ParticipantsThe study was conducted at three majorsites; at the Maudsley Trust, London; atAddenbrooke's Hospital Trust, Cambridge;and at Norfolk Mental Health Trust,Norwich. Participants were catchment areaclients who were recruited by asking forreferrals from community teams and inpatient units.

Criteria for referral were: at least onecurrent positive psychotic symptom (such asdelusions or hallucinations) that was distressing, unremitting (at least the past sixmonths) and medication resistant, that ishad not responded to a previous trial of atleast six months of appropriate neurolepticmedication. Clients prescribed clozapineneeded to have been stable on this for atleast one year (to allow time for all benefitto occur). People who had drug, alcohol ororganic problems as primary features wereexcluded.

Proceduresand randomisationOnce referred, all possible participants wereseen for a screening interview by an mdcpendent evaluator to establish whether theymet our criteria. Once this was confirmedand informed consent had been obtained,randomisation was carried out separatelywithin each treatment centre by the trialstatistician (G.D . ), using randomisedpermuted blocking (Pocock, 1983) and ablock size of six. Participants then enteredeither the control condition or the treatmentgroup, and baseline .assessments werecarried out. Considerable efforts were madeto collect data from all participants in thetrial from the time of randomisationonwards.

Prior power calculations, based on theresults of the pilot study (Garety et a!,1994), had indicated that a trial with a totalof 60 people would have a power of at least0.80 to detect an effect sizeof 0.516 using atwo-group t-test with a 0.05 two-tailedsignificance level.

Assessments

A wide range of assessments was adminis.tered to participants at baseline, threemonths, six months, and nine months(which was the end of the treatment condition). A follow-up at 18 months after entryinto the trial is still continuing. This paperwill present only the results of the treatmentphase of the trial. In a trial of psychological

Background A seriesofsmall, mainly

uncontrolled, studies have suggested

thattechniques adapted from cognitive

behavioural therapy (CBT) for depression

can improve outcome in psychosis, but no

large randomised controlled trial of inten

sive treatment for medication-resistant

symptoms ofpsychosis has previously

been published.

Method Sixtyparticipantswhoeachhad at leastone positive and distressing

symptom ofpsychosis that was medication

resistant were randomly allocated between

a CBTand standard care condition (n28)

and a standard care only control condition

(n32).Therapy was individualised, and

lasted for nine months. Multiple assessments

ofoutcome were used.

Results Overninemonths,improvement was significant only in the treatment

group, who showed a 25% reduction on

the BPRS.No otherclinical, symptomatic or

functioning measure changed significantly.

Participants had a low drop-out rate from

therapy (11%),and expressed high levels of

satisfaction with treatment (80%). Fifty per

cent ofthe CBT group were treatment

responders (one person became worse),

compared with 31%ofthe control group

(threepeoplebecameworseandanothercommitted suicide).

Conclusions CBT for psychosiscan

improve overall symptomatology.The

findings provide evidence that even a

refractory group of clients with a long

history of psychosis can engage in talking

about psychotic symptoms and their

meaning, and this can improve outcome.

Traditionally, it has been thought that theexperiences of psychosis were categoricallydifferent from normal experiences.Symptoms such as delusions have beendefined in terms of being unresponsive torational argument (Jaspers, 1963) and thusunamenable to ‘¿�talking'therapies.However, since the 19SOs a small numberof single case studies have indicated that ifspecific techniques are used, talking topeople about their psychotic experiencescan be productive and improve symptomatology (Beck, 1952; Watts et a!, 1973).More recently, other studies have shownthat gentle challenge of evidence,presenting alternative possible viewpoints,reality-testing and enhancing coping strategies may be helpful, particularly for thosewith distressing positive symptoms (Fowler& Morley, 1989; Chadwick & Lowe,1990; Kingdon & Turkington, 1991;Fowler, 1992; Tarrier et a!, 1993; Garetyet a!, 1994; Haddock et a!, 1996), even forthose with acute episodes (Drury et a!,1996).

Despite the promise of the studies sofar published (Bouchard et a!, 1996), therehave been only two randomised controlledtrials (RCTs) (Tarrier et a!, 1993; Drury eta!, 1996) neither of which offered longterm treatment. Therefore, we aimed tooffer cognitive—behavioural therapy (CBT)for a period of nine months to those withpersistent, distressing, medication-resistantsymptoms of psychosis, and to evaluateany changes in outcome using multiplecriteria, and an RCT design. A further aimwas to engage as many clients in therapyas possible, as we did in our earlier study(Garety et a!, 1994). This is becausefailure to engage clients in therapy, whilea well documented problem for this clientgroup, would obviously limit the usefulness of any psychological treatment forpsychosis. In this paper we present theresults of the treatment phase of a threecentre study, based in London and EastAnglia.

319

London—EastAnglia randomised controlled trial of METHOD

cognitive—behaviouraltherapy for psychosis

I: Effects ofthe treatment phase

ELIZABETH KUIPERS, PHILIPPA GARETY, DAVID FOWLER, GRAHAM DUNN,PAUL BEBBINGTON, DANIEL FREEMAN and CLARE HADLEY

treatment it is extremely difficult to makeassessments that are totally blind to thetreatment condition and this was notattempted. However, all assessments werecarried out by independent research workers(D. Freeman and C. Hadley) who were notinvolved in the treatments.

Measures

Our previous study had established thatparticipants were able to cope with ourassessments. These were administered overseveral sessions. We wished to look in detailat symptoms, functioning, and cognitive andemotional variables because we wanted notonly to monitor outcome, but also to findout whether there were predictors of good orpoor treatment response and correlationsbetween outcome measures. These will bereported separately (Garety et a!, 1997).

Symptom and functioning measures

We used the Present State Examination (PSE10; World Health Organization, 1992) toestablish psychotic symptomatology, usingthe associated CATEGO—Vprogramme toderive diagnostic categories according toDSM— HI—R (American PsychiatricAssociation, 1987) at baseline. The BriefPsychiatric Rating Scale (19-item, 0—6scale)(BPRS; Overall & Gorham, 1962) wasadministered to assess overall mental state(baseline and three-monthly). The researchworkers in the separate centres achieved ahigh level of interrater reliability on theBPRS (intra-class correlation coefficient=0.92). We used Personal Questionnaires

every three months to monitor changes inkey symptoms identified by the PSE, as thismethodology has proved to be both reliableand sensitive (Brett-Jones et a!, 1987). Fordelusions we measured conviction, preoccupation and distress; for hallucinations wemeasured frequency, intensity and distress.We also assessed hallucinations (Hustig &Hafner, 1990) three-monthly, and used theMaudsley Assessment of Delusions Schedule(MADS; Buchanan et a!, 1993) (baseline andnine months).

Insight (Amador et a!, 1993) wasmeasured at baseline and at nine months.The Beck Depression Inventory (BDI; Beck eta!, 1961) (three-monthly) the Beck AnxietyInventory (BA!; Beck et a!, 1988) (baselineand nine months), Beck Hopelessness Scale(BHS; Beck et a!, 1974) (baseline and ninemonths), and the Social Functioning Scale(Birchwood et a!, 1990) (baseline and ninemonths) were also completed.

Cognitive measures

The National Adult Reading Test (NART;Nelson, 1982) and the Quick Test (Ammons& Ammons, 1962) are estimates ofpremorbid and current IQ, respectively.Cognitive deficits and biases were investigated using tasks involving cognitive estimations (Shallice & Evans, 1978), verbalfluency (Miller, 1984) and probabilisticreasoning (Garety et a!, 1991). All of thesewere measured at baseline only.

Other measuresThe Self Concept Questionnaire (Robson,1989) is a measure of self-esteem (baseline

and nine months). The DysfunctionalAttitudes Scale investigates underlyingbeliefs about the self (described inWilliams, 1992) (baseline and nine months).The Autobiographical Memory Task(Williams & Dritschel, 1988) (baseline),the Recall of Adjectives Task (Bellew,1990) (baseline), and a Satisfaction with

Therapy Questionnaire (nine months) werealso completed.

Treatment condition

Participants randomised into the treatmentgroup received up to nine months ofindividual CBT for psychosis. Sessions wereconducted weekly initially, and then fortnightly, for up to an hour. Therapy wasdesigned to achieve the following aims:

(a) to reduce the distress and interferencethat can arise from the experience ofpsychotic symptomatology;

(b) to reduce emotional disturbance such asdepression, anxiety and hopelessness,and to modify dysfunctional schemas ifthey existed; and

(c) to promote the active participation ofthe individual in the regulation of theirrisk of relapse and social disability.

The methods used to achieve these aimshave been discussed in detail in our manual(Fowler et a!, 199S). Specific interventionswere individualised from the assessmentphase of the treatment. Initial sessions werefocused on facilitating engagement in treatment. Considerable effort was spent onbuilding and maintaining a good basictherapeutic relationship, and this relationship was characterised by considerable flexibility on the part of the therapist. Whennecessary, treatment was arranged in locations convenient to the client, includinghome visits and proactive outreach following

non-attendance. Within sessions the therapist was highly sensitive to changes in mentalstate and in particular the occurrence ofparanoia. Active attempts were made tomanage such problems so as to ensure thatclients did not feel unduly pressured and toprevent treatment from becoming aversive. Ifnecessary, sessions were cut short or rearranged. Difficult topics were discussedonly when clients felt able to do so. Duringthe early sessions the therapist conducted adetailed analysis of the client's problems.This involved eliciting in detail the client'sinterpretation of the development of theirproblems over time, in particular thedevelopment of delusional ideas and voicesfrom their first emergence, and the client'sappraisals of psychotic experiences occurringin different episodes. The therapist alsoattempted a detailed analysis of the currentproblems that the client had prioritised. Suchanalysis aimed to elucidate triggering factors,current coping responses and the context inwhich psychotic experiences were embedded.Following this, treatment involved the use ofseveral of the following strategies accordingto an individualised case formulation.

Improvingcopingstrategiesanddevelopingandpractisingnewones

This involved using a variety of widelyknown behavioural therapy techniques,including activity scheduling, relaxationand skills training. The aim was to buildon the client's own coping repertoire tomanage current problems. Such techniqueswere used primarily to assist clients toengage in functional activities such as goingshopping or socialising, or to manage thebehavioural consequences of psychoticsymptoms such as impulses deriving fromvoice commands, or self-harm.

Clients' own coping repertoires weredelineated, and they were encouraged touse strategies such as distraction or avoidance more specifically and consistently.Other strategies such as alcohol use andhigh levels of social withdrawal werediscussed in terms of their long-term costs,and discouraged. New strategies such asreading aloud to combat auditory hallucinations were suggested for the client to try outand report back on.

Developinga sharedmodelin collaborationwith the client

Therapists engaged in collaborative discussion with clients about the nature ofpsychotic symptoms and the effects on their

320

lives. Information was offered to enableclients to understand what had happened tothem in the context of current ideas aboutthe basis of psychotic experiences. The aimwas to assist clients to a new interpretationof their problems which had a less distressing meaning for them, and which was morelikely to lead to engaging in healthpromoting behaviour such as taking mcdication. Some clients wished for complexdiscussion of the interaction between biological, social and cognitive factors, otherspreferred using more basic concepts. The

therapists were open-minded about thedegree to which clients preferred or accepted

a medical model of events, and tolerant oftheir rejection of diagnoses. In cases whereclients were particularly resistant to changing delusional beliefs, therapists worked‘¿�within'the delusional system by fosteringless distressing and more functional specificmeanings while not directly tackling thedelusional belief itself.

Modifyingdelusionalbeliefsandbeliefsabouthallucinations

Therapists helped clients to review theevidence for their beliefs. Gentle challengeand the presentation of possible alternativeexplanations were used, together withreality-testing where appropriate. Beliefsheld with less conviction were discussedfirst. Beliefs about hallucinations were

looked at in detail in the same way, butmore particularly, the meaning attributed tovoices was examined. While links betweencurrent voices, especially distressing ones,and earlier incidents in clients' lives were notalways evident, they could sometimes beuncovered and this was often helpful. Ifpossible, voices and other hallucinationswere discussed as internal events, whichwere experienced as real by clients, but werenot part of the experience of others. Iffeasible, reality-testing was undertaken,such as testing out the actual rather thanthe feared consequences of not alwaysobeying a command hallucination.

Modifyingdysfunctionalschemas(Becket aL1979)In the context of a life review (Young, 1990)clients' negative views or dysfunctionalassumptions about themselves were identifled, and the evidence for their veracity wasre-examined in light of a current review oftheir circumstances; for example, was italways true that the individual was a ‘¿�badperson' or ‘¿�worthless'?

Managementofsocialdisabilityandrelapse

This included discussion of relapse signatures(Birchwood, 1996), issues of stigma and theneed to make changes as small as possible,given the clients' history of vulnerability. The

clients' ability to identify triggers that mightexacerbate psychotic phenomena wasdiscussed and they were encouraged to takeappropriate avoidant or coping measuressuch as increased medication, seeking help,supportive discussion or distraction techniques. These issues were often discussed lateron in treatment, after work on specificsymptoms had either been completed or hadproceeded as far as possible.

Therapists

Therapists in the trial were experiencedclinical psychologists. They met regularlyduring the treatment phase (at leastmonthly), either for peer supervision (E.Kuipers, P. Garety, D. Fowler and SteveJones) or for therapy supervision with D.Fowler (Nina Dick, Erik Kuiper, MichellePainter and Mark Westacott). Strenuousattempts were made at all times to followprocedures as laid down in the treatmentmanual (Fowler et a!, 1995).

All clients in the treatment conditionalso received routine care from their clinicalteams. In most instances, this included casemanagement and medication. Where theformer was not routinely available, thetherapist negotiated with the clinical teamto provide the monitoring and reviewappropriate to case management. Thus,participants in the trial had a designated

keyworker who saw them regularly and wasresponsible for coordinating their care.Clinical teams were asked, if at all possible,not to change clients' medication during thetrial, and to inform us of changes that wereunavoidable. This was monitored as closelyas possible.

Control conditionParticipants randomised into this condition

received routine care from their clinical team,which as part of our entry criteria consistedof case management and medication. Asabove, the research team negotiated withthe clinical team to ensure that clients had anallocated keyworker responsible for coordinating their care and setting goals for them.All control group keyworkers were also givenfeedback from the initial assessments, andwere encouraged to review the client'sprogress every three months.

Measures of contact with the clinicalteam, days spent in hospital and otheraspects of the costs of each group werecollected and will be presented separately.

StatisticalanalysisAll statistical analyses were carried out usingSPSS for Windows (Version 6.0) or BMDPPC—90.Changesover time (baseline,three,six and nine months) were assessed by theseparate estimation of a linear trend for each

person in the trial (using the data available for

that person, even though they may not haveprovided data for all four assessments). Theseestimates were then used as the responsevariable in further analyses, as recommendedby Matthews et a! (1990). The linear trendwas constructed (e.g. BPRS units/3 months)so that a negative value indicated an improvement (that is, there was an overall decrease inthe measured score, BPRS total, say, overtime) with a larger absolute value indicating agreater improvement ( —¿�5, for example,being a better outcome than —¿�3). Typicallywe used a two-way analysis of variance(ANOVA) with the ‘¿�experimental'sums ofsquares option in SPSS;the two explanatoryfactors being treatment centre (London,Cambridge or Norwich) and treatment group(CBTor control).

RESULTS

Participants

One hundred and fifty-two people werereferred for possible inclusion in the trial.Of these, 47 had no distressing positivesymptom present at the screening interview.Ten had not been stabilised on medication,and 26 were not suitable for practicalreasons such as living out of the catchmentarea. Nine people met the criteria but didnot agree to participate (9 of 69; 8%). Thusa total of 60 people met the criteria,consented, and were entered into the trial.Of these, all were on medication apart fromthree whose symptoms had been medication-resistant in the past, but who consistently refused to take any at the beginning ofthe trial. After randomisation, 28 peopleentered the treatment group and 32 thecontrol group. Demographic and clinicalinformation on participants in the twogroups are presented in Tables 1 and 2.

From Table 1 it can be seen that thetotal sample was middle-aged, had apreponderance of men, a long history ofillness, and average scores on an estimate ofcurrent IQ. The only difference between the

321

Variable CBTgroupControigroupn

Mean Range n MeanRangeAge(years)

28 38.5 9—65 32 41.88—63Durationofillness(years)

25 12.1 —¿�26 30 4.0—¿�33Numberofadmissions

24 5.2 0—30 29 4.30—12Predicted

IQ (NART) 25 02.9 69—I29 25 98.7 71—¿�I31Current

IQ (Quicklest) 25 99.8 72—130 29 91.5 70—I6GenderMale

IS23Female

39NART,

National AdultReadingTest.treatment

and control groups was that the withdrew from assessments over ninelatter had a somewhat lower current IQ. months, four (14%) from the CBT condition

The clinical data presented in Table 2 and seven (22%) from the control groupshow the scores on the symptoms and (five of whom withdrew immediately afterfunctioning measures that we used. As randomisation). Of the four people whoexpected, the group as a whole was sympto- dropped out of the treatment condition bymatic, with moderate levels of depression, nine months, only three (11%) attended foranxiety and hopelessness. By chance, the fewer than 10 sessions.

CBTgroup had lower levelsof self-esteemthan the control group at baseline, but bothgroups scored within the range for those Number of therapy sessionswith clinical problems. Social functioning received

was comparable to norms found in an The median number of therapy sessionsunemployed group of people with schizo- given to the treatment group was 15, andphrenia (Birchwood et a!, 1990). the mean was 18.6 (range 0—SO).One

Table 3 itemises the range of psychotic person did not attend any therapy appointsymptoms found in both groups, which did ments, one had fewer than five sessions, sixnot differ appreciably. Most participants had ‘¿�brieftherapy' (12 sessions or fewer).

had delusions, and a majority also had The rest of the treatment group (n=20) hadhallucinations, particularly in the treatment what we defined as ‘¿�fulltherapy' (more than

group. Diagnostic classification showed a 12 sessions). Treatment sessions usuallypreponderance of paranoid schizophrenia. lasted for an hour, but were kept flexible

(could be shortened) depending on a client'scurrent mental state. Most sessions were

Withdrawals conductedin out-patientclinicsettings,butOut of the 60 people who gave their consent some were home visits or ward visits tofor the trial, a total of 11 people (18%) maximise the likelihood of engagementin‘lible

2 Clinical data on participants who entered the therapytrialVariable

CBTgroupControlgroupn

Mean s.d. n Means.d.BPRS

27 26.4 6.5 26 24.57.1BDI

27 23.6 10.1 26 20.010.1BHS

27 I .6 4.8 27 9.85.2BAI

27 17.5 11.0 26 7.34.8Self-esteem

25 90.1 29.6 28 107.323.3Social

FunctioningScale 27 103.3 7.2 30 101.6 9.0

lible I Demographic data on participants who entered the therapy trial treatment, by offering sessions that were in aconvenient location for clients (e.g. oneclient was physically disabled and wasalways seen at home).

Outcome measures

Symptomsandfunctioning

All available data were analysed. The onlyparticipants who did not contribute either

refused assessments after randomisation(withdrawals) or provided only one assessment and thus could not contribute to atrend estimate. Of the 53 people who didprovide sufficient outcome data on the BPRSto estimate trends, two provided information on two time points, five providedthree of the four BPRS total scores, and theremainder (46) provided complete data (thatis, BPRS scores at all four times). The means

for the BPRS linear trends and raw scorescan be seen in Tables 4 and S.

The CBT group did significantly betterthan the controls (F1,47=7.41; P=0.009).There was little evidence that the differencebetween the CBT and control groupsdepended on centre (the test for the groupby centre interaction: Fz4@=0.S9; P0.561).Although there seemed to be an improvementin the control clients, particularly in

Norwich, neither the centre effect nor theoverall change in the control group wasstatistically significant. It should, however,be noted that there appears to be a lack ofhomogeneity of the standard deviations(Cochrane's Cg,6=0.46; P=0.06), with bothCambridge groups being considerably morevariable than the others. As there appears tobe no simple relationship between the meantrend and its standard deviation, a simpletransformation of the data would not removethis heterogeneity. The results appear to berobust, however, since dropping both theCambridge groups from the analysis yieldshomogeneous standard deviations in theremaining four groups, and a statisticallysignificant group effect remains (P=0.01).

The analysis was repeated after settingthe BPRS trend at zero (i.e. no change) forthose seven patients (two from the CBTgroup and five controls) for which a trendestimate was not available (i.e. they providedfewer than two of the repeated assessments)to provide a full ‘¿�decisionto treat' analysis

using the standard ‘¿�carryforward' method toimpute missing values. A two-group t-test forthe last row of Table S gave t= —¿�2.55withSi d.f. (P=0.014). Inserting zeros for missingparticipants and repeating the t-test gave

t= —¿�2.70with 58 d.f. (P=0.009). The meanBPRS.Brief Psychiatric Rating Scale;BDI,Beck Depression Inventory; BHS.Beck HopelessnessScale;BAI,Beck AnxietyInventory.

322

VariableCBT group

(n=27)Control

group

(n=27)DSM—Ill—R

diagnosisSchizophrenia(paranoidtype)920Delusional

disorder67'Schizoaffectivedisorder20(n=27)(n=29)2Positive

symptomsPSE:Oneormoredelusions2028PSE:

One or more hallucinations22I8PSE:

Perceptual disorders other thanhallucinationsI4PSE:

Subjective thought disorder and/or replacement ofwill57

Centre CBTgroupControlgroupn

Means.d.nMeans.d.London

14 —¿�1.491.57120.181.66Cambridge

6 —¿�2.903.998—0.353.08Norwich

6 —¿�2.371.607—1.671.21Combined

26 —¿�2.022.3127—0.462.15Significant

difference between CBTand control groups(P=O.009).Table

S Mean Brief PsychiatricRatingScalescoresAssessment

CBT groupControlgroupn

Means.d.nMeans.d.Initial

27 26.46.52624.57.1Three-month

25 22.28.22722.37.2Six-month

25 21.27.32722.96.2Nine-month

23 19.9 8.52422.77.6

Table 3 DSM—lll—Rdiagnoses and positive symptoms present in participants Clinicaloutcome

In comparison with others who have tendedto determine indices of clinical response onarbitrary grounds, we decided to adopt anapproach which aimed to take account ofthe degree of natural variability in BPRSscores over time. An estimate of the averagevariability in BPRS scores in the controlgroup was therefore calculated (taken as the

root mean sum of the squared standarddeviation of individual BPRS scores for eachcase in the control group). This equated tofive points. An improvement or worseningof greater than or equal to five points on theBPRSwas then taken as indicatinga reliableclinical change, and an improvement or

worsening of greater than or equal to 10points on the BPRS as indicating a largeclinical change. (A five-point change on theBPRS is similar to the criterion of a 20%improvement taken to be an index of clinicalresponse on the BPRS by Breier et a!(1994)). In these terms, 6/28 (21%) achieved

a large clinical improvement, and a further8/28 (29%) of the treatment group achieveda reliable clinical improvement. One personof the 28 (3%) in the treatment groupshowed a reliable worsening of symptomson the BPRS. In the control group, 1/32(3%) showed a large clinical improvementand 9/32 (28%) of cases achieved reliableclinical improvements. Three of the 32 (9%)of the control group showed a clinicallysignificant worsening of symptoms over thenine months.

If we widen the criteria to includeclinically significant response in the client'sprimary presenting problem as measured bythe Personal Questionnaires, then 18/28(64%) of treatment and 15/32 (47%) of

controls achieved clinically significantimprovements. If cases with no clear lineartrend in the scores are excluded (i.e. wherethere is little obvious evidence of trend),then the number in the control group whochanged drops to 12/32 (37%).

One person in the control condition hadcommitted suicide by the end of ninemonths. No one in the treatment conditionhad done this.

Medication

Medication regimes were complex andinformation was sometimes incomplete.We calculated chlorpromazine equivalents(CPZ)followingthe guidelinesin the BritishNational Formu!ary. Full data were available for the London participants, but datawere more limited for East Anglia. We

I. Thediagnosisfor onepersonwasconfirmedatthe three-monthassessmentbecauseofa lackof informationfromtheinitial assessment.2. Informationisincludedconcerningpositivesymptomspresentintwo participantswhowithdrew fromtheresearchpartwaythroughcollectionof PSE—lOdata.

Table 4 BriefPsychiatricRatingScalelineartrends ineachcentre

trend for the CBT group (n=28) was then—¿�1.87 (s.d. 2.89) and for the controls n=32)

—¿�0.38(s.d. 1.98).Visual inspection showed that items on

the BPRS which changed most in comparison with the control group were suspiciousness (ideas of reference and persecution),unusual thought content (delusional ideas)and hallucinations. The self-reported reduction in delusional conviction, measured bylinear trend, was —¿�0.65for the CBT group

and —¿�0.30for the controls. Delusionaldistress was —¿�0.61 for CBT and —¿�0.39for controls. The linear trend for thefrequency of hallucinations was —¿�0.24forthe CBT group and —¿�0.01 for the controlgroup. None of these reached significance atconventional levels, although all favouredthe CBT condition. Change on all othersymptom and functioning measures was notsignificantly different between conditions atthis stage of the trial.

323

CBT group

(n=20)

5

3

4

13

2

classified these into no medication, low (lessthan 300 mg CPZJday), medium (300—600 mg CPZiday) and high (more than600 mg CPZIday). We also divided participants into those receiving constant, fluctuating, decreasing or increasing doses. Fourclients were switched to clozapine before thefinal assessment. Three of these were in the

control group, and the change occurredbetween the three and six month assessments. One person was in the CBT groupand the change only occurred after the sixmonth assessment.

Inspection of the data at baseline inTable 6, suggests that there were noparticular differences in medication between

the treatment and control conditions.However, over the nine months of the trialmore control participants had their medication increased, whereas two of the CBTgroup and none of the controls had itdecreased. This meant that as the trialprogressed more of the control group movedinto the high-dose category and fewer ofthem received low doses or no medicationcompared with the CBT group.

@I@ble6 MedIcation levels based on chiorpromazine equivalents

Satisfaction

CBTgroup Control group Twenty of the 28 in the treatment groupcompleted a satisfaction with therapyquestionnaire at nine months. As can be

2 I seen in Table 7, 16 (80%) were satisfied5 4 or very satisfied with the therapy, 17 felt

3 I0 they had made some or much progress,8 5 and 17 felt they would be able to make

some or much progress in the future. OneI I I0 client reported that ‘¿�thingsgot worse', and

I 2 this person was also dissatisfied with

treatment.

Levelofneuroleptic doseat start of trial

None

Low

Medium

High

Changesin medicationduring trial

No changeFluctuating

Increasing 2

Decreasing 2

Levelof neuroleptic medicationthroughout the trial

None 4

Low 2

Medium 4

High 6

Levelsof neuroleptic medication: Low: lessthan 300 mg chlorpromazine; medium: 300 to 600 mg ofchlorpromazine;high: greater than 600 mg chlorpromazine.All availabledata on medication are included.These were predominantly from the London sample. which when consideredon itsowndidnothaveadiscerniblydifferentpattern.

How satisfied are you with the therapy?

VerysatisfiedSatisfied

Indifferent

Dissatisfied

During therapy how much progress do you feel you actually made?

Much progress

Someprogress

No progress

Things got worse

Infuture,howmuchprogressdoyouthinkyouwillbeableto makeindealingwithyour problem?'

Much progress

Someprogress

No progressThings will get worse

I. One participant did not answer this question.

7

0

0

•¿�lkbI.7 Satisfaction with cognitive—behaviouraltherapy

9

8

324

DISCUSSION

2 The results of this trial show that at the end8 of nine months of CBT it is possible to

9 improve the overall symptomatology of

_______ peoplewithmedication-resistant,distressingsymptoms of psychosis. This group stillexists even after the introduction of thenew neuroleptics (Kane, 1996). We showeda decrease in BPRS scores of 25% and thiswas produced mainly by changes in ourtargeted symptoms of delusions and hallucinations.Therewereno appreciableimprovements in the level of depression. Atthis stage, the specific improvements—¿�observedinconvictionfordelusionalideas,were not statistically significant, in contrastto the results from our previous waiting listcontrol trial. This is at least partly becauseof the more stringent methodology of anRCT design compared to uncontrolled orless well controlled trials. Further supportfor the specific effects of CBT is provided bya finding that only in the CBT group wasoutcome predicted by a cognitive measurelinked to delusional thinking (Garety et a!,

1997).

EngagementThe therapy was acceptable to clients, whoexpressed high levels of satisfaction and didnot show demonstrable negative consequences. Our results illustrate that psychological treatments can be offered to clients

even when they have long histories ofillness, and continuing distressing symptoms of psychosis. Our drop-out rates,which remained low, are particularlyencouraging, and suggest that if engagement issues are dealt with, then even thisclient group can accept demanding interventions. This has also been demonstratedin a recent study of compliance therapy(Kemp et a!, 1996), which addressed clientcentred concerns in a similar way, but hasoften previously been problematic (e.g.Tarrier et a!, 1993).

Rate of improvementThe rate of improvement in overall symptomatology (BPRS scores) that we havedemonstrated, is around the same level asthat found in studies on the effects onclozapine on clinically similar samples of

people who have failed to respond tostandard medication. While relatively fewRCTs of clozapine have been completed,those in the literature are usually six-weektrials, on large numbers of clients,comparing clozapine to chlorpromazine(Kane et a!, 1988) or to haloperidol (Breieret a!, 1994). These showed changes on theBPRS of between 11 and 26% in theclozapine group. Interestingly, the Kane eta! study (n=268) showed a 26% change inthe clozapine group, compared with an 8%change in the chlorpromazine group, that isboth groups improved with intensive monitoring and optimal medication.

The advantage of a talking therapy isthat it does not have physical side-effects,and can produce change that is bothclinically noticeable and significant. Thedisadvantages of such treatments includeintensive input by experienced clinicians(around 19 hours over nine months) andthe extra training and supervision that isrequired to support therapists. A furtherdisadvantage of this intervention so far, isthat at this stage we have shown changeonly in BPRS scores, not in social functioning or any of our other measures.

MethodologicalproblemsOur study suffered from methodologicalproblems that are common to virtually alltrials of psychological treatments. Mostproblematic was the view that we couldnot ensure our evaluators would remainblind to the treatment condition. Contactingand assessing 60 individuals over ninemonths requires considerable persistence and

sensitivity, and from the experience of our

pilot study we did not feel it wai realistic orreasonable to assume that we could preventdetails of therapy or control conditionsemerging during assessment meetings with

evaluators. We may have been mistakenabout this, and it does weaken the methodology. However, we decided before thetrial started that while we could maintainthe independence of our evaluators by notinvolving them in treatment, we would notattempt to keep them blind to treatmenttype. This decision is endorsed by Shapiro(1996), who comments that blind evaluationis virtually impossible with psychological

treatment trials: “¿�Personnelemployed tointerview patients to assess their progress

are seldom able to avoid exposure toinformation (especially within patients'accounts of their experiences) that givesaway the nature of the treatment they haveundergone― (p. 204).

Second, although we monitored mcdication and medication changes in bothgroups, it was not possible to keep allindividuals stable or on the same medication regime. It remains a possibility thatsome improvements were due to medication effects, particularly in the controlgroup. Clinicians increased standard mcdication or changed to clozapine more oftenfor control patients.

Third, the selection of a viable controlcondition can be seen as problematic. In

this study, we decided to compare CBTwith the best current routine treatment,

case management and medication. Thesecannot ethically be withdrawn from thetreatment group (nor would we wish it), soin effect we evaluated CBT as added to thebest standard treatment compared with thisstandard treatment alone. It is obviously apossibility that the extra 19 hours (mean)

of treatment effected improvement becauseof non-specific attention. The fact that ourresults showed specific symptomatic changepredicted by related cognitive measures(Garety et a!, 1997) mitigates against this,

but lack 9f change in delusional convictionor depression at this stage suggests that wecannot assume our CBT interventions werethe effective ingredients in improvement.On the other hand, there is a case forarguing that the detailed assessments thecontrol group received in itself comprisedelements of an attentional control condition. The control group had equallysustained contact with the assessors overtime; the assessors had to engage thepatients, and at the beginning patients

received a detailed assessment phase involving around six sessions. Patients then hadregular assessment contacts at three-monthintervals. Some patients reported that theybelieved the assessors were part of theirtreatment team. The process of conductingassessments involved specific discussion of

symptoms in a supportive atmosphere, andit is possible that such assessments werethemselves a minimal focusing intervention.However, this still does not control for thetherapy time in the CBT condition, andfurther controlled studies are required toclarify the issue of specific versus nonspecific effects of therapy.

Fourth, we did not aim in this study toexamine which aspects of CBT wereeffective, or to look at an optimal numberof sessions for this kind of treatment.Some patients would attend only for ‘¿�brieftherapy', despite efforts to visit them athome or reorganise appointments. Othersattended for maximum or even excessivenumbers of sessions, without muchapparent benefit. Thus, individuals variedconsiderably in how long they took toshow change, and this issue remains to beclarified.

Clinical implicationsFinally, it is clear that even though wedemonstrated changes in our treatmentgroup, only 50% of them were treatmentresponders. Assessing change in patientswith very long-standing and treatmentresistant psychotic symptoms poses theproblem that even clinically noticeablechange does not place people back in the‘¿�normal'range. On the other hand, evensmall changes in symptom levels mightsignify important differences in an individual's ability to cope with problems, orcope in the future. We will be able todiscuss any maintenance or preventioneffects when we have completed thefollow-up phase of the study. Predictorsof outcome are discussed in a subsequentpaper (Garety et a!, 1997). However, ourresearch seems to indicate that talking topatients about psychotic symptoms andtheir meaning to the individual is a skillthat clinicians working in this area shoulddevelop.

ACKNOWLEDGEMENTS

We thank all the patients who participated in thisstudy and the clinicians and clinical teams in London

and East Anglia who cooperated in referring and

325

Chadwkk, P. D.j. Ii Lows, C. F. (1990) Measurement and

modification of delusional beliefs.Journalof Consultingand

Qin,calPsychology.58, 225—232.

Drury,V., Birchwood, M., Cochran., R., .t al (1996)Cognitivetherapy andrecovery from acutepsychosis:acontrolled trial. I: Impacton psychoticsymptoms.BritishJournalof Psychiatry.169.593—601.

FowIer@D. (1992)Cognitivebehaviourtherapy in themanagementof patientswith schizophrenia:preliminarystudies.In Psychotherapyof Schizophrenia:FacilitatingandObstructiveFactors(edt A.@%@rbart& J.Cullberg).pp 145—153.Oslo:ScandinavianUniversityPress.

—¿� a Morley, S. (1989) The cognitive behavioural treatment

of hallucinationsand delusions:a preliminary study.Behaviouml

Psychotherapy,17,267—282.

—¿� , Gaiety, P. A. & Kulp.rs, L (1995) Cogn:twe Beha@our

Therapyfor Psychosis:Theoryand Practice.Chichester:Wiley.

Garety, P. A., H.msI.y, D. R. & Wsss&y, 5. (1991)Reasoningin deludedschizophrenicand paranoidpatients:biasesin performanceon a probabilisticinferencetask.Journalof NervousandMental 0.sease,179,194—201.

Garety, R, Kulpers, L, Fowler, D., t at (1994) Cognit@ebehaviouraltherapy for drug-resistantpsychosis.BritishJournalof MedicalPsychology.67,259—271.

—¿� . FowI.r, a,@ E., .t at (1997) The London—East

Angliarandomisedcontrolled trial of cognitive—behaviouraltherapy for psychosis.II:Predictorsof outcomein CBTforpsychosis.BritishJournalof Psychiatry,171,in press.

Haddock, G., Bentall, N. P. & Slad., R D. (19%)Psych@ogicaItreatment of auditory hallucinations:focusingordistraction?In Cognitive—BehavioumlInterventionswithPsychoticDisorders(edsG. Haddock & P.Slade),pp. 45—70.London:Routledge.

Hustig, H. H. & Hafn.r, U. (1999) Persistent auditoryhallucinationsandtheir relationshipto delusionsandmood.Journalof Nervousand MentalDisease,178,264—267.

ja*pers, K. (1963) Geneml Psychopathology(trans.J. Hoenig &M.Harndton).Manchester:ManchesterUniversity Press.

Kane, j. M. (1996)Treatmentresistantschizophrenicpatients.Journalof ClinicalPsychiatry,57 (suppl.9). 35—40.

—¿� , Honigf.Id, G., Singer, J., it aS (198$) Clozapine for

the treatment-resistantschizophrenic:a doubleblindcomparisonwithch@rpromazine.ArchivesofGeneralPsychiatry,45, 789—796.

Klngdon, D. G. & Thrklngton, D. (1991) The use of cognitivebehaviourtherapy with a normalisingrationale inschizophrenia.Journalof Nersous and Mental Disease, l@,

207—211.

K.mp, R., Hayward, P.,App1.WhaIt., G., .t ci (1996)Compliancetherapy in psychoticpatients: random controlledtrial. BritishMedscalJournal,312.345—349.

Matthews, J. N. 5., Altman, D@G., Campb.II, M.J., it at

(1990)Analysisofserialmeasurementsinmedicalresearch.BritishMedicalJoumal,300, 230—235.

MIlI.r@E. (1984) @rbalfluency as a function ofa measure ofverbal intelligenceand in relationto differenttypes of cerebralpathology.BntishJournalof ClinicalPsycho1og@U 53—57.

Nslson, H. E. (1982) TheNational AdultReadingTest.Wsndsor,Berkshire: NFER- Nelson.

Oy.raII, j. E. Si Gorham, D. R. (1962) The Brief PsychiatricRatingScale.PsychologicalReports,10,799-812.

Pocock, S. (1983)ClinicalTrials.Chichester:Wiley.

Robson, P. (1989) Development of a new self-reportquestionnalretOmeasureselfesteem.PsychologicalMedicine,I,, 513—518.

ShaIIke,T. a Evans, M. (1978) The involvement ofthe frontallobesincognitiveestimation.Cortex,14,292—303.

CLINICAL IMPLICATIONS

U The study did not control for attention effects.

U Assessors were independent but not blind to the treatment condition.

. Althoughchangesinthe treatment groupwere significantandclinicallyreliable,only50%of these participantswere treatment responders.

ELIZABETHKUIPERS,PhD,Department of Clinical Psychology.Institute of Psychiatry,London SES8AF; PHILIPPAGARETY, PhD, United Medical and Dental School, Department of Psychology, St Thomas' Hospital. London SEI;

DAVID FOWLER,MSC,Schoolof Health Policyand Practice,University of EastAnglia,Norwich NR4 7TJ;GRAHAM DUNN, PhD,Schoolof Epidemiologyand Health Sciences,University of Manchester,StopfordBuilding. Oxford Road, Manchester; PAUL BEBBING1ON, PhD, University College London, Archway Wing 1st

Floor,Whittington Hospital, Highgate Hill, London N19 5NF; DANIEL FREEMAN, BA, Department of Clinical

Psychology,Institute of Psychiatry,London SES8AF; GLAREHADLEY,MSc,Department of Clinical Psychology,LeedsUniversity, IS HydeTerrace,LeedsLS29LT

Correspondence: Dr E.Kuipers, Department of Clinical Psychology,Institute of Psychiatry, LondonSES8AF

(First received 8 November 1996, final revision 25 March 1997, accepted 14April 997)

supporting patients. We are grateful for the thera

peutic input of Steve Jones, Nina Dick, Erik Kuiper.

Michelle Painter and Mark Westacott.We alsothankthe members of our steering group for their support

during the trial. This research was supported by a

Researchand Development grant from the Department of Health.We are grateful for a charitable dona

tion from Janssen Pharmaceutica.

REFERENCES

Amado,@X. F.,Strauis, D. H.,'V@1.,5. A., .t at (I@@)Assessmentof insightin psychosis.AmericanJournalofPsychiatry,ISO,873—879.

Am.rlcan Psychiatric Association (1987) Diagnostic and

StatisticalManualofMencal D,so,riers(3rd edn, reveed)(DSM-lll-R).W@shington,DC: APA.

Ammosis. R. 8. Si Ammons.C. II. (1962)QuickTest.Missoula,MT:PsychologicalTestSpecialists.

R.ck. AT. (1952)Successfuloutpatient psychotherapywith aschizophrenicwith a deIus@onbasedon borrowed gutlt.PsychiatryIS, 305—312.

B.ck, A.T.. Rush,A.j., Shaw,B. F., it al (I9@) Cognit@eTherapyof Depression.New York:Guilford Press.

—¿� ,Ward, C. H.. Msndskon, M., .t at (IflI) An inventory

for measuringdepression.ArchivesofGeneralPsychiatry,4,561—571.

—¿� ,@iuman, A.W, L.ste,@ D., .t aS (1974) The

assessmentof pessimism:the HopelessnessScale.JoumolofConsultingandClinicalPsychoIog@42,861—865.

—¿� , Epitsin, N., Brown,G., .t aS(19U) An inventory for

measuringclinicalanxiety: Psychometricproperties.Journa/ofConsultingand ClinicalPsycho1og>@56, 893—897.

B.II.w, N. (IWO) informationProcessingB.asesandDepression.Universityof Keele,unpublishedPhD thesis.

Blrchwood, N. (I@96)Earlyintervention in psychoticrelapse:cognitive approaches to detection and management. In

Cogniti@e-BehcMoumIInterventionswith PsychoticDisossiers(eds G. Haddock & P.Slade), pp. 171—¿�21I. London: Routledge.

—¿� , Smith, J., Cochran., R.. it aS (IPFO) The social

functioning scale. British Journal of Psychiatry, 157, 853—859.

Bouchard, S.;VIIHSVSS,A., Ray,M., .t aI(I9@) Cognitiverestructuring inthe treatment ofpsychoticsymptomsinschizophrenia:acriticalanalysis.BehcMourTheropy,27,257-277.

Br@ A., Buchanan, R.W, Kirkpatrick, B., it aS(I@4)Effectsofclozapineonpositiveandnegativesymptomsinoutpatientswith schizophrenia.AmericanJournalof Psychtatry.151,20—26.

Br.tt.Jones, J., Gar.ty, P. A. & HemsI.y, D. (1987)

Measuringdelusionalexperiences:a method and itsapplication.BritishJournalof ClinicalPsychology.26 257—265.

Buchanan, A., R.d, A.,W@s,.Iy, S., .t at (1993) Acting ondelusions.2:Thephenomenologicalcorrelatesofactingondelusions.BritishJournalof Psychiatry.163,77—81.

326

. It ispossibleto improveoutcomein psychosisusinganadaptedform of CBT.

U Drop-out rates for therapy were low and satisfaction was high, suggesting that

clientswith psychosiswelcomethiskindof intervention.

I Effects on symptoms were similar to those found in RCTs of clozapine.

LIMITATIONS

5@iro. D. A. (1996) Outcome research. In Behavioural and Watts, F. N., POWSII,G. E. a Austin, S.V. (1973) The a D,ft.chd B. H. (198$) Emotional d@turbance and the

MentalHealth Research(2nd edn) (edsG. Parry & F.W@tts). modificationof abnormalbeliefs.BritishjournalofMedical specificityofautobiographicalmemory CognitionandEmotion,Hove: LawrenceErlbaum. Psychology,46, 359—363. 2, 221—234.

World H.slth Or@adoii (1992) SCANScheduiosfor

ThrrIe,@N., Bedc.tt, R.,Harwood,5., it aS(1993)A trialof ClinicalMsessment@ beuropsyshiotryGeneva:WHO.two cognitive-behavioural methods oftreating drug- WIlliami, J. N. G. (1992) The Psychological Treatment of

resistantpsychoticsymptomsin schizophrenicpatients.I: Depression.A GuidetoTheoryand Practiceof Cognitive Young,J. (1990) Cogrst,@eTherapyfor RersonahtyDisonierOutcome.BiitishJournalof Psychiatry162,524-532. BehonourTherapyLondon:Routledge. Sarasota.FL:ProfessionalResourceExchangeInc.

327