Why does exposure-based therapy fail in some individuals with obsessive-compulsive disorder?

KEYWORDS:

• Exposure and response prevention
• obsessive-compulsive disorder
• outcome prediction
• treatment failure
• cognitive behavioral therapy

1. Introduction

Exposure and response prevention (ERP) is a psychotherapeutic method in that individuals diagnosed with obsessive-compulsive disorder (OCD) (1) expose themselves to stimuli that trigger unwanted and intrusive thoughts, urges, or images (i.e. obsessions) and (2) refrain from actions they would usually engage in to reduce obsession-related distress or prevent-feared consequences of the exposure (i.e. compulsions). ERP is probably the best studied single method of psychotherapy in OCD and was shown to be effective both as stand-alone intervention and as part of cognitive behavioral therapies [1]. Individual treatment success is usually defined in terms of response and remission, ideally measured on the basis of symptoms decreasing to a minimum extent and falling below a defined cutoff score in reliable and valid symptom measures [2]. International consensus criteria require (amongst others) a minimum symptom reduction of at least 35% on the Yale-Brown Obsessive Compulsive Scale for response and a cutoff-score level of 12 or lower for remission [2]. Treatment failure would thus be indicated by not achieving symptom reduction of 35% or remaining at a symptom level of more than 12 at the end of therapy, depending on whether success is defined on the basis of response or remission. Although response without remission is typically not regarded as treatment failure, evaluation of treatments should consider that full remission may considerably increase the prospects for long-term recovery [3]. In addition, a particular type of failure is represented by premature discontinuation of therapy, which, however, typically includes non-achievement of criteria for response or remission.

Average success rates for exposure-based treatments in randomized controlled trials can be estimated to vary between 43% and 50% for remission and between 62% and 65% for response (including remission) [1]. Specific approaches can even reach higher scores. For example, in the Bergen 4-day-treatment (B4DT), a concentrated form of exposure treatment, 73% of participants achieved remission and another 22% achieved response at post-treatment [4]. Given the relatively low impact of common factors on OCD symptoms [5], the success rates of exposure-based therapy in OCD reflect specific treatment effects that are among the largest in the field of treatment for mental disorders. Nevertheless, there is still a substantial number of individual patients who do not sufficiently benefit from ERP-based treatments. Although several other treatments such as acceptance and commitment therapy, metacognitive therapy, or selective serotonin reuptake inhibitors have gained preliminary or robust evidence of efficacy [6], it is not clear to date whether and to what extent individuals without response to ERP benefit from those alternatives, either as subsequent or as concomitant treatment [6]. These limitations illustrate a need for further improvement of ERP, which in turn might depend on our understanding of reasons for treatment failure in some patients.

2. Relevant research

In our view, there are three lines of research that may contribute substantial knowledge to answer this question:

1. studies on outcome prediction by pre-treatment characteristics of patients and their families

2. studies on outcome prediction by psychological processes assessed in the course of therapy

3. studies on features of the setting or mode of treatment

Here, we will highlight selected empirical findings that might be particularly helpful to understand why the otherwise highly effective ERP method does not work for all patients with OCD, and subsequently provide a subjective evaluation on the basis of own clinical observations.

2.1. Pre-treatment characteristics of patients and families

While a number of socio-demographic and clinical variables (e.g. comorbid anxiety, depression or personality disorder, unemployment) have been associated with worse treatment outcomes in several but not all relevant studies, high initial symptom severity has been identified as obstacle for treatment success more consistently, particularly for achieving remission status [7–9]. As most prediction effects of individual characteristics identified so far are relatively small, a majority of patients with OCD have a generally high chance to benefit from ERP-based treatments irrespective of individual baseline characteristics. This also implies that research on prognostic properties of individual characteristics does not add much to understanding failure of ERP-based treatments, at least to date. Outcome prediction may improve, however, if family characteristics are included. Indeed, several studies showed that treatment outcome is worse if family accommodation (FA) is high, i.e. if family members of patients with OCD are highly engaged in helping their relative to avoid or alleviate distress related to their symptoms [10].

2.2. Psychological processes in the course of therapy

Although ERP is typically embedded in a varying number of other treatment components such as psychoeducation or other cognitive interventions, there is some evidence that the ERP method itself is crucial for patients with OCD in order to achieve treatment success. Firstly, receiving ERP appears to be a substantially stronger predictor of symptom reduction than common factors such as therapeutic alliance or expectancy [5]. Secondly, if ERP is conducted after several preparatory sessions, the largest rate of symptom reduction is typically seen after ERP sessions have started [11]. Thirdly, better treatment outcome is associated with ERP-related factors such as the number of exposure sessions [12] or adherence to ERP procedures, i.e. the extent to which patients actually confront their obsessive fears (adherence to the exposure component) and stop their compulsive responses (adherence to response prevention) [13–15]. Although adherence to ERP procedures relates to both within-session and between-session exposure and response prevention, the degree of success with between-session response prevention may be particularly important [13].

Despite the outstanding role of ERP in the treatment of OCD, there is surprisingly little evidence regarding the mechanisms that might mediate these effects. Cognitive-behavioral theories generally assume that exposure to threat associated stimuli and subsequent response prevention counteract avoidance and neutralizing behaviors that otherwise entail the maintenance of threat associations [16]. Specific assumptions exist on how preexisting threat associations might change during exposure. The Emotional Processing Theory (EPT), for example, posits that exposure acts via an extinction of threat associations and that a decline of fear, distress, or disgust during and between exposure sessions (within and between-session habituation) is a necessary prerequisite for this [16]. In contrast, the Inhibitory Learning Theory (ILT; recently updated and now referred to as Inhibitory Retrieval Approach) does not regard habituation as necessary for treatment success. It assumes that threat associations are inhibited by learning new associations [16,17]. This process is assumed to be promoted by measures that maximize the violation of expectancies and other treatment optimizations, which have been proposed on the basis of insights from experimental studies on learning mechanisms [16,17]. However, there is no evidence yet that the measures proposed to booster these experimentally derived mechanisms yield superior treatment effects [16]. Correlational findings on the relationship between process parameters and outcome are mixed for EPT and have rarely been studied for ILT [16]. In one study, Elsner et al. investigated parameters derived from both theories within the same exposure sessions. The results showed that habituation may contribute to treatment response. However, remission was only predicted by distress-related expectancy violation, i.e. experiencing lower distress during exposure than expected before [11]. Taken together, evidence is generally sparse, but habituation and expectancy violation may contribute to success of exposure-based treatment. Of note, these processes may work differently depending on whether patients predominantly experience anxiety or disgust during exposure sessions. There is at least some evidence that exposure-based treatment is less effective in reducing disgust compared to anxiety [18].

Whether ERP can be conducted effectively may also depend on therapist variables and the patient–therapist relationship. Although these factors are assumed to have relatively low impact on ERP for OCD [5], therapists may perceive organizational barriers (e.g. for leaving the office) or have dysfunctional attitudes (e.g. fear of adverse effects). If these do not prevent them from conducting ERP altogether, they may increase the risk of failure by impeding appropriate implementation of ERP (e.g. avoidance of maximally stress-inducing exposure items) [19]. Importantly, appropriate implementation includes therapist-guided as well as self-guided exposure exercises [20].

2.3. Setting or mode

There is some evidence that ERP is more effective in face-to-face vs. internet-based treatment [6]. However, such effects of treatment mode appear to be moderate, so that treatment failure may rather depend on inappropriate matching between treatment mode and individual characteristics or preferences of the patients. Unfortunately, research has not yet uncovered reliable criteria for differential indications of different treatment modes. The same applies to inpatient vs. outpatient setting, although clinical and economic criteria might favor outpatient treatment unless severity of symptoms or functional impairment, self-endangerment, or specific living conditions impede sufficient outpatient treatment. Regarding the intensity of treatment, recent research suggested that treatment failure is particularly low in the short but intensive B4DT [4], although superiority over lower intensity treatments remains to be proven in direct comparisons.

3. Expert opinion

Reviewing the empirical findings, ERP appears most likely to fail if patients are not able to adhere to ERP procedures, especially to sufficient response prevention. The reasons for non-adherence are not fully understood but are likely related to other predictors of treatment outcome, particularly characteristics of patients and their families, therapeutic processes, and treatment setting or mode. In the following, we discuss how these factors might influence treatment outcome, considering in particular a possible mediation by adherence to ERP procedures.

3.1. Personal and family characteristics

One of the most consistent predictors of outcome is high initial symptom severity. Its effect on outcome might be mediated by poor adherence because experienced or anticipated consequences of ERP are more distressing the more severe the symptoms are. In similar vein, FA might impair therapy response by reducing treatment adherence. If family members continue to provide reassurance, for example, this will likely impede the patient’s efforts to adhere to response prevention. High FA may therefore require active inclusion of family members in exposure-based treatments, in order to achieve satisfactory treatment gains [10].

Clinical observations also suggest that individual life circumstances may decrease motivation for ERP, e.g. family conflicts, poor perspective on life after long-term unemployment or high workload. This phenomenon has also been discussed – but rarely been investigated – with regard to hypothetical secondary functions of obsessive-compulsive symptoms, e.g. protection from aversive dealings with life tasks.

3.2. Therapeutic processes

Poor adherence to ERP was shown to mediate negative associations between poor patient-therapist agreement on tasks and goals on the one hand and treatment outcome on the other hand [14]. In addition, specific aspects of adherence might be related to hardly motivating exposure experiences. For example, patients may continue to exert compulsions, because they do not experience sufficient expectancy violation, i.e. distinct discrepancies between negative expectations and actual consequences of exposure exercises. This might be the case, for example, if negative emotions like anxiety or disgust continue to remain persistently strong although the patient had expected habituation. Or if an exposure exercise is conceptualized to test whether an event occurs (e.g. fire after not checking the stove), although the patient actually knows that this event is rather unlikely. These clinical examples show that exposure should not be introduced as exercises aiming at habituation or as experiments that test the probability of feared events. Exposure exercises should rather focus on expectations regarding distress or one’s ability to tolerate distress on the condition of response prevention [17,21].

Of note, the assumption of expectancy violation as a critical mechanism in exposure therapy does not necessarily mean that this process has to be made explicit. For example, the highly effective B4DT does not appear to encourage participants to focus on expectations and outcomes, but seems to rather work by motivating participants to ‘just doing it.’ Future research may show whether implicit expectancy violation contributes to its impressive long-term effects.

3.3. Setting and mode

The superior efficacy of face-to-face compared to internet-based therapy [6] and of treatments with vs. without therapist-guided exposure sessions [20] suggest that at least some patients benefit from direct therapist support in conducting exposure exercises and monitoring response prevention. If these patients receive insufficient support, they may have difficulties to adhere to ERP procedures properly, resulting in poor benefits. Similarly, highly intensive treatments like the B4DT may have large success rates partly because they facilitate patient adherence [15]. In low-intensity treatments, adherence is mainly challenged during between-session intervals when patients are supposed to conduct self-guided ERP. Here, digital tools proved to enhance adherence.

Although low adherence to response prevention is a major risk factor for failure of ERP treatment, patients may also continue to suffer from symptoms if they appear to adhere to ERP procedures outwardly. In these cases, emotional avoidance (not allowing feelings to arise) or mental compulsions (cognitive processes that neutralize obsessions, e.g. ruminating on the justification of obsessive fears) may play an important role. Hence, adherence might be impaired after all, but with respect to mental processes rather than overt behavior. Of note, recent research showed that patients with blunted emotional processing of threatening stimuli benefited less from subsequent ERP-based CBT [22].

4. Summary

Clinical considerations support the notion that patient non-adherence to ERP is a crucial risk factor for treatment failure. We suppose that poor adherence mediates the negative effects of patient and family characteristics (e.g. higher symptom severity, unemployment, family accommodation), unfavorable exposure processing (e.g. insufficient expectancy violation), and individually suboptimal setting or mode of ERP treatment (e.g. low therapist support or treatment intensity) on treatment success. In other words, we assume that specific patient and family characteristics, unfavorable exposure processing and suboptimal therapy settings negatively affect treatment success by leading to poor patient adherence to ERP procedures. However, this model has not been tested empirically so far. Future research should put more emphasis on investigating the complex interplay of predictors and processes in order to better understand why ERP is beneficial for many but not all individuals with OCD.

Declaration of interest

All authors are employed as academic staff and receive their salaries from their respective institutions (MSB Medical School Berlin and Humboldt-Universität zu Berlin). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors would like to thank J Klawohn for carefully reading the manuscript and providing valuable advice on the content and design of the manuscript.

Additional information

Funding

This paper was not funded.