Behaviour Research and Therapy 41 (2003) 633646

www.elsevier.com/locate/brat

Cognitive therapy versus applied relaxation as treatment of

generalized anxiety disorder
A. Arntz
Department of Medical, Clinical and Experimental Psychology, University of Maastricht, P.O. Box 616,
NL-6200 MD Maastricht, The Netherlands
Accepted 14 March 2002

Abstract

Cognitive therapy (CT) and applied relaxation (AR) as treatments of generalized anxiety disorder (GAD)
were compared in a sample of 45 patients of a community mental health center, randomly allocated to
condition. Patients were assessed before and after a 12-session treatment, and at one and six months follow-
ups. There was a 20% drop out from CT and 15% from AR (NS), with some drop outs being considerably
improved. Both completers and intention-to-treat analyses revealed that both treatments were effective (ESs
of composite and specific measures ranging from 0.53 to 1.14). At one-month follow-up AR tended to do
better than CT, with CT catching up with AR at six months. Recovery rates and proportions of patients
showing reliable change were comparable to other studies on AR and CT, with 55% of CT and 53.3% of
AR patients recovered on the STAI-trait at six-month follow-up. These results confirm that both CT and
AR are effective treatments for GAD, and also that there is still room for improvement.
2003 Elsevier Science Ltd. All rights reserved.

1. Introduction

Although a very common and chronic problem in the general population (lifetime prevalence
4.16.6%, Blazer, Hughes, George, Swartz, & Boyer, 1991; Kessler et al., 1994), generalized
anxiety disorder (GAD) is not so frequently seen in mental health centers (e.g. approx. 4.3% of
the patients from our center have GAD as a first diagnosis). Despite a relatively low prevalence
of GAD at our center, we decided thirteen years ago to start a clinical trial, comparing what then
seemed to be two of the most promising treatments for GAD, applied relaxation (AR) and cogni-

Tel.: +31-43-388-1225; fax: +31-43-388-4155.
E-mail address: arnoud.arntz@mp.unimaas.nl (A. Arntz).

0005-7967/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0005-7967(02)00045-1
634 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

tive therapy (CT). In the meantime, much more has become clear about the effects of various
treatments for GAD.
Several studies have compared CT and cognitive-behavior therapy (CBT), based on the work
of Beck and Emery (1985) with behavior therapy (BT) approaches (such as anxiety management
training). In general, C(B)T was more effective than BT, especially in the long run (Butler, Fen-
nell, Robson, & Gelder, 1991; Durham & Turvey, 1987; Durham, Murphy, Allan, Richard, Treliv-
ing, & Fenton, 1994). CBT has also been compared to benzodiazepines, and proved to be more
effective both in the short and in the long term (Power, Jerrom, Simpson, Mitchell, & Swanson,
1989; Power, Simpson, Swanson, & Wallace, 1990). A recently developed CBT package specifi-
cally focussing at worrying was only compared to a waitlist condition (Ladouceur et al., 2000).
AR, a treatment approach developed by several workers (Bernstein & Borkovec, 1973; O st, 1987)
also fared well in various clinical trials (Barlow, Rapee, & Brown, 1992; Borkovec & Costello,
1993), and has often been part of CBT programs (e.g. Borkovec & Costello, 1993; Power et al.,
1990). According to DeRubeis and Crits-Cristoph (1998), only C(B)T has proved to be effective
and specific, whereas AR has been proved to be effective. Only three studies directly compared
these two approaches. One found no statistically significant differences, but had low statistical
power (Barlow et al., 1992). The second found no significant differences immediately after treat-
ment, but at 12-months follow-up CT seemed superior to AR (Borkovec & Costello, 1993). The
third found the two approaches equally effective (O st & Breitholtz, 2000).
In a recent overview of six controlled trials, Fisher and Durham (1999), using the STAI-trait
as common measure, compared seven psychological treatments of GAD. Individual AR and indi-
vidual C(B)T were found to be the most effective, and the most consistent in their positive effects,
with about 63% recovered immediately after AR, and 48% after C(B)T. At a half year follow-
up, 60% of the AR, and 51% of the C(B)T patients were recovered. Non-directive therapy, group
C(B)T and group BT were moderately effective (3138% recovery at half year follow-up), and
individual BT (11%) and analytical psychotherapy the least (4% recovery). Thus, the present
study is helpful as an independent test of whether the good results of AR and C(B)T achieved
in other centers can be replicated, and as a much needed direct comparison of CT and AR.
A threat to external validity is the recruitment of subjects through media releases, and other
unnatural ways, as has been done in the typical GAD treatment study (e.g. recently by Ladouceur
et al., 2000; O st & Breitholtz, 2000). Two of the three previous studies directly comparing AR
and CT for GAD used advertisements to recruit subjects (Borkovec & Costello, 1993; O st &
Breitholtz, 2000). By contrast, the GAD patients in the present study were not especially recruited
for the study, but came from the regular population referred to a community mental health center.
A second threat to external validity is the use of stringent exclusion criteria. In the present study
most comorbid diagnoses were allowed, and many patients indeed had comorbid diagnoses. A
third threat to validity is the investigation of a treatment developed by the researchers. The finding
that researchers-alliance is related to stronger effects of the pertinent treatment, also called the
allegiance-effect, is not uncommon (McNally, 1996). Independent replication is the best safe-
guard against this bias (in general, independent replication is the most essential, but sometimes
undervalued, verification tool in science). A relevant example are three studies each comparing
Clark and Salkovskis CT for panic disorder with O sts AR package in the treatment of panic
disorder. Clark and coworkers found CT to be superior to AR (Clark et al., 1994), whereas O sts

group found AR and CT equally effective (Ost & Westling, 1995). An independent replication
 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646 635

of our group demonstrated CT to be more effective than AR in the treatment of panic disorder
(Arntz & van den Hout, 1996; see also McNally, 1996). O sts AR as treatment for GAD has only

been investigated by its developer and a coworker (Ost & Breitholtz, 2000). Thus, an independent
replication is needed to see whether their finding is equally effective as CT holds.
In sum, the present study aimed at comparing two psychological treatments for GAD, AR and
CT, with respect to their immediate and long-term effects, in a sample of GAD patients representa-
tive of the clinical population. The study is comparable to the study by O st and Breitholtz, 2000,
who compared the same treatment approaches. Although the comparison is not new, the present
study is valuable as an independent test, not biased by possible allegiance-effects, and because
the investigation was done in a clinical sample of non-recruited patients with high rates of comor-
bidity. Anxiety complaints, trait anxiety (notably heightened in these patients, cf. Fisher & Dur-
ham, 1999), and a general psychopathology index (a composite score of a large number of self-
report instruments) were the main outcome measures. Fisher and Durham (1999) have recently
published an analysis of recovery rates and clinically significant change in GAD across different
outcome studies. Their norms enable the results of the present study to be compared with those
of other studies.

2. Method

2.1. Patients

Patients of the Community Mental Health Center at Maastricht who were referred to the Aca-
demic Section Behavior Therapy participated if they met the following criteria: (1) primary diag-
nosis of GAD; (2) requesting treatment for their GAD; (3) older than 17 and younger than 70
years; (4) no depressive disorder preceding the current episode of GAD or requiring immediate
treatment; (5) no behavior therapy received for their GAD; (6) no evidence of organic mental
disorders accounting for the complaints, mental retardation, psychotic disorders, alcohol or drug
dependence. All patients gave informed consent.
The first 14 patients of the present study were screened with a self-developed interview, which
screened for presence of GAD symptoms as defined by the DSM-III-R. This interview also
instructed the screener to assess mood disorders, psychotic disorders, and possible organic factors
(notably hyperthyroidism, caffeine consumption). In 1992, the Dutch version of the Structured
Clinical Interview for DSM-III-R axis-I (SCID: Spitzer & Williams, 1986) became available, and
27 patients of the present study were screened with this instrument. The last four patients of the
present sample were screened with the Dutch DSM-IV version of the SCID. Of the 47 patients
with a first diagnosis of GAD on axis-I in the inclusion period of 10.5 years, two did not enter
the trial because they received a primary diagnosis on axis-II, and they were accordingly treated
for a personality disorder. The final sample consisted of 15 men and 30 women. Mean age was
35.9 years (range 2060), 71.1% were married or living with a partner, 38% had a paid job (others
being ill, disabled, unemployed, housewife/househusband, student, etc.). Mean duration of GAD
was 8.8 years (md = 5 years, range 135 years). Educational level was, in general, rather low:
44.4% had lower level education, 40% medium level, and 15.6% some form of higher education.
On Axis-I 22.2% had only a diagnosis of GAD, the others having secondary diagnoses (46.7%
636 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

one, 13.3% two, 13.3% three, 2.2% four, and 2.2% five secondary diagnoses). The mean number
of diagnoses (2.33) did not differ significantly between conditions (MannWhitney U =
220.5; Z = 0.717, p = 0.48). Anxiety diagnoses were most common as secondary diagnoses, 39
were diagnosed (17 panic disorders, 13 social phobias, 4 specific phobias, 3 obsessive-compulsive
disorders, 2 PTSD). There were 16 mood disorders (11 major depressions, 5 dysthymic), 3 somato-
form disorders, 3 eating disorders, 2 V-codes, 1 sexual disorder, and one adjustment disorder.

2.2. Treatments

Each treatment consisted of one pre-treatment session during which therapist and patient made
acquaintance, self monitoring of daily anxiety levels was introduced, and arrangements were made
about sessions and measurements. Twelve weekly sessions of one hour each followed. After one
and six months, follow-up meetings were arranged.

2.2.1. Cognitive Therapy

CT was based on the approaches of Beck and Emery (1985) and Butler et al. (1991). During
the first session the cognitive model was introduced by formulating the patients problem in cogni-
tive terms and sketching an idiosyncratic vicious circle involving catastrophic thoughts and anxi-
ety. Challenging dysfunctional thoughts and formulating rational thoughts was introduced during
session two. From the first session onwards patients monitored automatic thoughts
preceding/accompanying their anxiety. From the second session onwards, patients tried to chal-
lenge their automatic thoughts and to formulate rational thoughts in their diaries. From the third
session onwards, behavioral experiments during and between the sessions were introduced to test
catastrophic beliefs. Underlying assumptions were identified and tested from session 6 onwards.
Neither exposure in vivo nor relaxation training was part of CT. Behavioral experiments were
only used to test specific beliefs, not to habituate to feared situations.

2.2.2. Applied relaxation

AR followed the protocol developed by O st (1987). In session one an idiosyncratic vicious
circle stressing the role of bodily arousal in the maintenance of the problem was offered as an
explanation for the disorder. No reference was made to the possible role of automatic thoughts
and beliefs. Patients were told that it was possible to learn to reduce the autonomic arousal by
relaxation procedures, and that they would be helped to practice these procedures in gradually
more difficult situations. An overview of the 12-session program was given, showing patients
how they would learn to apply relaxation skills in still shorter times, and that they would learn
to apply them in difficult situations. During treatment patients were taught to identify early signs
of anxiety and to apply relaxation skills. Starting with progressive relaxation, the therapist taught
patients to relax increasingly quickly, to apply relaxation skills during daily activities such as
walking, typing, driving a car, etc., and to use these skills to counter anxiety as early as possible.
Patients were instructed to practice at least twice a day. During the last two weeks of treatment
graduated exposure to feared situations was introduced so that patients could practice their AR
skills in these situations.
 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646 637

2.3. Therapists

Thirteen therapists conducted both treatments. There were weekly meetings, during which each
case was presented in detail and adherence to the protocol was checked. Care was taken to exclude
cognitive techniques from the AR treatment, and exposure and relaxation techniques from the CT
treatment. AR was originally taught to our group in a workshop at our center by Lars-Go ran O st.
Three of the 13 therapists participated in this workshop. The others were trained by the present
author. CT was introduced by Gillian Butler to the present author, who trained the other therapists.

2.4. Measures.

On a diary, subjects indicated their average level of anxiety during every 3 h (while not asleep)
by filling in a figure between 0 and 100. These diaries were used in treatment to identify the
most anxious periods. (Originally, it was planned that the diaries should be used as a dependent
measure, but unfortunately, there were too many missing). A Dutch version of Spielbergers State
Trait Anxiety Inventory was used as a measure of state and trait anxiety (alphas 0.91; van
der Ploeg, Defares, & Spielberger, 1980). The SCL-90 was used as a multi-dimensional index of
psychopathological complaints (alphas of subscales 0.72; alpha of anxiety scale = 0.88; alpha
total score 0.95; Arrindell & Ettema, 1981). The Fear of Fear questionnaire was used to
measure fear of anxiety symptoms and has good validity (van den Hout, van der Molen, Griez, &
Lousberg, 1987). Of the Fear Questionnaire (Marks & Mathews, 1979) fear and avoidance scores
of the self-described main complaint, and of agoraphobic and social-phobic subscales were ana-
lyzed. The Bouman Depression Inventory (Bouman, 1987) is a Dutch parallel test of the Beck
Depression Inventory, with good psychometric properties (e.g. alphas = 0.92). There are three
subscales: mood, motivation and somatic complaints (mainly sleep problems).

2.5. Procedure and design.

At the end of the screening procedure the patient was informed about the study. Patients started
treatment after a wait of 36 months. They were randomly allocated to condition and to the next
available therapist. Questionnaires were administered one week before treatment, after the 12th
session (posttest), 4 weeks after the posttest (follow-up 1) and 6 months after the posttest (follow-
up 2). No additional treatment was given between posttest and first follow-up. For ethical reasons,
patients could receive further treatment after follow-up 1, either for remaining GAD problems,
or for other problems.

2.6. Data reduction and analysis.

Before analyzing treatment effects, questionnaire subscale scores were subjected to Principal
Component Analysis followed by Varimax rotation in order to reduce the number of dependent
variables, and to increase statistical power. Scree plots were inspected. It appeared that both a
one-factor and a four-factor solution were possible. With regards to the one-factor solution: all
dependent variables loaded on the factor, which accounted for 41.8% of the variance. (A four-
factor solution with factors reflecting anxiety, neuroticism/depression, interpersonal/phobic prob-
638 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

lems, and sleep problems was also explored in the analyses, and produced strongly similar results.
Therefore the results of the one-factor solution are reported here). A composite score was gener-
ated by standardizing each subscale score across all assessment points by converting to Z-scores,
and then averaging across the measures (see Arntz & van den Hout, 1996; Clark et al., 1994;
Hollon et al., 1992; Rosenthal & Rosnow, 1991). Changes from pretest to later tests of composite
scores were analyzed by ANCOVA using pretest levels as the covariate. Two measures were also
separately analyzed because they were judged to be especially relevant for GAD: the STAI-trait
scale and the anxiety scale of the SCL-90. Both completers and intention-to-treat analyses were
carried out. Degrees of freedom can vary slightly due to occasional missing values.

3. Results

3.1. Pretest differences

Two significant differences were found: 4 of the 25 CT subjects had no partner, as compared
to 9 of the 20 AR subjects, Fishers Exact Test p = 0.032; and educational level was lower in
the AR condition than in the CT condition, t(43) = 2.57, p = 0.014. There were no differences
between the groups on other variables, including sex, age, work situation, duration of complaints,
number of diagnoses, medication use, pretest panic frequency and pretest composite question-
naire score.

3.2. Drop-outs

There were five dropouts from CT (20%), and three from AR (15%), NS. Two of the five
dropouts from CT and one of the three from AR claimed that they were recovered and that they
did not need further therapy. One CT patient stopped because of a pain problem requiring immedi-
ate medical attention, and subsequently moved to another city, and one AR patient stopped
because of a severe depressive episode starting and worsening during treatment (the patient was
referred to another unit). Two CT patients were dissatisfied with treatment and one AR patient
stopped for unknown reasons and could not be contacted. Two of the (improved) dropouts, one
in each condition, agreed to continue to participate in the assessments and they were included in
the completers analysis.

3.3. SCL-90 Anxiety

The anxiety subscale of the SCL-90 was separately analyzed as the most direct index of general
anxiety complaints. Fig. 1 depicts the mean scores at the four assessment points. A completers
analysis revealed highly significant decreases from pretest to posttest (main effect), t (37) =
6.30, p 0.001, ES = 1.02; from pretest to first follow-up, t (37) = 5.90, p 0.001, ES =
0.96; and from pretest to second follow-up, t (34) = 6.77, p 0.001, ES = 1.14. In the completers
sample, ANCOVA using the pretest SCL-90 anxiety score as covariate revealed NS condition
effects on posttest, F(1, 35) = 0.35, p = 0.56; on first follow-up, F(1, 35) = 3.52, p = 0.07 (in
favor of AR); or on second follow-up, F(1, 32) = 0.06, p = 0.81. An intention to treat analysis
 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646 639

Fig. 1. Mean scores for AR and CT at the four assessments for completers and intention to treat samples separately.

revealed NS condition effects at posttest, F(1, 42) = 0.74, p = 0.39, and at second follow-up,
F(1, 42) = 0.02, p = 0.97. At first follow-up, however, AR was superior to CT, F(1, 42) =
4.28, p = 0.045.

3.4. STAI-trait

The STAI-trait scale (see Table 1 for statistics and Fig. 1 for illustration) was separately ana-
lyzed. A completers analysis revealed highly significant decreases from pretest to posttest (main
effect), t (37) = 3.29, p = 0.002, ES = 0.53; from pretest to first follow-up, t (38) = 3.52, p =
0.001, ES = 0.56; and from pretest to second follow-up, t (34) = 3.71, p = 0.001, ES = 0.63. A

Table 1
Means (and standard deviations) of the STAI-trait scale (completers)

Condition Pretest Posttest Follow-up 1 Follow-up 2

CT 57.5 (8.3) 52.7 (10.3) 51.4 (10.1) 47.0 (12.8)

AR 53.7 (10.2) 47.6 (13.0) 44.1 (14.9) 44.8 (17.1)
640 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

completers analysis by ANCOVA using pretest trait score as covariate revealed NS condition
effects on posttest, F(1, 35) = 0.17, p = 0.68; on first follow-up, F(1, 36) = 1.68, p = 0.20; or
on second follow-up, F(1, 32) = 0.04, p = 0.85. An intention to treat analysis (ANCOVA) revealed
NS condition effects at posttest, F(1, 42) = 0.19, p = 0.67, at first follow-up, F(1, 42) =
1.42, p = 0.24, or at second follow-up, F(1, 42)0.09, p = 0.77.

3.5. Composite score

First, a completers analysis was done. Paired t-tests revealed highly significant reductions in
the composite score at posttest, t(37) = 5.43, p 0.001, ES = 0.88; at first follow-up, t(37) =
4.94, p 0.001, ES = 0.80; and at second follow-up t(34) = 6.25, p 0.001, ES = 1.06.
ANCOVA using the pretest composite score as covariate revealed NS differences between con-
ditions at posttest, F(1, 35) = 0.21, p = 0.65; at first follow-up, F(1, 35) = 1.36, p = 0.25; or at
second follow-up, F(1, 32) = 0.22, p = 0.64. Next, an intention-to-treat analysis was carried out,
assuming similar improvement for the one improved CT drop-out who was not further assessed
after discontinuation, and no improvement for the unimproved drop-outs (see Fig. 1). Missing
values were estimated by inter/extrapolation. ANCOVA using the pretest composite score as
covariate revealed NS differences between conditions at posttest, F(1, 42) = 0.40, p = 0.53; at
first follow-up, F(1, 42) = 1.63, p = 0.21; or at second follow-up, F(1, 42) = 0.004, p = 0.95.

3.6. Recovery rates and clinical significant change

Fisher and Durhams (1999) criteria for recovery (45) and clinical significant change (8)
on the trait scale of the STAI were used to calculate proportions of patients meeting these criteria
at posttest and both follow ups. Table 2 shows the results. Although differences were NS, AR
tended to be superior to CT, especially at first follow-up, when recovery rates are considered.
Note, however, that patients in the CT condition had at pretest, on the average, higher scores
than patients in the AR condition. Using Dutch norms of nonpsychiatric controls (m = 36.9, sd
= 8.48, cf. van der Ploeg et al., 1980), and the statistics of the present GAD sample (pretest
STAI-trait m = 55.78, sd = 9.29), the cut-off point c was 46.12, so that recovery could be defined
slightly different (46). This influenced the results only for CT at second follow-up, Table 2,
but the difference between conditions remained NS.

3.7. Results controlled for educational level and living together

Since conditions differed significantly as to educational level and living together, the ANCO-
VAs were repeated with these two variables as additional covariates. Composite score and trait
anxiety: there were no significant differences between conditions at posttest and both follow-ups
after controlling for these demographic variables (both completers and intention to treat analyses).
SCL-90 Anxiety: there were no significant differences between conditions at posttest and second
follow-up after controlling for these demographic variables (both completers and intention to
treat analyses). As in the previous analyses, at first follow-up AR tended to be superior to CT,
completers F(1, 33) = 3.14, p = 0.086, intention to treat F(1, 40) = 4.50, p = 0.04. Living together
was positively related to outcome at first follow-up (completers F (1, 33) = 3.64, p = 0.065;
Table 2
Proportions of patients meeting criteria for reliably change and recovery at the trait scale of the STAI

Condition Posttest Follow-up 1 Follow-up 2

Reliably changed Recovered Reliably changed Recovered Reliably changed Recovered

CT 40.0% 35.0% 47.6% 28.6% 50.0% 50.0%a /55.0%b

AR 33.3% 44.4% 50.0% 50.0% 46.7% 53.3%

Note: None of the differences between conditions was significant. Results are identical for the Fisher and Durham (1999) norms and Dutch norms,
except for CT at FU2.
a
Fisher and Durhams norms.
b
Dutch norms.
A. Arntz / Behaviour Research and Therapy 41 (2003) 633646
641
642 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

intention to treat F(1, 40) = 3.48, p =0.07), and at second follow-up (completers F(1, 30) =
4.51, p = 0.04; intention to treat F(1, 40) = 5.36, p = 0.026).

3.8. Results at half-year follow-up controlled for additional treatment

Fifty-six percent of the patients received some form of additional treatment between the one-
month and 6 months follow-up, the number of sessions ranging from 1 to 20 (m = 3.7). Differences
between conditions were NS as to number of patients receiving additional treatment, and to num-
ber of additional sessions. Second follow-up data were re-analyzed with number of additional
sessions as an additional covariate, but differences between conditions remained NS. This suggests
that CTs catching up with AR at second follow-up, compared to first follow-up, was not related
to the amount of additional treatment received in the period between the two follow-ups

3.9. SCID diagnosed cases

The establishment of diagnoses (including exclusion criteria) might have been more reliable
and valid using the SCID. Cases not diagnosed with the SCID (n = 14) were therefore excluded
and the data were re-analyzed (for sample size reasons intention-to-treat only). Results were simi-
lar to those obtained with the complete sample: strong treatment effects, and no differences
between conditions at posttest or at second follow-up. At first follow-up, AR tended to do better
than CT, but the differences were far from significant. Since the pattern was similar to that of
the full sample, the loss of significance in case of the SCL-90 anxiety scale is probably caused
by reduction of power.

4. Discussion

As has been mentioned by others, there is a disturbing trend to understudy GAD (Dugas, 2000).
The present study is only the fourth that offers a direct comparison between two of the most
effective treatments for GAD, AR and CT. By and large, both approaches seemed equally effec-
tive. As recovery rates and rates of clinically significantly change are comparable to those found
in other studies (Fisher & Durham, 1999; O st & Breitholtz, 2000), the lack of difference between
AR and CT does not seem to be caused by a general ineffectiveness of our treatments.
Recovery rates at second follow-up were highly comparable to those found by Fisher and
Durham, who collapsed results of 6 randomized trials. Fisher and Durham found a recovery rate
at 6 months follow-up of 60% for AR (the present study 53.3%), and 51% for C(B)T (the present
study 50 or 55%, depending on whether Dutch or international norms were used). At posttest
recovery rates were lower than those calculated by Fisher and Durham, especially for CT, suggest-
ing a slower change process in the present study. The percentages are also highly similar to those
found by O st and Breitholtz (2000), which is a reassuring finding in that it demonstrates that O
st
and Breitholtz findings are not related to allegiance effects or to sampling biases.
Though not convincingly significant in a statistical sense, the present results suggest that AR,
more than CT, had a further anxiety reducing effect one month after treatment, whereas CT caught
up with AR at six months follow-up. It is unlikely that additional treatment during FU caused
 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646 643

CT to catch up with AR, since (i) the proportion of patients receiving additional treatment was
equal in both conditions, (ii) the number of additional sessions was equal, and (iii) covarying out
the effects of additional sessions did not change the conclusions at the half-year follow-up. With
hindsight, it would have been interesting to study the patients condition at longer follow-up
periods, to see whether CT would become superior to AR. Note that Borkovec and Costello
(1993) found CT to be superior to AR at a one year follow-up, but not at posttest. Similarly, in
direct comparisons C(B)T was superior to BT in the long run (Butler et al., 1991; Durham &
Turvey, 1987; Durham et al., 1994). On the other hand, O st and Breitholtz (2000) did not observe
such an effect at their 1 year follow-up.
There is good reason to believe that the present results are consistent with other findings.
However, before accepting their validity the following points should be considered. (1) The first
14 patients were diagnosed using a checklist rather than a SCID interview. This raises the possi-
bility (in our view unlikely) of false diagnosis, or of the improper assessment of exclusion diag-
noses. Reanalysis of the 34 cases who were diagnosed with the SCID showed that the results
were stable. (2) The dependent variables lacked an instrument specifically tailored to assess GAD
intensity (as is the case for most other studies). The development of an instrument that could
serve as an international standard would be most welcome. We tried to solve this problem by
using three dependent variables: the SCL-90 anxiety subscale as the most direct measure of
intensity/frequency of anxiety symptoms, the STAI-trait scale as a measure of the general tendency
to be overanxious and to worry too much, and a composite measure as an indication of general
psychopathology. (3) There was no formal treatment integrity check, for instance by having inde-
pendent raters rate audiotapes of the sessions, although checks were consistently done by having
therapists report every week about treatment in supervision sessions. (4) The study suffered from
dropout. Fortunately, two patients agreed to continue assessments, and an intention-to-treat analy-
sis showed no major differences compared with analysis of completers only. (5) Additional treat-
ment was offered when needed, sometimes for GAD, sometimes for other problems (note the
comorbidity in the sample), and this may have influenced the half-year follow-up assessment.
However, as was reasoned earlier, this was not likely to have influenced the difference between
CT and AR at that time. Additional treatment appeared to correlate with higher scores at half-
year follow-up, indicating that it was requested by those patients with more severe, or more
complex problems.
The particularly strong points of this study are the independence of our center as to treatment
preference, the relatively high number of subjects and the high external validity. The GAD patients
were not especially recruited for the study, but came from the regular population referred to the
institute; many patients had comorbid diagnoses; and patients educational level was relatively
low. The fact that 45 of the 47 patients in our center with GAD as first axis-1 diagnosis partici-
pated, the other two having primarily axis-2 problems, underlines the high external validity of
the present study.
Though both AR and CT appear to be good treatments for GAD, there is room for improvement.
Effect sizes appeared to be moderate to high (in the 0.961.14 range for SCL-90 anxiety score,
in the 0.801.10 range for the composite score, in the 0.530.63 range for trait anxiety), and
recovery rates seem fairly high, but comparison with the recent successes of CBT for panic
disorder, suggest that further increases could be achieved (similar comments were made by O st &
Breitholtz, 2000). At least three possibilities have been suggested. Borkovec and Newman (1999)
644 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

have drawn attention to interpersonal issues that may be important in GAD, suggesting that many
GAD patients suffer from an overly nurturant attitude to others, at the cost of healthy self-serving
attitudes and behaviors. A treatment specifically addressing these issues might be helpful. Others
have suggested that meta-cognitions and attentional processes play an important role in main-
taining the disorder, especially in fueling worry (Wells, 1995). Specific forms of CT addressing
these issues have been proposed. Only one study has tested a more specific CBT treatment and
found it to be highly effective, but this study compared the active treatment only to a wait list
control group and used a artificially recruited sample (Ladouceur et al., 2000). A third possibility
is that our treatments of GAD are generally not long enough. It is noteworthy that the one study
that found disappointing effects of CT used only 8 sessions of CT (Blowers, Cobb, & Mathews,
1987). Similarly, Durham et al., 1994 Fischer and Durham, 1999) found that a more intensive
form of CT (1620 sessions) was clearly superior to a less intensive form (with only 810
sessions), especially in the long run.
A fourth possibility may be added: that treatments so far were not maximally effective in
changing the trait-like aspects of GAD. It is widely recognized that GAD has strong trait-like
characteristics and, indeed, most patients report an early (and not exact) start of the problem and
a chronic course (Rapee, 1991). The interpersonal and cognitive trait-like characteristics of GAD
might be explained by an integrative model, hypothesizing that early maladaptive schemas, trig-
gered by current uncertainty, maintain the disorder. Early childhood experiences relating to uncer-
tainty about and threat to the availability and stability of caretakers may have led to a nurturant
attitude to the caretaker (instead of the other way around), a constant sense of threat, and the
feeling that the threat cannot be coped with. These childhood experiences may have led to the
development of schemas that make the person vulnerable to develop GAD once faced with current
uncertainty. GAD patients might profit from CT approaches that specifically address schema
issues, like those that have been developed for personality disorders, and approaches addressing
early learning experiences might be especially helpful (e.g. Arntz & Weertman, 1999).
In summary, the present study confirms earlier reports that both AR and CT are effective
treatments for GAD. There was some indication that AR showed a higher effectiveness in further
reducing GAD problems one month after treatment, but at six-months follow-up CT appeared to
be related to further reduction of problems, catching up with AR. Results remained stable when
factors potentially influencing the effects were investigated: drop-outs (intention-to-treat analyses);
educational level; marital status; nonSCID-diagnosed cases; and additional treatment during the
second follow-up period. Thus, the conclusions seem relatively reliable. Effects of treatment were
stronger on symptom level than on trait level. The results suggest that there is room for improve-
ment. New studies should focus on factors like duration of treatment, changing trait-like factors
(like core schemas), worry, meta cognition, attentional processes, and interpersonal issues.

Acknowledgements

Gillian Butler and Lars-Go ran O st are acknowledged for their help in preparing the study. The
centers research assistants are acknowledged for their help in acquiring, scoring and entering the
data. The therapists are thanked for treating the patients of this study. Gillian Butler gave valuable
comments on a previous version of this report.
 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646 645

References
Arntz, A., & van den Hout, M. A. (1996). Psychological treatment of panic disorder without agoraphobia: cognitive
therapy vs. applied relaxation. Behaviour Research and Therapy, 34, 113121.
Arntz, A., & Weertman, A. (1999). Treatment of childhood memories: theory and practice. Behaviour Research and
Therapy, 37, 715740.
Arrindell, W. A., & Ettema, J. H. M. (1981). Dimensional structure, reliability and validity of the Dutch version of
the Symptom Checklist (SCL-90). Nederlands Tijdschrift voor de Psychologie en haar Grensgebieden, 36, 77108.
Barlow, D. H., Rapee, R. M., & Brown, T. A. (1992). Behavioral treatment of generalized anxiety disorder. Behavior
Therapy, 23, 551570.
Beck, A. T., & Emery, G. (1985). Anxiety disorders and phobias: A cognitive perspective. New York: Basic Books.
Bernstein, D. A., & Borkovec, T. D. (1973). Progressive relaxation training. Champaign, IL: Research Press.
Blazer, D. G., Hughes, D., George, L. K., Swartz, M., & Boyer, R. (1991). Generalized anxiety disorder. In L. N.
Robins, & D. A. Reiger (Eds.), Psychiatric disorders in America: The epidemiologic catchment area study (pp.
180203). New York: The Free Press.
Blowers, C., Cobb, J., & Mathews, A. (1987). Generalized anxiety: a controlled treatment study. Behaviour Research
and Therapy, 25, 493502.
Borkovec, T. D., & Costello, E. (1993). Efficacy of applied relaxation and cognitive-behavioral therapy in the treatment
of generalized anxiety disorder. Journal of Consulting and Clinical Psychology, 61, 611619.
Borkovec, T. D., & Newman, M. G. (1999). Worry and generalized anxiety disorder. In A. S. Bellack, M. Hersen, &
P. Salkovskis (Eds.), Comprehensive clinical psychology, Vol. 6. Adults: Clinical formulation and treatment (pp.
439459). Amsterdam: Elsevier Science (A.S.B and M.H. (Series Eds) and P.S (Vol. Ed.)).
Bouman, T. K. (1987). The measurement of depression with questionnaires, PhD dissertation, University of Groningen,
Groningen, The Netherlands.
Butler, G., Fennell, M., Robson, P., & Gelder, M. (1991). Comparison of behaviour therapy and cognitive behaviour
therapy in the treatment of generalised anxiety disorder. Journal of Consulting and Clinical Psychology, 59, 167175.
Clark, D. M., Salkovskis, P. M., Hackman, A., Middleton, H., Anastasiades, P., & Gelder, M. (1994). A comparison
of cognitive therapy, applied relaxation and imipramine in the treatment of panic disorder. British Journal of Psy-
chiatry, 164, 759769.
DeRubeis, R. J., & Crits-Cristoph, P. (1998). Empirically supported individual and group psychological treatments for
adult mental disorders. Journal of Consulting and Clinical Psychology, 66, 3752.
Dugas, M. J. (2000). Generalized anxiety disorders publications: so where do we stand? Journal of Anxiety Disorders,
1, 3140.
Durham, R. C., Murphy, T., Allan, T., Richard, K., Treveling, L. R., & Fenton, G. W. (1994). Cognitive therapy, analytic
psychotherapy and anxiety management training for generalised anxiety disorder. British Journal of Psychiatry, 165,
315323.
Durham, R. C., & Turvey, A. A. (1987). Cognitive therapy vs. behaviour therapy in the treatment of chronic general
anxiety. Behaviour Research and Therapy, 25, 229234.
Fisher, P. L., & Durham, R. C. (1999). Recovery rates in generalized anxiety disorder following psychological therapy:
an analysis of clinical significant change in the STAI-T across outcome studies since 1990. Psychological Medicine,
29, 14251434.
Hollon, S. D., DeRubeis, R. J., Evans, M. D., Wiemer, M. J., Garvey, M. J., Grove, W. M., & Tuason, V. B. (1992).
Cognitive therapy and pharmacotherapy for depression: singly and in combination. Archives of General Psychiatry,
49, 774781.
Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes, M., Eshleman, S., Wittchen, H. U., & Kendler,
K. S. (1994). Lifetime and 12-months prevalence of DSM-III-R psychiatric disorders in the United States. Archives
of General Psychiatry, 51, 819.
Ladouceur, R., Dugas, M. J., Freeston, M. H., Le ger, E., Gagnon, F., & Thibodeau, N. (2000). Efficacy of a cognitive-
behavioral treatment for generalized anxiety disorder: evaluation in a controlled clinical trial. Journal of Consulting
and Clinical Psychology, 68, 957964.
Marks, I. M., & Mathews, A. M. (1979). Brief Standard self-rating scale for phobic patients. Behavior Research and
Therapy, 17, 263267.
646 A. Arntz / Behaviour Research and Therapy 41 (2003) 633646

McNally, R. J. (1996). Methodological controversies in the treatment of panic disorder. Journal of Consulting and
Clinical Psychology, 64, 8891.
st, L. G. (1987). Applied relaxation: description of a coping technique and review of controlled studies. Behaviour
O
Research and Therapy, 25, 397409.
st, L. G., & Breitholtz, E. (2000). Applied relaxation vs. cognitive therapy in the treatment of generalized anxiety
O
disorder. Behaviour Research and Therapy, 38, 777790.
st, L. G., & Westling, B. E. (1995). Applied relaxation vs. cognitive behavior therapy in the treatment of panic
O
disorder. Behaviour Research and Therapy, 33, 145158.
Power, K. G., Jerrom, D. W. A., Simpson, R. J., Mitchell, M. J., & Swanson, V. (1989). A controlled comparison of
cognitive-behavior therapy, diazepam and placebo in the management of generalized anxiety. Behavioral Psycho-
therapy, 17, 114.
Power, K. G., Simpson, M. B., Swanson, V., & Wallace, B. (1990). A controlled comparison of cognitive-behavior
therapy, Diazepam and Placebo, alone and in combination, for the treatment of generalised anxiety disorder. Journal
of Anxiety Disorders, 4, 267292.
Rapee, R. (1991). Generalized anxiety disorder: a review of clinical features and theoretical concepts. Clinical Psy-
chology Review, 11, 419440.
Rosenthal, R., & Rosnow, R. L. (1991). Essentials of behavioral research: Methods and data analysis (2nd ed.). New
York: McGraw-Hill.
Spitzer, R. L., & Williams, J. B. W. (1986). Structured clinical interview for DSM-III-R (SCID). New York: New York
State Psychiatric Institute, Biometrics Research.
Hout, M. A., van den, Molen, G. M., van der, Griez, E., & Lousberg, E. (1987). Specificity of interoceptive fear to
panic disorder. Journal of Psychopathology and Behavioral Assessment, 9, 99109.
Ploeg van der, H. M., Defares, P. B., Spielberger, C. D. (1980). Handleiding bij de Zelf-Beoordelings Vragenlijst, Een
Nederlandstalige bewerking van de Spielberger State-Trait Anxiety Inventory. Lisse: Swets and Zeitlinger.
Wells, A. (1995). Meta-cognition and worry: a cognitive model of generalized anxiety disorder. Behavioural and Cogni-
tive Psychotherapy, 23, 301320.