www.nature.

com/npp

REVIEW ARTICLE
Breakthroughs and challenges for generating brain network-
based biomarkers of treatment response in depression
2,3,4 ✉
Sapolnach Prompiengchai1 and Katharine Dunlop

© The Author(s), under exclusive licence to American College of Neuropsychopharmacology 2024

Treatment outcomes widely vary for individuals diagnosed with major depressive disorder, implicating a need for deeper
understanding of the biological mechanisms conferring a greater likelihood of response to a particular treatment. Our improved
understanding of intrinsic brain networks underlying depression psychopathology via magnetic resonance imaging and other
neuroimaging modalities has helped reveal novel and potentially clinically meaningful biological markers of response. And while
we have made considerable progress in identifying such biomarkers over the last decade, particularly with larger, multisite trials,
there are significant methodological and practical obstacles that need to be overcome to translate these markers into the clinic. The
aim of this review is to review current literature on brain network structural and functional biomarkers of treatment response or
selection in depression, with a specific focus on recent large, multisite trials reporting predictive accuracy of candidate biomarkers.
Regarding pharmaco- and psychotherapy, we discuss candidate biomarkers, reporting that while we have identified candidate
1234567890();,:

biomarkers of response to a single intervention, we need more trials that distinguish biomarkers between first-line treatments.
Further, we discuss the ways prognostic neuroimaging may help to improve treatment outcomes to neuromodulation-based
therapies, such as transcranial magnetic stimulation and deep brain stimulation. Lastly, we highlight obstacles and technical
developments that may help to address the knowledge gaps in this area of research. Ultimately, integrating neuroimaging-derived
biomarkers into clinical practice holds promise for enhancing treatment outcomes and advancing precision psychiatry strategies for
depression management. By elucidating the neural predictors of treatment response and selection, we can move towards more
individualized and effective depression interventions, ultimately improving patient outcomes and quality of life.

Neuropsychopharmacology; https://doi.org/10.1038/s41386-024-01907-1

INTRODUCTION effective for this individual with that specific problem, and under
which set of circumstances?” [2, 3] Auden’s eulogy to his physician
and friend poignantly underscores the tension between these two
“‘Healing,’ concepts: the generalizations drawn from neuropsychopharma-
Papa would tell me, cological research and the need to treat the individual. It is this
‘is not a science, tension that contemporary precision medicine and biological
but the intuitive art marker research in MDD seeks to resolve.
of wooing Nature.” Today, MDD is diagnosed according to the DSM-5-TR when an
– W.H. Auden, The Art of Healing individual possesses five of nine symptoms, including low mood,
anhedonia, sleep and appetite disturbances, and suicidal ideation
Published in 1969, W.H. Auden’s dedication to his physician, Dr. [4]. Consequently, numerous possible symptom combinations of
David Protetch, was written on the brink of a paradigm shift for MDD exist; for example, over 1000 unique symptom profiles were
modern psychiatry and our understanding of major depressive identified in 3703 outpatients with MDD [5]. Furthermore, the
disorder (MDD). Psychopharmacologist Joseph J. Schildkraut had DSM-5-TR includes clinical specifiers, such as MDD with sleep
recently published his seminal 1965 work on the catecholamine disturbances, and previous editions include other specifiers like
hypothesis of affective disorders [1], which set the stage for a melancholic MDD [6]. Unfortunately, these specifiers do not
biological understanding of MDD and second-generation phar- currently yield homogenous MDD subtypes [7] nor differentiate
macotherapies over the next three decades. At nearly the same antidepressant responses to many interventions, including phar-
time, psychotherapist Gordon L. Paul wrote on the challenges of maco- [8–12] and psychotherapies [13–15]. Unsurprisingly, two-
trial research in psychotherapy: “[i]n all its complexity, the thirds of patients do not remit to their first course of
question towards which all outcome research should ultimately pharmacological treatment, and roughly one-third develop
be directed is the following: What treatment, by whom, is most treatment-resistance [16]. Such highly variable symptomatology

1
University of Toronto, Scarborough, Toronto, ON, Canada. 2Centre for Depression and Suicide Studies, Unity Health Toronto, Toronto, ON, Canada. 3Keenan Research Centre for
Biomedical Science, Unity Health Toronto, Toronto, ON, Canada. 4Department of Psychiatry and Institute of Medical Science, University of Toronto, Toronto, ON, Canada.
✉email: katharine.dunlop@unityhealth.to

Received: 29 March 2024 Revised: 17 May 2024 Accepted: 13 June 2024

 S. Prompiengchai and K. Dunlop
2
suggests that antidepressant response for MDD is complex, and monosynaptic and complex polysynaptic connections of the brain
clinical characteristics alone are insufficient to predict response to [26, 27]. By studying IBNs, we can understand how network-like
a certain intervention or distinguish responses to different brain organization either produces or constrains cognitive
treatments [8–15]. function. Likewise, we can compare how structural and functional
Functional and structural neuroimaging can help to identify brain networks are altered in psychiatric populations to elucidate
pre-treatment brain network-based characteristics of antidepres- pathophysiology and disease mechanisms. Pre-treatment inter-
sant response. Recent neuroimaging studies from large, clinical individual variability in IBN structure or function present in a
trials have advanced our understanding of biological markers— particular clinical population may help to identify features
biomarkers—in brain networks implicated in MDD pathophysiol- predictive of response and/or inform treatment selection.
ogy. Such markers may guide patients through first-line anti- Large-scale structural networks provide an anatomical organiza-
depressant treatments and for second-line treatments in a tion of the brain that underlies cognition. Briefly, brain areas or
treatment-resistant population [17]. For the purposes of this “nodes” are determined using anatomical parcellation techniques.
review, we distinguish between two types of biomarkers of For instance, a participant could undergo an sMRI scan, and the
antidepressant response: prognostic and prescriptive [2]. Prog- structural imaging data could be parcellated into different regions
nostic biomarkers are biological characteristics that differentiate using standardized atlas (for review, see ref. [28]). We can
dichotomous response and nonresponse or correlate with change characterize the morphological relationships between nodes using
on a continuous primary clinical outcome measure to a single structural covariance, which uses measures like cortical thickness
treatment. Prescriptive biomarkers are biological characteristics and gray matter volume to identify spatially distributed but
that differentiate responses or improvements on a primary morphologically similar nodes [29]. These nodes are structurally
outcome measure between two or more treatments. While connected by the axons in the white matter. We can call these
numerous studies report promising prognostic and, to a lesser fiber tracts connecting each node as “edges”. Edges are identified
extent, prescriptive biomarkers, there currently is a remarkable in vivo using diffusion tensor [30, 31] and diffusion spectrum
paucity of prospective clinical trials that are critically needed to imaging [32, 33].
bring neuroimaging biomarkers into clinical practice. Our understanding of the structural organization of the brain
Although the current literature consists of a considerable helps to infer possible functional interactions within and across
amount of heterogeneity in the study design and analysis, thereby structural networks [23, 26, 34]. In other words, interconnected
yielding numerous prognostic biomarkers [17–19], our narrative brain areas that work together for a particular set of cognitive
review describes recent advances in structural and functional functions are a functional network. The nodes of the functional
brain network-based biological markers of treatment response network for a particular function are typically inferred by
and selection in MDD with a particular focus on reviewing concurrent activation or deactivation [28]. The functional edges
candidate neuroimaging biomarkers with the highest quality of between each node represent the statistical relationship between
evidence such as large clinical trials. Therefore, we conducted a brain regions’ activity using the time series data from fMRI,
sensitive search for peer-reviewed papers in PubMed and Web of electroencephalography, and magnetoencephalography. The
Science using synonyms of unipolar depression, neuroimaging/ quantified statistical relationships between nodes can be undir-
connectivity, and treatments (e.g., antidepressants, psychother- ected (e.g., correlation-based metrics, such as seed-to-voxel-based
apy, and neuromodulation). We extensively discuss studies that functional connectivity [35]) or directed/causal (e.g., Granger
employed task- and resting-state blood-oxygenation-level- causality analysis [36] and dynamic causal modeling [37]).
dependent (BOLD) functional magnetic resonance imaging (fMRI), On the other hand, large-scale functional networks can be
structural MRI (sMRI), and diffusion tensor imaging. Wherever discerned either during task-related or resting-state functional
relevant, we also discuss studies using arterial spin labeling (ASL) neuroimaging. Task-based paradigms measure how the brain
perfusion MRI, positron emission tomography (PET) and electro- responds to the cognitive processes in question while the resting-
encephalogram (EEG). In the final section, we examine the state paradigms measure the spontaneous activity of the brain in
obstacles hindering the clinical translation of current brain- a task-free environment. Using functional neuroimaging para-
based biomarkers and discuss aspects of study design and digms, IBNs have been identified via hypothesis-driven and data-
analysis worth considering for future biomarker research in driven approaches. For instance, seed-based functional connec-
MDD. Although we focus specifically on biomarkers in unipolar tivity analysis extracts the time series of an a priori region of
MDD, the challenges and future directions highlighted by this interest (ROI), which is then correlated with the activity in other
review are likely applicable to biomarker research in other brain regions to identify which areas might be functionally related
psychiatric disorders. Ultimately, biomarker development efforts to the seed region. A more data-driven approach like independent
are urgently needed for people diagnosed with MDD as this component analysis decomposes the fMRI signal into a set of
disorder is characterized by overwhelming disease burden [20, 21] statistically independent components, each representing a net-
and a frustrating trial-and-error approach to treatment [22]. work of distinct patterns of brain activity [38].
Given the fundamental role of IBNs in cognition and its close
relation to structural and task-based functional networks, under-
BRAIN NETWORKS RELATED TO TREATMENT RESPONSE AND standing interactions within and between intrinsic networks is key
SELECTION IN MDD to elucidating the pathophysiology of psychiatric disorders and
Network neuroscience is an approach to characterize complex searching for network-based biomarkers based on mechanistic
large-scale interactions between spatially distributed brain understanding of treatment response [17, 39]. While a recent
regions. In humans, neuroimaging techniques are used to gain study yielded inconclusive results regarding the identification of
insights into our structurally and functionally connected neuro- neuroimaging biomarkers [40], networks including but not limited
biological systems. For one, fMRI captures correlated patterns of to the default mode network (DMN), salience network (SN), central
neural activity in widely separated brain regions that occur executive network (CEN), and ventromedial limbic/affective net-
spontaneously or are evoked by stimuli. These interactions work (VMN) have emerged as focal points for potential biomarker
between different regions of the brain can be used to delineate discovery in the context of major depressive disorder [39, 41]
several distinct large-scale intrinsic brain networks (IBN) serving (Fig. 1). We summarize key nodes within each IBN and their
various aspects of human cognition [23–25]. IBNs can be hypothesized functions in Table 1. Dysfunctional interactions
discerned during rest-state or while performing a cognitive task, within and between IBNs have been associated with various
and notably, IBNs have been hypothesized to reflect both depressive symptoms such as increased negative emotional and

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
3
second-line treatments. Given that the psychopathology of
a Default mode network r = 1.0 heterogeneous MDD involves complex perturbations of large-
scale brain networks, patients with the same MDD diagnosis may
have varying aberrant structural and functional connectivity. This
heterogeneity implies that patients with certain structural or
functional networks might be more suitable for a particular
treatment compared to other patients. As we summarize break-
throughs in using pre-treatment brain networks or early changes
in the brain as prognostic biomarkers, we note that different
x=0 y = –54 z = +26 r = 0.4
factors such as types of study design, analysis methods, validation
techniques, and sample size will influence the robustness and
b Central executive network r = 1.0 clinical translatability of the biomarker. For instance, studies that
prospectively validated their biomarkers in large clinical trials and
an independent dataset within the same report (i.e., using the
same computational pipeline) will likely yield a reliable biomarker.
While most current neuroimaging biomarkers have not yet
undergone a successful prospective validation, we have made
substantial progress in the past decade. Some prognostic
x = +36 y = +18 z = +46
biomarkers particularly for pharmacotherapy and neuromodula-
r = 0.2 tion have been validated in an independent dataset within the
same study or at least replicated across different large clinical
c Salience network trials, while there are relatively less neuroimaging biomarkers
r = 1.0
predictive of psychotherapy response.

Pharmacotherapy
Since two-thirds of patients do not respond to their first
antidepressant medication [16], many trials aim to improve
prognostic biomarkers for commonly prescribed drugs like
selective serotonin reuptake inhibitors (SSRIs; e.g., escitalopram
x = +4 y = +12 z = +28 r = 0.2
and sertraline), and serotonin and norepinephrine reuptake
inhibitors (SNRIs; e.g., venlafaxine). Early clinical studies identified
promising neuroimaging biomarkers, which classified responders
d Ventromedial affect network r = 1.0 from non-responders above 80% [52, 53]. However, studies had a
modest sample size (<30), which posed substantial challenges in
applying these markers prospectively in a heterogeneous MDD
population. Recognizing the need for robust biomarkers, we are
making rapid progress in finding more reliable biomarkers using
larger, multisite datasets. Very large clinical trials are the most
common in pharmacotherapy studies. For instance, antidepres-
sant drug trials with neuroimaging and N > 70 include Interna-
x = +6 y = +32 z = –16 r = 0.2 tional Study to Predict Optimized Treatment for Depression
clinical trial (iSPOT-D) [54], Canadian Biomarker Integration Net-
Fig. 1 Four Intrinsic Brain Networks Implicated in MDD. Resting- work in Depression (CAN-BIND) [55, 56], and Establishing
state functional connectivity (rsFC) map of default mode network Moderators and Biosignatures of Antidepressant Response for
(a), central executive network (b), salience network (c), and Clinical Care for Depression (EMBARC) [57]. The details of major
ventromedial affect network (d), generated using NeuroSynth clinical trials are summarized in Table 2.
[281]. Colormap displays brain regions that are correlated with the Resting-state functional connectivity (rsFC) of the DMN has
seed voxel indicated with crosshairs (MNI coordinates). Red and been frequently reported as candidate biomarkers to first-line
yellow represent the minimum and maximum Pearson correlations antidepressants [58–66] across wide-ranging clinical trial settings,
(r) respectively. The seeds are nodes of each network: posterior
cingulate cortex (a), right dorsolateral prefrontal cortex (b), dorsal
types of drugs, and analysis methodologies (e.g., seeds-based vs
anterior cingulate cortex (c), and subgenual anterior cingulate connectomic approach). For example, increased DMN intra-
cortex (d). The coordinates of the seed voxel were obtained from connectivity was found to be associated with responders or
[282] and confirmed using the term-to-coordinate mapping in remitters in different clinical trials [58–62] such as iSPOT-D [58, 61]
NeuroSynth [281]. and EMBARC [60]. One example of strong evidence supporting
increased pre-treatment DMN intra-connectivity as prognostic
cognitive bias [39, 42], rumination [43–45], impaired emotional biomarker is by ref. [58]. In this study, they randomized 75 MDD
regulation [46, 47], cognitive deficits [48], and anhedonia [49] participants to escitalopram, sertraline, or venlafaxine, and found
(Box 1). Based on the current evidence, distinct connectivity that rsFC between the posterior cingulate cortex (PCC) and rostral
patterns, notably within the salience and DMNs, might underlie anterior cingulate cortex (rACC)/medial prefrontal cortex (mPFC)
multiple phenotypic expressions [50] rather than one-to-one predicted remission with greater than 75% accuracy [58].
relationships between brain networks and symptoms as pre- Remitters had relatively intact PCC-rACC/mPFC rsFC that closely
viously hypothesized [51]. resembles that of healthy controls. While [58] used ROI-based
analysis, ref. [61] used network-based statistics and similarly found
that greater baseline functional connectivity within the DMN
PROGNOSTIC BIOMARKERS OF MDD TREATMENT RESPONSE predicted treatment remission regardless of the type of medica-
We will first consider prognostic biomarkers from major clinical tion [62]. However, this within-DMN rsFC is likely a treatment non-
trials for first-line antidepressant treatments, followed by trials for specific biomarker as it did not differentially predict remission by

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
4
Table 1. Intrinsic brain networks: associated brain areas and cognitive function.
Intrinsic brain networks List of brain areas Associated function
Default Mode Network (DMN) Midline/core network: medial prefrontal cortex (mPFC) and Internally-directed cognition, integration of
posterior cingulate cortex (PCC) other DMN subsystems, introspection [2]
Dorsal network: dorsomedial prefrontal cortex (DMPFC), Self-referential or affective processes, effortful
rostral anterior cingulate cortex (rACC) emotion regulation [272, 273]
Temporoparietal network: bilateral inferior parietal cortex Autobiographical memory retrieval and scene
(IPL), medial/lateral temporal cortices, temporoparietal reconstruction [2]
junction, precuneus.
Frontoparietal Central Dorsolateral prefrontal cortex (DLPFC), frontal eye fields, Cognitive control/flexibility, action planning,
Executive Network (CEN) and superior parietal cortex sustaining attention, working memory [274, 275]
Salience Network (SN) Dorsal anterior cingulate cortex (dACC) and anterior insula Detection/integration of salient stimuli and
autonomic/emotional information. [276]
Ventromedial Affect and Subgenual anterior cingulate cortex (sgACC), nucleus Reward learning and valuation, affect
Reward Networks (VMN) accumbens (NAc), medial OFC (mOFC), and ventromedial processing [49, 277]
prefrontal cortex (VMPFC)
Limbic and Autonomic Amygdala, hippocampus, hypothalamus Emotional memory, fear conditioning, anxiety,
Circuitry autonomic/endocrine response to emotion
[278–280]

EMBARC study [66]. Overall, there is strong evidence that

Box 1. Intrinsic brain networks implicated in biomarkers of increased DMN intra-connectivity and the inter-connectivity
antidepressant response involving nodes of DMN could be potential prognostic biomarkers
of pharmacotherapy response. However, it remains a challenge to
Dysfunctional intra- and inter-connectivity among IBNs have been associated with
symptoms of depression. In brief, DMN (Fig. 1A, Table 1) is a network of brain
resolve different inter-connectivity biomarkers found across
regions that are active when an individual is at rest or engaged in internally different study designs and analysis pipelines. The need to
focused tasks such as self-reflection, mind wandering, and autobiographical address the heterogeneity across studies and subsequently
memory retrieval [283, 284]. Hyperconnectivity within the DMN is associated with prospectively validate these biomarkers will be the recurring
excessive rumination and negative self-referential processing, which are hallmark
features of depression [43–45]. However, several studies did not find rumination to
theme throughout.
be predictive of DMN hyperconnectivity [50, 285]. CEN (Fig. 1B) is involved in Nucleus accumbens (NAc) connectivity, an important area for
higher-order cognitive processes such as working memory, cognitive control, and reward neurocircuitry, has also been identified as a candidate
decision-making [274, 275]. Hypoconnectivity within CEN and decreased inter- biomarker given that SSRIs and augmentation therapies act on
connectivity between CEN and VMN/limbic network are related to deficits in
cognitive control [48] and impaired emotional regulation observed in MDD
dopaminergic systems. One CAN-BIND report found that baseline
participants [46, 47] respectively. SN (Fig. 1C) initiates a switch from DMN (self- NAc rsFC with nodes of the SN (ACC) and DMN (precuneus, PCC)
referential processing) to CEN (externally-oriented cognition) in response to a correlated with anhedonia improvement for escitalopram non-
salient stimulus [286]. Thus, impaired interactions between SN and DMN/CEN in responders who received adjunct aripiprazole, a dopamine D2
MDD, and the inability to disengage with DMN, result in increased negative
emotional and cognitive bias and rumination [49, 279]. Furthermore, impaired
receptor agonist [68]. However, another study (N = 128 MDD) did
functioning of VWN (Fig. 1D), SN, and limbic system has been associated with not find any difference in the baseline NAc rsFC between
heightened emotional reactivity to negative stimuli, impaired reward processing, responders and non-responders, but an increase in right NAc-
and anhedonia in MDD [49, 287]. right dorsal ACC (dACC) rsFC from pre- to post-treatment was able
to differentiate responders from non-responders, regardless of
antidepressant type [69]. So far, our understanding of the neural
medication [58]. Consistent with the above findings, an early mechanism by which different drugs decrease anhedonia remains
decrease in DMN intra-connectivity also predicted better sertraline limited. Based on the current evidence, anhedonia improvement
response in the EMBARC study [67]. in MDD may have unique prognostic biomarkers that integrate
In addition to DMN intra-connectivity, between-network con- the striatal regions related to reward processing with the DMN
nectivity of the DMN could differentiate remission. Using a larger, and SN.
multi-site sample [55, 56] of CAN-BIND-1 participants treated with Furthermore, baseline structural connectivity measures such as
open-label escitalopram (N = 129), ref. [65] found greater baseline fractional anisotropy of predefined white matter tracts yield
SN anterior cingulate cortex (ACC)-DMN PCC rsFC among early promising biomarkers [70–72]. For instance, the cingulum
remitters relative to non-remitters [65]. Furthermore, greater connects important nodes of DMN and SN within the cingulate
insular connectivity with anterior and posterior DMN nodes could cortex (e.g., PCC, dACC, and rACC) [73] while stria terminalis
differentiate early from late remitters. Remitters also had weaker subserves the limbic/affective network (e.g., amygdala and
intra-connectivity within the SN (left-right insula), and DMN (PCC- hypothalamus) [74]. Their dysfunctional structural connectivity
right superior temporal gyrus) [65]. Extending these findings, the has been implicated in MDD [75–77] and treatment response [70].
connectomic approach uncovers potential brain connectivity Fractional anisotropy of the cingulum and stria terminalis could
beyond the canonical brain networks. For instance, remitters in predict antidepressant treatment response up to 74% when
the iSPOT-D study had greater inter-connectivity among DMN, combined with sociodemographic variables like age [70].
somatomotor networks, attention networks, somatomotor net- Combining structural and functional data may also improve
works, and fronto-parietal networks. Adding the neuroimaging prediction accuracies. Tian et al. [78] combined structural and
connectome biomarkers to clinical variables significantly functional data (N = 106) from an open-label escitalopram trial
improved their cross-validated model accuracy from 61.5% to and found whole-brain biomarkers that predicted 69–72%
68.8% [62]. High baseline connectivity between the DMN and accuracy [78]. Recently, multi-modal connectomes trained on
other nodes, and low connectivity between the executive network 184 MDD patients in a naturalistic antidepressant monotherapy
and the rest of the brain predicted general treatment response to setting (SSRI or SNRI) could predict post-treatment depression
sertraline and placebo for 200 MDD patients who completed the

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
5
severity of the external validation dataset (N = 26 MDD) at
76.92% accuracy, as compared to using structural connectome

non-responders, sertraline for

Yes, buproprion for sertraline
(73.1%) and functional connectome (65.38%) alone [79]. Newer

Yes, aripiprazole for non-

placebo non-responders
ways are being developed to quantify combined structural and
functional connectome (e.g., anatomically weighted functional
Adjunctive therapy?

connectivity [80]), which may bolster prediction accuracy.

Neuroimaging data has also been incorporated into a multi-
modal dataset that combined clinical, behavioral, and molecular
responders

biomarkers [81], but its predictive value in this study was limited
(accuracy: 57%).
So far, neuroimaging studies have reported prognostic
No

No
biomarkers either by using group-level statistics to measure
the differences between responders and non-responders or by
evaluating predictive accuracy of their machine learning models
or classifiers. Particularly for those studies evaluating predictive
Duration of

accuracy, different choices of preprocessing steps, first-level

treatment

12 weeks
8 weeks

8 weeks

8 weeks

statistical modeling, and machine learning methodologies could

yield drastically different prediction accuracies. For example,
there are different approaches to performing anatomical
parcellations, quantifying functional connectivity, identifying
the most predictive connectomic features, and classifying
Number of

responders and non-responders. Using CAN-BIND data, ref. [82]

tested the performance of 240 different models using either the
sites

baseline or the post-treatment resting-state connectome

20

(N = 144 MDD) and none were capable of predicting response

6

above chance [82]. Comparing different machine learning

classifiers on a pre-treatment sMRI dataset for an open-label
Double-blind (between
SSRI and SNRI), Open-

escitalopram trial (N = 79 MDD), ref. [83] found cross-validated

accuracies for predicting treatment response ranging from 0.46
(support vector machine) to 0.62 (random forest) [83]. However,
Blinding level

Double-blind

the classifiers trained on their dataset were not generalizable to

Open-label

Open-label

label (CBT)

an independent dataset, albeit for a different SSRI (sertraline).

These findings contradict smaller single-site studies reporting
>80% accuracy [52, 53] as well as the single-site iSPOT-D study
[58, 61]. Interestingly, ref. [82] reported that the change in
functional connectivity from baseline to week 2 showed a
predicted accuracy of up to 69.6% for the best-performing
Randomize?

model [82]. This aligns with quantitative electroencephalography

studies indicating that early changes predict antidepressant
response at >70% accuracy [84, 85]. Collecting neuroimaging
Yes

Yes

Yes
No

data at baseline and early on during the treatment may have

some clinical utility, as it may capture early changes in the brain
that might be predictive of a delayed symptomatic change [86].
Escitalopram, sertraline,

Sertraline and placebo

Thus far, we have made substantial progress in finding

duloxetine, and CBT
Treatment groups

baseline intrinsic functional connectivity as prognostic network-

or venlafaxine-XR

based biomarkers for treatment response to antidepressant

Escitalopram,
Escitalopram

pharmacotherapy. Interactions within DMN and inter-

List of large clinical trials for biomarkers discovery.

connectivity between DMN and the rest of the brain appeared

to be biomarkers that emerged from both seeds-based and
connectomic approaches, while the reward neurocircuitry might
play a role in predicting anhedonia response. Recent studies
have shown that combining structural and functional connec-
tomes, and novel computational techniques could improve the
predictive value of neuroimaging biomarkers. However, it is not
Pharmacotherapy and

clear which of these biomarkers is truly suitable for real-world

prediction of antidepressant response. Future studies should
Pharmacotherapy

Pharmacotherapy

Pharmacotherapy

rigorously validate the accuracy of our current biomarkers using

psychotherapy

a larger independent dataset to capture the heterogeneity

Treatment

of MDD.

Psychotherapy
Psychotherapy is a first-line psychological treatment for MDD
with comparable efficacy to pharmacotherapy [87–89], and
Clinical trial

CAN-BIND-1

includes cognitive behavioral therapy (CBT) and interpersonal

therapy. Like pharmacotherapy, 45% of MDD patients do not
EMBARC

PReDICT
iSPOT-D
Table 2.

respond to CBT [89, 90]. Yet, there is relatively less research on

prognostic biomarkers of psychotherapy treatment response
compared to that of pharmacotherapy [91].

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
6
Prognostic studies for psychotherapy in MDD have predomi- may reflect maladaptive functional compensation. Thus, these
nantly involved task-based fMRI during emotional processing or findings suggest that those who exhibit compensatory mechan-
reward learning. The rationale for these tasks stems from the ism might best respond to psychotherapy. These findings also
hypothesized mechanistic action of psychotherapy. For instance, support the notion that certain treatments may be best suited to
the development of cognitive-based therapies for depression is address subgroups of patients exhibiting similar symptomatotic
geared towards identifying and correcting dysfunctional thinking and rsFC profiles (i.e., behavioral activation treatment for
patterns implicated in processing negative emotional information individuals presenting with high anhedonia and normalized
and reward-based learning [92–94]. From a brain network MTG rsFC).
perspective, these cognitive processes relate to the dual-process Other psychotherapy biomarkers are consistent with the idea
model of emotional regulation: prefrontal emotional regulation that psychotherapy addresses maladaptive rumination and top-
regions inhibit limbic nodes implicated in emotional reactivity down emotional regulation. Straub et al. [122] found that higher
[91, 95, 96]. Specifically, nodes of the CEN (e.g., dorsolateral baseline amygdala-left DLPFC, amygdala/left anterior insula, and
prefrontal cortex or DLPFC) inhibit limbic/affective regions right sgACC-right DLPFC rsFC predicted greater CBT improvement
(amygdala) via the SN (dACC), and via VMN and DMN nodes [122]. Thus far, current literature suggests a few potential
implicated in emotional appraisal/evaluation like the orbitofrontal prognostic biomarkers for psychotherapy response possibly
prefrontal cortex (OFC), subgenual ACC (sgACC), and rACC involving (i) increased rsFC between SN (insula) and DMN (MTG),
[97–100]. Thus, for MDD participants, limbic areas such as the (ii) insula hypermetabolism at rest, (iii) hyperactivity in sgACC and
amygdala are hyperactivated in response to negative stimuli dACC (emotional regulation system) in response to negative
[101–104] likely due to decreased top-down control from dorsal emotional processing. However, future research should explore
prefrontal regions [105, 106]. data-driven and connectomic approaches using a larger sample
Based on this model, it is believed that pharmacotherapy and size to uncover depressive brain states that are responsive to
psychotherapy have distinct mechanistic actions. Antidepressant psychotherapy. Particularly with resting-state studies, it is unclear
drugs act via decreasing amygdala hyperactivity directly while if early brain changes can be used as prognostic biomarkers for
psychotherapy increases PFC hypoactivity, which has a top-down psychotherapy. It is also unknown whether these neuroimaging
effect on normalizing amygdala dysfunction [107]. Thus, we would biomarkers are general to every psychotherapy or simply
expect amygdala hyperactivity to be a prognostic marker for cognitive/behavioral therapy, or with improvements in specific
response to both psychotherapy and pharmacotherapy because symptoms. Lastly, since psychotherapy may reduce the risk of
both treatments ultimately normalize amygdala activity. While relapse or recurrence [123, 124], future research should address
converging evidence suggests amygdala hyperactivation during whether these biomarkers can predict long-term remission and
negative emotional reactivity to be a prognostic biomarker of relapses.
pharmacotherapy response [108, 109], such trends are mixed for
psychotherapy. Two studies have reported baseline amygdala Neuromodulation and second-line treatments
hyperactivity during negative emotional processing [110] or reward Treatment-resistant depression is typically defined as non-
learning [111] as predictive of response to CBT. However, most task- response to two or more trials of pharmaco- or psychotherapies
based studies report nonsignificant amygdala findings [112–115] [125, 126], and affects 35% of people diagnosed with MDD [127].
including a follow-up study [116] of Siegle et al. [110]. Neuromodulation therapies are indicated for those who do not
Alternatively, we would expect areas related to emotion respond to or cannot tolerate first-line treatments. These
regulation and evaluation to be a possible biomarker for treatments modulate brain activity using the delivery of a stimulus
psychotherapy. A few fMRI studies found that baseline hypoactiv- such as electrical currents in electroconvulsive therapy (ECT),
ity in sgACC [110, 116] and dACC [112] during emotional magnetic pulses in repetitive transcranial magnetic stimulation
processing tasks predicted psychotherapy response. Similarly, a (rTMS) targeting the DLPFC [128], or electrical stimuli via
PET study found that baseline sgACC hyperactivity sgACC at rest implanted electrode leads in deep brain stimulation (DBS). The
predicted non-response to CBT [117]. However, a recent CBT task- neuromodulatory effects of these treatments can locally affect the
fMRI study (N = 32 MDD) did not find any baseline emotion stimulation site and further propagate to other remote areas,
regulation or emotional reactivity-related brain activity signifi- targeting specific IBN implicated in MDD psychopathology [129].
cantly associated with depression improvement [115]. Although it Treatment responses to neuromodulation-based treatments vary
could be that more research with a larger sample size is needed to greatly across individuals [130]. Searching for neuroimaging
resolve the conflicting literature, newer models [118, 119] have biomarkers, particularly for neuromodulation-based treatments,
moved beyond a simplistic binary interaction between the will guide the development of neuromodulation techniques that
prefrontal cortex and limbic regions to better account for optimally engage with brain regions or circuits implicated in MDD
treatment mechanisms and biomarkers. based on identified biomarkers, and generate individualized care.
To date, there is relatively less research on rsFC predictors of This targeted and personalized approach increases the likelihood
psychotherapy treatment response. Investigating four canonical of therapeutic efficacy while minimizing potential side effects.
resting-state networks (DMN, SN, dorsal attention network, CEN), Pre-treatment sgACC connectivity is implicated in responses to
ref. [120] found that increased baseline right insula (SN)–middle numerous neuromodulation-based treatments. For DLPFC-rTMS,
temporal gyrus (MTG/DMN) rsFC was predictive of improvements baseline hyperconnectivity between the sgACC and DMN at rest,
in anhedonia following behavioral activation treatment, which is a but not between DMN and CEN, predicted response 10Hz-rTMS
psychotherapy aimed at improving engagement with rewarding over the left DLPFC [131, 132]. Of recent interest, greater
stimuli [120]. However, this biomarker did not predict improve- anticorrelation between the DLPFC stimulation site and sgACC is
ments in overall depression severity. Interestingly, this rsFC was associated with better outcomes to treatment [133–135]. How-
hypoconnected in all MDD participants relative to controls, but ever, one study reported that stronger baseline sgACC and right
individuals with relatively normalized right insula–MTG connec- DLPFC connectivity predicted non-response to high-frequency
tivity had worse baseline anhedonia and better response to rTMS over the left DLPFC [136]. Other groups have indicated that
treatment [120]. Although this study used resting-state fMRI, MDD this biomarker may be sensitive to modeling choices and
participants excessively recruit MTG during the generation of may have a relatively modest effect size [137]. Future studies
negative affect as a compensatory mechanism for inefficient are suggesting that symptom-specific manipulations of brain
activation of dACC and supplementary motor area during the activity via rTMS may yield more robust and personalized care
down-regulation of emotions [121]. Increased insula-MTG rsFC using this intervention [138, 139].

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
7
Similarly implicating the sgACC, DBS for MDD targets this expected rsFC map for each patient’s stimulation site. The
region, which alters blood flow locally at the stimulation site, but connectivity of each stimulation site is usually derived from a
also in other remote areas within the limbic and prefrontal large functional connectome dataset of healthy participants
networks that are functionally connected to sgACC [140–142]. This [160, 161] to improve the reliability of the connectomic
normalization of sgACC activity in TRD is associated with a biomarkers [159]. Preliminary evidence in retrospective datasets
reduction in depressive symptoms [140, 141]. Given this proposed suggests that a connectomic approach to defining regions
mechanism, DBS response is associated with the white matter functionally related to stimulation sites may help identify distinct
tracts and structural connectivity of the stimulation site [143]. targets for different clusters of depressive symptoms [162].
Individuals with DBS leads implanted near white matter tracts
connecting the sgACC to the mPFC, rACC, dACC, and subcortical
nuclei responded best to treatment. In a prospective, open-label, PRESCRIPTIVE BIOMARKERS OF MDD TREATMENT SELECTION
follow-up study, ref. [143] used this knowledge to select the Much progress has been made in finding prognostic biomarkers
optimal stimulation site using tractography, finding that 82% of that predict treatment response or remission in MDD. Many of
patients responded one year post-surgery [143]. This response those studies, however, have used reported prognostic biomar-
rate is higher than previous trials from the same group [144], kers based on a single intervention prescribed open-label or
demonstrating the potential of this marker in personalizing care relative to a placebo [65, 82, 163, 164]. While such prognostic
for this population. biomarkers lay the foundation for understanding responsive
Despite the stimulation non-specificity of ECT, this treatment depressive brain states specific to those treatments, they are
also implicates sgACC structure and function. Greater baseline limited in their clinical utility of guiding whether such biomarkers
sgACC volume predicts improved outcomes to ECT [145]. At rest, can differentially predict response better than other treatments.
greater variability in sgACC activity is correlated with ECT response, Relatively few neuroimaging studies compared more than two
which decreased after treatment [146]. Another study found that differing treatment options to search for prescriptive biomarkers
sequential changes in sgACC, DLPFC, and amygdala rsFC correlated that can predict the optimal type of intervention. However, many
with ECT improvement [147]. Furthermore, a classifier trained on a of those comparative studies were only able to find treatment
multimodal dataset with features from sMRI, BOLD rs-fMRI, and non-specific, prognostic biomarkers, rather than prescriptive
arterial spin-labeled fMRI predicted ECT response with a balanced biomarkers, for the treatment options in question. We summarize
accuracy of 68% [148]. Although the sample size in this study was below advances toward finding prescriptive neuroimaging bio-
modest (N = 46), the classification models consistently highlighted markers for MDD.
left DLPFC and sgACC, as well as the connectivity between motor Earlier studies involving modest sample sizes found baseline
and temporal networks around electrodes used in electroconvul- neuroimaging predictors between two different classes of
sive therapy (ECT), as predictive features. antidepressant medication [165, 166]. For instance, ref. [166]
In addition to sgACC, connectivity within and between different found that bupropion responders (n = 6) showed cerebellar
IBNs, such as DMN and CEN, has been implicated in non-invasive hypermetabolism, whereas venlafaxine responders (n = 7) exhib-
brain stimulation response [149–154]. For instance, within-DMN ited bilateral temporal and basal ganglia hypometabolism.
hypoconnectivity and connectivity between the DMN and CEN are However, these findings have been challenging to replicate. Even
associated with ECT response [151]. Van Waarde et al. [152] used a among larger randomized controlled trials for pharmacotherapy,
classifier trained on either resting-state fMRI or sMRI to predict ECT prescriptive biomarkers have been difficult to identify. For
remission in TRD [152]. Their approach identified two functional example, studies from the iSPOT-D trial sought to identify whether
networks that predicted remission at >80%: one network involving rsFC can be used as prescriptive biomarkers for optimizing
the mPFC, DLPFC, OFC, and PCC, and another centered on the treatment selection to escitalopram, sertraline, or venlafaxine-XR
dACC, DLPFC, sensorimotor cortex, parahippocampal gyrus, and [58, 69]. Both seed-based [58] and connectomic [61] approaches
midbrain. Notably, structural data was not significant predictors, did not yield any prescriptive biomarkers. Given the mechanism of
contradicting other studies that found structural biomarkers such antidepressant drugs is to modulate monoaminergic neurotrans-
as hippocampal subfields [155], striatum [156], and sgACC [145] to mission [167], identifying neuroimaging biomarkers with sufficient
be predictive of ECT outcomes. predictive value and specificity to individual classes of antide-
One limitation to identifying biomarkers of treatment response pressant medications remains a significant challenge.
is that MDD is a symptomatically heterogeneous condition. On the other hand, the EMBARC study yielded fMRI and EEG
Therefore, the connectomic signatures of MDD widely vary across biomarkers that differentially predicted sertraline and placebo.
individuals, making it potentially advantageous to incorporate Using the seed-based functional connectivity analysis, Chin Fatt
biological subtypes of depression with the predictive connectomic et al. [60] found that higher DMN-ECN interconnectivity predicted
biomarkers into classifiers for predicting response. Using this better sertraline and worse placebo response [60]. Higher
approach, ref. [157] found that classifiers based on both resting- interconnectivity of hippocampus with VMN/ECN/attention net-
state connective features and biotype diagnosis can accurately work, and limbic network with SN and somatomotor network
predict MDD responders to rTMS over the mPFC (90%), which predicted better placebo and worse sertraline response. The ASL
significantly outperformed classifiers based on the connectomic perfusion study did not find opposite-direction predictions but
features (79%) or clinical features (64%) alone [157]. They found found several regions in the DMN, SN, VWN, and limbic network
that areas across DMN (PCC, mPFC), CEN (left DLPFC), and limbic/ similar to the rsFC study that predicted treatment-specific
reward system (amygdala, OFC) had the strongest discriminant responses [168]. And while we would expect that early changes
effects [157]. These methods were recently updated to address in perfusion patterns might yield potential biomarkers similar to
limitations related to overfitting, feature selection, and multisite that of rsFCs [82] since cerebral blood flow has a higher signal-to-
variability in a sample of MDD patients. We found that symptom- noise ratio than resting-state fMRI, a recent perfusion study did
RSFC and subtypes were not only stable and generalizable in not find any perfusion patterns predictive of sertraline response
unseen data using an updated approach, but also stratified [169]. However, ASL may emerge as an important modality when
individuals treated with rTMS targeting the DLPFC by response considered with other neuroimaging biomarkers since ASL data
[158]. Another approach is to cluster symptom-response maps and contributed the most to the recent multimodal MRI biomarker
use individual depression symptoms since most studies used total predicting sertraline-specific response [170].
depression scores [133–135, 159]. The symptom-response maps Furthermore, an initial attempt to find connectomic prescriptive
are the correlations between the symptom changes and the biomarkers using the same EMBARC dataset did not succeed [66].

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
8
However, ref. [171] used individual-specific rsFC from the EMBARC role in initiating network switching between DMN and CEN [39].
dataset and successfully found connectomic signatures that were However, when the clinical utility of the prescriptive biomarker
specific to either sertraline (e.g., DMN-somatomotor network was prospectively tested by using the anterior insula metabolism
connectivity) or the placebo arm (e.g., somatomotor network and to assign MDD participants to either CBT or escitalopram, ref. [186]
visual network connectivity) [171]. In other words, models trained did not find anterior insula beneficial in predicting remission rates
on the sertraline arm did not predict placebo treatment outcome (38%) [186]. The findings emphasize the importance of prospec-
(r = 0.00, p = 0.97) and vice-versa (r = 0.08, p = 0.55). Furthermore, tively validating current biomarkers and integrating our under-
in comparison to using the raw rsFC data, the individualization of standing of large-scale brain networks in depression into the
FCs improved the prediction power and changed connectivity development of novel network-based biomarkers.
weights of important nodes of DMN (middle/inferior temporal Despite the extensive body of research dedicated to under-
cortex) and SN (insula) in sertraline; the left superior temporal standing the etiology, pathophysiology, and treatment mechan-
cortex and right middle cingulate cortex in the placebo [171]. ism of MDD, the identification of neuroimaging biomarkers
These regions were also the most predictive of sertraline and capable of guiding personalized treatment approaches is yet to
placebo response respectively, suggesting that individualized be successful. Based on the current literature, potential prescrip-
precision functional mapping might be more sensitive to subtle tive biomarkers for predicting different classes of pharmacother-
differences in varying connectomic features that might contain apy, or between pharmacotherapy and psychotherapy could be
critical information for predicting treatment selection. Addition- based on interactions among DMN (e.g., lateral temporal cortex,
ally, EEG represents a more accessible modality that could precuneus), SN (anterior insula), and VMN (e.g., sgACC). However,
potentially inform treatment selection. Similar to fMRI, most EEG many of the biomarkers identified for the same intervention but in
studies could not find reproducible, prescriptive biomarkers different clinical trials consist of many overlapping brain regions
[172–174] partly because EEG faces the challenge of signal within the well-known IBN implicated in MDD, likely reflecting the
smearing [175], the risk of overfitting high-dimensional EEG [176], need to account for heterogeneity in the MDD participants, study
and optimizing feature identification [177]. However, a recent design, eligibility criteria, operationalization of treatment out-
EMBARC study developed a novel rs-EEG computational model comes, and computational methodologies in analyzing the data
and found the latent signatures from alpha frequency range rs- and calculating the predictive utility of biomarkers. These
EEG data that were specific to either sertraline or placebo [178]. challenges are extensively discussed in the next section below.
Similarly, the CAN-BIND group was able to classify open-label
escitalopram response using baseline and treatment week 2 EEG
recordings with roughly 80% accuracy [179]; this finding is FUTURE DIRECTIONS AND CLINICAL IMPLICATIONS
currently being prospectively validated (NCT05017311). Lastly, The reviewed studies demonstrate that significant strides have
EEG-based biomarkers have been prospectively replicated in been made over the last decade toward unraveling the complex
patients treated with agomelatine/ALTO-300 [180]. These results underpinnings of antidepressant response, with the strong
point to EEG as an accessible, and potentially effective tool to potential of informing personalized treatment strategies. However,
identifying and validating biomarkers of existing and novel challenges remain on the path to clinical translation. For example,
treatments. the therapeutic window defining treatment response can vary
Another potential avenue for treatment selection is to compare considerably from one biomarker study to the next (discussed in
two different interventions with highly divergent hypothesized [17]). Response and remission rates also differ slightly depending
mechanisms: pharmacotherapy and psychotherapy. The most on the therapeutic window considered, with longer trials having
compelling biomarkers thus far come from the PReDICT study, higher rates [187]. Therefore, it is unsurprising that biomarkers of
where participants were randomized to either CBT or pharma- response qualitatively and substantially differ when investigating
cotherapy [181]. Dunlop et al. [182] found baseline sgACC rsFC 6-week or 12-week outcomes [17]. Another area of concern is
with the left frontal operculum, left ventromedial prefrontal publication bias, which further limits our ability to independently
cortex, and dorsal midbrain as prescriptive biomarkers for CBT and validate and prospectively use biomarkers. Open data, and pre-
pharmacotherapy treatment selection [182]. They then summed registered, freely-available method will be paramount to further-
up the mean functional connectivity between SCC and each of the ing our confidence in efforts to translate findings from the scanner
three regions. Positively summed rsFC accurately predicted CBT into the clinic [188, 189].
remission (78%) and pharmacotherapy non-remission (75%) while In this section, we highlight considerations on study design,
negatively summed rsFC accurately predicted CBT non-remission methodological, and implementation that, if addressed in future
(89%) and pharmacotherapy remission (72%) [182]. These findings work, will improve our understanding of and aid in implementing
suggest that baseline sgACC rsFC could potentially be used to biomarkers of treatment response and selection.
recommend CBT or pharmacotherapy, but more replication and
prospective validation must be done before implementing this Clinical trial design
biomarker in real-world psychiatric care. Study design can significantly influence response and remission
In addition to the rsFC, resting-state metabolic activity rates to antidepressants. For example, response and remission rates
measured using PET has also been explored. In a randomized are higher in open-label relative to placebo-controlled trials [187],
controlled trial of CBT and escitalopram (n = 38), ref. [183] found likely due to expectancy effects [190]. Furthermore, the choice of
six regions in the SN (right anterior insula), DMN (right inferior intervention impacts willingness to participate in research and
temporal cortex and precuneus), limbic network (left amygdala), therefore may bias participant selection, as individuals are willing to
and somatomotor network (left premotor cortex and right motor participate in head-to-head or placebo-controlled studies using
cortex) as potential prescriptive biomarkers [183]. Of note, the psychotherapy over head-to-head drug trials [191]. It is therefore
right anterior insula showed the highest discriminative ability, with likely that these aspects of study design could impact the
insula hypermetabolism being associated with escitalopram identification and validation of biomarkers of response.
remission and CBT non-remission, and insula hypometabolism Researchers have used a variety of different study designs in an
being associated with CBT remission and escitalopram non- effort to identify biomarkers of antidepressant response. For studies
remission [183]. This finding, along with studies showing whose primary aim was to evaluate biological modulators of
heterogenous insula metabolism profiles among MDD participants response, some groups, such as EMBARC [57], used a placebo-
[184, 185], led to the hypothesis that anterior insula metabolism controlled study of a single SSRI, while others, like PReDICT [181] and
might be a viable treatment selection biomarker [119] due to its iSPOT-D [54], randomized participants into one of three active

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
9
treatment arms. Others, including EMBARC and CAN-BIND-1 [55], Furthermore, more inclusive study criteria may help converge
have used a staged (mono- and adjunctive therapy) approach, our biomarker study efforts across mental illnesses, especially for
which more closely mirrors the trial-and-error approach to therapies that are indicated for multiple disorders other than
antidepressant treatment and other studies evaluating efficacy unipolar MDD. For example, researchers have reported convergent
[22]. Heterogeneity in study design could negatively impact our mechanisms and predictors of psychotherapy and pharmacother-
ability to pool studies for meta-analysis [192], potentially limiting the apy response in mood and anxiety disorders despite strong
generalizability of findings. Reassuringly and in contradiction to this between-study heterogeneity [91, 208]. Future research should
notion, recent efforts pooling CAN-BIND, PReDICT, EMBARC, and consider convergent and divergent biomarkers of treatment
other trial data have identified sMRI dimensions of MDD that response considering the high comorbidity of mental illness and
stratified response to SSRIs and placebo [193]. Regardless, the transdiagnostic indications of many psychotropic
idiosyncrasies in study design will need to be carefully considered interventions.
in upcoming efforts to pool studies in larger consortia including the
Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA) Defining treatment response
Consortium [194, 195] and the COORDINATE-MDD Consortium [196]. Defining treatment response, the dependent variable, is not trivial.
Moving forward, research is needed to clarify the impact of trial Studies frequently acquire a clinician-rated or self-reported
design on biological markers of treatment response, and different depression severity scale, like the Hamilton Rating Scale for
study designs may be better equipped to answer specific questions Depression [209] or Beck Depression Inventory [210], pre-
on the robustness or utility of candidate biomarkers. For one, the treatment, at regular intervals during treatment, and post-
goal of placebo-controlled trials is to unravel the treatment-specific treatment to track improvement. Selecting which scale and how
biomarkers, while controlling for psychophysiological factors con- to assess response (i.e., as a continuous or dichotomous variable),
tributing to the placebo response [197]. These comparisons between as previously discussed [17], varies considerably across trials,
biomarkers of active and placebo are helpful both from a limiting our ability to generalize findings.
mechanistic and prognostic perspective. That said, placebo Further complicating this issue, these scales often evaluate
responses are high but variable across a variety of different multiple symptom dimensions [211]. This likely negatively impacts
interventions [198], and will be present in prospective, real-world our ability to detect biomarkers of response and distinct
translation efforts. Some have argued that the placebo response biomarkers of specific treatments, when evaluating the change
could be harnessed to bolster treatment efficacy or refine treatment in total score on these scales due to heterogenous baseline
options, which would be particularly relevant to biomarkers research symptom profiles and symptom-specific change. Confirming this
[197]. Indeed, both biomarkers of antidepressant and analgesic notion, specific pre-treatment symptom clusters are associated
placebos are associated with networks pertinent to MDD, including with differential response to both first- and second-line treatments
the SN and DMN [171, 199], so more research is needed to [158, 212]. Furthermore, researchers have shown that different
understand the clinical utility of these markers. interventions yield distinct patterns of symptom-specific response
Another area needing more investigation is research comparing [213–215]. For example, ref. [216] found that individuals treated
two or more active treatments to better understand predictors of with escitalopram or duloxetine experienced greater improve-
specific interventions and aid in treatment selection. While this ments in melancholic symptoms relative to those treated with
has shown helpful in disentangling biomarkers of psycho- vs. psychotherapy [216]. Second-line treatments for MDD may also
pharmacotherapy [117, 182, 183] and has ethical benefits by impact specific symptom dimensions, including ketamine [217]
randomizing patients to receive at least one efficacious treatment, and rTMS [162]. Given that specific symptom domains have
there are relatively few direct comparisons of predictors to specific distinct underlying biology [157, 158, 218], it is possible that
drugs. Furthermore, studies investigating active interventions change in these domains are a fruitful way forward to identifying
have randomized participants into arms irrespective of patient treatment-specific and robust biomarkers.
preference. This line of research suggests that patient preference How else might we characterize response? Incorporating
does not influence overall treatment efficacy but does impact multiple scales may be one path forward, particularly if they
dropout rates [200, 201], and we do not yet know whether or how assess symptoms not fully realized in one scale (e.g., anhedonia in
patient preference impacts biomarkers of response/selection. the Hamilton Depression Rating Scale [209]). Alternatively, quality
Partially randomized patient preference trials could improve the of life and functional impairment scales are commonly acquired in
internal and external validity of biomarkers by assigning patients biomarkers studies but typically considered a secondary outcome
their preferred treatment, which would emulate real-world measure. Scales such as the Sheehan Disability Scale [219] could
treatments while retaining response and remission rates. also yield important biomarkers of clinically meaningful improve-
ment, especially if aggregated across many trials. Early improve-
Study population ments in functional impairment predict later depressive response
Another factor to consider is study population. To date, most large [220, 221], and structural and functional connectivity related to
biomarker trials recruit adults diagnosed with unipolar MDD functional impairment in MDD implicates nodes of depression-
[54–57], with fewer biomarker trials occurring in adolescence and related networks [222, 223]. Another consideration is identifying
in later life. Studies within the unipolar space vary considerably on biomarkers predicting distinct trajectories of treatment response,
inclusion/exclusion criteria, particularly on factors like antidepres- which will consider all available timepoints. Linear growth mixture
sant treatment history, comorbid mental illness, and a depression modeling is one method that stratifies subgroups of participants
severity cutoff [202, 203]. Nearly all studies exclude individuals by trajectory, which has shown high prediction accuracy for
with a high suicide risk, but the methods to determine risk differ response to psychotherapy in late-life depression [224] and has
(e.g., clinician judgement, score on scales acquired during been used to characterize response to psychotherapy and
screening). While these criteria help to ensure a homogenous pharmacotherapy in adults [225, 226]. Early work incorporating
sample/population, comorbid mental illnesses are common trajectories with neuroimaging is promising [227, 228].
[204, 205] and, therefore, limit the external validity of biomarkers.
Given that individuals suffering from comorbid mental illness Neuroimaging and modeling considerations
experience more MDD recurrence, greater treatment resistance, Another nontrivial consideration is selecting the neuroimaging
and slower recovery than those without [206, 207], future studies paradigm and acquisition parameters. First, there is considerable
should “lean in” than “shy away” from this aspect of clinical variability in biomarker modality across studies (e.g., sMRI, resting-
heterogeneity to maximize external validity. state and/or task-based fMRI), again limiting our ability to

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
10
generalize findings. Resting-state and sMRI are popular options, as using normative modeling [263]. The rationale for this approach
they are relatively simple to acquire and do not rely on task lies in the notion that patient heterogeneity should be linked to
compliance or performance. However, interpreting functional biological dysfunction [50, 51]. However, it is possible brain-based
connectivity unconstrained by a task is complicated by the fact features correlated with symptom heterogeneity and antidepres-
that individuals may experience the resting-state scan differently, sant response need not be atypical relative to controls
and this variability is inadequately accounted for [229]. Further, [120, 158, 264]. Alternative methods to characterize structure or
task-based fMRI may be better suited to characterizing inter- function in a patient sample relative to controls, such as brain
individual differences in behavior over resting-state. In a aging [265] or approaches that exploit larger control databases
nondepressed sample, task-based functional connectivity is [162], may also help to distinguish prescriptive markers of
superior at characterizing variability in a number of cognitive response.
domains, including reading comprehension [230], attention [231],
behavioral inhibition [232], and fluid intelligence [233, 234]. In
MDD, tasks that have been used to distinguish antidepressant CONCLUSION
response have included tasks related to emotional reactivity and To conclude, significant progress has been made over the last
social cognition [109, 112], appetitive and avoidance behaviors decade to identify biologically based predictors of antidepressant
[111, 235–237], emotion regulation [115, 238, 239], and working treatments. That said, inability to replicate candidate biomarkers
memory/cognitive control [240–242]. Another untapped area is and considerable variability in model performance by study
passive movie watching, which outperforms resting-state fMRI in design, preprocessing, and analysis remains a significant challenge
predicting individual differences in cognition and emotional traits in this field [81, 82, 186]. Studies optimally designed to compare
[243] without risking omitting data due to task performance. treatment-specific biomarkers, with pre-registered methods, will
Lastly, alternative modalities, such as ASL or EEG, may yield robust be paramount to uncovering rigorous prescriptive markers of
or scalable biomarkers of treatment response or selection. It is response.
therefore possible that such modalities or using task-based or Even once replicable and robust markers are found, scalability
naturalistic viewing (movie watching) may be preferred functional and equitable access to care are significant impediments to
methods in future biomarker efforts. translating any neuroimaging biomarker on a large scale [266]. For
Due to the inherently low signal-to-noise ratio of fMRI data example, Canada has one of the lowest MRI scanners per capita
[244], it also remains a challenge to use individual data rather than and longest wait times amongst high income countries [267, 268].
the averaged group-level imaging data to create individual- In the United States, delays in acquiring an MRI are associated with
specific functional connectomes for a personalized targeting individuals with public insurance, who identify as female, or live in
approach. Newer scan acquisition parameters, such as multi-echo a low socio-economic neighborhood [268]. It will be imperative to
fMRI, may also yield novel insights. Multi-echo fMRI acquires advocate for equitable access to care as neuroimaging biomarkers
multiple echoes, imaging readouts, per volume, which help are integrating into clinical practice. Further, identifying analogue
distinguish blood oxygen-level dependent signal from nonneur- and more accessible markers of response, perhaps using
onal noise [245–247]. Optimally combining and denoising these molecular, genetic or ecological momentary assessment measures
multiple echoes [248] has been shown to improve the signal-to- [81, 269] may help improve the scalability of markers of response
noise ratio and reliability of findings [249–251], and improve the to first-line interventions. It could be that MRI-based biomarkers in
interpretability of task-based fMRI results [252, 253]. Multi-echo psychiatry, at least at first, are reserved for second-line treatments
may also improve test-retest reliability of precision functional like DBS where pre-surgical tractography can be used to
mapping [254, 255], a method modeling individualized rsFC maps individualize care.
that can provide clinically meaningful variation in cognition and Despite these challenges, there are exciting prospective studies
behavior [256–258]. Further, machine learning approaches should currently recruiting. For example, Optimized Predictive Treatment
carefully consider individual cases where the model fails. Greene In Medications for Unipolar Major Depression (OPTIMUM-D;
and colleagues [259] recently demonstrated, in healthy samples, NCT05017311) is a follow-up to CAN-BIND, and seeks to test a
that misclassification in models predicting individual heterogene- previously identified prescriptive biomarker [270]. In the study,
ity in behavior using fMRI is non-random, related to specific participants will be assigned to one of two arms. The active,
phenotypic profiles and generalizable across datasets [259]. Future personalized arm will receive open-label escitalopram and
biomarker studies should similarly assess model misclassification placebo-controlled brexpiprazole; active brexpiprazole or placebo
cases to better understand the robustness and refine predictive will be assigned using a predictive algorithm. The placebo arm will
models. Other considerations include a larger sample size, be randomized to receive escitalopram/placebo or escitalopram/
evaluating biomarkers using an independent cohort, and validat- brexpiprazole. Using a similar trial design, Biomarker-guided rTMS
ing the predictive accuracy across different models (e.g., parcella- for Treatment Resistant Depression (BioTMS; NCT04041479) will
tions) and algorithms. assign different rTMS targets (DLPFC or mPFC) based on rsFC-
Another important factor is how to model brain-based based subtype assignment in treatment-resistant depression
measures. There is substantial variation across studies in [271]. These studies, if successful, will be monumental to reducing
preprocessing methods, and statistical analyses, and these the frustration of treatment, by either enabling patients to bypass
modeling decisions should be carefully considered. Systematically monotherapy trial-and-error and proceed with adjunctive treat-
testing these factors is one employed solution, particularly for ment, or personalizing stimulation sites based on co-occurring
specific resting-state preprocessing steps (e.g., global signal symptoms and rsFC. Time will tell if these studies are indeed
regression) and parcellation selection [135, 260]. Large, whole- successful, but they are certainly on the right track to integrating
brain univariate correlations have been criticized due to limita- neuroimaging data to help inform the “intuitive art of wooing
tions in statistical power, multiple comparisons, and reproduci- Nature” in the context of treating MDD.
bility [261]. Multivariate whole-brain associations might be one
way forward to developing predictive models without requiring
tens of thousands of individuals in the sample [262], and we REFERENCES
recently modeled the impact of sample size on the robustness of 1. Schildkraut JJ. The catecholamine hypothesis of affective disorders: a review of
multivariate methods in a depressed sample [158]. Another supporting evidence. Am J Psychiatry. 1965;122:509–22.
proposed approach is to identify biomarker regions exhibiting 2. Cohen ZD, DeRubeis RJ. Treatment selection in depression. Annu Rev Clin
Psychol. 2018;14:209–36.
atypical structure or function relative to nondepressed controls

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
11
3. Paul GL. Strategy of outcome research in psychotherapy. J Consult Psychol. 30. Oishi K, Zilles K, Amunts K, Faria A, Jiang H, Li X, et al. Human brain white matter
1967;31:109–18. atlas: identification and assignment of common anatomical structures in
4. American Psychiatric Association. Diagnostic and statistical manual of mental superficial white matter. Neuroimage. 2008;43:447–57.
disorders. DSM-5-TR. Washington, DC: American Psychiatric Association Pub- 31. Gong G, He Y, Concha L, Lebel C, Gross DW, Evans AC, et al. Mapping anatomical
lishing; 2022. connectivity patterns of human cerebral cortex using in vivo diffusion tensor
5. Fried EI, Nesse RM. Depression is not a consistent syndrome: An investigation of imaging tractography. Cereb Cortex. 2009;19:524–36.
unique symptom patterns in the STAR*D study. J Affect Disord. 2015;172:96–102. 32. Schmahmann JD, Pandya DN, Wang R, Dai G, D’Arceuil HE, de Crespigny AJ,
6. American Psychiatric Association, American Psychiatric Association, editors. et al. Association fibre pathways of the brain: parallel observations from diffu-
Diagnostic and statistical manual of mental disorders: DSM-IV-TR. 4th ed. sion spectrum imaging and autoradiography. Brain. 2007;130:630–53.
Washington, DC: American Psychiatric Association; 2000. 33. Hagmann P, Cammoun L, Gigandet X, Meuli R, Honey CJ, Wedeen VJ, et al.
7. Lorenzo-Luaces L, Buss JF, Fried EI. Heterogeneity in major depression and its Mapping the structural core of human cerebral cortex. PLoS Biol. 2008;6:e159.
melancholic and atypical specifiers: a secondary analysis of STAR*D. BMC Psy- 34. Honey CJ, Sporns O, Cammoun L, Gigandet X, Thiran JP, Meuli R, et al. Predicting
chiatry. 2021;21:454. human resting-state functional connectivity from structural connectivity. Proc
8. Kennedy SH, Lam RW, McIntyre RS, Tourjman SV, Bhat V, Blier P, et al. Canadian Natl Acad Sci USA. 2009;106:2035–40.
Network for Mood and Anxiety Treatments (CANMAT) 2016 clinical guidelines 35. Hampson M, Peterson BS, Skudlarski P, Gatenby JC, Gore JC. Detection of
for the management of adults with major depressive disorder: section 3. functional connectivity using temporal correlations in MR images. Hum Brain
Pharmacological Treatments. Can J Psychiatry. 2016;61:540–60. Mapp. 2002;15:247–62.
9. Uher R. Genes, environment, and individual differences in responding to 36. Roebroeck A, Formisano E, Goebel R. Mapping directed influence over the brain
treatment for depression. Harv Rev Psychiatry. 2011;19:109–24. using Granger causality and fMRI. Neuroimage. 2005;25:230–42.
10. Arnow BA, Blasey C, Williams LM, Palmer DM, Rekshan W, Schatzberg AF, et al. 37. Friston KJ, Harrison L, Penny W. Dynamic causal modelling. Neuroimage.
Depression subtypes in predicting antidepressant response: a report from the 2003;19:1273–302.
iSPOT-D Trial. Am J Psychiatry. 2015;172:743–50. 38. Damoiseaux JS, Rombouts SARB, Barkhof F, Scheltens P, Stam CJ, Smith SM,
11. Stewart JW, McGrath PJ, Fava M, Wisniewski SR, Zisook S, Cook I, et al. Do et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad
atypical features affect outcome in depressed outpatients treated with citalo- Sci USA. 2006;103:13848–53.
pram? Int J Neuropsychopharmacol. 2010;13:15–30. 39. Menon V. Large-scale brain networks and psychopathology: a unifying triple
12. McGrath PJ, Khan AY, Trivedi MH, Stewart JW, Morris DW, Wisniewski SR, et al. network model. Trends Cogn Sci. 2011;15:483–506.
Response to a selective serotonin reuptake inhibitor (citalopram) in major 40. Winter NR, Leenings R, Ernsting J, Sarink K, Fisch L, Emden D, et al. Quantifying
depressive disorder with melancholic features: a STAR*D report. J Clin Psy- deviations of brain structure and function in major depressive disorder across
chiatry. 2008;69:1847–55. neuroimaging modalities. JAMA Psychiatry. 2022;79:879.
13. Simon GE, Perlis RH. Personalized medicine for depression: can we match 41. Dutta A, McKie S, Deakin JFW. Resting state networks in major depressive dis-
patients with treatments? Am J Psychiatry. 2010;167:1445–55. order. Psychiatry Res Neuroimaging. 2014;224:139–51.
14. Driessen E, Hollon SD. Cognitive behavioral therapy for mood disorders: efficacy, 42. Kaiser RH, Andrews-Hanna JR, Spielberg JM, Warren SL, Sutton BP, Miller GA,
moderators and mediators. Psychiatr Clin North Am. 2010;33:537–55. et al. Distracted and down: neural mechanisms of affective interference in
15. Parikh SV, Quilty LC, Ravitz P, Rosenbluth M, Pavlova B, Grigoriadis S, et al. subclinical depression. Soc Cogn Affect Neurosci. 2015;10:654–63.
Canadian Network for Mood and Anxiety Treatments (CANMAT) 2016 clinical 43. Cooney RE, Joormann J, Eugène F, Dennis EL, Gotlib IH. Neural correlates of
guidelines for the management of adults with major depressive disorder: sec- rumination in depression. Cogn Affect Behav Neurosci. 2010;10:470–8.
tion 2. Psychological treatments. Can J Psychiatry. 2016;61:524–39. 44. Burkhouse KL, Jacobs RH, Peters AT, Ajilore O, Watkins ER, Langenecker SA.
16. Rush AJ, Trivedi MH, Wisniewski SR, Nierenberg AA, Stewart JW, Warden D, et al. Neural correlates of rumination in adolescents with remitted major depressive
Acute and longer-term outcomes in depressed outpatients requiring one or disorder and healthy controls. Cogn Affect Behav Neurosci. 2017;17:394–405.
several treatment steps: a STAR*D report. Am J Psychiatry. 2006;163:1905–17. 45. Lemogne C, Le Bastard G, Mayberg H, Volle E, Bergouignan L, Lehéricy S, et al. In
17. Dunlop K, Talishinsky A, Liston C. Intrinsic brain network biomarkers of anti- search of the depressive self: extended medial prefrontal network during self-
depressant response: a review. Curr Psychiatry Rep. 2019;21:87. referential processing in major depression. Soc Cogn Affect Neurosci.
18. Tura A, Goya-Maldonado R. Brain connectivity in major depressive disorder: a 2009;4:305–12.
precision component of treatment modalities? Transl Psychiatry. 2023;13:196. 46. Sheline YI, Price JL, Yan Z, Mintun MA. Resting-state functional MRI in depression
19. Gerlach AR, Karim HT, Peciña M, Ajilore O, Taylor WD, Butters MA, et al. MRI unmasks increased connectivity between networks via the dorsal nexus. Proc
predictors of pharmacotherapy response in major depressive disorder. Neuro- Natl Acad Sci USA. 2010;107:11020–5.
image Clin. 2022;36:103157. 47. Beck AT. The evolution of the cognitive model of depression and its neuro-
20. GBD 2019 Mental Disorders Collaborators. Global, regional, and national burden biological correlates. Am J Psychiatry. 2008;165:969–77.
of 12 mental disorders in 204 countries and territories, 1990–2019: a systematic 48. Snyder HR. Major depressive disorder is associated with broad impairments on
analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry. neuropsychological measures of executive function: a meta-analysis and review.
2022;9:137–50. Psychol Bull. 2013;139:81–132.
21. Greenberg PE, Fournier A-A, Sisitsky T, Simes M, Berman R, Koenigsberg SH, 49. Cooper JA, Arulpragasam AR, Treadway MT. Anhedonia in depression: biological
et al. The economic burden of adults with major depressive disorder in the mechanisms and computational models. Curr Opin Behav Sci. 2018;22:128–35.
United States (2010 and 2018). PharmacoEconomics. 2021;39:653–65. 50. Goldstein-Piekarski AN, Ball TM, Samara Z, Staveland BR, Keller AS, Fleming SL,
22. Rush AJ, Fava M, Wisniewski SR, Lavori PW, Trivedi MH, Sackeim HA, et al. et al. Mapping neural circuit biotypes to symptoms and behavioral dimensions
Sequenced treatment alternatives to relieve depression (STAR*D): rationale and of depression and anxiety. Biol Psychiatry. 2022;91:561–71.
design. Control Clin Trials. 2004;25:119–42. 51. Williams LM. Precision psychiatry: a neural circuit taxonomy for depression and
23. Sporns O. The human connectome: a complex network. Ann N. Y Acad Sci. anxiety. Lancet Psychiatry. 2016;3:472–80.
2011;1224:109–25. 52. Cohen SE, Zantvoord JB, Wezenberg BN, Bockting CLH, Van Wingen GA. Mag-
24. Kelly AMC, Uddin LQ, Biswal BB, Castellanos FX, Milham MP. Competition netic resonance imaging for individual prediction of treatment response in
between functional brain networks mediates behavioral variability. NeuroImage. major depressive disorder: a systematic review and meta-analysis. Transl Psy-
2008;39:527–37. chiatry. 2021;11:168.
25. Shannon BJ, Raichle ME, Snyder AZ, Fair DA, Mills KL, Zhang D, et al. Premotor 53. Gao S, Calhoun VD, Sui J. Machine learning in major depression: from classi-
functional connectivity predicts impulsivity in juvenile offenders. Proc Natl Acad fication to treatment outcome prediction. CNS Neurosci Ther.
Sci USA. 2011;108:11241–5. 2018;24:1037–52.
26. Greicius MD, Supekar K, Menon V, Dougherty RF. Resting-state functional con- 54. Williams LM, Rush AJ, Koslow SH, Wisniewski SR, Cooper NJ, Nemeroff CB, et al.
nectivity reflects structural connectivity in the default mode network. Cereb International Study to Predict Optimized Treatment for Depression (iSPOT-D), a
Cortex. 2009;19:72–78. randomized clinical trial: rationale and protocol. Trials. 2011;12:4.
27. Hermundstad AM, Bassett DS, Brown KS, Aminoff EM, Clewett D, Freeman S, 55. Lam RW, Milev R, Rotzinger S, Andreazza AC, Blier P, Brenner C, et al. Discovering
et al. Structural foundations of resting-state and task-based functional con- biomarkers for antidepressant response: protocol from the Canadian biomarker
nectivity in the human brain. Proc Natl Acad Sci USA. 2013;110:6169–74. integration network in depression (CAN-BIND) and clinical characteristics of the
28. Bressler SL, Menon V. Large-scale brain networks in cognition: emerging first patient cohort. BMC Psychiatry. 2016;16:105.
methods and principles. Trends Cogn Sci. 2010;14:277–90. 56. Kennedy SH, Lam RW, Rotzinger S, Milev RV, Blier P, Downar J, et al. Sympto-
29. Spreng RN, Turner GR. Structural covariance of the default network in healthy matic and functional outcomes and early prediction of response to Escitalopram
and pathological aging. J Neurosci. 2013;33:15226–34. Monotherapy and sequential adjunctive aripiprazole therapy in patients with

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
12
major depressive disorder: a CAN-BIND-1 report. J Clin Psychiatry. 79. Wang X, Xue L, Shao J, Dai Z, Hua L, Yan R, et al. Distinct MRI-based functional
2019;80:18m12202. and structural connectivity for antidepressant response prediction in major
57. Trivedi MH, McGrath PJ, Fava M, Parsey RV, Kurian BT, Phillips ML, et al. Estab- depressive disorder. Clin Neurophysiol. 2024;160:19–27.
lishing moderators and biosignatures of antidepressant response in clinical care 80. Ayyash S, Davis AD, Alders GL, MacQueen G, Strother SC, Hassel S, et al.
(EMBARC): rationale and design. J Psychiatr Res. 2016;78:11–23. Assessing remission in major depressive disorder using a functional-structural
58. Goldstein-Piekarski AN, Staveland BR, Ball TM, Yesavage J, Korgaonkar MS, data fusion pipeline: a CAN-BIND-1 study. IBRO Neurosci Rep. 2024;16:135–46.
Williams LM. Intrinsic functional connectivity predicts remission on anti- 81. Sajjadian M, Uher R, Ho K, Hassel S, Milev R, Frey BN, et al. Prediction of
depressants: a randomized controlled trial to identify clinically applicable ima- depression treatment outcome from multimodal data: a CAN-BIND-1 report.
ging biomarkers. Transl Psychiatry. 2018;8:57. Psychol Med. 2023;53:5374–84.
59. Ye Y, Wang C, Lan X, Li W, Fu L, Zhang F, et al. Baseline patterns of resting 82. Harris JK, Hassel S, Davis AD, Zamyadi M, Arnott SR, Milev R, et al. Predicting
functional connectivity within posterior default-mode intranetwork associated escitalopram treatment response from pre-treatment and early response resting
with remission to antidepressants in major depressive disorder. NeuroImage state fMRI in a multi-site sample: A CAN-BIND-1 report. NeuroImage Clin.
Clin. 2022;36:103230. 2022;35:103120.
60. Chin Fatt CR, Jha MK, Cooper CM, Fonzo G, South C, Grannemann B, et al. Effect 83. Beliveau V, Hedeboe E, Fisher PM, Dam VH, Jørgensen MB, Frokjaer VG, et al.
of intrinsic patterns of functional brain connectivity in moderating anti- Generalizability of treatment outcome prediction in major depressive dis-
depressant treatment response in major depression. AJP. 2020;177:143–54. order using structural MRI: a NeuroPharm study. NeuroImage Clin.
61. Korgaonkar MS, Goldstein-Piekarski AN, Fornito A, Williams LM. Intrinsic con- 2022;36:103224.
nectomes are a predictive biomarker of remission in major depressive disorder. 84. Iosifescu DV, Greenwald S, Devlin P, Mischoulon D, Denninger JW, Alpert JE,
Mol Psychiatry. 2020;25:1537–49. et al. Frontal EEG predictors of treatment outcome in major depressive disorder.
62. DeMaster D, Godlewska BR, Liang M, Vannucci M, Bockmann T, Cao B, et al. Eur Neuropsychopharmacol. 2009;19:772–7.
Effective connectivity between resting-state networks in depression. J Affect 85. Leuchter AF, Cook IA, Marangell LB, Gilmer WS, Burgoyne KS, Howland RH, et al.
Disord. 2022;307:79–86. Comparative effectiveness of biomarkers and clinical indicators for predicting
63. Wu H, Liu R, Zhou J, Feng L, Wang Y, Chen X, et al. Prediction of remission outcomes of SSRI treatment in major depressive disorder: results of the BRITE-
among patients with a major depressive disorder based on the resting-state MD study. Psychiatry Res. 2009;169:124–31.
functional connectivity of emotion regulation networks. Transl Psychiatry. 86. Godlewska BR, Harmer CJ. Cognitive neuropsychological theory of anti-
2022;12:391. depressant action: a modern-day approach to depression and its treatment.
64. Martens MAG, Filippini N, Harmer CJ, Godlewska BR. Resting state functional Psychopharmacology. 2021;238:1265–78.
connectivity patterns as biomarkers of treatment response to escitalopram in 87. Spielmans GI, Berman MI, Usitalo AN. Psychotherapy versus second-generation
patients with major depressive disorder. Psychopharmacology. 2022;239:3447–60. antidepressants in the treatment of depression: a meta-analysis. J Nerv Ment
65. Van Der Wijk G, Harris JK, Hassel S, Davis AD, Zamyadi M, Arnott SR, et al. Dis. 2011;199:142–9.
Baseline functional connectivity in resting state networks associated with 88. Cuijpers P, Sijbrandij M, Koole SL, Andersson G, Beekman AT, Reynolds CF. The
depression and remission status after 16 weeks of pharmacotherapy: a CAN- efficacy of psychotherapy and pharmacotherapy in treating depressive and
BIND report. Cereb Cortex. 2022;32:1223–43. anxiety disorders: a meta-analysis of direct comparisons. World Psychiatry.
66. Fan S, Nemati S, Akiki TJ, Roscoe J, Averill CL, Fouda S, et al. Pretreatment brain 2013;12:137–48.
connectome fingerprint predicts treatment response in major depressive dis- 89. Amick HR, Gartlehner G, Gaynes BN, Forneris C, Asher GN, Morgan LC, et al.
order. Chronic Stress. 2020;4:247054702098472. Comparative benefits and harms of second generation antidepressants and
67. Nemati S, Akiki TJ, Roscoe J, Ju Y, Averill CL, Fouda S, et al. A unique brain cognitive behavioral therapies in initial treatment of major depressive disorder:
connectome fingerprint predates and predicts response to antidepressants. systematic review and meta-analysis. BMJ. 2015;351:h6019.
iScience. 2020;23:100800. 90. Cuijpers P, Karyotaki E, Weitz E, Andersson G, Hollon SD, van Straten A. The
68. Vaccarino SR, Wang S, Rizvi SJ, Lou W, Hassel S, MacQueen GM, et al. Functional effects of psychotherapies for major depression in adults on remission, recovery
neuroimaging biomarkers of anhedonia response to escitalopram plus adjunct and improvement: a meta-analysis. J Affect Disord. 2014;159:118–26.
aripiprazole treatment for major depressive disorder. BJPsych Open. 91. Marwood L, Wise T, Perkins AM, Cleare AJ. Meta-analyses of the neural
2024;10:e18. mechanisms and predictors of response to psychotherapy in depression and
69. Fischer AS, Holt-Gosselin B, Fleming SL, Hack LM, Ball TM, Schatzberg AF, et al. anxiety. Neurosci Biobehav Rev. 2018;95:61–72.
Intrinsic reward circuit connectivity profiles underlying symptom and quality of 92. Yoshimura S, Okamoto Y, Onoda K, Matsunaga M, Okada G, Kunisato Y, et al.
life outcomes following antidepressant medication: a report from the iSPOT-D Cognitive behavioral therapy for depression changes medial prefrontal and
trial. Neuropsychopharmacol. 2021;46:809–19. ventral anterior cingulate cortex activity associated with self-referential pro-
70. Korgaonkar MS, Williams LM, Song YJ, Usherwood T, Grieve SM. Diffusion tensor cessing. Soc Cogn Affect Neurosci. 2014;9:487–93.
imaging predictors of treatment outcomes in major depressive disorder. Br J 93. Brown VM, Zhu L, Solway A, Wang JM, McCurry KL, King-Casas B, et al. Rein-
Psychiatry. 2014;205:321–8. forcement learning disruptions in individuals with depression and sensitivity to
71. Grieve SM, Korgaonkar MS, Gordon E, Williams LM, Rush AJ. Prediction of symptom change following cognitive behavioral therapy. JAMA Psychiatry.
nonremission to antidepressant therapy using diffusion tensor imaging. J Clin 2021;78:1113.
Psychiatry. 2016;77:e436–43. 94. Beck AT, Rush AJ. Cognitive therapy of depression. New York: Guilford Press;
72. Davis AD, Hassel S, Arnott SR, Harris J, Lam RW, Milev R, et al. White matter 1979.
indices of medication response in major depression: a diffusion tensor imaging 95. Drevets WC, Raichle ME. Suppression of regional cerebral blood during emo-
study. Biol Psychiatry Cogn Neurosci Neuroimaging. 2019;4:913–24. tional versus higher cognitive implications for interactions between emotion
73. Bubb EJ, Metzler-Baddeley C, Aggleton JP. The cingulum bundle: anatomy, and cognition. Cognit Emot. 1998;12:353–85.
function, and dysfunction. Neurosci Biobehav Rev. 2018;92:104–27. 96. Phillips ML, Drevets WC, Rauch SL, Lane R. Neurobiology of emotion perception
74. Mori S, Kageyama Y, Hou Z, Aggarwal M, Patel J, Brown T, et al. Elucidation of II: implications for major psychiatric disorders. Biol Psychiatry. 2003;54:515–28.
white matter tracts of the human amygdala by detailed comparison between 97. Etkin A, Egner T, Kalisch R. Emotional processing in anterior cingulate and
high-resolution postmortem magnetic resonance imaging and histology. Front medial prefrontal cortex. Trends Cogn Sci. 2011;15:85–93.
Neuroanat. 2017;11:16. 98. Pizzagalli DA. Frontocingulate dysfunction in depression: toward biomarkers of
75. Murphy ML, Frodl T. Meta-analysis of diffusion tensor imaging studies shows treatment response. Neuropsychopharmacology. 2011;36:183–206.
altered fractional anisotropy occurring in distinct brain areas in association with 99. Davidson RJ. Affective neuroscience and psychophysiology: toward a synthesis.
depression. Biol Mood Anxiety Disord. 2011;1:3. Psychophysiology. 2003;40:655–65.
76. Korgaonkar MS, Cooper NJ, Williams LM, Grieve SM. Mapping inter-regional 100. Ray JP, Price JL. The organization of projections from the mediodorsal nucleus
connectivity of the entire cortex to characterize major depressive disorder: a of the thalamus to orbital and medial prefrontal cortex in macaque monkeys. J
whole-brain diffusion tensor imaging tractography study. Neuroreport. Comp Neurol. 1993;337:1–31.
2012;23:566–71. 101. Stuhrmann A, Dohm K, Kugel H, Zwanzger P, Redlich R, Grotegerd D, et al.
77. Cole J, Chaddock CA, Farmer AE, Aitchison KJ, Simmons A, McGuffin P, et al. Mood-congruent amygdala responses to subliminally presented facial expres-
White matter abnormalities and illness severity in major depressive disorder. Br sions in major depression: associations with anhedonia. J Psychiatry Neurosci.
J Psychiatry. 2012;201:33–9. 2013;38:249–58.
78. Tian S, Sun Y, Shao J, Zhang S, Mo Z, Liu X, et al. Predicting escitalopram 102. Suslow T, Konrad C, Kugel H, Rumstadt D, Zwitserlood P, Schöning S, et al.
monotherapy response in depression: the role of anterior cingulate cortex. Hum Automatic mood-congruent amygdala responses to masked facial expressions
Brain Mapp. 2020;41:1249–60. in major depression. Biol Psychiatry. 2010;67:155–60.

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
13
103. Klug M, Enneking V, Borgers T, Jacobs CM, Dohm K, Kraus A, et al. Persistence of based care in STAR*D: implications for clinical practice. Am J Psychiatry.
amygdala hyperactivity to subliminal negative emotion processing in the long- 2006;163:28–40.
term course of depression. Mol Psychiatry. 2024. https://doi.org/10.1038/ 128. George MS, Wassermann EM, Williams WA, Callahan A, Ketter TA, Basser P, et al.
s41380-024-02429-4. Daily repetitive transcranial magnetic stimulation (rTMS) improves mood in
104. Dannlowski U, Ohrmann P, Bauer J, Kugel H, Arolt V, Heindel W, et al. Amygdala depression. NeuroReport. 1995;6:1853–6.
reactivity to masked negative faces is associated with automatic judgmental 129. Tik M, Hoffmann A, Sladky R, Tomova L, Hummer A, Navarro De Lara L, et al.
bias in major depression: a 3 T fMRI study. J Psychiatry Neurosci. 2007;32:423–9. Towards understanding rTMS mechanism of action: Stimulation of the DLPFC
105. Fales CL, Barch DM, Rundle MM, Mintun MA, Snyder AZ, Cohen JD, et al. Altered causes network-specific increase in functional connectivity. NeuroImage.
emotional interference processing in affective and cognitive-control brain cir- 2017;162:289–96.
cuitry in major depression. Biol Psychiatry. 2008;63:377–84. 130. Cao X, Deng C, Su X, Guo Y. Response and remission rates following high-
106. Siegle GJ, Thompson W, Carter CS, Steinhauer SR, Thase ME. Increased amygdala frequency vs. low-frequency Repetitive Transcranial Magnetic Stimulation
and decreased dorsolateral prefrontal BOLD responses in unipolar depression: (rTMS) over right DLPFC for treating Major Depressive Disorder (MDD): a meta-
related and independent features. Biol Psychiatry. 2007;61:198–209. analysis of randomized, double-blind trials. Front Psychiatry. 2018;9:413.
107. DeRubeis RJ, Siegle GJ, Hollon SD. Cognitive therapy versus medication for 131. Baeken C, Marinazzo D, Wu G-R, Van Schuerbeek P, De Mey J, Marchetti I, et al.
depression: treatment outcomes and neural mechanisms. Nat Rev Neurosci. Accelerated HF-rTMS in treatment-resistant unipolar depression: insights from
2008;9:788–96. subgenual anterior cingulate functional connectivity. World J Biol Psychiatry.
108. Canli T, Cooney RE, Goldin P, Shah M, Sivers H, Thomason ME, et al. Amygdala 2014;15:286–97.
reactivity to emotional faces predicts improvement in major depression. Neu- 132. Liston C, Chen AC, Zebley BD, Drysdale AT, Gordon R, Leuchter B, et al. Default
roReport. 2005;16:1267–70. mode network mechanisms of transcranial magnetic stimulation in depression.
109. Williams LM, Korgaonkar MS, Song YC, Paton R, Eagles S, Goldstein-Piekarski A, Biol Psychiatry. 2014;76:517–26.
et al. Amygdala reactivity to emotional faces in the prediction of general and 133. Fox MD, Buckner RL, White MP, Greicius MD, Pascual-Leone A. Efficacy of
medication-specific responses to antidepressant treatment in the randomized transcranial magnetic stimulation targets for depression is related to intrinsic
iSPOT-D trial. Neuropsychopharmacology. 2015;40:2398–408. functional connectivity with the subgenual cingulate. Biol Psychiatry.
110. Siegle GJ, Carter CS, Thase ME. Use of fMRI to predict recovery from unipolar 2012;72:595–603.
depression with cognitive behavior therapy. Am J Psychiatry. 2006;163:735–8. 134. Weigand A, Horn A, Caballero R, Cooke D, Stern AP, Taylor SF, et al. Prospective
111. Queirazza F, Fouragnan E, Steele JD, Cavanagh J, Philiastides MG. Neural cor- validation that subgenual connectivity predicts antidepressant efficacy of
relates of weighted reward prediction error during reinforcement learning transcranial magnetic stimulation sites. Biol Psychiatry. 2018;84:28–37.
classify response to cognitive behavioral therapy in depression. Sci Adv. 135. Cash RFH, Zalesky A, Thomson RH, Tian Y, Cocchi L, Fitzgerald PB. Subgenual
2019;5:eaav4962. functional connectivity predicts antidepressant treatment response to tran-
112. Fu CHY, Williams SCR, Cleare AJ, Scott J, Mitterschiffthaler MT, Walsh ND, et al. scranial magnetic stimulation: independent validation and evaluation of per-
Neural responses to sad facial expressions in major depression following cog- sonalization. Biol Psychiatry. 2019;86:e5–7.
nitive behavioral therapy. Biol Psychiatry. 2008;64:505–12. 136. Ge R, Downar J, Blumberger DM, Daskalakis ZJ, Vila-Rodriguez F. Functional
113. Costafreda SG, Khanna A, Mourao-Miranda J, Fu CHY. Neural correlates of sad connectivity of the anterior cingulate cortex predicts treatment outcome for
faces predict clinical remission to cognitive behavioural therapy in depression. rTMS in treatment-resistant depression at 3-month follow-up. Brain Stimulation.
NeuroReport. 2009;20:637–41. 2020;13:206–14.
114. Ritchey M, Dolcos F, Eddington KM, Strauman TJ, Cabeza R. Neural correlates of 137. Elbau IG, Lynch CJ, Downar J, Vila-Rodriguez F, Power JD, Solomonov N, et al.
emotional processing in depression: changes with cognitive behavioral therapy Functional connectivity mapping for rTMS target selection in depression. Am J
and predictors of treatment response. J Psychiatr Res. 2011;45:577–87. Psychiatry. 2023;180:230–40.
115. Rubin-Falcone H, Weber J, Kishon R, Ochsner K, Delaparte L, Doré B, et al. Neural 138. Roalf DR, Figee M, Oathes DJ. Elevating the field for applying neuroimaging to
predictors and effects of cognitive behavioral therapy for depression: the role of individual patients in psychiatry. Transl Psychiatry. 2024;14:87.
emotional reactivity and regulation. Psychol Med. 2020;50:146–60. 139. Siddiqi SH, Khosravani S, Rolston JD, Fox MD. The future of brain circuit-targeted
116. Siegle GJ, Thompson WK, Collier A, Berman SR, Feldmiller J, Thase ME, et al. therapeutics. Neuropsychopharmacology. 2024;49:179–88.
Toward clinically useful neuroimaging in depression treatment: prognostic uti- 140. Mayberg HS, Lozano AM, Voon V, McNeely HE, Seminowicz D, Hamani C, et al.
lity of subgenual cingulate activity for determining depression outcome in Deep brain stimulation for treatment-resistant depression. Neuron.
cognitive therapy across studies, scanners, and patient characteristics. Arch Gen 2005;45:651–60.
Psychiatry. 2012;69:913. 141. Elias GJB, Germann J, Boutet A, Loh A, Li B, Pancholi A, et al. 3T MRI of rapid
117. McGrath CL, Kelley ME, Dunlop BW, Holtzheimer PE, Craighead WE, Mayberg HS. brain activity changes driven by subcallosal cingulate deep brain stimulation.
Pretreatment brain states identify likely nonresponse to standard treatments for Brain. 2022;145:2214–26.
depression. Biol Psychiatry. 2014;76:527–35. 142. Huebl J, Brücke C, Merkl A, Bajbouj M, Schneider G-H, Kühn AA. Processing of
118. Messina I, Bianco F, Cusinato M, Calvo V, Sambin M. Abnormal default system emotional stimuli is reflected by modulations of beta band activity in the
functioning in depression: implications for emotion regulation. Front Psychol. subgenual anterior cingulate cortex in patients with treatment resistant
2016;7:202607. depression. Soc Cogn Affect Neurosci. 2016;11:1290–8.
119. Dunlop BW, Mayberg HS. Neuroimaging-based biomarkers for treatment 143. Riva-Posse P, Choi KS, Holtzheimer PE, Crowell AL, Garlow SJ, Rajendra JK, et al.
selection in major depressive disorder. Dialogues Clin Neurosci. 2014;16:479–90. A connectomic approach for subcallosal cingulate deep brain stimulation sur-
120. Crowther A, Smoski MJ, Minkel J, Moore T, Gibbs D, Petty C, et al. Resting-state gery: prospective targeting in treatment-resistant depression. Mol Psychiatry.
connectivity predictors of response to psychotherapy in major depressive dis- 2018;23:843–9.
order. Neuropsychopharmacol. 2015;40:1659–73. 144. Lozano AM, Giacobbe P, Hamani C, Rizvi SJ, Kennedy SH, Kolivakis TT, et al. A
121. Davis EG, Foland-Ross LC, Gotlib IH. Neural correlates of top-down regulation multicenter pilot study of subcallosal cingulate area deep brain stimulation for
and generation of negative affect in major depressive disorder. Psychiatry Res treatment-resistant depression. J Neurosurg. 2012;116:315–22.
Neuroimaging. 2018;276:1–8. 145. Redlich R, Opel N, Grotegerd D, Dohm K, Zaremba D, Bürger C, et al. Prediction
122. Straub J, Metzger CD, Plener PL, Koelch MG, Groen G, Abler B. Successful group of individual response to electroconvulsive therapy via machine learning on
psychotherapy of depression in adolescents alters fronto-limbic resting-state structural magnetic resonance imaging data. JAMA Psychiatry. 2016;73:557.
connectivity. J Affect Disord. 2017;209:135–9. 146. Argyelan M, Lencz T, Kaliora S, Sarpal DK, Weissman N, Kingsley PB. et al.
123. Hollon SD, Thase ME, Markowitz JC. Treatment and prevention of depression. Subgenual cingulate cortical activity predicts the efficacy of electroconvulsive
Psychol Sci Public Interest. 2002;3:39–77. therapy. Transl Psychiatry. 2016;6:e789
124. Zhou D, Zhou X, Lin Q, Wang W, Lv Z, Chen X, et al. Nonpharmacological 147. Cano M, Cardoner N, Urretavizcaya M, Martínez-Zalacaín I, Goldberg X, Via E,
interventions for relapse prevention in unipolar depression: a network meta- et al. Modulation of limbic and prefrontal connectivity by electroconvulsive
analysis. J Affect Disord. 2021;282:1255–62. therapy in treatment-resistant depression: a preliminary study. Brain Stimula-
125. Berlim MT, Turecki G. Definition, assessment, and staging of treatment—resis- tion. 2016;9:65–71.
tant refractory major depression: a review of current concepts and methods. 148. Leaver AM, Wade B, Vasavada M, Hellemann G, Joshi SH, Espinoza R, et al.
Can J Psychiatry. 2007;52:46–54. Fronto-temporal connectivity predicts ECT outcome in major depression. Front
126. Nemeroff CB. Prevalence and management of treatment-resistant depression. J Psychiatry. 2018;9:92.
Clin Psychiatry. 2007;68:17–25. 149. Moreno-Ortega M, Prudic J, Rowny S, Patel GH, Kangarlu A, Lee S, et al. Resting
127. Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al. state functional connectivity predictors of treatment response to electro-
Evaluation of outcomes with citalopram for depression using measurement- convulsive therapy in depression. Sci Rep. 2019;9:5071.

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
14
150. Sun H, Jiang R, Qi S, Narr KL, Wade BS, Upston J, et al. Preliminary prediction of 174. Pizzagalli DA, Webb CA, Dillon DG, Tenke CE, Kayser J, Goer F, et al. Pretreatment
individual response to electroconvulsive therapy using whole-brain functional rostral anterior cingulate cortex theta activity in relation to symptom
magnetic resonance imaging data. NeuroImage Clin. 2020;26:102080. improvement in depression: a randomized clinical trial. JAMA Psychiatry.
151. Pang Y, Wei Q, Zhao S, Li N, Li Z, Lu F, et al. Enhanced default mode network 2018;75:547.
functional connectivity links with electroconvulsive therapy response in major 175. Tenke CE, Kayser J. Generator localization by current source density (CSD):
depressive disorder. J Affect Disord. 2022;306:47–54. implications of volume conduction and field closure at intracranial and scalp
152. Van Waarde JA, Scholte HS, Van Oudheusden LJB, Verwey B, Denys D, Van resolutions. Clin Neurophysiol. 2012;123:2328–45.
Wingen GA. A functional MRI marker may predict the outcome of electro- 176. Müller K-R, Tangermann M, Dornhege G, Krauledat M, Curio G, Blankertz B.
convulsive therapy in severe and treatment-resistant depression. Mol Psychiatry. Machine learning for real-time single-trial EEG-analysis: From brain–computer
2015;20:609–14. interfacing to mental state monitoring. J Neurosci Methods. 2008;167:82–90.
153. Ge R, Downar J, Blumberger DM, Daskalakis ZJ, Lam RW, Vila-Rodriguez F. 177. Singh AK, Krishnan S. Trends in EEG signal feature extraction applications. Front
Structural network integrity of the central executive network is associated with Artif Intell. 2023;5:1072801.
the therapeutic effect of rTMS in treatment resistant depression. Prog Neuro- 178. Wu W, Zhang Y, Jiang J, Lucas MV, Fonzo GA, Rolle CE, et al. An electro-
Psychopharmacol Biol Psychiatry. 2019;92:217–25. encephalographic signature predicts antidepressant response in major
154. Fan J, Tso IF, Maixner DF, Abagis T, Hernandez-Garcia L, Taylor SF. Segregation depression. Nat Biotechnol. 2020;38:439–47.
of salience network predicts treatment response of depression to repetitive 179. Zhdanov A, Atluri S, Wong W, Vaghei Y, Daskalakis ZJ, Blumberger DM, et al. Use
transcranial magnetic stimulation. NeuroImage Clin. 2019;22:101719. of machine learning for predicting escitalopram treatment outcome from
155. Cao B, Luo Q, Fu Y, Du L, Qiu T, Yang X, et al. Predicting individual responses to electroencephalography recordings in adult patients with depression. JAMA
the electroconvulsive therapy with hippocampal subfield volumes in major Netw Open. 2020;3:e1918377.
depression disorder. Sci Rep. 2018;8:5434. 180. ACNP 62nd Annual Meeting: Poster Abstracts P1 – P250. Neuropsycho-
156. Wade BSC, Joshi SH, Njau S, Leaver AM, Vasavada M, Woods RP, et al. Effect of pharmacol. 2023;48:63–210.
electroconvulsive therapy on striatal morphometry in major depressive dis- 181. Dunlop BW, Binder EB, Cubells JF, Goodman MM, Kelley ME, Kinkead B, et al.
order. Neuropsychopharmacol. 2016;41:2481–91. Predictors of remission in depression to individual and combined treatments
157. Drysdale AT, Grosenick L, Downar J, Dunlop K, Mansouri F, Meng Y, et al. (PReDICT): study protocol for a randomized controlled trial. Trials. 2012;13:106.
Resting-state connectivity biomarkers define neurophysiological subtypes of 182. Dunlop BW, Rajendra JK, Craighead WE, Kelley ME, McGrath CL, Choi KS, et al.
depression. Nat Med. 2017;23:28–38. Functional connectivity of the subcallosal cingulate cortex and differential
158. Dunlop K, Grosenick L, Downar J, Vila-Rodriguez F, Gunning FM, Daskalakis ZJ, outcomes to treatment with cognitive-behavioral therapy or antidepressant
et al. Dimensional and categorical solutions to parsing depression hetero- medication for major depressive disorder. Am J Psychiatry. 2017;174:533–45.
geneity in a large single-site sample. Biol Psychiatry. 2024:S0006322324000556. 183. McGrath CL, Kelley ME, Holtzheimer PE, Dunlop BW, Craighead WE, Franco AR,
159. Fox MD. Mapping symptoms to brain networks with the human connectome. N. et al. Toward a neuroimaging treatment selection biomarker for major
Engl J Med. 2018;379:2237–45. depressive disorder. JAMA Psychiatry. 2013;70:821–9.
160. Thomas Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, 184. Delaveau P, Jabourian M, Lemogne C, Guionnet S, Bergouignan L, Fossati P.
et al. The organization of the human cerebral cortex estimated by intrinsic Brain effects of antidepressants in major depression: a meta-analysis of emo-
functional connectivity. J Neurophysiol. 2011;106:1125–65. tional processing studies. J Affect Disord. 2011;130:66–74.
161. Van Essen DC, Ugurbil K, Auerbach E, Barch D, Behrens TEJ, Bucholz R, et al. The 185. Fitzgerald PB, Laird AR, Maller J, Daskalakis ZJ. A meta‐analytic study of changes
human connectome project: a data acquisition perspective. Neuroimage. in brain activation in depression. Hum Brain Mapp. 2008;29:683–95.
2012;62:2222–31. 186. Kelley ME, Choi KS, Rajendra JK, Craighead WE, Rakofsky JJ, Dunlop BW, et al.
162. Siddiqi SH, Taylor SF, Cooke D, Pascual-Leone A, George MS, Fox MD. Distinct Establishing evidence for clinical utility of a neuroimaging biomarker in major
symptom-specific treatment targets for circuit-based neuromodulation. Am J depressive disorder: prospective testing and implementation challenges. Biol
Psychiatry. 2020;177:435–46. Psychiatry. 2021;90:236–42.
163. Greenberg T, Fournier JC, Stiffler R, Chase HW, Almeida JR, Aslam H, et al. 187. Rutherford BR, Sneed JR, Roose SP. Does study design influence outcome? The
Reward related ventral striatal activity and differential response to sertraline effects of placebo control and treatment duration in antidepressant trials.
versus placebo in depressed individuals. Mol Psychiatry. 2020;25:1526–36. Psychother Psychosom. 2009;78:172-81.
164. Chin Fatt CR, Cooper C, Jha MK, Aslan S, Grannemann B, Kurian B, et al. Dor- 188. Hannon K, Bijsterbosch J. Challenges in identifying individualized brain bio-
solateral prefrontal cortex and subcallosal cingulate connectivity show pre- markers of late life depression. Adv Geriatr Med Res. 2023;5:e230010.
ferential antidepressant response in major depressive disorder. Biol Psychiatry 189. Nichols TE, Das S, Eickhoff SB, Evans AC, Glatard T, Hanke M, et al. Best practices
Cognit Neurosci Neuroimaging. 2021;6:20–8. in data analysis and sharing in neuroimaging using MRI. Nat Neurosci.
165. Frodl T, Scheuerecker J, Schoepf V, Linn J, Koutsouleris N, Bokde ALW, et al. 2017;20:299–303.
Different effects of mirtazapine and venlafaxine on brain activation: an open 190. Rutherford BR, Wall MM, Brown PJ, Choo T-H, Wager TD, Peterson BS, et al.
randomized controlled fMRI study. J Clin Psychiatry. 2011;72:448–57. Patient expectancy as a mediator of placebo effects in antidepressant clinical
166. Little JT, Ketter TA, Kimbrell TA, Dunn RT, Benson BE, Willis MW, et al. Bupropion trials. Am J Psychiatry. 2017;174:135–42.
and venlafaxine responders differ in pretreatment regional cerebral metabolism 191. Gaudiano BA, Ellenberg SR, Schofield CA, Rifkin LS. A randomized survey of the
in unipolar depression. Biol Psychiatry. 2005;57:220–8. public’s expectancies and willingness to participate in clinical trials of anti-
167. Perez-Caballero L, Torres-Sanchez S, Romero-López-Alberca C, González-Saiz F, depressants versus psychotherapy for depression. Prim Care Companion CNS
Mico JA, Berrocoso E. Monoaminergic system and depression. Cell Tissue Res. Disord. 2016;18 https://doi.org/10.4088/PCC.15m01879.
2019;377:107–13. 192. Fountoulakis KN, McIntyre RS, Carvalho AF. From randomized controlled trials
168. Cooper CM, Chin Fatt CR, Jha M, Fonzo GA, Grannemann BD, Carmody T, et al. of antidepressant drugs to the meta-analytic synthesis of evidence: metho-
Cerebral blood perfusion predicts response to sertraline versus placebo for major dological aspects lead to discrepant findings. Curr Neuropharmacol.
depressive disorder in the EMBARC trial. EClinicalMedicine. 2019;10:32–41. 2015;13:605–15.
169. Dang Y, Lu B, Vanderwal T, Castellanos FX, Yan C-G. Early-treatment cerebral 193. Fu CHY, Antoniades M, Erus G, Garcia JA, Fan Y, Arnone D, et al. Neuroanato-
blood flow change as a predictive biomarker of antidepressant treatment mical dimensions in medication-free individuals with major depressive disorder
response: evidence from the EMBARC clinical trial. Psychol Med. 2024:1–10. and treatment response to SSRI antidepressant medications or placebo. Nat
170. Poirot MG, Ruhe HG, Mutsaerts H-JMM, Maximov II, Groote IR, Bjørnerud A, et al. Ment Health. 2024;2:164–76.
Treatment response prediction in major depressive disorder using multimodal 194. Schmaal L, Pozzi E, C. Ho T, Van Velzen LS, Veer IM, Opel N, et al. ENIGMA MDD:
MRI and clinical data: secondary analysis of a randomized clinical trial. Am J seven years of global neuroimaging studies of major depression through
Psychiatry. 2024;181:223–33. worldwide data sharing. Transl Psychiatry. 2020;10:172.
171. Zhao K, Xie H, Fonzo GA, Tong X, Carlisle N, Chidharom M, et al. Individualized 195. Thompson PM, Stein JL, Medland SE, Hibar DP, Vasquez AA, Renteria ME, et al.
fMRI connectivity defines signatures of antidepressant and placebo responses in The ENIGMA consortium: large-scale collaborative analyses of neuroimaging
major depression. Mol Psychiatry. 2023;28:2490–9. and genetic data. Brain Imaging Behav. 2014;8:153–82.
172. Korb AS, Hunter AM, Cook IA, Leuchter AF. Rostral anterior cingulate cortex 196. Fu CHY, Erus G, Fan Y, Antoniades M, Arnone D, Arnott SR, et al. AI-based
theta current density and response to antidepressants and placebo in major dimensional neuroimaging system for characterizing heterogeneity in brain
depression. Clin Neurophysiol. 2009;120:1313–9. structure and function in major depressive disorder: COORDINATE-MDD con-
173. Widge AS, Bilge MT, Montana R, Chang W, Rodriguez CI, Deckersbach T, et al. sortium design and rationale. BMC Psychiatry. 2023;23:59.
Electroencephalographic biomarkers for treatment response prediction in major 197. Enck P, Bingel U, Schedlowski M, Rief W. The placebo response in medicine:
depressive illness: a meta-analysis. Am J Psychiatry. 2019;176:44–56. minimize, maximize or personalize? Nat Rev Drug Discov. 2013;12:191–204.

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
15
198. Jones BDM, Razza LB, Weissman CR, Karbi J, Vine T, Mulsant LS, et al. Magnitude 224. Solomonov N, Lee J, Banerjee S, Flückiger C, Kanellopoulos D, Gunning FM, et al.
of the placebo response across treatment modalities used for treatment- Modifiable predictors of nonresponse to psychotherapies for late-life depression
resistant depression in adults: a systematic review and meta-analysis. JAMA with executive dysfunction: a machine learning approach. Mol Psychiatry.
Netw Open. 2021;4:e2125531. 2021;26:5190–8.
199. Wager TD, Atlas LY. The neuroscience of placebo effects: connecting context, 225. Lewis CC, Simons AD, Kim HK. The role of early symptom trajectories and
learning and health. Nat Rev Neurosci. 2015;16:403–18. pretreatment variables in predicting treatment response to cognitive behavioral
200. Dunlop BW, Kelley ME, Aponte-Rivera V, Mletzko-Crowe T, Kinkead B, Ritchie JC, therapy. J Consult Clin Psychol. 2012;80:525–34.
et al. Effects of patient preferences on outcomes in the Predictors of Remission 226. Kelley ME, Dunlop BW, Nemeroff CB, Lori A, Carrillo-Roa T, Binder EB, et al.
in Depression to Individual and Combined Treatments (PReDICT) study. Am J Response rate profiles for major depressive disorder: Characterizing
Psychiatry. 2017;174:546–56. early response and longitudinal nonresponse. Depress Anxiety.
201. Wasmann KA, Wijsman P, van Dieren S, Bemelman W, Buskens C. Partially ran- 2018;35:992–1000.
domised patient preference trials as an alternative design to randomised con- 227. Frässle S, Marquand AF, Schmaal L, Dinga R, Veltman DJ, van der Wee NJA, et al.
trolled trials: systematic review and meta-analyses. BMJ Open. 2019;9:e031151. Predicting individual clinical trajectories of depression with generative
202. Zimmerman M, Balling C, Chelminski I, Dalrymple K. Have treatment studies of embedding. Neuroimage Clin. 2020;26:102213.
depression become even less generalizable? Applying the inclusion and 228. Schmaal L, Marquand AF, Rhebergen D, van Tol M-J, Ruhé HG, van der Wee NJA,
exclusion criteria in placebo-controlled antidepressant efficacy trials published et al. Predicting the naturalistic course of major depressive disorder using
over 20 years to a clinical sample. Psychother Psychosom. 2019;88:165–70. clinical and multimodal neuroimaging information: a multivariate pattern
203. Zimmerman M, Balling C, Chelminski I, Dalrymple K. Applying the inclusion/ recognition study. Biol Psychiatry. 2015;78:278–86.
exclusion criteria in placebo-controlled studies to a clinical sample: a compar- 229. Gonzalez-Castillo J, Kam JWY, Hoy CW, Bandettini PA. How to interpret resting-
ison of medications. J Affect Disord. 2020;260:483–8. state fmri: ask your participants. J Neurosci. 2021;41:1130–41.
204. Kessler RC, Merikangas KR, Wang PS. Prevalence, comorbidity, and service uti- 230. Jiang R, Zuo N, Ford JM, Qi S, Zhi D, Zhuo C, et al. Task-induced brain con-
lization for mood disorders in the United States at the beginning of the twenty- nectivity promotes the detection of individual differences in brain-behavior
first century. Annu Rev Clin Psychol. 2007;3:137–58. relationships. Neuroimage. 2020;207:116370.
205. Kaufman J, Charney D. Comorbidity of mood and anxiety disorders. Depress 231. Rosenberg MD, Finn ES, Scheinost D, Papademetris X, Shen X, Constable RT,
Anxiety. 2000;12:69–76. et al. A neuromarker of sustained attention from whole-brain functional con-
206. Wiethoff K, Bauer M, Baghai TC, Möller H-J, Fisher R, Hollinde D, et al. Prevalence nectivity. Nat Neurosci. 2016;19:165–71.
and treatment outcome in anxious versus nonanxious depression: results from 232. Zhao W, Makowski C, Hagler DJ, Garavan HP, Thompson WK, Greene DJ, et al.
the German Algorithm Project. J Clin Psychiatry. 2010;71:1047–54. Task fMRI paradigms may capture more behaviorally relevant information than
207. Richards D. Prevalence and clinical course of depression: a review. Clin Psychol resting-state functional connectivity. NeuroImage. 2023;270:119946.
Rev. 2011;31:1117–25. 233. Greene AS, Gao S, Scheinost D, Constable RT. Task-induced brain state manip-
208. Lueken U, Zierhut KC, Hahn T, Straube B, Kircher T, Reif A, et al. Neurobiological ulation improves prediction of individual traits. Nat Commun. 2018;9:2807.
markers predicting treatment response in anxiety disorders: a systematic review 234. Elliott ML, Knodt AR, Cooke M, Kim MJ, Melzer TR, Keenan R, et al. General
and implications for clinical application. Neurosci Biobehav Rev. 2016;66:143–62. functional connectivity: Shared features of resting-state and task fMRI drive
209. Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. reliable and heritable individual differences in functional brain networks. Neu-
1960;23:56–62. roimage. 2019;189:516–32.
210. Beck AT, Steer RA, Brown G. Manual for the Beck depression inventory-II. San 235. Dunlop K, Rizvi SJ, Kennedy SH, Hassel S, Strother SC, Harris JK, et al. Clinical,
Antonio, TX: Psychological Corporation; 1996. behavioral, and neural measures of reward processing correlate with
211. Bagby RM, Ryder AG, Schuller DR, Marshall MB. The hamilton depression rating escitalopram response in depression: a Canadian Biomarker Integration Net-
scale: has the gold standard become a lead weight? Am J Psychiatry. work in Depression (CAN-BIND-1) report. Neuropsychopharmacology.
2004;161:2163–77. 2020;45:1390–7.
212. Williams LM. Defining biotypes for depression and anxiety based on large-scale 236. Admon R, Kaiser RH, Dillon DG, Beltzer M, Goer F, Olson DP, et al. Dopaminergic
circuit dysfunction: a theoretical review of the evidence and future directions for enhancement of striatal response to reward in major depression. Am J Psy-
clinical translation. Depress Anxiety. 2017;34:9–24. chiatry. 2017;174:378–86.
213. Grunebaum MF, Keilp JG, Ellis SP, Sudol K, Bauer N, Burke AK, et al. SSRI versus 237. Kraus C, Klöbl M, Tik M, Auer B, Vanicek T, Geissberger N, et al. The pulvinar
bupropion effects on symptom clusters in suicidal depression: post hoc analysis nucleus and antidepressant treatment: dynamic modeling of antidepressant
of a randomized clinical trial. J Clin Psychiatry. 2013;74:872–9. response and remission with ultra-high field functional MRI. Mol Psychiatry.
214. Stewart JG, Harkness KL. Symptom specificity in the acute treatment of major 2019;24:746–56.
depressive disorder: a re-analysis of the treatment of depression collaborative 238. Walsh EC, Eisenlohr-Moul TA, Minkel J, Bizzell J, Petty C, Crowther A, et al.
research program. J Affect Disord. 2012;137:87–97. Pretreatment brain connectivity during positive emotion upregulation predicts
215. Uher R, Maier W, Hauser J, Marušič A, Schmael C, Mors O, et al. Differential decreased anhedonia following behavioral activation therapy for depression. J
efficacy of escitalopram and nortriptyline on dimensional measures of depres- Affect Disord. 2019;243:188–92.
sion. Br J Psychiatry. 2009;194:252–9. 239. Karim HT, Wang M, Andreescu C, Tudorascu D, Butters MA, Karp JF, et al. Acute
216. Dunlop BW, Cole SP, Nemeroff CB, Mayberg HS, Craighead WE. Differential trajectories of neural activation predict remission to pharmacotherapy in late-
change on depressive symptom factors with antidepressant medication and life depression. Neuroimage Clin. 2018;19:831–9.
cognitive behavior therapy for major depressive disorder. J Affect Disord. 240. Crane NA, Jenkins LM, Bhaumik R, Dion C, Gowins JR, Mickey BJ, et al. Multi-
2018;229:111–9. dimensional prediction of treatment response to antidepressants with cognitive
217. Ballard ED, Yarrington JS, Farmer CA, Lener MS, Kadriu B, Lally N, et al. Parsing control and functional MRI. Brain. 2017;140:472–86.
the heterogeneity of depression: an exploratory factor analysis across com- 241. Miller S, McTeague LM, Gyurak A, Patenaude B, Williams LM, Grieve SM, et al.
monly used depression rating scales. J Affect Disord. 2018;231:51–57. Cognition-childhood maltreatment interactions in the prediction of anti-
218. Fried EI, Nesse RM. Depression sum-scores don’t add up: why analyzing specific depressant outcomes in major depressive disorder patients: results from the
depression symptoms is essential. BMC Med. 2015;13:72. iSPOT-D trial. Depress Anxiety. 2015;32:594–604.
219. Sheehan DV, Harnett-Sheehan K, Raj BA. The measurement of disability. Int Clin 242. Hack LM, Tozzi L, Zenteno S, Olmsted AM, Hilton R, Jubeir J, et al. A cognitive
Psychopharmacol. 1996;11:89–95. biotype of depression and symptoms, behavior measures, neural circuits, and
220. Soares CN, Endicott J, Boucher M, Fayyad RS, Guico-Pabia CJ. Predictors of differential treatment outcomes: a prespecified secondary analysis of a rando-
functional response and remission with desvenlafaxine 50 mg/d in patients with mized clinical trial. JAMA Netw Open. 2023;6:e2318411.
major depressive disorder. CNS Spectr. 2014;19:519–27. 243. Finn ES, Bandettini PA. Movie-watching outperforms rest for functional
221. Soares CN, Zhang M, Boucher M. Categorical improvement in functional connectivity-based prediction of behavior. Neuroimage. 2021;235:117963.
impairment in depressed patients treated with desvenlafaxine. CNS Spectr. 244. Welvaert M, Rosseel Y. On the definition of signal-to-noise ratio and contrast-to-
2019;24:322–32. noise ratio for fMRI data. PLoS one. 2013;8:e77089.
222. Fu CHY, Costafreda SG, Sankar A, Adams TM, Rasenick MM, Liu P, et al. Multi- 245. Kundu P, Inati SJ, Evans JW, Luh W-M, Bandettini PA. Differentiating BOLD and
modal functional and structural neuroimaging investigation of major depressive non-BOLD signals in fMRI time series using multi-echo EPI. Neuroimage.
disorder following treatment with duloxetine. BMC Psychiatry. 2015;15:82. 2012;60:1759–70.
223. Zhou R, Chen J, Zhao G, Wang Z, Peng D, Xia W, et al. Neural biomarker of 246. Kundu P, Brenowitz ND, Voon V, Worbe Y, Vértes PE, Inati SJ, et al. Integrated
functional disability in major depressive disorder: a structural neuroimaging strategy for improving functional connectivity mapping using multiecho fMRI.
study. Prog Neuropsychopharmacol Biol Psychiatry. 2021;111:110337. Proc Natl Acad Sci USA. 2013;110:16187–92.

Neuropsychopharmacology
 S. Prompiengchai and K. Dunlop
16
247. Kundu P, Voon V, Balchandani P, Lombardo MV, Poser BA, Bandettini PA. Multi- 273. Scharnowski F, Nicholson AA, Pichon S, Rosa MJ, Rey G, Eickhoff SB, et al. The
echo fMRI: a review of applications in fMRI denoising and analysis of BOLD role of the subgenual anterior cingulate cortex in dorsomedial prefrontal-
signals. Neuroimage. 2017;154:59–80. amygdala neural circuitry during positive-social emotion regulation. Hum Brain
248. DuPre E, Salo T, Ahmed Z, Bandettini P, Bottenhorn K, Caballero-Gaudes C, et al. Mapp. 2020;41:3100–18.
TE-dependent analysis of multi-echo fMRI with tedana. J Open Sour Softw. 274. Niendam TA, Laird AR, Ray KL, Dean YM, Glahn DC, Carter CS. Meta-analytic
2021;6:3669. evidence for a superordinate cognitive control network subserving diverse
249. Steel A, Garcia BD, Silson EH, Robertson CE. Evaluating the efficacy of multi-echo executive functions. Cogn Affect Behav Neurosci. 2012;12:241–68.
ICA denoising on model-based fMRI. Neuroimage. 2022;264:119723. 275. Dosenbach NUF, Fair DA, Cohen AL, Schlaggar BL, Petersen SE. A dual-networks
250. Cohen AD, Chang C, Wang Y. Using multiband multi-echo imaging to improve architecture of top-down control. Trends Cogn Sci. 2008;12:99–105.
the robustness and repeatability of co-activation pattern analysis for dynamic 276. Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, et al. Dis-
functional connectivity. Neuroimage. 2021;243:118555. sociable intrinsic connectivity networks for salience processing and executive
251. Power JD, Plitt M, Gotts SJ, Kundu P, Voon V, Bandettini PA, et al. Ridding fMRI control. J Neurosci. 2007;27:2349–56.
data of motion-related influences: removal of signals with distinct spatial and 277. Mayberg HS, Liotti M, Brannan SK, McGinnis S, Mahurin RK, Jerabek PA, et al.
physical bases in multiecho data. Proc Natl Acad Sci USA. 2018;115:E2105–14. Reciprocal limbic-cortical function and negative mood: converging PET findings
252. Gonzalez-Castillo J, Panwar P, Buchanan LC, Caballero-Gaudes C, Handwerker in depression and normal sadness. Am J Psychiatry. 1999;156:675–82.
DA, Jangraw DC, et al. Evaluation of multi-echo ICA denoising for task based 278. RajMohan V, Mohandas E. The limbic system. Indian J Psychiatry. 2007;49:132.
fMRI studies: block designs, rapid event-related designs, and cardiac-gated fMRI. 279. Sergerie K, Chochol C, Armony JL. The role of the amygdala in emotional pro-
Neuroimage. 2016;141:452–68. cessing: a quantitative meta-analysis of functional neuroimaging studies. Neu-
253. Gilmore AW, Agron AM, González-Araya EI, Gotts SJ, Martin A. A comparison of rosci Biobehav Rev. 2008;32:811–30.
single- and multi-echo processing of functional MRI data during overt auto- 280. Rolls ET. Limbic systems for emotion and for memory, but no single limbic
biographical recall. Front Neurosci. 2022;16:854387. system. Cortex. 2015;62:119–57.
254. Lynch CJ, Power JD, Scult MA, Dubin M, Gunning FM, Liston C. Rapid precision 281. Yarkoni T, Poldrack RA, Nichols TE, Van Essen DC, Wager TD. Large-scale auto-
functional mapping of individuals using multi-echo fMRI. Cell Rep. mated synthesis of human functional neuroimaging data. Nat Methods.
2020;33:108540. 2011;8:665–70.
255. Lynch CJ, Elbau I, Liston C. Improving precision functional mapping routines 282. Shirer WR, Ryali S, Rykhlevskaia E, Menon V, Greicius MD. Decoding subject-
with multi-echo fMRI. Curr Opin Behav Sci. 2021;40:113–9. driven cognitive states with whole-brain connectivity patterns. Cereb Cortex.
256. Kong R, Li J, Orban C, Sabuncu MR, Liu H, Schaefer A, et al. Spatial topography of 2012;22:158–65.
individual-specific cortical networks predicts human cognition, personality, and 283. Andrews-Hanna JR, Reidler JS, Sepulcre J, Poulin R, Buckner RL. Functional-
emotion. Cereb Cortex. 2019;29:2533–51. anatomic fractionation of the brain’s default network. Neuron.
257. Seitzman BA, Gratton C, Laumann TO, Gordon EM, Adeyemo B, Dworetsky A, 2010;65:550–62.
et al. Trait-like variants in human functional brain networks. Proc Natl Acad Sci 284. Smallwood J, Bernhardt BC, Leech R, Bzdok D, Jefferies E, Margulies DS. The
USA. 2019;116:22851–61. default mode network in cognition: a topographical perspective. Nat Rev
258. Wang D, Li M, Wang M, Schoeppe F, Ren J, Chen H, et al. Individual-specific Neurosci. 2021;22:503–13.
functional connectivity markers track dimensional and categorical features of 285. Tozzi L, Zhang X, Chesnut M, Holt-Gosselin B, Ramirez CA, Williams LM. Reduced
psychotic illness. Mol Psychiatry. 2020;25:2119–29. functional connectivity of default mode network subsystems in depression:
259. Greene AS, Shen X, Noble S, Horien C, Hahn CA, Arora J, et al. Brain-phenotype meta-analytic evidence and relationship with trait rumination. NeuroImage Clin.
models fail for individuals who defy sample stereotypes. Nature. 2022;609:109–18. 2021;30:102570.
260. Arslan S, Ktena SI, Makropoulos A, Robinson EC, Rueckert D, Parisot S. Human 286. Sridharan D, Levitin DJ, Menon V. A critical role for the right fronto-insular cortex
brain mapping: a systematic comparison of parcellation methods for the human in switching between central-executive and default-mode networks. Proc Natl
cerebral cortex. NeuroImage. 2018;170:5–30. Acad Sci USA. 2008;105:12569–74.
261. Marek S, Tervo-Clemmens B, Calabro FJ, Montez DF, Kay BP, Hatoum AS, et al. 287. Northoff G. Psychopathology and pathophysiology of the self in depression —
Reproducible brain-wide association studies require thousands of individuals. Neuropsychiatric hypothesis. J Affect Disord. 2007;104:1–14.
Nature. 2022;603:654–60.
262. Spisak T, Bingel U, Wager TD. Multivariate BWAS can be replicable with mod-
erate sample sizes. Nature. 2023;615:E4–7.
263. Sun X, Sun J, Lu X, Dong Q, Zhang L, Wang W, et al. Mapping neurophysiological
AUTHOR CONTRIBUTIONS
Authors KD and SP conceived, wrote, and approved the manuscript.
subtypes of major depressive disorder using normative models of the functional
connectome. Biol Psychiatry. 2023;94:936–47.
264. Buch AM, Vértes PE, Seidlitz J, Kim SH, Grosenick L, Liston C. Molecular and
network-level mechanisms explaining individual differences in autism spectrum FUNDING
disorder. Nat Neurosci. 2023;26:650–63. KD is supported by a University of Toronto Department of Psychiatry Academic
265. Jha MK, Chin Fatt C, Minhajuddin A, Mayes TL, Trivedi MH. Accelerated brain Scholars Award.
aging in adults with major depressive disorder predicts poorer outcome with
sertraline: findings from the EMBARC study. Biol Psychiatry Cogn Neurosci
Neuroimaging. 2023;8:462–70.
266. Fusar-Poli P, Manchia M, Koutsouleris N, Leslie D, Woopen C, Calkins ME, et al.
COMPETING INTERESTS
The authors declare no competing interests.
Ethical considerations for precision psychiatry: a roadmap for research and
clinical practice. Eur Neuropsychopharmacol. 2022;63:17–34.
267. Stein LA. Making the best use of radiological resources in Canada. Healthpapers.
2005;6:18–23. ADDITIONAL INFORMATION
268. Lacson R, Pianykh O, Hartmann S, Johnston H, Daye D, Flores E, et al. Factors Correspondence and requests for materials should be addressed to
associated with timeliness and equity of access to outpatient MRI examinations. Katharine Dunlop.
J Am Coll Radio. 2024;S1546-1440:00001–2.
269. Siegel JS, Pearson C, Lenze EJ. Better biomarkers, faster drugs, stronger models: Reprints and permission information is available at http://www.nature.com/
progress towards precision psychiatry. Mo Med. 2023;120:292–8. reprints
270. ClinicalTrials.gov Identifier: NCT05017311: Optimized Predictive Treatment In
Medications for Unipolar Major Depression (OPTIMUM-D) (CAN-BIND-17). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
https://clinicaltrials.gov/study/NCT05017311. in published maps and institutional affiliations.
271. ClinicalTrials.gov Identifier: NCT04041479: Biomarker-guided rTMS for Treatment
Resistant Depression (BioTMS). https://clinicaltrials.gov/study/NCT04041479.
272. Ochsner KN, Knierim K, Ludlow DH, Hanelin J, Ramachandran T, Glover G, Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to
et al. Reflecting upon feelings: an fMRI study of neural systems supporting this article under a publishing agreement with the author(s) or other rightsholder(s);
the attribution of emotion to self and other. J Cogn Neurosci. author self-archiving of the accepted manuscript version of this article is solely
2004;16:1746–72. governed by the terms of such publishing agreement and applicable law.

Neuropsychopharmacology