Clinical Psychology and Psychotherapy

Clin. Psychol. Psychother. 15, 1–14 (2008)

Published online in Wiley InterScience ([Link]). DOI: 10.1002/cpp.556

Cognitive–Behavioural,
Pharmacological and Psychosocial
Predictors of Outcome During
Tapered Discontinuation
of Benzodiazepine
Kieron O’Connor,1,* André Marchand,2 Lucie Brousseau,1
Frederick Aardema,1 Nicole Mainguy,1
Pierre Landry,2 Pierre Savard,3 Cathy Léveillé,1
Valérie Lafrance,1 Sonia Boivin,1 Denise Pitre,1
Sophie Robillard1 and Donald Bouthillier3
1
Fernand-Seguin Research Centre, Louis-H. Lafontaine Hospital, University
of Montreal
2
Louis-H. Lafontaine Hospital
3
Research Centre, Sacré-Cœur Hospital

Eighty-six participants wishing to stop benzodiazepine and who met

DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th ed.
American Psychological Association, 1994) criteria for anxiety disorder
or insomnia were assessed pre- and post-taper on clinical, pharmaco-
logical and psychosocial measures. An initial cohort of 41 participants
received treatment as usual (taper only) plus physician counselling in
the same clinic setting. A second cohort of 45 participants were ran-
domly allocated to group cognitive–behavioural therapy (CBT) plus
taper, or group support (GS) plus taper. At 3 months follow-up, the out-
comes in both the CBT and the GS subgroups were equivalent. Inten-
tion to treat analysis revealed a slight advantage to the CBT over the GS
group and the CBT group showed higher self-efficacy post-taper.
Over all 86 participants, a high-baseline level of psychological dis-
tress, anxiety and dosage predicted a poor outcome, but increase in
self-efficacy contributed to a successful outcome particularly in those
with initially poor baseline predictors. Although there was a decrease
in positive affect during preliminary stages of tapered discontinua-
tion compared to baseline, there was no significant overall increase in
negative affect. Copyright © 2008 John Wiley & Sons, Ltd.

INTRODUCTION withdrawal symptoms, where new (usually time

limited) sensations appear; and reoccurrence of
Distress from benzodiazepine (BZD) discontinu- the original problem for which the medication
ation syndrome can be related to rebound, where was originally prescribed (O’Brien, 2005). There
the initial problem (anxiety, sleep) is intensified; are wide individual differences in the experience
of these symptoms, but up to 80% of people who
discontinue, subsequently relapse apparently due
* Correspondence to: Kieron O’Connor, Ph.D., Fernand-
Seguin Research Centre, Louis-H. Lafontaine Hospital, 7331
to discontinuation syndrome (Klosko, 1990). Even
Hochelaga St., Montréal, Québec H1N 3V2, Canada. with tapering, a significant number of people expe-
E-mail: [Link]@[Link] rience distress. Ashton (2001) reports that between

Copyright © 2008 John Wiley & Sons, Ltd.

2 K. O’Connor et al.

10–15% of long-term users present withdrawal cal and clinical factors, and targeting self-efficacy,
effects even months after stopping. There are few specifically prior to discontinuation, could poten-
indicators of what might predict relapse, but per- tially boost emotional and psychological tolerance
sonality and dosage of BZD have been postulated to withdrawal.
as mediating factors (Schweizer, Rickels, DeMarti-
nis, Case, & García-España, 1998). For some time,
there has been a consensus that BZD discontinu-
ation can benefit from non-medical interventions AIM OF THE STUDY
and psychotherapy (Juergens, 1993), in particular The aim of the current study was threefold: first,
some form of social support (Fraser, Peterkin, to examine the entire sample of 86 participants
Gamsu, & Baldwin, 1990) or counselling (Nagy, with a range of psychosocial measures pre- and
1989), and especially cognitive behaviour therapy post-taper including self-efficacy and to compare
(CBT) (Otto, Bruce, & Deckersbach, 2005). the contribution of pharmacological, psychosocial
Previous work applying CBT during BZD dis- and clinical measures both at baseline and post-
continuation has tended to focus on specialized treatment to a successful outcome. Second, to
withdrawal packages implementing CBT while examine, in a randomized subset of participants,
withdrawing from BZD (e.g., for panic disorder the effect of enhancing self-efficacy in aiding suc-
(Hegel, Ravaris, Lewis, & Ahles, 1994; Otto, Pollack, cessful discontinuation of BZD, compared with a
Sachs, Reiter, Meltzer-Brody, & Rosenbaum, 1993); non-specific active comparison group receiving
insomnia (Baillargeon, Landreville, Verreault, only group support but without any direct cogni-
Beauchemin, Grégoire, & Morin, 2003); (Morin, tive–behavioural intervention. Third, to examine
Bastien, Guay, Radouco-Thomas, Leblanc, & longitudinally the evolution of mood, sleep quality
Vallières, 2004); and a mixed anxiety group (Oude and intellectual wellbeing during the taper. Dete-
Voshaar, Gorgels, Mol, van Balkom, Mulder, & rioration in intellectual and physical form, mood
Lisdonk, 2003, 2006). All of these studies reported instability and sleep disturbance are cited as prin-
superior outcome for CBT in combination with cipal effects of withdrawal from BZD (see First &
tapering, with the exception of Oude Voshaar et al. Tasman, 2004), but there has been little systematic
(2003, 2006), where CBT was less specialized with study of the evolution of these factors over the
a high dropout rate. The CBT delivery among the taper period.
above studies varied considerably regarding the
type of CBT administered, the duration of the CBT
and the number of sessions, and whether CBT was
administered in group or individual format. One MATERIALS AND METHOD
approach to standardizing CBT treatment targets
Recruitment
is to base a discontinuation programme around
an intervention addressing psychosocial factors The total sample consisted of 130 French Cana-
known to influence successful discontinuation. dian adults, aged 21–64, wishing to discontinue
O’Connor et al. (1999, 2004) reported that psycho- BZD and who suffer from non-psychotic anxiety or
social factors such as level of education, negative insomnia who were recruited via media announce-
life events, high performance inhibition, lack of ments, clinic publicity and referrals. A subset of
social support, and personality factors such as neu- 41 participants were consecutively referred to
roticism and trait anxiety predicted distress during tapering and treatment-as-usual on an individual
withdrawal, and that low baseline performance basis (individual tapering with physician counsel-
inhibition, low state anxiety and low psychologi- ling). Of the 61 people eligible for referral in this
cal distress predicted successful outcome, whilst group, 15 did not meet inclusion criteria and five
self-efficacy in coping without BZD was associated refused treatment. The other 48 participants were
with less relapse. Self-efficacy has also been shown recruited separately approximately 1 year later
to positively influence outcome in other areas of and at inclusion were randomly allocated to either
dependence (Stuart, Borland, & McMurray, 1994). a group CBT programme or non-directive group
If high confidence in discontinuing BZD and high support condition by the recruitment coordinator
confidence in dealing with performance difficulties on the basis of a random sequence generator. Sixty-
are vital factors in achieving a successful outcome, nine people were eligible for the programme, 17
then self-efficacy should better predict outcome were excluded after assessment and four refused
compared to other psychosocial, pharmacologi- treatment after assessment. Three of the remain-

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 3

Assessed for eligibility (n=61) Assessed for eligibility (n=69)

Excluded (n=15) Excluded (n=17)
Refusers (n=5) Refusers (n=4)

Consecutive referral (n=41) Randomised patients (n=48)

Received treatment-as-usual (n=41) Received group support only (n=22) Received cognitive behavior therapy
T0 - Baseline T0 – Baseline + group support (n=23)
Abandon (n=2) T0 – Baseline
Abandon (n=1)

20-week (n=31) 20-week (n=13) 20-week (n=18)

T1 succeeders (n=12) T1 succeeders (n=11) T1 succeeders (n=15)
T1- non-succeeders (n=19) T1- non-succeeders (n=2) T1- non-succeeders (n=3)
Abandon (n=10) Abandon (n=9) Abandon (n=5)

T2 3-month follow-up 3-month follow-up 3-month follow-up

T2 succeeders (n=10) T2 succeeders (n=11) T2 succeeders (n=15)
T2- non-succeeders (relapse) (n=2) T2- non-succeeders (relapse) (n=0) T2- non-succeeders (relapse) (n=0)

26-month follow-up (n=26) 11-month follow-up (n=10) 11-month follow-up (n=10)

Relapse (n=1) Relapse (n=1) Relapse (n=1)

Figure 1. CONSORT (Consolidated Standard for Reporting Clinical Trials) flow diagram for the three treatment
groups

ing 48, however, abandoned prior to the baseline order, social phobia, specific phobia, and adapta-
measures and were not replaced (see Figure 1). tional problems with anxious mood. Insomnia and
The study was conducted at two Montreal sites: anxiety are the disorders for which BZDs are most
Sacré-Coeur Hospital and Louis-H. Lafontaine commonly prescribed, and both are reported to
Hospital. The study protocol was approved by the experience equivalent withdrawal distress (Morin
hospital ethics committee and signed informed et al., 2004; O’Connor et al., 1999).
consent was obtained from all participants. Exclusion criteria were (1) taking any other
All groups were comparable in terms of recruit- psychotropic medication, (2) substance or alcohol
ment and evaluation procedures, diagnostic crite- abuse, (3) presence of any other axis I diagnosis
ria, demographics and BZD use and had received (except insomnia), (4) presence over the course
taper completing the same measures with identical of the last year of a major medical or physical
taper time lines from the same clinicians in the problem, (5) receiving any other psychotherapy for
same clinic setting within 12–15 months of each at least 3 months and (6) not intending to consult
other (O’Connor et al., 2004). Clinical and demo- for other therapy.
graphic characteristics are given in Table 1. Diagnosis was on the basis of a psychiatric inter-
Inclusion criteria were (1) aged between 21– view conducted according to DSM-IV criteria by
65 years, (2) taking BZD for at least 2 years, (3) one of the psychiatrists (P.L., N.M., P.S.) plus an
having a diagnosis of BZD dependence for at least independent semi-structured ADIS-IV (The Anxiety
2 months, (4) presenting an anxiety problem and/ Disorders Interview Schedule for DSM-IV) interview
or insomnia for at least 3 months and (5) meeting (Brown, Di Nardo, Lehman, & Campbell, 2001) for
DSM-IV criteria for one of the following: insomnia, anxiety disorders conducted by a trained health
panic with agoraphobia, generalized anxiety dis- professional. Although principally conceived as

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
4 K. O’Connor et al.

Table 1. Clinical and demographic data

Characteristics Treatment-as-usual Group support Cognitive–behavioral therapy

group (n = 41) only (n = 22) + group support (n = 23)
Sex
Male 17 13 11
Female 24 9 12
Age
Minimum 31 27 25
Maximum 64 64 61
Mean (SD) 48.22 (9.5) 48.19 (9.8) 47.35 (9.9)
Diagnostic
Anxiety disorder 33 22 23
Insomnia 8 0 0
Anxiety and insomnia 6 2 3
Half-life
Short 23 14 14
Long 17 9 9
Years of schooling
Mean (SD) 13.33 (3.9) 12.63 (2.7) 12.29 (2.6)
Dosage in diazepam equivalent units*
Mean (SD) 14.77 (12.6) 11.84 (13.2) 12.98 (10.6)
Number of years of BZD use 11.45 (9.96) 6.55 (4.99) 7.84 (8.01)
See: Bezchlibnyk-Butler, K. Z., & Jeffries, J. J. (Eds). (2002). Clinical Handbook of Psychotropic Drugs (pp. 154–158.). Toronto, ON,
Canada: Hogrefe & Huber Publishers.

a measure of anxiety disorders, the ADIS-IV also through developing a belief in capacity to function
assesses affective disorder, somatoform problems, autonomously from BZD.
substance abuse, mania, psychotic problems, and The PASS programme began with a 4-week
medical problems. Only if there was concordance period of preparation which preceded the tapered
between diagnosticians on principal Axis I diag- withdrawal schedule. The preparation period
nosis according to inclusion/exclusion criteria was involved psychoeducation and cognitive restruc-
the participant accepted into the study. turing through providing realistic information on
withdrawal and addressing personal or subcul-
tural beliefs or myths about cessation. The taper
programme was explained and perceived diffi-
CBT (PASS) Programme
culties were discussed together with individual
The PASS programme (Programme d’Aide au Succès motivations to cease BZD, withdrawal expectan-
du Sevrage; Programme Aimed at Successful Sever- cies, and the future vision of obstacles impeding
ance of benzodiazepines) attempts to boost confi- functioning. Developing a positive preparation
dence in the initial phase of withdrawal regardless involved basing anticipation on the consideration
of diagnosis. The PASS group programme and the of resources and abilities of the person by improv-
group support programme were both delivered in ing self-efficacy to cope with difficult everyday
a manualized form to be administered in group situations without BZD.
format over a 20-week period. Both manuals were As well as confronting anticipations, interpre-
divided into three sections covering (a) prepara- tations and prejudices about the risk of failure,
tion; (b) severance; (c) maintaining abstinence. The other exercises addressed the importance of the
PASS programme was built around insights gained attribution of sensations and their significance. In
from the previous studies, and the literature, on the particular, the role of extreme vigilance to unusual
psychosocial profiles likely to influence outcome. physical sensations in amplifying discomfort was
The therapy aimed to enhance self-efficacy prin- discussed, as was the tendency to misinterpret any
cipally through normalising expectancies of with- discomfort experienced during discontinuation as
drawal and attributions of withdrawal; through due to withdrawal or rebound.
boosting confidence in (a) coping without BZD, Discontinuation began at the fifth week, and
(b) coping with anxious inhibiting situations; and passed through four stages, each of 4 weeks dura-

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 5

tion: getting started; keeping going; nearly there; staying The principal difference was in the lack of any
there. As well as the weekly group meetings, par- specific directions for changing thoughts and
ticipants also attended at three weekly intervals behaviours. In this GS condition, no CBT strategies
for a consultation with a treating physician who were presented and exchanges took the form of
controlled the taper regime. open-ended discussion on themes such as ‘What is
In order to help coping with getting started, the anxiety?’ No direct action or strategy to deal with
participants had recourse to 10 resource docu- any problems was suggested. Any requests for
ments which could be discussed in the PASS group specific help were deflected back to the group. Fol-
under the direction of the facilitators. The resource lowing open discussion, participants noted the key
documents included cognitive–behavioural strate- points of discussion and also continued to reflect
gies to deal with withdrawal symptoms; normal- on the themes throughout the following week.
izing reactions and anxiety sensitivities; stress and Each week a different theme was discussed and
mood management; muscle relaxation; problem any personal request for a strategy was referred
resolution; regulating lifestyle; quality of life; sleep for group discussion.
hygiene; social support; motivation; self-efficacy.
These resources were employed on an as-needed
basis and their use in the PASS context was to Taper Protocol
deal with specific difficulties encountered during All treating physicians included in the study were
withdrawal. Each participant, with the aid of the blind as to group membership of the referred
resource document, group discussion and thera- participants. The coordinating psychiatrist (N.M.)
pist input, decided on a plan of action for the fol- assured uniformity of taper procedure. There is no
lowing week. Upon the return, the following week, standard recommended taper procedure but the
the effectiveness of the plan and its consequences procedure adopted here was a weekly reduction
were evaluated in the group. in 1/4 of the initial dose over 16 weeks, up to a
In the second step of keeping going, the prog- maximum of 113 days and depending on the half-
ress and lessons learned so far were consolidated life of the medication. Where tablets did not allow
and medium term difficulties in adaptation were precise dose reductions, either tablets were cut in
addressed, for example, social support; poor life half or spaced over days as required. Participants
quality; changing lifestyle; and reinforcing confi- were evaluated 1 month before taper (T0), between
dence in functioning in daily activities. 2 and 14 days after cessation depending on the
The third step, nearly there, focused on overcom- half-life of medication (T1), and three months after
ing thoughts or feelings likely to sabotage absti- stopping (T2). Participants consulted treating phy-
nence, again working on confidence and coping sicians at the 6th, 8th, 11th, 14th, 17th, and 20th group
skills in the group with the aid of the resource treatment session. All treating physicians (N.M.,
documents. At this stage, participants set the date P.L., P.S.) adhered to the same taper protocol
for their final severance. and, apart from answering questions pertinent
In the final stage, staying there, the participants to the medication, they engaged in no form of
dealt with the initial stages of complete severance. psychotherapy.
In particular they revised and rehearsed useful
coping strategies, envisioned how they would deal
with new difficult situations, and reviewed how Self-Monitoring
they had so far successfully adapted to and tol-
erated symptoms. Specific strategies to deal with In order to record fluctuations in well-being over
relapse were rehearsed, such as thinking ahead, the withdrawal period, all participants in both
learning from mistakes; not jumping to conclu- PASS and GS groups kept a daily diary that moni-
sions; making future decisions consistent with a tored variations in mood, intellectual and physical
non-BZD user. Successful completers received a well-being and sleep quality over the month-long
PASS-PORT at follow-up to signal their autonomy preparation period and, subsequently, during the
from BZD use. 16 weeks of severance.

The Group Support (GS) Programme Treatment Delivery and Integrity

The GS group met at the same regularity as the Participants were allocated at random to either
PASS group and followed identical taper regimes. PASS or GS condition. Two groups in each condi-

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
6 K. O’Connor et al.

tion were run and alternated between Mondays taper was labelled T1− and in the case of relapse
and Wednesdays to control for weekly period. The at three-month follow-up, T2−. In accordance with
presence of BZD and its metabolites was verified the hypotheses, three types of measures were
by urine analyses in the hospital biochemistry lab- administered and the instruments were divided
oratory (EMIT Test) on all participants at baseline into: quantitative measures of withdrawal symptoms,
(T0), post-taper (T1) and three month follow-up psychological distress, mood and well-being; and mea-
(T2). Throughout tapering, two individuals were sures of hypothesized psychological and psychosocial
picked from each group at random to test for pres- mediating factors.
ence of BZD each week. The sessions of all groups Quantitative measures of withdrawal symptoms, dis-
were videotaped to ensure fidelity of treatment tress, mood and well-being included: The Benzo-
delivery and compliance. diazepine Withdrawal Symptom Questionnaire
All therapists leading the groups received train- (Tyrer, Murphy, & Riley, 1990) a self-administered
ing on a pilot group of four people to familiarize withdrawal questionnaire of 20 items, classifying
them with the intervention protocol and the group withdrawal symptoms according to the appear-
therapy. Therapists were balanced across groups. ance of new symptoms following withdrawal.
One of the therapist participated in both GS and It has shown clinical validity and scores follow
PASS groups, whilst the other two therapists were the evolution of withdrawal and predict relapse
allocated exclusively either to the GS or the PASS (Couvee & Zitman, 2002). Psychological distress was
group. The aim here was at the same time to ensure measured by the Psychological Distress Inventory,
comparability of non-specific factors between a 29-item self-report questionnaire covering diffi-
groups and also control for treatment delivery and culties in diverse areas of health (Préville, Boyer, &
integrity in each group. All participants completed Potvin, 1992). Affect was measured by asking par-
forms rating therapist and treatment satisfaction ticipants to fill in a short form characterizing their
post-cessation and there were no significant dif- mood over the past week at each time point (T0, T1,
ference between GS and PASS groups in satisfac- T2). However, in addition, all participants in the
tion with group therapy, psychiatric consultations, GS and PASS groups kept a daily diary rating on a
documentation, impact of programme and practi- Likert-type scale (range: −2 to +2) : mood, intellec-
cal utility. tual and physical form and sleep quality over the
entire period of preparation and discontinuation.
The diary served the clinical purpose of aiding self-
Treatment as Usual
monitoring of mood before and after withdrawal
The treatment-as-usual (TAU) reference group in a systematic fashion. The Systematic Quality of
received individual taper and counselling under Life Inventory (Duquette, Dupuis, & Perrault, 1994),
medical supervision but no other form of group is an instrument developed in Quebec that mea-
or individual therapy. This TAU group had dis- sures quality of life according to the capacity of
continued through individual physician counsel- the person to succeed in 30 life domains. Three
ling over the same duration of taper at the same subscales were included here: actual state, goal state,
clinic with the same physician and completed the and the difference between actual and desired goals of
same measures as the randomized subgroups at quality of life.
approximately one year prior to the randomized Personality and psychosocial factors were measured
trial. Although not comparable in terms of ran- by: The Eysenck Personality Inventory (Eysenck &
domization, the control group constituted a com- Eysenck, 1975) which is a standard measure of
parable cross-sectional cohort in that they were introversion–extraversion and neuroticism. The
an equivalent group who experienced an identi- Life Experience Survey (Sarason, Johnson, & Siegel,
cal event (BZD taper) at a distinct period of time 1978) measures negative and positive life events as
(Glenn, 2005). well as their impact over 57 life domains. The Spiel-
berger State-Trait Anxiety (Spielberger, Gorsuch, &
Lushene, 1983) is a standardized measure of state
Questionnaires
and trait anxiety. The Specific Performance Worries–
Questionnaire measures assessing psychological Inhibition Questionnaire (Widlocher & Pull, 1988) is
and psychosocial adaptation were also adminis- a questionnaire of 10 items enumerating difficulties,
tered at baseline (T0), post-taper (T1), and 3-month discomfort and degree of inhibition in functioning
follow-up (T2). In the case of participants not suc- in everyday life. The Social Support Questionnaire
ceeding in withdrawal, the measure point post- (Sarason, Shearin, Pierce, & Sarason, 1987) has

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 7

two scales measuring the amount of social support A completer was defined as a participant who
and degree of satisfaction with social support. Self- completed the entire 20-week taper programme.
efficacy was measured (0–100) following guidelines A succeeder (or responder) was defined as a par-
from Condiotte and Lichtenstein (1981) as (a) con- ticipant who had ceased medication at 20 weeks.
fidence in coping with difficult everyday situations A relapse was defined as retaking medication (of
(averaged across situations); and (b) confidence in any dose) at 3-month follow-up. The criteria for
discontinuing BZD. success was total abstinence of BZD, and a further
follow-up was performed at approximately 7–15
months post T2 on those who had successfully
ANALYSIS tapered (see Figure 1). Abandon was defined as
ceasing to discontinue and participate in the group
Multivariate analyses (version 13 SPSS Inc.,
at any time following evaluation. Two participants
Chicago, Ill, USA) between T0 and T1 and sub-
in the GS and one in the PASS group dropped
sequently between T1 and T2 were conducted
out prior to initial interview and their data were
separately. Power and effect sizes, represented
excluded. Clinical data was available on 86 partici-
by partial etas (squared conventions: small, 0.01;
pants. Questionnaire data at T0 was available on a
medium, 0.06; large, 0.14), are given for main
maximum of 73 participants, 57 at T1, and 34 at T2
significant outcome findings. Complete data on
due to abandonments and missing data. Question-
all measures (Means and Standard Deviations)
naire data for the full sample is given in Tables 2,
for all time periods is given in Tables 2–4. Three
3 and 4 but pre-post analysis was conducted only
types of analyses were conducted: (a) Changes pre-
on completers.
post-taper across all participants and between randomized
treatment groups over T0 to T1 and, subsequently,
between post-taper and 3-month follow-up T1 RESULTS
to T2; (b) differences between succeeders and non-
Baseline
succeeders over all groups; and (c) logistic regression
of variables predicting successful outcome includ- All three treatment conditions were equivalent in
ing all participants. The primary outcome measure terms of baseline demographic, clinical, personal-
was a successful/non-successful taper at T1 which ity, and other psychosocial factors (see Tables 1, 2
was further assessed by continued success or and 3). Although the reference TAU group appear
failure (no relapse) 3 months post-taper at T2. to have a longer history of BZD use, this difference

Table 2. Psychosocial variables, mood and withdrawal effects over T0, T1, T2

Group support only (GS)

T0 (n = 17) T1 (n = 12) T2 (n = 10)
Mean (SD) Mean (SD) Mean (SD)
Neuroticism 13.00 (4.83)
Extraversion 12.94 (3.17)
Lie scale 4.18 (1.90)
Life events over the last 12 months 7.71 (3.90) 6.60 (3.41)
Impact of life event −4.88 (9.99) −3.00 (6.83)
Spielberger: state 43.06 (11.87) 43.18 (9.66) 41.90 (9.55)
Spielberger: trait 49.71 (11.01) 45.40 (9.29) 42.56 (7.75)
Self-efficacy in coping with difficult situations 51.28 (25.93) 73.12 (21.69) 74.65 (25.85)
Number of difficult situations 7.29 (2.62) 5.00 (2.83) 5.38 (2.88)
Self-efficacy in coping without benzodiazepines 60.88 (33.05) 77.08 (28.16) 83.74 (26.95)
Number of withdrawal symptoms 10.82 (6.56) 8.64 (4.12) 7.22 (3.15)
Performance inhibition scale 13.82 (7.32) 8.62 (7.33) 12.78 (6.42)
Psychological distress 58.31 (18.35) 49.50 (6.09) 54.40 (12.74)
Quality of life : Current state 6.75 (3.15) 8.35 (0.72)
Goal 9.49 (1.02) 10.38 (0.91)

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
8 K. O’Connor et al.

Table 3. Psychosocial variables, mood and withdrawal effects over T0, T1, T2

Cognitive–behavioral therapy (PASS) + group support

T0 (n = 18) T1 (n = 14) T2 (n = 11)
Mean (SD) Mean (SD) Mean (SD)
Neuroticism 12.11 (4.42)
Extraversion 13.33 (3.66)
Lie scale 3.39 (1.04)
Life events over the last 12 months 7.62 (4.21) 7.73 (3.59)
Impact of life event −3.12 (11.48) −0.27 (9.71)
Spielberger: state 43.50 (10.11) 43.29 (12.15) 35.33 (10.14)
Spielberger: trait 45.56 (10.03) 42.07 (11.74) 39.00 (11.22)
Self-efficacy in coping with difficult situations 51.27 (22.87) 78.65 (18.64) 76.69 (25.85)
Number of difficult situations 7.50 (2.53) 4.42 (3.09) 4.82 (3.09)
Self-efficacy in coping without benzodiazepines 52.38 (28.59) 88.75 (12.81) 96.23 (4.96)
Number of withdrawal symptoms 7.17 (5.78) 7.57 (4.48) 7.67 (5.32)
Performance inhibition scale 10.61 (6.02) 9.29 (7.73) 8.89 (7.03)
Psychological distress 54.06 (13.29) 52.57 (14.08) 44.44 (12.70)
Quality of life : Current state 6.13 (4.04) 8.40 (1.88)
Goal 7.79 (4.48) 10.03 (1.51)

Table 4. Psychosocial variables, mood and withdrawal effects over T0, T1, T2

Treatment-as-usual Group (TAU)

T0 (n = 38) T1 (n = 31) T2 (n = 13)
Mean (SD) Mean (SD) Mean (SD)
Neuroticism 13.45 (4.82)
Extraversion 11.93 (3.36)
Lie scale 4.28 (1.68)
Life events over the last 12 months 6.85 (4.22) 7.59 (10.84)
Impact of life event −6.12 (8.53) −6.22 (9.27)
Spielberger: state 41.24 (10.29) 43.24 (11.84) 34.13 (10.48)
Spielberger: trait 46.53 (9.59) 49.38 (9.52) 45.67 (17.21)
Self-efficacy in coping with difficult situations 48.07 (23.32) 57.33 (28.37) 72.44 (24.20)
Number of difficult situations 6.56 (2.90) 5.69 (3.19) 3.89 (1.17)
Self-efficacy in coping without benzodiazepines 59.24 (26.92) 55.93 (38.83) 77.22 (35.28)
Number of withdrawal symptoms 8.11 (5.50) 8.80 (5.95) 6.08 (4.62)
Performance inhibition scale 10.43 (6.75) 9.90 (7.76) 8.00 (4.97)
Psychological distress 56.70 (15.27) 54.71 (19.35) 45.58 (8.79)

was not significant. Insomnia as a comorbid condi- stopped compared to 83% in the PASS group
tion was over-represented in the TAU group, but a and 39% in the TAU condition. The difference
diagnosis of insomnia seems not to influence suc- between the TAU group and the PASS and GS
cessful outcome (O’Connor et al., 2004). Amongst treatments was significant (X2 [2] = 13.33; p < 0.002).
demographic variables, completers were more likely However, the outcome comparison of interest
to be unmarried (p < 0.05) than non-completers, was between the randomized groups. An inten-
but there were no other demographic differences tion to treat analysis looking at success compared
between completers and non-completers. to abandons in each group was significant (one-
tailed) (X2 [2] = 8.23, p < 0.02) favouring the PASS
over the GS group (X2 [1] = 2.74; p < 0.05). The
Outcome
PASS group had less abandons. Forty-six of the
At T2, 3 months after the programme cessation original completers at T1 (26 TAU, 10 GS, 10 PASS)
period, 85% of completers in the GS group had agreed to a further follow-up interview which took

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 9

place between 7–15 months post-T2 (for the GS and the PASS group only showed a significant decrease
PASS groups [x̄ = 11.1 (2.7)]) and between 22 and in number of difficult situations post-taper (F[2,46]
29 months for the TAU group [x̄ = 26.4 (2.4)]). The = 3.71, p < 0.03, effect size = 0.14, power = 0.65).
follow-up sample was representative of the origi- There were no significant differences for any other
nal sample. All groups showed minimal relapse questionnaire measures.
(one per group) amongst those completers who
had discontinued at T2. There were no differences Analysis of Daily Self-Monitoring Diaries (GS and
in satisfaction ratings post-therapy between the GS PASS groups only)
and PASS groups. The daily diaries of the randomized groups (20
participants in the GS group and 21 in the PASS
group) were analysed by repeated measures
ANOVA. The fluctuations in mood and physical,
Changes Post-Taper and Between-treatment
intellectual and sleep quality were averaged over
groups (T0, T1, T2)
five different time points before, during and after
Withdrawal Symptoms withdrawal: time period 1: 1st to 5th week, prepara-
Over all participants, there was no decrease tion period; time period 2: 6th to 10th week, getting
across time but those who completed to T2 showed started; time period 3: 11th to 15th week, keep going;
a significant decrease in reported withdrawal time period 4: 16th to 20th week, nearly there; time
symptoms (F[2,23] = 3.85, p < 0.04, effect size = period 5: 21st to 25th week, end of programme:
0.25, observed power = 0.64). There were no dif- maintaining gains.
ferences between any groups in the number of Repeated analysis of variance revealed no dif-
withdrawal symptoms reported either at baseline ferences in fluctuations between those who suc-
or at T1 or T2. ceeded or not, for any time period or for any of
the four measures (see Figure 2). There were no
Quality of Life group differences in any of the changes over the
Both GS and PASS groups showed a significant four time periods for any of the four diary mea-
increase in state of quality of life between T0 and sures between the GS and PASS groups, so the
T2 (F[1,19] = 12.30, p < 0.002, effect size = 0.39, groups were combined for the analysis over dis-
power = 0.91) but there were no group-by-time tinct time points. Repeated measures ANOVA
interaction effects. The same improvement was revealed differences for each of the measures over
found in both groups for the goal subscale (F[1,19] the five time periods: mood (F[4,38] = 5.10, p <
= 8.34, p < 0.009, effect size = 0.31, power = 0.78), 0.05); physical well-being (F[4,41] = 3.03, p < 0.05);
and the gap between current and ideal quality of intellectual well-being (F[4,39] = 2.85, p < 0.05);
life subscale (F[1,19] = 4.82, p < 0.04, effect size = sleep quality (F[4,39] = 3.70, p < 0.05). Planned post
0.20, power = 0.55). hoc t-test comparisons between means of the differ-
ent time periods revealed the following significant
Performance Inhibition. differences: decrease in reported positive mood:
There was a significant improvement in perfor- between time periods 2 and 3 (p < 0.02), between
mance inhibition over all groups over time between time periods 2 and 4 (p < 0.05), between time period
T0 to T1 (F[1,56] = 6.07; p < 0.02, effect size = 0.10, 2 and 5 (p < 0.01); decrease in reported positive
power = 0.68). There was no difference between the physical well-being: between time period 2 and 5
PASS and GS treatments. (p < 0.04); decrease in reported positive intellectual
well-being: between time period 2 and 5 (p < = 0.04);
Self-Efficacy in Coping and Self-Efficacy and decrease in reported sleep quality: between time
in Discontinuation period 2 and 5 (p < 0.05).
There was a significant increase at T1 in self-
efficacy about coping with difficult situations for
all groups (F[1,46] = 33.30, p < 0.001, effect size =
Differences Between Succeeders and
0.45, power = 1.00). There was a similar overall
Non-Succeeders at T1
increase for self-efficacy in achieving or maintain-
ing BZD discontinuation (F[2,44] = 8.86, p < 0.005, Comparisons on all measures were made between
effect size = 0.17, power = 0.83) but only the PASS succeeders and non-succeeders over all groups.
group showed significant improvement (F[2,44] = There were no differences in success between
4.72, p < 0.01, effect size = 0.18, power = 0.76). Also diagnostic groups (panic, generalized anxiety, and

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
10 K. O’Connor et al.

1,5
Positive scale
(0 - 2)

Mood
1
Physical form
Intellectual form
Sleep quality
0,5

0
1 2 3 4 5
Time periods

Figure 2. Variations in daily diary mean ratings on a continuous scale (+2 to −2) of mood, physical form, intel-
lectual form and sleep quality over the five time periods of discontinuation (weeks: 1–5, 6–10, 11–15, 16–20, 21–25)
for the group support and PASS (Programme d’Aide au Succès du Sevrage; Program Aimed at Successful Severance of
benzodiazepines) therapy groups
Note. All Mean ratings stayed in the positive range (0 to +2).

90 (N=24) Logistic Regression of T0 and Change

(N=24)
80 Measures on T1
70
Logistic regression on outcome revealed baseline
60 (N=24)
dosage, duration of medication, state anxiety, psy-
50 Success chological distress, and performance inhibition
(N=23)
40 (N=23) Failure as significant predictors. Baseline self-efficacy was
30 not a predictor of outcome. However, severity of
20 baseline psychological distress, dosage and dura-
10
tion of medication (β = 1.09; r2 = 0.38) emerged as
the three main predictors of outcome with a clas-
0
sification rate at 79.2% for succeeders and 73.9% for
T0 T1 OR T1` T2 or T2`
non-succeeders. The duration of medication added
Figure 3. Self-efficacy in coping with difficult situations minimally to the variance and dosage and distress
for succeeders and non-succeeders showing significant were the principal predictors accounting for an
differences at T1 (p < 0.001) overall 76.6% prediction rate without duration. In
a logistic regression of change scores, between T0
to T1, on outcome, self-efficacy in coping with dif-
insomnia). Those reporting successful discontinua- ficulties and self-efficacy in coping without BZDs
tion reported less withdrawal symptoms at T1 than emerged as the two variables which successfully
those who failed to taper (F[1,50] = 10.02; p < 0.003). predicted outcome (β = 0.96; r2 = 0.64) account-
This interaction effect was more pronounced at T1, ing for 89.5% correct classification of succeeders
but even for those relapsing at T2 there is a trend and 75.0% correct classification of non-succeeders
towards greater withdrawal severity (t[28] = −1.55; for an overall classification rate of 83.9%. Change
p < 0.08). Succeeders reported less psychological in self-efficacy in coping with difficult situations
distress (t[50] = −2.49, p < 0.02), less state anxiety alone accounted for 52% of the variance (β = 0.95;
(t[50] = −2.64, p < 0.01) and trait anxiety (t[50] = r2 = 0.52) and predicted 84.2% of succeeders and
−5.07, p < 0.001), less performance inhibition (t[51] 66.7% of non-succeeders for an overall predic-
= −4.36, p < 0.001), more social support (t[50] = tion rate of 77.4%. Since baseline distress, dosage
2.22, p < 0.03), and reported higher self-efficacy in and duration were independent of change in self-
coping without BZD (t[46] = 3.77, p < 0.001) and efficacy, mediation analysis was not possible since
higher efficacy in coping with difficult situations correlation amongst variables is a requirement
(t[46] = 4.98, p < 0.001) (see Figure 3). (Baron & Kenny, 1986), but partial correlations

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 11

(Nunnally & Bernstein, 1994) of the three baseline therapist contact, even time of day of group ses-
predictors controlling for change in self-efficacy in sion. Also the criteria for discontinuation was
coping with difficult situations rendered the corre- strictly abstinence. Amongst completers, there was
lations with outcome non-significant, respectively no difference in satisfaction ratings between GS
for baseline dosage (p < 0.22), duration (p < 0.10) and PASS groups; however, the GS group had a
and distress (p < 0.09). higher rate of abandon. There were also differences
between the GS and PASS groups on both self-
efficacy measures to the advantage of the PASS
Separate Analysis of Self-Efficacy in group. Hence, we may conclude that change in con-
Succeeders Showing High Distress, High fidence in coping with difficult situations played
Dosage and High Severity of Withdrawal a more central role in the PASS, than in the other
at Baseline (According to Median Split) groups, where it related to successful outcome.
These differences apart, it must be concluded that
There were no differences in baseline self-efficacy
the ingredients of group support, regular meetings
between succeeders high or low in distress, dosage
and other non-specific psychosocial factors seemed
and withdrawal severity. Change in self-efficacy of
as effective as specific CBT strategies in successful
coping with difficult situations was significantly
taper. Such group social support can also be offered
greater in those who succeeded whilst reporting
successfully in non-specialist settings.
severe withdrawal symptoms (t[18] = 3.37, p <
A limitation of the group comparisons was that
0.003) as compared to those who succeeded but
questionnaire measures were not available on all
reported few or no withdrawal symptoms (t[23] =
participants at follow-up. Also, the two random-
2.24, p < 0.04). In the high-baseline anxiety group,
ized groups were relatively small and so some of
change in self-efficacy in coping with difficult situ-
the comparisons may have been underpowered.
ations was significantly greater (t[19] = 4.42, p <
The current sample did not cover the complete
0.001) between those who succeeded and those
range of anxiety disorders and were without sig-
who did not. In the low-baseline anxiety group,
nificant comorbidity, which might limit generaliz-
there were no differences (t[23] = 1.52, p < 0.14)
ing the effectiveness of the programme to a clinic
in changes in self-efficacy between succeeders and
population. Also, the results may not generalize to
non-succeeders. Likewise, those who succeeded
a population suffering only from insomnia.
with initial high dosage and high psychological
A further finding of the current study is that suc-
distress also showed a more significant increase in
ceeders who taper by different means (GS, PASS,
self-efficacy in coping with difficult situations than
TAU) show positive changes in psychological dis-
those with low baseline scores (high dosage, t[15] =
tress, anxiety, performance inhibition and increased
2.74, p < 0.01; high distress, t[16] = 3.39, p < 0.004;
self-efficacy both in coping without BZD and in
low dosage, t[25] = 2.38, p < 0.03; low distress, t[25]
coping with difficult situations. Furthermore, logis-
= 1.91, p < 0.07).
tic regression revealed that although, as has been
reported in other studies, baseline distress, dosage
and duration of medication predicted successful
DISCUSSION
outcome, change in self-efficacy in coping without
The results from the comparison of the random- BZD and in coping with difficulties also predicted
ized groups suggest that group support can be as outcome and when baseline predictors were con-
effective as a more targeted CBT in helping attain trolled for change in self-efficacy, through partial
and maintain BZD discontinuation. The PASS correlations, their baseline relationship to outcome
targeted psychological and psychosocial variables became non-significant. The crucial contribution
previously shown to be associated with successful of increased self-efficacy to successful outcome
outcome, regardless of primary presenting diagno- was further underlined in the significantly greater
sis. The outcome post-taper of >80% success com- change in self-efficacy in those succeeders whose
pares favourably with previous studies of CBT. baseline distress and dosage were originally likely
Although other studies have shown a superior- to predict unsuccessful outcome.
ity for CBT over comparison conditions, generally In support of the external validity of the results,
these conditions have been treatment-as-usual. findings from previous studies were largely repli-
The present study controlled rigorously for all cated. The same psychosocial correlates of outcome
non-specific factors which might contribute to emerged as in previous publications (O’Connor
out-come: group format, manualized delivery, et al., 1999, 2004), in particular baseline anxiety

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
12 K. O’Connor et al.

Baseline predictors Contributing variable Outcome

Low dosage

Successful outcome

Low psychological
distress Increase in self-
efficacy in coping with
difficult situations

High dosage

Unsuccessful outcome

High psychological
distress

Figure 4. Schema of baseline predictors and contributing variable predicting successful outcome

level, baseline performance inhibition, psychologi- tolerance to withdrawal may play a role in influ-
cal distress and initial dosage. Also, as reported encing outcome, particularly when baseline pre-
in the previous studies, dosage was unrelated to dictors (anxiety, dosage, distress) are not optimal,
psychosocial factors, suggesting independent but as shown in Figure 4, and it suggests the utility
interacting biological and psychosocial influences. of addressing self-efficacy in this context. Indeed,
Again, such baseline factors proved more power- change in self-efficacy was a significant predictor
ful predictors of outcome than the diagnostic clas- of outcome in the high-baseline anxiety groups but
sification. Typically, three pharmacological factors not in the low baseline anxiety groups. Perhaps
are thought to influence successful withdrawal, future work should address development of psy-
namely dosage, duration and half-life (O’Brien, chological tolerance as a facilitating factor in over-
2005), with these three factors interacting together. coming immediate problems of discontinuation
In our study, dosage and duration proved predic- regardless of diagnosis.
tors of outcome, but the association with outcome Another important contribution of the current
was considerably reduced when change in self- study concerns mood change during discontinua-
efficacy was partialled in. Change in self-efficacy tion. The transversal measure of affect rated at times
also proved an important contributing variable to T0, T1, T2 showed no change over time. In addi-
the relationship between baseline psychological tion, change in affect, apart from being minimal
distress and anxiety as well as withdrawal sever- over time across groups, did not predict outcome.
ity and outcome. The relative contributions were But when assessed longitudinally in the form of a
assessed by partial correlation since requirements daily diary, there were significant changes in the
for mediation analysis were not met. Hence in our daily diary for the GS and PASS groups—monitor-
findings, self-efficacy represents a contributing ing mood, intellectual form and sleep quality—but
rather than mediating or moderating influence. the change showed a decrease in positive rather
Although successful outcome was associated than an increase in negative ratings (see Figure 3)
with less withdrawal symptoms, success seems and did not differ between succeeders and non-
not to depend exclusively on low reported with- succeeders. The decrease in positive ratings was
drawal symptoms. Whilst degree of physical most pronounced at the beginning of withdrawal
dependence is manifested by a drug class-specific as compared to other time periods. The results
withdrawal syndrome, psychological dependence revealed similar variations over time in all the five
may be gauged more by perceived ability to cope measures of the daily diary during the withdrawal
in problem-related life domains. Psychological period. The present results were obtained using a

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
CBT and Benzodiazepine Withdrawal 13

very slow tapered withdrawal, whereas a more Brown, T.A., Di Nardo, P.A., Lehman, C.L., & Campbell,
rapid withdrawal might produce more deleterious L.A. (2001). Reliability of DSM-IV anxiety and mood
changes in mood. disorders: Implications for the classification of emo-
tional disorders. Journal of Abnormal Psychology, 110,
In conclusion, baseline levels of psychologi- 49–58.
cal distress prior to taper predicted outcome as Condiotte, M.M., & Lichtenstein, E. (1981). Self-efficacy
well as baseline dosage, and these factors were and relapse in smoking cessation programs. Journal of
independent. However, increased self-efficacy in Consulting and Clinical Psychology, 49, 648–658.
coping with difficult everyday situations, spe- Couvee, J., & Zitman, F. (2002). Psychometric evalua-
cifically targeted in the CBT (PASS) group, was tion during a discontinuation program in depressed
chronic benzodiazepine users in general practice.
associated with successful outcome, particularly Addiction, 97, 337–345.
in cases where baseline levels of distress, anxiety Duquette, R.L., Dupuis, G., & Perrault, J. (1994). A new
and dosage predicted unsuccessful outcome. So approach for quality of life assessment in cardiac
increased self-efficacy in coping without BZD, patients: Rationale and validation of the quality of life
but in particular increased self-efficacy in coping systemic inventory. Canadian Journal of Cardiology, 19,
with difficult situations, contributes to successful 106–112.
Eysenck, H., & Eysenck, S. (1975). Manual of the Eysenck
outcome when tapering BZD. The role of self-effi-
personality questionnaire. London: Hodder and
cacy may also have implications for the combined Stoughton.
use of BZD and CBT. Additive effects of therapy First, M.B., & Tasman, A. (2004). Substance-related dis-
and medication can be beneficial when treating orders: General approaches to substance and polysub-
acute anxiety, but subsequent discontinuation of stance use disorders/other substance use disorders. In
BZD can result in loss of efficacy, due to the impor- M.B. First, & A. Tasman (Eds.), DSM-IV-TRTM Mental
tance of attributing successful coping to personal Disorders. Diagnosis, etiology and treatment (pp. 349–374).
Chichester, UK: John Wiley & Sons, Ltd.
resources rather than medication (Otto et al., 2005). Fraser, D., Peterkin, G., Gamsu, C., & Baldwin, P. (1990).
Increasing confidence in coping with difficulties Benzodiazepine withdrawal: A pilot comparison of
prior to discontinuation may be useful to build up three methods. British Journal of Clinical Psychology, 29,
emotional and psychological tolerance to anxiety 231–233.
and withdrawal distress during discontinuation. Glenn, N.D. (2005). Cohort analysis. Thousand Oaks, CA:
Sage Publications, Inc.
Hegel, M.T., Ravaris, C.L., Lewis, C., & Ahles, T.A.
ACKNOWLEDGEMENTS (1994). Combined cognitive–behavioral and time-
limited alprazolam treatment of panic disorder.
The authors acknowledge the Fonds de la Recher- Behavior Therapy, 25, 183–195.
che en Santé du Québec and the Conseil Québécois Juergens, S. (1993). Benzodiazepines and addiction.
de la Recherche Sociale for financial support for Recent Advances in Addictive Disorders, 16, 75–86.
this project (grant 961227). Klosko, J.S. (1990). A comparison of alprazolam and
behavior therapy in treatment of panic disorder.
The PASS manual is available from the first Journal of Consulting and Clinical Psycholog, 58, 77–
author on request. 84.
Morin, C.M., Bastien, C., Guay, B., Radouco-Thomas, M.,
REFERENCES Leblanc, J., & Vallières, A. (2004). Randomized clinical
trial of supervised tapering and cognitive behaviour
American Psychiatric Association. (1994). Diagnostic therapy to facilitate benzodiazepine discontinuation in
and statistical manual of mental disorders (4th ed.). older adults with chronic insomnia. American Journal of
Washington, DC. Psychiatry, 161, 332–342.
Ashton, C.H. (2001). Benzodiazepines: How they work and Nagy, L.M. (1989). Clinical and medication outcome after
How to withdraw. Newcastle Upon Type, UK: News- short-term alprazolam and behavioral group treat-
castle University. ment in panic disorder. Archives of General Psychiatry,
Baillargeon, L., Landreville, P., Verreault, R., Beauche- 46, 993–999.
min, J.P., Grégoire, J.P., & Morin, C.M. (2003). Discon- Nunnally, J.C., & Bernstein, I.H. (1994). Psychometric
tinuation of benzodiazepines among older insomniac theory. New York: McGraw-Hill, Inc.
adults treated with cognitive–behavioural therapy O’Brien, C.P. (2005). Benzodiazepine use, abuse, and
combined with gradual tapering: a randomized trial. dependence. Journal of Clinical Psychiatry, 66, 28–
Canadian Medical Association Journal, 169, 1015–1020. 33.
Baron, R.M., & Kenny, D.A. (1986). The moderator- O’Connor, K.P., Bélanger, L., Marchand, A., Dupuis, G.,
mediator variable distinction in social psychological Elie, R., & Boyer, R. (1999). Psychological distress and
research: Conceptual, strategic and statistical consid- adaptational problems experienced during discon-
erations. Journal of Personality and Social Psychology, 51, tinuation of benzodiazepines. Addictive Behaviors, 24,
1173–1182. 537–541.

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp
14 K. O’Connor et al.

O’Connor, K.P., Marchand, A., Bélanger, L., Mainguy, Préville, M., Boyer, R., & Potvin, L. (1992). La démoralisa-
N., Landry, P., Savard, P. et al. (2004). Psychological tion: détermination de la validité et de la fiabilité de la
distress and adaptational problems associated with mesure utilisée dans l’Enquête Santé Québec. Cahier de
benzodiazepine withdrawal and outcome. A replica- Santé Québec, 7, 0–0.
tion. Addictive Behaviors, 29, 583–593. Sarason, B.R., Shearin, E.N., Pierce, G.R., & Sarason, I.G.
Otto, M.W., Bruce, S.E., & Deckersbach, T. (2005). Ben- (1987). Social support. Journal of Personality and Social
zodiazepine use, cognitive impairment, and cogni- Psychology, 52, 813–832.
tive–behavioral therapy for anxiety disorders: Issues Sarason, I.G., Johnson, J.H., & Siegel, J.M. (1978). Assess-
in the treatment of a patient in need. Journal of Clinical ing the impact of life changes: Development of the life
Psychiatry, 66, 34–38. experiences survey. Journal of Consulting and Clinical
Otto, M.W., Pollack, M.H., Sachs, G.S., Reiter, S.R., Meltzer- Psychology, 46, 932–946.
Brody, S., & Rosenbaum, F. (1993). Discontinuation Schweizer, E., Rickels, K., DeMartinis, G., Case, G., &
of benzodiazepine treatment: Efficacy of cogni- García-España, F. (1998). The effect of personality on
tive–behavioral therapy for patients with panic withdrawal severity and taper outcome in benzodiaz-
disorder. American Journal of Psychiatry, 150, 1485– epine dependent patients. Psychological Medicine, 28,
1490. 713–720.
Oude Voshaar, R.C., Gorgels, W.J.M.J., Mol, A.J.J., van Spielberger, C.D., Gorsuch, R., & Lushene, R. (1983).
Balkom, A.J.L.M., Mulder, J., Van De Lisdonk, E.H. Manual for the state-trait anxiety inventory. Palo Alto,
et al. (2006). Long-term outcome of two forms of CA: Consulting Psychology Press.
randomised benzodiazepine discontinuation. British Stuart, K., Borland, R., & McMurray, N. (1994). Self-
Journal of Psychiatry, 188, 188–189. efficacy, health locus of control, and smoking cessa-
Oude Voshaar, R.C., Gorgels, W.J.M.J., Mol, A.J.J., van tion. Addictive Behaviors, 19, 1–12.
Balkom, A.J.L.M., Van De Lisdonk, E.H., Breteler, Tyrer, P., Murphy, S., & Riley, P. (1990). The benzodiaz-
M.H.M. et al. (2003). Tapering off long-term ben- epine withdrawal symptom questionnaire. Journal of
zodiazepine use with or without group cognitive– Affective Disorders, 19, 53–61.
behavioural therapy: Three-conditioned, randomised Widlocher, D., & Pull, C.B. (1988). The W.P. scale, a
controlled trial. British Journal of Psychiatry, 182, 498– rating scale for the assessment of inhibition in anxiety
504. states. Encéphale, 14, 139–146.

Copyright © 2008 John Wiley & Sons, Ltd. Clin. Psychol. Psychother. 15, 1–14 (2008)
DOI: 10.1002/cpp