Evidence-based pharmacotherapy of obsessive-compulsive disorder

Abstract

Pharmacological strategies for the treatment of obsessive–compulsive disorder (OCD) continue to develop apace but deficiencies remain. We present an updated literature review of the evidence supporting available strategies. We aim to answer key questions including: (1) What are the first-line treatments? (2) Does pharmacotherapy improve health-related quality of life? (3) How do we evaluate clinical response and relapse? (4) How long should treatment continue? (5) Can we predict treatment outcomes? (6) What is the management of treatment-refractory OCD? Selective serotonin reuptake inhibitors (SSRIs) remain the pharmacological treatment of choice for most patients and are associated with improved health-related quality of life. However, discontinuation is associated with relapse and loss of quality of life, implying treatment should continue long term. A substantial minority of patients fail to respond to SSRI. Such patients may respond to strategies such as dose elevation or adjunctive antipsychotic, although long-term trials validating the effectiveness and tolerability of these strategies are relatively lacking. Newer compounds targeting other neurotransmitter systems, such as glutamate, are undergoing evaluation.

Health-related quality of life, neuropsychopharmacology, obsessive–compulsive disorder

Introduction

Obsessive–compulsive disorder (OCD), with its own distinctive pathophysiology and pharmacology, is an enduring, lifespan illness and was considered untreatable prior to the 1960s. Epidemiological surveys using DSM-III, DSM-III-R and DSM-IV criteria have shown lifetime prevalence to range between 1 and 3% of the worldwide population (Robins et al.1984; Weissman et al.1994, Wittchen & Jacobi, 2005). Despite this relatively high prevalence, only a fraction of those with OCD present for treatment and the diagnosis is often missed. OCD is less common in men than women (1.0:1.5). The mean age of onset is 20 yr, with peaks at 12–14 yr and 20–22 yr (Rasmussen & Eisen, 1990). Rare in the very young, it increases to adult rates at puberty and affects around 1% of children and adolescents overall (Heyman et al.2001). Untreated OCD has been little studied and is thought to pursue an unremitting, fluctuating course, with the highest prevalence in the early years of middle adult life. Major comorbidity with Axis I and Axis II disorders has been identified (Hollander et al.1998), including depression in about two-thirds of clinical cases, simple phobia (22%), social phobia (18%), eating disorder (17%), alcohol dependence (14%), panic disorder (12%) and Tourette's syndrome (7%; Pigott et al.1994), as well as increased rates of suicidal behaviour. Individuals with OCD report substantial impairment in health-related quality of life (HR-QoL) and social functioning (Hollander et al.2010) and children with early onset OCD are particularly badly affected (Piacentini & Langley, 2004).The cost of OCD to society, in terms of human suffering, diminished individual potential and lost revenue, is high (Hollander & Wong, 1998). Severe, chronic OCD is associated with high levels of hospitalization (Drummond, 1993).

OCD remains poorly recognized and under-treated. Although surveys suggest that the time between the onset of symptoms and diagnosis may be decreasing, it is often only when depressive symptoms emerge that treatment is started. The average duration of untreated illness has been estimated at 17 yr (Hollander & Wong, 1998). Treatment success depends, to a large extent, on the condition not being overlooked. Better recognition of the disorder has been cited as a public health priority (National Collaborating Centre for Mental Health, 2006 ). OCD responds preferentially to drugs that powerfully inhibit the reuptake of serotonin at the synapse, i.e. clomipramine and the selective serotonin reuptake inhibitors (SSRIs). Drugs without potent serotonin reuptake inhibitor (SRI) activity, when used as monotherapy, have been shown to be ineffective in controlled trials. This selective pharmacological response has generated hypotheses about the role of serotonin in the aetiology of OCD but no unifying theory has emerged, thus far, and the mechanisms by which SSRIs exert anti-obsessional effects remain poorly understood. Indeed, it is widely believed that OCD encompasses a heterogeneous group of illnesses and that other neurotransmitters such as dopamine, noradrenaline and glutamate (Fineberg et al.2006b, 2010) are involved in its pathophysiology.

Method

This paper represents a revision and update of the original published in 2005 (Fineberg & Gale, 2005) and mainly consists of data derived from psychopharmacological treatment trials published subsequent to that review. Our narrative review is based, wherever possible, on randomized controlled trials (RCTs) and addresses clinical questions including: (1) What are the first-line treatments? (2) Does pharmacotherapy improve HR-QoL? (3) How do we evaluate clinical response and relapse? (4) How long should treatment continue? (5) Can we predict treatment outcomes? (6) What is the management of treatment-refractory OCD? In addition, we draw attention to methodological issues that we consider may enlighten clinical practice or advance future trial design in this field. Uncontrolled studies are cited where systematic data are lacking and meta-analyses are cited where adequate head-to-head comparator studies do not exist. Expert consensus guidelines are considered and practical suggestions made for the clinical setting.

A systematic search of electronic databases [EMBASE (1974–date), Medline (1966–date), PsycINFO (1987–date)] was run using a combination of the terms obsessive compulsive (randomized or control or clinical trial or placebo or blind) and (systematic or review or meta-analysis), as well as individual drug names. This was complemented by consulting with colleagues in the field and reviewing data presented at international, peer-reviewed symposia. There is a shortage of long-term and relapse-prevention studies in OCD and the majority of published studies present short-term data. OCD is a lifespan disorder and effective treatment early on may prevent the problems of long-term chronicity. Available data suggest that children may respond rather like adults. In this paper, we include an analysis of the limited studies that have been conducted in children with OCD. Unfortunately, there is almost no research into OCD in the elderly.

First-line treatments for OCD

Clomipramine

SRIs represent the mainstay of first-line pharmacotherapy for OCD. Evidence for the efficacy of clomipramine as an anti-obsessional agent in adults and children and its superiority over tricyclic antidepressants and monoamine oxidase inhibitors was reviewed by Fineberg & Gale (2005) .

SSRIs (Table 1)

In addition to clomipramine's powerful SRI activity, its active metabolite has strong noradrenergic properties. Given that the more highly selective SSRIs are also beneficial (see below) and show a similar slow, incremental effect on OCD symptoms, their anti-obsessional actions are likely to be related to SRI activity. A review of the early evidence for SSRIs (Fineberg & Gale, 2005) demonstrates the efficacy of fluvoxamine, sertraline, fluoxetine, paroxetine and citalopram.

Table 1

Rate of clinical response in placebo-controlled studies of SSRIs for patients with OCD

SSRI, Selective serotonin reuptake inhibitor; OCD, obsessive–compulsive disorder; CGI-I, Clinical Global Impression – Improvement; YBOCS, Yale–Brown Obsessive Compulsive Scale; C-YBOCS, Children's YBOCS.

Placebo-controlled studies of escitalopram

Escitalopram was investigated in a 24 wk, active-referenced (paroxetine), placebo-controlled multicentre study (Stein et al.2007). Patients received 10 mg (n=112), 20 mg (n=114) escitalopram, 40 mg (n=116) paroxetine or placebo (n=113). The primary endpoint was 12 wk and the study continued for a further 12 wk under double-blind conditions. Doses of 20 mg escitalopram and paroxetine were superior to placebo at 12 wk and all three active treatments were superior to placebo at 24 wk. When compared to placebo, the 20 mg escitalopram dose produced an earlier onset of action and was more effective than the 10 mg dose across the duration of the trial. This study demonstrates the importance of continuing treatment for an adequate duration (24 wk). Escitalopram was well tolerated. A post-hoc factor analysis by Stein et al. (2008) showed escitalopram to be effective for most symptom dimensions of OCD, but the hoarding/symmetry subtype was associated with a relatively poor response.

Have changes in study populations affected treatment trial design?

The magnitude of the observed treatment effect has diminished from 40 to 50% average reduction in baseline scores in the clomipramine studies to around 30% in later SSRI studies. This may result from changes in the characteristics of patients entering the trials over time. Based upon efficacy data, SSRIs and clomipramine were rapidly accepted as first-line pharmacological treatments for OCD (National Collaborating Centre for Mental Health, 2006) leading to challenges in recruiting treatment-naive patients into clinical trials. Thus, greater numbers of treatment-refractory individuals may have been included in more recent studies. Concurrently, exclusion of comorbid illnesses, such as depression, may have led to the inclusion of milder cases with lower capacity for symptomatic improvement. Indeed increasingly stringent inclusion criteria, often driven by regulatory authority requirements, may contribute to the changing pattern of response rates across disorders, since treatment groups are less heterogeneous. Increased placebo-response rates in recent OCD trials, in some cases exceeding 20% improvement in baseline scores, have also been observed. The reasons for this are likely to be complex, e.g. changes in patient expectation of improvement or a shift toward the inclusion of milder, atypical cases, more likely to undergo spontaneous remission. In the face of recruitment difficulties, baseline severity scores may become artificially inflated to fulfil entry criteria and, by diminishing to their real value after study entry, may additionally contribute towards increased response rates irrespective of treatment allocation. Rising placebo response rates caution against drawing conclusions about efficacy from open, naturalistic reporting or wait-list controls and emphasize the crucial importance of controlled investigation (Fineberg et al.2006b). The net effect of these changes has been to reduce the statistical power of RCTs, so that larger numbers are now needed to test efficacy of new treatments. Meta-analyses of existing studies can, to some extent, compensate by pooling data, but may be misleading if they fail to take these changes into account.

Direct head-to-head comparisons of SRIs in OCD

SSRI vs. SSRI

SSRIs differ from one another in terms of the selectivity and potency of effect at the serotonin transporter and their secondary pharmacological actions (Stahl, 2008). Consequently, one might predict differences in clinical efficacy in OCD. Stein et al. (2007) compared escitalopram (10 and 20 mg) with paroxetine (40 mg) and placebo. Whereas symptomatic improvements on 20 mg escitalopram and 40 mg paroxetine appeared similar from the 12 wk primary endpoint onwards, improvement in the Yale–Brown Obsessive Compulsive Scale (YBOCS) score was significantly better than placebo as early as week 6 in the 20 mg escitalopram group only. These results are not strong enough to support the superior efficacy or tolerability of any one SSRI, even in conjunction with meta-analysis data. Therefore, treatment selection should probably take into account other factors, such as potential drug×drug interactions with co-administered compounds and potential effects on the cytochrome system.

SSRI vs. clomipramine

While some meta-analyses report a smaller effect size for SSRIs relative to clomipramine, head-to-head studies tend to demonstrate equivalent efficacy but more favourable acceptability and tolerability for SSRI (reviewed in Fineberg & Gale, 2005). Meta-analyses may provide a more objective and quantifiable measure of treatment effect than narrative reviews such as this. However, problems arise in controlling for between-study differences such as dose, duration, blinding, method of assessment and population changes, so results must be viewed cautiously (Pigott & Seay, 1999). The UK National Institute for Health and Clinical Excellence (National Collaborating Centre for Mental Health, 2006) systematically accessed all available published and unpublished RCTs. In terms of efficacy, clomipramine and SSRIs were indistinguishable. However, clomipramine was associated with higher rates of adverse event-related, premature trial discontinuation when compared to SSRI. The meta-analysis of SSRI vs. placebo by Soomro et al. (2008) included 17 studies (3097 participants) and unequivocally demonstrated the efficacy of SSRIs in OCD. Analysis of 13 studies (2697 participants) indicated that SSRIs are nearly twice as likely as placebo to produce a clinical response (⩾25% reduction in YBOCS from baseline). Watson & Ress (2008) performed a meta-analysis of 13 RCTs in young people (aged ⩾19 yr) with OCD. Ten studies compared pharmacotherapy; five compared cognitive behavioural therapy (CBT) to control. The results show that pharmacotherapy and CBT are significantly more effective than placebo in controlling OCD symptoms in young people. The strength of this study is limited as there was no search for unpublished or non-English material nor was the validity of the included studies formally assessed.

Improved safety and tolerability of SSRIs over clomipramine offer considerable advantages for the long-term treatment of OCD and indicate that the SSRIs should usually be considered the treatment of choice, with clomipramine reserved for those who cannot tolerate or who have failed to respond to them.

Suicide in children with OCD receiving SSRI

A meta-analysis (Bridge et al.2007) examined the effects of SSRIs in children aged 6–18 yr, following warnings from the American Food and Drug Administration that SSRIs in the young may increase the risk of suicidal thoughts and behaviours. They identified 27 RCTs of SSRI, of which six were in OCD. There were no completed suicides and the pooled absolute rates of either suicidal ideation/suicide attempt (treatment vs. placebo) in OCD (1% vs. 0.3%) compared favourably with the pooled absolute clinical response rates (treatment vs. placebo; 52% vs. 32%). For patients with OCD, the data suggest a number needed to treat (NNT) of six compared to a number needed to harm (NNH) of 200. The authors concluded that the benefits of SSRI probably outweigh the risks in the OCD paediatric population. March et al. (2006) calculated the NNT and NNH for multicentre trials of sertraline in children and adolescents with major depressive disorder and OCD. NNT ranged from 2 to 10 with no apparent age effect in OCD. No patients reported suicidality in the two OCD trials, giving a NNH approaching infinity. The authors concluded a positive benefit to risk ratio for sertraline in paediatric OCD, with the doctor–patient relationship playing an important role.

What is the most effective dose?

Traditionally, it has been thought that OCD requires treatment with higher doses of medication than depression or anxiety. Fluoxetine, paroxetine, sertraline, citalopram and escitalopram have each been investigated in fixed dose studies (see Fineberg & Gale, 2005; Table 2). Positive dose–response relationships were clearly demonstrated for paroxetine and fluoxetine and less so for sertraline and citalopram. Stein et al. (2007) demonstrated a sustained advantage for 20 mg escitalopram over the 10 mg dose, which continued until the 24-wk endpoint. A dose of 20 mg escitalopram was superior to placebo on the YBOCS from 6 wk onwards and on secondary endpoints including remission, whereas 10 mg escitalopram separated from placebo only at 16 wk on secondary outcome measures only. Bloch et al. (2010) conducted a meta-analysis of nine SSRI studies to determine dose-related differences in efficacy and tolerability using a fixed-effects model. Higher doses of SSRI (within formulary limits) were associated with improved treatment efficacy than low or medium doses using YBOCS score or proportion of responders as outcome measures. Higher SSRI dose was not associated with ‘all cause’ drop-out rates but was associated with higher rates of drop-outs ‘due to side-effects’.

Table 2

Placebo-controlled comparator studies of fixed doses of SSRI

^{SSRI, Selective serotonin reuptake inhibitor; RCT, randomized controlled trial.}

a

Marginally significant benefit for medium and higher doses on primary analysis (total Yale–Brown Obsessive Compulsive Scale, p=0.059); significant on ‘responder’ analysis (p<0.05).

Strategies for dose titration in OCD

Exacerbation of anxiety in the early stages of treatment appears to be rare in OCD, as demonstrated in acute studies of SRIs, which show a slow, gradual treatment effect. Irrespective of dose, it can take several weeks or months for improvements to become established and it may help to inform patients about this. Early signs of improvement may be noticed by an informant before the patient, who may have problems recognizing their own progress (authors’ own observation). Observer-rated scales such as the YBOCS may thus be helpful to detect small improvements in the clinical setting. Moreover, clinicians may feel pressurized to change treatments or increase SSRI doses prematurely. A balance must be struck between tolerability and rate of dose increase. The British Association for Psychopharmacology Expert Consensus Guidelines (Baldwin et al.2005) suggest waiting for 12 wk before assessing efficacy and upwards dose titration in the face of insufficient clinical response. In the absence of evidence supporting a sustained advantage for rapid dose titration (Bogetto et al.2002), the arguments for slower dose increases are persuasive, particularly in children and the elderly, as slow titration can ameliorate early SSRI-related adverse events such as nausea and agitation. Longer-term, dose-related side-effects such as sleep disturbance and headache also need to be monitored. Sexual dysfunction is a common cause of drug discontinuation and, if necessary, strategies such as dose reduction, short drug holidays or adjunctive use of drugs with restorative potency (e.g. mianserin, Aizenberg et al.1999; agomelatine, Eser et al.2010; sildenafil, Farre et al.2004; mirtazapine, Lee et al.2010) can be considered in stable cases. Further guidance on dose titration is available in the American Psychiatric Association treatment guideline (Koran et al.2007). Pulse loading with i.v. and oral clomipramine has been examined under open conditions (Koran et al.1998) and double-blind conditions (Koran et al.2006) and warrants further investigation.

Do SSRIs improve HR-QoL?

Several peer-reviewed studies measured the impact of OCD on HR-QoL, of which a small number were treatment trials (reviewed in Hollander et al.2010). Among the several measures used, the 36-Item Short-Form Health Survey (SF-36; Koran et al.1996) was applied most frequently (Bobes et al.2001; Eisen et al.2006; Koran et al.1996; Moritz et al.2005; Rodriguez-Salgado et al.2006). The relationship between HR-QoL and symptom severity in OCD appears complex. While one study suggested a linear correlation between HR-QoL and YBOCS (Goodman et al.1989a, b), others found no correlation, implying that QoL and symptomatic change represent independent variables and indicating their separate evaluation in OCD treatment trials. Moreover, obsessions, compulsions and comorbid depression may impact differently upon HR-QoL; correlations with HR-QoL being most pronounced for depression severity and the number of OCD symptoms (WHO, 1999). These findings suggest that OCD-related HR-QoL may be most meaningfully assessed in samples where depression is not strictly excluded. It should be borne in mind that improvements in quality of life depend on the patient making life changes once OC symptoms have begun to diminish. Thus, differences in rates of change in symptoms may lead to weak correlation with changes in HR-QOL measures in short-term studies and long-term trials are needed to fully evaluate HR-QOL.

Hollander et al. (2010) analysed function and HR-QOL measures [Sheehan Disability Scale (SDS), SF-36, respectively] in two prospective, randomized, placebo controlled trials of escitalopram (Fineberg et al.2007b; Stein et al.2007). In the fixed dose, paroxetine referenced study by Stein et al. (2007), by week 12 there were statistically significant improvements in the social life and family life subscales of the SDS in all the active treatment groups and by 24 wk there were statistically significant improvements in all three subscales. The 20 mg escitalopram group showed significant improvements in the work subscale from week 6 onwards. Further, statistically significant improvements in all four of the mental health domains of the SF-36 were seen for escitalopram and paroxetine at 12 wk and were sustained through to 24 wk. There was also a statistically significant and clinically relevant advantage for responders, defined as ⩾25% improvement in YBOCS from baseline, compared to non-responders, based on mean scores for SDS and all domains of SF-36 (with the exception of bodily pain). Furthermore, at the 12 wk endpoint there were significant correlations between the YBOCS and SDS total scores, as well as the four mental health domains of the SF-36, implying a direct relationship between OCD symptom severity, function and HR-QoL. These results suggest that SSRI is associated with clinically relevant improvement in HR-QoL in OCD.

Criteria for treatment response and relapse

Substantial improvement can be achieved in many patients, but for approximately 50% the treatment response is incomplete (Table 3). The problem of partial response is beset by the lack of universally agreed definitions for response, relapse and remission and is an area that has received little controlled investigation. Pallanti et al. (2002a) point out that methodological issues such as the absence of a definition of non-response contribute to non-generalizability of study outcomes. Likewise, Simpson et al. (2006, 2008) describe the impact of varying statistical methods and outcome criteria on the interpretation of study results. The use of standardized operational criteria across treatment trials has been advocated. Stein et al. (2007) proposed a stringent remission criterion, requiring a YBOCS total score of ⩽10. Storch et al. (2010b) proposed response and remission criteria for children with OCD, arguing that a Children's YBOCS (C-YBOCS) reduction of 25% constitutes treatment response, a reduction of 45–50% constitutes symptom remission and a C-YBOCS total score of 14 represents remission after treatment. However, the definition of a meaningful clinical response and the concept of ‘relapse’ continue to spark debate and can be difficult to apply to an illness that naturally runs a chronic, fluctuating course and shows partial response to longer-term treatment. Other proposed relapse criteria include a worsening of post-baseline YBOCS of ⩾50%, a 5-point worsening of YBOCS, total YBOCS score ⩾19, Clinical Global Impression – Improvement Scale (CGI-I) scores of ‘much’ or ‘very much worse’ (Fineberg et al.2007a, b).

Table 3

Double-blind discontinuation studies of relapse prevention in OCD

^{OCD, Obsessive–compulsive disorder.}

a

In children and adolescents.

b

Survival analysis performed.

Hollander et al. (2010) attempted to validate the otherwise empirical responder and relapse criteria by correlating functional disability and HRQoL with YBOCS changes. They found a statistically significant and clinically relevant distinction between responders and non-responders, based on SDS and SF-36 scores, when response was defined as at least 25% improvement in YBOCS score relative to baseline. This indicates that a 25% improvement in YBOCS is clinically relevant and represents at least a partial response. Likewise, relapse, defined as a 5-point worsening of YBOCS, positively correlated with a deterioration in HR-QoL and social function, implying that this degree of symptom change signifies clinical relapse.

How long should treatment be continued?

Long-term efficacy studies

Given the protracted course of OCD, treatments need to maintain their efficacy over the longer term. Early, uncontrolled studies of SRI did not signal the development of tolerance; however, evidence from controlled studies would be more convincing (Fineberg & Gale, 2005).

In the study by Stein et al. (2007), patients received placebo (n=113), 40 mg paroxetine (n=116) or 10 mg (n=112) or 20 mg escitalopram (n=114) for up to 24 wk. The higher escitalopram dose and paroxetine were superior to placebo at the primary, 12-wk rating point and by 24 wk all three active treatments were superior. At 24 wk, the 20 mg escitalopram dose was more effective than the 10 mg dose when compared to placebo. This study highlights the advantage of continuing treatment for a sufficient period of time and suggests that, whereas both doses of escitalopram can be effective, the 20 mg dose has an earlier onset of action and superior long-term efficacy.

Relapse-prevention studies (Table 3)

Controlled studies that randomize patients to continuing or discontinuing drug treatment provide important information on the effect of continuation treatment on relapse. However, the design and interpretation of these studies is not always straightforward. The patients under test are usually selected to be treatment responders and are therefore not representative of all treatment-receiving patients.

Existing relapse-prevention studies have produced mixed results and at least some of the negative studies appear to reflect design flaws (see Fineberg & Gale, 2005). Overly rigorous ‘relapse criteria’ have sometimes been imposed, e.g. some studies require more than two criteria to be endorsed (Romano et al.2001), or to persist over several visits (Koran et al.2002), which may have compromised the sensitivity to detect outcome differences. Studies also need to differentiate between gradual re-emergence of OCD and acute ‘discontinuation effects’ occurring soon after drug termination, which are more likely to be related to the pharmacological properties of the compound.

Fineberg et al. (2007b) assessed the efficacy and tolerability of escitalopram in the prevention of relapse in patients with non-comorbid OCD. Patients were first treated with open label escitalopram (10 mg or 20 mg). Of the 468 entrants, 320 were classified as responders (YBOCS – decreased by ⩾25%) and were entered into the 24-wk relapse-prevention phase. Patients were randomized to escitalopram at the assigned dose or to placebo. Escitalopram was well tolerated and group improvements in symptoms were sustained throughout the extension phase. Relapse was defined as a worsening from baseline of ⩾5 YBOCS points. Patients randomized to placebo relapsed more quickly than those remaining on escitalopram. Moreover, 52% of the placebo group compared to 23% of the escitalopram group relapsed overall; this difference was statistically significant. Relapse also correlated with deterioration in function; patients who relapsed had statistically significantly worse outcomes that those who did not on SDS and SF-36 (Hollander et al.2010). These results suggest that worsening by 5 YBOCS points holds credibility as a threshold for relapse and illustrate the damaging effect of symptomatic relapse associated with treatment discontinuation on quality of life.

A meta-analysis detected overall superiority of SSRIs to placebo in preventing relapse amongst adult treatment responders (Fineberg et al.2007a). Viewed collectively, these results suggest that medication, as long as it is continued, probably confers protection against relapse, SSRIs are effective long-term treatments and relapse prevention represents a rational treatment target for OCD. Greist et al. (2003) recommend continuation of pharmacotherapy for a minimum of 1–2 yr in treatment responsive individuals and emphasize the importance of long-term treatment from the outset. Prior to discontinuation, patients should be advised to look out for the early signs of relapse so that pharmacotherapy can be reinstated; hopefully achieving the same level of improvement, although this cannot be guaranteed (Ravizza et al.1998). Discontinuation should be gradual to minimize discontinuation effects. In many cases, life-long medication may be the best option until clear predictors of relapse are available.

What is the best dose for long-term treatment?

The study by Romano et al. (2001), in which a dose of 60 mg fluoxetine appeared to be the most effective over the 24-wk placebo-controlled extension phase, to some extent supports the continuation of treatment at higher dose levels. Indeed, there is little evidence to support dose reduction as a strategy in the long-term management of OCD and therefore most experts recommend continuing treatment at the effective dose for optimal relapse prevention.

Predictors of treatment response

Roughly half of patients with OCD fail to make a good response to SSRI. The literature on pharmacological response predictors is sparse and inconsistent (Fineberg & Gale, 2005). A 9-yr follow-up study of 142 children and adolescents with OCD (Micali et al.2010) found that the main predictor for persistent OCD was the duration of illness at assessment. Garcia et al. (2010) analysed data from the study by March (2004) and found that youth with lesser symptom severity and OCD-related functional impairment, greater insight, fewer comorbid externalizing symptoms and poorer family accommodation showed greater improvement across treatment conditions than their counterparts. In the study by Storch et al. (2008b), children with comorbid illnesses such as attention deficit hyperactivity disorder, tic disorder and oppositional defiant disorder showed a poorer response. In a small follow-up study by Bloch et al. (2009), female gender, earlier age at childhood assessment, later age of OCD onset, more severe OCD symptoms and comorbid oppositional defiant disorder were associated with persistence of OCD into adulthood. A long-term follow-up study of treatment non-responders (Reddy et al.2010) also reported that poor outcome was predicted by later age of onset, as well as poorer quality of life at onset, shorter duration of follow-up and not receiving CBT during the interval period.

The identification of symptom ‘dimensions’ in OCD attempts to produce more homogeneous subgroups that might improve the prediction of treatment outcome. Stein et al. (2008) analysed data on 466 patients from a placebo-controlled RCT of escitalopram. Five factors (contamination/cleaning, harm/checking, hoarding/symmetry, religious/sexual and somatic/hypochondriacal) were identified in an exploratory factor analysis of individual YBOCS items. The hoarding/symmetry dimension was associated with a poorer response to escitalopram. In the study by Bloch et al. (2009), the presence of hoarding symptoms was also associated with the persistence of OCD symptoms. Landeros-Weisenberger et al. (2010) found that 60% of OCD patients with predominant symptoms in the harm/checking dimension were ‘much or very much improved’ in response to SRI treatment, particularly if treated with fluoxetine and fluvoxamine. Too few patients reported symptoms of hoarding for reliable conclusions to be drawn. The symmetry dimension was associated with poor response to SRI and with comorbid tic disorder, which may be independently associated with a poor treatment response (Leckman et al.1994; although not according to Bloch et al.2009).

In summary, we do not have reliable, clinical or demographic response predictors, although there is a growing consensus that a long period of untreated illness as well as a later childhood onset and symptoms of hoarding and symmetry may predict a poorer response to SSRI. Strategies for early identification and treatment are likely to be cost-effective and merit exploration. Research into potential OCD biomarkers, such as genes or neurocognitive changes, may provide new scope for optimizing treatment for individual patients on a more personalized basis.

Preferred options if first-line treatment fails

Switch the SRI

Practice guidelines produced by the American Psychiatric Association (Koran et al.2007) recommend a variety of approaches for partial, little or no response to initial SRI therapy. Switching SSRIs remains the preferred option for many clinicians. However, it may be appropriate to persist with a given SRI, even in patients showing little improvement, since delayed response may occur after more sustained treatment (Miguel et al.2009).

Increase the SRI dose

Uncontrolled case studies, small randomized studies (e.g. Ninan et al.2006) and a systematic, retrospective case-note survey (Pampaloni et al.2009) suggest that increasing SSRI doses beyond formulary limits can produce better effects in some patients. The American Psychiatric Association Practice Guideline (Koran et al.2007) provides an empirically derived table of starting, usual, maximum and ‘occasionally prescribed’ SRI doses (Table 4).

Table 4

Dosing of SRIs in OCD (adapted from Koran et al.2007)

^{SRI, Serotonin reuptake inhibitor; OCD, obsessive–compulsive disorder.}

a

Lower doses may be required to avoid side-effects.

b

Used for rapid metabolizers, patients with no or mild side-effects or inadequate clinical response following treatment at the usual maximum dose.

c

Better absorbed with food.

d

Combined plasma level of clomipramaine and desmethylclomipramine 12 h after the dose should be kept <500 ng/ml to minimize the risk of seizures and cardiac conduction delay.

Altering mode of delivery

Altering mode of administration of SRI from oral to i.v. may improve response but is often impractical (reviewed in Fineberg & Gale, 2005). Pallanti et al. (2002b) investigated the use of i.v. citalopram in an open label study in a selected population of resistant patients. While this study demonstrated good efficacy, tolerability and a fast response to i.v. citalopram, strong evidence supporting this mode of delivery is as yet unavailable.

Combining SRIs and drugs exerting antidepressant or anxiolytic properties

Based mainly on the results of small studies and open case series, the evidence supporting adjunctive antidepressant or anxiolytic drugs is rather limited (see Fineberg & Gale, 2005).

Combining SRIs with drugs with antipsychotic properties

There are no positive studies of antipsychotic monotherapy in OCD that meet current RCT standards.

First generation antipsychotics

McDougle et al. (1990) reported benefit from adding open-label pimozide (6.5 mg) in 17 patients unresponsive to fluvoxamine. A subsequent double-blind, placebo-controlled study demonstrated a significant improvement with haloperidol (mean dose 6.2 mg) added to fluvoxamine with a preferential response seen in patients with comorbid tics (McDougle et al.1994).

Second generation antipsychotics (Table 5)

Adjunctive risperidone

Second generation antipsychotics that modulate serotonin and dopamine neurotransmission also offer promise and constitute a lower risk of extra-pyramidal effects. Positive reports from case series were confirmed by McDougle et al. (2000), who reported the first double-blind, placebo-controlled study showing efficacy for adjunctive risperidone in 36 patients unresponsive to 12 wk SRI. Risperidone (mean dose 2.2 mg) was superior to placebo in reducing YBOCS scores as well as anxiety and depression, was well tolerated and there was no difference between those with and without comorbid tics or schizotypy. A smaller, double-blind study by Hollander et al. (2003) examined patients failing to respond to at least two trials of SRI. Four out of 16 patients randomized to risperidone (0.5–3.0 mg) were responders at the 8 wk end-point, compared to none of the six patients randomized to placebo. In an 8 wk, double-blind, placebo-controlled study, Li et al. (2005) compared the benefits of 2-wk adjunctive risperidone (1 mg), haloperidol (2 mg) and placebo in 16 SRI-resistant OCD patients. Risperidone and haloperidol were both significantly superior to placebo in reducing obsessions and anxiety. In addition, risperidone but not haloperidol improved depressed mood. Selvi et al. (2011) demonstrate superiority of risperidone over aripiprazole as adjunctive treatment in OCD patients who failed to respond to 12 wk of treatment with SSRI. Refractory patients (n=41) were randomized to receive either risperidone (3 mg/d, n=20) or aripiprazole (15 mg/d, n=21) in addition to continuing SSRI. Of the risperidone group, 13 (72%) met response criteria (⩾35% decrease in YBOCS from baseline) compared to eight (50%) in the aripiprazole group, suggesting risperidone's superiority over aripiprazole.

Table 5

Adjunctive second generation antipsychotics

a

⩾35% improvement in Yale–Brown Obsessive Compulsive Scale (YBOCS).

b

⩾25% improvement in YBOCS.

c

Final YBOCS score <16.

d

CGI of “much improved” or “very much improved”.

e

Consensus opinion of investigators.

f

Completer analysis.

Adjunctive olanzapine

Bystritsky et al. (2004) conducted a 6 wk, double-blind, placebo-controlled study of adjunctive olanzapine in 26 OCD patients who had failed to respond to SRI monotherapy. The olanzapine group (mean dose 11.2 mg) showed a mean improvement in YBOCS of 4.2 points, compared to a mean increase of 0.54 points in the placebo group. Six of the 13 patients in the olanzapine group met response criteria compared to none in the placebo group. Olanzapine was well tolerated with only two patients dropping out because of side-effects. Shapira et al. (2004) conducted a similar study in 44 patients with OCD who had failed to respond to 8 wk open-label fluoxetine (mainly 40 mg). As the investigators had failed to wait for 12 wk, it is possible that these subjects were not truly resistant. Moreover, both groups showed significant improvement from baseline with no significant difference between them.

Adjunctive quetiapine

There have been four randomized, double-blind, placebo-controlled trials of quetiapine in OCD, one pooled analysis and one meta-analysis. Carey et al. (2005) investigated subjects who had responded inadequately to 12-wk open-label SRI. Adjunctive quetiapine failed to show superiority to placebo after 6 wk. Eight (40%) of the quetiapine group and 10 (47.6%) of the placebo group were classed as responders. Again, the high placebo response rate seen in this study may reflect the fact that patients were not truly SSRI resistant at the point of study entry. In contrast, the double-blind, placebo-controlled study by Denys et al. (2004) showed clear evidence of efficacy for 8 wk adjunctive quetiapine (<300 mg) in 20 SRI-refractory patients, producing a mean decrease of 31% on baseline YBOCS, compared to 20 patients on adjunctive placebo, who showed only 6% improvement. Fineberg et al. (2006a) randomized 11 patients who had failed to respond to SRI monotherapy over the preceding 6 months to adjunctive quetiapine and 10 to placebo. Quetiapine (mean 215 mg) produced a mean reduction in YBOCS scores of 3.4 points (14%) compared to 1.4 (6%) for placebo and three of the quetiapine-treated subjects achieved a response compared to one on placebo. These differences did not reach statistical significance. Kordon et al. (2008) conducted a 12-wk, double-blind, randomized, placebo-controlled study of adjunctive quetiapine in 40 patients with severe OCD who had not responded to treatment with 12 wk SRI. Between weeks six and 12, the dose of quetiapine was titrated up to 600 mg. The reduction in mean YBOCS from baseline was not significantly different between the groups (5.2 for quetiapine and 3.9 for placebo). The latter two studies may have been compromised by inadequate sample size. Alternatively, doses of quetiapine beyond 400 mg may be too high.

A meta-analysis (Fineberg et al.2006a) of data from three qualifying RCTs showed evidence of efficacy for adjunctive quetiapine (<400 mg/d) on the primary efficacy criterion (changes from baseline in total YBOCS), although the results were limited by between-study heterogeneity. Denys et al. (2007) additionally assessed the impact of the type and dose of SRI when augmenting with quetiapine in a pooled analysis of the study data. Quetiapine was found to be superior to placebo in reducing the mean YBOCS score (6.8 vs. 3.9). The benefit of quetiapine was seen mainly in patients taking the lowest SRI doses, compared to those on the median and highest doses. This may suggest that the same pharmacological mechanism underpins the anti-obsessional effect of SRI with or without adjunctive antipsychotic. The best responses for quetiapine were achieved in combination with clomipramine, fluoxetine and fluvoxamine.

Another double-blind, randomized placebo-controlled study has investigated the efficacy of quetiapine (300–450 mg/d) in combination with citalopram (60 mg/d) in 76 OCD patients who were treatment naive or medication free (Vulink et al.2009). Thirty-one received quetiapine and 35 placebo. When compared to placebo, quetiapine showed a significantly greater reduction on the YBOCS (11.9 vs. 7.8) and CGI-I (2.1 vs. 1.4) scales. However, more of the quetiapine-treated patients withdrew due to adverse effects. If replicated, this result would suggest that adjunctive quetiapine may be a more efficacious first-line treatment than SSRI monotherapy.

Adjunctive aripiprazole

In a small, open-label 12-wk study, Pessina et al. (2009) showed a significant improvement from baseline in 12 SSRI-resistant OCD patients co-administered flexible doses of aripiprazole (up to 20 mg). More recently, Muscatello et al. (2011) reported a highly significant advantage for adjunctive aripiprazole (15 mg) in a 16-wk, randomized, double-blind, placebo-controlled study involving 40 patients with treatment-resistant OCD (YBOCS total score; p<0.001). Patients randomized to aripiprazole showed a mean reduction from baseline YBOCS of nearly 7 points, whereas the placebo group barely changed. Of the 16 aripiprazole completers, 11 met response criteria of 25%, seven met criteria between 25 and 34% and four >35% reduction from baseline. In a double-blind RCT of aripiprazole vs. risperidone as adjunctive treatment to SSRI in patients who had failed to respond to 12 wk treatment with SSRI, aripiprazole was outperformed by risperidone with 8/16 (50%) vs. 13/18 (72%) meeting response criteria (⩾35% decrease in YBOCS from baseline) respectively (Selvi et al.2011).

Which antipsychotic?

Notwithstanding the small number of studies of small size, there is now at least one positive RCT of adjunctive haloperidol, risperidone, olanzapine, quetiapine and aripiprazole in SRI-resistant OCD. Skapinakis et al. (2007) conducted a meta-analysis of antipsychotic augmentation (excluding aripiprazole) of SRI therapy, which concluded that OCD patients randomized to the antipsychotic arm of such trials were up to three times more likely to respond than those randomized to placebo. Taken together, the evidence supporting the efficacy of antipsychotics as adjunctive treatments in OCD appears convincing. However, there are insufficient data upon which to determine the most effective compound or the optimal dose. The suggestion of a stronger antidepressant effect for risperidone compared to haloperidol (Li et al.2005) may be relevant for treatment-resistant OCD, where depression commonly supervenes. An 8-wk, single-blind, RCT (Maina et al.2008) compared adjunctive risperidone (1–3 mg) with olanzapine (2.5–10 mg) in 50 treatment-resistant OCD patients. Both groups responded similarly to augmentation, with 44 and 48% respectively of the risperidone and olanzapine group achieving responder status. However, risperidone was associated with amenorrhoea whereas olanzapine was associated with weight gain. A small, retrospective, non-placebo, comparator study (Savas et al.2008), which demonstrated clinical improvement in 80% of the patients treated with adjunctive quetiapine (n=15) compared to 44% of those treated with ziprasidone (n=9), hints that quetiapine may be superior to ziprasidone. In a recent randomized, open-label trial (Diniz et al.2010), the effectiveness of adjunctive quetiapine (<200 mg) was compared to that of adjunctive clomipramine (<75 mg) in 21 SSRI-resistant patients. There was a non-significant advantage for quetiapine on the YBOCS and four (36%) patients in the quetiapine group showed a clinical response, compared to just one (10%) in the clomipramine group, suggesting possible superiority for adjunctive quetiapine.

Long-term adjunctive antipsychotic

We do not have randomized controlled studies to allow a full evaluation of the long-term effectiveness of adjunctive antipsychotic in OCD. In a naturalistic study, Marazziti et al. (2005) assessed the long-term efficacy of adjunctive olanzapine (2.5–10 mg) in 26 treatment-resistant individuals. At 1 yr, 17 patients (68%) showed a reduction in YBOCS of at least 35%. In another 12-month trial of antipsychotic augmentation, Matsunaga et al. (2009) randomly allocated 44 SSRI-resistant patients to adjunctive olanzapine, quetiapine or risperidone plus CBT. Another group of SSRI responders (n=46) continued on SSRI and CBT for 1 yr. Compared to the resistant group, the responder group had lower YBOCS scores, both at the start and the end of the study (25.8–13.7 vs. 29.3–19.3), suggesting that relatively modest long-term improvements accrue in the more refractory cohort when adjunctive antipsychotic is prescribed.

Some authors report emergent obsessions during treatment of schizophrenia with second generation antipsychotics (Kim et al.2009) and this may be particularly evident for clozapine (Sa et al.2009); although, there have also been reports of obsessive symptoms in patients with schizophrenia improving with clozapine (Kumar & Howie, 2003) or with the addition of aripiprazole (Englisch et al.2009). These effects may be related to the mixed receptor profile of the drugs, combined with the neurological changes associated with schizophrenia and are hard to reconcile with the evidence of efficacy for the same drugs when used to treat OCD. New evidence is emerging from gene association and pharmacogenetic studies that variants of the glutamate transporter gene, SLC1A1, are associated with transmission of OCD traits, particularly in males (Arnold et al.2006), and that these variants may be associated with susceptibility to the emergence of atypical antipsychotic-induced OCD in schizophrenia (Kwon et al.2009).

Altogether, these results favour the use of adjunctive risperidone, quetiapine, olanzapine, aripiprazole and haloperidol as a strategy for resistant OCD. In view of the long-term adverse effects associated with these drugs, despite their evident efficacy, it remains controversial as to whether they should be used in preference to increasing the SSRI dose or augmenting with another agent, such as clomipramine. Head-to-head studies comparing the relative efficacy and tolerability of these interventions at different fixed doses and over the long term, as well as relapse-prevention trials, are much needed in this area. It also remains uncertain as to how long patients need to remain on augmented treatment. A small retrospective study by Maina et al. (2003) showed that the majority of patients (15 of 18), who had responded to the addition of an antipsychotic to their SRI, subsequently relapsed when the antipsychotic was withdrawn.

Combining SRIs with other agents

Serotonin receptor agonists/antagonists

The 5-HT₃ receptor antagonist ondansetron is theoretically interesting as it modulates serotonin and dopamine neurotransmission. Ondansetron was explored in a small open-label study of treatment-resistant OCD patients and showed a positive signal for clinical response and tolerability (Pallanti et al.2009). In an 8-wk double-blind, placebo controlled study (Soltani et al.2010), 42 patients were randomized to receive either fluoxetine (20 mg/d) and ondansetron (4 mg/d) or fluoxetine (20 mg/d) and placebo. Both groups improved and by week eight the mean YBOCS score in the ondansetron group was significantly lower (5 vs. 15; baseline in both groups was 35) suggesting possible efficacy in OCD.

Glutamatergic antagonists

Glutamate is a ubiquitous neurotransmitter throughout the central nervous system, including the corticostriatal–thalamic circuits considered to be important for the generation of compulsive behaviour. Preliminary findings of Pasquini & Biondi (2006), reporting an immediate beneficial effect of augmentation with the glutamatergic compound memantine, in one of two cases, led to the exploration of adjunctive memantine in a number of small uncontrolled studies that have shown promise. In a 12-wk open-label trial (Aboujaoude et al.2009), adjunctive memantine (5–20 mg) was given to 15 SRI-resistant patients. Almost half the subjects had a meaningful improvement. Memantine was well tolerated and no patient left the study early because of an adverse event. In the study by Feusner et al. (2009), 10 OCD and seven generalized anxiety disorder (GAD) subjects received 12 wk of open-label memantine (20 mg/d), as either monotherapy or augmentation of their existing medication. The OCD group experienced a significant mean 40.6% reduction in YBOCS scores at endpoint. Three of 10 of OCD subjects were classified as responders, whereas none of the GAD subjects was a responder. Again, memantine was well tolerated and there were no serious adverse effects. Stewart et al. (2010) conducted a further single-blind, case–control study of memantine in OCD inpatients. Twenty-two subjects receiving memantine in addition to standard care were matched with 22 OCD inpatient controls. The mean reduction in the YBOCS was 7.2 (27%) for memantine and 4.6 (16.5%) for standard care, providing further evidence for the possible efficacy of memantine in OCD.

Riluzole, another glutamate-modulating compound, was used as an adjunctive agent in a small (n=13) open-label OCD trial (Coric et al.2005). Seven patients improved and five of them were classified as ‘recovered'. These results are supported by two further small open-label studies of riluzole in children with OCD, which reported improvement on the YBOCS without major adverse effects (Grant et al.2007; Pittenger et al.2006) and by preliminary results of an ongoing trial (Grant et al.2010).

D-cycloserine is a partial agonist of the N-methyl-D-aspartate (NMDA) receptor and has been used to enhance exposure therapy outcome in paediatric OCD in a randomized, double-blind, placebo-controlled augmentation trial (Storch et al.2010a). Although not statistically significant, the active treatment arm showed small-to-moderate treatment effects. There have been mixed results with d-cycloserine augmentation of exposure and response prevention (ERP) therapy in adults. Three small studies (Kushner et al.2007; Storch et al.2008a; Wilhelm et al.2008) each used different doses (100, 125 and 250 mg) and different timings (1, 2 and 4 h before ERP). Only the study by Wilhelm et al. (2008) yielded some positive results, although these were at the mid- rather than the end-point of the study. Overall the results are not convincing.

Treatment with glycine, another NMDA receptor modulator, over a 5-yr period in a patient with OCD and body dysmorphic disorder (BDD) led to robust reduction of BDD and OCD symptoms and partial relapses during treatment cessation (Cleveland et al.2009). In a 12-wk placebo-controlled RCT of adjunctive glycine (<60 mg; Greenberg et al.2009), patients completing glycine treatment (n=5) experienced a statistically non-significant mean reduction in YBOCS of 6 points compared to 1 point in the placebo group (n=9). Two patients taking glycine were classed as responders; however, the compound was poorly tolerated and 10 dropped out, mainly because of its aversive taste or nausea.

Anti-epileptics

In a retrospective, open-label case series in 16 patients who were non-responsive or partially responsive to SRI with or without antipsychotics, 11 patients prescribed adjunctive topiramate (up to 400 mg/d; mean dose 253 mg) were found to be globally clinically improved (van Ameringen et al.2006). However, a 12-wk, double-blind, placebo-controlled, parallel group trial, in which topiramate (up to 400 mg/d; mean dose 178 mg; n=18) was poorly tolerated compared to placebo (n=18), produced largely negative results, suggesting that, at best, SRI augmentation with topiramate may be beneficial for compulsions (YBOCS compulsions subscale, p=0.014) but not for obsessions or OCD as a whole (YBOCS total, p=0.11; Berlin et al.2011). A single case report suggests a possible role for adjunctive lamotrigine titrated up to 150 mg/d (Uzun, 2010).

Other agents as monotherapy

Mirtazapine monotherapy was evaluated in a double-blind, placebo-controlled discontinuation study (Koran et al.2005b). Thirty patients (15 treatment-naive and 15 treatment-experienced) entered the 12-wk open-label 30–60 mg mirtazapine phase. At 12 wk, 15 of 16 responders were randomly allocated to receive either ongoing mirtazapine or placebo. The mean YBOCS score in the mirtazapine group fell by 2.6 and in the placebo group rose by 9.1, indicating mirtazapine's potential in treating OCD. Other agents, supported by small, double-blind RCTs, which warrant further investigation, include dextroamphetamine, caffeine (Koran et al.2009) and once weekly oral morphine (Koran et al.2005a).

Conclusions

The pharmacological evidence base for the treatment of OCD is developing into new areas. Treatment with SSRIs and clomipramine remains uncontroversial and improvements are sustained over time. SSRIs are usually preferred over clomipramine, in view of their improved tolerability, and relapse prevention, achieved through sustained treatment, is a realistic goal for most SSRI responders (Fineberg et al.2007a). Moreover, SSRIs are associated with improved HR-QoL, which may be lost if symptomatic relapse occurs. On the other hand, SRI response is often incomplete and a substantial proportion of cases fail to reach remission. Better predictors of treatment outcome, which could be used to plan personalized care, are needed. Treatment-resistant OCD is receiving systematic evaluation. Options include continuing the SSRI at maximal dose levels for a longer period, increasing the dose beyond formulary limits, switching the SSRI or augmenting the SSRI with a first or second generation antipsychotic, while novel treatments such as compounds acting on serotonin receptors or glutamate neurotransmission are under evaluation.

Acknowledgements

The authors thank Christine Lawes, Vivienne Eldridge and Lesley Gibbon of the Postgraduate Medical Centre Library and Kiri Jefferies, research assistant in the Research and Development Department, Hertfordshire Partnership Foundation Trust for their tireless support in locating and obtaining papers on our behalf. Without their help, this project would have taken twice as long and been half as interesting to the reader.

Statement of Interest

Naomi Fineberg has consulted for Lundbeck, GlaxoSmithKline, Servier, Transept and Bristol–Myers Squibb and has received research support from Lundbeck, GlaxoSmithKline, AstraZeneca, Wellcome, Cephalon, Servier and ENCP and has received honoraria for lecturing at scientific meetings from Janssen, Jazz, Lundbeck, Servier, AstraZeneca and Wyeth. Financial support to attend scientific meetings has been received from Janssen, Bristol-Myers Squibb, Jazz, Lundbeck, Servier, AstraZeneca, Wyeth, Cephalon and the International College of OC Spectrum Disorders. Between 2001 and 2006 Angus Brown was co-investigator at Roche Products Ltd. Phase I Clinical Pharmacology Unit in Welwyn Garden City and has subsequently participated in research sponsored by Servier. Samar Reghunandanan has participated in research sponsored by Servier. Ilenia Pampaloni has consulted for Lundbeck. Financial support to attend scientific meetings has been received from Eli-Lilly and Pfizer.