Do NMDA-R Antagonists Re-Create Patterns of Spontaneous Gamma-Band Activity in Schizophrenia? A Systematic Review and Perspective

Bianciardi and colleagues situate their review within the debate over the role of N-methyl-D-aspartate receptor (NMDA-R) hypofunction in the pathophysiology of schizophrenia. Earlier research has shown that NMDA-R antagonists such as ketamine and phencyclidine can transiently produce a range of schizophrenia-like symptoms and cognitive deficits in healthy volunteers, and post-mortem, genetic and imaging data implicate NMDA-R abnormalities—including reductions in NR1 subunits and effects on interneurons enriched for GAD67—in the disorder. Because parvalbumin-positive GABAergic interneurons and NMDA/AMPA receptor interactions are central to generating gamma-band oscillations (approximately 30–200 Hz), the authors reason that NMDA-R hypofunction could perturb excitation/inhibition (E/I) balance and thereby alter spontaneous gamma-band amplitude and network organisation. To test whether pharmacological NMDA-R hypofunction recreates the gamma-band signatures reported in schizophrenia, the study reports a systematic review comparing effects of NMDA-R antagonists on spontaneous gamma-band power and functional connectivity in preclinical and human studies with resting-state EEG/MEG findings from clinical high-risk, first-episode and chronic schizophrenia samples. The primary aim was to establish consistency between antagonist-induced changes and patient data; a secondary aim examined effects on functional connectivity. The review included 24 preclinical studies, 9 human NMDA-R antagonist studies, and 27 resting-state EEG/MEG studies in schizophrenia patients.

The investigators conducted literature searches in PubMed and Google Scholar up to August 2019 using tailored search terms for the three domains: preclinical NMDA-R antagonist studies (terms such as Ketamine, PCP, MK801, NMDA, Gamma), human ketamine studies (Human, Ketamine, NMDA, Gamma), and resting-state EEG/MEG studies (Resting state, EEG/MEG, EEG/LFP/EcoG). Reference lists of relevant articles were also screened. Titles and abstracts were screened and duplicates removed; reviews, meta-analyses, case reports and task-based studies were excluded, as the focus was strictly on spontaneous resting-state activity. Eligibility criteria differed by study type. Preclinical inclusion required in vivo administrations of ketamine, PCP or MK-801 at subanaesthetic dosages with EEG/LFP/ECoG recordings. Human ketamine studies needed EEG/MEG recordings in medication-free, psychiatrically and substance-abuse-free participants with sample sizes of at least 10. Schizophrenia resting-state studies required EEG/MEG recordings, a healthy control group, and at least 10 patients. The authors excluded studies that only examined baseline during tasks and studies with inadequate artefact control (for example, those relying solely on visual artefact correction were excluded because ocular and muscular artefacts overlap with high-frequency neuronal activity). Data extraction was performed by BB (first author) and supervised by PU; the authors contacted study authors when effect sizes were missing. For statistical analysis, non-parametric Kruskal–Wallis and Fisher–Freeman–Halton exact tests were used to probe study-level factors associated with direction of gamma-band effects in schizophrenia. Effect sizes for schizophrenia studies were calculated as Hedges' g using Comprehensive Meta-Analysis software when means or inferential statistics were available; standardised mean differences with 95% confidence intervals were plotted in R. Funnel plots and Egger's regression tested for publication bias. Risk-of-bias assessments followed established tools: Cochrane guidelines for randomized human studies, SYRCLE for animal studies, and ROBINS-I for matched-cohort schizophrenia studies. The authors note that pooled effect-size analyses were restricted to schizophrenia studies because preclinical and human ketamine studies lacked sufficient comparable data for meta-analysis.

Study selection yielded 24 eligible preclinical studies (mostly rats, with one monkey and two mice), 9 human NMDA-R antagonist studies (ketamine, MK-801, PCP in healthy volunteers), and 27 resting-state EEG/MEG studies in schizophrenia (including chronic, first-episode and clinical high-risk groups). Preclinical work predominantly used male animals, mostly freely moving recordings, and acute drug administration; recordings included LFP, ECoG and intracranial EEG across cortical and subcortical sites. Human ketamine studies were mainly crossover designs using subanaesthetic infusions (0.25–0.65 mg/kg) with EEG or MEG, and schizophrenia studies varied in patient stage, medication status and recording parameters (eyes-closed more common). Risk-of-bias assessment indicated that among preclinical studies 14 were judged low risk, 5 unclear and 5 high risk. For human ketamine studies, 5 were low risk, 2 high and 2 unclear. In the schizophrenia literature 14 studies were low risk, 6 unclear and 7 high risk. Egger's regression did not indicate publication bias for the schizophrenia effect-size dataset (t = 0.2439, df = 20, p = 0.8098). Preclinical findings: the large majority of preclinical studies (reported as n = 20 of the included set) found that NMDA-R antagonists increased gamma-band power. A smaller number reported no effect or mixed, dose-dependent outcomes: several studies described increases at lower dosages and decreases at higher dosages, and at least one study contrasted acute increases with decreases following chronic administration. Connectivity analyses were reported in six preclinical studies, five of which found increased gamma-band connectivity while one found no change. No systematic differences emerged between antagonist types (ketamine, MK-801, PCP) in their effects on gamma-band amplitude, nor clear dose-related patterns across the literature; definitions of ‘‘gamma’’ varied between lower (30–60 Hz) and higher (>60 Hz) ranges. Human ketamine studies: eight out of nine studies reported increases in gamma-band power following ketamine administration. One study using a Hidden Markov model observed a heterogeneous pattern of both increases and decreases in the 25–45 Hz range across discrete brain states. Four studies reported effects in both low and high gamma ranges, localisation was sometimes cortical and sometimes both cortical and subcortical, and two studies reported connectivity analyses with opposing findings (one reduced fronto-parietal effective connectivity, one increased functional connectivity). Correlations between ketamine-induced gamma changes and psychopathology were inconsistent across three studies. Schizophrenia resting-state studies: results were heterogeneous. Ten out of 27 studies reported gamma-band upregulation (most in chronic patients), seven reported reductions in gamma power, eight reported no group differences, and two reported mixed regional or frequency-dependent results. Connectivity outcomes across 14 studies were mixed: one reported increased sensor-level connectivity, four reported decreases, six reported no difference, and three reported mixed patterns contingent on region or illness subgroup. Fifteen studies examined symptom correlations: six found associations with positive symptoms (two of these negative correlations), one found a positive correlation with negative symptoms, seven reported no association, and some studies linked high-gamma abnormalities with cognitive deficits. The authors attempted to identify moderators (illness stage, duration, medication, recording method, frequency range, eyes-open/closed) but found no systematic differences explaining the divergent schizophrenia findings.

Bianciardi and colleagues interpret the evidence as showing a relatively consistent effect of acute NMDA-R antagonism—across species and recording modalities—to increase spontaneous gamma-band power and, in preclinical studies, to increase gamma-band connectivity. They note this consistency across ketamine, MK-801 and PCP despite methodological heterogeneity, and they emphasise that effects were observed across cortical and subcortical structures and in both low and high gamma ranges. These findings align with models in which NMDA-R hypofunction on interneurons shifts E/I balance and disinhibits pyramidal cells, producing elevated high-frequency activity. However, the authors stress a central disjunction: schizophrenia resting-state EEG/MEG findings do not uniformly mirror the antagonist-induced pattern. Patient studies are heterogeneous—showing increases, decreases or no change in gamma power—so acute pharmacological NMDA-R hypofunction does not map neatly onto the empirical signatures reported in patients. The investigators argue that several factors may contribute to this mismatch, including differences between acute and chronic NMDA-R hypofunction (with some evidence that chronic hypofunction may reduce gamma), methodological variability across patient studies, and the possibility that the pharmacological effects may not be spectrally identical to true oscillatory signals observed in patients. They therefore recommend that future work should characterise whether drug- and patient-related high-frequency changes reflect bona fide narrowband oscillations or broadband, non-rhythmic activity, because the functional and mechanistic interpretations differ. The authors acknowledge important limitations: NMDA-R antagonists have off-target effects on other neurotransmitter systems, so specificity to NMDA-R hypofunction is not guaranteed; the preclinical and human literatures lack sufficient chronic-administration studies to model long-term pathophysiology; and heterogeneous artefact rejection and analysis methods in EEG/MEG studies complicate comparisons. Notwithstanding these caveats, they note prior evidence implicating NMDA-R 2A subunits on parvalbumin interneurons as a plausible circuit mechanism linking NMDA-R antagonism to gamma upregulation, but they emphasise that further targeted research—especially contrasting acute versus chronic administration, task-related versus resting-state oscillations, and rigorous spectral characterisation—is required. Ultimately, the authors conclude that current data call into question a simple mapping from NMDA-R hypofunction to the resting-state gamma abnormalities reported in schizophrenia, with implications for E/I-balance models of the disorder.