www.thelancet.com/psychiatry Vol 2 April 2015
Lancet Psychiatry 2015;
March 4, 2015
See Comment page 282
Department of Psychiatry,
University of Oxford, Oxford,
UK (Prof D Freeman PhD,
H Startup DPhil, K Pugh DClinPsy,
E Černis MSc); Centre for
Biostatistics, Institute of
Population Health, University
of Manchester, Manchester, UK
and Medical Research Council
NorthWest Hub for Trials
Department of Biostatistics,
University of Liverpool,
(Prof G Dunn PhD);
Sussex Partnership NHS
Foundation Trust, Research
and Development Department,
Mill View Hospital, Hove, UK
(H Startup, K Pugh); and
Academic Department of
Psychiatry, Faculty of Medicine,
University of Southampton,
(J Cordwell DClinPsy,
H Mander DClinPsy,
G Wingham BSc, K Shirvell MSc,
Prof D Kingdon MD)
Prof Daniel Freeman, Department
of Psychiatry, University of
Oxford, Warneford Hospital,
Oxford OX3 7JX, UK
Eﬀ ects of cognitive behaviour therapy for worry on
persecutory delusions in patients with psychosis (WIT):
a parallel, single-blind, randomised controlled trial with
a mediation analysis
Daniel Freeman, Graham Dunn, Helen Startup, Katherine Pugh, Jacinta Cordwell, Helen Mander, Emma Černis, Gail Wingham,
Katherine Shirvell, David Kingdon
Background Worry might be a contributory causal factor in the occurrence of persecutory delusions in patients with
psychotic disorders. Therefore we postulated that reducing worry with cognitive behaviour therapy (CBT) would
reduce persecutory delusions.
Methods For our two-arm, assessor-blinded, randomised controlled trial (Worry Intervention Trial [WIT]), we recruited
patients aged 18–65 years with persistent persecutory delusions but non-aﬀ ective psychosis from two centres: the Oxford
Health National Health Service (NHS) Foundation Trust (Oxford, UK) and the Southern Health NHS Foundation Trust
(Southampton, UK). The key inclusion criteria for participants were a score of at least 3 on the Psychotic Symptoms
Rating Scale (PSYRATS) denoting a current persecutory delusion; that the delusion had persisted for at least 3 months; a
clinical diagnosis of schizophrenia, schizoaﬀ ective disorder, or delusional disorder; and a clinically signiﬁ cant level of
worry. We randomly assigned (1:1) eligible patients, using a randomly permuted block procedure with variable block sizes
and division by four strata, to either six sessions of worry-reduction CBT intervention done over 8 weeks added to standard
care (the CBT-intervention group), or to standard care alone (the control group). The assessors were masked to patient
allocations and did their assessments at week 0 (baseline), 8 weeks (end of treatment), and 24 weeks, follow-up. The
primary outcomes were worry measured by the Penn State Worry Questionnaire (PSWQ) and delusions measured by the
PSYRATS-delusion scale; we did the analyses in the intention-to-treat population, and also did a planned mediation
analysis. This trial is registered with the ISRCTN Registry (number ISRCTN23197625) and is closed to new participants.
Findings From Nov 1, 2011, to Sept 9, 2013, we recruited 150 eligible participants and randomly assigned 73 to the CBT
intervention group, and 77 to the control group. 143 patients (95%) provided primary outcome follow-up data.
Compared with standard care alone, at 8 weeks the CBT intervention signiﬁ cantly reduced worry (mean diﬀ erence
6·35 [SE 1·56] PSWQ units, 95% CI 3·30–9·40; p<0·001) and persecutory delusions (2·08 [SE 0·73] PSYRATS units,
95% CI 0·64–3·51; p=0·005). The reductions were maintained to 24 weeks follow-up. The mediation analysis
suggested that the change in worry accounted for 66% of the change in delusion. No patients died or were admitted
to secure units during our study. Six suicide attempts (two in the CBT intervention group, and four in the control
group) and two serious violent incidents (one in each group) were noted, but no adverse events were deemed related
to the treatments or the assessments.
Interpretation To our knowledge, this is the ﬁ rst large trial focused on persecutory delusions. We have shown that
long-standing delusions were signiﬁ cantly reduced by a brief intervention targeted on worry, although the limitations
for our study include no determination of the key elements within the intervention. Our results suggest that worry
might cause paranoia, and that worry intervention techniques might be a beneﬁ cial addition to the standard treatment
Funding Eﬃ cacy and Mechanism Evaluation programme, which is a UK Medical Research Council and National
Institute of Health Research partnership.
Copyright © Freeman et al. Open Access article distributed under the terms of CC BY.
Treatments for psychotic disorders such as schizophrenia
need substantial improvement. Our approach is to study
single psychotic experiences such as persecutory
delusions, establish a theoretical model, and translate
the knowledge gained into treatment. To build the
treatment, one putative causal factor is taken at a time,
changed, and the eﬀ ect on the psychotic occurrence
examined.1 This approach is called an interventionist-
causal model approach.2 In this Article, we report the
eﬀ ects of targeting one causal factor—worry—in
patients with persecutory delusions.
www.thelancet.com/psychiatry Vol 2 April 2015
Worry is an expectation of the worst happening. It
consists of repeated negative thoughts about potential
adverse outcomes, and is a psychological component of
anxiety. Worry brings implausible ideas to mind, keeps
them there, and increases the level of distress. Therefore
we have postulated that worry is a causal factor in the
development and maintenance of persecutory delusions,
and have tested this theory in several studies.
showed that levels of worry in patients with persecutory
delusions are similar to those noted in generalised
anxiety disorder;3 a dose-response association exists
between levels of worry and paranoia;4 worry is a
predictor of the occurrence and persistence of non-
clinical paranoia in the general population5,6 and in
experimental settings;7 and levels of worry predict the
persistence of persecutory delusions.8,9 Other study
groups are also replicating and extending these
ﬁ ndings.10,11 We have translated this knowledge into
treatment and shown in a pilot trial12 that a brief
intervention of worry-reduction added to standard care
might lead to reductions in both worry and persecutory
delusions. In the terminology of the scientiﬁ c literature,
worry in delusions is a so-called inus condition—“an
insuﬃ cient but non-redundant part of an unnecessary
but suﬃ cient disorder.”13 Persecutory delusions arise
from a combination of causes, with each causal factor
increasing the probability of such fears occurring.
We planned our trial as a rigorous test of these
mechanistic links to inform both theory and treatment.
A key mechanism (worry) was targeted. The appropriate
control condition was a standard care group to establish
that the mechanism had been successfully targeted,
which would then allow examination of the eﬀ ects of the
mechanism change on the central clinical occurrence
(persecutory delusions). We planned an elaborate
mediation analysis to substantiate the postulated
mechanism of delusion change. The aim of our study
was to investigate whether the intervention with
cognitive behaviour therapy (CBT) would reduce levels
of worry in patients with persecutory delusions and
reduce the delusions themselves; the improvements
would be maintained at follow-up; and the reduction in
worry would mediate changes in persecutory delusions.
Study design and participants
We did a randomised, controlled, single-blind trial in two
UK centres: the Oxford Health National Health Service
(NHS) Foundation Trust, Oxford, and the Southern
Health NHS Foundation Trust, Southampton. These
large mental health services cover populations of about
1·2 million people each. The trial received a favourable
opinion from an NHS Research Et hics Service Commit tee,
and the trial protocol has been published.14 We sou ght
referrals of patients aged 18–65 years with persecutory
delusions from both centres. The inclusion criteria were:
a current persecutory delusion as deﬁ ned by Freeman and
Garety,15 scoring at least 3 on the conviction scale of the
Psychotic Symptoms Rating Scale (PSYRATS);16 that the
delusion had persisted for at least 3 months; a clinical
diagnosis of schizophrenia, schizoaﬀ ective disorder, or
delusional disorder (ie, a diagnosis of non-aﬀ ective
psychosis); and a clinically signiﬁ cant level of worry, as
shown by a score of more than 44 on the Penn State Worry
Questionnaire (PSWQ).17 Where major changes in drugs
were going to be made, entry to the investigation would
not occur until at least 1 month after stabilisation of
dosage. Criteria for exclusion were: a primary diagnosis of
alcohol or substance dependency or personality disorder;
an organic syndrome or learning disability; a command of
spoken English that was inadequate for engaging in
therapy; and currently having individual CBT. All patients
provided written informed consent.
Randomisation and masking
We randomly assigned (1:1) eligible patients, after a
baseline assessment, to either six sessions of CBT
worry-reduction intervention done over 8 weeks added
to standard care (the CBT intervention group), or to
standard care alone (the control group). We used a web-
based randomisation system, written by the Oxford
Clinical Trials Unit for Mental Illness with a stratiﬁ ed
randomisation procedure including four strata and a
randomly permuted block procedure with variable
block sizes. We did the stratiﬁ cation on the basis of
centre and level of worry (deﬁ ned as moderate when
the PSWQ worry score was 44–62, and high when the
score was ≥63).
The assessors were masked to patients’ treatment
allocations, but all patients were informed of their
allocation by a trial therapist. Precautionary strategies
included thinking about the best room to use and diary
arrangements; patients being reminded by the assessors
not to talk about allocation; and, after the initial
assessment, the assessors did not look at clinical notes. If
an allocation was revealed to the assessor, then remasking
occurred, by use of another rater, which happened
11 times. However, if an allocation was revealed during
an assessment session then these ratings were used: two
8-week assessments (both with the intervention) and
four 24-week assessments (three with the intervention)
were done unmasked.
We aimed to provide the CBT worry-reduction inter-
vention in six sessions over 8 weeks. Each session lasted
roughly an hour and took place in NHS clinics or at
patients’ homes. Therapy was delivered individually.
Before therapy began the clinician met the patient for an
initial introduction and assessment. The assessments of
outcome measures were completed at 0 weeks (baseline),
8 weeks (end of therapy), and at 24 weeks (follow-up).
Three graduate psychologists (EČ, GW, and KS) did the
enrolment and assessments.
www.thelancet.com/psychiatry Vol 2 April 2015
The highly detailed intervention is designed to provide
clear and simple messages for patients to take into their
day-to-day lives. We wrote a set of six session booklets,
shared by the patient and therapist. The worry reduction
strategies included have been shown to be eﬀ ective at
reducing worry and do not challenge the delusion itself.
The main techniques were psychoeducation about
worry, identiﬁ cation and reviewing of positive and
negative beliefs about worry, increasing awareness of
the initiation of worry and individual triggers, use of
worry periods, planning activity at times of worry (which
could include relaxation), and learning to let go of worry.
We formulated a so-called worry cycle early in the
intervention: feeling under threat leads to activation of
positive beliefs about worry and hence engagement in
this thinking style, resulting in dwelling on the worst
outcomes and an increase in the initial feelings of
threat. The worry cycle was discussed in relation to a
recent bout of worry by the patient. Tasks were set
between sessions—eg, imple mentation of worry
periods. Whenever patients agreed, the trial therapists
telephoned or texted them between sessions, to
encourage them to try the new strategies. We helped
patients to learn that they had understandable positive
beliefs about worry (eg, that worry kept them safe) that
meant that they engaged with this thinking style. They
were helped to see the skewed view that worry provides
and how it exacerbates fears. The two main practical
techniques to reduce worry were then introduced: the
use of worry periods (conﬁ ning worry to about a 20 min
set period each day) and planning of activities at peak
worry times. Worry periods were implemented ﬂ exibly.
For example, most patients set up one worry period a
day, but they could choose to have two worry periods a
day or, in severe instances, patients instead aimed for a
worry-free period. Ideally, the worry period was then
substituted with a problem-solving period. Our general
approach and techniques are also described in a
Three clinical psychologists provided therapy (KP, JC,
and HM), and were supervised each week by DF and HS.
One of the therapists provided the intervention for all
participants in Oxford (KP). The trial began with another
therapist (JC) providing all therapy in Southampton,
although in the latter part of the trial a third therapist
took over (HM). We recorded therapy sessions when
patients gave permission. To assess treatment ﬁ delity,
12 recordings, chosen randomly, were rated on the
Cognitive Therapy Scale—Revised (CTSR)19 by an
independent clinical psychologist who was skilled in
CBT for psychosis. All chosen recordings were rated as
providing at least satisfactory cognitive therapy (ie, a
mean score of at least 3).
Standard care was delivered according to national and
local service protocols and guidelines. This usually
consists of prescription antipsychotic drugs, visits from a
community mental health worker, and regular outpatient
appointments with a psychiatrist. It was recorded with
the Client Service Receipt Inventory.20
The pre-speciﬁ ed primary outcome measures were levels
of worry assessed by the PSWQ21 and levels of persecutory
delusions assessed by the PSYRATS-delusions scale.16
High scores on these scales indicate high levels of worry
and delusions, respectively. Secondary outcome measures
were delusion distress measured by the PSYRATS-distress
scale; total psychiatric symptoms measured by the Positive
and Negative Syndromes Scale (PANSS);22 paranoia
Figure 1: Trial proﬁ le
PSWQ=Penn State Worry Questionnaire. CBT=cognitive behavioural therapy.
73 allocated to worry-reduction
CBT intervention plus
64 received allocated
9 did not receive allocated
intervention (deﬁned as
three or fewer sessions)
3 lost to
4 lost to
77 allocated to control
(standard care alone)
70 attended 8 week follow-up 73 attended 8 week follow-up
68 attended 24 week follow-up 73 attended 24 week follow-up
73 analysed (intention-to-treat
77 analysed (intention-to-treat
5 lost to
4 lost to
276 assessed for eligibility
441 participants referred
113 were excluded
8 had insuﬃcient score on PSWQ
102 had no current persecutory delusion
2 had insuﬃcient capacity to consent
1 was too acutely unwell to engage
13 were suitable but declined to participate
163 declined to be screened
2 excluded because of high forensic risk
www.thelancet.com/psychiatry Vol 2 April 2015
measured by the Green et al Paranoid Thoughts Scale
(GPTS);23 rumination measured by the Perseverative
Thinking Questionnaire (PTQ);24 an adapted service user-
led measure of patient outcomes (CHOICE)25 assessing—
eg, self-conﬁ dence, having coping strategies, and a sense
of being in control; and wellbeing measured by the
Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS).26
High scores on these scales indicate delusion distress,
higher overall levels of psychiatric symptoms, paranoia,
rumination, patient satisfaction, and psychological
wellbeing. We tested interrater reliability for the two
interviewer-rated assessments, with two-way mixed, one-
measure intraclass correlations (ICC).
At baseline, to examine additional moderators of out-
come, participants completed assessments of intel lectual
functioning (the Wechsler Adult Intelligence Scale
[WAIS]),27 illicit drug use (the Maudsley Addiction Proﬁ le),28
illness and treatment representations,29 probabilistic
reasoning,30 and working memory (appendix).31,32
During the trial, we recorded any adverse event that
came to our attention. We also checked medical notes at
the end of the trial for the following events prespeciﬁ ed
as adverse: all deaths, suicide attempts, serious violent
incidents, admissions to secure units, and formal
complaints about therapy.
Our target sample size was 150 patients, split equally
between the two centres. We wanted to detect moderate
or large eﬀ ects. A simple two-tailed t-test with 60 people
per group would provide 90% power to detect an eﬀ ect
size of 0·60 at a signiﬁ cance level of 0·05, and would
have 80% power to detect an eﬀ ect size of 0·52. In
practice, further power would be gained by use of
multiple regression. Therefore, conservatively allowing
for a 20% dropout, 150 people would need to be recruited
to enable full data to be obtained from 120 participants.
We did all main analyses at the end of the last follow-up
assessments at week 24 (ie, we did not do any interim
analyses) with Stata version 13,33 in the intention-to-treat
population, with due consideration being given to potential
biases arising from loss to follow-up. Random or mixed
eﬀ ects models (with Stata’s xtreg command) were ﬁ tted to
the repeated measures to estimate treatment eﬀ ects for
outcomes, controlling for stratum (treatment centre
crossed by the initial level of worry; ie, moderate or high),
and the corresponding baseline assessment for the
outcome being investigated. To ﬁ nd out whether the
intervention eﬀ ects diﬀ ered at 8 weeks compared with
24 weeks (ie, whether eﬀ ects were maintained), we also
tested treatment by follow-up time interactions; this
analysis tested whether diﬀ erences in the intention-to-treat
eﬀ ects at the two follow-up times were signiﬁ cant. We
allowed for the presence of missing outcome data under
the assumption that the data were missing at random.34 We
calculated standard eﬀ ect sizes (Cohen’s d) by dividing the
estimated treatment eﬀ ects by the pooled SD at follow-up.
We did all mediation analyses using the structural
equation modelling package Mplus Version 7 (appendix).35
Our mediation analysis strategy was similar to that
advocated by Baron and Kenny36—ie, we tested for
intervention eﬀ ects on the outcome (delusions) and on
the proposed mediator, then ﬁ tted a full model to estimate
the direct and indirect eﬀ ects of the intervention on
outcome—but with statistical models that account for the
repeated measures of both mediator and outcome (ie, a
parallel process model),37 acknowledge that confounding
of the eﬀ ect of mediator on outcome is probable,38 and
allow for the fact that the mediator and outcome are
subject to substantial measurement error.39
Age (years) 40·9 (10·5) 42·1 (12·2)
Male 42 (58%) 44 (57%)
Female 31 (42%) 33 (43%)
White 68 (93%) 69 (89%)
Black 1 (1%) 0 (0%)
Chinese 0 (0%) 2 (3%)
Indian 0 (0%) 3 (4%)
Other 4 (6%) 3 (4%)
Unemployed 55 (75%) 51 (66%)
Part-time employed 8 (12%) 6 (7%)
Full-time employed 3 (4%) 10 (13%)
Self employed 1 (1%) 2 (3%)
Retired 2 (3%) 6 (8%)
Student 1 (1%) 2 (3%)
Housewife or husband 3 (4%) 0 (0%)
Intelligence quotient 100·3 (19·0) 101·8 (18·2)
Schizophrenia 58 (79%) 53 (69%)
Schizoaﬀ ective disorder 5 (7%) 6 (7%)
Delusional disorder 4 (5%) 6 (7%)
Psychosis NOS 6 (8%) 12 (16%)
Outpatient 71 (97%) 76 (99%)
Inpatient 2 (3%) 1 (1%)
Inpatient admission in
previous 6 months
10 (14%) 8 (10%)
dose of antipsychotic drug
523·2 (394·3) 475·5 (420·6)
Time in contact with services
<1 year 5 (8%) 7 (9%)
1–5 years 12 (16%) 17 (22%)
6–10 years 16 (22%) 12 (16%)
11–20 years 18 (25%) 26 (34%)
>20 years 21 (29%) 15 (19%)
Data are n (%) or mean (SD). NOS=not otherwise speciﬁ ed.
Table 1: Baseline characteristics of the intention-to-treat population
See Online for appendix
www.thelancet.com/psychiatry Vol 2 April 2015
We used data from both treatment groups in these
analyses—essentially assessing what proportion of the
intention-to-treat eﬀ ect of the worry intervention on
delusions is attributed to its eﬀ ect on worry. All statistical
testing was two-tailed.
Level of worry was assumed to be the mediator and
severity of paranoia the ﬁ nal outcome (rather than vice
versa)—primarily motivated by the fact that the
intervention was speciﬁ cally targeted on worry as the
mechanism of change. The parameters of the chosen
model were then estimated assuming the underlying
validity of the model.
We started with two simple measurement or factor
analysis models—the ﬁ rst for worry and the second for
delusions. In each case, the loadings for 8 and 24 weeks
were constrained to be 1, the intercept term for each
timepoint was constrained to be 0, and the variances of the
measurement errors were equal for the two timepoints.
We assumed that measurements at the follow-up times
were parallel measures of a stable underlying latent
variable.35 The measurement errors for worry and
delusions were correlated at 8 weeks and 24 weeks.
We estimated the eﬀ ects of the intervention on the worry
outcome factor, the delusions outcome factor, and the
eﬀ ect of the worry factor on the outcome factor, allowing
for a direct eﬀ ect of the intervention on the outcome. In
practice, the worry and delusions outcome factors were
assessed in a joint structural equation model, allowing for
the residual (ie, not accounted for by the intervention and
baseline covariates) variation in worry and delusion to be
correlated (as would be expected if mediation were
present). For the eﬀ ect of worry on the outcome, we jointly
modelled the eﬀ ect of the intervention on worry and the
eﬀ ect of worry and the intervention on outcome (this time
not allowing the residuals to be correlated).
We allowed for confounding mainly by inclusion of the
baseline values of both worry and delusions (in addition
to the stratifying factors) in all the structural equation
model analyses. In the intention-to-treat analyses there
was no diﬃ culty of confounding and the covariates were
included to strengthen precision. In the mediation
analysis we looked at a non-randomised comparison
(neither mediator or outcome are under the direct control
of the investigator), and confounding might therefore be
present. A major source of such confounding is likely to
be the correlation between the baseline values of worry
and delusions (estimated here to be 0·51).
The main mediation analysis model was essentially
equivalent to an analysis of covariance model for the
eﬀ ects on the intervention on the latent outcome common
to 8 weeks and 24 weeks outcome, conditioning on the
corresponding latent mediator and baseline covariates.
An alternative approach to the analysis might have been
through the use of latent change score models37—but, if
no changes were shown in either mediator or outcomes
between 8–24 weeks follow-up, the results of ﬁ tting an
appropriately parameterised and constrained latent
change model would yield identical results (ie, identical
goodness-of-ﬁ t indices and identical parameter estimates
for the direct and indirect eﬀ ects of the intervention;
appendix). A data monitoring and ethics committee
oversaw our study. This trial is registered with the
ISRCTN Registry, number ISRCTN23197625.
Role of the funding source
The funder of the study reviewed the application for the
trial and monitored the progress of trial milestones (eg,
recruitment). The funder had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report. The corresponding author had full
access to all the data in the study and had ﬁ nal
responsibility for the decision to submit for publication.
Between Nov 1, 2011, to Sept 9, 2013, with the last
assessments completed on March 10, 2014, we assessed
276 participants, of whom 150 were eligible, gave
imformed consent, and were randomly assigned to
either the CBT intervention group (n=73) or to the
control group (n=77; ﬁ gure 1). As with other studies of
persistent psychotic occurrences, both groups had a
Panel 1: Patient comments on the intervention
“The discussions about preventing worry and reducing worry were extremely helpful to
me. It made me see my worry as something real. The breakthrough was that I was able to,
with the help of my psychologist, come up with a strategy—that is, when worry [was]
gripping me I would say “’excuse me worry, while I do…’” or “’excuse me worry, I need to
interrupt you because….’” I sometimes worry about people trying to harm me but now
I can interrupt my worry and do something else. I challenge myself to do this because
I know it works for me.”
“I had no conﬁ dence in who I was and felt I avoided everyone because of my thoughts and
being in company was really frightening. I found the therapy challenging and sometimes
very diﬃ cult. But it was eye-opening as I didn’t realise how much I worried and where the
worry was coming from. To see it on paper made it more straightforward and made my
life more clear. I do feel that I now try to take time out, whether that’s a cup of tea or
going to the shops. Just doing things that I actually enjoy doing and building on them as
I was so wrapped up in anxiety I was lost. I am more relaxed at certain times of the day
where I was once completely stressed. I still ﬁ nd it hard around people but I feel I can still
build on the skills you gave me and if it’s slightly better, that’s good.”
“The therapy was very rewarding. There wasn’t anything I didn’t like. I needed that kind of
therapy at the time because if I didn’t have that therapy at that time, I wouldn’t be here. It
was therapeutic talking about things. I listened to what you had to say and wrote down
how I felt. I also tried relaxing to the tape and I ignored people when they were horrible to
me. It was hard becoming disciplined but we worked as a team, that’s what I liked about it.
You don’t get nowhere in this world if you don’t work as a team. I was having a hard time
and you was doing your best to stop me having a hard time. That’s what I call team work;
I couldn’t have been able to do it by myself, no way. I thought a lot about what I thought
the therapy did—it decreased my worrying but in other ways it built my conﬁ dence.”
www.thelancet.com/psychiatry Vol 2 April 2015
slightly higher preponderance of men than women, the
mean age was around 40 years, most were unemployed,
and the main psychiatric diagnosis was schizophrenia.
All but nine patients were taking antipsychotic drugs
(one in the CBT group, eight in the control group). Most
patients had been in contact with mental health services
for many years (table 1).
The mean number of sessions received was 5·5 (SD 1·8);
51 patients attended six sessions. In the interest of
ﬂ exibility, for a few patients the intervention was provided
in seven (n=7) or eight sessions (n=2) during the 8 week
period. Two patients attended no therapy sessions. The
remainder of the patients attended one (n=5), two (n=1),
three (n=1), four (n=3), or ﬁ ve (n=1) sessions. Panel 1
shows patient comments about the intervention. An
analysis of the eﬀ ects of increasing compliance with
therapy had been proposed in the published trial protocol14
but, in the event, compliance with the allocated intervention
was so high that such an analysis was deemed unnecessary.
The therapist in Oxford provided the intervention to
37 participants. The two therapists in Southampton
provided the intervention to 22 and 14 participants,
respectively. The number of trial participants that can be
used as controls for each of these three therapists was
37 for Oxford, and 23 and 13 for Southampton. In the
sensitivity analyses allowing for therapist eﬀ ects described
in the section on mediation, trial participants were, in
eﬀ ect, stratiﬁ ed by therapist instead of centre. For inter-
rater reliability tests, when rater 1 attended 23 assessments
with rater 2, their reliability ratings were PSYRATS total
ICC=0·99, PANSS total ICC=0·83. Rater 1 attended
18 assessments with rater 3 and their reliability ratings
were PSYRATS total ICC=0·98, PANSS total ICC=0·75.
When compared with standard care alone, the CBT
intervention led to a signiﬁ cant reduction in levels of
worry (table 2). The estimated mean diﬀ erence in PSWQ
scores at 8 weeks between the CBT-intervention group
and the control group was 6·35 (SE 1·56; 95% CI
3·30–9·40; p<0·001). Persecutory delusions were also
reduced in the CBT-intervention group compared with
the control group; the estimated mean diﬀ erence in
PSYRATS scores at 8 weeks in the intervention group
compared with the standard care group was 2·08 (SE 0·73;
95% CI 0·64–3·51; p=0·005). The mean treatment by
follow-up time (8 and 24 weeks) interactions were
estimated to be –2·43 PSWQ (SE 1·57; p=0·121) and
0·86 PSYRATS (SE 0·68; p=0·205), suggesting that at
24 weeks, the treatment eﬀ ects were smaller for PSWQ,
but larger for PSYRATS. However, neither of these
interactions were signiﬁ cant and the statistical models
were reﬁ ned to estimate treatment eﬀ ects (ie, diﬀ erences
in average outcome between the two randomised groups)
that were assumed to be common to both follow-up
times. The resulting treatment-eﬀ ect estimates were
5·15 (SE 1·35; 95% CI 2·50–7·79; p<0·001; Cohen’s
d=0·47) and 2·50 (SE 0·65; 95% CI 1·22–3·78; p<0·001;
Cohen’s d=0·49). No substantial temporal trends in the
mediator or the outcome between 8 and 24 weeks were
noted, substantially simplifying the statistical models
needed for the analysis of the associations between
changes in the mediator and the corresponding changes
in clinical outcome.
Signiﬁ cant improvements were noted with the CBT
treatment for all the secondary outcome measures. There
were no signiﬁ cant treatment by follow-up time inter-
actions (ie, intention-to-treat eﬀ ects did not signiﬁ cantly
diﬀ er between 8 weeks and 24 weeks), and therefore
treatment estimates common to both follow-ups were
made. Compared with standard care alone, CBT inter-
vention reduced mean PSYRATS distress scores (0·85,
CBT intervention group Control group
n Mean (SD) n Mean (SD)
0 weeks 73 64·8 (8·6) 77 64·5 (9·5)
8 weeks 70 54·8 (10·5) 73 61·0 (12·2)
24 weeks 68 56·1 (9·7) 73 59·8 (11·0)
0 weeks 73 18·7 (3·0) 77 18·0 (3·0)
8 weeks 70 14·3 (4·8) 73 15·9 (5·1)
24 weeks 68 13·6 (5·6) 72 16·4 (4·8)
Delusion distress (PSYRATS-distress)
0 weeks 73 6·4 (1·4) 77 6·5 (1·3)
8 weeks 70 5·1 (1·9) 73 5·8 (2·1)
24 weeks 68 5·0 (2·2) 72 6·1 (1·8)
Total symptoms (PANSS)
0 weeks 73 82·0 (13·6) 76 79·0 (13·5)
8 weeks 69 70·7 (12·4) 73 75·3 (16·0)
24 weeks 68 71·5 (15·4) 71 76·3 (16·7)
0 weeks 73 115·9 (27·3) 77 110·8 (27·8)
8 weeks 70 90·0 (32·2) 73 102·3 (31·7)
24 weeks 67 92·5 (32·7) 73 105·6 (32·4)
0 weeks 70 44·3 (9·7) 72 44·9 (9·8)
8 weeks 68 37·7 (9·7) 70 41·0 (11·7)
24 weeks 64 37·3 (10·5) 71 42·7 (10·6)
Patient outcomes (CHOICE)
0 weeks 71 49·4 (17·3) 75 49·5 (18·5)
8 weeks 67 64·4 (17·1) 69 51·7 (21·1)
24 weeks 66 61·6 (21·4) 70 52·5 (22·4)
0 weeks 73 36·4 (9·6) 77 34·5 (9·2)
8 weeks 68 41·5 (9·1) 73 36·5 (11·3)
24 weeks 67 40·2 (10·8) 73 36·6 (10·5)
CBT=cognitive behavioural therapy. PSWQ=Penn State Worry Questionnaire. PSYRATS=Psychotic Symptoms Rating Scale.
PANSS=Positive and Negative Syndromes Scale. GPTS=Green et al Paranoid Thoughts Scale. PTQ=Perseverative Thinking
Questionnaire. CHOICE=CHoice of Outcome In Cbt for psychosEs. WEMWBS=Warwick-Edinburgh Mental Wellbeing Scale.
Table 2: Primary and secondary outcome measures
www.thelancet.com/psychiatry Vol 2 April 2015
SE 0·25, p=0·001, Cohen’s d=0·41), PANSS psychiatric
symptom scores (6·16, SE 1·69, p<0·001, Cohen’s
d=0·42), paranoia GPTS scores (14·68, SE 4·18, p<0·001,
Cohen’s d=0·45), and rumination PTQ scores
(3·51, SE 1·43, p=0·014, Cohen’s d=0·32). We noted
improvements in the intervention group versus standard
care group in psychological wellbeing WEMWBS scores
(2·40, SE 1·11, p=0·03, Cohen’s d=0·23) and patient
chosen outcomes CHOICE scores (10·45, SE 2·42,
p<0·001, Cohen’s d=0·52).
Treatment eﬀ ects were not moderated by centre,
therapist, level of worry or delusions, intellectual func-
tioning, illicit drug use, illness perceptions, reasoning, or
working memory (p>0·05).
Figure 2 provides an overview of the mediation analysis.
The CBT intervention reduced the worry factor by a mean
of 5·66 (SE 1·32, 95% CI 3·08–8·24; p<0·001) and the
delusions factor by a mean of 2·33 units (SE 0·64, 95% CI
1·08–3·58; p<0·001). The intervention directly reduced
the delusion factor by a mean of 0·80 (SE 0·65, 95% CI
–0·70 to 2·07; p=0·214). Each unit reduction in the worry
factor produced a 0·27 change in the delusions factor
(SE 0·06, 95% CI 0·15–0·39; p<0·001). The estimated
indirect (mediated) eﬀ ect of the intervention on the
delusions factor was a reduction of 1·53 (SE 0·49, 95% CI
0·57–2·48; p=0·002). The proportion of the eﬀ ect of the
intervention on outcome (delusions) that is mediated by
changes in worry is therefore 1·53/2·33=66%. The
structural equation model ﬁ tted the data as shown by a
χ² score of 20·03 with 17 degrees of freedom (p=0·273), a
root mean square error of approximation of 0·035, and
comparative ﬁ t index of 0·992.
One concern about the validity of the estimate of eﬀ ect
of change in worry on change in delusions came from
the possibility of confounding arising from diﬀ erential
therapist eﬀ ects. However, when we used therapist
identity as a covariate in the models instead of treatment
centre (but not including a worry stratum by therapist
interaction), the estimated eﬀ ect of worry on delusions
was unchanged: 0·27 (SE 0·06). The further addition of
the therapists by treatment interactions (acknowledging
that diﬀ erences might occur in the eﬀ ectiveness of the
therapists) as covariates produced identical results.
The standard care provided for each group was similar
between groups (table 3). Data for the number of days in
hospital is skewed for the CBT treatment group, because
one patient was in hospital for 2 years before entering
the trial, although they were discharged 3 months into
Two patients did not give us permission to check
medical notes at the end of the trial. No deaths,
admissions to secure units, or formal complaints about
therapy occurred during the trial. There were six suicide
attempts (two in the treatment group, four in the control
group) and two serious violent incidents (one in each
allocation group). None of the adverse events were
related to therapy or the assessments.
The results of the planned analysis were entirely
consistent with the inference that treating worry in
patients with persecutory delusions leads to reductions
in delusions. With the psychological treatment, patients
also had several other important outcomes, such as a
reduction in overall levels of psychiatric symptoms and
general level s of paranoid think ing, and an improvement
in psychological wellbeing (panel 1, 2).
Traditionally, a fundamental divide has been made
between neurosis and psychosis. Worry was studied and
treated in emotional disorders, but not in psychosis.
Ironically, our WIT study, to our knowledge,40 is the
largest trial so far of a psychological treatment for
patients with clinical worry, but it was undertaken in
patients with diagnoses of psychosis. Our study was
based on a theoretical understanding of the role of worry
in delusions, empirical studies that suggested an
important link, and the results of a promising pilot
study.12 The group given treatment had severe persecutory
delusions that had not responded suﬃ ciently to other
treatments. The main outcomes were very clear. A brief
cognitive behavioural intervention for worry, compared
with treatment as usual, led to signiﬁ cant reductions in
both worry and the persecutory delusions.
Patients liked the focus on worry, seen in the high
uptake of the therapy sessions. They agreed that they had
this problem; nonetheless, by reducing their preoccupation
with threat and increasing activity levels, the persecutory
delusions were implicitly challenged. Some patients, by
being more active with the goal of dealing with worry,
learned that they were safer outside than they had feared.
Only eight patients with persecutory delusions were
excluded from entering the trial on the basis of reporting
insuﬃ cient worry. The intervention was deliberately
highly detailed to help with later dissemination. The
length of therapy was remarkably short to achieve such
change in long-standing delusional beliefs. Agreeing to
six sessions help both the patient and therapist to initiate
Figure 2: Mediation analysis
Rectangles or squares represent measured variables. Ellipses or circles represent
latent variables (including random errors or residuals). Single headed arrows
represent predisposing eﬀ ects; bold arrows represent main ones of interest.
Double-headed arrows represent correlations. W8=worry measures at 8 weeks.
W24=worry measures at 24 weeks. D8=delusion measures at 8 weeks.
D24=delusion measures at 24 weeks. e1 and e2=random residuals (worry).
e3 and e4=random residuals (delusions).
Total eﬀect=–0·80–5·66 × 0·27=–2·33
Mediated eﬀect=–5·66 × 0·27=–1·53 (66%)
www.thelancet.com/psychiatry Vol 2 April 2015
active techniques early, and keeps therapy precisely
focused. Nevertheless, we do not envisage that the worry
intervention is suﬃ cient psychological help for these
patients; they still had signiﬁ cant levels of worry and
paranoia at 24 weeks follow-up—therefore the beneﬁ ts
need to be enhanced and maintained over longer periods.
We are now beginning to test the worry intervention in
combination with modular interventions targeting other
key causal factors, such as sleep disturbance, reasoning
biases, and low self-esteem.1 The intervention will
probably have wider applicability—eg, to patients at high
risk of psychosis,41 patients at ﬁ rst episode of psychosis,
and to patients with other disorders for which worry is a
putative contributory cause.
Our investigation had three main limitations. We did
not include a condition to control for therapist contact;
however, this was because the most important aspect in
this explanatory study was to show a change in the
putative causal factor—worry—so that any eﬀ ects on
delusions could be assessed. In this mechanistic trial,
change in the worry thinking style needed to be
established, not the components of therapy that might
achieve this. For example, although we think it highly
unlikely that befriending or supportive counselling
would have such persistent eﬀ ects on worry and
delusions, this possibility will have to be tested
speciﬁ cally in this patient group. Importantly, sub-
stantial limitations exist in what can be established
deﬁ nitively with regard to mediation. In our investi-
gation, we could not rule out the possibility that the
intervention has merely created non-speciﬁ c change in
a range of outcomes; against this possibility, the largest
eﬀ ect sizes for psychiatric symptoms were for the two
that were targeted—worry and persecutory delusions.
These positive eﬀ ects of six sessions of therapy persisted
at 6 months. Worry is a transdiagnostic process, and
therefore many beneﬁ ts could probably be gained by
reducing worry. (The control group showed some
improvement, which is typical with the monitoring that
occurs during a clinical trial.) We did not aim to
measure temporal associations between changes in
worry and changes in the delusion. Although the worry
style was the target of intervention, and not the content
of the delusions, the statistical models cannot
deﬁ nitively rule out reverse causation—indeed, a
reciprocal association between worry and paranoia is
plausible—or possible hidden confounding (particularly
those arising from experiences and life-events that
occurred during the trial but were assumed to be
unrelated to the trial intervention). Overall, we note the
advice of Bullock and colleagues42 “to think of mediation
analysis as a cumulative enterprise”. The study cannot
deﬁ nitively show mediation, but the results are
consistent with reports in the theoretical and empirical
scientiﬁ c literature and the focus of the intervention
techniques. Finally, follow-up was only roughly
4–6 months after the end of treatment, though we
regard this time as appropriate for such a short
intervention. In clinical practice, booster sessions
should be added. We hope to see further clinical trials
that focus speciﬁ cally on persecutory delusions.
DF, GD, HS, and DK designed the trial. DF took the main
responsibility for drafting the study report. DF was the main lead
for the trial, and led research in Oxford; DK led the investigation in
Southampton. GD did the analyses of the trial outcome and
mediation. HS was the trial coordinator. DF, HS, and DK provided the
training and supervision for the trial therapists and research workers.
KP, JC, and HM provided the therapy. EČ, GW, and KS did the data
collection. All authors contributed to, read, and approved the ﬁ nal
Declaration of interests
We declare no competing interests.
CBT intervention group Control group
n Mean (SD) n Mean (SD)
6 months before the trial
Number of days in hospital 73 7·4 (26·8) 77 2.·8 (9·5)
Meetings with psychiatrist 72 2·4 (3·9) 77 2·8 (4·2)
Meetings with community psychiatric nurse 72 12·3 (9·9) 76 10·5 (10·1)
Meetings with counsellor or therapist 72 1·5 (6·2) 77 1·1 (4·7)
Visits to day-care centre 72 0·8 (4·3) 77 1·7 (10·6)
GP meetings 73 3·8 (4·8) 77 2·6 (3·2)
6 months during the trial
Number of days in hospital 73 3·5 (15·0) 77 0·2 (1·6)
Meetings with psychiatrist 65 1·6 (1·9) 71 1·8 (2·2)
Meetings with community psychiatric nurse 65 11·2 (11·3) 71 9·2 (13·9)
Meetings with counsellor or therapist (outside
of the trial)
61 1·0 (3·6) 66 1·1 (3·4)
Visits to day-care centre 65 0·4 (2·6) 71 1·0 (6·3)
GP meetings 65 2·6 (2·6) 71 2·6 (2·5)
Data are n, mean (SD). CBT=cognitive behavioural therapy. GP=general practitioner.
Table 3: Standard care provided in the CBT intervention group and the control group
Panel 2: Research in context
We searched the ISRCTN trial registry and the PubMed database with the search terms
“worry”, “delusions”, “persecutory”, “paranoia”, and “schizophrenia” without date restrictions,
for English-language publications of randomised controlled trials investigating the treatment
of worry in patients with persecutory delusions. Other than our pilot investigation12 there
were no other such clinical trials in the medical literature. We also examined published
meta-analyses on standard cognitive behavioural therapy (CBT) for persistent delusions or
hallucinations, or both.43
After our pilot study,12 we have shown for the ﬁ rst time that treating worry leads to
reductions in persecutory delusions. This is new for the evidence-base and consistent
with (though not deﬁ nitive for) a role for worry as a contributory cause of paranoia.
The eﬀ ect size is similar to standard CBT for persistent psychotic occurrences,
according to meta-analyses.43 Standard CBT for psychosis does not include standard
worry-intervention techniques, but our intervention is much briefer. The trial shows the
promise of taking a focused theoretically driven approach to the treatment of psychosis.
www.thelancet.com/psychiatry Vol 2 April 2015
18 Freeman D, Freeman J. How to keep calm and carry on. Harlow:
19 Blackburn I, James I, Milne D, et al. (2001). The revised cognitive
therapy scale (CTS-R). Behav Cogn Psychother 2001; 29: 431–46.
20 Beecham J, Knapp M. Costing psychiatric interventions. In:
Thornicroft G, Brewin CR, Wing JK, eds. Measuring Mental Health
Needs. London: Gaskell, 1992: 163–84.
21 Meyer TJ, Miller ML, Metzger RL, Borkovec TD. Development and
validation of the Penn State Worry Questionnaire. Behav Res Ther
1990; 28: 487–95.
22 Kay SR. Positive and negative syndromes in schizophrenia. New York:
23 Green C, Freeman D, Kuipers E, et al. Measuring ideas of persecution
and reference: the Green et al Paranoid Thought Scales (G-PTS).
Psychol Med 2008; 38: 101–11.
24 Ehring T, Zetsche U, Weidacker K, Wahl K, Schönfeld S, Ehlers A.
The Perseverative Thinking Questionnaire (PTQ): validation of a
content-independent measure of repetitive negative thinking.
J Behav Ther Exp Psychiatry 2011; 42: 225–32.
25 Greenwood K, Sweeney A, Williams S, et al. CHoice of Outcome In
Cbt for psychosEs (CHOICE): the development of a new service
user–led outcome measure of CBT for psychosis. Schizophr Bull 2010;
26 Tennant R, Hiller L, Fishwick R, et al. The Warwick-Edinburgh
Mental Well-being Scale (WEMWBS): development and UK
validation. Health Qual Life Outcomes 2007; 5: 63.
27 Wechsler D. (1999). Wechsler Abbreviated Scale of Intelligence
(WASI). San Antonio TX: The Psychological Corporation, 1991.
28 Marsden J, Gossop G, Stewart D, et al. The Maudsley Addition Proﬁ le
(MAP). Addiction 1998; 93: 1857–67.
29 Weinman J, Petrie KJ, Moss-Morris R, Horne R. The illness pq.
Psychol Health 1996; 11: 431–45.
30 Garety PA, Freeman D, Jolley S, et al. Reasoning, emotions and
delusional conviction in psychosis. J Abnorm Psychol 2005; 114: 373–84.
31 Wechsler D. Wechsler Adult Intelligence Scale-III/Wechsler
Memory Scale. Third edition technical manual. San Antonio, TX:
The Psychological Corporation, 1997.
32 Freeman D, Startup H, Dunn G, et al. Understanding jumping to
conclusions in patients with persecutory delusions: working memory
and intolerance of uncertainty. Psychol Med 2014; 44: 3017–24.
33 StataCorp Stata Statistical Software: Release 13. StataCorp LP,
College Station, TX, (2013).
34 Little RJA, Rubin DB. Statistical Analysis with Missing Data, 2nd ed.
New York: Wiley, Chichester and New York, 2002.
35 Muthén LK, Muthén BO. Mplus User’s Guide, 7th edn. Los Angeles,
CA: Muthén & Muthén, 1998–2012.
36 Baron R, Kenny DA. The moderator-mediator variable distinction in
social psychological research. J Pers Soc Psychol 1986; 5: 1173–82.
37 McArdle JJ. Latent variable modelling of diﬀ erences and changes in
longitudinal data. Annu Rev Clin Psychology 2009; 60: 577–605.
38 Emsley R, Dunn G, White IR. Modelling mediation and moderation
of treatment eﬀ ects in randomised controlled trials of complex
interventions. Stat Meth Med Res 2010; 19: 237–70.
39 Dunn G. Statistical evaluation of measurement errors: design and
analysis of reliability studies, 2nd edn. London: Arnold, 2004.
40 Hanrahan F, Field A, Jones F, Davey G. A meta-analysis of cognitive
therapy for worry in generalised anxiety disorder. Clin Psychol Rev
2013; 33: 120–32.
41 van der Gaag M, Smit F, Bechdolf A, et al. Preventing a ﬁ rst episode
of psychosis: a meta-analysis of randomized controlled prevention
trials of 12 month and longer-term follow-ups. Schizophr Res 2013:
42 Bullock J, Green D, Ha S. Yes, but what’s the mechanism? (don’t
expect an easy answer). J Pers Soc Psychol 2010; 98: 550–58.
43 van der Gaag M, Valmaggia L, Smit F. The eﬀ ects of individually
tailored formulation-based cognitive behavioural therapy in auditory
hallucinations and delusions: a meta-analysis. Schizophr Res 2014;
This project (09/160/06) was awarded by the Eﬃ cacy and Mechanism
Evaluation (EME) Programme, and is funded by the UK Medical Research
Council (MRC) and managed by the UK NHS National Institute for
Health Research (NIHR) on behalf of the MRC-NIHR partnership. DF is
supported by a UK MRC Senior Clinical Fellowship (G0902308). Methods
research by GD is supported by the MRC (MR/K006185/1). We thank all
the trial participants; the independent members of the Trial Steering
Committee (Thomas Craig, Anthony Morrison, John Norrie) and the Data
Monitoring and Ethics Committee (Douglas Turkington, Paul French,
Sabine Landau); Katie Ashcroft for rating the therapy tapes; and the
clinical teams in Oxford Health NHS Foundation Trust and Southern
Health NHS Foundation Trust for their support of the trial. The views
expressed in this publication are those of the authors and not necessarily
those of the Medical Research Council, National Health Service, National
Institute of Health Research, or the Department of Health.
1 Freeman D, Garety P. Adv ances in understanding and treating
persecutory delusions. Soc Psychiatry Psychiatr Epidemiol 2014;
2 Kendler KS, Campbell J. Interventionist causal models in psychiatry:
repositioning the mind-body problem. Psychol Med 2009; 39: 881–87.
3 Freeman D, Garety PA. Worry, worry processes and dimensions of
delusions. Behav Cogn Psychother 1999; 27: 47–62.
4 Freeman D, Pugh K, Vorontsova N, Antley A, Slater M. Testing the
continuum of delusional beliefs: an experimental study using virtual
reality. J Abnorm Psychol 2010; 119: 83–92.
5 Freeman D, Stahl D, McManus S, et al. Insomnia, worry, anxiety and
depression as predictors of the occurrence and persistence of
paranoid thinking. Soc Psychiatry Psychiatr Epidemiol 2012;
6 Freeman D, Thompson C, Vorontsova N, et al. Paranoia and
post-traumatic stress disorder in the months after a physical assault:
a longitudinal study examining shared and diﬀ erential predictors.
Psychol Med 2013; 43: 2673–84.
7 Freeman D, Pugh K, Antley A, et al. A virtual reality study of paranoid
thinking in the general population. Br J Psychiatry 2008; 192: 258–63.
8 Startup H, Freeman D, Garety PA. Persecutory delusions and
catastrophic worry in psychosis: developing the understanding of
delusion distress and persistence. Behav Res Ther 2007; 45: 523–37.
9 Vorontsova N, Garety P, Freeman D. Cognitive factors maintaining
persecutory delusions in psychosis. J Abnorm Psychol 2013;
10 Morrison AP, Wells A. Relationships between worry, psychotic
experiences and emotional distress in patients with schizophrenia
spectrum diagnoses and comparisons with anxious and non-patient
groups. Behav Res Ther 2007; 45: 1593–1600.
11 Hartley S, Haddock G, Vasconcelos E, Emsley R, Barrowclough C.
An experience sampling study of worry and rumination in psychosis.
Psychol Med 2014; 44: 1605–14.
12 Foster C, Startup H, Potts L, Freeman D. A randomised controlled
trial of a worry intervention for individuals with persistent
persecutory delusions. J Behav Ther Exp Psychiatry 2010; 41: 45–51.
13 Mackie JL, ed. The cement of the universe: a study of causation.
Oxford, Oxford University Press, 1974.
14 Freeman D, Dunn G, Startup H, Kingdon D. The eﬀ ects of reducing
worry in patients with persecutory delusions: study protocol for a
randomized controlled trial. Trials 2012; 13: 223.
15 Freeman D, Garety PA. Comments on the content of persecutory
delusions: does the deﬁ nition need clariﬁ cation? Br J Clin Psychol
2000; 39: 407–14.
16 Haddock G, McCarron J, Tarrier N, Faragher FB. Scales to measure
dimensions of hallucinations and delusions: the psychotic symptom
rating scales (PSYRATS). Psychol Med 1999; 29: 879–89.
17 Startup HM, Erickson TM. The Penn State Worry Questionnaire
(PSWQ). In: GCL Davey, A Wells, eds. Worry and its psychological
disorders Chichester: Wiley, 2006: 101–20.