three sources of evaluationwere the patient, the therapist, and the independent assessor.The outcome scores are listed according to source in Table1. Some measures were provided by all three sources,whereas others were unique to each source. Most wereassessed as current status scores, although a few were as-sessed as rated benefit scores.

The first two scores provided by the patient were derivedfrom the Interpersonal Behavior Scale (IBS), a 30-itemquestionnaire, which focuses on the degree to which thepatient engages in a number of positive behaviors withpeople he or she interacts with regularly (Piper, Debbane,& Garant, 1977). The first score (IBSP) represents thepatient's perception of his or her present level of functioningand the second score (IBSD) represents the discrepancybetween the patient's present level and preferred level. Thethird score (CAT) was Cattell's H Scale, a 13-itcm subscaleof the Sixteen Personality Factor Questionnaire that focuseson social shyness and inhibition (Cattell & Eber, 1956).The fourth score (COR) was the total score of the CornellIndex, a 100-item symptom checklist heavily representedby psychosomatic content (Weider, Wolff, Brodman, Mit-tleman, & Wechsler, 1948). The fifth through eighth scoresconcern the patient's target objectives. The patient, withthe assistance of the independent assessor, attempted to

formulate a written set of several, clear, realistic, and non-overlapping objectives. The average rating of severity ofdisturbance (TSP) and the average rating of change (TCP)for the objectives were calculated. In addition, because thepatient also rated the relative importance of each objective,ratings of severity (TSPI) and change (TCPI) for the mostimportant objective were also used as scores. The ninthscore was a global rating of the overall usefulness of ther-apy (OUP).

The therapist was given the written list of target objectivesand was also asked to provide severity of disturbance andchange ratings. A parallel set of four scores were calculated(TST, TSTI, TCT, TCTI). The therapist also provided hisglobal rating of the overall usefulness of therapy (OUT).Nine scores were provided by the independent assessor.The first two scores were parallel ratings of severity ofdisturbance for the target objectives (TSIA, TSIAI). Thethird score (DA) was the Depression-Anxiety subscale ofthe Psychiatric Status Schedule (Spitzer, Endicott, & Cohen,1967), which was presented as a structured interview, au-diotaped, and then rated. Two independent raters achievedan average interrater reliability of 89.4% perfect agreementfor 20 randomly chosen patients (range = 75% to 100%).The fourth through ninth scores were the subscale scoresand the average of the subscale scores for a modificationof the Structured and Scaled Interview to Assess Mal-adjustment (SSIAM), a structured interview that assessesinterpersonal functioning in five areas: work-school (W),friends (FR), family of origin (FO), partner-children (PC),and sexual adjustment (SA; Gurland, Yorkston, Frank, &P'liess, 1972). This interview was also audiotaped and rated.Pearson product-moment correlations were calculated forthe ratings of two independent raters for the same set of20 randomly chosen patients. The average correlation was.85, with a range from r(18) = .59, p < .01, to r(18) =.98, p < .001. Prior to conducting the major statisticalanalyses with the 23 outcome scores, a correlation matrixwas conducted with the remainers' pre- to end-of-therapychange scores or end-of-therapy postscores depending onthe measure. Only a relatively small number of the cor-relations was high (r > .50), and these involved scores thatwere intuitively similar in content. On the basis of thispreliminary correlation matrix it was concluded that about10 relatively independent outcome variables were repre-sented by the 23 outcome scores. The 10 variables werebest represented by IBSD, CAT, COR, TSP, OUP, TST,TCT, DA, SSIAM, and TSIA. Thus, when multivariateoutcome analyses were conducted only these 10 were en-tered, and when interpreting univariate outcome analysesthe notion of 10 relatively independent variables ratherthan 23 was emphasized.

Results

Overall, there were 79 remainers (75%) and27 dropouts (25%). The numbers and per-centages for each form of therapy are presentedin Table 2. Chi-square analyses indicated nosignificant differences in number of dropoutswithin the main dimensions of type of therapy,duration of therapy, therapist, or within formof therapy. Remainers did not significantly dif-

COMPARATIVE STUDY OF PSYCHOTHERAPY 271

Table 1Outcome Score Means at the Three Assessment Times

Short-term individual

Score

PatientIBSPIBSDCAT"CORTSPTSPITCP"TCPI"OUP

TherapistTSTTSTITCT"TCTI"OUT"

Independentassessor

TSIATSIAIDAWFRFOPCSASSIAM

Pre

125.241.324.826.73.84.1

3.94.1

4.44.3

13,43.73.83.45.54.03.9

Post

130.335.726.820.82.83.08.68.65.0

3.33.38.89.04.2

2.22.68.83.23.53.64.33.43.6

FU

128.037.627.019.52.63.08.78.85.2

2.52.59.03.12.63.23.03.63.4

Short-term group

Pre

123.042.320.629.04.04.1

4.34.4

4.44.6

12.34.23.83.14.83.93.8

Post

125.536.422.324.6

3.43.67.97.83.6

3.93.97.06.92.6

3.33.4

11.24.13.93.84.63.83.9

FU

119.240.323.821.9

2.62.78.18.03.3

2.72.78.53.23.23.83.63.13.4

Long-term individual

Pre

127.437.824.622.2

3.53.6

3.93.8

4.24.3

10.63.53.73.94.64.23.9

Post

131.432.825.616.62.83.48.38.44.6

3.43.58.89.03.8

2.32.49.23.43.53.43.63.43.4

FU

124.637.525.315.22.32.68.68.54.4

2.02.27.73.23.03.42.93.33.2

Long-term group

Pre

108.949.421.430.7

3.94.3

4.34.4

4.44.4

12.43.94.83.04.24.74.1

Post

124.532.826.219.22.22.29.09.15.0

3.74.08.48.43.5

2.62.68.73.44.53.32.94.03.8

FU

123.731.425.718.62.22.39.09.04.7

1.51.47.33.13.13.62.52.93.1

Note. IBSP = Interpersonal Behavior Scale (present functioning); IBSD = Interpersonal Behavior Scale (discrepancybetween present and ideal functioning); CAT = Cattell's H Scale; COR = Cornell Index; TSP, TSPI = severity for alltarget objectives and most important objective; TCP, TCPI = change for all target objectives and most importantobjective; OUP = overall usefulness of therapy; TST, TSTI = severity for all target objectives and most importantobjective; TCT, TCTI = change for all target objectives and most important objective; OUT = overall usefulness oftherapy; TSIA, TSIAI = severity for all target objectives and most important objective; DA = depression-anxietysubscale; W, FR, FO, PC, SA, SSIAM = work, friends, family of origin, partner-children, sexual adjustment, andmean subscale scores for Structured and Scaled Interview to Assess Maladjustment (SSIAM), respectively. Pre = prescore,Post = postscore, FU = follow-up.a High scores are desirable.

fer from dropouts on the five demographiccharacteristics of age, sex, marital status, em-ployment, and education. The overall designof the project was also well-balanced on thesefive demographic characteristics. Statisticalanalyses (chi-square; analysis of variance, AN-OVA; and t test) revealed no significant differ-ences within the main dimensions of type, du-ration, and therapist, or within form of therapy.This was true for all 106 patients who startedtherapy and for the 79 remainers on whomthe outcome analyses were based.

Before-Therapy to End-of-Therapy Analyses

A three factor ( 2 X 2 X 3 ) analysis of co-variance (ANCOVA) or ANOVA was conductedwith all but one of the 23 outcome scores. Thethree factors were type of therapy (type), du-ration of therapy (term), and therapist (ther).The sample size used in the analysis for thetherapist factor was the number of patients,not the number of therapists. It is commonin therapy outcome research to find significantlinear relationships between prescores (before

272 PIPER, DEBBANE, BIENVENU, AND GARANT

Table 2Number and Percentage ofRemainers and Dropouts

Therapist

Short-termindividual

R D

Short-termgroup

R D

Long-termindividual

R D

Long-termgroup

R D

Note. R = remainers; D = dropouts.

Subtotal

R D

123

Subtotal%

885

2178

123

622

775

1983

013

417

767

2067

532

1033

676

1973

322

727

282823

7975

9810

2725

therapy) and postscores (end of therapy). Thiswas the case for 12 of the 17 outcome measuresthat had both prescores and postscores. For11 of these scores an ANCOVA was conductedusing the prescores as covariates and the post-scores as dependent scores. In the case of the12th score (FO) the slopes of the regressionequations for the four forms of therapy werenot homogeneous. Therefore a four-factor re-peated-measures ANOVA was conducted wherethe fourth factor was the prescore-postscoredimension. For the other five measures thathad prescores and postscores an ANOVA wasused with the postscores because there wereno significant relationships between prescoresand postscores. For the remaining six measuresthat were assessed only as postscores ANOVASwere conducted. One other question that wasaddressed prior to performing the major anal-yses was whether there were significant pre-

Table 3MANOVA for Before-Therapy to End-oj-TherapyOutcome Scores"

Effect Fbdf

TypeTermTherType X TermType X TherTerm X TherType X Term X Ther

3.310.681.981.891.431.830.94

10, 4510, 4520,9210,4520,9220,9210,45

.01ns.02.08ns.03ns

" The ten scores entered were IBSD, CAT, COR, TSP,OUP, TST, TCT, TSIA, DA, and SSIAM. (For an expla-nation of these abbreviations see the note to Table 1.)Ther = therapist.b Pillais approximate F.

score differences on the outcome scores amongthe four forms of therapy. A series of one-factor (forms) ANOVAS on the 17 prescores re-vealed that only one was significant at p <.03. This was the TSIAI score, which was an-alyzed with an ANCOVA. In addition to theunivariate analyses, a three-factor multivariateanalysis of variance (MANCOVA) was conductedon the subset of 10 outcome scores that werefound to be independent. The multivariate re-sults are presented in Table 3.

In interpreting the major analyses severalconventional guidelines were followed. In thecase of a score that had both a significant maineffect and a significant interaction effect thatincluded the main effect factor, the interactioneffect was always emphasized. In examiningoutcome scores for a given effect, the numberof significant (p = .05) scores that one couldexpect due to a Type I error, was consideredas well as the corresponding MANOVA effect.Finally, the absolute level of each outcomescore was considered by examining where themean ratings were located on the rating scalesand by examining prescore to postscore t tests.

Evidence for interaction effects. There wasno significant evidence for a triple (Type XTerm X Ther) interaction effect at either theunivariate or multivariate levels of analysis.In contrast, there was significant evidence fora Type X Term interaction effect. This effectrepresents the importance of the form of ther-apy. Three of the nine patient-rated scores weresignificant, CAT, F(l, 66) = 4.59, p < .04;TSPI, F(l, 67) = 3.95, p = .05; OUP, F(l,63) = 5.42, p < .03, and one approached sig-nificance, TCP, F(l, 67) = 3.15,p< .09. Threeof the five therapist-rated scores were signifi-

COMPARATIVE STUDY OF PSYCHOTHERAPY 273

cant, TCT, F(l, 66) = 8.64, p < .01; TCTI,F(l,66) = 7.82,£<.01;andOUT,F(l,67) =6.07, p < .02. One of the independent assessor-rated scores approached significance, DA, F(l,66) = 3.03, p < .09. Thus, out of 23 univariatescores, six evidenced a significant Type X Terminteraction effect and two approached signif-icance. Multiple comparison tests conductedwith the significant scores (see Table 4) re-vealed a consistent pattern. From the patient'sperspective LTG therapy was rated best, STItherapy was a close second, LTI therapy wasthird, and STG therapy was a distant fourth.From the therapist's perspective STI therapywas rated best, followed closely by LTI andLTG therapy, and STG therapy was similarlya distant fourth. There were usually significantdifferences between STG therapy and LTGtherapy, between STG therapy and STI ther-apy, and in a few cases significant differencesbetween STG and LTI therapy. In addition tothe univariate effects, the multivariate, Type XTerm interaction effect approached signifi-cance, F(10, 45) = 1.89, p < .08.

Next, considering evidence for a Term XTher interaction effect, the MANOVA effect wassignificant, F(20, 92) = 1.83, p < .03. At theunivariate level two of the independent asses-sor-rated scores were significant, SSIAM, F(2,66) = 4.42, p < .02; FO, F(2, 67) = 3.72, p <.03, and two of the patient-rated scores ap-proached significance, TCP, F(2, 67) = 2.68,p < .08; TCPI, F(2, 67) = 2.90, p < .07. Thus,two of 23 univariate scores were significant,two approached significance, and the multi-variate effect was significant. In an effort tounderstand this interaction effect we raised twoquestions. Which therapist(s) was associatedwith better outcome for short-term or long-term therapy? Was there a consistent outcomeorder among the therapists? Unfortunately, theanswers to these two questions varied fromscore to score. This made it impossible to de-duce what therapist characteristic might beinteracting with length of therapy. For ex-ample, in regard to the characteristic of ther-apist experience, in the case of SSIAM, the mostexperienced therapist and the least experiencedtherapist had better results with long-termtherapy, whereas the intermediately experi-enced therapist had better results with short-term therapy. However, in the case of TCP,the most experienced therapist and the inter-

Table 4Multiple Comparisons Among the Four Formsof Therapy

Score

Patient-ratedscores

CAT

OUP

TSPI

TypeXTerm

interactionP

.04

.03

.05

Form

LTGSTILTISTG

LTGSTILTISTG

LTGSTILTISTG

Rank

1234

1.51.534

1234

Post-scoremean"

26.6,26.32

25.322.8,,2

5.0,5.02

4.63.6,,2

2.4,,23.03.4,3.52

Therapist-ratedscores

TCT

TCTI

OUT

.01

.01

.02

STI 1.5LTI 1.5LTG 3STG 4

STILTILTGSTG

STILTILTGSTG

8.8,8.82

8.43

9.0,8.92

8.43

6.9,,2,3

4.2,3.82

3.52

2.61,2.3

Note. Identical subscripts for two means indicate a sig-nificant (p < .05) difference using Duncan's multiple-rangetest. CAT = Cattail's H Scale; OUP = overall usefulnessof therapy; TSPI = severity for most important targetobjective; TCT, TCTI = change for all target objectivesand most important objective; OUT = overall usefulnessof therapy; LTG = long-term group; STI = short-termindividual; LTI = long-term individual; STG = short-termgroup." Adjusted ANCOVA postscore means are reported for CAT.

mediately experienced therapist had better re-sults with long-term therapy, whereas the leastexperienced therapist had better results withshort-term therapy. Such examples led to theconclusion that the significant, multivariateTerm X Ther effect was reflecting a set of in-consistent univariate effects. Finally, consid-

274 PIPER, DEBBANE, BIENVENU, AND GARANT

Table 5MANOVA for Before-Thempy to Follow-UpOutcome Scores"

Effect df

TypeTermTherType X TermType X TherTerm X TherType X Term X Ther

1.921.191.351.340.490.711.15

8, 328, 32

16,668,32

16, 6616,668,32

.09nsnsnsnsnsns

" The eight scores entered were IBSD, CAT, COR, TSP,OUP, TSIA, DA, and SSIAM. (For an explanation of theseabbreviations see note to Table 1.) Ther = therapist.b Pillais approximate F.

ering evidence for a Type X Ther interactioneffect, only one score was significant, TCPI,F(2, 67) = 5.80, p < .01, and one score ap-proached significance, TST, F(2, 65) = 2.66,p < .08.

Evidence for main effects. Nearly all evi-dence for significant main effects (univariateand multivariate) was overshadowed by sig-nificant interaction effects, particularly theType X Term interaction effect. The only ex-ception for a significant type main effect wasFO, F(l, 67) = 5.57,p< .03, where individualtherapy (pre-post mean decrease = 0.13) wasfavored over group therapy (pre-post meanincrease = 0.47). There were no exceptionsfor a significant term main effect. The onlyexception for a significant ther main effect wasOUT, F(2, 67) = 5.53, p < .01, where themeans in order of increasing therapist expe-rience were 3.9, 2.9, and 3.8. Overall, the ev-idence for main effects was quite minimal.

Evidence for improvement. Examination ofprescore to postscore t tests for the four formsof therapy indicated an overall pattern of pa-tient improvement. Eight of the 10 relativelyindependent scores used in the MANOVA hadprescores and postscores. Significant improve-ment was found on six of these scores for LTGtherapy, IBSD, t(\8) = 2.74, p < .02; CAT,r(18) = 3.31, p < .01; COR, r(18) = 4.07, p <.001; DA, ?(18) = 2.22, p < .05; TSP, /(11) =4.52, p < .001; and TSIA, *(18) = 6.59, p <.001, on 4 of these scores for STI therapy,COR, f(20) = 3.19, p < .01; DA, f(20) = 4.42,p < .001; TSP, *(20) = 3.38, p< .01; andTSIA, f(20) = 8.53, p < .001, on 3 of these

scores for LTI therapy, COR, t( 1 9) = 2.52, p <.03; TSP, /(1 4) = 2.20, p < .05; and TSIA,/(1 9) = 7.63, p < .001, and on 3 of these scoresfor STG therapy, TSP, J(18) = 2.33, p < .04;TST, t(l%) = p< .04; and TSIA, t(\S) = 3.30,p < .01. Evidence for significant worseningwas found on one score for STG therapy, FO,

= 2.60, p < .02.

Before-Therapy to Follow-Up Analyses

Sixty-nine of the 79 remainers (87%) pro-vided data at the 6-month follow-up assess-ment. The percentages of patients lost at fol-low-up for the four forms of therapy were 5%(STI), 16% (STG), 10% (LTI), and 21% (LTG).Only a few patients reported taking psycho-tropic medication or participating in formaltherapy during the follow-up period. Outcomescores were available from the patient and theindependent assessor, but not the therapist,who had no contact with the patient duringthe follow-up period. The follow-up data wereanalyzed in the same way as the end of therapydata. For the MANOVA, eight of the 10, pre-viously used, independent scores were enteredbecause the two therapist scores were notavailable. The multivariate results are pre-sented in Table 5.

Evidence for interaction effects. No signif-icant MANOVA interaction effects were found.At the univariate level two significant triple(Type X Term X Ther) interaction effects werefound, TCP, F(2, 55) = 3.85, p < .03, andTCPI, F(2, 55) = 3.94, p < .03. For bothscores patients of the most experienced ther-apist reported greater improvement for LTItherapy over STI therapy and STG therapyover LTG therapy, whereas the patients of thelesser experienced therapists reported the re-verse. In the case of the Type X Term inter-action, only one significant effect, OUP, F( 1 ,55) = 10.36, p < .01, and one near significanteffect, IBSD, F(l, 55) = 3.60, p < .07, werefound. The pattern was the same as that foundwith the end of therapy scores. The worst out-come was associated with STG therapy andthe best outcome with LTG therapy and STItherapy. No significant Term X Ther inter-action effects or Type X Ther interaction effectswere found.

Evidence for main effects. Similar to thebefore- to end-of-therapy data nearly all evi-

COMPARATIVE STUDY OF PSYCHOTHERAPY 275

dence for significant main effects (univariateand multivariate) was overshadowed by sig-nificant interaction effects. The only exceptionfor a significant type main effect was FO, F( 1,53) = 8.27, p < .01, where individual therapy(pre-follow-up mean decrease = 0.37) was fa-vored over group therapy (pre-follow-up meanincrease = 0.72). There were no exceptionsfor a significant ther main effect. The two ex-ceptions for a significant term main effect wereTSIA, F(l, 53) = 9.93, p < .01, and TSIAI,F(l, 52) = 8.86, p < .01, where long-termtherapy (pre-follow-up mean decreases = 2.55,2.58) was favored over short-term therapy (pre-follow-up mean decreases = 1.79, 1.88). Over-all, the evidence for main effects was quiteminimal.

Evidence for continued improvement. Four-factor (Type X Term X Ther X Trials) ANOVAS,where trials referred to the end of therapy tofollow-up assessment dimension, were con-ducted with each outcome score. Significanttrial effects indicating continued improvementemerged for three scores, DA, F(l, 53) = 5.33,p < .03; SSIAM, F(l, 53) = 11.96, p < .01;andTSP,jF(l,55) = 4.17,p<.05.Threeotherscores approached significance, COR, F(l,55) = 3.64, p < .07; W, F(l, 44) = 4.03, p <.06; and SA, F(l, 52) = 3.04, p < .09. Thus,three of 18 trials effects were significant andthree approached significance. Few other sig-nificant effects were found, including no sig-nificant Type X Term X Trials effect. The pat-tern was clear. Patients tended to either main-tain their end of therapy scores or evidenceadditional improvement.

Discussion

The outcome analyses indicated minimalevidence of main effects for type of therapyor duration of therapy, which is consistent withthe findings of previous studies. However, thecurrent study provided considerable evidencefor a Type X Duration (term) interaction effect,which highlights the importance of these twodimensions when taken in combination. Inthe case of group therapy the outcome resultsstrongly favored the long-term form, whereasin the case of individual therapy the outcomeresults favored the short-term form. The bestoutcomes were associated with LTG therapyand STI therapy and the worst outcomes with

STG therapy. Stated in somewhat differentterms, what was found to be important wasthe particular form of therapy received, notthe general type (individual, group) or thegeneral duration (short-term, long-term) oftherapy.

The pattern that emerged, which empha-sized the interaction between type of outcomeand duration of therapy, was strongly evidentat the time that therapy ended. The patternwas also present, but in considerably weakerform, at the time of follow-up. Examinationof the follow-up mean scores indicated almostwithout exception that STG patients had thepoorest scores. They, like most patients, con-tinued to improve during the follow-up period.However, their outcome scores remained atthe bottom relative to the other patients. Lesssignificant effects at follow-up is not an un-common finding in therapy outcome studies.Although follow-up data is important to ex-amine, it is often beset with ambiguities. Inthe present study there were differential lossesamong the four forms of therapy, ranging from5% (STI) to 21% (LTG). How the missing pa-tients would have scored is unknown. The factthat therapist outcome ratings were not avail-able also limits the strength of the follow-updata.

Although the poorest outcome results wereassociated with STG therapy, it would be amistake to view its outcome effects as disas-trous. Evidence for negative effects was min-imal and the dropout rate for STG therapywas not particularly high. If STG therapy hadbeen studied alone its outcome results wouldhave seemed more favorable. It was only whenthe full range of outcome scores were examinedand when the results for all four forms of ther-apy were considered that the unimpressivequality of the STG therapy outcome resultsbecame clear.

Considerations of Cost-Effectiveness

The four forms of therapy differed substan-tially in the amount of therapist time that wasrequired for each patient. As indicated in Table6 the therapist time per patient ratio for thefour forms of therapy was approximately 1(STG) to 4 (LTG) to 5 (STI) to 20 (LTI). LTItherapy was by far the most costly and STG

276 PIPER, DEBBANE, BIENVENU, AND GARANT

therapy was the most economical of the fourforms of therapy. Time is much easier toquantify and to compare meaningfully in aratio than the concept of effectiveness, whichin the present study was based on an entireset of outcome scores. To represent the effec-tiveness of one form of therapy as a multipleof another is possible but is unappealing dueto the lack of a sound logical basis for deter-mining the ratio. Despite this limitation it ap-pears that several conclusions can be madeconcerning cost-effectiveness.

First, it can be argued that STI therapy andLTG therapy were more cost-effective than LTItherapy. In terms of outcome, STI and LTGtherapy produced results that were as good(usually better) than LTI therapy. In terms oftherapist time STI and LTG therapy requiredabout one quarter of the time. Second, it canbe argued that STI and LTG therapy weremore cost-effective than STG therapy. Thisconclusion is more open to debate becauseproponents of STG tnerapy may argue thatthe low cost of STG therapy and the presenceof some improvement indicate that it is cost-effective (i.e., small investment, small gain).We considered but rejected this conclusion for

several reasons. In absolute terms a therapistinvestment of 36 hours and a time period of6 months is not a small investment. In ad-dition, the magnitude of rated improvementand usefulness was minimal for most outcomescores, not unlike what one could expect tofind among a set of waiting-list control or at-tention-placebo patients.

It is also possible to consider the time re-quired of the patient as a cost. The time re-quired of the patient also differed substantiallyamong the four forms of therapy. This timeratio (see Table 6) was approximately 1 (STI)to 2 (STG) to 4 (LTI) to 7 (LTG). In this casethere is no doubt that STI was the most cost-effective form of therapy. However, becausepatients assigned the best outcome ratings toLTG therapy, they may have felt that the extratime was worth it. In addition, the structuralcharacteristics of STI and LTG therapy wereconsiderably different. Although STI tnerapyoffered a sense of "privacy" and "personal at-tention from the therapist," LTG therapy of-fered a sense of "continuity over time" and"peer relationships." Personal preference mayalso override the issue of how much time isrequired.

Table 6Therapist Time per Patient Ratio and Patient Time Ratio for the Four Forms of Therapy

Form Calculation

Therapist time per patient

STGLTGSTILTI

Therapist time per patient ratio(approximate)

24 sessions X 1.5 hours - 8 patients = 4.5 hours96 sessions X 1.5 hours - 8 patients = 18.0 hours24 sessions X 0.9 hours - 1 patient =21.6 hours96 sessions X 0.9 hours — 1 patient = 86.4 hours

1 (STG) :4 (LTG) :5 (STI) :20 (LTI)

Patient time

STISTGLTILTG

Patient time ratio (approximate)

24 sessions X 0.9 hours = 21.6 hours24 sessions X 1.5 hours = 36.0 hours96 sessions X 0.9 hours = 86.4 hours96 sessions X 1.5 hours = 144.0 hours

1 (STI) :2 (STG) :4 (LTI) :7 (LTG)

Note. Individual sessions lasted 55 min. Group sessions lasted 90 min. Short-term therapy involved 24 sessions. Long-term therapy involved 96 sessions. Groups consisted of 8 patients. STG = short-term group; LTG = long-term group;STI = short-term individual; LTI = long-term individual.

COMPARATIVE STUDY OF PSYCHOTHERAPY 277

Considerations Concerning theProcess of Therapy

The emphasis in the data on the particularform of therapy suggests that each possessedunique features. Short-term individual therapywas not simply less of long-term individualtherapy, and long-term group therapy was notsimply more of short-term group therapy.From the perspective of the therapists a num-ber of distinctive qualities characterized eachform of therapy. Although they were not pres-ent in every case, they were present oftenenough to permit the following general con-ceptions.

Short-term individual therapy. A facilitativeatmosphere of time pressure was common toboth the patient and the therapist. Both partiesfelt the need to work hard and relativelyquickly. Attention was concentrated and fo-cused. Affective involvement was high. Al-though the range of problems explored wasnot extensive there was depth associated withthose that were explored. At completion mostpatients felt that they had received somethingvaluable from the therapist.

Long-term individual therapy. Althoughthe patients seemed satisfied with the structureand process of therapy, the therapists were lesssatisfied. The length of time available coupledwith the frequency of one session per weekseemed to favor an increase in resistance anda decrease in working through. Thus, the pa-tient tended to behave as if there was alwaysplenty of time to work later. Patients dutifullycame to sessions but found it easy to defendagainst affective involvement. They tended tocontrol regression and restrict expression oftransference, which made these processes moredifficult to clarify and interpret.

Short-term group therapy. Both the patientsand the therapists experienced difficulties withthis form of therapy. Its structure appeared toheighten the patients' anxiety about obtainingrelief for their problems to the exclusion ofallowing themselves to engage in an explora-tion of relationship difficulties experienced inthe group. Initial anxiety about working onsensitive issues in the presence of others wassoon followed by anxiety about ending thegroup. An atmosphere of deprivation pre-vailed. The therapists felt burdened with the

task of trying to treat 7-8 patients togetherfor their presenting problems using a psycho-analytically oriented approach in the limitedtime and situation that was available. It is pos-sible that the structure of these groups wouldbe more suitable for other orientations of short-term group therapy, for example those thatconsist of highly structured exercises and/orattempt to minimize rather than arouse anx-iety during sessions. It is also possible that adifferent technical application of psychoana-lytically oriented therapy in short-term groupswould prove to be more successful.

Long-term group therapy. This form oftherapy was characterized by a high degree ofinvolvement and attentiveness by both patientsand therapists. The presence of a group ofpeople provided continual stimulation andthere was ample time to deal with importantissues that were not immediately and directlyassociated with the patient's presenting symp-tomatology. The clinical material that emergedwas viewed by the therapists as relatively richand varied. The frequency and duration ofsessions of this form is characteristic of psy-chodynamic group therapy as it is practicedin North America.

Admittedly the therapists' conceptions ofprocess were provided retrospectively withknowledge of the outcomes. Nevertheless, theyprovide some suggestions about advantagesand difficulties associated with the structureof each of the four forms of therapy that mayhave been related to outcome. The structureof both LTG and STI therapy seemed to fa-cilitate active involvement and attentivenessfor both parties and an optimal mobilizationof affect on the patient's part. The patientsappeared to evidence greater regression, stron-ger transference reactions and less resistance,processes crucial to an interpretive, psycho-analytic approach. In contrast, the structuresof LTI therapy and STG therapy seemed toresult in levels of involvement, attentiveness,and affect that were not optimal. LTI therapywas permeated with a sense of too much timeand STG therapy with a sense of too littletime.

An appropriate question concerning theoutcome differences among the four forms oftherapy is whether they may have been due todifferences in therapist experience, bias, or

278 PIPER, DEBBANE, B1ENVENU, AND GARANT

skill. It would be naive to assume that thetherapists of the present study, or any thera-pists, could have exactly the same experience,bias, or skill for four forms of therapy. Nev-ertheless, we do not believe that these factorsplayed a significant role in the present study.In regard to the experience issue, the threetherapists had previously worked together fora number of years in a unit that provided bothindividual and group therapy of different du-rations. Provision of the four forms of therapydid not appear to be a large step from the workwith which they were quite familiar. In regardto the bias issue, if an initial outcome biasexisted in the form of differing outcome ex-pectations, it probably favored the long-termtherapies. However, the data indicated minimalevidence for a term main effect and the resultsfor LTI therapy and STI therapy actually rancounter to such a bias. A more serious formof bias concerns the possibility that the ther-apists prejudicially wished to demonstrate thesuperiority of one or more of the forms oftherapy. We can only emphasize that we werenot aware of such a bias. The commitment toprovide something useful to each patient wastaken seriously and permeated all forms oftherapy. The issue concerning the possibilityof differential skill is a more open one at thistime because we unfortunately do not haveprocess ratings that indicate whether each ofthe four forms was conducted in an equallyskillful manner.

The therapist dimension per se had a rel-atively small impact (main or interaction ef-fects) on the outcome data. The multivariateresults appeared to mask inconsistent uni-variate effects, which themselves were not thatsubstantial. We believe that the commonalitiesamong the three therapists outweighed the dif-ference. All three were male psychiatrists, whohad completed their training and who sub-scribed to a similar conceptual framework.Although differences in length of experienceand some differences in technical style prob-ably existed, they did not appear to exert amajor impact on the outcomes of the fourforms of therapy.

One other finding that deserves highlightingwas the absence of significant differences indropout rates among the four forms of therapyand in particular between the group and in-

dividual forms. In the past it has been sus-pected that group therapy is associated withhigher dropout rates than individual therapy.High dropout rates for group therapy have beenreported in a number of studies (Baekeland& Lundwall, 1975; Berne, 1955; Scher &Johnson, 1964; Sethna & Harrington, 1971;Nash, Frank, Gliedman, Imber, & Stone,1957). Dropping out often has a demoralizingeffect on the departing patient, the remainingpatients, and the therapist. The potential thatit may precipitate a chain of additional drop-outs that will eventually result in the demiseof the group is real. The potential for seriousmultiple repercussions in a therapy group mayhave resulted in an exaggerated perception ofits incidence relative to individual therapy.

In conclusion, we believe that the presentstudy has provided evidence that favors LTGtherapy and STI therapy relative to LTI therapyand STG therapy. In terms of quality of out-come, LTG therapy and STI therapy receivedthe best outcome ratings, although LTI therapywas not far behind. In terms of cost-effective-ness and the quality of therapy process asviewed by the therapists LTG therapy and STItherapy were regarded as superior. These con-clusions need to be replicated. No single studycan provide definitive answers. If the findingsof the present study hold, they could have sig-nificant implications for the use of professionalresources in the treatment of outpatients withpsychotherapy.

References

American Psychiatric Association. (1968). Diagnostic andstatistical manual of mental disorders (2nd ed.). Wash-ington, DC: Author.

Baekeland, E, & Lundwall, L. (1975). Dropping out oftreatment: A critical review. Psychological Bulletin, 82,738-783.

Berne, E. (1955). Group attendance: Clinical and theo-retical considerations. International Journal of GroupPsychotherapy. 5, 392-403.

Budman, S. H. (Ed.). (1981). Forms of brief therapy. NewYork: Guilford Press.

Cattell, R. B., & Eber, H. W. (1956). Sixteen personalityfactor questionnaire. Champaign, IL: Institute for Per-sonality and Ability Testing.

Davanloo, H. (1978). Principles and techniques of short-term dynamic psychotherapy. New \fork: Spectrum Press.

Gurland, B. J., Yorkston, M. J., Frank, J. D., & Fliess,J. L. (1972). The structured and scaled interview toassess maladjustment (SSIAM): I. Description, rationale

COMPARATIVE STUDY OF PSYCHOTHERAPY 279

and development. Archives of General Psychiatry, 27,259-263.

Luborsky, L., Singer, B., & Luborsky, L. (1975). Com-parative studies of psychotherapies. Archives of GeneralPsychiatry, 32, 995-1008.

Malan, D. H. (1976). The frontier of brief psychotherapy.New York: Plenum Press.

Nash, E. H., Frank, J. D., Gliedman, L. H., Imber, S. D.,& Stone, A. R. (1957). Some factors related to patientsremaining in group psychotherapy. InternationalJournalof Group Psychotherapy. 7, 264-274.

Piper, W. E., Debbane, E. G., & Garant, J. (1977). Anoutcome study of group psychotherapy. Archives of Gen-eral Psychiatry. 34, 1027-1034.

Scher, M., & Johnson, M. H. (1964). Attendance fluctu-ations in an aftercare group. International Journal ofGroup Psychotherapy. 14, 223-224.

Sethna, E. R., & Harrington, J. A. (1971). A study of

patients who lapsed from group psychotherapy. BritishJournal of Psychiatry, 119, 59-69.

Sifneos, P. E. (1979). Short-term dynamic psychotherapy:Evaluation and technique. New \fork: Plenum Press.

Small, L. (1979). The briefer psychotherapies. New"Vbrk:Bruner/Mazel.

Smith, M. L., Glass, G. V., & Miller, T. I. (1980). Thebenefits of psychotherapy. Baltimore, MD: The JohnsHopkins University Press.

Spitzer, R. L., Endicott, J. E., & Cohen, G. M. (1967).Psychiatric status schedule. New Vbrk: New York StateDepartment of Mental Hygiene.

Weider, A., Wolff, H. G., Brodman, K., Mittleman, B., &Wechsler, D. (1948). Cornell index. New York: The Psy-chological Corporation.

Received September 9, 1983Revision received October 11, 1983 •