Exploring aperiodic activity in first episode schizophrenia spectrum psychosis: A resting-state EEG analysis ☆

.


Introduction
Psychosis is characterised by impaired reality testing, reflecting an inability to distinguish between the external and internal world, as evidenced by hallucinations (perception in the absence of external stimulus) and/or delusions (false belief despite contradicting evidence; Arciniegas, 2015;Bhati, 2013).Psychosis is the defining symptom of a group of disorders, termed by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5; American Psychiatric Association, 2013) as Schizophrenia Spectrum Psychosis (SSP); however it is not the sole symptom (Bhati, 2013).Alongside positive, psychotic symptoms (i.e., hallucinations and delusions), SSP disorders are also characterised by disorganised speech, grossly disorganised or aberrant motor behaviour, and negative symptoms, which include functional/behavioural impairments such as reduced emotional expression (Bhati, 2013).SSP disorders are associated with a range of adverse outcomes including poor physical health, reduced life expectancy (Laursen et al., 2014), impaired social and occupational functioning (Rabinowitz et al., 2013), and high rates of hospital admission and unemployment (Morgan et al., 2014).Early intervention strategies are likely to be important for SSP disorders, as a shorter duration of untreated psychosis is predictive of improved long-term outcomes (Lambert et al., 2010).Hence, detecting changes in neural activity patterns early in the illness course may prove beneficial for establishing novel therapeutic strategies that target disrupted functional brain circuits (Hoy et al., 2021;Scangos et al., 2023;Siddiqi et al., 2023).In this context, the exploration of neural activity in first episode SSP (FESSP) using electroencephalography (EEG) is likely to be particularly valuable, as it provides an opportunity to assess for functional brain changes at very early stages of the illness.This has the potential to promote early identification of disrupted neural dynamics and promote timely treatment strategies (Perrottelli et al., 2021).In addition to allowing the exploration of early illness markers, FESSP patients are also often either medication-naïve or have received only minimal pharmacotherapy, making their results less susceptible to possible medication-related confounds.
In a healthy brain, levels of excitation and inhibition are tightly maintained under homeostatic control (Heinze et al., 2021;Le Roux et al., 2006).Although the precise ratios exhibit variability (Alvarez & Destexhe, 2004), higher levels of inhibition relative to excitation is widely established as characteristic of regulated excitation/inhibition (E/I) balance (Alvarez & Destexhe, 2004;den Boon et al., 2015;Le Roux et al., 2006).Maintaining E/I balance is crucial for optimal brain function, including the regulation of neuronal oscillations, information transmission (Foss-Feig et al., 2017), and the maintenance of neuronal homeostasis (Turrigiano & Nelson, 2004), which collectively contribute to efficient information processing within the brain (Foss-Feig et al., 2017).Past research has implicated disrupted E/I balance in SSP disorders.Both human (Merritt et al., 2016) and animal (Liu et al., 2021;Yizhar et al., 2011) research supports an elevated E/I ratio in schizophrenia, indicative of heightened excitation and/or reduced inhibition in this disorder (For recent reviews see: Howes & Shatalina, 2022;Liu et al., 2021).Further support for the elevated E/I ratio hypothesis of schizophrenia arises from studies examining γ-aminobutyric acid (GABA) and glutamate, the primary inhibitory and excitatory neurotransmitters in the brain, respectively (Foss-Feig et al., 2017;Gao & Penzes, 2015;Liu et al., 2021).For example, recent work by Howes and Shatalina (2022) found evidence of impaired GABA signalling in SSP patients, as demonstrated by lower levels of cortical GABAergic markers relative to healthy controls (HC).However, findings have been inconsistent, with other studies demonstrating unaltered GABA concentrations in SSP populations (Grent-'t-Jong et al., 2018).
More consistent research indicates that hypofunction of the glutamate receptor, N-methyl-D-asparate (NMDA) can produce elevated glutamate release, leading to enhanced excitatory tone (Adams & Moghaddam, 1998;Foss-Feig et al., 2017;Kehrer et al., 2008).Several studies support this notion, revealing increased glutamate levels in clinically high-risk and first-episode schizophrenia patients (Grent-'t-Jong et al., 2018;Merritt et al., 2016).In combination with GABA findings, these observations suggest elevated excitation and reduced inhibition in SSP disorders.Furthermore, studies administering ketamine, an NMDA antagonist, to neurotypical adults found that it can induce both positive and negative schizophrenia-like symptoms (Curic et al., 2019;Krystal et al., 1994).Together, these findings highlight E/I imbalance as a likely neural mechanism, and possible treatment target, for schizophrenia (Foss-Feig et al., 2017;Howes & Shatalina, 2022;Selten et al., 2018).Nevertheless, current findings need to be interpreted considering key limitations.Firstly, methods such as optogenetics, which are often implemented in E/I animal models (McNally et al., 2020;Yizhar et al., 2011), are invasive and not suitable for use in humans.Second, non-invasive neuroimaging modalities used to measure excitatory and inhibitory neuro-metabolites (i.e., glutamate and GABA concentrations) humans have relatively poor temporal resolution as a result of typically long acquisition times, thus potentially limiting their ability to infer 'real-time' changes in neural function (Harris et al., 2017).As such, studies using additional neurophysiological approaches might aid in assessing the elevated E/I ratio hypothesis to ascertain its usefulness as a potential neural marker for SSP disorders.
Analysing the aperiodic component of the neural signal recorded using EEG provides a novel approach to assessing putative E/I (im) balance in humans.EEG is a beneficial technique as it is non-invasive, widely accessible to researchers, and provides a direct measurement of neural activity with high temporal resolution and low cost (Cohen, 2017;Sun et al., 2011).The aperiodic signal, reflecting the broadband neural activity across all frequency bands, follows a 1/f-like power distribution marked by an exponential decrease in spectral power with increasing frequency (He, 2014;Ostlund et al., 2022).The aperiodic signal consists of two components: the exponent (or spectral slope), which represents the pattern of power across frequencies, and the offset, which represents the broadband spectrum shift (Donoghue et al., 2020;Ostlund et al., 2022).Animal and computational studies have indicated the aperiodic exponent to be a reliable measure of E/I balance (Gao et al., 2017;Wiest et al., 2023), with an elevated E/I ratio resulting in a 'flatter' slope (reduced exponent), while a reduced E/I ratio results in an increased exponent and 'steeper' slope (Donoghue et al., 2020;Gao et al., 2017;Ostlund et al., 2022).This has been further supported by human pharmacological intervention research demonstrating respective steepening/flattening of the aperiodic slope following either propofolor ketamine-induced anaesthesia, which produce net increases/decreases in neural inhibition, respectively (Waschke et al., 2021).
To date, few studies have investigated aperiodic activity in SSP disorders.However, a recent study by Molina et al. (2020) found that patients with schizophrenia displayed significantly larger exponent values (steeper aperiodic slope) relative to HCs when EEG was recorded during a passive auditory oddball paradigm.Additionally, Peterson et al. (2023) found steeper aperiodic slopes in patients with schizophrenia undergoing a visual target detection task.In contrast, however, Spencer et al. (2023) reported reduced exponents (i.e., flatter slope) in schizophrenia patients during auditory steady-state stimulation.
To our knowledge, no study has yet examined spontaneous aperiodic activity in in schizophrenia or SSP more broadly.As such, this study used a resting-state EEG paradigm to examine aperiodic neural activity in FESSP.Resting-state data are advantageous as they can be acquired quickly and require only minimal participant input making them conducive to the acquisition of neural data in participants who might be particularly unwell and/or unable to complete more cognitively demanding task-based paradigms.
In addition to examining the aperiodic exponent, the present study also explored the aperiodic offset.The precise neural underpinnings of this metric remain uncertain, however, it has been associated with neuronal spiking (Miller et al., 2012), which is known to be regulated, in part, by GABAergic neurons (Pouille & Scanziani, 2001).Given impaired GABA functioning in SSP patients, investigating the aperiodic offset in these patients could therefore offer complimentary insights into E/I balance and underlying GABA dysfunction.The aperiodic offset and exponent have also been shown to be strongly correlated (Hill et al., 2022;Merkin et al., 2023), highlighting a potential shared mechanism (E/I balance).
Here, we conduct an analysis of EEG data collected from individuals with FESSP and HCs to assess differences in the aperiodic signal between these two groups and explore its potential as a putative physiological marker of disrupted neural activity FESSP.We further examine associations between aperiodic activity and clinical symptom severity in FESSP using the Brief Psychiatric Rating Scale (BPRS), Scale for the Assessment of Negative Symptoms (SANS), and the Scale for the Assessment of Positive Symptoms (SAPS).We hypothesised that individuals with FESSP would show a reduced aperiodic exponent (i.e., flatter 1/f slope) compared to HCs, consistent with the elevated E/I ratio hypothesis of schizophrenia (Foss-Feig et al., 2017).We further predicted that individuals with FESSP would show alterations in aperiodic offset compared to HCs, without a predefined direction due to very limited prior research exploring the aperiodic offset.Finally, we anticipated that the EEG-derived exponent and offset would show associations with clinical symptoms, as measured by the BPRS, SANS, and SAPS.

Participants
This study utilised a pre-collected dataset, accessed via the Open-Neuro website (Dataset: ds003944; Salisbury et al., 2022).A detailed overview of the experimental methodology is provided in Phalen et al.

R.J. Earl et al.
(2020).The sample included 72 participants aged between 12 and 35 years.However, during data screening, one participant was removed due to excessive artifacts on the EEG record.Of the remaining 71 participants, 43 were FESSP and 28 were HC.All FESSP patients were diagnosed using the DSM, Fourth Edition (DSM-IV; American Psychiatric Association, 1994).Fifteen FESSP participants were not taking antipsychotic medications, 26 were taking at least one antipsychotic at the time of investigation (mean Chlorpromazine equivalent dose: 208.66 mg [SD = 123.88]),and two had no recorded medication data.Demographic details of the FESSP and HC groups are presented in Table 1.All participants provided informed consent prior to data collection.Ethical approval for the original data collection was provided by the University of Pittsburgh's International Review Board.Exemption from ethical review was also provided by the Deakin University Human Research Ethics Committee for secondary analysis of these data (2023-109).

EEG data acquisition
Resting-state eyes-open EEG data were recorded for five minutes using a Neuromag Vectorview system (Elekta, Helsinki, Finland), with a sampling rate of 1 KHz.Participants sat upright in a magnetically shielded room and were instructed to focus their gaze on a central fixation cross for the duration of the EEG acquisition.EEG data were collected using a low-impedance 10-10 system 60-channel cap.Magnetoencephalography (MEG) recordings were also acquired but are not reported here (see: Phalen et al., 2020).

EEG data pre-processing
Raw EEG files were pre-processed in MATLAB (version 9.10.0;The MathWorks Inc 2021) using the EEGLAB toolbox (Delorme & Makeig, 2004) and custom scripts.The data were cleaned using the open-source Reduction of Electroencephalographic Artifacts (RELAX) software, which uses an automated approach to remove artifacts from the EEG data while preserving the neural signal (Bailey et al., 2023a;Bailey et al., 2023b).As part of the RELAX pipeline, data were bandpass filtered (0.5-80 Hz; fourth-order Butterworth filter with zero-phase) and were also notch filtered (57-63 Hz) to remove line noise.Bad channels were removed using a multi-step process incorporating the 'findNoi-syChannels' function from the PREP pipeline (Bigdely-Shamlo et al., 2015).On average, 4.58 (SD = 3.52) electrodes were rejected for the FESSP group and 4.22 (SD = 3.34) for the HC group.Any removed electrodes were subsequently interpolated back into the data using EEGLAB's spherical interpolation function.Multi-channel Wiener filters (Somers et al., 2018) were used to initially clean blinks, muscle activity, horizontal eye movement and drift, followed by robust average rereferencing and wavelet-enhanced ICA (Castellanos & Makarov, 2006) with components for cleaning identified using IClabel (Pion-Tonachini et al., 2019).

Calculation of aperiodic exponent and offset
Power spectral density was calculated independently for each EEG electrode for each participant, using Welch's method in MATLAB.The FOOOF Python toolbox (version 1.0.0;Donoghue et al., 2020) was then used to parameterize the spectral data into the periodic and aperiodic components.The power spectra were parameterised across a frequency range between 3 and 50 Hz to obtain the broadband aperiodic exponent and offset.Spectral parameterisation settings for the algorithm were: peak width limits = [1, 8], maximum number of peaks = 6, peak threshold = 2, minimum peak height = 0.1.For each participant, aperiodic exponent and offset values were obtained across all electrodes and then averaged to produce a 'global' brain-wide value for the offset and exponent that were used in the subsequent statistical analysis (Hill et al., 2022).This approach was chosen as we had no specific a priori hypotheses regarding the spatial distribution of the aperiodic signal across the cortex; while past research has also indicated the aperiodic slope to be topographically widespread (Gyurkovics et al., 2022;Hill et al., 2022;Merkin et al., 2023).However, to further examine the spatial features of any potential differences in aperiodic activity, additional non-parametric cluster-based permutation analyses were also performed across all electrodes (further detailed in the Statistical Analysis section).The performance of the FOOOF algorithm was evaluated through the assessment of two 'goodness of fit' measures: R 2 and Error (taken as the average across all electrodes; Donoghue et al., 2020).R 2 represents the explained variance, while Error represents the total error of the model fit.Good model fit is indicated by high Mean R values, and low Mean Error values (Ostlund et al., 2022).Overall, good model fits were observed for the FOOOF algorithm in both the FESSP group (R 2 = 0.99, SD = 0.00; Error = 0.04, SD = 0.01) and the HC group (R 2 = 0.99, SD = 0.00; Error = 0.05, SD = 0.01), similar to those reported in previous studies (Hill et al., 2022;McSweeney et al., 2023;Merkin et al., 2023).

Assessment of clinical symptoms
As part of the original study protocol, FESSP participants were assessed using several clinical measures.Here, we focus on the Brief Psychiatric Rating Scale (BPRS), Scale for the Assessment of Negative symptoms (SANS), and Scale for the Assessment of Positive symptoms (SAPS).

BPRS
The 18-item BPRS (Overall & Gorham, 1988) purports to measure several aspects of SSP disorders, including positive symptoms (e.g., hallucinatory behaviour), negative symptoms (e.g., emotional withdrawal), and affective symptoms (e.g., anxiety).There are a total of domains measured by the BPRS, with each factor represented by a single item.Questions are scored during clinical interview on a 7-point Likert scale ranging from 1 (Not Present) to 7 (Extremely Severe).Values from each item can be summed to produce a score of symptom severity, with higher scores indicating greater severity.The BPRS has been shown to have good internal consistency (α = 0.85;He et al., 2021) and inter-rater reliability as measured by the intraclass correlation coefficient (ICC = 0.79; Schützwohl et al., 2003).

SANS
The SANS (Andreasen et al., 1984) consists of 25 items measured via clinical interview on a 6-point Likert scale from 0 (None) to 5 (Severe).The SANS measures five domains of negative symptomology: affective blunting, alogia, avolition/apathy, anhedonia/asociality, and attention.An overall score is calculated by summing the scores from each question, with a higher score indicating more severe negative symptoms.The SANS has good internal consistency (α = 0.89) and inter-rater reliability (ICC = 0.72; Peralta et al., 1995).

SAPS
The SAPS (Andreasen and King, 1984) consists of 34 items measured via clinical interview on a 6-point Likert scale from 0 (None) to 5 (Severe).The SAPS measures four domains of positive symptomology: hallucinations, delusions, bizarre behaviour, and formal thought disorder.An overall score is calculated by summing the scores from each question, with a higher score indicating more severe positive symptoms.

Statistical analysis
Statistical analyses were conducted using Jamovi (version 2.0.0;The Jamovi Project, 2023), with additional post-hoc Bayesian analyses performed in JASP (version 0.17.1;JASP team, 2023).Bayesian analyses utilised BF 01 , which indicates the Bayes factor in favour of the null hypothesis (H 0 ) over the alternative hypothesis (H 1 ) (Wagenmakers et al., 2018).BF 01 values from 1 to 3 indicate anecdotal evidence for the null hypothesis, values from 3 to 10 indicate moderate evidence, and values from 10 to 30 indicate strong evidence (Wagenmakers et al., 2018).
Independent samples t-tests were used to compare the aperiodic exponent and offset values between the FESSP and HC groups, with group allocation (FESSP, HC) entered as the grouping variable, and average brain-wide aperiodic activity (exponent, offset) entered as the dependent variable.Normality was assessed using the Shapiro-Wilk method, and the Levene's test was used to assess homogeneity of variance.False discovery rate (FDR) corrections were applied to adjust for the two separate comparisons (i.e., exponent and offset; Benjamini & Hochberg, 1995).In addition to assessment of the average global EEG signal across all electrodes, separate non-parametric cluster-based permutation analyses (Maris & Oostenveld, 2007) were also conducted in MATLAB (Version 9.10.0;The MathWorks Inc, 2021) to identify potential between-group differences in aperiodic signal taking into account all cortical regions (electrodes), while controlling for multiple comparisons.For these analyses, a cluster was defined as two or more neighbouring electrodes.The Monte Carlo method was used to calculate the significance probability using 5000 iterations, with alpha set at 0.05 (two-tailed t-test).
Linear regression models were used to assess the relationship between the brain-wide aperiodic offset and exponent values and psychiatric symptoms measured using the BPRS, SANS, and SAPS.For these tests, the BPRS, SANS, and SAPS were entered as dependent variables, and the offset and exponent were added in separate models as the predictor variables.Separate models for the offset and exponent were used due to previous studies reporting strong correlations between the offset and exponent (Hill et al., 2022;Merkin et al., 2023).FDR corrections were again applied to adjust for the six models.All non-significant findings across the t-tests and regression models were further explored through Bayesian analysis, to ascertain the strength of evidence in support of H 0 (Biel & Friedrich, 2018).As a final exploratory analysis, we also ran additional separate regression models for participants listed as either taking or not taking antipsychotic medication.

Descriptive Statistics
Clinical scores for the FESSP group are provided in Table 2. Overall, participants in the FESSP group displayed scores in the 'very mild' or 'questionable' ranges indicating low symptom severity.Furthermore, the maximum score among FESSP participants did not pass the 'moderate' category for any symptom scales.Participants in the FESSP group also displayed, on average, lower aperiodic exponent (M = 1.57,SD = 0.29) and offset (M = 0.73, SD = 0.47) values to participants in the HC group (exponent: M = 1.67,SD = 0.29; offset: M = 1.00,SD = 0.52).

Between-group differences in aperiodic activity
Results of the between-group comparisons for aperiodic exponent and offset are provided in Table 3.All independent samples t-test assumptions were met.Analyses revealed no difference in exponent values between FESSP and HC participants (p uncorrected = 0.139).A difference was observed between the two groups for offset, with a medium effect size, however this failed to remain significant following FDR correction (p uncorrected = 0.028).Two-sided Bayesian t-tests were further implemented to provide additional context given the non-significant findings.As shown in Table 3, these analyses revealed only anecdotal support for the null hypothesis.An overview of the aperiodic findings is provided in Fig. 1.
In addition to the t-tests comparing the global average exponent and offset values across all electrodes, a further cluster-based approach was also used to assess if there were any between-group differences across specific brain regions.These analyses revealed no significant betweengroup differences for the aperiodic exponent.However, there was a difference between the FESSP and HC groups for aperiodic offset (p = 0.012).Electrodes contributing to the negative cluster are visually represented in Fig. 2.This indicates lower offset values in the FESSP group compared to the HC group, and encompasses bilateral frontal, central, temporal, and parietal electrodes.

Association between aperiodic activity and clinical symptoms
As the aperiodic offset and exponent were highly correlated (r = 0.89, p < 0.001; Fig. 1D), separate regression analyses were run for each outcome measure to avoid potential confounds associated with multicollinearity.No violations of the normality of residuals (as assessed by quantile-quantile plots and Shapiro-Wilk test) were observed in the data.There was also no evidence of autocorrelation (Durbin-Watson value falling between one and three), and Cooks distance was < 1 indicating no outliers.Two FESSP participants did not have BPRS data and were therefore not included in the regression analyses, reducing the sample size to 41.As displayed in Table 4, all linear regression models revealed non-significant relationships between the brain wide aperiodic signal and the clinical measurement scales (scatterplots of the association between aperiodic activity and clinical scores is provided in the Supplementary Materials, Fig. S1).
Further Bayesian regression analyses were run to add context to the non-significant frequentist results, with an uninformed uniform prior [P (M)] of 0.5 set for each model.Results of the Bayesian analyses are presented in Table 5.Overall, the Bayes factors showed either anecdotal or moderate support for the null hypothesis.Finally, as the cluster-based Note.BPRS = brief psychiatric rating scale.SANS = scale for the assessment of negative symptoms.SAPS = scale for the assessment of positive symptoms.Note.FESSP = first episode schizophrenia spectrum psychosis.HC = healthy controls.Both models report between-group differences in the aperiodic exponent and aperiodic offset.The BF 01 factor indicates the likelihood of the null hypothesis being correct, with values from 1 to 3 indicating anecdotal evidence for the null hypothesis, 3-10 indicating moderate evidence, and 10-30 indicating strong evidence (Wagenmakers et al., 2018).
permutation analyses had identified a significant difference in aperiodic offset between the two groups, we also ran associations with clinical scores using the electrodes forming the significant cluster.These results are presented in the Supplementary Materials (Table S1).In alignment with the global analyses, none of these associations reached significance.
As a final exploratory analysis, we also ran the regression models separately for participants who were either taking antipsychotic medication or were medication naive.Results are presented in Table 6 and Fig. 3.The two participants with no recorded medication data were excluded from these analyses.There were no significant associations between aperiodic exponent and any of the clinical measures for either  the medicated or unmedicated participants.However, after FDR correction, aperiodic offset was able to significantly predict BPRS and SANS scores in the unmedicated sub-group.

Discussion
The current study aimed to assess if the aperiodic signal is altered in FESSP, and also if aperiodic activity is associated with clinical symptom severity.We predicted that FESSP participants would show widespread alterations in aperiodic exponent and offset compared to HCs.While there was no evidence of significant alterations in aperiodic exponent in FESSP, non-parametric cluster-based analyses revealed the aperiodic offset to be significantly reduced in this cohort, relative to HCs, encompassing electrodes spanning a relatively broad scalp distribution.Despite these findings, no associations between aperiodic activity and symptom severity were observed.However, when separating participants based on antipsychotic medication status, aperiodic offset was shown to predict BPRS and SANS scores in unmedicated participants.
In relation to the aperiodic exponent, it was hypothesised that FESSP participants would display reduced exponent values compared to the HC group.An independent samples t-test failed to uncover significant between group differences.Further cluster-based analysis supported this conclusion, with no regions (electrode clusters) showing significant between-group differences in aperiodic exponent.Previous research has highlighted the aperiodic exponent as an emerging putative marker of E/I balance in the brain (Donoghue et al., 2020;Gao et al., 2017;Muthukumaraswamy & Liley, 2018;Waschke et al., 2021).Specifically, greater exponent values, indicating a steeper 1/f-like broadband spectral slope, suggest higher contributions of inhibitory GABAergic to glutaminergic signalling indicative of greater inhibitory tone, while lower exponent values (flatter slopes) suggest enhanced excitatory tone (Gao et al., 2017;Gyurkovics et al., 2022;Waschke et al., 2021).Hence, within this framework, the present findings do not provide support for disrupted E/I balance in this cohort.
However, previous work by Molina et al. (2020) did report significant alterations in the aperiodic exponent in patients with schizophrenia compared to HCs when data were recorded during an oddball paradigm.More broadly, our present findings also appear to deviate from previous work using other modalities to asses E/I systems in schizophrenia, such as magnetic resonance spectroscopy (Marsman et al., 2014;Merritt et al., 2016), and high-frequency gamma-band oscillations (Ferrarelli et al., 2008;Shaw et al., 2020).However, to our knowledge, no study has yet examined aperiodic activity in FESSP specifically.It is therefore possible that our results might reflect a lack of sensitivity to detect subtle differences between FESSP and HC groups using the aperiodic exponent as a putative marker of E/I balance.While accumulating evidence supports the aperiodic exponent as a potential non-invasive measure of E/I balance (Colombo et al., 2019;Donoghue et al., 2020;Gao et al., 2017;Waschke et al., 2021), it's reliability as a marker of neural E/I ratio, and its sensitivity for differentiating clinical from neurotypical cohorts, requires additional research (Pani et al., 2022;Salvatore et al., 2024).Our additional Bayesian analysis further revealed only anecdotal evidence in favour of the null hypothesis, thus also largely supporting our frequentist results.
It is also possible that using a global brain-wide approach that averaged the aperiodic signal across all EEG electrodes might have limited the ability of the current study to find significant between-group differences in instances where activity might have been more localised (i.e., confined to discrete cortical regions).However, this seems unlikely given that our additional cluster-based analyses, which account for the spatial pattern of results across all electrodes, also supported these nonsignificant findings.The 'goodness of fit' measures (R 2 and Error values) obtained after spectral parameterisation in the current dataset also make it unlikely that the non-significant findings would be related to model fitting issues during spectral parameterisation.Finally, we also note that the FESSP participants in the current study presented with relatively low symptom scores across the BPRS, SAPS and SANS (i.e., falling within the 'questionable ' and 'very mild' categories).It is therefore possible that this might have restricted our ability to observe between-group differences with the HC group given the mild illness severity of the FESSP cohort.Overall, these findings emphasise the need for further replication studies in a separate FESSP cohort.
Results partially supported our prediction of a between-group  Note.BPRS = Brief Psychiatric Rating Scale.SANS = Scale for the Assessment of Negative Symptoms.SAPS = Scale for the Assessment of Positive Symptoms.The BF 01 factor indicates the likelihood of the null hypothesis being correct, with values from 1 to 3 indicating anecdotal evidence for the null hypothesis, 3-10 indicating moderate evidence and 10-30 indicating strong evidence (Wagenmakers et al., 2018).

Table 6
Association between the aperiodic signal and symptom severity in medicated and unmedicated FESSP patients.difference in aperiodic offset.Specifically, although there were no significant differences observed when comparing the global signal averaged across all electrodes, cluster-based analysis did reveal a relatively widespread cluster of electrodes that displayed significantly reduced aperiodic offset values in the FESSP group relative to HCs.Currently, very limited research has explored the aperiodic offset; hence, its precise functional significance remains limited (Ostlund et al., 2022;Pani et al., 2022).However, given the strong correlation between the aperiodic exponent and offset, as reported by both previous studies (Hill et al., 2022;Merkin et al., 2023) and the current study, it is possible that they are measuring similar underlying neural processes.Furthermore, aperiodic offset has been associated with broadband neuronal spiking (Manning et al., 2009;Miller, 2010;Miller et al., 2012) and might possibly represent, in part, a measure of underlying GABAergic activity, as this plays a vital role in controlling neuronal spike generation and timing (Pouille & Scanziani, 2001).Combined with our present results, this may indicate a degree of disruption to GABA-mediated inhibitory processes in FESSP, which would be in alignment with the broader literature (Chiu et al., 2018;Orhan et al., 2018).However, we caution that further studies are needed to explore these possible mechanisms in detail.Nevertheless, the present findings provide an important addition to the psychosis literature, as the aperiodic offset has received very limited attention.Our results indicate that the aperiodic offset might represent a potential neural marker in FESSP, which could be further investigated in future studies.We did not find any significant associations between aperiodic activity and clinical scores when regression models were run across all FESSP participants.While speculative, it is possible that this result may be attributable to the relatively low symptom severity experienced by most participants in the current study.Specifically, FESSP symptom scores as measured by the BPRS, SANS and SAPS were on average in the 'questionable' or 'very mild' range, with no scores surpassing the 'moderate' category.It is possible that the lower levels of symptom severity in our experimental group, experiencing only recent symptom onset, might have limited the ability to observe any associations with the neurophysiological data.Indeed, previous studies have identified that lower levels of GABA are associated with higher symptom severity in first episode psychosis patients (Orhan et al., 2018), suggesting that brain abnormalities and symptom severity might increase concurrently.Future longitudinal studies exploring possible links between the aperiodic signal and symptomology across a sustained illness course might help to illuminate the role of symptom severity and disease progression on measures of E/I (im)balance.Finally, while non-significant, negative associations were observed between the aperiodic offset and all clinical scores.This was most apparent for the SANS, which only lost significance after FDR corrections were applied, and while the present study had a moderate sample size, it is possible that it might still have been underpowered to detect any subtle brain-behaviour relationships.Future larger studies are therefore needed to further explore possible links between aperiodic offset and clinical symptoms.
However, our additional exploratory sub-group analyses, which separated participants based on whether or not they were being treated with an antipsychotic medication did reveal significant associations between aperiodic offset and BPRS and SANS scores for the unmedicated sub-group.Specifically, offset was able to predict BPRS and SANS scores, with lower offset values associated with higher BPRS and SANS scores in unmedicated participants.Although antipsychotic medications can affect the EEG record (Aiyer et al., 2016), the specific role of these medications in affecting EEG-derived aperiodic activity remains uncertain.Our present results suggest that accounting for antipsychotic medication status might be an important consideration when exploring associations between aperiodic activity and symptom severity in psychosis.
There were several limitations of the current study.First, we chose an exploratory, brain-wide approach to assessing the aperiodic signal, rather than attempting to select specific a priori regions of interest.We chose this approach due to several factors such as the widespread nature of the aperiodic signal (Gyurkovics et al., 2022;Hill et al., 2022;Merkin et al., 2023) and a lack of past aperiodic literature identifying potential regions of disrupted activity.Second, while any potential medication confounds would have likely been minimized through investigation of an FESSP participant group (i.e., compared to a cohort of patients with chronic illness), 60.5 % of participants in the current study were taking at least one antipsychotic medication, which may have influenced the results.Several studies have suggested that antipsychotic medication may affect EEG recordings (Aiyer et al., 2016;Jackson & Seneviratne, 2019;Yoshimura et al., 2007), however, to our knowledge the effects of these medications on the aperiodic signal remains to be systematically investigated.Third, the final sample size consisted of 71 participants, this moderate sample size might have meant that our study was underpowered to detect significant results.Future studies should aim to recruit a larger cohort, particularly a larger sample of FESSP participants, to better identify potential associations between the offset and exponent with symptom severity.Finally, the current study was limited by the fact that it did not include a chronic SSP experimental group in addition to the FESSP and HC groups.This prevented any comparisons between FESSP participants and individuals experiencing more chronic illness.
Despite these limitations, the current study represents an important initial step in exploring a novel neurophysiological measure (the aperiodic signal) in FESSP.Our results indicate that FESSP and HCs do not differ in terms of the aperiodic exponent but do reveal betweengroup differences in the aperiodic offset, a component of the neural signal that has received only limited attention in the EEG literature.We also found associations between aperiodic offset and symptom severity on the BPRS and SANS in unmedicated participants, specifically.Future research would benefit from further exploring aperiodic activity in larger schizophrenia spectrum samples, including longitudinal designs to track changes in neural activity across the illness course.Extension of this approach to other psychiatric and neurodevelopmental disorders linked to alterations in E/I balance such as autism and depression (Foss-Feig et al., 2017;Hu et al., 2023), would also be valuable.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Aperiodic activity in the FESSP and HC groups.A) The 1/f-like aperiodic signal (average across all electrodes) plotted in log-log space.Thick lines represent the mean signal for each group, while thin lines represent the signal for each individual participant.B) Boxplots with individual data overlaid of the aperiodic exponent (left) and offset (right) for the FESSP and HC groups, averaged across all electrodes.C) Topographic plots showing the aperiodic exponent and offset across all electrodes for the FESSP and HC groups.D) Association between aperiodic exponent and offset (average over all electrodes).FESSP = first episode schizophrenia spectrum psychosis.HC = healthy control.

Fig. 2 .
Fig. 2. Visual representation of cluster-based analysis results.A) Topographic map displaying the difference in aperiodic offset between the first episode schizophrenia spectrum psychosis (FESSP) and healthy control (HC) groups.White dots represent electrodes forming the significant negative cluster following the nonparametric cluster-based permutation analysis.B) Box plots showing the difference between the FESSP and HC groups as an average of all electrodes forming the significant cluster highlighted in (A).

Fig. 3 .
Fig. 3. Associations between the aperiodic exponent and offset and clinical scores run separately for medicated and unmedicated participants.BPRS = Brief Psychiatric Rating Scale.SAPS = Scale for the Assessment of Positive Symptoms.SANS = Scale for the Assessment of Negative Symptoms.

Table 2
Symptom scores within the FESSP group.

Table 3
Differences in aperiodic signal between the FESSP and HC groups.

Table 4
Association between the aperiodic signal and symptom severity in FESSP patients.Note.BPRS = Brief Psychiatric Rating Scale.SANS = Scale for the Assessment of Negative Symptoms.SAPS = Scale for the Assessment of Positive Symptoms.FESSP = First Episode Schizophrenia Spectrum Psychosis.

Table 5
Bayesian Regression Results.