Brain-based correlates of antidepressant response to ketamine: A comprehensive systematic review of neuroimaging studies

Abstract

Background:

Ketamine is an effective treatment for depressed patients who have failed to respond to conventional treatments. However, there is marked variability in response to ketamine and its precise mechanism of action is unclear. Neuroimaging has the potential to provide predictive and mechanistic insights but current findings are limited by small sample sizes and the lack of a comprehensive synthesis across studies. This systematic review focuses on neuroimaging studies investigating baseline (pre-treatment) and longitudinal (post-treatment) brain-based biomarkers of antidepressant response to ketamine.

Methods:

We performed searches in five electronic databases, which were complemented by manual searches (cutoff date: April 26^th, 2022). We included clinical trials (open label and/or randomized controlled) that investigated adults with major depressive disorder or bipolar depression who received at least one treatment with ketamine and/or esketamine. All neuroimaging modalities were included.

Outcomes:

We identified 2,315 articles of which, ultimately, 69 studies were included yielding a combined sample of 1,751 participants. The number of studies investigating esketamine was small (n=2), therefore, this systematic review mainly summarized the evidence relating to racemic ketamine. Overall, there was substantial methodological heterogeneity and there was not a well-replicated neuroimaging biomarker. However, we found convergence across some significant results, particularly in longitudinal biomarkers. The most consistent neuroimaging biomarkers were 1) post-treatment increases in gamma power in frontoparietal regions in electrophysiological studies, 2) post-treatment increases in functional connectivity within the prefrontal cortex, and 3) post-treatment increases in the functional activation of the striatum. Each of these three associations were observed in three independent samples.

Interpretation:

Although a well-replicated neuroimaging biomarker of ketamine response was not identified, there are longitudinal biomarkers that warrant further investigation. Post-treatment increases in gamma power in frontoparietal regions may be particularly promising given its consistency, strong translational potential, and occurrence in early stages of antidepressant response.

Introduction

Major depressive disorder (MDD) and bipolar depression are associated with substantial disability and significantly reduced quality of life.¹ Conventional treatments for these disorders (such as selective serotonin reuptake inhibitors, and serotonin and norepinephrine inhibitors) are moderately effective but have significant limitations including limited rates of remission (particularly in cases with treatment-resistant depression – TRD), a similar mechanism of action (modulation of monoamines), and a delayed onset of action requiring weeks to achieve significant therapeutic effects.^2,3 The discovery of rapid antidepressant effects from (R,S)-ketamine (ketamine) and its (S)-ketamine enantiomer (esketamine) can potentially overcome some of the limitations of conventional treatments as ketamine/esketamine have been shown to be effective even in cases with TRD with an onset of action within hours.^2,3

Nevertheless, there is substantial individual variability in the response to ketamine/esketamine with most studies reporting response rates between 35 and 60%.^2,4 Currently, there are no established clinical or biological predictors of response to ketamine/esketamine and, as a result, clinical use of these rapid-acting agents still relies on a trial-and-error approach. Prior studies have investigated clinical variables and biomarkers associated with response to ketamine/esketamine but mostly with small sample sizes and methodological heterogeneity that have led to mixed and inconclusive results. For example, there is conflicting evidence indicating that clinical variables such as positive personal and/or family history of alcohol use disorder,^5–8 higher body mass index,^5,7,9–11 and the absence of a history of suicide attempt and/or psychiatric hospitalization^5,8,10,12,13 are linked to better response to ketamine/esketamine. Similarly, blood-based biomarkers have thus far shown limited utility in predicting response to ketamine/esketamine.⁴ A systematic review and meta-analysis found that the most consistent association in blood-based biomarkers was longitudinal treatment-associated increases in blood brain-derived neurotrophic factor (BDNF) and response to ketamine/esketamine,⁴ albeit with a modest effect size (Cohen’s d (95% confidence interval) of .26 (.03, .48))⁴ that, by itself, is unlikely to be clinically actionable.

The current limited ability to predict response to ketamine/esketamine with clinical variables and blood-based biomarkers provide further impetus to examine the usefulness of other predictive modalities that may more directly reflect illness-related dysfunction in the brain. Several neuroimaging modalities can be applied to study the structure and function of the central nervous system (CNS) each with intrinsic strengths, limitations and different specificity/sensitivity for investigating specific aspects of the CNS (Table 1).¹⁴ Structural neuroimaging (such as structural magnetic resonance imaging (sMRI) and diffusion tensor imaging (DTI)) obtain anatomical data with high spatial resolution while functional magnetic resonance imaging (fMRI) can measure brain activity as well as functional connectivity (FC) within and between brain regions. Molecular imaging, defined as “the use of neuroimaging techniques to detect and characterize molecular processes other than water”,¹⁵ include approaches such as positron emission tomography (PET) and magnetic resonance spectroscopy (MRS) which enable the examination of specific molecules within the CNS and can provide unique neurochemical information but have limited spatial and temporal resolution. Finally, electrophysiological modalities such as magnetoencephalography (MEG) and electroencephalography (EEG) measure the brain’s electrical activity with high temporal resolution; however, these modalities have limited spatial resolution. A summary of the available studies on structural and functional neuroimaging correlates of response to ketamine/esketamine may not only allow a better prediction of response to these rapid-acting antidepressants but can also be particularly insightful about neurobiological therapeutic mechanisms of ketamine/esketamine.

Table 1.

Overview of the main characteristics of structural and functional neuroimaging modalities

Neuroimaging modality	Mechanism / measurement	Advantages	Disadvantages
Structural neuroimaging: modalities specialized in visualization and analysis of static anatomical data
Structural magnetic resonance imaging (sMRI)	Produces a strong magnetic field that aligns the body’s natural protons and then detects their magnetic properties. Measures the amount of water in different tissues to produce anatomical images	Great spatial resolution (millimeters) Widely available No exposure to ionizing radiation	Contraindications (e.g., metallic/dental devices, pacemaker) Potentially uncomfortable Potential physiological and movement artifacts
Diffusion tensor imaging (DTI)	Uses magnetic resonance imaging techniques to detect the pattern of water diffusion, which is quantified by measures such as fractional anisotropy (FA), radial diffusivity (RD), and apparent diffusion coefficient (ADC)	Great assessment of white matter organization Great spatial resolution (millimeters) No exposure to ionizing radiation	Contraindications (e.g., metallic/dental devices, pacemaker) Potentially uncomfortable Potential physiological and motion artifacts
Functional magnetic resonance imaging (fMRI): modalities that use magnetic resonance imaging to estimate, at rest or during tasks, functional connectivity and activity of brain regions
Resting-state/task-based fMRI	Detects changes in blood oxygenation (through measurement of deoxyhemoglobin concentration) to estimate changes in neuronal activity	Great spatial resolution (millimeters) Widely available No exposure to ionizing radiation	Potential physiological and motion artifacts Contraindications (e.g., metallic devices, pacemaker) Potentially uncomfortable
Arterial spin labeling (ASL)	Uses fMRI techniques to label arterial blood water protons, a freely diffusible intrinsic tracer. ASL is primarily used to measure blood flow of the brain	Great assessment of the blood flow of the brain Reliable and reproducible No exposure to ionizing radiation or other exogenous contrasts	Poor standardization of ASL techniques across studies Low signal-to-noise ratio Contraindications (e.g., metallic/dental devices, pacemaker)
Electrophysiological modalities: modalities that non-invasively examine neural electrical activity
Magnetoencephalography (MEG)	Detects magnetic fields generated by the electrical currents to study the brain electrical activity. One of the advantages of MEG over EEG is that MEG provides the tridimensional location of the electrical activity	Great temporal resolution (milliseconds) Not affected by signal distortion from the scalp and CSF No need to place electrodes on the scalp	Poor spatial resolution (centimeters) High cost / low availability Contraindications (e.g., metallic/dental devices, pacemaker)
Electroencephalography (EEG)	Detects the electrical activity of the brain though non-invasive electrodes placed on the scalp. The currents measured are divided in bands, which are believed to represent different functional components	Great temporal resolution (milliseconds) Low cost / widely available Great portability	Poor spatial resolution (centimeters) Does not accurately assess deep brain structures Lengthy preparation
Molecular modalities: modalities that detect and analyze specific molecules (other than water)
Magnetic resonance spectroscopy (MRS)	Detects radio frequency electromagnetic signals produced by the atomic nuclei within the molecules. Different molecules can be examined and, as a result, tissue metabolites are identified and quantified in vivo	Great sensitivity to specific molecules Provides metabolic data No exposure to ionizing radiation	Poor spatial resolution (centimeters) Poor temporal resolution (minutes) Contraindications (e.g., metallic/dental devices, pacemaker)
Positron emission tomography (PET)	Detects radiation issued by a radioactive substance (radiotracer) to visualize and measure intracellular biochemical changes	Great sensitivity to specific molecules Provides metabolic data Measures variations in the course of the scan	Poor temporal resolution (minutes) Poor spatial resolution Ionizing radiation exposure (injection of radioactive tracer)

Previous reviews partially summarized the available studies examining neural correlates of response to ketamine/esketamine.^16–19 However, not all reviews have examined all neuroimaging modalities such as electrophysiological studies,^17,19 structural neuroimaging^18,19 or molecular neuroimaging,^18,19 and they have generally not focused on the convergence or replicability of the findings in independent samples.^16,18,19 As a result, there has not, to our knowledge, been a comprehensive updated systematic review, which is crucial given the often small and heterogeneous studies, often with overlapping samples.

Hence, in this report we have performed a systematic review of all neuroimaging data focused on ketamine/esketamine and response to treatment. In synthesizing the data, we have focused on: (1) whether baseline (pre-treatment) neuroimaging biomarkers are associated with antidepressant response to ketamine/esketamine in individuals with MDD or bipolar depression; and (2) whether longitudinal (post-treatment) neuroimaging biomarkers are associated with antidepressant response to ketamine/esketamine in individuals with MDD or bipolar depression.

Methods

This review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (PRISMA Checklist available as Supplemental Table 1).

Search strategy

Searches were conducted in five electronic databases: Medline (PubMed), Embase, PsycINFO, The Cochrane Library, and Web of Science. Searches (conducted by CT) were initially performed on April 8, 2020 and were later updated on April 26, 2022. The search terms used are detailed in Supplemental Table 2. Manual searches (conducted by GCM and MM) and studies obtained through personal communication complemented the electronic searches.

Inclusion and exclusion criteria

This systematic review included studies that 1) examined adult human subjects who were 18 years old or older, 2) included patients who were in a major depressive episode (i.e. had active depressive symptoms) in MDD or bipolar disorder, 3) were published in English (there were no restrictions on dates), 4) reported on clinical trials (open label and/or randomized controlled) that administered at least one dose of intravenous or intranasal ketamine and/or esketamine, 5) measured the depressive symptoms using a standardized depression instrument/scale, 6) investigated the participants with any structural or functional neuroimaging modality, and 7) measured the association between change in depressive symptoms after ketamine/esketamine treatment, and pre-treatment and/or post-treatment neuroimaging findings (i.e., manuscripts that measured pre-treatment and/or post-treatment neuroimaging findings but did not examine their relationship with changes in depressive symptoms were not included). We excluded studies that 1) investigated non-human animals, 2) were based solely on conference abstracts, case reports/case series and editorials,²⁰ 3) assessed healthy volunteers and/or individuals with depressive symptoms due to other disorders besides MDD and BD, 4) examined the association of pre-treatment and/or post-treatment neuroimaging findings with only a specific depressive symptom (e.g., suicidality, anhedonia) and not an overall depression scale, 5) included concurrent neuromodulatory treatments including vagal nerve stimulation, electroconvulsive therapy, and repetitive transcranial magnetic stimulation (psychotherapy and concurrent medications were not exclusion criteria), 6) examined individuals with serious/unstable comorbid neurological and/or medical diseases, and 7) studied subjects in surgical/perioperative settings.

Manuscript selection

The manuscript selection process was comprised of an initial screening of titles and abstracts to choose the articles warranting a full-text review, followed by a selection of the manuscripts that were ultimately included in this systematic review. The manuscript selection team consisted of four investigators (GCM, MM, SM and FSG) and, in each of the two selection phases, a manuscript was independently reviewed by two investigators. Disagreements between the two independent reviewers were resolved by consensus discussion. If two or more manuscripts reported on the same results on the same dataset, we only included the most comprehensive report in this systematic review.

Data extraction and data report

The list of variables extracted for each manuscript and the database used are available as supplemental material.

In order to appropriately answer our research questions, main findings focused on the association between neuroimaging findings (baseline and/or longitudinal) and change in depressive symptoms in individuals who were treated with ketamine/esketamine. Improvement of depressive symptoms included continuous measures of improvement in depression scores (such as percentage and absolute changes in depression scores) and/or categorical measures of improvement (such as response and remission). The data extraction team consisted of three investigators (GCM, MM and ID) and the data for each individual study was independently extracted by two investigators.

Results

Our search identified 2,315 articles (database searching: 2,277; hand searching: 35; personal communication: 3). After the screening of titles and abstracts, 168 were considered potentially eligible for inclusion in this systematic review and, therefore, underwent full-text review (Figure 1). Sixty-nine manuscripts were ultimately included yielding a combined sample of 1,751 participants. Most of the included studies used only one neuroimaging modality (n=64, 93%) and investigated only individuals with MDD (n=64, 93%). All 69 studies treated individuals with intravenous racemic ketamine and two of these studies (3%) had a sub-sample of participants who were treated with intravenous esketamine.^21,22 Since the two studies that included individuals treated with esketamine did not conduct separate analyses for participants who received only esketamine, this systematic review mostly comprises of evidence on neuroimaging correlates of antidepressant response to racemic ketamine.

Figure 1.

PRISMA flow diagram of systematic searches for studies assessing neuroimaging biomarkers of antidepressant response to ketamine/esketamine

^a Databases used in the searches: Medline (PubMed), Cochrane Library, Embase, PsycInfo, and Web of Science.

^b Hand searching included screening of review articles, grey literature, included papers and manuscripts that cited the included papers.

Ten (15%) studies investigated structural neuroimaging (sMRI and DTI), twenty-nine (42%) studies reported on fMRI (resting-state fMRI, task-based fMRI and ASL), 20 (29%) studies examined electrophysiological modalities (MEG and EEG), and 15 (22%) studies used molecular modalities (PET and MRS). Thirty-three (48%) articles reported on baseline (pre-treatment) neuroimaging findings while 59 (86%) reported on longitudinal (post-treatment) neuroimaging findings. Fifty-five (80%) articles reported statistically significant associations between neuroimaging findings and response to ketamine. The main brain regions involved in statistically significant findings were the PFC (16 studies),^21,23–37 the ACC (12 studies),^{21,28,34,38–46} and the striatum (nine studies).^{23,27,29,37,47–51}

Table 2 describes the main characteristics of the included studies.

Table 2.

Main characteristics of studies using neuroimaging to examine ketamine/esketamine antidepressant effects (n=69 studies)

Variables	n (%) or median (25th percentile - 75th percentile)
Number of neuroimaging modalities used in individual studies
One modality	64 (93%)
Two or more modalities	5 (7%)
Neuroimaging modality used
Structural modalities (Structural magnetic resonance imaging (sMRI) / Diffusion tensor imaging (DTI))	10 (15%)
- sMRI	7 (10%)
- DTI	3 (4%)
Functional magnetic resonance imaging (fMRI)	29 (42%)
- Resting-state fMRI	19 (28%)
- Task-based fMRI	7 (10%)
- Arterial spin labeling	3 (4%)
Electrophysiological modalities (Magnetoencephalography (MEG) / Electroencephalography (EEG))	20 (29%)
- Resting-state MEG	4 (6%)
- Task-based MEG	8 (12%)
- Resting-state (awake) EEG	4 (6%)
- Task-based EEG	2 (3%)
- Sleep EEG	2 (3%)
Molecular modalities (Magnetic resonance spectroscopy (MRS) / Positron emission tomography (PET))	15 (22%)
- MRS	8 (12%)
- PET	7 (10%)
Medication used
Only ketamine	67 (97%)
Both ketamine and esketamine	2 (3%)
Number of infusions
Single	59 (86%)
Multiple (two or more)	10 (14%)
Number of participants in individual studies	22 (16 – 32)
Average age of participants in individual studies (n=66 studies)	41.5 (36.1 – 45.0)
Average percentage of women in individual studies (n=66 studies)	55.5 (41.25 – 61.0)
Primary diagnosis included in the studies
Only major depressive disorder (MDD)	64 (93%)
Only bipolar depression	3 (4%)
Combination of individuals with MDD and bipolar depression	2 (3%)
Association studied
Only longitudinal neuroimaging findings	36 (52%)
Both baseline and longitudinal neuroimaging findings	23 (33%)
Only baseline neuroimaging findings	10 (15%)
Reported statistically significant association(s)a
Yes	55 (80%)
No	14 (20%)
Clinical trial design
Randomized controlled trial (RCT)	35 (51%)
Open-label trial	32 (46%)
Combination of different designs	2 (3%)
Primary depression outcome measure
Montgomery-Asberg Depression Rating Scale (MADRS)	49 (71%)
Hamilton Depression Rating Scale (HDRS)	14 (20%)
More than one measure	4 (6%)
Beck Depression Inventory (BDI)	2 (3%)
Allowed concurrent antidepressants
Yes	27 (39%)
No	42 (61%)
Number of failed antidepressant trials to meet inclusion criteria	2 (1 – 2)
Average current prevalence of any anxiety disorder in individual studies (n=15 studies)	70% (55% - 75%)

^aAssociation(s) between response to ketamine/esketamine (analyzed categorically and/or continuously), and baseline and/or longitudinal neuroimaging findings in the total sample (i.e., statistically significant associations in in sub-samples were not included)

Structural Neuroimaging (sMRI and DTI): 10 studies

Ten studies examined structural neuroimaging biomarkers of response to ketamine (Supplemental Table 3). Seven studies used sMRI^{40,44,47,52–55} and three studies used DTI.^41,56,57 Eight of the ten (80%) studies reported statistically significant associations between neuroimaging findings and response to ketamine.

Baseline (pre-treatment) findings in structural neuroimaging

Nine studies reported on baseline structural neuroimaging biomarkers of response to ketamine,^{40,41,44,52–57} six using sMRI and three using DTI. There was convergence in two types of findings. In sMRI studies, there was convergence between two studies that found an association between smaller baseline hippocampal volume and greater improvement of depressive symptoms.^44,52 In DTI studies, there was convergence in the results of two studies that found an association between greater baseline fractional anisotropy (FA) in the cingulum^56,57 and greater improvement of depressive symptoms (Supplemental Results).

Longitudinal (post-treatment) findings in structural neuroimaging

Six studies reported on longitudinal (post-treatment) structural biomarkers of response to ketamine^{40,44,47,53,55,57} five using sMRI and one using DTI. Three studies reported statistically significant findings,^47,55,57 two of them (using two independent samples) involving the hippocampus.^47,55 However, there was no significant convergence between positive findings (Supplemental Results).

fMRI (resting-state fMRI, task-based MRI and ASL): 29 studies

Twenty-nine studies examined the correlates of response to ketamine using fMRI^{21–25,27,29,35,37,38,41,44,48,50,58–72} (Supplemental Table 4), of which nineteen studies used blood-oxygenation level-dependent (BOLD) resting-state fMRI,^{21,23–25,27,35,38,41,44,50,58–66} seven studies used task-based BOLD fMRI^{37,48,67–71} and three studies used arterial spin labeling (ASL) during resting-state.^22,29,72 Twenty studies investigated FC and eleven studies examined measures of fMRI brain activation. The tasks used significantly varied with two studies using the same task (NoGo/Go task) and each of the other five studies using a distinct task (Supplemental Table 4).^70,71 Since ASL measures cerebral blood flow, which is hypothesized to be correlated with brain activation, these studies were also included.⁷³ Twenty-two of the 29 (76%) studies using fMRI-based modalities reported statistically significant associations between neuroimaging findings and response to ketamine.

Baseline (pre-treatment) findings in fMRI

Fourteen studies reported baseline neural correlates of response to ketamine,^{21,22,24,27,41,44,48,50,64,66,68,70–72} of which ten studies examined resting-state and four investigated neural correlates during tasks. Ten studies investigated FC, of which eight had statistically significant findings. Five studies investigated brain activation, of which three had statistically significant findings. There was no significant convergence between positive findings relating to FC or brain activation.

Longitudinal (post-treatment) findings in fMRI

Twenty-seven studies examined the association between longitudinal neuroimaging findings and response to ketamine,^{21–25,29,35,37,38,44,48,50,58–72} of which twenty studies examined resting-state and seven investigated neural correlates during tasks. Twenty studies investigated longitudinal changes in FC, of which nine had statistically significant findings. Eleven studies examined the association between longitudinal changes in brain activation and response to ketamine, of which eight found statistically significant associations. In resting-state studies, there was convergence in findings associating response to ketamine with post-ketamine increases in FC in the PFC (Supplemental Results).^23,25 Six of the seven task-based studies reported statistically significant findings. However, there was no significant convergence between the positive results.

Electrophysiological modalities (MEG and EEG): 20 studies

Twenty studies investigated the neural correlates of response to ketamine with electrophysiological neuroimaging modalities (Supplemental Table 5).^{26,28,31,33,34,39,42,43,61,74–84} Twelve studies used MEG, of which four used resting-state MEG^33,34,39,80 and eight used task-based MEG,^{42,43,74–79} while eight studies used EEG, of which four used wake resting-state EEG,^{26,28,31,61,81} two used sleep EEG^82,83 and two used task-based EEG.⁸⁴ In the ten task-based studies (eight using MEG and two using EEG), the task used significant varied with four studies using passive somatosensory stimulation task and each of the other six studies using a distinct task (Supplemental Table 5). Seventeen of the 20 (85%) studies using electrophysiological modalities reported statistically significant associations between neuroimaging findings and response to ketamine. There was significant sample overlap between the studies that used MEG, with all twelve studies ultimately coming from two independent samples. The eight EEG studies were comprised of five independent samples.

Baseline (pre-treatment) findings in electrophysiological modalities

Six electrophysiological studies reported on baseline biomarkers of response to ketamine.^{26,28,33,42,43,83} All six studies reported statistically significant findings but the specific electrophysiological correlates highlighted were variable, and there was no significant convergence between positive results.

Longitudinal (post-treatment) findings in electrophysiological modalities

Seventeen studies investigated the association between longitudinal changes in electrophysiological parameters and response to ketamine.^{26,28,31,33,34,39,61,74–82,84} All studies used distinct analytical approaches, complicating comparisons between studies. With this caveat in mind, there was convergence in two types of findings: 1) response to ketamine was associated with post-ketamine increases in power in the gamma band in frontoparietal regions,^28,31,76 an EEG frequency range (typically between 30 and 50 Hz) that is closely related with the generation of action potential by cortical neurons,^85,86 (observed in three studies using three independent samples), and 2) response to ketamine was associated with post-ketamine decreases in theta cordance, an electrophysiological measure that correlates with brain consumption of energy⁸⁷ (observed in two studies with two independent samples) (Supplemental Results).^26,28

Molecular modalities (MRS and PET): 15 studies

Fifteen studies examined the neural correlates of response to ketamine with molecular neuroimaging modalities (Supplemental Table 6).^{30,32,36,41,45,46,49,51,53,88–93} Eight studies used MRS^{32,36,41,45,46,88–90} and seven studies used PET.^{30,49,51,53,91–93} Nine studies investigated glutamatergic function,^{32,36,41,45,46,88–90,93} six studies investigated GABAergic function,^{32,36,45,46,88,90} five studies (all using PET) examined [18F]-fluorodeoxyglucose^{30,49,53,91,92} to measure cerebral glucose metabolism, and one study examined serotonergic function (changes in 5-HT1b receptor availability).⁵¹ Eleven of the 15 (73%) studies using molecular modalities reported statistically significant associations between molecular/biochemical function and response to ketamine.

Baseline (pre-treatment) findings in molecular modalities

Eight studies using molecular neuroimaging reported on baseline biomarkers of response to ketamine.^{36,41,46,51,53,88,89,92} Four studies reported statistical significance but there was no significant convergence in positive results.

Longitudinal (post-treatment) findings in molecular modalities

Twelve studies examined the association between longitudinal changes in molecular/biochemical function and response to ketamine.^{30,32,45,49,51,53,88–93} Eight studies reported statistically significant results but there was no significant convergence between the positive findings (Supplemental Results).

Convergence in findings across neuroimaging modalities

A cross-modality comparison (including structural, fMRI-based, electrophysiological and molecular modalities) revealed significant disparities across the results, with no “strict” convergence (replication) between study-specific positive findings. However, at a broader level, some findings showed convergence in two or more studies, warranting further consideration.

In the baseline (pre-treatment) neuroimaging analyses, there were associations between response to ketamine and 1) smaller baseline hippocampal volume in sMRI studies^44,52 and 2) greater baseline FA in the cingulum in DTI studies.^56,57 However, an important caveat is that although these results showed convergence in two independent samples, the association with smaller baseline hippocampal volume was much less consistent since there were four negative studies^40,53–55 while there were no negative studies for the association with greater baseline FA in the cingulum.

Six longitudinal (post-treatment) neuroimaging biomarkers showed convergence in independent samples, of which three biomarkers were observed in three independent samples while the other three biomarkers were observed only in two independent samples. All convergent longitudinal findings were related to functional neuroimaging studies. Findings that broadly replicated in three independent samples included: 1) post-treatment increases in gamma power in frontoparietal regions, a putative marker of cortical excitability and synaptic potentiation, in electrophysiological studies (observed in two EEG studies^28,31 and one MEG study,⁷⁶ with only one negative study)⁶¹; 2) post-treatment increases in FC within the PFC (observed in two fMRI-based studies^23,25 and one MEG study,³⁴ with three negative studies);^60,61,80 and 3) post-treatment increases in activation within the striatum (observed in two fMRI-based studies^29,37 and one PET study,⁴⁹ with two negative studies).^38,48 The other three convergent associations (that were observed in two independent samples) were the relationship between response to ketamine and 1) post-treatment increases in FC within the striatum (observed in two fMRI-based studies,^23,48 with one negative study),⁸⁰ 2) post-treatment increases in activation within the PFC (one fMRI-based study³⁷ and one PET study,³⁰ with three negative studies);^29,68,91 and 3) decreases in theta cordance (observed in two EEG studies,^26,28 with no negative studies).

For findings that were convergent in at least two independent samples (26 studies), we compared the characteristics of the studies that had convergent findings (18 studies) versus those without convergent findings (8 studies). There were no statistically differences in any characteristics including number of ketamine infusions, ketamine dose, study design and primary diagnosis (Supplemental Table 7).

Discussion

We performed a systematic review of 69 studies that examined baseline (pre-treatment) and longitudinal (post-treatment) neuroimaging biomarkers. Although ketamine and esketamine may have overlapping mechanisms of action, the number of studies investigating esketamine was small (n=2), therefore, this systematic review mainly summarized the evidence relating to racemic ketamine. While we found no systematically replicated results due to the substantial methodological heterogeneity across studies, promising convergence was seen in longitudinal (post-treatment) biomarkers including the association between response to ketamine and 1) post-treatment increases in gamma power in frontoparietal regions, 2) post-treatment increases in FC within the PFC, and 3) post-treatment increases in activation within the striatum. Each of these three associations was observed in three independent samples; however, the most consistency was seen in studies of increased gamma power in frontoparietal regions.

The heterogeneity in the findings included in this systematic review might be explained by several factors. First, most studies had relatively small sample sizes with the median (25%; 75%) sample size being 22 (16; 32) participants. Small sample sizes are at the same time statistically underpowered⁹⁴ and susceptible to yielding inaccurate effect sizes that replicate poorly.⁹⁵ In addition, the fact that 55 of the 69 studies (80%) included in this systematic review reported statistically significant findings together with the inconsistency of results and the small sample sizes suggests the likelihood of publication bias. Third, there was significant heterogeneity of analytical methods and statistical strategy, including variable approach to correction for potential confounders (e.g., age, sex, body mass index – Supplemental Table 8).⁹⁴ Fourth, there was substantial variability in study characteristics including factors such as inclusion criteria, sample characteristics such as prevalence of concurrent psychiatric disorders, treatment schedule, and timepoints when neuroimaging outcomes were obtained. Fifth, MDD is an intrinsically heterogeneous disorder.

In addition, this systematic review has several limitations that need to be taken into account. In particular details of each study’s statistical analyses (specific tests used, adjustment for multiple comparisons) were not always described in detail and limited our ability to perform a more quantitative comparison of the literature. Similarly, given the lack of available individual-level data, we were not able to reliably conduct meta-analytical calculations. Third, we reported the brain regions as named in the original studies, which does not necessarily mean they have the exact same brain coordinates based on the atlas used. Finally, neuroimaging studies usually do not include individuals with severe active symptoms (such as acute suicidality), which may limit the generalizability of our findings.

Despite these caveats, an encouraging finding was the association between post-ketamine increased gamma power in frontoparietal regions and response to ketamine. This association was investigated in four independent samples and, in three of them, this association was statistically significant in individuals receiving the standard antidepressant dose of ketamine (0.5 mg/kg). Increases in gamma power have a strong translational potential since they may be captured by EEG, a widely available neuroimaging modality, and are observed in early stages of antidepressant response (within minutes to hours after treatment with ketamine). Another result seen broadly in three independent samples was the association between response to ketamine and post-treatment increases in FC within the PFC, which may counteract the hypoconnectivity of the PFC previously found in MDD and bipolar depression.^3,96 Finally, post-treatment increases in the activity within the striatum, a brain region involved in reward processing, is consistent with previous findings that depressed individuals have hypoactive striatum and that ketamine, a medication with significant anti-anhedonic action, may normalize this hypoactivity.⁹⁷

The three brain-based correlates of response to ketamine that were most widely convergent are broadly consistent with ketamine’s proposed main mechanisms of action, which may underlie our findings. Pre-clinical studies conclude that ketamine is a N-methyl-D-aspartate (NMDA) receptor antagonist that preferentially blocks GABAergic inhibitory interneurons.⁹⁸ This blockage in GABA transmission results in a downstream increase of neuronal firing and resulting glutamate release (disinhibition hypothesis). As the EEG gamma band is closely related to the generation of action potentials by cortical neurons,^85,86 increases in gamma power may reflect the strengthening of glutamatergic neurotransmission as a result of ketamine. In addition, this increase in glutamatergic transmission triggers intracellular pathways that ultimately increase levels of neurotrophins, such as BDNF, that leads to synaptogenesis.⁹⁸ This increase in synaptic strength may be seen in functional neuroimaging as increases in FC such as the one observed within the PFC in three independent samples.^23,25,34 The strengthening of glutamatergic transmission and increased synaptogenesis, as well as ketamine’s potential effect on dopamine, either direct of via glutamatergic connections, may also explain the increases in the activity within the striatum.

Interestingly, six of the eight replicated neuroimaging biomarkers were longitudinal (post-treatment), which raises the question as to why there were fewer replicated baseline (pre-treatment) neuroimaging biomarkers. First, there were fewer publications that addressed baseline neuroimaging biomarkers compared to longitudinal studies. Although data on baseline neuroimaging biomarkers were obtained in all 69 studies included in this systematic review, less than half of the studies (n=33/69, 48%) reported on them while 86% (n=59/69) reported on longitudinal neuroimaging biomarkers. In addition, baseline biomarkers may be intrinsically more difficult to identify due to prominent inter-subject heterogeneity as opposed to longitudinal biomarkers, where each individual can effectively act as their own control in a more controlled analytical framework. As a result, the identification of inter-subject (baseline) biomarkers will likely require larger sample sizes and more effective subtyping to reduce heterogeneity.

Given the inconsistencies and possible publication bias of the current literature, future studies investigating brain-based biomarkers of response to ketamine should address the limitations of previous studies (see also Supplemental Discussion). Table 4 summarizes some recommendations for improving the reproducibility/replicability and the clinical utility of neuroimaging studies. First, it important to investigate larger samples see also reduce the noise due to sampling variability, which may be particularly pronounced when studying a highly heterogeneous condition such as depression. This may be achieved with development of multisite collaborations and open sharing of neuroimaging data. Second, the preregistration in publicly available repositories can improve research transparency, accountability and credibility of the results, and decrease multiple testing and selective reporting. Third, it is crucial to appropriately control for common confounders (Supplemental Table 8). Finally, modalities with good temporal resolution, such as EEG and MEG, may be particularly useful to examine rapid-changing phenomena such as those during ketamine or immediately after ketamine infusion. Studies using modalities highly susceptible to physiological and motion artifact, such as sMRI and fMRI, should incorporate protocols that predefine exclusion motion threshold and conduct post hoc corrections.

Table 4.

Recommendations to improve the reproducibility and replicability of neuroimaging studies investigating neural correlates of antidepressant response to ketamine/esketamine

Recommendations	Rationale	Actions that facilitate the implementation of the recommendations
Investigation of larger samples	Most neuroimaging studies examine small samples, which are underpowered, more likely to provide inaccurate/inflated effect sizes and false positive findings. Larger samples decrease the noise due to pronounced sampling variability, which may be particularly pronounced when studying a highly heterogeneous condition such as depression	Development of neuroimaging consortiums and multisite collaborations Open sharing of neuroimaging data as, for example, in free and open-source platforms
Preregistration in publicly available repositories	Most neuroimaging studies are not preregistered, which may facilitate the use of inappropriate analytical designs and of multiple testing. Preregistration allows for a better distinction between exploratory and confirmatory studies, improve research transparency, accountability and credibility of the results, and may decrease selective reporting	Use of free and open-source platforms that allow the researchers to register and share their studies (e.g., Open Science Framework (OSF) and the AsPredicted platform) Journal/editorial policies that require and/or prioritize preregistered neuroimaging studies such as publishing preregistered studies even if the results are negative
Appropriate controlling of confounders	Most studies do not describe/perform controlling for confounders. Procedural (such as motion) and patient-related confounders (such as age, gender, body mass index, other medications/substances, other psychiatric comorbidities) may significantly affect neuroimaging results	Study protocols that decrease motion by minimizing time, increasing comfort and familiarization in the scanner. Predefine exclusion motion threshold and conduct post hoc corrections Conduct and describe controlling of main patient-related confounders Research design that minimizes the impact of medications and substances When combining distinct samples, use harmonization techniques such as the ComBat algorithm

Conclusion

A well-replicated neuroimaging biomarker of ketamine response was not identified in this systematic review. However, we found convergence across some significant results, particularly in longitudinal (post-treatment) biomarkers, that warrant further investigation. Post-treatment increases in gamma power in frontoparietal regions may be particularly promising given its consistency, strong translational potential, and occurrence in early stages of antidepressant response.

Contributors

GCM wrote the first draft of the report, set up the database, conducted manual searches, selected included studies, and extracted, verified, analyzed and interpreted the data. MM assisted in the writing of the first draft of the report, conducted manual searches, selected included studies, extracted and verified the data, interpreted the data and critically reviewed/edited the report for important intellectual content. ID assisted in the writing of the first draft of the report, participated in the data extraction, interpreted the data and critically reviewed/edited the report for important intellectual content. MJR interpreted the data and critically reviewed/edited the report for important intellectual content. SM selected included studies, interpreted the data and critically reviewed/edited the report for important intellectual content. CT conducted the systematic searches in the electronic databases, interpreted the data and critically reviewed/edited the report for important intellectual content. GSS interpreted the data and critically reviewed/edited the report for important intellectual content. TDG assisted in the writing of the first draft of the report, interpreted the data and critically reviewed/edited the report for important intellectual content. CAZJ interpreted the data and critically reviewed/edited the report for important intellectual content. FSB provided input into the design and conceptualization of the study, interpreted the data and critically reviewed/edited the report for important intellectual content. FSG designed and conceptualized the study, provided supervision, assisted in the writing of the first draft of the report, verified the data, interpreted the data and critically reviewed/edited the report for important intellectual content.

All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Declaration of interests

GSS conducted an investigator-initiated study that used medication (vortioxetine) provided without charge by Lundbeck. TDG is listed as co-author on patent and patent applications related to the pharmacology and use of (2R,6R)-hydroxynorketamine in the treatment of depression, anxiety, anhedonia, suicidal ideation, and post-traumatic stress disorder. He has assigned his patent rights to the University of Maryland Baltimore, but will share a percentage of any royalties that may be received. TDG has received research funding from Allergan and Roche Pharmaceuticals. CAZJ is listed as a co-inventor on a patent for the use of ketamine in major depression and suicidal ideation; as a co-inventor on a patent for the use of (2R,6R)-hydroxynorketamine, (S)-dehydronorketamine, and other stereoisomeric dehydroxylated and hydroxylated metabolites of (R,S)-ketamine metabolites in the treatment of depression and neuropathic pain; and as a co-inventor on a patent application for the use of (2R,6R)-hydroxynorketamine and (2S,6S)-hydroxynorketamine in the treatment of depression, anxiety, anhedonia, suicidal ideation, and post-traumatic stress disorders. He has assigned his patent rights to the U.S. government but will share a percentage of any royalties that may be received by the government. FSB is a scientific advisor for WavePaths, Ltd, and MindState Design Labs, Inc. FSG has received research grant support from Janssen Therapeutics. GCM, MM, ID, MJR, SM, CT, and GSS do not have conflicts of interest to report.

Data sharing

The database with the data of the individual studies included in this systematic review and a data dictionary were made available as supplemental material.

Table 3:

Convergence between neuroimaging biomarkers of antidepressant response to ketamine.

Convergent Finding	Number of independent samples where the finding was observed	Number of independent samples where the finding was not observed
Baseline (pre-treatment) findings
Response to ketamine was associated with greater baseline fractional anisotropy in the cingulum	Two independent samples Vasavada et al. 2016 (DTI); Sydnor et al. 2020 (DTI)	None
Response to ketamine was associated with smaller baseline hippocampal total volume	Two independent samples Abdallah et al., 2015 (sMRI); Siegel et al., 2021 (sMRI)	Four independent samples Ortiz et al., 2015 (sMRI); Niciu et al., 2017 (sMRI); Zhou et al., 2020 (sMRI); Herrera-Melendez et al., 2021 (sMRI)
Longitudinal (post-treatment) findings
Response to ketamine was associated with increased gamma power	Three independent samples Nugent et al., 2019c (MEG); de la Salle et al., 2022 (EEG); Lijffijt et al., 2022 (EEG)	One independent sample McMillan et al., 2020 (EEG)
Response to ketamine was associated with post-ketamine increases in functional connectivity within the prefrontal cortex	Three independent samples Abdallah et al., 2017a (rs-fMRI); Abdallah et al., 2018 (rs-fMRI); Nugent et al., 2020 (MEG)	Three independent samples Nugent et al., 2016 (MEG); Kraus et al., 2020 (rs-fMRI); McMillan et al., 2020 (rs-fMRI)
Response to ketamine was associated with post-ketamine increases in activation within the striatum	Three independent samples Nugent et al., 2014 (PET); Sterpenich et al., 2019 (tb-fMRI); Gonzalez et al., 2020 (ASL)	Two independent samples Murrough et al., 2015 (tb-fMRI); Downey et al., 2016 (rs-fMRI);
Response to ketamine was associated with post-ketamine increases in functional connectivity within the striatum	Two independent samples Murrough et al., 2015 (tb-fMRI); Abdallah et al., 2017a (rs-fMRI);	One independent sample Nugent et al., 2016 (MEG)
Response to ketamine was associated with post-ketamine increases in activation within the prefrontal cortex	Two independent samples Li et al., 2016 (PET); Sterpenich et al., 2019 (tb-fMRI);	Three independent samples Carlson et al., 2013 (PET); Gonzalez et al., 2020 (ASL); Loureiro et al., (2020) (tb-fMRI)
Response to ketamine was associated with post-ketamine decreases in theta cordance in the prefrontal cortex	Two independent samples Cao et al., 2018 (EEG); de la Salle et al., 2022 (EEG)	None

Abbreviations: ASL = arterial spin labeling; DTI = diffusion tensor imaging; EEG = electroencephalography; MEG = magnetoencephalography; PET = positron emission tomography; rs-fMRI = resting-state functional magnetic resonance imaging; sMRI = structural magnetic resonance imaging; tb-fMRI = task-based functional magnetic resonance imaging