Predicting treatment outcome in obsessive-compulsive disorder using neuroimaging biomarkers

Treatment for OCD is based on stepped care, in which patients initially receive
pharmacological treatment with selective serotonin reuptake inhibitors (SSRIs)
orpsychological treatment with cognitive-behavioral therapy (CBT) (van Balkom
et al. 2013). Both these treatments are effective, but 40-60% of patients do
not benefit sufficiently (Pallanti et al. 2002; Eddy et al. 2004). Although
multiple treatment steps ensure that the majority of patients do receive
effective treatment, approximately 50% of patients are exposed to one or more
ineffective treatments, and therefore to prolonged treatment trajectories and
avoidable disease burden, side effects, and risks. One of the main priorities
of the European Commission and National Institute of Mental Health therefore is
to move away from trial-and-error based clinical practice and develop
biomarkers that enable personalized treatment (Insel 2009; Olatunji et al.
2013). Yet, despite the efforts taken, there are still no reliable markers to
guide individual treatment decisions in psychiatry. Recently, using a machine
learning technique called support vector classification combined with
leave-one-out cross-validation, we discovered that a resting-state network
centered around the dorsomedial prefrontal cortex could predict recovery from
depression with 84% sensitivity and 85% specificity (van Waarde et al. 2015).
To optimize treatment for OCD and reduce the burden and costs associated with
unsuccessful therapy, we aim to discover treatment outcome biomarkers for OCD
by combining neuroimaging with machine learning methods. Similar to studies on
the prediction of treatment outcome, research on the longitudinal effects of
treatment are scarce. In OCD, the disbalance between the ventral and the dorsal
cortico-striato-thalamo-cortical circuit leads to increased anxiety, repetitive
behaviors and the inability to modulate responses (van den Heuvel et al. 2015).
The most common findings in neuroimaging studies investigating the effects of
treatment are decreased activity in the ventral circuits and increased activity
in the dorsal circuits (Thorsen et al. 2015). In addition to our other aim, we
will investigate the longitudinal effects of OCD treatment on functional and
structural neuroimaging. The analysis of CBT and SSRI-related changes at the
level of brain areas and circuits will provide more perspective on the
pathophysiology of OCD and the response to different treatments. In order to
relate the alterations in functional imaging to actual treatment-induced
changes instead of time-related changes or test-retest reliability, the
comparison with a healthy population is crucial.

Our primary objectives are 1) to discover and validate a treatment selection
fMRI biomarker for allocating OCD patients to CBT or SSRIs, and 2) to determine
the divergent longitudinal effects of SSRIs and CBT on functional and
structural brain measures in OCD.

Patients in the first cohort will be treated with SSRI or CBT to develop and
validate a treatment selection fMRI biomarker for allocating OCD patients and
to determine the divergent longitudinal effects on brain measures of treatment
in patients with OCD. Treatment will be performed as usual and in accordance
with the national guidelines. In the second cohort, patients will be allocated
to SSRI or CBT based on fMRI biomarkers identified in the first cohort.

The subjects will be randomized between SSRI treatment and CBT. Treatment is as
usual, consisting of a high dosed SSRI or group sessions CBT on a weekly basis
for 16 weeks.

As the participates will be treated according to the national guidelines, the
burden and risk of treatment will be the same as usual. The additional risk for
participation in this study is limited to MRI scanning, which can be considered
negligible. Randomization to SSRIs or CBT and participating in the neuroimaging
study imposes an additional burden to patients which can be considered minimal.
Given that our approach is expected to provide a large benefit for patients in
the future, we consider this additional burden well justified.