B R I T I S H I O U R N A L O F PSYCHIATRY (1999). 175.

188-185 PRELIMINARY REPORT

Group treatment of auditory hallucinations resistant auditory hallucinations which

were not the result of an organic disorder.
Exploratory study of effectiveness
Main outcome measures
TIL WYKES. ANN-MARIE PARR and SABINE LANDAU Auditory Haffuanations Rating Scak (PSY-
RATS; Haddock et al, 1999). A self-report
scale with proven reliability and validity
which measures emotional content,
physical characteristics and cognitive
interpretation.
Expanded Brief Psychiatric Rating Scale
Background Cognitive- behavioural Despite high doses of medication a sign&- (BPRSE; Ventura et al, 1993), completed
therapy has been shownto be effective in cant number of people with schizophrenia by an independent clinician who was not
still experience distressiig auditory halluci- blind to treatment status.
reducingpsychotic symptoms, butfew
nations. These not only affect the quality of Self-Report Insight Scak for Psychosis (IS;
patients have access to these services. their lives but are probably instrumental in Birchwood et al, 1994). A self-report ques-
Group cognitivetreatment may providea maintaining depression and low self- tionnaire which provides measures of
less costly service with similar benefits. esteem. The efficacy of psychological treat- symptom attribution, awareness of illness
ment in addition to pharmacotherapy has and acceptance of the need for treatment.
Alms To explore the effectivenessof been established in a number of randomised
group cognitive-behavioural therapy on controlled trials (e.g. Kuipers et af, 1997;
Specific clinical outcome measures
insight and symptoms, particularly Tarrier et a1, 1998). However, because of
the shortage of trained therapists and the About Voices Qmsti-i~e
auditory hallucinations. length of treatment, these thetapies are un- (BAVC; Chadwick Birchwood, 1995).
Method Twenty-one DSM-IV
likely to become widely available in the
health services in the near future. An alter-
wing strate+,,. A simple self-report
diagnosedpatientswith schizophrenia qu*onnaire in the participants
native is to Present the therapy in a group listed their coping strategies in mpo- to
with treatment-resistant, distressing format, which has already been shown to the voices and rated the of
auditory hallucinationswere referred to a be feasible (Gledhill et al, 1998), and which these strategies.
group programmeconsistingof six offers the likelihood of a more general
availability of psychological treatment at a
sessions of cognitivetreatment following a Ancillary outcome measures
lower cost. The present study describes
strict protocol which emphasised an evaluation of this alternative moup Beck Deplesion Inventory (BDI; Beck et
- -
individualpower and control as well as presentation. al, 1961).
coping strategies. Beck Anxiety Itwentory (BAI; Beck et al,
1988).
Results There were significantchanges METHOD
Rosenberg's Self-Esteem Scak (Rosenberg,
in all three main outcome measures
Design
followingtreatment; those changes were
This study adopted a waiting-list control
maintained at follow-up and were greater design with repeated measures within sub- Intervention
than changes over the waiting-list period. jects. However, a few people were only Sessions were based on a cognitivebehav-
Specifically, there were changes in referred when the groups started; these i o d approach, followed a semi&
perceivedpower and distress as well as referrals were also included in the study. format and lasted for an hour. Each session
The efficacy of treatment was evaluated dealt with a particular theme:
increasesinthe number and effectiveness
using seven symptom measures and one
ofthe coping strategies. Week 7 - sharing of information about
measure of coping. Participants were allo-
the voices
cated to one of three groups through a roll-
Conclusions Group treatment for
ing programme of referrals based on service Week 8 - models of psychosis
auditory hallucinations needs further contact with three community teams. Each Week 9 - models of hallucinations
investigationbut does look promisingand of the groups met weekly, for one hour,
over a six-week period, with a follow-up Week 10 - effeaive coping strategies
may providea less costly alternativeto
session three months later. Outcome mea- Week 11- improving self-esteem
individual cognitivetreatment.
suses were on four different oc- Week 12 - an model of coping
Declarationof interest None. casions: referral (week 0); pre-intervention with voices
(week 6); post-intervention (week 12);
three-month follow-up (week 24). Week 24 - follow-up session.
Subjects were included in the study if Each session followed a detailed protocol
they experienced distressing, medication- containing the aims of the session,
 GROUP TREATMENT O F AUDITORY H A L L U C I N A T I O N S

examples of interventions and model AvPikblanumbcrofoboervrcknsathcvvburu~tarenttimesfammporisomdrmjorovtcanet

responses for the therapist. (out of 21 d d)

Sale Cases with any Complete Baseline and b a n d post- btreatment Control
Statistical analysis observation cases pre-treatment treatment and followup comparison
For the analysis of the main and ancillary
BPRS-E 16 6 8 13 10 8
outcomes, unbalanced repeated-measures
PSYRATS 20 7 II I2 10 8
models assuming an unstructured covar-
iance matrix were employed to impute miss-
ing observations (see for example Everitt,
1998). Models were fitted into the software
package BMDP, release 7 (Dixon, 1992)
using the method of restricted maximum
likelihood. The effect of time was tested remainder were prescribed standard neuro- si@cant differences between the groups
using the Wald statistic; when this was leptics with an average chlorpromazine on any of the variables.
sigdicant, specific Wald tests were carried dose equivalent of 343.18 (sd. 205).
out over three periods: waiting, treatment Thirteen of the 21 people referred to the
and follow-up. Finally, if there was evidence study completed the course of treatment Main outcomes
of a treatment effect we compared it with and some post-treatment assessment, Expanded Brief Psychiatric Rating Scok
the changes over the waiting period as a although only erght of these entered the (BPRS-E)
control condition. Even when it could be trial during the waiting-list period. This Figure 1shows the estimated mean BPRS-E
argued that the comparisons between the led to considerable differences in the scores; a reduction in score indicates a
treatment phase and the control phase (wait- number of observations available for the symptom improvement. The BPRS-E score
ing time) were supported by a specific pairwise comparisons, as shown in Table was affected by the assessment time (Wald
hypothesis we decided to adopt the most 1 for the main outcomes. test: x2=37.7, d.f.=3, P<0.0001). There
conservative approach, so all tests were was no significant difference over the wait-
two-tailed. ing period (estimated mean difference
Drop out (e.m.d.)=5.8, 95% confidence interval
The dropout mechanism was investigated (CI) - 1.9 to 13.5) but scores were si@-
RESULTS by comparing socio-demographic, clinical cantly reduced over the treatment period
or outcome variables at the pre-treatment (Wald test: $=26.02, d.f.=l, P<0.0001,
Patient sample stage for the group of participants whose e.m.d.=7.9, 95% CI 4.9 to 10.9). This re-
The participants were representative of PSYRATS totals were missing at post- duction was not maintained at follow-up
those who continue to attend mental health treatment andlor follow-up stage (n=ll) (e.m.d.=2,95% CI -2.8 to 6.8). The con-
services with resistant psychotic symptoms. with the remaining participants for whom trol test supported a treatment effect larger
All 21 had a diagnosis of schizophrenia scores existed (n=10). There were no than expected from the waiting period
according to DSM-TV criteria (American
Psychiatric Association, 1994). They had
high scores on the BPRS-E with both nega-
tive and positive symptoms, and moderate
levels of depression on the BPRS-E, which
were about the same as estimates obtained
from the recent cognitive-behavioural ther-
apy (CBT) study reported by Kuipers et a1
(1997). They generally were middle-aged
(average age 40 years) and had a long dura-
tion of illness (mean duration of hearing
voices 14 years), and about half were living
in psychiatric residences. Threequarters
experienced voices daily, these voices
mostly having a negative content (95%),
and 75% also reported that the voices
caused at least a moderate amount of dis-
ruption to their lives. All subjects were on
stable doses of neuroleptic medications:
48% were prescribed one of the novel anti- Time (weeks)
psychotic medications, the average dose
(British National Fomuhry guidelines Fig. I Estimrad mean total symptom Expanded Brief Psychiatric Rating Scale scores (bas represent
percentage) being 56.5% (s.d. 23.2); the standard ertur of the man).
W Y K E S E T AL

considerably skewed (Wald test: 2=3.38,

d.f.=l, P=0.066, e.m.d.-3.2, 95% CI
-0.2 to 6.5).

Insight Scak Psychosis (IS)

Figure 3 shows the estimated mean IS
scores; an increase in score indicates an
improvement in insight. IS score was
affected by the assessment time (Wald test:
2=15.46, d.f.=3, P=0.0015). Scores were
significantly reduced over the waiting
period (Wald test: 2=4.2, d.f.=l,
P=0.04, e.m.d.=0.7, 95% CI 0.03 to 1.4)
but were significantly increased over the
treatment period (Wald test: 2=6.97,
d.f.=l, P=0.0083, e.m.d.=l.S, 95% CI
0.4 to 2.5). This increase was maintained
at follow-up (Wald test: 2=7, d.f.=l,
P=O.OO82, e.m.d.=l.S, 95% CI 0.4 to
2.7). The control test supported a treat-
ment effect larger than expected from
the waiting period (Wald test: 2=11.99,
d.f.=l, P=0.0005, e.m.d.=2.2, 95% CI
(Wald test: Xf=7.31, d.f.=l, P=0.007, was no significant difference over the 1 to 3.4).
e.m.d.=13.7, 95% CI 3.8 to 23.6). waiting period (e.m.d.=l.S, 95% CI - 1 Table 2 summarises the results for our
to 4), but scores were sigdkantly reduced three main outcomes. The table indicates
over the treatment period (Wald test: treatment success in terms of all three main
Auditory HaIIucinations Rating Scak $=15.43, d.f.=l, P=0.0001, e.m.d.=4.7, scores. For PSYRATS and IS scores there is
(PMRAK) 95% CI 2.3 to 7), and this reduction was evidence that this success was maintained
F i 2 shows the estimated mean maintained at follow-up (Wald test: at follow-up.
PSYRATS scores; a reduction in score indi- ?=5.48, d.f.=l, P=0.019, e.m.d.=6.6,
cates an improvement in the experience of 95% CI 1.1 to 12.1). The control test
auditory hallucinations. PSYRATS score (which was two-tailed) was not statistically
was affected by the assessment time (Wald significant at the 5% level, but the P value Specific clinical outcomes
test: 2=21.7, d.f.=3, P=0.0001). There was small and the confidence interval The treatment programme was designed to
target a number of specific issues relating to
voices. These specific targets were mea-
sured by individual items or factor scores,
except for the coping strategy measure.
These data were analysed with paired t-
tests over each of the three assessment
periods. The means and confidence inter-
vals for the mean differences are given in
Table 3.

Scale Waiting Travnanr MaintawKa

PSYRATS NS - -
IS - + +
+.blaeamouarpabd; -.-
~p~s-E.Exppnded&klRychba*~k.b;
PMRATS.--mwk.b:
rsSdl-Rcpatlnrbhr-f'JrRlchodr.
 GROUP TREATMENT O F AUDITORY HALLUCINATIONS

lWe 3 Meam and confidena intemls ( a s ) for the rpcdfic clinical ourmmes over the thm assessment periods

Clinical musure Pre-treatmentbaseline (n= II) k-treatment-pmtreatment (n= 12) Followup-pre-treatment (n= 10)

Baseline &treat- 95%CI Re-treat- k-veat- 95%CI Pre-treat- Follow-up 95%CI

ment difference ment ment difference ment difference

Ownershipdvoices 2.3 2.4 -0.54 to 0.73 2.4 1.9 - 1.0 to 0.006' 2.5 1.6 -1.9to0.14
Control over voices 3.6 3.5 -0.65 to 0.47 3.6 2.8 -0.22 to 1.52' 3.5 2.6 -2.18t00.38
Perceived power of the 0.8 0.8 -0.30 to 0.30 0.8 0.6 -0.54 to 0.04' 0.8 0.6 -0.50to0.10
vdat
Amount of distress 3.0 2.8 -0.77 to 0.41 2.7 2.3 - 1.05 to 0.22 2.6 2.0 - 1.37 to 0.17
Severity ddistress 2.8 3.0 - 0 . 5 4 ~0.91 2.8 2.1 - 1.08to -0.25' 2.6 2.0 - 1.29 to 0.09'
Physical characteristics 10 8.5 -2.78to -0.13' 9.0 8.0 -2.05 to 0.05' 8.6 7.2 -3.79 to 0.99
(duration, loudness,
frcqucncy, bcatiloc)
DkrupioncauKdby 2.1 2.0 -0.45 to 0.27 1.7 1.2 -0.83to -0.171 1.6 1.3 -0.78t00.18
the voices
Number of coping 1.6 1.6 -0.29to0.11 1.5 2.8 0.92 to 1.7S2 1.6 2.7 0.47 to 1.73'
strategies reported
p~ - -

I.a which k dwed. indicating a vsnd in rhs hypothrskeddirection.

2. Cl which does not indude zcm

There is only one significant change strategies) achieved significance, although d.f.=3, P=0.11) or depression (BDI:
over the waiting period - a decrease in the there was a trend for reduced distress sever- 2=4.04, d.f.=3, P=0.26). The Rosenberg
physical attributes of the voices. There were ity. The data for these individual items are Self-Esteem Scale was not significantly af-
statistically significant differences in the not as robust as for the total scores, sug- fected by the assessment time (2=7.34,
predicted direction over the treatment peri- gesting that treatment effects may be differ- d.f.=3, P=0.062), but since this P value
od for three measures, and for a further ent for each individual. was small, pairwise time point comparisons
four the confidence intervals were extre- were carried out. There was no significant
mely skewed. These data, therefore, indi- difference in scores over the waiting period
cate a targeted treatment effect. At follow- Ancillary outcomes (e.m.d.=O.6, 95% CI -0.8 to 2), but self-
up, however, the size of the improvements The Wald tests did not indicate an effect of esteem improved over the treatment period
was reduced and only one (increasedcoping assessment time for anxiety (BAI: 2=6.1, (Wald test: 2=5.12, d.f.=l, P=0.024,
e.m.d.=2.2, 95% CI 0.3 to 4.1), and this
improvement was maintained to follow-up
W e4 Coping scntq$as used by patienu (Wald test: 2=5.73, d.f.=l, P=0.017,
e.m.d.=ld, 95% CI 0.3 to 2.5).
coping Number of patients using strategy

Baseline Pre-treatment k-treatment Follow-up

(n=I2)I (n=17)' (n=I2)I (,=lo)l Coping strategies
A list of coping strategies is shown in Table
Hummingor whistling 0 0 4 3 4. The most frequently used strategies be-
Take medication 0 I I I fore treatment were watching television or
Talking to others I I 3 3 listening to music, although less than half
Talk inwardly to yourself I I I 0 reported receiving any benefit. After the
WatchN 4 6 3 3 therapy the participants reported a wider
Goout 0 0 4 2 . range of strategies (see above analyses)
Listento the radio/music and increased effectiveness (percentage
Readthe Bible or pray
effective: baseline (n=ll) 73%, pre-treat-
ment (n=lS) 54%, post-treatment (n=12)
R u u m with the voices
92%, follow-up (n=9) 78%; sign test pre-
Tell the voices to go away
post-treatment P c 0.016, two-tailed).
Ask the voices to come back later 0 0 4 2
There were no differences during the wait-
Passm
i straegies, e.g. ignoringthe voices 2 2 0 2 ing period and the change in effectiveness
I. Number of prtkmr canpkting the coping stnmgks quarionnaire. disappeared at follow-up.
W Y K E S ET AL

Associations with improvement

Multiple stepwise (forward) regression ana- C L l N l C A L IMPLICATIONS
lyses were carried out, with change in each
outcome measure over the treatment period Gmup psychological treatment for distressing auditory hallucinations may reduce
as the dependent variable and demo- symptoms and increase insight.
graphic, clinical and other pre-treatment
outcome measures as independent vari- Patients who experience distressing voices express high levels of satisfaction with
ables. No significant predictors emerged the treatment.
for any of the main outcomes. The effect of treatment was similar to that obtained by individual therapy but was
Biichwood & Chadwick (1997) suggest
considerably less expensive.
that there is a relationship between the per-
ceived power of the voices and the effect LTMITATIONS
generated by the voices. Reductions in
power should, therefore, have a beneficial B The study used a waiting-list control desi~n,which is not as powerful as a
effect on distress. There was a significant randomised control design.
partial correlation in the predicted direction
B Several people dropped out of the assessments and/or treatment.
between the change in perceived power and
the post-treatment levels of distress after Although symptom assessments were independent of the treatment they could
controlling for pre-treatment levels (partial not be blind, as everyone received treatment.
r=0.63, P=O.04).

DISCUSSION

The study described in this paper explored TILWYKES,PhD. ANN-MARIE PARR. SABINE LANDAU. PhD. Department of Rychobgy, Institute of
the usefulness of a less expensive version Ryrhiatry.London
of cognitively oriented treatment for audi-
tory hallucinations. The treatment was Correspondence:T.Wykes.Departmentof Psychology,Institute of Psychiatry,De Crespigny Park.
similar to that in other, more intensive London SES 8AF. e-mail: [email protected]
individual treatments. However, it was also
(First received 23 September 1998, final revision 23 February 1999, accepted 2 March 1999)
hypothesised that presenting this treatment
in a group format would add value because
the group processes themselves are
particularly powerful. The participants
for PSYRATS. Improvements in the main current data. Larger subject numbers
were representative of those who attend
outcomes could not be explained by would allow covariance modelling, which
most psychiatric services regularly. They
changes in specific measures, which leads might elucidate them.
had medication-resistant symptoms at the
us to assume that individuals make differ-
levels described in a similar study of indivi-
ent adjustments following treatment which
dual CBT (Kuipen et al, 1997). Although
are only reflected in the total scores. There
people did drop out of the study it was Is treatment clinically use!ful?
were some relationships between the
possible to use the data from all 21 people
changes over therapy and the changes in Participants had experienced treatment-
recruited.
key outcome measures. For example, resistant distressing
- hallucinations for an
changes in voice powerfulnets did reduce average of 14 years. Although many re-
Does treatment produce distress, and half the group who completed ported that medication had helped, most
significant effects? treatment improved their perceived control said that they had never been free of these
The analyses suggest that the main outcome over the voices. This is an important result, experiences even when taking adequate
measures of global symptoms do change as people with schizophrenia who feel less doses. All the participants for whom we
over time in the direction predicted. Symp- control over their voices are more likely to have complete data expressed a high degree
toms changed little over the waiting period, be violent (Cheung et a1, 1997). Although of satisfaction with the group. Talking
and improved over the treatment phase; it has been suggested that increasing coping about the voices with others who had simi-
this improvement was maintained at follow- strategies might be beneficial (Lee et al, lar experiences was reported as being parti-
up. Clearly, the study may have benefited 1993; Carter et al, 1996), it was not di- cularly beneficial, and many patients
from a comparison with a control group rectly related to the outcome measures in commented on how easily they were able
that did not receive treatment. However, the present study. Rather, it seems that to communicate within the group. Many
the waiting-time control is a conservative the ability to engage in coping strategies patients said they were 'comforted' by the
measure of 'no treatment' as it includes might influence the person's perception of fact that they were not alone in their experi-
the effects of treatment expectancy. Signifi- control over their experience and the ences. Among the most frequently reported
cant treatment effects relative to control distress associated with the voices. benefits were the educational aspects of the
were found for both BPRS and IS, and there Unfortunately, many of these interesting therapy, particularly with regard to medi-
was a strong trend in the required direction possibilities cannot be tested with the cation, and the learning of new coping
 GROUP TREATMENT O F AUDITORY H A L L U C I N A T I O N S

skdh (such as " ~ 0 I I f r 0 n ~the

g voices and less dependence on expert therapist time
asking them to come back later"). and which could improve the prognosis of
many people with treatment-resistant
How does group treatment psychotic symptoms. Errlcr.B. (Im)Analysis dbng'iinal data. &yond
~ m w ~ . ~ r i t i s h j - d o1n.7-10.
f m
compare with individual CBT?
Group treatment produces reductions in the REFERENCES -.&LabbuqF.&-w(lm)Grovp
CBT for people with schizophrenia: a preliminary
total BPRS-E score (7.9 points) similar to

-
eValuation.&hoviocmland~Ifychahempy,26,
those achieved in individual treatment stu- -hlchhbic-(~)Dicgnostic
o n d ~ ~ o f ~ ~ ( 4 t h e d 63-75.
n )
dies (e.g. Kuipers et a1, 1997). However, (OW-N)Washington DC: APA.
the cost ratio when compared with indivi- -G,McC.rron.J..-.kr.l(Im
Back, A.T...wa&.C. H.. H e d m h n . n..(d(1961) M i Hallucinations Rating Scak. Scales to measure
dual care is 1:14. That is, one person himentory formeasvringdeprrssion ArdvLesof dimensions dhallucinationsand Mudom: the Psychotic
completing individual treatment for 14 G~IXIU! 41 561-567. Ratrng (rnJ.9 Rychdogicd
Mediche, in pen.
completing group treatment. It is, of -.~H.Bmwn.b.rd(I~Aninvendwy
course, possible that this form of therapy for measuringclinical anxiety: p@ometk properties.
lUprrkG=-%r..-.~..(d(lm
has fewer long-term benefits than indivi- I Consultmgond Uiniwl psych dog)^ Sb. 893-897.
J u u n ~of London-East Angli rMdomaed contrdled trial d
dual CBT. The follow-up in this study B W - V - m V w . ( d ( I 9 W A
cognrtive-behmbwal therapy for psycpsychosis.I: Efkcts d
self-report imight scak for psycho&: d i l i t y , validity the treatment phase British jwnd of Psychiou)c l7l.
was very short, and even then there was a
fall-off in the treatment effects. Alterna-
and sensitivity to change. Acta Rychiovro .- 319-327.
89.6247.
tively, group treatment could be considered Lw,Rw~U.h-twt,Sm~n.tal(tm)
as a supplement to individual CBT, intre
- &Chdrrldr.R(l~Theomnipotenadvoias: Coping strategiesd x h i i h h patients
. .
~ and their
tating the v a l i i y da cognitive model. P @ k g d ~ptooutcw.BritishjwndofRyrhioq1~
ducing the participants to the processes in- Medicine, 27,1345- 1353. 177- 182.
volved in individual work. It may then Cvtr,an.-A.&c~pdor.nL(lm) nosdq,n(tru)bwrycndtheAdckentw-
help to reduce the duration of individual F'atiem*strategiesbr coping with d i bnage. Prinaeton. Nj: Rinaeton Univadty Ras
treatments or even haease the effective- hallucinatiom)wndof k w n s cnd Mentol Disease,
184,159- 164.
ness of individual CBT. -,H.WL.-C.(d(Im)
Randomaed control trial of intensive cognitive
The current exploratory study does, of ~ R & b i d w o o d , ~ ( l W S ) T h e
m i p o t e n a d w i a s . 1l:The Beliefs Abwt UDias behaviourt~brpaticntswithchmnic
course, require replicating in a more stand- schizophrrnia. British MediwI JLWK$, 317.303-307.
Qumomire (BAVQZ BritishJarmd of PsyrhioV)( 166,
ard randomised controlled mal, especially 773-776.
as results are promising. Group treatment -J..Grm.nF..-&.(d(w
a u l l l r * k ~ L ~ n . . ( d ( 1 m Trsining and quality assurance with the Brief f'sychiatric
for hallucinations may be a practical alter- V I in s c h i m role of hallucinationsand Rating Scak. knwwtmrd Jcumal of M s in
native psychological treatment which has delusions. schi2Dphnmr Rcswrrh. 16,181-190. Pzychatric Reseadl, 1.221 -244.