ERP Therapy for OCD: How It Works and What the Trials Show

ERP therapy is a structured behavioral treatment for obsessive-compulsive disorder (OCD) in which a person deliberately faces the thoughts, objects or situations that trigger obsessions and then refrains from the compulsion that usually follows. NIMH describes ERP as "a specific type of CBT" and says it reduces compulsive behaviors even in people who do not respond well to medication^[1].

The American Psychiatric Association's practice guideline lists CBT and serotonin reuptake inhibitors (SRIs) as "safe and effective first-line treatments for OCD"^[6]. The guideline was approved in 2006 and published in 2007, so its session counts reflect expert consensus from that period^[6].

ERP is not a general talk therapy. Sessions are built around practice: psychoeducation, planned exposures and resisting rituals^[5]. Prevalence data for the condition itself are on our OCD statistics page, and the obsession and compulsion types ERP targets are covered in OCD symptoms.

Treatment combines three elements: psychoeducation, exposure and response prevention^[5]. NIMH's example is a person who is gradually exposed to touching dirty objects and then prevented from handwashing^[1].

Types of exposure

Exposures can be in vivo (real situations), imaginal (imagined situations) or interoceptive (the physical sensations of anxiety)^[5]. Exposure is introduced gradually^[1], typically planned as a ranked list from least to most distressing, often called an exposure hierarchy.

How exposure is delivered matters. A meta-analysis cited in the 2019 review found that exposure under therapist supervision, complete abstention from rituals, and a mix of in vivo and imaginal exposure were each associated with greater symptom improvement than the alternatives^[5].

Response prevention and homework

Response prevention means not performing the ritual, including mental compulsions such as silently repeating words, which therapists can overlook because they are invisible^[5]. Early compliance with between-session homework has repeatedly predicted better short- and long-term outcomes^[5].

Why it works: habituation vs. inhibitory learning

The older explanation was habituation: fear fades with repeated exposure when escape is blocked. Later research found that the drop in anxiety within a session does not predict treatment outcome, and that people can improve without it^[5].

The current inhibitory learning model holds that the original fear association stays intact while a new, non-threatening one is learned and competes with it. The review's example is "I will get sick if I touch a dirty counter" being countered by "the chances of getting sick from touching a dirty counter are low"^[5].

Across trials, about 50% to 60% of patients who complete ERP show clinically significant improvement in OCD symptoms, and gains tend to hold over the long term^[5]. The same review notes that roughly half of patients do not improve significantly^[5].

The best-known head-to-head test is a 2005 placebo-controlled trial of 122 adult outpatients at three centers^[2]. It compared 4 weeks of intensive ERP followed by 8 weekly maintenance sessions, 12 weeks of clomipramine, the two combined, and pill placebo^[2].

Among people who completed treatment, 86% responded to ERP, 79% to ERP plus clomipramine, 48% to clomipramine and 10% to placebo; rates among all treated patients were 62%, 70%, 42% and 8%^[2]. The authors concluded that intensive ERP "may be superior to clomipramine and, by implication, to monotherapy with the other SRIs"^[2].

Pooled data point the same way. A network meta-analysis of 61 randomized trials and 3,710 patients found every CBT format for OCD, a category that included ERP-only protocols, significantly more effective than control conditions^[4]. Individual CBT's effect against a waitlist was a standardized mean difference (SMD) of -1.25, and -0.96 against psychological placebo^[4].

OCD treatment generally means ERP, medication or both. NIMH says the most common medications are antidepressants that target serotonin, mainly SSRIs, that they can take 8 to 12 weeks to start working, and that OCD may require higher doses than depression^[1]. The APA guideline recommends an SRI alone for people who cannot take part in CBT or who prefer medication, and combined treatment when one approach alone has not worked well^[6].

In adults, the 2019 review concluded that ERP is as efficacious as, if not more efficacious than, first-line SRIs, and that combining the two was not significantly better than ERP alone in the 2005 trial^[5]. It also reports that 45% to 89% of patients see symptoms return after stopping an SRI, while improvement after ERP tends to persist^[5].

Adding ERP when medication is not enough

For adults still symptomatic on an SRI, a 2013 trial randomized 100 patients to add ERP, the antipsychotic risperidone or placebo for 8 weeks^[3]. Response rates were 80%, 23% and 15%, and 43%, 13% and 5% reached minimal symptoms^[3]. The authors wrote that such patients "should be offered EX/RP before antipsychotics"^[3].

Children and teens

The evidence in young people favors combining therapy and medication more than in adults^[5].

In the Pediatric OCD Treatment Study of 112 patients aged 7 to 17, 12-week remission rates were 53.6% for CBT plus sertraline, 39.3% for CBT alone, 21.4% for sertraline alone and 3.6% for placebo^[7]. Combined treatment did not differ significantly from CBT alone on remission, and the authors concluded young people should start with combined treatment or CBT alone^[7]. Broader data on this age group are on our adolescents page.

Other OCD treatments

For severe OCD that has not responded to other treatment, the FDA approved deep repetitive transcranial magnetic stimulation (rTMS) in 2018 and extended the approval to standard TMS devices in 2022^[1]. Deep brain stimulation is available for severe OCD under a Humanitarian Device Exemption but is still considered experimental^[1].

The APA guideline says the number, length and duration of an adequate ERP trial "have not been established," but expert consensus recommends 13 to 20 weekly sessions for most patients^[6]. It calls for sessions at least once a week, notes that sessions have run from under 1 hour to 2 hours, and suggests booster sessions for more severely ill patients and for those who have relapsed or show early signs of relapse^[6].

Frequency can be raised. The guideline notes that five ERP sessions a week may beat weekly sessions but are not necessarily better than twice-weekly ones^[6]. The 2013 augmentation trial used 17 twice-weekly 90-minute sessions over 8 weeks^[3].

The guideline treats 13 to 20 weeks of weekly CBT, or 3 weeks of daily CBT, as the point to reassess if response is poor^[6].

Intensive ERP: the Bergen 4-day format

The Bergen 4-day treatment compresses concentrated exposure into four consecutive days, delivered to groups of 3 to 6 patients with the same number of therapists^[8]. In a 2018 follow-up of 65 outpatients, 60.0% of whom had severe to extreme OCD, no one dropped out during treatment^[8].

At 12 months, 83.1% had responded and 67.7% were classified as recovered; 89% rated the treatment very good^[8]. The study had no control group, so these results show feasibility and durability rather than a randomized comparison with weekly ERP^[8].

Standard ERP has been weekly and face to face, but the 2026 network meta-analysis compared seven delivery formats across 61 trials^[4]. It found no significant differences in effectiveness among individual, remote (video or telephone), guided self-help, time-intensive and family-involved formats^[4].

Group therapy was less effective than individual, remote and family-involved formats (SMDs of -0.38 to -0.60)^[4]. Unguided self-help, with no clinician support, still beat a waitlist (SMD -0.67) but was less effective than individual, remote, family-involved and intensive formats, and less acceptable than most other formats^[4].

Two caveats apply. The analysis grouped ERP with other CBT variants, including cognitive therapy and acceptance and commitment therapy, and most control groups were waitlists, against which effects were larger than against placebo conditions^[4]. Small trials adding mindfulness or ACT components to ERP found no outcome advantage over standard ERP^[5]; more on that approach is in ACT therapists.

ERP deliberately provokes anxiety. NIMH notes that this initial anxiety creates a risk of dropping out, though compulsions decrease for most people who continue^[1].

Estimates of attrition vary with how dropout is defined. The 2019 review puts it at 25% to 30%^[5], while the Bergen authors cite a figure of about 15% across ERP studies, similar to other OCD treatments^[8]. In one sample of 202 adults with OCD, about 30% either never started recommended CBT or stopped early^[5].

Concerns that exposure makes symptoms worse have been examined; one analysis cited in the review concluded exposure therapy is safe and tolerable with minimal risk of harm^[5].

Who tends to respond less

Lower baseline severity and better insight predict more improvement in both adults and children^[5]. Some research finds poorer response when OCD centers on taboo thoughts or on "not just right" feelings rather than a feared consequence, and high family accommodation predicts worse outcomes in children^[5]. Severe depression may also blunt response, though even this group showed moderate to significant gains^[5].

The review describes ERP as underused despite its record^[5]. It cites a gap of roughly 10 to 17 years between OCD onset and the start of treatment^[5].

Clinician reluctance is part of the picture. In a large therapist survey, 37.3% agreed that conducting exposure was strenuous for them, 14.7% did not feel competent delivering it for OCD, and 10% cited fear of harming patients^[5]. Geography adds to it: one study found only 23% of communities in the lowest income quartile offered a specialized mental health treatment resource^[5].

Lower-cost models are being tested. In one trial, a stepped-care approach that began with bibliotherapy and moved to standard ERP only for non-responders was efficacious, with significantly lower patient costs than standard ERP^[5]. Workforce and cost data are in therapist shortage statistics and cost of therapy statistics, and how exposure-based CBT is used for a related condition is covered in CBT for social anxiety.