V. Review and Synthesis of Available
The following sections review evidence on the outcomes of
pharmacological treatments for OCD. Unless otherwise indicated,
outcomes are given for the intent-to-treat (ITT) sample with the
last-observation-carried-forward (LOCF) method. ITT results inform
the clinician of what outcome to expect when considering all patients
exposed to a treatment. Results reported for patients who completed
the study, by contrast, indicate what to expect for those exposed
to completed durations of treatment. Finally, visitwise outcome
results indicate likely outcomes for patients exposed to those particular
durations of treatment. In considering "responder" rates
reported in OCD pharmacotherapy studies, it may be helpful to keep
in mind the responder rates reported in subjects in the placebo
arms of such studies. As noted previously in Section II.B.6, "responders" are
variously defined as subjects who experience a 25% or
35% decrease in Y-BOCS score or a CGI-I score of 1 (very
much improved) or 2 (much improved), usually after 12 weeks of treatment.
In one analysis of studies published before 1997, responder rates
in placebo subjects ranged from 0% to 35% (mean
15%), with later studies generally reporting higher placebo
response rates (70). In the interest of brevity, the
responder criteria specified above are symbolized in this
section as follows: YBOCS-25%, YBOCS-35%, and
1. Efficacy of Clomipramine
Clomipramine is a mixed serotonin and norepinephrine reuptake
inhibitor (and cholinergic and histaminic blocking agent) introduced
in Europe in 1966 for treating depression. It was subsequently used
for OCD and was approved by the FDA in 1989 for the treatment of
OCD in the United States.
Randomized controlled studies have found clomipramine significantly
superior to placebo in the treatment of OCD. However, no adequate
studies have determined the minimally effective or optimal clomipramine
dose. Trials directly comparing clomipramine with certain SSRIs
(e.g., fluoxetine, fluvoxamine, and paroxetine) report equal effectiveness.
However, in some studies the SSRIs appear to have better tolerability.
Sample sizes limited the power of most of these studies to detect
differences, and most studies did not include a placebo comparison
group. Thus, although clomipramine is recommended for treating OCD,
safety and tolerability issues favor the SSRIs.
The Clomipramine Collaborative Study (390), the
first large, double-blind, placebo-controlled trial in the United
States of a pharmacotherapy for OCD, was a landmark for the treatment
of OCD in general, and clomipramine treatment specifically. This
10-week, double-blind, placebo-controlled, multicenter study included
260 subjects in each group. Subjects who had not received prior
CBT or clomipramine and scored 16 on the Y-BOCS and 7 on the
National Institute of Mental Health Obsessive-Compulsive Scale (NIMH-OC)
were included. Subjects took clomipramine at a dose of at least
200 mg/day or matched placebo, with the opportunity to
increase the dose to 300 mg/day. The mean Y-BOCS score
in the clomipramine group decreased 40% compared with 4% in
the placebo group. The YBOCS-35% responder rate in the
clomipramine group (55%) far exceeded that in the placebo
group (7%). In a 1-year extension (391) in
a subsample (n = 134)
of participants in the Clomipramine Collaborative Study, OCD symptoms
fell to the subclinical range in 50% of the clomipramine
group, compared with 4% of the placebo group. CGI-I:1,2
responder rates for clomipramine were 50% versus 7% for
placebo. Adverse reactions, however, led 17 of the 129
(13%) subjects taking clomipramine to drop out of the study.
Many additional randomized, double-blind, placebo-controlled
studies or active-comparator studies support the effectiveness of
clomipramine treatment of OCD (70, 392, 392b).
Taken together with the findings of the Clomipramine Collaborative
Study, these studies indicate that clomipramine at doses of up to
250 mg/day is an effective treatment for OCD. Adverse effects,
especially anticholinergic and cardiovascular effects and weight
gain, are common; however, dropout rates are not high except in
one study (393). Clomipramine may elevate levels of
liver transaminases, and a potential for seizures exists at doses
exceeding 250 mg/day.
Several studies have compared clomipramine with other medications
(also see Section V.A.2). Pigott et al. (394) compared
clomipramine 250 mg/day (n = 5)
and fluoxetine 40 mg/day (n = 6)
in a small crossover trial of 10 weeks on each drug with a 4-week
washout period interposed. The Y-BOCS score decreased significantly
in both groups, with no between-group significant difference. Lopez-Ibor
et al. (395) found no difference in Y-BOCS score decrease
in an 8-week, double-blind comparison of clomipramine 150 mg/day
(n = 25) and fluoxetine
40 mg/day (n = 30).
The YBOCS-25%, but not the YBOCS-35%, responder
rate was significantly higher for clomipramine. The drugs did not
differ in dropout rates. This study was, however, small, had no placebo
control arm, and used low doses of both medications.
A 10-week, multicenter, randomized controlled trial (396)
compared clomipramine (n = 34;
maximum dose = 250 mg) and fluvoxamine (n = 30;
maximum dose = 250 mg). All patients had a Y-BOCS score
16 at baseline, and half of the patients had had prior treatment.
Improvement in Y-BOCS score was equivalent (fluvoxamine mean Y-BOCS
score decrease = 8.6; clomipramine decrease = 7.8).
Both medications were well tolerated, but the clomipramine group
experienced more sexual dysfunction. In a 10-week, multicenter,
randomized controlled trial, patients scoring 16 on the Y-BOCS
were treated with clomipramine (n = 65;
maximum dose = 300 mg, mean = 206 mg) or fluvoxamine
(n = 37; maximum dose = 300
mg, mean = 212 mg). The drugs were equally effective (YBOCS-25% responder
rates for clomipramine and fluvoxamine were 56% and 54%,
respectively). Both medications were well tolerated, with no difference
in dropout rates due to adverse events (397). Another
10-week, multicenter, randomized controlled trial enrolling patients
scoring 16 on the Y-BOCS compared clomipramine (n = 42;
maximum dose = 250 mg, mean = 255 mg) with fluvoxamine
(n = 37; maximum dose = 300
mg, mean = 201 mg). The drugs were equally effective, but
fluvoxamine was better tolerated; constipation and dry mouth were
problematic in the clomipramine group (398).
In a 12-week, double-blind, flexible-dose trial that included
a placebo arm, clomipramine (n = 99;
150–250 mg/day, mean = 113 mg/day)
and paroxetine (n = 201;
20–60 mg/day, mean = 37 mg/day)
produced equal Y-BOCS-25% responder rates (55%), which
were significantly higher than that associated with placebo (35%)
(393). Paroxetine was significantly better tolerated
than clomipramine; withdrawal rates for treatment-emergent adverse
events were 17% for clomipramine, 9% for paroxetine,
and 6% for placebo. However, the placebo group had more
subjects with "serious" treatment-emergent adverse
events (6.1%) than did the clomipramine (2.5%)
or placebo (2.0%) groups.
At least five meta-analyses have evaluated randomized, double-blind,
controlled studies comparing clomipramine with SSRIs. A meta-analysis
of studies comparing clomipramine and fluoxetine reported a greater
effect size (Cohen's d) for clomipramine (1.84) than for
fluoxetine (1.34) (399), but with fewer adverse events
for fluoxetine. Dropout rates did not differ. A later meta-analysis (400)
found the effect size (Cohen's d) for fluoxetine (3.45)
in seven studies to be greater than that for clomipramine (3.24)
in 12 studies, with a lower dropout rate for fluoxetine subjects.
Using data on subjects who completed the studies, Abramowitz (69)
found a modestly greater effect size (Cohen's d) for clomipramine
than for certain SSRIs (effects sizes: clomipramine vs. placebo,
1.31/0.66 [clinician rating/patient rating],
fluvoxamine vs. placebo, 1.28/0.37; sertraline vs. placebo,
0.37/ 1.09; fluoxetine vs. placebo, 0.68/[no
patient rating done]). When the difference in side-effect
profiles between clomipramine and placebo was statistically adjusted
to zero, the superiority of clomipramine over the SSRIs disappeared.
Eddy et al. (59) also found a greater effect size (Cohen's
d) for clomipramine in analyzing 32 randomized controlled studies
(18 involving clomipramine) published between 1980 and 2001, enrolling
3,500 subjects. The pre-post clomipramine effect size was 1.55;
the sertraline effect size (largest among the SSRIs) was 1.36. This
result must be viewed with caution because more subjects in the
clomipramine trials were treatment naive, one-third of potential
subjects were excluded from the studies, and only 80% completed
the trials. A meta-analysis using meta-regression (effect-size modeling
using least-squares regression) applied to 25 randomized controlled
trials published between 1989 and 1997 found that the superiority
of clomipramine over fluoxetine, fluvoxamine, and sertraline in
placebo-controlled trials persisted after heterogeneity effects
were controlled for (70). There was no significant
difference among the SSRIs in comparisons with placebo-controlled trial
Several explanations have been proposed for this disparity
in results between placebo-controlled and clomipramine-SSRI direct
comparison studies. Abramowitz (69) suggested that
clomipramine's apparent superiority may have resulted from
its more obvious side effects, thus diminishing the integrity of
the blind in placebo-controlled studies. Further doubt is cast on
the larger effect size of clomipramine compared with the effects
of the SSRIs by the fact that double-blind trials directly comparing
clomipramine with fluvoxamine, fluoxetine, and paroxetine showed
no difference (70), and a double-blind comparison with
sertraline found sertraline more effective (401). This
latter result was strongly influenced, however, by inappropriately
high starting doses of clomipramine (50 mg/day), which
led to a high dropout rate, and by low maximum clomipramine doses.
A meta-analysis using meta-regression (70) found that
age at onset, pre-trial OCD severity, date of publication, trial
length, and length of single-blind prerandomization each affected
the magnitude of the treatment effect; but after these predictive
factors were controlled for, clomipramine still appeared superior
in comparisons across placebo-controlled trials.
Clomipramine has been compared, albeit in methodologically
limited studies, with the MAOIs clorgyline and phenelzine. A small
crossover study with 6-week drug treatment periods found a significant
effect for clomipramine but not for clorgyline (402).
A 12-week randomized trial comparing clomipramine 225 mg/day
(n = 16) with phenelzine
75 mg/day (n = 14)
found no difference, but the study was very underpowered and used
nonstandard outcome measures (403).
Foa et al. (123) compared clomipramine (n = 36),
intensive CBT consisting of ERP (n = 29),
clomipramine plus intensive ERP (n = 31),
and placebo (n = 26).
In this 12-week randomized trial enrolling subjects with Y-BOCS
16, no major depression, and no prior adequate treatment with
clomipramine or ERP, clomipramine was more effective than placebo.
Intensive ERP combined with clomipramine was more effective than clomipramine
alone but was not more effective than intensive ERP alone.
a. Intravenous Clomipramine
A few investigators have studied the effects of intravenously
administered clomipramine, which produces higher immediate plasma
levels by avoiding first-pass liver metabolism. However, this treatment
is not available in the United States. In controlled trials, intravenous
clomipramine has been shown to be superior to placebo in treatment-resistant
patients (404). Pulse-loaded intravenous clomipramine
was more rapidly effective than identical oral doses in a double-blind
pilot study (405), but a larger study did not confirm
this finding (85). Both studies reported therapeutic
effects in some patients with very treatment-resistant OCD, suggesting
that rapid escalation of oral doses may help such patients. Pulse-loaded
intravenous clomipramine was more effective than gradually increased
intravenous clomipramine, and more rapidly so in a study in which
intravenous treatment was followed by treatment with orally administered
b. Clomipramine as an Augmentation
The strategy of adding clomipramine to an SSRI or vice versa
is supported by expert opinion (140, 175)
and several open-label trials. In a randomized, open-label, 90-day
trial that compared adding clomipramine or nothing to citalopram
in patients who had failed to benefit from adequate 16-week trials
of both clomipramine and fluoxetine, nine of nine patients in the
clomipramine augmentation group were YBOCS-35% responders,
versus only one of seven patients assigned to citalopram alone (176).
In patients with an inadequate response to 6 months of clomipramine
150 mg/day, Ravizza et al. (407) reported
a better response and fewer side effects when sertraline 50 mg/day
was added to clomipramine 150 mg/day than when the clomipramine
dose was raised to 250 mg/day.
Double-blind, placebo-controlled, and active-comparator studies
indicate that fluvoxamine is significantly more effective than placebo
and equal in efficacy to clomipramine and certain SSRIs (citalopram,
paroxetine), although the sample size for the latter comparison
was small. Compared with clomipramine, fluvoxamine showed fewer
anticholinergic side effects and better tolerability. Whether the
combination of fluvoxamine and CBT (particularly ERP) is more effective
than CBT alone is uncertain because of methodological shortcomings
in available studies.
An early double-blind, placebo-controlled trial (408)
reported positive findings when 42 OCD patientshalf of
whom also had depressive symptomswere randomly assigned
to receive fluvoxamine (up to 300 mg/day, mean final dose = 255 mg/day)
or placebo for 6–8 weeks. Nine of 21 fluvoxamine patients
were CGI-I:1,2 responders (mean Y-BOCS score decrease from baseline = 42%)
versus none in the placebo group. The majority of week 6 partial
responders became full responders at week 8 of fluvoxamine treatment, suggesting
that at least 8 weeks of treatment are needed to detect a full clinical
A 10-week, double-blind trial (409) randomly
assigned 40 OCD subjects to receive fluvoxamine (up to 300 mg/day,
mean maximum dose = 294 mg/day) or placebo and
reported a statistically significant greater improvement for fluvoxamine
than for placebo on Y-BOCS and NIMH-OC but not CGI measures.
Two pivotal 10-week, multicenter, double-blind,
placebo-controlled studies with identical study protocols provide
convincing evidence for the therapeutic efficacy of fluvoxamine
in OCD (410, 411). Subjects met DSM-III-R
criteria for OCD of at least 12 months' duration and had
an NIMH-OC scale score of 7 and a 17-item
Hamilton Depression Rating Scale (Ham-D) score of 19.
Seventy-nine (410) and 78 (411) fluvoxamine-treated
subjects and 78 (410) and 80 (411) placebo-treated
subjects completed the studies. Fluvoxamine was flexibly titrated
to 100–300 mg/day. At week 10, the mean fluvoxamine
doses were 245 and 251 mg/day, at which time the mean Y-BOCS
score had fallen 21% in the fluvoxamine groups compared
with 7% in the placebo groups (among patients who received
at least one postbaseline rating). A statistically significant difference
between the two groups was first observed at week 6. CGI-I:1,2 response
(ITT) was achieved in 33% and 38% of fluvoxamine
subjects compared with 9% and 15% of placebo subjects.
In the largest double-blind, placebo-controlled fluvoxamine
trial (412), 253 OCD subjects were randomly assigned
to receive fluvoxamine controlled release or placebo (efficacy analyses: n = 237,
117 fluvoxamine controlled release, 120 placebo). After 12 weeks,
fluvoxamine was significantly more effective than placebo on all
efficacy measures, including the Y-BOCS and CGI scales. YBOCS-25% and
YBOCS-35% response rates were significantly higher in the
fluvoxamine group, as were remission rates (44% vs. 31% and
18% vs. 8%, with Y-BOCS score definitions of
respectively). Therapeutic effects were evident at week 2, which
is earlier than reported in other fluvoxamine versus placebo studies,
with the earlier onset of effects perhaps attributable to the higher
starting dose (100 mg/day). Fluvoxamine, although having
more side effects (e.g., insomnia, nausea, somnolence) than placebo,
was safe and generally well tolerated.
In double-blind, active-comparator studies, fluvoxamine was
superior to desipramine and as efficacious as other SRIs (clomipramine
and some SSRIs); however, the lack of placebo control groups prevents
calculating the net drug effect (i.e., active drug effect minus
In an 8-week trial (413), OCD subjects were randomly
assigned to receive fluvoxamine (up to 300 mg/day, mean
final dose = 214 mg/day) or desipramine (up to
300 mg/day, mean final dose = 223 mg/day).
Forty subjects completed at least 2 weeks of treatment and were
included in the efficacy analysis. The mean Y-BOCS score decreased
29% from baseline in the fluvoxamine group and was virtually unchanged
in the desipramine group.
In a 12-week trial (414), 12 OCD subjects were
randomly assigned to receive fluvoxamine or clomipramine (both up
to 200 mg/day). For the 10 subjects who completed the study,
the Y-BOCS score decreases were similar in the treatment groups;
however, the small number of subjects limits the power to detect differences.
In a 10-week multicenter trial (396), fluvoxamine
(up to 250 mg/day, mean final dose = 200 mg/day)
was as effective as clomipramine (up to 250 mg/day, mean
final dose = 200 mg/day). Endpoint Y-BOCS scores
among the 64 randomly assigned subjects with at least one postbaseline
rating did not significantly differ between the two treatment groups.
Fluvoxamine produced fewer anticholinergic side effects and less
sexual dysfunction than clomipramine. These findings were replicated
in a subsequent 10-week study (397) that involved 79
OCD subjects randomly assigned to receive fluvoxamine (up to 300
mg/day, mean final dose = 225 mg/day)
or clomipramine (up to 300 mg/day, mean final
dose = 201 mg/day). Among the 73 subjects with
at least one postbaseline rating, the percentage of Y-BOCS-25% responders
in the two groups showed no difference at any time. The mean Y-BOCS
score decrease was 30% in both treatment groups. The fluvoxamine
group experienced fewer anticholinergic side effects.
In a large, 10-week, multicenter, double-blind trial, 227
OCD subjects were randomly assigned to receive fluvoxamine (up to
300 mg/day) or clomipramine (up to 300 mg/day)
(415). Both groups experienced a marked improvement
in OCD as evidenced by Y-BOCS, NIMH-OC, and CGI scores. Fluvoxamine
was better tolerated primarily because troubling anticholinergic side
effects were more common in the clomipramine group. In a small,
10-week, single-blind trial, 30 OCD patients were randomly assigned
to receive fluvoxamine (up to 300 mg/day, mean final
dose = 290 mg/day), paroxetine (up to 60 mg/day,
mean final dose = 53.3 mg/day), or citalopram
(up to 60 mg/day, mean final dose = 50.9
mg/day); all patients completed the study (416).
At trial endpoint the percentage of responders (with response defined
as YBOCS-35% and a CGI-I
improved]) showed no statistically significant differences,
suggesting similar effectiveness. However, the small number of subjects
in each group severely limited the power to detect differences between
Two double-blind studies compared the efficacy of fluvoxamine
combined with different forms of CBT to the efficacy of CBT alone
or of CBT combined with placebo. In a combined single- and double-blind
trial, 60 patients were randomly assigned to receive fluvoxamine
and antiexposure therapy, fluvoxamine and ERP, or placebo and ERP
(165). Pharmacotherapy (fluvoxamine up to 300 mg/day,
mean dose = 282 mg/day) lasted for 24 weeks. The
medication was then tapered over a 4-week period and discontinued,
and patients were then free to seek treatment as desired. Evaluations
were conducted after 2 months of active treatment (n = 50)
and at the end of active treatment (6 months, n = 44).
Follow-up evaluations by a blinded rater were done at 1 year (n = 37)
and 18 months (n = 33).
Fluvoxamine with ERP and fluvoxamine with antiexposure therapy yielded
greater reduction in rituals at week 8 than placebo with ERP, but
this superiority disappeared at 1 year. By week 24, all
treatments had reduced OCD symptoms, with no significant between-group
differences. Fluvoxamine plus antiexposure and fluvoxamine plus
ERP had more effect on depressive measures than did ERP plus placebo.
However, the lack of a standard response measure (Y-BOCS), the small
number of subjects in each treatment group, and varying treatments
subjects received during follow-up limit interpretation of the results.
Hohagen et al. (164) randomly assigned 60 OCD
inpatients to receive 10 weeks of either fluvoxamine (up to 300 mg/day,
mean dose = 288 mg/day) plus CBT or placebo plus
CBT. The CBT consisted of therapist-aided ERP plus cognitive restructuring.
In the 49 patients who completed the study, both treatments significantly
reduced OCD symptoms. However, there were significantly more YBOCS-35% responders
in the fluvoxamine plus CBT group (87.5%) than in the placebo
plus CBT group (60%). Post hoc analyses suggested that
patients with OCD and depression benefited more from fluvoxamine
plus CBT than from placebo plus CBT. However, this conclusion must
be viewed cautiously, as no information was given on the two groups' degree
of response to prior treatments, and the analyses excluded nine
subjects in order to equalize the two groups' baseline
Van Balkom et al. (61) randomly assigned 117
outpatients to five treatment conditions: fluvoxamine plus cognitive therapy,
fluvoxamine plus self-guided ERP, cognitive therapy alone, self-guided
ERP alone, or an 8-week wait-list control. Fluvoxamine was titrated
to 300 mg/day, with a mean endpoint dose in the two drug
groups of 197 mg/day. Pharmacotherapy lasted 16 weeks,
and a naturalistic follow-up measurement was made at 6 months. Completer
and ITT analyses posttreatment revealed no differences in effects
(Y-BOCS, SCL-90, BDI) between the four active treatment conditions; however,
this result may be due to inadequate power. Overall, 36% of
subjects who completed the study were responders (Y-BOCS score
and 6-point improvement). No evidence was found that the combination
of fluvoxamine with cognitive therapy or ERP was superior to the
cognitive therapy or ERP alone. However, neither the ERP nor the
fluvoxamine dosing was optimized. Moreover, because of the absence
of a group treated with fluvoxamine alone and of a control group
for the duration of the study, the differential efficacy of fluvoxamine,
cognitive therapy, or self-guided ERP at week 16 cannot be determined.
A recent follow-up study (66) assessed 62 OCD
subjects who completed controlled trials (23 treated with CBT consisting
of ERP alone, 24 with SRI alone [fluvoxamine or clomipramine],
15 with ERP plus medication) and found that most subjects showed
long-term improvement following either ERP or medication treatment.
The small number of patients in each group, however, limited the
power to detect differences between the groups.
One fluvoxamine study supports the hypothesis that OCD with
co-occurring chronic tic disorders may be a clinically meaningful
subtype. An 8-week open-label trial (223) assessed
the efficacy of fluvoxamine in 66 OCD patients, of whom 33 had tic
disorders. Of the OCD patients with co-occurring chronic tic disorders,
21% were fluvoxamine YBOCS-35% and CGI-I:1,2
responders compared with 52% of the OCD patients without
co-occurring chronic tics. The authors concluded that fluvoxamine
monotherapy may be less efficacious in OCD patients with tics than
in those free of this condition. A clomipramine study, however,
found no reduction in effectiveness in OCD patients with tics (224).
Three randomized, double-blind, placebo-controlled studies
show that fluoxetine is significantly more effective than placebo.
In addition, double-blind active-comparator studies suggest fluoxetine
is comparable in efficacy to clomipramine and sertraline and superior
in efficacy to phenelzine. Compared with clomipramine, fluoxetine
exhibited fewer side effects in one study. In other studies, fluoxetine
was well tolerated, with side effects comparable to those of the
Two large, randomized, double-blind, placebo-controlled studies
demonstrated the effectiveness of fluoxetine in the treatment of
adults with DSM-III-R OCD. In an 8-week double-blind study
(417), 214 subjects were randomly assigned to receive
fluoxetine 20, 40, or 60 mg/day or placebo. Fluoxetine
response (defined as YBOCS-25% and CGI-I:1,2)
rates were significantly higher in the 40 mg/day
and 60 mg/day groups (48% and 47%, respectively)
than in the placebo group (26%), but the response rate
in the 20 mg/day group (36%) was not. In a 16-week
extension, subjects who had not responded to 20 mg/day
or 40 mg/day and who took fluoxetine 60 mg/day
experienced a highly significant decrease in Y-BOCS scores. Fluoxetine
and placebo dropout rates did not differ significantly.
A 13-week, randomized, double-blind trial (83)
assessed the effects of fluoxetine 20, 40, or 60 mg/day
versus placebo in 355 outpatients with OCD. At each dose, fluoxetine
was significantly superior to placebo on the Y-BOCS and other efficacy
measures, with statistical significance reached by week 5. The fluoxetine
groups had YBOCS-35% response rates of 32%, 32%,
and 35%, respectively, with a trend for greater improvement
in the 60 mg/day group. The YBOCS-35% response
rate in the placebo group was 8.5%. The safety and efficacy
of fluoxetine in the acute treatment of OCD are further supported by
open trials (418–421).
Two studies compared fluoxetine with clomipramine in the treatment
of DSM-III-R OCD without using a placebo control group (394, 395).
In the first study, involving crossover designs with 10 weeks of
treatment, 4 weeks of drug washout, and samples of 6 and 20 subjects,
fluoxetine up to 80 mg/day was as effective as clomipramine
up to 250 mg/day (394). Both drugs produced a significant
decrease in the Y-BOCS score, although clomipramine was associated
with more adverse events. The second study, an 8-week, double-blind,
randomized trial, compared fluoxetine 40 mg/day (n = 30)
with clomipramine 150 mg/day (n = 25)
(395). The two drugs appeared equally effective over this short
treatment period. The YBOCS-25% responder rate, but not
the YBOCS-35% responder rate, was higher with clomipramine.
The discontinuation rates for adverse events were 3% for
fluoxetine and 4% for clomipramine.
A 24-week, randomized, double-blind trial compared the efficacy
and tolerability of fluoxetine (mean dose = 57 ± 23 mg/day)
and sertraline (mean dose = 140 ± 59 mg/day) in
outpatients with DSM-IV OCD (422). Equivalent and significant
improvement was found at week 24 in Y-BOCS and NIMH-OC scale scores.
Remission rates (defined as Y-BOCS score 11 and CGI-I:1,2)
at weeks 12 and 24 were significantly higher for sertraline (36% vs.
22% at week 24). Subjects treated with sertraline showed
an earlier improvement on some, but not all, efficacy measures.
Both medications were well tolerated; rates of discontinuation due
to adverse events were 14% for fluoxetine and 19% for
A 10-week randomized trial compared fluoxetine 80 mg/day,
phenelzine 60 mg/day (both doses achieved by the
end of week 3), and placebo in 64 adults with DMS-III-R OCD (423).
Fluoxetine was superior to placebo at weeks 6 and 10 as well as
to phenelzine at week 10. Symmetry obsessions and lower baseline
Y-BOCS scores were significantly more common in phenelzine responders
than in fluoxetine responders; however, this post hoc analysis provides only
weak evidence for a phenelzine effect in this subgroup.
The long-term treatment of OCD with fluoxetine has been examined
to a limited extent. In a continuation of the 13-week, double-blind,
placebo-controlled, fixed-dose fluoxetine study (83),
treatment responders continued their blinded treatment, whereas
nonresponders began a 24-week open-label trial of maximally tolerated
doses up to 80 mg/day (81). Among acute-phase
responders, all three doses of fluoxetine (20, 40, and 60 mg/day)
were associated with further Y-BOCS improvement. The acute-phase
nonresponders benefited from upward dose titration, with two-thirds
achieving a YBOCS-35% response. Another study assessed
the efficacy and safety of 52 weeks of fluoxetine or placebo treatment
in patients with DSM-IV OCD who had responded to single-blind fluoxetine
for 20 weeks (201). Patients who received fluoxetine
had numerically lower relapse rates compared with those who received
placebo, although the difference was not significant (see Section
V.E for details).
Three double-blind placebo-controlled trials show paroxetine
to be more effective than placebo acutely; an additional double-blind
study shows the superiority of paroxetine relative to placebo in
maintaining response over 6 months of continuation treatment.
A double-blind active-comparator study suggests that paroxetine
is comparable in efficacy to clomipramine. Compared with venlafaxine,
the relative efficacy of paroxetine is less clear, as findings vary
with the definition of treatment response. Paroxetine's
tolerability is comparable to that of other SSRIs. Some evidence (424),
but not all (80), suggests paroxetine is more likely
to be associated with significant weight gain. Paroxetine is more
likely to induce anticholinergic side effects than are other SSRIs
(118, 425). It also carries a greater
risk of an unpleasant withdrawal syndrome, comparable to the risk
associated with venlafaxine (426).
In a 12-week double-blind trial (80), OCD patients
without co-occurring major depression, tics, or Tourette's
disorder were randomly assigned to receive paroxetine 20 mg/day
(n = 88), 40 mg/day
(n = 86), 60 mg/day
(n = 85), or placebo (n = 89).
A little more than half of subjects had had a prior SRI trial. Endpoint
response rates (defined as YBOCS-25% or Clinical
Global Impression–Severity [CGI-S] score
decrease of 2 points) for paroxetine 40 mg/day
(25%) and 60 mg/day (29%), but not 20
mg/day (16%), were significantly greater than
for placebo (13%). In a 12-week, double-blind, flexible-dose
study (427), 191 subjects were randomly assigned to
receive placebo or paroxetine, with the dose increasing from 20
mg/day to 40 mg/day by week 3 and up
to 50 mg/day from week 8 onward. The CGI-I:1,2 response
rate was significantly greater in the paroxetine (50%)
than in the placebo group (24%). A significantly greater
response rate was similarly observed in subjects randomly assigned to
receive 12 weeks of flexibly dosed paroxetine 20–60 mg/day
(mean dose = 37 mg/day) (n = 201)
or placebo (n = 99) (393).
More than half (55%) of the paroxetine subjects were YBOCS-25% responders
compared with 35% of placebo subjects. The active comparator,
flexibly dosed clomipramine (150–250 mg/day, mean
dose = 113 mg/day), produced the same responder
rate as paroxetine.
A 12-week, randomized, double-blind, flexible-dose study (172)
comparing paroxetine with venlafaxine found no significant difference
in YBOCS-35% responder rates for paroxetine at doses up
to 60 mg/day (44%) (n = 76)
and venlafaxine at doses up to 300 mg/day (37%)
(n = 75), although the
YBOCS-25% responder rate was higher for paroxetine (66%)
than for venlafaxine (49%). When the medication for nonresponders
in this study was switched to the alternative medication in a double-blind
fashion, a higher YBOCS-25% responder rate was observed
for paroxetine (56% [15/27])
than for venlafaxine (19% [3/19])
Long-term effectiveness of paroxetine has been observed in
one study. Responders to paroxetine in a 12-week double-blind study
and its 6-month open-label, flexible-dose extension phase (N = 105)
were randomly assigned to receive 6 months of double-blind
paroxetine or placebo (80). Relapse was defined as
a return to the baseline Y-BOCS score or an increase of 1 point
in the CGI-S score for more than one visit. Subjects assigned to
placebo had a significantly higher relapse rate (59%) than
those assigned to paroxetine (38%). The mean time to relapse
was 29 days in the placebo group and 63 days in the paroxetine group.
Two double-blind, placebo-controlled trials demonstrated the
efficacy of sertraline in treating OCD. In double-blind active-comparator
studies, sertraline appeared comparable in efficacy to fluoxetine.
Sertraline was superior in efficacy to clomipramine (although methodological
shortcomings influenced the latter comparison) and, in subjects
with co-occurring depression, was superior to desipramine. Finally,
in an open-label trial, subjects who responded to 1 year
of treatment with sertraline experienced further small but noticeable
decreases in symptoms when treatment was extended to 2 years.
In a 12-week, randomized, fixed-dose trial (82),
subjects were assigned to sertraline 50 mg/day (n = 80),
100 mg/day (n = 81),
200 mg/day (n = 80),
or placebo (n = 84). Sertraline
at doses of 50 mg/day and 200 mg/day was significantly
superior to placebo with regard to change in Y-BOCS, NIMH-OC, CGI-S,
and CGI-I scores, but at 100 mg/day sertraline was only
superior in terms of the NIMH-OC, probably because of the high dropout
rate (33%) in this group. At endpoint, CGI:1,2 responder
rates were 39% for sertraline and 30% for placebo.
A 12-week, double-blind, randomized study of flexibly-dosed sertraline
50–200 mg/day (mean maximum dose at endpoint = 165
± 55 mg/day) found the drug (n = 86)
more effective than placebo (n = 81)
with regard to change in Y-BOCS, NIMH-OC, and CGI-S scores (428).
The CGI-I:1,2 responder rate was numerically but not significantly
higher for sertraline (41%) than for placebo (23%).
In a 24-week, double-blind, randomized, flexible-dose comparison
of sertraline 50–200 mg/day (mean endpoint dose = 140
± 59 mg/day) (n = 77)
versus fluoxetine 20–80 mg/day (mean endpoint
dose = 57 ± 23 mg/day), the differences in CGI-I:1,2
responder rates (60% and 60%) and remission (defined
as CGI-I:1,2 plus Y-BOCS < 12)
rates (36% vs. 22%) were not significant (422).
A 16-week double-blind study compared sertraline and clomipramine
50 mg/day for 4 weeks followed by flexible increases in
dose to 200 mg/day (mean final dose = 132 mg/day
for sertraline and 101 mg/day for clomipramine) (401).
The sertraline group had significantly greater improvement as measured
by the Y-BOCS, NIMH-OC, and CGI-S. Inappropriately high starting
doses of clomipramine (50 mg/day), producing a high dropout
rate and low maximum clomipramine dose, strongly influenced the
comparative result. Among subjects treated for at least 4 weeks,
the two drugs produced equal results, but the mean final clomipramine
dose was relatively low. The Y-BOCS-35% responder rates
were 72% for sertraline and 65% for clomipramine.
In another double-blind, flexible-dose study (429), OCD patients
with co-occurring depression were randomly assigned to receive sertraline
50–200 mg/day (mean endpoint dose = 160
± 50 mg/day) or desipramine 50–300 mg/day
(mean endpoint dose = 194 ± 90 mg/day). Sertraline
(n = 79) was more effective
than desipramine (n = 85)
in bringing about "robust improvement in OCD symptoms" (Y-BOCS
score decrease 40%).
A second completed year of continued treatment with open-label
sertraline flexibly dosed from 50 mg/day to 200 mg/day
was associated with a mean decrease in Y-BOCS scores from about
12 to about 9 in 38 subjects (430) who had been CGI-I:1,2
responders in a 1-year, fixed-dose, double-blind study (431).
Finally, after 1 year of single-blind treatment, sertraline
responders rarely relapsed over 28 weeks regardless of whether they
were maintained on flexibly dosed sertraline (50–200 mg/day)
(3/108, or 3%) or switched over 2 weeks
to placebo (5/113, or 4%) (200) (see
Section V.E. for details).
A double-blind, placebo-controlled trial showed citalopram
to be more effective than placebo, with a trend for greater efficacy
and more rapid response at a higher dose. Several open trials suggest
efficacy for citalopram in individuals whose OCD has not responded
to other SRIs. In addition, several open-label trials suggest comparable
efficacy to other SRIs. The active isomer in citalopram, escitalopram,
is now marketed as a separate SSRI in the United States.
In the only double-blind, placebo-controlled, randomized trial,
12 weeks of treatment with fixed-dose citalopram 20 mg/day
(n = 102), 40 mg/day
(n = 98), or 60 mg/day
(n = 100) produced higher
YBOCS-25% response rates (57%, 52%, and
65%, respectively) than did placebo (n = 101)
(37%) (432). There were trends for the highest
dose to be associated with a more rapid response.
A small open-label trial suggests that citalopram (n = 11;
mean dose = 51 mg/day) and paroxetine (n = 9;
mean dose = 53 mg/day) bring about similar YBOCS-35% responder
rates (40% and 45%, respectively) in inpatients
(416). The response rate to fluvoxamine (n = 10;
mean dose = 290 mg/ day) (60%) was numerically
but not statistically significantly higher. An open-label, random-assignment,
flexible-dose study utilizing a blinded rater found no significant
difference in YBOCS-35% responder rates to 12 weeks of
citalopram 40–60 mg/day (n = 23)
(48%), fluvoxamine 200–300 mg/day (n = 83)
(55%), clomipramine 150–250 mg/ day (n = 37)
(48%), or paroxetine 40–60 mg/day (n = 16)
A single case report described a patient whose OCD was unresponsive
after 3 months of citalopram 80 mg/day but subsequently
responded to 160 mg/day, which was well tolerated over
several months (436). Intravenous citalopram (unavailable
in the United States) was well tolerated in one study at doses of
20–80 mg/day and may have a faster onset of action
than oral citalopram (437).
Escitalopram was as effective as paroxetine for OCD in a European
multicenter double-blind, active-comparator trial (437a) and was
superior to placebo in preventing OCD relapse in a second large
European double-blind trial (437b).
Venlafaxine is a serotonin-norepinephrine reuptake inhibitor
(SNRI) that does not have an FDA indication for OCD. A small, double-blind,
placebo-controlled trial with venlafaxine was negative, but several
open-label trials showed robust responses in OCD symptoms at doses
of at least 225 mg/day. In addition, double-blind active-comparator
studies suggest venlafaxine is comparable in efficacy to clomipramine
and perhaps to paroxetine. Venlafaxine has been generally well tolerated.
In the only double-blind, placebo-controlled trial (438),
30 OCD patients were randomly assigned to receive placebo or venlafaxine
(up to 225 mg/day) for 8 weeks. At endpoint, there were
no statistically significant differences in response, although there
was a trend for greater response in the venlafaxine group. The study's
small sample size, short trial length, low venlafaxine dose, and
lack of standard outcome measures (Clinical Global Impression, ratings
of avoidance) severely constrain interpretation.
In a 12-week, double-blind, active-comparator study (172),
150 OCD patients were randomly assigned to receive venlafaxine XR
(up to 300 mg/day) or paroxetine (up to 60 mg/day).
Full response was defined as a Y-BOCS score decrease of 50% and
partial response as a decrease of 35%. An ITT LOCF analysis
demonstrated no significant differences in responder rates (full response:
24% venlafaxine vs. 22% paroxetine; partial response:
37% venlafaxine vs. 44% paroxetine). Only a small
percentage of patients (5%) dropped out because of adverse
effects. The study's methodological limitations include
the absence of a placebo control group and venlafaxine doses not
exceeding 300 mg/day. In addition, the venlafaxine group
had undergone more unsuccessful medication trials.
Nonresponders (Y-BOCS decrease < 25%) in this
study (n = 43) were treated
for 12 additional weeks with the alternative medication (171).
A significantly higher proportion of those whose medication was
switched to paroxetine (56%, 15/27) were YBOCS-25% responders
compared with those whose medication was switched to venlafaxine
XR (19%, 3/16). However, the small sample size,
the lack of a placebo control group, and a less stringent response
criterion are methodological limitations in this second study.
In a 12-week double-blind trial (173), 73 OCD
subjects were randomly assigned to receive venlafaxine (225–350
mg/day, mean dose = 265 mg/day) or clomipramine
(150–225 mg/day, mean dose = 168 mg/day).
Visitwise and LOCF analyses at study end revealed no statistically
significant difference between the groups in YBOCS-35% and CGI-I:1,2
responder rates (visitwise responder rates: venlafaxine 36% vs.
clomipramine 50%; LOCF responder rates: venlafaxine 35% vs.
clomipramine 43%). The investigators concluded that venlafaxine
at these doses may be as effective acutely as clomipramine, with
fewer side effects. However, confidence in these results is again
limited by the lack of a placebo control group, the small size of
the study, and by the relatively low mean clomipramine dose.
An open, naturalistic, retrospective study examined treatment
results for 39 OCD patients (29 who were "nonresponders" [undefined] to
one or more SRI trials) after treatment for a mean of 18 months
(range 1–56 months) with venlafaxine up to 450 mg/day
(mean final dose = 230 mg/day) (439).
At study end, 69% of subjects entering the study were CGI-I:1,2
responders. Of note, 76% of the "nonresponders" to
one or more SRI trials, and 82% of the "nonresponders" to
two or more SRI trials were sustained responders. Venlafaxine even
at the higher end of the dosing range was well tolerated.
An 8-week open-label trial in which 12 OCD patients were treated
with venlafaxine 150–300 mg/day reported responder
rates of 75% (YBOCS-35%) and 35% (CGI-I:1,2)
(440) without substantial side effects. A 12-week open-label
trial utilizing venlafaxine 150–350 mg/day in 10
OCD patients reported responder rates of 30% (YBOCS-35%)
and 40% (CGI–I:1,2), with a more robust response
in treatment-naive patients (441). Marazziti (442)
reported five patients whose OCD was resistant to SSRIs
who improved (Y-BOCS, Ham-D, and other clinical evaluations) for
at least 1 year with venlafaxine 150–225 mg/day.
3. Implementation of SRIs
Available trial data suggest that higher SSRI doses produce
a somewhat higher response rate and somewhat greater magnitude of
symptom relief (79–82) (Table 5).
Table 5. Effects of Higher Selective Serotonin Reuptake Inhibitor
(SSRI) Doses in Fixed-Dose Trials on Obsessive-Compulsive Disorder
Among nonresponders, raising the dose of an SSRI is associated
with enhanced response (Table 6). The literature does not allow
specification of the chance of response as a function of the number
of previously failed adequate SRI trials. Attempts to interpret
the clinical trial data are limited by differences in the number
of failed trials in patients included in a given study, by absence
of information about the number of failed adequate trials, by differences
in the definition of "failed," and by the small,
highly selected samples. However, clinical experience suggests that
patients who do not respond to one SRI may still respond well to
another (Table 7). With SRIs, response rates to a second trial are
close to 50% but may fall off as the number of failed adequate
trials increases. A switch to venlafaxine at doses of 225–350
mg/day is also supported by active-comparator trials and
open-label studies that suggest its effectiveness in treating OCD.
Table 6. Effects of Raising the Dose of Selected SSRIs in Nonresponders
| Add to My POL
Table 6. Effects of Raising the Dose of Selected SSRIs in Nonresponders
|Initial Phase||Continuation Phase|
|Drug||Duration||Dose||Duration||Dose||Responder Ratea (n)|
|Fluoxetineb||13 weeks||20 mg||26 weeks ||60 mg||80%c (8/10)|
|80 mg ||46%c (15/33)|
|40 mg||60 mg||20%c (1/5)|
|80 mg||53%c (17/32)|
|60 mg||80 mg||50%c (15/30)|
|Sertralined||16 weeks|| 200 mg||12 weeks|| 200 mg||34%e (10/29)|
| 377 mg||52%e (11/21)|
Table 7. Chance of Responding to the Next Serotonin Reuptake Inhibitor
(SRI) After Failure to Benefit From the Previous SRI
a. Monoamine Oxidase Inhibitors
There is only very weak support for the use of MAOIs in OCD.
In one small, double-blind, placebo- and fluoxetine-controlled study,
the Y-BOCS score decrease for subjects completing a 10-week trial
of phenelzine 60 mg/day (n = 17)
was not significantly greater than for those completing placebo
treatment (n = 18), in
contrast to the Y-BOCS score decrease produced by fluoxetine 80
mg/day (n = 19)
(423). In a post hoc analysis, the authors suggested
that symmetry obsessions might be a strong predictor of phenelzine "response" (undefined).
In a blinded, 12-week, random-assignment, open-label comparison
of phenelzine 75 mg/day (n = 12
completers) and clomipramine 225 mg/day (n = 14
completers), with both doses reached by week 5, no significant difference
was found on either the Maudsley Obsessional-Compulsive Inventory
(MOCI) or a nonstandard scale (403). The absence of
a placebo group, use of nonstandard rating scales, and small sample
size limit this study's evidentiary weight. A case series
(443) and isolated case reports add only minimal evidence
of the effectiveness of MAOIs. The presence of severe anxiety or
panic attacks or of symmetry obsessions has been a positive predictor
in some case reports.
The side-effect burden of MAOIs can be significant and includes
cardiovascular problems and weight gain, as well as potentially
severe drug-drug interactions and dietary restrictions associated
with nonselective MAOIs or high-dose selective MAOIs (444, 445).
This burden, combined with the relative lack of evidence for MAOI
efficacy, argues against the use of these medications except in
severely ill OCD patients who have failed most or all first-line treatments
and most second-line treatments.
b. Tricyclic Antidepressants
Limited investigations of TCAs other than clomipramine have
found no evidence for their efficacy in treating OCD.
A randomized controlled trial comparing nortriptyline, clomipramine,
and placebo with eight subjects in each treatment group found that
clomipramine, but not nortriptyline, was superior to placebo in
reducing interview-based ratings of OCD severity (446).
However, there was no significant difference in effectiveness between
clomipramine and nortriptyline.
In a placebo-controlled trial that divided OCD subjects into
a "high depression" group (Beck Depression Inventory [BDI] score
21) and a "low depression" group (BDI score £ 20),
imipramine (mean dose = 233/mg/day) reduced
depression over 6 weeks in the highly depressed patients (n = 37)
but did not affect the obsessive-compulsive symptoms in either depressed
In another study, 38 patients were divided into moderately
and mildly depressed groups according to their Beck Depression Inventory
scores (232). One half of each group received imipramine
and the other half received placebo for 6 weeks followed by 3 weeks
of daily CBT consisting of ERP and then 12 weekly sessions of supportive
psychotherapy. Although imipramine improved depressive symptoms
in the depressed patients, it did not affect obsessive-compulsive
symptom severity. ERP reduced OCD symptom severity, but imipramine
did not potentiate ERP effects. Response of OCD to therapy did not
differ in moderately depressed versus mildly depressed patients.
Case reports and case series (448) suggest that
trazodone at doses of at least 250 mg/day may warrant a
trial in OCD patients who have not responded to first- and second-line
treatments. In one case series (N = 5),
augmentation of an SSRI with trazodone 300–600
mg/day was helpful in alleviating OCD and anxiety as well
as sleep disturbance, gastrointestinal distress, and sexual dysfunction
(449). However, a 10-week, double-blind, placebo-controlled
trial with 11 patients who completed the trazodone trial (mean dose = 235
mg/day) and 6 patients who completed the placebo trial
found no evidence of efficacy (450). Nevertheless,
the trial may have been too short (6 weeks at 250 mg/day)
with too small a sample to allow definitive conclusions. If trazodone
is used, sedation is likely to be a limiting side effect, and torsades
de pointes has been reported on rare occasions (451).
Males must be warned of the risk of priapism, which may occur in
from 1/1,000 to 1/10,000 men (451).
Few studies have examined the efficacy of antipsychotics as
monotherapy for OCD, and the available evidence does not support
such use. An early study (452) examined chlorpromazine
in a mixed population of patients (those with "psychoneuroses
or personality disorders with some symptoms of an obsessive or compulsive
type") and did not use standardized assessment instruments.
Case studies using haloperidol were inconclusive (e.g., references 453 and 454),
although a case report described an OCD patient who responded well
to loxapine (455).
An open 10-week trial involving 12 OCD patients examined the
possible efficacy of clozapine 300–600 mg/day
(456). The patients had not responded (Y-BOCS score
decrease 35% or score < 16 and a CGI-I:1,2) to prior
SRI trials. Two patients dropped out because of side effects (i.e.,
sedation and hypotension). Among the 10 patients who completed the
study, none had a response to the trial; the mean Y-BOCS reduction
was 10%. The authors concluded that clozapine is ineffective
as monotherapy in patients who have not responded to prior SRI treatment.
Most recently, Connor et al. (457) examined the
efficacy of aripiprazole in eight OCD patients over 8 weeks. Seven
patients took aripiprazole at a dose of 10–30 mg/day,
but two dropped out due to side effects (i.e., akathisia, nausea).
Among the five completers, three experienced a Y-BOCS decrease of
30%; two subjects were rated much or very much improved.
The authors concluded that some patients may benefit from aripiprazole
monotherapy. However, the small sample and open-label design preclude
In the many OCD patients who have either no response or a
partial response to SRI treatment, antipsychotic medication has
been used to augment treatment with an SRI. Randomized, placebo-controlled
augmentation trials of both first-generation (haloperidol) and second-generation
(risperidone, olanzapine, quetiapine) antipsychotic medications
have yielded response rates in the range of 40% to 55% within
4–6 weeks (Table 8).
Table 8. Results of Second-Generation Antipsychotic Augmentation
in Treatment-Resistant Obsessive-Compulsive Disorder (Double-Blind,
| Add to My POL
Table 8. Results of Second-Generation Antipsychotic Augmentation
in Treatment-Resistant Obsessive-Compulsive Disorder (Double-Blind,
|Medication||Mean Final Dose (mg/day)||Final Dose Range (mg/day)||Active Drug: Respondersa/Total
(N)||Percentage of Responders, Drug/Placebo|
|Carey et al. 2005 (458)b||Quetiapine||169||25–300||8/20||40%/48%|
|Fineberg et al. 2005 (459)||Quetiapine||215||50–400||3/11||27%/10%|
|Denys et al. 2004 (158)||Quetiapine||c||200–300||8/20||40%/10%|
|Erzegovesi et al. 2005 (157)||Risperidone||0.5||0.5||5/10d||50%/20%|
|McDougle et al. 2000 (155)||Risperidone||2.2||1–4||9/20||45%/0%|
|Hollander et al. 2003 (156)||Risperidone||2.3||NA||4/10||40%/0%|
|Bystritsky et al. 2004 (159)||Olanzapine||11.2||5–20||6/13||46%/0%|
|Shapira et al. 2004 (460)b||Olanzapine||6.1||5–10||9/22||41%/41%|
Other controlled trials did not find significant differences
between antipsychotic and placebo augmentation (458–460);
however, methodological limitations in these studies likely contributed
to the negative findings. In two of these studies (458, 460),
in which SSRI monotherapy was limited to 8 weeks, the failure of
active drug to separate from placebo was probably due to a high
rate of response to continued SSRI monotherapy in the placebo augmentation
group. One 16-week, double-blind, placebo-controlled trial (459)
found no significant difference between quetiapine (mean final dose = 50–400
mg/day) and placebo. Again, the dosing may have been low.
The long-term effects of antipsychotic augmentation have not
been systematically studied. A retrospective chart review (160)
found that 15 of 18 patients (83%) who responded to antipsychotic
augmentation relapsed within 1 year after the antipsychotic
was discontinued. Thirteen of the 15 who relapsed did so by the
eighth week after discontinuation.
Many questions about antipsychotic augmentation in OCD remain
unanswered, including the optimal dose for each of the agents, their
long-term tolerability, and the reasons some patients benefit but
others do not. In addition, the relative efficacy of the different
agents remains to be examined.
In the first double-blind, placebo-controlled study of antipsychotic
augmentation (154), 34 patients with OCD resistant to 8 weeks of
fluvoxamine (defined as less than a YBOCS-35% decrease
or a score 16 and not having a CGI-I:1,2) were randomly
assigned to receive 4 weeks of adjunctive haloperidol (n = 17)
or placebo (n = 17). Adjunctive
haloperidol (initiated at 2 mg and increased to a maximum of 10
mg/day) was significantly more effective than placebo.
Eleven of the 17 haloperidol patients responded versus none of the
patients receiving placebo, but akathisia requiring propranolol
treatment was common. Response was defined as 1) YBOCS-35% and score
< 16; 2) CGI-I:1,2; and 3) consensus of the treating clinician
and two of the primary investigators. Of the 11 responders, 7 met
all three criteria, and 4 met two criteria. All 8 subjects with co-occurring
tics responded to haloperidol, versus 3 of 9 without tics. No subject
with tics responded to placebo. The authors concluded that OCD patients
with a chronic tic disorder might benefit from adjunctive haloperidol
but that it should not be used indiscriminately because of the risk
of tardive dyskinesia.
A 9-week, double-blind, placebo-controlled, crossover study
compared 2 weeks of adjunctive treatments with risperidone 1 mg/day,
haloperidol 2 mg/day, or placebo in 16 patients with Y-BOCS
scores of 16 after at least 12 weeks of therapeutic SRI doses
(461). Haloperidol augmentation, but not risperidone
augmentation, reduced the Y-BOCS score significantly more than did
placebo augmentation. Both drugs were significantly better than
placebo at reducing Y-BOCS obsession scores. Four subjects dropped
out of the study before receiving neuroleptic treatment. Of the
12 subjects (75%) who completed the risperidone arm, 5
(42%) discontinued haloperidol for adverse events. The
low dose of risperidone that was used constrains interpretation
of the study results (see Section V.A.5.d).
Three double-blind, placebo-controlled studies, albeit of
modest size, and several open-label studies support the safety and
effectiveness of risperidone augmentation of SRI treatment of OCD. McDougle
et al. (155) randomly assigned 36 patients whose OCD
was resistant to 12 weeks of SRI treatment (< YBOCS-35% decrease
or a score 16 and CGI-I:1,2) to 6 weeks of adjunctive risperidone
(n = 20) or placebo (n = 16).
Risperidone was initiated at 1 mg/day, and the dose was
increased by 1 mg weekly. The mean final risperidone dose was only
2.2 mg/day (SD = 0.7 mg/day; range = 1–4).
Among patients who completed the trial (risperidone, n = 18;
placebo, n = 15), risperidone
was significantly superior to placebo (Y-BOCS reduction: 32% for
risperidone; 9% for placebo). Nine (50%) of the
18 patients who completed the risperidone trial were responders
compared with none of the 15 patients who completed the placebo
trial. Response was defined as 1) YBOCS-35% and score
< 16; 2) CGI-I:1,2; and 3) consensus of the treating clinician
and two of the primary investigators. There was no difference in
outcome between OCD patients with and without co-occurring tic disorder
or schizotypal personality disorder. Risperidone was well tolerated,
with mild transient sedation being the most prominent adverse effect;
one risperidone patient dropped out in the first week because of
In a smaller controlled study (156), 16 OCD patients
who had "failed" (i.e., no more than minimally
improved) at least two 12-week SRI trials were randomly assigned
to receive 8 weeks of adjunctive risperidone (n = 10)
or placebo (n = 6). Risperidone
was started at 0.5 mg/day, and the dose was increased by
0.5 mg weekly to a maximum of 3 mg/day; the mean risperidone
dose was 2.25 mg/day (SD = 0.86), with no difference
between responders and nonresponders. In the ITT sample, the risperidone
group had a numerically larger mean Y-BOCS score decrease (25%)
than the placebo group (5%). Four of 10 (40%)
risperidone patients and none of six (0%) placebo patients
were YBOCS-25% responders. Three subjects discontinued
(risperidone, n = 1; placebo, n = 2)
because of unsatisfactory clinical response. Risperidone was generally
well tolerated; only four risperidone patients experienced side
effects (i.e., sedation, dizziness, dry mouth).
A randomized controlled trial (157) examined
the efficacy of adding risperidone 0.5 mg/day versus placebo
in OCD patients who had either responded or not responded
to their first SRI trial (12 weeks of fluvoxamine, maximum dose = 300
mg/day, final doses not provided). Responders (defined
as those with YBOCS-35% and CGI-I:1,2)
and "nonresponders" were then randomly assigned
to receive added risperidone 0.5 mg/day or placebo for
6 weeks. Among the 39 patients completing the trial, added risperidone significantly
reduced OCD symptoms in the 10 fluvoxamine nonresponders but not
in the 9 fluvoxamine responders (Y-BOCS reduction for fluvoxamine
nonresponders: risperidone 26%, placebo 7%; Y-BOCS
reduction for fluvoxamine responders: risperidone 4%, placebo
28%). Among the fluvoxamine nonresponders, 5 of 10 (50%)
risperidone patients and 2 of 10 (20%) placebo patients
became YBOCS-35% responders. The study's limitations
include the small sample size, the potential for ceiling effects
in the fluvoxamine responders, the low dose of risperidone, and the
lack of information about whether the treatment groups received
similar fluvoxamine doses prior to augmentation.
The safety and effectiveness of adjunctive olanzapine in OCD
have been examined in two randomized, placebo-controlled trials
and several open-label trials. Bystritsky et al. (159)
randomly assigned 26 OCD patients who had not "improved" (undefined)
after at least two 12-week SRI trials and at least one ERP trial
to 6 weeks of adjunctive olanzapine (n = 13)
or placebo (n = 13). OCD
patients with current co-occurring Axis I disorders were excluded.
The mean olanzapine dose was 11.2 mg/day (SD = 6.5;
range: 5–20 mg/day). In the ITT sample, adjunctive
olanzapine was significantly superior (Y-BOCS reductions: olanzapine
17%; placebo 2%). Six (46%) of 13 olanzapine
patients were YBOCS-25% responders compared with none in
the placebo group. Two olanzapine patients (15%) discontinued
because of the side effects (sedation: n = 1;
weight gain: n = 1).
Shapira et al. (460) randomly assigned OCD nonresponders
(< 25% decrease) or partial responders (YBOCS-25% but
score 16) after 8 weeks of fluoxetine (40 mg/day in 42
subjects, 20 mg/day in 1 subject), to 6 weeks of adjunctive
olanzapine (n = 22), or
placebo (n = 22). Olanzapine
was started at 5 mg/day, and the dose was increased to
a maximum of 10 mg/day. Both treatment groups improved
significantly, with no significant difference between them; the
proportions of responders were similar (YBOCS-25% = 41% for
both groups; YBOCS-35% = 23% for olanzapine,
18% for placebo). The authors concluded that adding olanzapine
was not superior to extending the 8-week fluoxetine monotherapy
trial. However, as they noted, the patients were unlikely to have
attained full benefit from the SSRI before the olanzapine trial began,
thus obscuring any olanzapine effect. Olanzapine patients gained
a mean of 2.8 (± 3.1) kg compared with 0.5 (± 1.8) kg for placebo
The safety and effectiveness of quetiapine augmentation of
SRI treatment in OCD have been evaluated in three randomized, double-blind,
placebo-controlled trials; one randomized, single blind, placebo-controlled
trial; and several open-label studies.
Denys et al. (158) randomly assigned 40 OCD patients
without co-occurring diagnoses who were unresponsive (Y-BOCS decrease
< 25%) after at least two SRI trials (at maximum tolerated
dose for 8 weeks) to receive adjunctive quetiapine (n = 20)
or placebo (n = 20) for
8 weeks. Quetiapine was started at 50 mg/day, and the dose
was increased, following a fixed dosing schedule, to a maximum of
300 mg/day. In the ITT sample, adjunctive quetiapine was
significantly superior to placebo (Y-BOCS reduction: quetiapine
32%; placebo 7%). Eight (40%) quetiapine
patients were responders (defined as YBOCS-35% and CGI-I:1,2),
compared with only two (10%) placebo patients. The most
common side effects of quetiapine were somnolence (95%),
dry mouth (55%), weight gain (30%), and dizziness
Carey et al. (458) randomly assigned 41 patients
who had not responded (defined as not CGI-I:1,2 or not YBOCS-25%)
after 12 weeks of SRI treatment to 6 weeks of flexibly dosed adjunctive
quetiapine (n = 20) or
placebo (n = 21). The
mean final quetiapine dose was 169 (± 121) mg/day. Both
quetiapine and placebo led to a significant reduction in mean Y-BOCS
scores (Y-BOCS reduction: quetiapine 27%; placebo 26%).
Responder (defined as CGI-I:1, 2 and YBOCS-25%) rates were
40% and 48% for quetiapine and placebo, respectively.
Two quetiapine patients dropped out because of severe sedation, and
75% complained of sedation (vs. 33% of placebo
patients). Quetiapine augmentation was no more effective than placebo,
but the study was limited in that patients had received their maximum
SRI dose for only 6 weeks before randomization.
Fineberg et al. (459) randomly assigned 21 adult
OCD patients with minimal response (defined as <25% Y-BOCS decrease)
after 12 weeks of an SRI at the maximum tolerated dose, to receive
either adjunctive quetiapine (n = 11;
mean final dose = 215 mg/day, range = 50–400
mg/day) or placebo (n = 10).
After 16 weeks of augmentation, there was no difference in the ITT
sample between the two groups (Y-BOCS reduction: quetiapine 14%;
placebo 6%). Three of 11 quetiapine-treated patients were
YBOCS-25% responders compared with 1 of 10 placebo patients.
The authors suggested that exclusion of co-occurring Axis I disorders
and tic disorders may have led to their negative findings.
In a single-blind placebo-controlled study (462),
27 OCD patients with no response to at least one 12-week SRI trial
(defined as no more than minimal improvement, a Y-BOCS score of
18, and agreement of three of the authors) were randomly assigned
to receive adjunctive quetiapine (n = 14)
or placebo (n = 13) for
8 weeks. Of the 14 patients randomly assigned to 50–200
mg/day of quetiapine, 10 (71%) experienced improvement
(defined as Y-BOCS decrease 30%), in comparison to none
of the placebo patients. Nine quetiapine-treated patients reported
side effects (nausea, n = 6;
sedation, n = 3; dizziness, n = 1).
The study's main limitation was the single-blind design.
g. Other Antipsychotic Agents
Double-blind, placebo-controlled studies are not available
for ziprasidone or aripiprazole. Published case reports weakly suggest
that ziprasidone augmentation of SRI pharmacotherapy may be effective
(463). A small (n = 8),
open-label, 8-week, flexible-dose study of aripiprazole (10–30
mg/day) monotherapy reported that three subjects (38%)
experienced a 30% or greater decrease in Y-BOCS score (457).
Pindolol, a beta-blocker and serotonin1A (5-HT1A)
presynaptic receptor antagonist, increases serotonergic transmission
through its effect on the presynaptic 5-HT1A receptor.
It has been suggested that pindolol can be given once or twice daily
for augmentation in the treatment of psychiatric disorders (464).
In OCD, small studies have produced mixed results regarding its
possible efficacy as an augmentation agent. An 8-week, double-blind,
placebo-controlled trial examined pindolol augmentation of fluvoxamine
in 15 patients (180). No differences between the two
treatment groups were noted either in symptomatic response or in
the latency of response to fluvoxamine. A double-blind, placebo-controlled
trial enrolled 14 patients with DSM-IV OCD who had not responded
to paroxetine and at least two other SRIs (179). Augmentation
with pindolol 2.5 mg three times daily was associated with significant
decrease in the Y-BOCS score after the fourth week of treatment.
The greatest improvement was noted in the ability to resist compulsions.
No group differences were found in pulse rate or blood pressure. An
open-label study found beneficial therapeutic effects from combining
pindolol and a serotonergic antidepressant, but only after tryptophan
was added (465). Another open-label study found that
one of eight patients with treatment-resistant OCD responded to
pindolol augmentation (466).
A double-blind crossover comparison trial with 6-week drug
periods found clonidine (maximum dose = 1.0 mg/day)
ineffective in 28 patients with DSM-III-R OCD (467).
Evidence for beneficial effects of benzodiazepines as monotherapy
for OCD is primarily limited to case reports with clonazepam and
alprazolam. For clonazepam, negative results from a double-blind,
placebo-controlled trial (468) and an open trial (469)
cast serious doubt on the positive results of a double-blind, multiple-crossover
trial (467). In the latter study, effectiveness seems
to have necessitated doses that were poorly tolerated and produced
serious adverse events. Modest doses of benzodiazepines may relieve
anxiety and distress in OCD without directly diminishing the frequency
or duration of obsessions or compulsions. However, case reports
have noted an onset of action within 1–3 weeks. Among patients with
histories of substance abuse or dependence, benzodiazepine use may
aggravate symptoms and should be prescribed cautiously (234, 255).
Thus, given their limited evidence for efficacy, benzodiazepines
cannot be recommended as monotherapy for OCD, except in those rare
individuals who are unable or unwilling to take standard anti-OCD
Small and methodologically limited studies provide inconsistent
results regarding the possible effectiveness of buspirone 60 mg/day
as monotherapy and no substantial evidence of its effectiveness
as an augmenting agent.
A 6-week double-blind comparison of buspirone titrated to
60 mg/day (n = 10)
and clomipramine titrated to 250 mg/day (n = 10)
suggested equal effectiveness (470). In a 4-week, double-blind,
placebo-controlled, crossover trial (n = 13),
however, buspirone was no better than placebo (471).
In addition, an 8-week open trial of buspirone at a dose of 60 mg/day
for the last 5 weeks of the trial resulted in virtually no decrease
in the mean Y-BOCS score of the 10 subjects who completed the study
(472). However, these trial durations were too short
to allow a robust test of buspirone's possible effectiveness.
In a small (n = 14),
10-week, double-blind, placebo-controlled trial, which involved
a 2-week placebo lead-in followed by 10 weeks of buspirone augmentation,
buspirone did not differ from placebo, although 29% (4/14)
of buspirone subjects experienced a YBOCS-25% response
(473). A 6-week, double-blind, placebo-controlled trial
of buspirone 60 mg/day as an augmentation of fluvoxamine
was negative, but this study assessed the efficacy of buspirone
augmentation in patients with treatment-resistant OCD (<35% decrease
in Y-BOCS score and rated "unimproved" by the
investigators) rather than in partial responders for whom further
improvement was sought (474).
Inositol, a precursor in the phosphatidylinositol cycle, has
been studied in two double-blind, placebo-controlled, crossover
studies (as monotherapy in one and as an augmenting agent in the other)
and an open augmentation trial. Inositol appears to be well tolerated.
In addition, the data weakly suggest that inositol may benefit a
minority of OCD patients but do not support a recommendation that
it be routinely tried.
In a double-blind, crossover study, inositol (18 gm/day)
and placebo were administered for 6 weeks in each condition in 13
subjects free of major depression and with variable responses to SSRIs
(475). There was a small but significantly greater
decrease in the Y-BOCS score in the inositol condition (5.9 points)
compared with the placebo condition (2.8 points). A double-blind,
placebo-controlled, crossover augmentation study with 6 weeks in
each condition and no washout between conditions found no difference
in Y-BOCS score decrease between inositol and placebo in the first
drug condition, but the study groups were small (six inositol and
four placebo) (476). Subjects had been taking a stable
SRI dose for at least 8 weeks before randomization but showed greater
improvement in the study's first 6 weeks than in its second
6 weeks, regardless of which blinded drug was administered first.
In an open study (477), inositol (18 gm/day)
was added for 6 weeks in 10 subjects who had been rated
minimally improved on the CGI-I after 12 weeks or more of stable
SSRI treatment. Mean Y-BOCS scores fell significantly from 23.6
(± 4.4) to 17.6 (± 4.6), but only three subjects (30%)
achieved CGI-I scores of much improved.
Case reports suggest that lithium monotherapy may deserve
further study in trials that utilize an adequate serum level (
0.6 mEq/L) and an adequate duration of treatment ( 10
weeks). However, findings from an 8-week, double-blind, crossover
augmentation study (n = 16)
with a mean serum level of 0.54 mEq/L (478)
and a 4-week, double-blind, placebo-controlled, augmentation trial
(n = 10) with a mean serum
level of 0.77 mEq/L (479) were negative. The
4-week lithium treatment period in these studies may have been too
short to fully evaluate lithium's potential utility as
an augmentation agent in OCD. The utility of lithium in the treatment
of co-occurring bipolar disorder is clearly established (193).
A study combining an open-label first phase with a double-blind
discontinuation phase suggests that mirtazapine may be effective
for OCD in patients who have not received SRI treatment or have
not responded to only one adequate SRI trial (174).
However, the small sample size (15 treatment-naive patients and
15 not responding to exactly one adequate SRI trial) makes these
results suggestive rather than strong evidence for mirtazapine's
effectiveness. Significant weight gain was observed in more than
30% of patients in the first 12 weeks of treatment. A small
pilot study provides slight additional support (480).
Additional double-blind, placebo-controlled trials utilizing a parallel
groups design are indicated.
A case series (481) and a double-blind, placebo-controlled
crossover study (182) suggest that once-weekly oral
morphine sulphate 30–45 mg may be useful as an augmentation strategy
for resistant OCD. In the double-blind study, 7 of 23 subjects (30%)
experienced a YBOCS-25% response to morphine versus none
in the placebo condition. In addition, a small (n = 8),
6-week, open-label study found evidence for the effectiveness of
tramadol monotherapy 254 ± 119 mg/day in the six patients
who completed at least 2 weeks of treatment (183).
The dose-limiting side effect was sedation.
Some anticonvulsants (valproate, oxcarbazapine, carbamazepine,
gabapentin, topiramate) have been reported to help individual patients
either as monotherapy or as augmentation agents. A small (n = 9),
8-week, open-label carbamazepine trial (482) and a
small case series (n = 5)
(483) each reported only one positive response. A 6-week
open-label study of gabapentin augmentation (mean dose = 2520
mg/day) in five patients who had a partial response to
fluoxetine suggested some benefit (484), but a 6-week,
double-blind, placebo-controlled, crossover trial of gabapentin
3600 mg/day added to fluoxetine found no benefit (485).
An open case series reviewing at least 14 weeks of topiramate augmentation
(mean daily dose = 253 mg) in 16 patients who had a partial
response or no response to SRI treatment reported 11 of 16 (68.8%)
CGI-I:1,2 responders. The mean time to response was 9 weeks (486).
l-Tryptophan 3–9 gm/day
combined with nicotinic acid 1 gm two times daily and pyridoxine
200 mg two times daily was reported to be beneficial in early case
reports that predated modern diagnostic criteria and measurement
instruments (487, 488). However, some
patients became violent. Adding l-tryptophan
at doses exceeding 1–2 gm/day to an SRI can induce
the serotonin syndrome.
d-Amphetamine 30 mg, studied
in a single-dose, double-blind, placebo-controlled trial, was associated
with a significant decrease in self-rated symptoms about 6 hours
after the dose, independently of effects on mood (184). d-Amphetamine
had an acute anti-OCD effect in 11 of 12 subjects (92%).
With placebo, neither the self-ratings nor the blinded observer's
ratings decreased significantly. Two patients continued d-amphetamine
at a dose of 10–20 mg/day for "several
weeks" with continued response. In a small (n = 11),
double-blind, placebo-controlled, crossover study of single doses
of methylphenidate 40 mg and d-amphetamine
30 mg, both taken orally, the latter drug was associated with a
significantly greater reduction in OCD symptom rating than was placebo
(185). Five of the 11 subjects (45%) had a
50% decrease in their OCD scores after d-amphetamine,
two (18%) after methylphenidate, and only one (9%)
after placebo. In both studies, the decrease in OCD symptoms was
independent of mood effects. Open-label methylphenidate, 40 mg once
orally, produced no significant effect on OCD or mood 4 hours later
in a small study (n = 13),
although four patients had a 50% decrease in an OCD rating
scale score (489). Case reports exist of OCD benefit
after treating co-occurring attention-deficit disorder with stimulants.
The presence of tics or Tourette's disorder does not contraindicate
the use of stimulants to treat ADHD co-occurring with OCD, although
methylphenidate appears to be better tolerated in this situation
than d-amphetamine (490).
Hallucinogens have been reported to alleviate OCD in individual
cases (491, 492). Since hallucinogens
are not a practical treatment modality or recommended, studies of
safer serotonin2A,C (5-HT2A,C)
receptor agonists may be warranted.
Ondansetron 1 mg three times daily was associated with a significant
decrease in Y-BOCS scores in a small (n = 8),
8-week, open-label study (493).
St. John's wort (450 mg of 0.3% hypericum
two times daily), a weak serotonin-reuptake inhibitor, was associated
with CGI-I:1,2 response in 5 of 12 (42%) subjects in a
12-week open-label trial (494). However, a 12-week,
flexible-dose, placebo-controlled trial enrolling 60 subjects found
St. John's wort to be no better than placebo (181).
In addition, St. John's wort predisposes to photosensitivity
and interacts with anti-HIV medications (495), cyclosporin
(496, 497), and birth control pills (498),
among other medications (499).
Bupropion titrated from 150 mg/day to 300 mg/day
after 2 weeks had no mean effect on Y-BOCS scores in an open trial
involving 12 patients (500). However, 2 patients were
YBOCS-25% responders; 4 patients "improved," with
a mean Y-BOCS decrease of 31%, but 8 patients experienced
a worsening of symptoms, with a mean Y-BOCS increase of 21%.
A 12-week open-label study adding riluzole 50 mg two times
a day to SSRIs and other augmenting medications reported that 7
of 13 (54%) patients with treatment-resistant OCD were
YBOCS-35% responders (501). However, these
results must be viewed cautiously because ratings were not blinded,
other augmenting medications were present, and prior treatment regimens
were stable for only 4 weeks before riluzole was added. Riluzole
was well tolerated, although one patient experienced an asymptomatic
increase in liver enzyme (ALT) to a level more than nine times normal,
which decreased despite continued treatment.
B. Other Somatic Therapies
Somatic therapies used in the treatment of OCD include deep
brain stimulation and other forms of neurosurgery, transcranial
magnetic stimulation, and electroconvulsive therapy. Plasmapheresis has
been investigated only in childhood OCD. None of these therapies
are considered first-line treatments for OCD, and their use is limited
to patients with treatment-resistant OCD.
Specific descriptions of the criteria used to establish treatment
resistance and to determine eligibility for surgical treatments
(stereotactic lesion procedures and DBS) in OCD patients have been published
elsewhere (152, 502–504).
Data regarding treatment of OCD with these somatic therapies
are quite limited, and there is an understandable absence or paucity
of double-blind trials; as a result, no definitive conclusions can be
drawn. The majority of available reports are case series and open-label
trials. Although stereotactic lesion procedures have a more abundant
database (197) than the other somatic therapies in
patients with treatment-resistant or intractable OCD, the cost, irreversibility,
and lack of a clear relationship between specific anatomic lesions
and successful outcomes continue to limit their use.
1. Transcranial Magnetic Stimulation
Findings of the four published trials of repetitive TMS (rTMS)
are inconsistent, perhaps because the studies differed in design,
stimulation sites, duration, and stimulation parameters. The available
results and the technique's non-invasiveness and good tolerability
should encourage future research, but the need for daily treatment
may limit the use of TMS in practice.
In a TMS study of possible frontal lobe involvement in OCD
(505), 12 OCD patients (6 patients with past or current
major depression) were randomly assigned to receive one session
of active right-side or left-side or sham (occipital) rTMS. Blinded
raters made ratings during the stimulation and 30 minutes and 8
hours after the session. Compulsive urges, but not obsessions, decreased
significantly, and positive mood increased moderately, with right
lateral prefrontal rTMS (during stimulation and 30 minutes and 8
hours after stimulation), but not after left rTMS or occipital rTMS.
Stimulation was well tolerated, with two patients reporting mild headache
after stimulation. Although the report was not intended as a treatment
study (only one administration per anatomical site was used), methodological
limitations include the treatment duration, lack of Y-BOCS outcome
measures, small sample size, and the lack of a control group. Another
non-treatment TMS study reported altered cortical excitability in
subjects with OCD compared with controls (506).
A 6-week trial (three sessions per week) (507)
found no advantage for low-frequency right rTMS (n = 10)
over sham stimulation (n = 8).
No dropouts were reported, and side effects, consisting of headache
and cognitive difficulties, were modest and transient. The use of
a relatively nonfocal, teardrop-shaped coil limits the interpretation
of the results.
In a 10-session, single-blind, 2-week trial (508),
12 patients with treatment-resistant OCD were randomly assigned
to receive active right-side or left-side rTMS. Ten subjects taking
medications were maintained on a constant dose for 8 weeks prior
to and during the study. Evaluations after 2 weeks of stimulation
and 4 weeks later showed significant reduction in obsessions and
compulsions in both groups, with no significant difference between
right and left stimulation. Four patients (two receiving left and
two receiving right rTMS) had a clinically significant improvement
(Y-BOCS reduction >40%); one patient relapsed but responded
somewhat to repeat treatment. No dropouts were reported, and stimulation
was well tolerated, with three patients reporting headache. However,
interpretation is limited by the absence of a placebo control group
and the presence of concurrent pharmacotherapy.
In an open-label trial (509), 10 treatment-resistant
patients (5 with OCD, 3 with Tourette's syndrome, and two
with both) received low-frequency rTMS for 2 weeks. Medication doses
were stable for at least 12 weeks before and throughout rTMS and
the follow-up period. Eight patients completed the study; no dropouts
due to side effects were reported. CGI scores decreased significantly
at the end of the first and second weeks of treatment, with benefit
maintained at the 1-month and 3-month follow-ups. Three of the five
patients with pure OCD had a clinically significant improvement,
with a > 40% reduction in Y-BOCS scores, and two Tourette's
patients had a complete remission at the second week. Six subjects
(60%) had clinical improvement that persisted at 3-month
follow-up. The study was limited by the open design and small sample
2. Electroconvulsive Therapy
The literature on ECT in treatment-resistant OCD includes
a case series and several individual cases, with some reported degree
of effectiveness. However, the frequent Axis I comorbidity among
subjects, lack of standard outcome measures, absence of blinded
trials, need for repeated anesthesia, and the side effects of ECT
preclude it from being considered for treatment-resistant OCD uncomplicated
by co-occurring conditions (191).
The case series (190) describes a retrospective
study of 32 patients with DSM-III-R treatment-resistant OCD (19
not depressed and 13 depressed; 14 with primarily checking rituals,
13 with primarily cleaning rituals, and four with both) who received
ECT between 1979 and 1991. The patients were treated with bilateral
frontotemporal ECT (average three to five seizures per session over
2–3 weeks) and were evaluated 2 days before treatment and
at 5 days and 6 and 12 months after the end of treatment. Comparison
of baseline scores (Beck Hopelessness Scale, Maudsley Obsessional-Compulsive
Inventory) with data 5 days after treatment yielded highly significant
pre-post paired t test results
(P < 0.001) that were still
significant at 6 months posttreatment. However, by 12 months only
the MOCI scores remained significantly different from those prior
to ECT. In addition, methodological limitations include the absence
of blinded ratings and standard outcome measures, the use of medications
during the long-term follow-up, and the presence of co-occurring
depression in a significant proportion of patients (13/32).
Furthermore, the number of seizures per session is not considered
standard ECT treatment (191).
In addition, several single case reports (510–514)
suggest possible efficacy of ECT in treatment-resistant OCD. However,
the unblinded ratings, frequent Axis I comorbidity (schizophrenia,
depression, Tourette's disorder), and differing ECT parameters
limit the confidence that can be placed in the findings.
3. Deep Brain Stimulation
Two small, double-blind trials and several case reports have
investigated the efficacy of DBS in OCD. Given the preliminary promising
results in treatment-resistant OCD, the procedure's reversibility
and adjustability in comparison with ablative neurosurgery, and
the absence to date of serious adverse events, DBS deserves investigation
in severe, treatment-resistant OCD. Nonetheless, DBS is an invasive
procedure, and the risks of brain hemorrhage, infection, and new
onset of seizures must be kept in mind (195).
The first report (515, 516) concerns
six subjects with OCD refractory to various antiobsessional treatments
and to CBT (Y-BOCS 30, GAF 45,
both for a minimum of 5 years) who were treated with DBS. Quadripolar
electrodes were implanted bilaterally in the anterior limbs of the
internal capsules with a double-blind stimulation-off condition
representing the placebo condition throughout the 21 months of evaluation.
Four patients completed a blinded crossover trial (i.e., stimulator
on for 3 months and stimulator off for 3 months, or vice versa,
in random order). Three were YBOCS-35% responders during
the stimulation-on condition, with CGI-I scores much improved. Responders
reported clinically meaningful improvement in the first week of
stimulation. Side effects included fatigue and memory disturbances.
The persistence of the effect for at least 21 months argues against
a placebo effect. Patient 5 received a different electrode placement,
one in each dorsomedial thalamic nucleus and one in each internal
capsule (517). The dorsomedial thalamic nucleus did
not seem as effective a target for this patient, who was also considered
a nonresponder with longer-term internal capsule stimulation. Patient
6 experienced a major improvement (more than 50% decrease
in postoperative tests) in his aggressive, intrusive thoughts and
mood when stimulation was turned on.
One small trial enrolled four subjects with DSM-IV OCD who
did not respond to at least four antiobsessional medications and
CBT (Y-BOCS 25; GAF 44)
(518). The patients, who had been following stable
medication regimens for at least 6 weeks before surgery and during
the blinded phase of the study, received DBS with quadripolar electrodes placed
stereotactically in the anterior limb of each internal capsule;
stimulation-off was the control. The double-blind study consisted
of four consecutive 3-week periods (in an alternating on-off design),
followed by an open phase in which stimulation, medication, and
CBT were adjusted to optimize response for up to 1 year. During
the double-blind phase, two of the four patients showed clinically
meaningful responses but one of these patients also responded during
periods without stimulation. One subject experienced mood elevation
in response to stimulation. Positron emission tomography (PET) scans
showed orbitofrontal deactivation only in the two patients with a
positive clinical response. Side effects included tingling, nausea,
and diarrhea. Beneficial effects appeared over variable times, ranging
from 3 weeks to several months.
In addition, four recent case reports of open-label DBS (519–522)
suggest the efficacy of DBS in treatment-resistant patients with
OCD. In a recent case-series report (523), three of four patients
with severe OCD and anxiety disorders who received DBS in the shell
of the right nucleus accumbens experienced significant reduction
in severity of symptoms.
4. Neurosurgical Stereotactic Lesion
Neurosurgical treatment for psychiatric disorders has a long
and controversial history. Today this approachincluding
cingulotomy, capsulotomy (also performed via radiosurgery and known
as "gamma-knife" capsulotomy), subcaudate tractotomy,
limbic leucotomy, central lateral thalamotomy/anterior
medial pallidotomyis a highly selective treatment performed
for relatively few patients with severe, treatment-refractory affective,
anxiety, or obsessive-compulsive disorders (197). The
availability of reversible and adjustable DBS may lead to a decrease
in the use of ablative neurosurgical procedures. However, these
procedures still represent a potentially efficacious alternative
for a few carefully selected patients with very severe OCD.
In view of the changes in neurosurgical techniques in the
last decade, only trials using these advanced techniques with adequate
numbers of patients and specified inclusion criteria and outcome
measures are reviewed.
A prospective unblinded study of bilateral anterior capsulotomy
in 15 subjects with treatment-refractory OCD (mean duration = 18.1
± 5.6 years; mean total Y-BOCS score = 29.7) observed positive
results at 1 and 12 months postsurgery: at 12 months, the mean Y-BOCS
score decrease was 39%, with 53%, 29%,
and 17% of the patients experiencing a 33% decrease,
50% decrease, and 66% decrease in Y-BOCS score,
respectively (524). No cognitive deficit was evident
in neuropsychological screening tests. Complications were observed
in three case subjects: one with transitory hallucinations, one
with a single epileptic seizure, and one who developed a progressive
behavior disorder (not further specified) that became permanent.
In an unblinded study (525) enrolling 44 subjects
with DSM-III-R OCD refractory to at least three SRIs trials, at
least one SRI augmentation trial, and a trial of CBT (Y-BOCS 25,
GAF 60, both for 5
years), one or more cingulotomies (electrodes positioned in each
cingulate gyrus with magnetic resonance imaging [MRI] stereotactic
guidance) were performed. Follow-up evaluations were carried out
a mean of 32 months after neurosurgery. At the first follow-up (a
mean of 6.7 months after the first cingulotomy), the mean Y-BOCS
score decrease was 20%; 5 patients (11%) met full
responder criteria (YBOCS-35% and CGI-I:1,2),
and 4 patients (9%) met partial responder criteria (either
YBOCS-35% or CGI-I:1,2 or this
degree of improvement not attributed to cingulotomy). At the 32-month
follow-up, the mean Y-BOCS score decrease was 29%, with
32% of patients classified as responders and 14% as
partial responders. At the most recent follow-up for the 18 patients
who received multiple cingulotomies, 5 (28%) were responders
and 2 (11%) were partial responders. These results suggest
that cingulotomy may benefit some patients with severe treatment-refractory
OCD. Although 20% of patients reported at least one adverse
effect after cingulotomy (e.g., memory disturbances, apathy, urinary
disturbances), only 2 patients (5%) reported enduring sequelae
(i.e., seizure disorder and hydrocephalus).
Fourteen subjects with treatment-refractory and medically
intractable OCD were evaluated up to 12 months after bilateral anterior
cingulotomy (526). At the 6- and 12-month follow-ups,
the mean Y-BOCS scores decreased 29% and 36%,
respectively, with 4 of 14 (29%) and 6 of 14 (43%)
having a response (defined as YBOCS-35% and CGI-I:1,2).
No significant changes in cognitive functions or memory were reported
at the 12-month evaluation compared with preoperative scores. Adverse
effects (headache, insomnia, weight-gain/loss) persisted
no more than 3 months after cingulotomy.
In a retrospective unblinded study, 15 subjects with treatment-refractory
OCD were followed up for approximately 1 year after bilateral cingulotomy
(527). Four of the 15 patients were YBOCS-35% responders.
However, only one had sustained benefit lasting more than 1 year.
Minor postoperative symptoms included headache, nausea/ vomiting, and
urinary incontinence, which resolved after several months. Postoperative
MRIs revealed no clear relationship between lesion location and
significant clinical improvement.
The efficacy of limbic leucotomy was evaluated in a small
unblinded study of 21 subjects, of whom 15 had treatment-refractory
OCD and 6 had refractory major depression (528). Patients
were evaluated after a mean of 26 months. Five of 12 OCD patients
(42%) with physician ratings were rated responders (defined
as physician-rated CGI-I:1,2), and 8 of the 13 (62%) OCD
patients with self-ratings rated themselves as responders (defined
as PGI-I:1,2). One OCD patient, who had a history of a suicide attempt,
died by suicide after the surgical procedure. Minor postoperative
symptoms of headache, low-grade fever, and nausea/vomiting
were common but generally lasted less than 48 hours. Among the 21
subjects, transient postoperative somnolence and apathy were noted
in 29% and 24% of patients, respectively. These
results are comparable to those in the limited previous reports
regarding limbic leucotomy in intractable OCD (529).
1. Exposure and Response Prevention
Historically, CBT for OCD has been divided into two forms:
1) CBT that relies primarily on behavioral techniques, such as ERP
(57), and 2) CBT that relies primarily on cognitive
therapy techniques, such as identifying, challenging, and modifying
faulty beliefs (530, 531). Certain variants
of exposure therapy routinely include some informal cognitive therapy
techniques (e.g., a discussion of fear-related thoughts and beliefs),
and many variants of cognitive therapy include behavioral experiments,
which can be similar to exposure techniques. Therefore, these two
forms of CBT, as administered in treatment trials, often overlap.
In clinical practice, these two forms are often combined.
Studies that have examined CBT consisting of ERP for adults
with OCD are reviewed in this section (see also reference 532).
a. Randomized Controlled Trials Comparing
ERP With a Nonactive Treatment
Several randomized controlled trials have
examined whether ERP is superior to a nonactive treatment. Although
the studies used different variants of ERP (e.g., therapist-supervised
or self-controlled exposures), different formats (e.g., individual
sessions vs. group therapy), different intensity (e.g., frequency
and length of sessions per week), and different control groups,
all studies concluded that ERP is efficacious for the treatment
of adults with OCD.
Several early controlled studies reported that ERP was superior
to progressive muscle relaxation (PMR) (e.g., see references 533, 534).
These studies were limited by their small samples and the lack of
standard diagnostic and outcome measures.
A 12-week trial compared the effectiveness of random-assignment
individual ERP (n = 31),
group ERP (n = 30), or
PMR (n = 32), which was
included as an attention control condition (132). Subjects
were free of major depression and Axis II disorders and were treatment
naive. Individual ERP was delivered in 1-hour sessions two times
a week; group ERP consisted of 2-hour sessions two times a week
delivered in groups of 10. In both cases, treatment consisted of
exposure (imaginal and in vivo) and response prevention. In those
subjects who completed treatment, both individual and group ERP
were superior to PMR but not to each other in reducing OCD and depressive
symptoms (Y-BOCS score reduction in subjects who completed the study:
40% for individual; 46% for group; 9% for
PMR). However, group ERP took much less staff time. Patients receiving
either individual or group treatment maintained their gains at 6-month
follow-up; however, whether patients received additional treatment
during follow-up is unclear.
In a 3-week trial in which all patients completed the study,
18 patients were randomly assigned to receive either ERP (n = 9)
or anxiety management (n = 9)
(535). Both treatments involved about 15 hours of therapy.
ERP consisted of graded exposure to feared situations and ritual
prevention both in sessions and as homework. Anxiety management consisted
of breathing retraining, PMR, and problem solving. ERP was significantly
superior to anxiety management for both OCD and depressive symptoms
(e.g., Y-BOCS: 62% reduction for ERP; 6% increase
for anxiety management). During the study, five patients continued
SRIs that had brought "no improvement" for at
least 12 months. Study limitations include the small sample, lack
of independent raters, and very short-term duration of observation.
In a large multisite trial (139), 218 patients
were randomly assigned to 10 weeks of ERP guided by a therapist
(n = 69), ERP guided by
a computer and workbook (n = 74),
or systematic relaxation guided by an audiotape and manual (n = 75).
ERP guided by a therapist consisted of 11 weekly 1-hour sessions
in which therapists told patients how to conduct exposures at home
but did not supervise in-session exposures. ERP guided by computer
used BT STEPS, a nine-step computer-driven interactive voice response
system that allows patients to telephone from home at any time of
day or night and progress through a self-paced workbook that allows
the patient to design and implement an individualized plan of behavior
therapy. Relaxation therapy consisted of an audiotape and manual
that directed the patient to practice PMR 1 hour daily. Of those
who entered, 55 (80%) completed ERP, 55 (74%) completed
BT STEPS, and 66 (88%) completed PMR. In ITT analyses,
ERP was superior to BT STEPS, and both were superior to PMR (Y-BOCS
reduction: 32% for ERP; 23% for BT STEPS; 7% for
PMR). Significantly more patients receiving ERP were CGI-I:1,2 responders
(60%), compared with patients receiving either BT STEPS
(38%) or PMR (14%). Patients who adhered to ERP
(either guided by clinician or computer) had a larger mean decrease
in their symptoms than the overall mean decrease for that condition.
Patients were followed for an additional 14 weeks. However, the
groups received different follow-up treatment, so overall group
comparisons were confounded. Of note, nonresponders to BT STEPS
who were switched to clinician-guided ERP had a significant decrease
in OCD severity, whereas nonresponders to ERP who were switched
to BT STEPS did not. The authors concluded that minimal ERP (i.e.,
11 weekly 1-hour sessions in which a therapist provides instructions
only) is superior to BT STEPS and that PMR is ineffective for OCD.
However, the reduction in OCD symptoms was substantially less than
that seen in other studies in which therapists supervised in-session
exposures (e.g., compare ITT results in Lindsay et al.  and
Foa et al. ).
In a Brazilian controlled trial of group ERP (134),
OCD patients were randomly assigned to 12 weeks of either group
ERP (n = 23) or wait-list
control (n = 24). Although
45% of patients were taking medications at the time of
therapy, they had been taking the medication at a stable dose for
at least 3 months. The group ERP was conducted in weekly 2-hour
sessions in groups of seven or eight; it included in-session exposure,
ritual prevention, and cognitive restructuring. In ITT analyses,
group ERP was significantly superior to the wait-list condition
(Y-BOCS reduction: group ERP 43%; wait-list 6%);
70% of group ERP patients were YBOCS-35% responders.
Twenty-two ERP patients reevaluated after 3 months had maintained
their gains. However, whether patients received additional treatment
during follow-up was not stated.
A multisite trial (123) compared the efficacy
of 12 weeks of ERP, clomipramine, their combination, and pill placebo
in adults with OCD and without co-occurring depression. Of 122 entering
randomly assigned treatment, 87 (71%) completed the study.
ERP was delivered intensively for the first 4 weeks (i.e., five 2-hour
sessions per week); it consisted of exposure (imaginal and in vivo),
ritual prevention, and relapse prevention. During exposures, feared
consequences and dysfunctional beliefs were discussed. For the remaining
8 weeks, patients received 45-minute maintenance sessions in which
no in-session exposures were conducted. Patients randomly assigned
to receive clomipramine or placebo were seen for 30 minutes
weekly by a research psychiatrist. They received clomipramine titrated
up to 200 mg/day or placebo in the first 5 weeks, with
an optional increase to 250 mg/day if needed. In ITT and
completer analyses, all active treatments were superior to placebo.
ITT and completer response rates (CGI-I:1,2) were, respectively,
62% and 86% for intensive ERP, 42% and
48% for clomipramine, 70% and 79% for
combination treatment, and 8% and 10% for placebo.
Post hoc analyses (126) demonstrated that intensive
ERP with or without clomipramine produced a significantly higher proportion
of patients (> 50%) with minimal symptoms (i.e., Y-BOCS 12)
than either placebo (0%) or clomipramine alone (25%).
However, high study refusal rates (71% of those meeting
entry criteria refused to participate) and dropout rates (18% at
randomization and an additional 23% before completion)
may limit the widespread applicability of this study's
In a Japanese study (536), 31 OCD outpatients
were randomly assigned to 12 weeks of single-blind ERP plus pill
placebo; fluvoxamine 150–200 mg/day (dose reached
no later than week 5) plus autogenic training (psychotherapy placebo);
or autogenic training and placebo. Among the 28 patients (90%)
who completed treatment, ERP was significantly more effective than
fluvoxamine in reducing Y-BOCS scores. Ten of 10 (100%)
patients receiving ERP were Y-BOCS-35% and CGI:1,2 responders compared
with 3 of 10 (30%) fluvoxamine patients and 0 of 8 (0%)
placebo patients. The study is limited by small sample size, lack
of double-blind ratings, and low doses of fluvoxamine.
In addition to these randomized controlled trials, several
controlled trials compared ERP and different variants of cognitive
therapy with and without SRI medication. These studies are reviewed in
the section on cognitive therapy for OCD (Section V.C.2). Overall,
these studies also found that ERP significantly reduced OCD symptoms,
further supporting its efficacy for OCD. However, most of these
studies lacked a placebo or nonactive control condition. Two used
a wait-list control, with one (61) reporting that gradual
self-controlled ERP was significantly superior to 8 weeks of wait-list
control, and the other (133) finding that group ERP
was superior to wait-list control.
Open trials of ERP (totaling hundreds of patients) also support
the efficacy of ERP for OCD (e.g., see references 229, 537–539).
b. Factors That Affect Outcome From
Several factors appear to affect the outcome of ERP. These
include patient adherence to the ERP procedures (540, 541),
the patient's degree of insight into the irrationality
of his or her fears (with poor insight leading to worse outcome
in some [208, 542, 543],
but not all , studies), and
certain co-occurring conditions. For example, severe depression
and some co-occurring anxiety disorders (e.g., posttraumatic stress
disorder, generalized anxiety disorder) (233, 544, 545)
may negatively impact outcome. Other predictors of treatment outcome
have also been reported, but none is sufficiently accurate to apply
in the individual case (532, 546). Because
some data suggest that certain subtypes of OCD patients (e.g., patients
with severe hoarding behavior, patients without overt compulsions)
may not benefit as much from ERP, modifications of standard ERP
for these subtypes are being investigated (e.g., see references 121, 150).
Whether the addition of formal cognitive therapy elements can improve
the outcome from ERP (either for all OCD patients or for patients
with co-occurring conditions such as depression) is also under investigation
(Sections II.B.1, II.B.3, II.B.4, and V.C.2).
Some data suggest that ERP is most effective when delivered
intensively (e.g., five sessions per week [547, 548]).
On the other hand, one study found that twice-weekly ERP was comparable
to intensive treatment (141), and good results have
been achieved with weekly, 1-hour sessions (63). To
date, no study has compared once- or twice-weekly treatment with
intensive treatment in a randomized controlled design.
Whether the effects of ERP achieved in randomized controlled
trials can be reproduced in routine clinical practice remains unclear.
However, several large case series found that ERP (even when delivered
weekly) can produce robust effects in fee-for-service settings,
as long as the therapists are skilled (or supervised by those skilled)
in ERP for OCD (537, 539).
c. Long-Term Outcome From ERP
Studies examining the long-term outcome of
adult OCD patients after ERP have generally concluded that most
patients remain treatment responders at follow-up (67, 132–134, 165, 202, 549–552).
However, these findings are inconclusive for reasons that include
design limitations in some studies (e.g., uncontrolled studies and/or
naturalistic follow-up), methodological limitations in others (e.g.,
lack of standardized assessment instruments and/or blind
ratings), and/or inconsistencies in whether additional
treatment was received during follow-up. Recently, Simpson et al.
(67) examined the posttreatment effects of intensive
ERP with relapse prevention after sustained treatment discontinuation,
using evaluators blind to original treatment assignment. Twelve
weeks after treatment discontinuation, the relapse rate was significantly
lower, and the time to relapse was significantly longer, for ERP
responders (with or without concomitant clomipramine; n = 33;
12% relapse rate) than for responders to clomipramine alone
(n = 11; 45% relapse
rate). Limitations include the small sample size and short period
of observation after treatment discontinuation.
Incorporating relapse prevention procedures into exposure
therapy appears to improve the long-term outcome of ERP. A small
randomized trial (124) examined whether ERP with (n = 10)
or without (n = 10) relapse
prevention produced different outcomes. Both groups received 3 weeks
of intensive ERP (i.e., 15 daily sessions with 45 minutes devoted
to imaginal exposure and 45 minutes to in vivo exposure sessions)
followed by four 90-minute sessions of either relapse prevention
(i.e., discussion of stressors likely to trigger OCD, meeting with
a significant other to discuss maintenance of gains, and cognitive
restructuring) or associative therapy (i.e., free association about OCD
symptoms) combined with progressive muscle relaxation (AT-PMR).
Those subjects receiving relapse prevention also received nine 15-minute
phone calls over 12 weeks of follow-up. Outcome was evaluated
after intensive ERP with relapse prevention and after 6 months of
follow-up. Both groups had dramatic decreases in OCD symptoms after
intensive ERP, without a significant difference (Y-BOCS decreases
in subjects who completed the study: 66% for relapse prevention; 60% for
AT-PMR). However, at 6-month follow-up the relapse prevention group
showed significantly lower relapse rates than the AT-PMR group on
most measures and a trend in this direction on the Y-BOCS. The authors
concluded that relapse prevention helps patients maintain gains
from ERP. Relapse prevention techniques are part of some standard
ERP protocols (142).
d. Cognitive-Behavioral Therapy as
an Augmentor of SRI Response
Data from open trials (i.e., Simpson et al.  [n = 6],
Tolin et al.  [n = 20])
and a completed randomized trial (i.e., Tenneij et al.  [n = 96])
indicate that CBT consisting of ERP can successfully augment a partial
response to an adequate SRI trial. A small (n = 14)
open trial (553) also suggests that the addition of CBT consisting
of exposure, response prevention, and cognitive therapy can help
SRI nonresponders, although this trial lacked a control group continuing
on medication alone. No data are available regarding the effect
of adding cognitive therapy elements alone in attempts to augment
This section reviews studies that have examined the efficacy
of cognitive therapy (CT) in the treatment of adults with OCD. For
the purposes of this review, CT includes those variants of CBT that
rely primarily on cognitive therapy techniques as described in Section
V.C.1. The studies address three issues: whether CT without ERP
is effective, whether CT is as effective as ERP and/or
medication, and whether the addition of cognitive procedures to
ERP leads to a better outcome.
a. Efficacy of Cognitive Therapy
A small randomized trial involving patients who had previously
failed to respond to CBT consisting of ERP compared the efficacy
of group cognitive therapy to a wait-list control for OCD patients
with contamination concerns (554). Eleven patients
received 9 weeks of Danger Ideation Reduction Therapy (DIRT), which
consisted of eight 1-hour group therapy sessions and included cognitive
restructuring, expert testimony and corrective information, attentional
focusing, DIRT proscribed exposure, response prevention, and behavioral
experiments. Although DIRT led to significant greater changes than the
wait-list control on several self-report OCD and depression measures,
the effects were small (e.g., a 20% mean reduction in OCD
symptoms on the MOCI). Study limitations include the small sample
size, self-report measures, and use of a wait-list control.
An open trial enrolling 15 patients (555) found
that individual CT in the absence of prolonged exposure led to significant
improvement on several self-report OCD and depression measures (including
the Y-BOCS, Obsessional Beliefs Questionnaire, and BDI). CT consisted
of 14 weekly 60-minute sessions and included psychoeducation, CT
procedures following Beck's method, and relapse prevention.
Behavioral experiments were used to test and correct a patient's
belief but did not involve prolonged exposure.
In sum, there is limited evidence to support the efficacy
of pure CT (i.e., cognitive restructuring without exposure or "behavioral
experiments") in the treatment of OCD.
b. Efficacy of Cognitive Therapy
Most direct comparisons of CT and ERP have found that these
treatments produce similar results. However, strong conclusions
are difficult to reach for a number of reasons. First, some variants
of ERP include informal cognitive techniques (e.g., relapse prevention,
cognitive restructuring during exposures) and some forms of cognitive
therapy include informal exposures (i.e., behavioral experiments [behaviors
that test the validity of the patient's obsessional beliefs,
e.g., "If I think about a disease, my daughter will get
it."]), blurring the distinction between these
treatments. Second, the published studies have differed in their
designs and treatment procedures (e.g., duration and frequency of
treatment sessions). Third, only some studies formally monitored
therapist adherence to the treatment protocols. Fourth, some studies
had limited power to detect differences. Fifth, no study included
a placebo group other than a wait-list control. Therefore, the treatment
recommendations gleaned from these studies are necessarily specific
to the procedures used and limited by these methodological weaknesses.
Together, the data suggest that CT that includes behavioral experiments
has similar efficacy to ERP based solely on habituation (i.e., without
discussion of feared consequences or dysfunctional beliefs).
Two early studies (556, 557) examined
the efficacy of rational-emotive therapy (RET) based on the work
of Ellis (558); this CT program included the identification
and challenge of irrational beliefs but no behavioral experiments.
Both studies found that therapist-administered RET helped OCD patients
and was similar in efficacy to self-controlled ERP. One study (557)
concluded that RET followed by self-controlled ERP was no better
than self-controlled ERP alone. However, both studies had small
samples (i.e., n
per condition) and used a less than optimal ERP format (e.g., self-controlled
as opposed to therapist-guided exposure). Another small open trial
in six subjects compared RET with ERP and thought stopping and found
that RET was more effective than ERP for purely obsessional patients who
had no covert rituals, and that thought stopping was not helpful
at all (151).
A larger randomized study by van Oppen et al. (559)
compared outcomes in 28 patients who received 16 weeks of CT based
on the model of Beck (530) and Salkovskis (531),
and 29 patients who received self-controlled ERP. This CT program
targeted dysfunctional beliefs considered to be central to OCD (i.e.,
the overestimation of danger and inflated personal responsibility),
and specifically included behavioral experiments. For both conditions,
therapy was delivered weekly in 45-minute sessions. Among the subjects
who completed the trial, both treatments led to significant and
clinically meaningful improvement in OCD. Although there were no
significant group differences, CT appeared somewhat superior (Y-BOCS
reduction: 45% for CT; 32% for ERP). However,
a less than optimal ERP format (i.e., self-controlled exposure)
was utilized, and the CT program used behavioral experiments. Half
the patients in this study participated in a multicenter trial comparing
fluvoxamine, ERP, and CT, described below.
Van Balkom et al. (61) randomly assigned 117
patients to one of five conditions: fluvoxamine plus ERP, ERP, CT,
fluvoxamine plus CT, or wait-list control. At baseline, patients
on average had mild depressive symptoms. Full results are presented
for the 70 patients who completed 16 weeks of active treatment
and the 16 patients who completed the 8-week wait-list condition.
The mean dose at week 16 in both fluvoxamine conditions was 197
(± 82) mg/day. All therapy sessions were 45 minutes long.
CT and ERP were delivered as in the van Oppen study (559)
described above. CT targeted dysfunctional OCD beliefs and included
behavioral experiments. ERP consisted of gradual self-controlled
exposure in vivo with gradual self-imposed response prevention.
At week 16, all active treatments led to a significant decrease
on all OCD measures, with no significant differences between the
treatment groups (Y-BOCS reduction for subjects who completed the
trial: 46% for CT [n = 19];
32% for ERP [n = 19];
43% for fluvoxamine plus CT [n = 14];
49% for fluvoxamine plus ERP [n = 18]).
The authors concluded there was no reason to combine SRIs and CBT
(either CT or ERP) in OCD adults without severe co-occurring mood
disorder. However, neither the ERP nor the fluvoxamine treatments
were optimized. Moreover, the combination groups received only 10
therapy sessions that started after 8 weeks of fluvoxamine treatment,
whereas the groups receiving ERP or CT alone received 16 therapy
sessions. Some of these patients were followed naturalistically
for 6 months, but the fact that they received varied treatment during
this follow-up period precludes strong conclusions (549).
A French multisite randomized trial (62) compared
the outcome of patients who received 20 hours of either CT (n = 32)
or ERP (n = 33) over 16
weeks. CT treatment consisted of twenty 1-hour individual sessions
following the Beck and Salkovskis model (i.e., challenging of dysfunctional
beliefs); behavioral experiments to "confront feared situations
to modify thoughts" were also used. ERP consisting of therapist-aided exposure
with response prevention was delivered in an intensive phase (4
weeks of two 2-hour individual sessions per week) and a maintenance
phase (12 weeks of one 40-minute booster session every 2 weeks).
In those patients who completed treatment, both treatments led to
significant and substantial reductions in OCD symptoms (Y-BOCS reduction:
44% for CT; 42% for ERP), and many patients were
YBOCS-25% responders (77% for CT; 70% for
ERP). There were no significant group differences. Patients were
followed after treatment to week 52, but 26% of the patients
were lost to follow-up, and whether those followed received additional
treatment is unclear.
In a Canadian randomized trial (133), OCD patients
(48% who were taking a stable dose of medication and 50% of
whom had a co-occurring Axis I disorder) were randomly assigned
to 12 weeks of CT (n = 18),
ERP (n = 16), or wait-list
control (n = 33). Patients
in the wait-list condition were randomly assigned to receive CT
or ERP after the 12-week delay, and their data were pooled with
the data from those who received active treatment initially. Treatment
was delivered in groups and consisted of 2.5-hour sessions delivered
once per week. Based on the work of Salkovskis (560),
Freeston et al. (64), and van Oppen and Arntz (561),
the CT focused on challenging appraisals of intrusive thoughts;
behavioral experiments were used to collect evidence for and against
alternative appraisals. ERP consisted of in-session and between-session
exposure and ritual prevention focused on producing habituation,
and relapse prevention; discussion of cognitive beliefs was proscribed.
In those patients who completed treatment, both treatments were
superior to wait-list control; however, group ERP was superior to group
CT in the pooled sample (Y-BOCS decrease: 26% for CT; 39% for
ERP). Of the 63 patients who completed treatment, 16% of
CT patients and 38% of ERP patients recovered (defined
as a Y-BOCS score decrease 6 points and a total score < 12).
Of note, 12 of 49 CT and 2 of 44 ERP patients dropped out of the
study after learning of their randomization and before starting
treatment. Moreover, nearly twice as many patients taking medication
received ERP than received CT. At 3-month follow-up, there was still
a significant advantage for ERP over CT. However, these data are
limited by the fact that patients were not prohibited from obtaining
treatment during follow-up.
In another Canadian controlled trial (63), patients
were randomly assigned to receive individual CT (n = 37)
or ERP (n = 34). Both
treatments consisted of 12 weekly 60-minute sessions and followed
the same format as in the study of McLean et al. (133)
described above. Specifically, the CT focused on challenging dysfunctional
beliefs but included behavioral experiments; the ERP included relapse
prevention but no cognitive restructuring. In those who completed
treatment (n = 59, 83%),
OCD severity improved significantly in both treatment groups, with
no significant group differences (Y-BOCS score reduction: 56% for
CT; 52% for ERP). Of those patients who completed the study,
67% of CT and 59% of ERP patients recovered (defined
as a Y-BOCS score decrease 6 points and a total score < 12).
Both groups also had a significant decrease in depressive symptoms
and in dysfunctional beliefs. The authors concluded that individual
ERP utilizing only habituation was similar in efficacy to individual
CT that included behavioral experiments, and that both treatments
reduce OCD symptoms. Additional CT techniques without exposure are described
in case reports (129, 130).
c. Adding Cognitive Therapy to ERP
Some data suggest that ERP is more effective
if it includes not only habituation but also discussion of feared
consequences and dysfunctional beliefs (120, 121).
One small study suggests that the addition of relapse prevention
helps patients maintain their gains (124). Whether
CT added to ERP helps OCD patients with co-occurring conditions
(e.g., depression) is under investigation.
An early study (562) found that both ERP and
ERP plus "self-instructional" talk (i.e., emitting
more positive statements after imagining being exposed to some feared
stimulus) led to clinically significant improvement. However, the
small sample (n = 8 vs. n = 7)
and the uncertain diagnostic validity of the sample (given the lack
of standard criteria at that time) preclude strong conclusions.
A Norwegian randomized trial (122) examined whether
the ERP protocol of Kozak and Foa (142), which includes
discussion of feared consequences and dysfunctional beliefs during
exposures, could be improved by the addition of formal CT elements
based on Beck's work (530). Patients were
randomly assigned to receive ERP plus CT, ERP plus relaxation training,
or wait-list control for 6 weeks. After 6 weeks, those assigned
to the wait-list control were randomly assigned to receive one of
the active treatments. The final sample consisted of 16 patients
who received ERP plus CT and 19 patients who received ERP plus relaxation
training; 34% were taking a stable dose of medication during
the trial. Therapy consisted of 2-hour sessions twice weekly, for
a total of 12 sessions. Each session consisted of 1.5 hours of ERP
and 30 minutes of either CT (based on the Beck model and focused
on either co-occurring conditions or faulty OCD beliefs) or relaxation
training (progressive muscle relaxation). ITT analysis indicated
that both active groups did significantly better than wait-list
controls, with no differences between the two active conditions
(Y-BOCS reduction: 33% for ERP plus CT; 28% for
ERP plus relaxation training). However, more patients completed
ERP plus CT (15/16) than ERP plus relaxation (12/19). The
authors concluded that the addition of CT to ERP (delivered as outlined
in Kozak and Foa ) may reduce the dropout
rate but does not necessarily enhance efficacy. However, the sample
was relatively small, and there were other methodological limitations
(e.g., wait-list design, many ratings done by the therapist, more
exposure time for patients who received ERP plus relaxation training).
Recent studies have also examined whether CT added to ERP
can improve outcome (120, 122).
3. Group and Multifamily Behavioral
A limited number of studies have investigated group and multifamily
behavioral treatments for OCD.
A 12-week unblinded random-allocation study (132)
compared group (n = 30)
and individual ERP (n = 31)
with an active control (progressive muscle relaxation; n = 32).
Patients with major depression or Axis II disorders were excluded.
Both active treatments were superior to the control treatment, as
reflected in changes in Y-BOCS, BDI, and Social Adjustment Scale
scores. However, response to treatment was faster with individual
behavior therapy. The authors concluded that group therapy ERP was
useful for less severe OCD. From an efficiency standpoint, the individual
treatment consumed 720 staff hours compared with 48 hours in the
group therapy condition.
A randomized trial (133) (consisting of 12 weeks
of weekly 2.5-hour groups) compared group CT with behavioral experiments
(n = 33) and group ERP
(n = 40) against a wait-list
control. The control subjects were later randomly assigned to either active
treatment. Subjects were diagnosed with DSM-IV OCD of greater than
1 year's duration and had been taking medication on a stable
regimen for 3 months. Ninety-three subjects entered the trial,
76 (82%) began treatment, and 63 (68%) completed
treatment. Both active treatments were significantly superior to
the control condition as reflected in change in Y-BOCS scores. The effect
sizes (Cohen's d) were 1.62 for ERP (n = 16),
and 0.98 for CT (n = 18).
For those completing treatment, the difference in the proportions
of "recovered" subjects (defined as Y-BOCS score
decrease 6 points and final
score < 12) was not significant: ERP (38%, n = 32)
and CBT (16%, n = 31).
At 3-month follow-up, ERP was associated with a significantly greater
recovery rate among subjects who completed the treatment: ERP (45%)
and CBT (13%). The ERP group included more subjects using
medication, but in the analyses this did not affect improvement.
This study suggested that group ERP was marginally superior to group
CT. Caution in interpreting these results is warranted, because
more subjects in the ERP group were taking medication, and because
CT was characterized by a higher refusal rate among subjects accepted
A single-blind, randomized trial compared 12 weeks of weekly
2-hour group therapy sessions combining ERP with cognitive therapy
and homework assignments (group CBT) (n = 23)
against a wait-list control (n = 24)
(134). The group CBT included exposure, ritual prevention,
and cognitive restructuring. Seventy percent of those in group CBT
were YBOCS-35% responders compared with 4% of
controls. The Y-BOCS score effect sizes (not Cohen's
d) in the ITT groups were 1.33 for the group CBT group and 0.43
for the controls. The therapeutic gains were maintained at 3-month
follow-up. Group CBT was associated with a significant improvement
in the quality of life as measured by the Abbreviated WHOQOL. Nearly half
the subjects were maintained on stable medication regimens, but
this did not appear to influence the effect of active treatment.
These and other reports suggest the utility of group behavioral therapies
in the treatment of OCD; however, additional study is warranted.
An uncontrolled, double-blind study without random assignment
compared group CBT (including ERP) (n = 17)
and multifamily CBT (n = 19)
(136). In each treatment there were ten to twelve 2-hour
sessions monthly. This small study found that both modalities were
effective, with a significant decrement in the Y-BOCS score in both
components, and a significant decrease in the Sheehan Disability
Inventory in the group CBT component. This study provides support
for both group approaches. There are too few studies of multifamily
CBT to provide sufficient evidence to recommend its use.
In an unblended, random allocation study in India, patients
who had had no prior CBT but who had failed pharmacological treatment
were assigned to either a treatment in which family members functioned
as co-therapists conducting desensitization and ERP (n = 15)
or the same treatment without a family member co-therapist (n = 15)
(135). The treatment group with a family member co-therapist
showed significantly greater improvement at 12 weeks and at 1-month
follow-up on the MOCI and the Global Assessment of Severity scale.
This small study suggests the utility of family co-therapists, but
the results may not be generalizable to other cultures and are limited
by the absence of a control group and the variability to be expected
In a 12-week study (563), subjects with a primary
DSM-III-R diagnosis of OCD, a minimum Y-BOCS score of 15, and no medical
contraindications to yoga were randomly assigned to 12 weekly 2-hour
group sessions of Kundalini yoga (n = 12)
or 1-hour group sessions of mental mindfulness and relaxation response
management (n = 10). Both
groups were instructed to practice at home daily. Seven subjects
completed each treatment arm. The subjects in the yoga group who
completed the trial experienced a mean Y-BOCS score decrease of
38% (9.4 points), compared with a 14% decrease
(2.9 points) for the control group. In the ITT analyses, the yoga
group's mean decrease in Y-BOCS score was significant (5.5
points), but the mean decrease for the control group (2.0 points)
was not. In both groups, an unspecified but equal number of subjects
had been taking stable doses of anti-OCD medications for at least
3 months before randomization. Yoga was apparently well tolerated,
but the reasons for dropout are not given. This small study requires
replication by an independent group before any conclusion about
the effectiveness of yoga can be drawn.
Only six randomized trials have directly addressed whether
the combination of an SRI and CBT consisting of ERP is superior
to either treatment alone in adults with OCD, and limitations in
their designs and/or procedures prevent definitive conclusions.
Some studies also compared SRI monotherapy with ERP monotherapy.
Few studies had adequate sample sizes to detect small differences
between treatments, even if such differences did exist. Moreover,
some studies excluded patients with significant comorbidity even
though comorbidity is common in OCD and such patients may be those
who would benefit most from combination treatment. Finally, certain design
decisions (e.g., how the treatments were delivered) may have prevented
the detection of important differences between combination treatment
Despite these problems, the available data support the idea
that combination therapy can be superior to monotherapy in some
OCD patients but that it is not necessary for all OCD patients (166, 167).
In particular, one study found that combination therapy is superior
to ERP monotherapy in OCD patients with co-occurring depression
(164). In OCD patients without co-occurring depression,
another study found that combination therapy and intensive ERP therapy
alone were each superior to SRI monotherapy (123).
Marks et al. (564) compared the outcome of 40
OCD patients randomly assigned to receive oral clomipramine or pill
placebo plus 30 sessions of CBT consisting of ERP. The study design
was complex: during weeks 0–4, patients received either
clomipramine or placebo; during weeks 4–10, patients also received
either 30 sessions of ERP or 15 sessions of relaxation training
followed by 15 sessions of ERP. A direct comparison of the effects
of clomipramine plus ERP, ERP plus placebo, clomipramine plus relaxation
training, and placebo plus relaxation training could only be made
at week 7. At that time, patients receiving clomipramine plus ERP
(n = 10) had more improvement
in rituals than patients receiving ERP plus placebo (n = 10)
or clomipramine plus relaxation training (n = 10);
however, there was no significant interaction of clomipramine and
ERP. The complex design, small sample, lack of standard OCD measures,
and treatment procedures (e.g., the effects of clomipramine were
likely underestimated at week 7) preclude strong conclusions.
A separate study (565) compared the outcome of
49 patients randomly assigned to receive one of four treatments: 6 months
of clomipramine and 23 weeks of antiexposure instructions (clomipramine
plus antiexposure); 6 months of clomipramine and 23 weeks of self-controlled
ERP (clomipramine plus ERP [self]); 6 months of
clomipramine and 8 weeks of self-controlled ERP, followed by therapist-aided
ERP from weeks 8 to 23 (clomipramine plus ERP [self and
therapist]); or 6 months of placebo and 8 weeks of self-controlled
ERP, followed by therapist-aided ERP from weeks 8 to 23 (placebo
plus ERP [self and therapist]). At week 8, clomipramine
plus ERP (self, n = 25)
produced significantly more improvement in rituals and depression
than placebo plus ERP (self, n = 12).
However, at week 23, clomipramine plus ERP (self and therapist, n = 10)
showed no superiority over placebo plus ERP (self and therapist)
plus placebo (n = 8).
At week 8, clomipramine plus ERP (self, n = 13)
also produced significantly more improvement than clomipramine plus
antiexposure (n = 12)
in rituals and depression. Only three of the subjects in the clomipramine
plus antiexposure group improved enough to continue, precluding
further comparisons. The authors concluded that the combination
of clomipramine and ERP therapy had a small transitory
additive effect compared with placebo plus ERP therapy. However,
the complex design, small sample, lack of standard OCD measures,
and treatment procedures (e.g., the clomipramine groups only achieved doses
ranging from 127–157 mg/day) preclude strong conclusions.
Cottraux et al. (165) randomly assigned 60 adult
patients with OCD to 24 weeks of fluvoxamine with ERP (fluvoxamine
plus ERP), fluvoxamine with instructions not to engage in ERP (fluvoxamine
plus antiexposure), or placebo plus ERP. Of the 60 patients who
entered, 44 (73%) completed all 24 weeks (n = 16,
13, and 15, respectively); of these 44, 19 (43%) entered
with a major depressive or dysthymic disorder (mean 17-item Ham-D
score = 19). ERP therapy consisted of eight weekly sessions that
included imaginal ERP during sessions and self-controlled ERP, followed
by 16 weeks of therapist-guided ERP. Among those patients who completed
treatment, all groups improved on some measures of rituals and depression,
but only the fluvoxamine plus ERP group improved on all measures
at week 24. The combined fluvoxamine treatment groups had the largest
percent reduction in the duration of rituals per day (fluvoxamine
plus ERP = 46%; fluvoxamine plus antiexposure = 42%;
placebo plus ERP = 25%) and the largest percentage
of patients who, by self-report, had more than a 30% reduction
in rituals per day (fluvoxamine plus ERP = 69%;
fluvoxamine plus antiexposure = 54%; placebo plus
ERP = 40%); however, these group differences were
not statistically significant. Study limitations include the small
sample, the lack of standard OCD measures, and the limited information
about the treatment procedures (e.g., the ERP protocol consisted
of weekly sessions of an unspecified duration, and other psychosocial
interventions were provided "as needed").
Hohagen et al. (164) randomly assigned 60 adults
with OCD to receive fluvoxamine plus CBT or placebo plus CBT. Many
patients had co-occurring mood, anxiety, and personality disorders
(mean 21-item Ham-D score = 19), and many (92%)
had received prior treatment (84% had taken medication,
35% had prior CBT). The mean dose of fluvoxamine was 288
mg/day (range = 250–300 mg/day).
CBT was conducted weekly for at least 3 hours and included therapist-aided
exposure as well as cognitive restructuring. After 9 weeks of treatment,
both groups showed significant reductions in OCD severity. However,
there were significantly more YBOCS-35% responders in the
fluvoxamine plus CBT group (87.5%) than in the placebo
plus CBT group (60%). Post hoc analyses revealed that 1)
both groups improved significantly and comparably on compulsions,
but the fluvoxamine plus CBT group improved significantly more on
obsessions; and 2) patients with co-occurring depression fared better
if they received fluvoxamine plus CBT. The authors concluded that
combination therapy should be used when obsessions dominate the
clinical picture or when a secondary depression is present. Limitations
include the lack of data on whether the two groups differed in their
response to prior treatment, the fact that only 49 patients entered
the final analysis because 9 were removed to equate baseline Y-BOCS
scores in the two groups, and the fact that two patients dropped
out for clinical reasons.
Van Balkom et al. (61) randomly assigned 117
patients to receive one of five conditions: fluvoxamine plus ERP,
ERP, CT, fluvoxamine plus CT, or wait-list control. At baseline,
patients on average had mild depressive symptoms. Full results are
presented for the 70 patients who completed 16 weeks of active treatment
and the 16 patients who completed the 8-week wait-list condition.
The mean fluvoxamine dose at week 16 in both fluvoxamine conditions
was 197 (± 82) mg/day. All therapy sessions were 45 minutes
long. ERP consisted of gradual self-controlled exposure in vivo
with gradual self-imposed response prevention. At week 16, all active
treatments led to a significant decrease on all OCD measures,
with no significant differences between the treatment groups. The
authors concluded there was no reason to combine SRIs and CBT (either
ERP or CT) in OCD adults without severe co-occurring mood disorder.
However, neither the ERP nor the fluvoxamine dosing was optimized.
Moreover, the combination group received only 10 ERP sessions, whereas
the group receiving ERP alone received 16 sessions.
Foa et al. (123) compared treatment outcome in
122 adult OCD patients randomly assigned to receive intensive ERP,
clomipramine, clomipramine plus intensive ERP, or placebo. Patients
with major depressive disorder (and a Ham-D 18) were excluded.
Mean daily doses of clomipramine during the last study week for
all who entered and all who completed the trial, respectively, were
196 and 235 mg/day for clomipramine patients and 163 and
194 mg/ day for clomipramine plus intensive ERP patients.
ERP was delivered intensively for the first 4 weeks (i.e., two information-gathering
sessions, fifteen 2-hour sessions conducted over 3 weeks, two home
visits); this intensive phase was followed by eight 45-minute weekly
maintenance sessions over the next 8 weeks. Patients receiving clomipramine
plus intensive ERP began both treatments simultaneously. At week
12, all active treatments were superior to placebo at reducing OCD
symptoms. In addition, clomipramine plus intensive ERP and ERP did
not significantly differ from each other, but each was superior
to clomipramine alone. For all who entered treatment and all who
completed the 12-week trial, the CGI-I:1,2 response rates were,
respectively, 70% and 79% for clomipramine plus
intensive ERP, 42% and 48% for clomipramine, 62% and
86% for ERP, and 8% and 10% for placebo.
The authors concluded that clomipramine, intensive ERP, and their
combination are all efficacious treatments for OCD. In addition,
in OCD patients without co-occurring depression, intensive ERP was
superior to clomipramine. The authors noted several factors that
may have limited their ability to detect a superiority of combination
treatment over intensive ERP monotherapy (e.g., the exclusion of
patients with co-occurring depression, the potency of intensive
ERP, the fact that the combination group did not achieve maximum
clomipramine doses). Study limitations include the lack of data
on patients who dropped out after randomization but before treatment
started and the lack of systematic data on subjects' prior
treatment history. Interpretation of these results (123, 126)
is also limited by uncertainty as to whether the treatment groups
were equally treatment resistant at baseline and by high study refusal
rates and dropout rates.
In the absence of definitive data, combination treatment (SRI
medication plus CBT consisting of ERP) is appropriate in clinical
situations in which there are co-occurring disorders that are SRI responsive
or there has been a partial response to monotherapy (163),
and in efforts to reduce the chance of relapse when medication is
E. Discontinuation of Active Treatment
Four double-blind SRI discontinuation studies in adults with
OCD have been published. Each concerned a different SRI and used
a different design (e.g., length of observation and method of placebo
substitution) and a different relapse definition. Each produced
Pato et al. (199) found that 89% of
the 18 patients who had responded to clomipramine had "substantial
recurrence" (not further defined) of OCD 7 weeks after
they were blindly switched (over 4 days) to placebo. Romano et al.
(201) found no significant differences in 1-year estimates
of relapse rates among patients who had responded after 5 months
of fluoxetine treatment between 36 patients who continued to take
fluoxetine (21%) and 35 patients who were switched to placebo
(32%); relapse was defined as a 50% loss of
improvement on the Y-BOCS, a Y-BOCS score of 19, and a CGI-I rating
of much or very much worse relative to the end of treatment. Koran
et al. (200) found that it was uncommon for patients
who responded to sertraline to discontinue the medication because
of relapse or insufficient response over 28 weeks, regardless of
whether they continued to take sertraline (3% [3/108])
or were switched over 2 weeks to placebo (4% [5/113]); relapse
was defined as a Y-BOCS increase of 5 points, a total Y-BOCS score
of 20, and a 1-point increase in CGI-I score relative to the
end of acute treatment at three consecutive visits at 2-week intervals.
Rates of relapse or discontinuation because of insufficient response
were 9% for sertraline and 24% for placebo, a
significant difference. Finally, Hollander et al. (80)
found that patients who had responded to 9 months of paroxetine
treatment who continued to take paroxetine for 6 months
had a significantly lower relapse rate (37.7%) than those
switched immediately to placebo (58.8%) and a longer time
to relapse (62.9 days vs. 28.5 days); relapse was defined as a return
to the pretreatment Y-BOCS score or a 1-point increase on the
CGI-severity (CGI-S) scale (566) relative to the end
of treatment. In summary, using different SRIs, different study
designs, and different relapse criteria, double-blind discontinuation
studies reported SRI relapse rates ranging from a low of 4% over
28 weeks (200) to a high of 89% over 7 weeks
An unblinded study of 130 OCD patients (177),
in which drug discontinuation was instituted after response to 6
months of open treatment, reported significantly higher 6-month
relapse rates for the patients whose medication was discontinued:
8% (clomipramine 150 mg/day) versus 46% (no
drug), 0% (fluoxetine 40 mg/day) versus 40% (no
drug), and 8% (fluvoxamine 300 mg/day) versus
62% (no drug) (177). Equally large or larger
disparities were present after 1 and 2 years of treatment versus
no treatment, and relapse rates were higher. Relapse was defined
as a Y-BOCS increase of 25% plus a
CGI-I score indicating much or very much worse relative to the end
of treatment. These data suggest both that SRIs may not fundamentally
differ in the long-term durability of treatment response after treatment
discontinuation and that most patients will eventually relapse after
stopping SRI treatment.
In a review of 16 CBT studies that used ERP, Foa and Kozak
(202) concluded that patients receiving ERP (with and
without concomitant medication) did well long-term. Of 376 treated
patients, 76% were responders at follow-up (mean = 29
months; range = 6–72 months). Of responders to
acute ERP treatment (with or without medication), the proportion losing
their response during follow-up was 20% or less in most
studies. While suggestive, these findings are inconclusive because
of 1) design limitations in some studies (e.g., uncontrolled studies
with naturalistic follow-up); 2) methodological limitations in others
(e.g., lack of evaluators blind to original treatment assignment);
3) differences in determining "response"; 4) inconsistencies
in whether treatment during follow-up was permitted (and reported);
and 5) differences in length of follow-up. Further complicating
any comparison with SRI relapse rates is the fact that the relapse
definition employed in this review (i.e., loss of response) differs
from that used in the SRI studies.
A multi-site study that compared the effects of clomipramine
and intensive ERP after 12 weeks of treatment (123)
and again 12 weeks after treatment discontinuation (67)
found that subjects who responded to intensive ERP (with or without
concomitant clomipramine) had a significantly lower relapse rate
(12%) and longer time to relapse after treatment discontinuation
than did subjects who responded to clomipramine alone (45%);
relapse was defined as a return to baseline severity on the CGI-S
scale. Post hoc analyses of these data generally supported these
findings, since most of the relapse criteria examined produced the
same outcomealbeit with substantial variabilitydepending
on the specific criteria used for relapse (203). However,
high study refusal rates (71% of those meeting entry criteria
refused to participate) and dropout rates (18% at randomization
and an additional 23% before completion) may limit the widespread
applicability of this study's findings.
Together, these data suggest that ERP treatment response may
be more durable, at least in the short run, than response to some
SRIs after they are discontinued. However, the observed differences
could be explained by other factors, including clinical characteristics
of the subjects studied, differences in the length of follow-up,
the intensity of treatment prior to treatment discontinuation, the
rate of medication taper, and the relapse criteria. Because of these
differences, no definitive conclusions about the relative durability
of SRI and ERP treatment effects can be drawn from these studies.