Study on Neuroimaging predictors of Outcome in ECT Patients

Severe depressive symptomatology occurs commonly in the population and is
associated with significant functional impairment. Depressive symptomatology is
seen in unipolar major depressive disorder (MDD) and bipolar disorder, type 1
as well as type 2. Many depressed patients can be treated successfully with
antidepressants, moodstabilizers such as lithiumcarbonate and/or psychotherapy.
However, the largest clinical trial in the United States showed that at least
one third of the patients do not recover even after four successive steps of
pharmacological/psychotherapeutic treatment, and that 21% of depressed patients
meet the criteria for chronic depression (duration longer than 2 years) (Rush
et al., 2006). Electroconvulsive therapy (ECT) is an important treatment option
for such medication resistant patients; depending on the response criteria,
about 40-60% of these resistant patients recover (Prudic et al., 1996; van den
Broek, de Lely, Mulder, Birkenhager, & Bruijn, 2004). However, this implies
that approximately 50% of these patients do not respond or respond
insufficiently. Furthermore, of the patients who initially recover on ECT,
40-80% relapse within half a year, which is in part dependent on the type of
continuation treatment given after ECT (Sackeim et al., 2001).

Predicting treatment outcome using neuroimaging techniques
Previous studies have shown that neuroimaging might be used to predict
treatment outcome. Functional magnetic resonance imaging (fMRI) and
quantitative electroencephalogram (qEEG) have been used to predict treatment
outcome to various antidepressant treatments at group level (for review see:
Arns & Olbrich, 2014; Pizzagalli, 2011) such as antidepressants (Arns et al.,
2015a, 2015b), rTMS (Arns et al., 2012) and ECT (ten Doesschate et al., 2014).
Furthermore, the cognitive side-effects of ECT have also been related to EEG
parameters at group level (Sackheim et al., 2000; Ten Doesschate et al, 2015).
However, these methods only allow inferences to be made at group level,
limiting the ability to use these findings in the clinical practice. Recent
advances in neuroscientific analysis techniques opened up the possibility of
predicting treatment outcome for individual patients and has potential to serve
as prognostic biomarkers to guide personalized treatment decisions. That is, in
a previous study using a machine learning approach, achieving remission after
ECT was accurately predicted by analyzing pre-treatment fMRI data (J. Van
Waarde et al., 2015). Similar findings have been reported using fMRI in smaller
samples for antidepressant pharmacotherapy (Fu et al., 2008; Korgaonkar et al.,
2014) and cognitive behavioural therapy in depression (Costafreda et al.,
2009). Other studies predicted the treatment outcome of rTMS and
pharmacotherapy using a machine learning approach to EEG data (Erguzel et al.,
2015; Khodayari-Rostamabad et al., 2013).
Although the use of neuroimaging as a prognostic biomarker in the treatment of
depression seems promising, findings need to be replicated and extended before
they can be put into clinical practice.

The primary objective of this study is to predict the antidepressant treatment
outcome of ECT in patients with severe depressive symptomatology using
neuroimaging. ECT itself is conducted as usual according to national guidelines
(Nederlandse Vereniging voor Psychiatrie, 2010). fMRI and EEG resting-state
data will be acquired within two weeks prior to initiation of the ECT course. A
machine learning approach will be used to predict remission rates. Thereby, we
replicate one of our previous studies in which we successfully predicted ECT
outcome using fMRI (J. Van Waarde et al., 2015). Furthermore, we aim to extend
these findings using EEG data. Compared to fMRI, EEG data can be acquired at
lower costs and the equipment for EEG acquisition is more widely available
throughout psychiatric clinics. Thereby, taking a machine learning approach to
EEG data bears great potential to become a widely used prognostic biomarker for
the prediction of ECT outcome.
The second objective of this study is to predict the cognitive side-effects of
ECT for individual patients. The ability to predict the potential cognitive
side-effects, weighted to the expected antidepressant efficacy, of ECT may
guide physicians and patients in determining the best course of treatment.
Finally, because ECT seems to change brain functioning, after a course of ECT
the neuroimaging parameters are repeated to explore for changes due to
treatment and at a follow-up period of three months after ECT.

We will use a prospective cohort design following a group of severely depressed
patients, indicated for a course of ECT. Depressive symptoms and cognitive
measures will be established at pre-ECT baseline and post-ECT endpoint. BDI-II
depression scales will be established biweekly, if the patient is capable to do
so. MRI and EEG measures will be performed at pre-ECT baseline, as well as
after the index-course of ECT (endpoint) and after three months (follow-up).

The burden of ineffectively treated depression is high. The burden of ECT is
considerable but is warranted because of successful
treatment, and its risk can be considered negligible. Importantly, only the
regular ECT-population will be recruited. The additional
burden for participating in this study is minimal and the additional risk
negligible. The additional
burden for participating in the neuroimaging study can be considered minimal,
and the additional risk for eligible candidates is
negligible.