For clinicians

ERP vs. I-CBT for OCD: a clinician's guide to the evidence

Inference-based cognitive behavioral therapy (I-CBT) treats OCD through reasoning, not exposure. How it differs from exposure and response prevention (ERP), what the randomized trials actually found, and where I'd put it while the evidence matures.

What is I-CBT, and how does it differ from ERP?

Inference-based cognitive behavioral therapy (I-CBT) treats obsessive-compulsive disorder by correcting the reasoning that produces the obsessional doubt, without deliberate, prolonged exposure exercises. Exposure and response prevention (ERP) enters from the other end, through planned contact with feared thoughts and situations while the compulsion is deliberately not performed. They differ in where they enter the obsessional sequence, and in how much evidence stands behind each.

The cognitive models most of us trained on are appraisal-based. They work on how a person interprets an intrusive thought once it has already arrived. The I-CBT model, developed by Kieron O'Connor, Frederick Aardema, and colleagues, goes a step earlier and holds that the obsessional doubt itself is manufactured by a faulty reasoning process. The person crosses from what their senses and common knowledge actually show into an imagined possibility, and the imagined possibility comes to feel real. I-CBT calls that crossing inferential confusion, and the whole treatment aims at it.

The 2022 multicenter trial defines it as "a specialized psychological treatment for obsessive-compulsive disorder (OCD) without deliberate and prolonged exposure and response prevention (ERP) that focuses on strengthening reality-based reasoning and correcting the dysfunctional reasoning giving rise to erroneous obsessional doubts and ideas."

ERP needs less introduction. It's a graded plan of contact with feared thoughts, objects, and situations, with the compulsive response deliberately withheld, and the International OCD Foundation's treatment guide calls it "a first-line treatment" that's "recommended for most people with OCD," with a typical course of 12 to 20 sessions of about an hour each.

So I-CBT works at the formation of the doubt, and ERP works at the response to it. That single difference runs through the rest of the comparison, and the table below is mostly that difference worked out in practice.

What does the evidence actually show?

Randomized trials show I-CBT reduces obsessive-compulsive disorder severity. A 2022 multicenter RCT (111 participants) found no significant difference between I-CBT and appraisal-based CBT. A 2024 multisite trial (197 participants) against CBT that included ERP also found no significant between-group differences, but the confidence intervals exceeded the non-inferiority margin, so the result stays inconclusive.

Three randomized controlled trials carry this comparison, and each one gets closer to the question you actually care about.

Start in 2015. Visser and colleagues randomized 90 patients with OCD and poor insight to 24 sessions of CBT or 24 sessions of the inference-based approach (IBA), the treatment now usually called I-CBT. Both groups improved: "In both conditions, a significant OCD symptom reduction was reached, but no condition effects were established." The interesting part showed up in a post hoc subgroup. Among the 23 patients with the worst insight, those treated with the IBA improved significantly more, and the authors wrote that the approach "might be more promising than CBT for patients with more extreme poor insight." That's 23 patients, found after the fact. Hold it lightly.

Then 2022. Aardema and colleagues randomized 111 participants across three arms, I-CBT, appraisal-based CBT (A-CBT), and an adapted mindfulness-based stress reduction program as a non-specific active control. The paper's summary is one sentence: "All treatments significantly reduced general OCD severity and specific symptom dimensions without a significant difference between treatments." Before that result settles anything for you, hold onto two things. The comparator was appraisal-based cognitive-behavioral therapy, not ERP. And finding no significant difference between groups is a weaker claim than demonstrating equivalence, which is exactly the gap the next trial tried to close.

The 2024 trial by Wolf and colleagues was built to test equivalence head on. It randomized 197 participants to 20 sessions of I-CBT or CBT, with the CBT arm including ERP, and set a non-inferiority margin of 2 points on the trial's primary measure, the standard structured scale of OCD symptom severity. No statistically significant between-group differences turned up at any assessment point, but "the confidence intervals exceeded the non-inferiority threshold, making the results inconclusive."

The authors' conclusion deserves to be read whole: "While both CBT and I-CBT are effective for OCD, whether I-CBT is non-inferior to CBT in terms of OCD symptom severity remains inconclusive. Nevertheless, I-CBT offers better tolerability and warrants consideration as an alternative treatment for OCD."

Now set all of that against the ERP literature. The IOCDF's guide credits ERP with a "very strong evidence base," a "vast body of literature" that "has consistently demonstrated that ERP can produce large OCD symptom reductions," and reports that on average patients achieve a reduction of 60% of OCD symptoms.

That asymmetry is the summary, and it isn't a mark against I-CBT so much as a description of where the field is. I-CBT's trials are encouraging and few. ERP's are consistent and many.

Who might I-CBT suit?

The clearest candidates are patients who decline exposure-based treatment or leave it early. The 2024 trial noted that many patients refuse CBT or drop out partly because of anxiety about the ERP exercises, and it measured I-CBT as significantly more tolerable. The signals of benefit in overvalued ideation and poor insight come from secondary and post hoc analyses.

Every clinician who treats OCD has met the patient who wants help and won't go near exposure work, or who started it and left, sometimes more than once. The 2024 trial opens with that reality: "many patients refuse CBT or drop out prematurely, partly because of anxiety regarding exposure and response prevention (ERP) exercises."

The natural objection, that an exposure-averse patient wasn't going to do ERP anyway, is the reason the tolerability result matters. The trial measured tolerability rather than assuming it. Posttreatment tolerability scores were significantly higher with I-CBT, and the authors concluded that I-CBT "offers better tolerability." For the patient who has already declined exposure work, that means there's now a structured, manualized treatment they may actually stay in, and that's worth a real conversation.

The second candidate group is at the insight end of the spectrum. In the 2022 trial, "I-CBT led to significantly greater improvement in overvalued ideation, as well as significantly higher rates of remission as compared to MBSR at mid-test," and its authors listed "effectiveness for overvalued ideation" among the areas where I-CBT "may have an edge." In the 2015 poor-insight trial, the post hoc subgroup with the worst insight did better with the inference-based approach.

Both signals come from secondary outcomes and post hoc subgroups, which keeps them hypothesis-generating rather than established indications. It also makes them the corner of the I-CBT literature I'd watch most closely.

Is I-CBT a replacement for ERP?

No. Exposure and response prevention remains the first-line psychological treatment for obsessive-compulsive disorder, recommended by the International OCD Foundation for most people with OCD on the strength of its evidence base. I-CBT is a promising alternative with maturing evidence, worth discussing when a patient declines exposure-based care or can't stay in it.

None of the three trials found I-CBT superior to a CBT comparator on its primary outcome, and the one trial built to establish non-inferiority couldn't close the question on its own terms. The 2024 authors picked their word carefully, that I-CBT "warrants consideration as an alternative treatment for OCD," and nothing in these trials makes it more than that word says. An alternative is a different claim than a successor.

That said, this isn't a fringe idea anymore. The International OCD Foundation hosts an I-CBT special interest group among its treatment-guide pages, and the treatment is manualized in a clinician handbook written by its developers.

Recognition isn't an evidence base, though, and certainly not one the size of ERP's. The position that survives a case discussion holds both facts at once, that I-CBT is promising and that its evidence is still maturing.

How should clinicians decide between them?

A reasonable default is exposure and response prevention first, because its evidence base is deepest. Discuss I-CBT when a patient declines exposure work or has left it before, and present it accurately. Randomized trials support benefit and tolerability appears higher, but equivalence to CBT with ERP isn't yet established.

For what it's worth, I'd sequence rather than blend. ERP first when the patient will take it, since it's the first-line treatment with the deepest support and nothing in the I-CBT literature unseats that. I-CBT enters the conversation on specific facts about the patient in front of you. Exposure-based care refused or abandoned, or, more tentatively, marked overvalued ideation or very poor insight.

Whatever you offer, the patient deserves precision about what's known. A framing that matches the trials:

• Both treatments reduced OCD severity in randomized studies.
• Participants rated I-CBT more tolerable.
• Whether it works as well as CBT with ERP isn't settled.

A patient can weigh that without being sold anything.

Take two cautions with this, though. First, the trials tested each treatment delivered whole and on protocol. They say nothing about an improvised blend of the two, and I-CBT is its own manualized model, not ERP with the exposures taken out. Second, the larger problem is upstream of this choice. Most adults with even severe OCD who reach a professional aren't receiving treatment specific to OCD at all, a gap quantified in the treatment gap data. Either of these treatments, delivered properly, beats that default.

So the Monday-morning version is simple enough. Offer ERP first. When a patient needs an alternative, name I-CBT plainly, and say out loud what's settled and what isn't. And keep an eye on this literature, because it's moving.