ERP: The Treatment That Feels Backwards but Works (Part 4)

OCD says relief comes from avoidance or ritual. Exposure and response prevention reverses that pattern — gradually, safely, and with consent — so uncertainty becomes manageable.

Part 4 of a five-part series, OCD Across the Lifespan. Where this sits: OCD is closely related to anxiety and is filed here with our anxiety writing, but current diagnostic systems classify it separately. DSM-5-TR places it in obsessive-compulsive and related disorders, and ICD-11 gives it its own code, 6B20.

What to know

• ERP is collaborative, individualized, and one of the best-supported OCD treatments. “Best-supported” does not mean a guaranteed cure.
• SSRIs are first-line; clomipramine usually comes later. Verify treatment quality before using the word “resistant.” Some people need combined or intensive care.
• Certain FDA-authorized TMS devices carry an adult OCD add-on indication; devices and protocols are not interchangeable. Surgery and emerging options belong in specialist or research settings.

Zoom

Why the treatment can feel backwards

Avoiding a feared object, replaying a memory, washing again, or asking for reassurance can lower distress quickly. That relief teaches the brain, “The ritual protected me,” so the next doubt feels more urgent.

Exposure and response prevention (ERP) does not try to prove that nothing bad could ever happen. It helps a person approach a safe trigger while reducing the ritual that promises impossible certainty.

• Exposure means intentionally practicing contact with a safe situation, thought, image, sensation, decision, or uncertainty that OCD has made difficult.
• Response prevention means choosing not to perform the compulsion — or reducing it according to the plan. That includes hidden rituals such as mental review, self-reassurance, neutralizing, confession, internet research, or checking how one feels.

This is different from “face your worst fear.” Ethical ERP is gradual, consent-based, and built around an individual formulation. It never requires actual danger, illegal conduct, abandonment of necessary hygiene or medical care, or violation of a person’s core faith practice.

The aim is not to force anxiety down on schedule. Anxiety may rise, fall, or remain for a while. Progress means learning that the trigger can be handled without the ritual — and recovering time, choices, and participation in life.

What happens during ERP

1. Assessment and a shared map

The clinician asks about obsessions, visible and mental compulsions, avoidance, reassurance, family accommodation, insight, safety, and effects on school, work, sleep, relationships, and self-care. They also assess conditions that can change sequencing: depression, bipolar disorder, psychosis, substance use, eating disorders, autism, ADHD, and tic disorders.

2. A practice plan — not a surprise

Patient and therapist list restricted situations, then choose safe, feasible practices. They discuss pacing, consent, culture, disability, medical limits, and values.

Imagine Lena photographs her stove, then asks her partner to confirm it is off. A first practice might be one ordinary check, leaving without a photo, and allowing doubt to travel with her. Her partner follows an agreed response instead of supplying another guarantee. Later practices vary. The goal is to rejoin her day without building a larger checking system — not to feel perfectly certain.

3. Response prevention and new learning

The clinician identifies replacement rituals. If someone stops checking the stove but reviews it in memory, the loop continues. Practice includes declining visible and mental checking and learning: “I can act with ordinary care without total certainty.”

4. Practice between sessions

Planned practice at home, school, work, or in the community makes learning portable. The therapist reviews learning, not whether fear disappeared.

5. Measurement, maintenance, and boosters

Progress includes symptom measures, fewer rituals, less accommodation, and restored functioning. Relapse plans identify warning signs and useful practices; boosters can help when symptoms drift.

How long does it take? A standard outpatient ERP course is commonly 12 to 20 sessions, though intensive formats compress this and complex presentations may need more. Ask any prospective clinician what a typical course looks like in their practice and how they will measure whether it is working.

What the evidence actually shows

Randomized evidence supports ERP across ages. A 2024 U.S. Agency for Healthcare Research and Quality review of OCD in children found strong evidence versus waitlist and meaningful evidence versus active behavioral care; therapist-supported remote ERP also helped.

The adult evidence deserves an honest summary. A 2021 meta-analysis of 36 randomized trials found a large advantage for ERP over waitlist and over psychological placebo, a small and non-significant advantage over adequately dosed medication, and no detectable advantage over other active psychological therapies. It also flagged serious methodological limits: only a minority of trials were rated at low risk of bias, and trials with suspected researcher allegiance to ERP reported far larger effects than trials without it.

ERP is strongly supported against doing nothing. The honest claim against other active, well-delivered psychotherapies is that it is at least as good — not that it is proven better.

Children, families, school, and telehealth

For younger children, ERP may use drawings, games, rewards, shorter practices, and simple language. Parents reduce accommodation, support homework, and praise flexibility. In a randomized trial of 127 children ages 5 to 8, family-based CBT with ERP outperformed family-based relaxation (72 percent versus 41 percent response).

Schools can coordinate responses to checking, treatment time, and return to avoided tasks while preserving educational and disability rights.

Telehealth can bring a therapist into the home context. Pediatric comparative evidence is encouraging. In adults, however, a direct trial could not conclusively prove guided internet CBT was noninferior to face-to-face CBT, and unguided care performed significantly worse. Remote, guided ERP is not the same as an app generating exposures without clinical oversight.

Families working on the reassurance side of the plan will find the detail in Why Reassurance Makes OCD Worse (Part 3).

How medication fits

Selective serotonin reuptake inhibitors (SSRIs) and ERP are the main first-line options in international guidance. Choice depends on age, severity, preference, safety, prior response, and access.

Severe impairment or partial response often favors combined treatment. In children, the AHRQ review found ERP plus an SSRI better than an SSRI alone — but no better than ERP alone for average symptom reduction. Medication should not be automatic when competent ERP is working.

U.S. OCD labeling is age- and formulation-specific:

The formulation distinctions are not pedantry. Paroxetine CR is approved for depression, panic disorder, social anxiety disorder, and PMDD — but not OCD. Fluvoxamine CR carries an adult OCD indication only; the pediatric approval belongs to the immediate-release tablet.

Because OCD symptoms may improve slowly, prescribers often use longer trials and higher supervised targets than for depression. An adequate SSRI trial for OCD is generally at least 8 to 12 weeks, including several weeks at the target dose, before it is called inadequate. Partial improvement often appears around weeks 4 to 6 and continues to build.

Benefit can level off while adverse effects rise — dose-response research in OCD shows efficacy increasing and then flattening or declining while adverse-effect discontinuation keeps climbing. Never change a target dose yourself. A fair trial includes time, adherence, tolerability, interactions, age, organ function, and measured change. Average short-term adult SSRI benefit is modest: 2.65 Y-BOCS points over placebo across 11 regulatory trials in 2,372 adults, with about seven people needing treatment for one additional responder. That is roughly half the 5-to-6-point change usually treated as the smallest difference patients actually notice — an argument for combining treatments and measuring response, not for skipping medication.

Clomipramine can help, but dry mouth, constipation, sedation, cardiac and seizure risks, interactions, and overdose toxicity usually place it after SSRIs. Its pediatric maximum is weight-capped, and seizure risk is dose-dependent. Older indirect comparisons suggest greater benefit than SSRIs, and that advantage persists in some analyses even after bias correction — but direct head-to-head comparisons do not consistently confirm it.

All antidepressants carry an FDA boxed warning: antidepressants increased the risk of suicidal thoughts and behaviors in pediatric and young adult patients in short-term studies. In practice this means everyone under roughly age 25 warrants especially close monitoring, and everyone starting or changing an antidepressant should be watched for clinical worsening, agitation or activation, mania, and new or worsening suicidal thoughts — most intensively in the first weeks of treatment and after any dose change. Prescribers should also screen for bipolar disorder before starting, since an antidepressant alone can destabilize it. Discontinuation plans belong with the prescriber.

Pregnancy or breastfeeding requires a risk–risk discussion about prior response, untreated illness, pregnancy stage, infant exposure, and switching. Older adults need review of falls, sodium, bleeding, cardiac risk, dry mouth, constipation, cognition, anticholinergic burden, and other medicines.

Benzodiazepines are not a core OCD treatment. They may have another short-term indication, but they do not treat the central obsession–compulsion cycle.

Before calling treatment resistant

Partial response is common. Before adding a riskier treatment, ask:

1. Is the diagnosis correct, and has current safety been reassessed?
2. Were mental rituals, avoidance, reassurance, and accommodation identified?
3. Was ERP actually delivered with supported practice and response prevention — and for enough sessions?
4. Was the medication taken consistently for at least 8 to 12 weeks at a tolerated, appropriate target?
5. Are interactions, adverse effects, cost, transportation, privacy, language, or clinician access blocking care?
6. Are depression, bipolar disorder, psychosis, substance use, eating pathology, autism, ADHD, or tics changing the plan?

Often the next step is to improve ERP, add it to medication, combine treatments, or try another first-line medicine. After an adequate serotonin reuptake inhibitor course, adding ERP beat stress management in one adult trial; in another, adding ERP substantially outperformed adding risperidone — and in that trial risperidone was no better than placebo.

Adding an antipsychotic may help some adults after adequate SRI and ERP care, but none is FDA-approved as an OCD add-on. Meta-analysis of 14 double-blind trials in 491 people found a moderate pooled benefit (Hedges’ g = 0.64), with individually significant effects for aripiprazole, haloperidol, and risperidone, and non-significant results for quetiapine and olanzapine. Overall dropout did not differ from placebo, but discontinuation specifically due to adverse effects was more than twice as common on antipsychotic. Weight or blood-sugar changes, movement problems, hormonal effects, cardiac risks, and sedation require specialist monitoring. Evidence in children is sparse.

What is the role of TMS?

TMS uses magnetic pulses without surgery, but device, coil, brain target, schedule, provocation, age, and FDA status all matter.

On August 17, 2018, the FDA granted a Class II De Novo marketing authorization (DEN170078) to the BrainsWay Deep TMS System as an adjunct — an add-on — for adults with OCD. Its pivotal protocol used an H7 deep coil over the medial prefrontal and anterior cingulate network, clinician-guided symptom provocation, and 29 sessions. Brief provocation is not a full ERP plan or homework program.

In the FDA’s modified analysis, which did not include every randomized participant, active treatment improved Yale–Brown Obsessive Compulsive Scale (Y-BOCS) scores about 2.8 points more than sham. A response, defined as at least a 30 percent reduction in Y-BOCS score, occurred in 38.1 percent versus 11.1 percent.

Three limitations belong in the same paragraph. The analysis including everyone randomized was not statistically significant. Remission did not differ between groups. And controlled follow-up lasted only four weeks.

Later devices received 510(k) clearance based on substantial equivalence, often without a new sham-controlled OCD trial. Evidence does not transfer to every machine, coil, or faster schedule.

Meta-analysis finds short-term benefit on average but mixes protocols and has weak long-term evidence. In the United Kingdom, NICE guidance (HTG548, migrated from IPG676) remains research-only: it judged safety acceptable but efficacy evidence “inadequate in quantity and quality.” U.S. and UK systems ask different questions — a device authorization is not the same as a health-technology recommendation — so this is a difference in remit, not a factual contradiction.

TMS is a specialty add-on, not first-line care and not an ERP replacement. Headache or scalp discomfort is common; seizure is rare but possible. Programs screen for metal, implants, and seizure risk and use hearing protection. Travel, cost, and coverage matter; authorization does not guarantee benefit or payment.

Intensive care, DBS, and lesion procedures

An intensive outpatient or partial-hospital program can provide frequent ERP while a medically stable person returns home. Residential care can add a 24-hour therapeutic environment when home accommodation or severe functional collapse blocks treatment. Inpatient care may be needed for imminent danger, grave self-neglect, medical instability, psychosis or mania, or inability to perform basic care. The program should actually provide OCD-specific ERP; a bed alone does not retrain the loop.

For rare, profoundly disabling adult OCD, a highly specialized team may review deep-brain stimulation (DBS). In February 2009 the FDA authorized the Medtronic Reclaim system under a Humanitarian Device Exemption (HDE H050003). Its narrow indication covers bilateral stimulation of the anterior limb of the internal capsule, as an adjunct to medications and as an alternative to anterior capsulotomy, for adults with chronic, severe, treatment-resistant OCD who have failed at least three SSRIs.

HDE authorization rests on probable benefit outweighing risk for a small population; it is not ordinary proof of effectiveness, and HDE devices also require institutional review board oversight at the treating center. That is a signal this sits outside ordinary care. A 2025 individual-patient analysis of nine small sham-controlled trials in 91 patients favored active DBS, but certainty was rated low and brain targets varied.

DBS entails surgery, hardware, programming, and hemorrhage, infection, seizure, device, and psychiatric risks. Capsulotomy, cingulotomy, radiosurgery, laser procedures, and focused ultrasound create lesions; evidence is mostly observational, and some methods are irreversible or investigational.

Keep emerging care separate from established care

Off-label means using an approved product for a condition not on its FDA label; it does not mean FDA-approved for OCD.

Do not sell emerging options as cures. ECT is not standard treatment for core OCD, though it may be used for a separate urgent condition such as severe depression or catatonia.

Safety box — before you change anything

Do not start, stop, or change medication without the prescriber. Self-directed exposures are not appropriate when they involve actual hazards, severe instability, psychosis, mania, current suicidality, or a medically unsafe task. When an eating disorder or another condition makes a proposed exposure unsafe, use coordinated specialist care.

If danger is imminent, call 911 or go to the nearest emergency department. In the United States, call or text 988, or chat at 988lifeline.org, for any mental-health crisis.

What you can do next

Ask one prospective provider these five questions:

1. “Do you regularly assess and treat OCD?”
2. “Is ERP a central part of your work?”
3. “How do you measure progress?”
4. “How do you involve family when helpful?”
5. “What is your plan if first-line care is not enough?”

For a TMS center, add: “Which OCD device, coil, target, and protocol do you use, and what evidence and FDA authorization apply to that exact protocol?”

FAQs

Is ERP just flooding?

No. Ethical ERP is gradual, individualized, and consent-based. It targets safe triggers, never actual danger or surprise exposure.

Do I have to feel calm before an exposure is successful?

No. Anxiety can rise, fall, or remain. Success is practicing a different response and returning to meaningful activity without completing the ritual.

Is medication always needed with ERP?

No. Some people respond to ERP alone; others need combined care because of severity, comorbidity, preference, or partial response.

How long before an OCD medication is working?

Generally at least 8 to 12 weeks, including several weeks at the target dose. Partial improvement often appears around weeks 4 to 6. OCD typically needs longer trials and higher doses than depression, which is one reason medicines get abandoned too early.

When is TMS considered for OCD?

In the United States, a program may consider a specific FDA-authorized device and its labeled adult add-on protocol, usually after first-line ERP and medication. TMS is not universal first-line care, and neither benefit nor insurance coverage is guaranteed.

What does “treatment resistant” mean?

Definitions vary. Before using the label, clinicians should verify diagnosis, risk, adherence, sufficient supervised medication trials, competent ERP, family accommodation, comorbidities, and access barriers.

The rest of this series

• Start with the basics: No, You’re Not “a Little OCD” — What OCD Actually Is (Part 1)
• If thoughts feel too shameful to disclose: The Thoughts You’re Afraid to Say Out Loud (Part 2)
• If reassurance is feeding the loop: Why Reassurance Makes OCD Worse (Part 3)
• You are here — Part 4: ERP: The Treatment That Feels Backwards but Works
• If symptoms do not resemble the stereotype: The Many Faces of OCD You Might Not Recognize (Part 5)

Educational disclaimer

This article is for education and does not diagnose OCD or replace care from a licensed clinician. Do not start, stop, or change medication without your prescriber. If there is imminent danger or you cannot keep yourself or someone else safe, call 911 or go to the nearest emergency department. In the United States, call or text 988, or chat at 988lifeline.org, for any mental-health crisis. New or worsening hallucinations, delusions, mania, severe confusion, command hallucinations, major behavioral change, or inability to care for yourself or an infant requires prompt professional assessment — urgently or emergently when safety or basic care is impaired. Regulatory content — FDA label indications, device authorizations, and NICE positions — reflects records retrieved August 12, 2026.

References

1. National Institute for Health and Care Excellence. Obsessive-compulsive disorder and body dysmorphic disorder: treatment (CG31).
2. National Institute for Health and Care Excellence. Transcranial magnetic stimulation for obsessive-compulsive disorder (HTG548, migrated from IPG676).
3. Bandelow B, et al. World Federation of Societies of Biological Psychiatry guidelines for treatment of anxiety, OCD and PTSD, version 3, part II. World J Biol Psychiatry. 2023. PMID 35900217.
4. Steele DW, et al. Diagnosis and Management of Obsessive Compulsive Disorders in Children. AHRQ Comparative Effectiveness Review No. 276. December 2024. (Pediatric scope.) PMID 39836793.
5. Reid JE, et al. CBT with exposure and response prevention in OCD: systematic review and meta-analysis of RCTs. Compr Psychiatry. 2021. PMID 33618297.
6. Freeman J, et al. Family-based treatment of early childhood OCD (POTS Jr). JAMA Psychiatry. 2014. PMID 24759852.
7. Lundström L, et al. Effect of internet-based vs face-to-face cognitive behavioral therapy for adults with OCD. JAMA Netw Open. 2022. PMID 35285923.
8. Cohen SE, et al. Efficacy of SSRIs in obsessive-compulsive disorder: individual participant data meta-analysis of regulatory trials. Br J Psychiatry. 2025. PMID 40369939.
9. Xu J, et al. Optimal dose of serotonin reuptake inhibitors for obsessive-compulsive disorder. 2021. PMID 34630180.
10. Dold M, et al. Antipsychotic augmentation of serotonin reuptake inhibitors in treatment-resistant OCD: an updated meta-analysis. Int J Neuropsychopharmacol. 2015.
11. Simpson HB, et al. A randomized controlled trial of cognitive-behavioral therapy for augmenting pharmacotherapy in OCD. Am J Psychiatry. 2008. PMID 18316422.
12. Simpson HB, et al. Cognitive-behavioral therapy vs risperidone for augmenting serotonin reuptake inhibitors in OCD. JAMA Psychiatry. 2013. PMID 24026523.
13. U.S. Food and Drug Administration. De Novo Classification DEN170078 (BrainsWay Deep TMS), decision date August 17, 2018.
14. Carmi L, et al. Efficacy and safety of deep transcranial magnetic stimulation for OCD: a prospective multicenter randomized double-blind placebo-controlled trial. Am J Psychiatry. 2019. PMID 31109199.
15. Steuber ER, McGuire JF. A meta-analysis of transcranial magnetic stimulation in obsessive-compulsive disorder. Biol Psychiatry Cogn Neurosci Neuroimaging. 2023. PMID 37343662.
16. U.S. Food and Drug Administration. Humanitarian Device Exemption H050003 (Medtronic Reclaim DBS), approved February 10, 2009.
17. Cohen SE, et al. Deep brain stimulation for obsessive-compulsive disorder: individual patient data meta-analysis. Mol Psychiatry. 2025. PMID 40579425.
18. Moreno FA, et al. A randomized clinical trial of repeated doses of psilocybin for obsessive-compulsive disorder. J Psychopharmacol. 2026. PMID 41825921.
19. Bloch MH, et al. Effects of ketamine in treatment-refractory obsessive-compulsive disorder. Biol Psychiatry. 2012.
20. U.S. FDA prescribing information: Prozac, Zoloft, fluvoxamine maleate tablets, Luvox CR, Paxil, Paxil CR, Anafranil, Celexa, Lexapro.

If you or someone you know is in crisis

• Call 911 or go to your nearest emergency room for any life-threatening emergency.
• 988 Suicide & Crisis Lifeline — call or text 988, available 24/7. En español: marque 988 y oprima 2. Veterans: 988 and press 1, or text 838255.
• Crisis Text Line — text HOME to 741741.
• The Trevor Project (crisis support for LGBTQ+ young people) — call 1-866-488-7386, or text START to 678-678.
• National Sexual Assault Hotline (RAINN) — call 1-800-656-HOPE (4673) or text HOPE to 64673; free, confidential, 24/7. Online chat at RAINN.org/hotline.
• National Domestic Violence Hotline — call 1-800-799-SAFE (7233) or text START to 88788; 24/7, help in 200+ languages. Online chat at TheHotline.org. If your phone or computer may be monitored, calling from a safer device is an option.
• Riverside County — Inland SoCal Crisis Helpline 951-686-HELP (4357), 24/7 (Inland SoCal United Way / 211+, in partnership with RUHS-BH); Community Access, Referral, Evaluation and Support (CARES) Line 800-499-3008, 24/7.
• San Bernardino County — Access Unit (Behavioral Health Helpline) 888-743-1478, 24/7; Mobile Crisis/CCRT 800-398-0018 (24/7, all ages) or text 909-420-0560. Arrowhead Regional Medical Center (ARMC) has a dedicated walk-in adolescent psychiatric ER (ages 13–17).
• Children under 13 — call 911 for immediate danger, contact your county's mobile crisis team (they respond to all ages), or go to the nearest pediatric emergency room.
• California Peer-Run Warm Line (non-crisis — someone to talk to) — call or text 1-855-600-WARM (9276); daytime and evening hours, not a 24/7 line.
• NP Fady (non-emergency) — for routine scheduling or questions, call (909) 707-6261. This line is not monitored for emergencies.