Levels of cytokines in the cerebrospinal fluid of patients with psychotic disorders compared to individually matched healthy controls

Background: Increased peripheral cytokine levels have been observed in patients with psychotic disorders; however, large high-quality studies with individually matched healthy controls have been lacking regarding cytokines in cerebrospinal fluid (CSF) of individuals with psychotic disorders. Methods: Patients diagnosed with a non-organic, non-affective psychotic disorder (ICD-10: F20/22 – 29) within a year prior to inclusion and individually age-and sex-matched healthy controls were included by identical in-and exclusion criteria ’ s except for the psychiatric diagnoses. All participants were aged 18 – 50 years and individuals with neurological or immunological disorders were excluded. CSF cytokines were analyzed with MesoScale V-PLEX neuroinflammation panel. Co-primary outcomes were CSF interleukin-6 (IL-6) and IL-8. Results: We included 104 patients and 104 healthy controls, matching on age, sex and BMI. No significant differences were found for the primary outcomes IL-6 (relative mean difference (MD): 0.97, 95 %CI: 0.84 – 1.11, p = 0.637) or IL-8 (MD: 1.01, 95 %CI: 0.93 – 1.09, p = 0.895). Secondary analyses found patients to have higher IL-4 (MD: 1.30, 95 %CI: 1.04 – 1.61, p = 0.018), a trend towards higher IFN-γ (MD: 1.26, 95 %CI: 0.99 – 1.59, p = 0.056), and lower IL-16 (MD: 0.83, 95 %CI: 0.74 – 0.94, p = 0.004) than healthy controls, though not significant after correction for multiple testing. IL-8 and IL-16 were found positively associated with CSF white blood cells and CSF/serum albumin ratio. The study was limited by 77.9 % of the patients being on antipsychotic treatment at time of intervention, and that levels of nine of the 26 cytokines were below lower limit of detection (LLOD) in > 50 % of samples; however, for the primary outcomes IL-6 and IL-8 more than 99.5 % of the samples were above LLOD and for IL-8 all samples exceeded the lower limit of quantification (LLOQ). Conclusions: We found no evidence of increased IL-6 and IL-8 in patients with recent-onset psychotic disorders in contrary to previous findings in meta -analyses of CSF cytokines. Secondary analyses found indication of higher IL-4, decreased IL-16, and borderline increased IFN-γ in patients, neither of which have previously been reported on in CSF analyses of individuals with psychotic disorders.


Introduction
The etiology of psychotic disorders is vastly unknown and substantial heterogeneity exists regarding symptoms, clinical course and prognosis, possibly due to different pathophysiological mechanisms.A hypothesis receiving increased scientific interest in the last decades, is that immunological alterations are present in some patients suffering from psychotic disorders.Prior meta-analyses have shown increased levels of several pro-inflammatory cytokines in blood from patients with firstepisode psychosis (FEP) or with acute relapse of psychoses, (Miller et al., 2011) which seem to normalize with antipsychotic treatment and stabilization of symptoms.(Çakici et al., 2021;Zhou et al., 2021) Due to the blood-brain barrier (BBB), peripheral levels of these cytokines only provide limited information on disorders of the brain, while investigation of the cerebrospinal fluid (CSF) offers greater insight into neuroinflammatory processes Fig. 1.
A recent meta-analysis found increased levels of both IL-8 and IL-6 in CSF from patients with psychotic disorders compared to healthy controls.(Orlovska-Waast et al., 2018) The most investigated cytokine in CSF in the context of psychotic disorders is IL-6, which has been reported on in 10 studies comprising a total of 257 patients, with the largest individual study including 61 men with schizophrenia.(van Kammen et al., 1999) The largest study on CSF levels of IL-8 to date comprises 94 patients with schizophrenia.(Hidese et al., 2021) Other significant findings of cytokines in the CSF include increased interferon-beta (INFβ) and decreased IL-12 in one study, (Hidese et al., 2021) and increased IL-1β in two studies.(Söderlund et al., 2009;Gallego et al., 2018) However, most prior studies investigating cytokines in CSF of patients with psychosis compared to healthy controls, were small with the majority comprising only around 10 to 20 subjects, (Söderlund et al., 2009;Gallego et al., 2018;Coughlin et al., 2016;Coughlin et al., 2013;el-Mallakh et al., 1993;Schwieler et al., 2015) and only one study on IL-6 and one on IL-8 included more than 50 patients.(van Kammen et al., 1999;Hidese et al., 2021) In addition, most prior studies lacked crucial information on limitations of cytokine measurements and how this was handled statistically.With many undetectably low levels of several cytokines in the CSF, proper handling of missing values is of high importance.
This report represents the largest study to date investigating differences between levels of different cytokines in the CSF of patients with psychotic disorders compared to individually matched healthy controls.Furthermore, it represents a high-quality study regarding inclusion of patients and individually matched controls, with focus on psychopathological evaluation of cases and controls, and thorough description of limitations of laboratory analyses, as well as statistical handling of censored measurements.

Methods
The study is approved by The Regional Committee on Health Research Ethics in the Capital Region of Denmark (j.no: H-16030985) and The Danish Data Protection Agency (j.no:RHP-2016-020, I-Suite no.: 04945).The study is part of the larger Psych-Flame biobank.Data on population characteristics, and data on routine CSF and blood measurements, have been published elsewhere (Jeppesen et al., 2022).The methods have previously been described in detail in our pre-published protocol (Jeppesen et al., 2021).

Participants
From June 2017 until September 2021, patients with psychotic disorders were recruited via psychiatric in-and outpatient clinics in the Capital Region of Denmark, and mentally and physically healthy controls matched on sex and age were recruited via a Danish website, forsøgsperson.dk.Patients and controls came from the same community.All participants were aged 18-50 years at time of inclusion.Inclusion criteria for patients was a diagnosis of a non-affective psychotic disorder (ICD-10: F20/F22-29) within a year prior to inclusion.The supplementary eMethods provides a detailed overview of exclusion criteria, including rationale.In brief, exclusion criteria were a prior diagnosis of a psychotic disorder, prior or current organic psychiatric disorder or severe neurological disorder, general medical conditions with known impact on the immune system, contraindications to lumbar puncture, regular use of anti-inflammatory medication, treatment with electroconvulsive therapy within three months prior to inclusion, current abuse of alcohol or drugs, and pregnancy.Healthy controls were additionally excluded when any prior or current psychiatric disorder could not be ruled out.All participants were thoroughly screened as described in our protocol.(Jeppesen et al., 2021) Informed consent was obtained for all participants prior to inclusion.

Clinical assessment
Patients and controls underwent an interview using WHO Schedules for Clinical Assessment in Neuropsychiatry (SCAN, version 2.1).The interview was used to validate diagnoses given to patients by clinical doctors prior to inclusion and to rule out prior or current mental disorders in the healthy controls.The Positive and Negative Symptom Scale (PANSS) was used to rate psychopathology for all participants.Functioning was rated for all participants using Global Assessment of Functioning (GAF), while cognition was assessed for all participants using the Brief Assessment of Cognition in Schizophrenia (BACS).For the first 24 patients, SCAN interview, rating of PANSS, and BACS were not performed, as these measures were added to the study protocol at a later point.
Data on time since diagnosis and medication status for patients were collected from electronic health records and validated by self-report.Height, weight, smoking status and alcohol intake was self-reported by participants.

Biological samples
Lumbar puncture was carried out according to current consensus guidelines for lumbar puncture (Engelborghs et al., 2017).All CSF samples were collected between 09:34 am and 01:10 pm, with 90 % in the time span of 09:58 am to 12:06 pm.Immediately following the lumbar puncture, samples were transported via taxi to the laboratory (Rigshospitalet, Copenhagen, Denmark, approximate transportation time; 20 min), where they were centrifuged, aliquoted and frozen.Mean time from initiation of sampling to final freezing of CSF was 75 min, with 90 % in the range of 61 to 95 min and no significant difference between patients and controls.Further details on the lumbar puncture procedure can be found in the supplementary eMethods.

Laboratory analyses
Mean storage time prior to analyses was 1.53 years (range 0.04-4.70),with a significant difference between patients and controls (mean 1.80 versus 1.26, p = 0.001), since patient inclusion started before recruitment of healthy controls.Samples were analyzed at Section for Biomarkers, Immunology and Antibodies, Center for Neonatal Screening, Statens Serum Institut, Artillerivej 5, DK-2300 Copenhagen, Denmark.All laboratory personnel were blinded to all clinical information, including patient/control status.Analyses of cytokines were performed using the V-PLEX Neuroinflammation Panel 1 (#K15210D, Meso Scale Diagnostics (MSD), Rockville, USA) on the Meso QuickPlex SQ 120 (MSD), according to the manufacturer's instructions (supplementary eMethods).The first aliquot collected after centrifugation was used for all analyses.To ensure consistent results, all samples were thawed once in order to be placed on the multiplex panels, before the panels were placed back in the freezer as soon hereafter as possible.CSF was analyzed undiluted.See Supplementary Table 1 for intra-and interassay coefficients of variation.

Statistical analyses
Demographic data were compared using two-sample t-tests for continuous outcomes, and χ 2 test for categorical ones.Outcomes measured as concentrations were log-transformed prior to analyses as they were not normally distributed.As we expected varying numbers of censored measurements for different cytokines, differences in primary and secondary outcomes between patients and healthy controls were analyzed using the censored Gaussian model, adjusted for age, sex, and plate.The censored Gaussian model (also known as the Tobit model) has been found to be superior in reducing bias in cases of left-censored data compared to substitution of censored values with e.g.lower limit of detection (LLOD) or LLOD/2 (Lynn, 2001).Previous publication has highlighted the importance of correctly establishing these values (Singh et al., 2023).Details on how LLOD was established at the laboratory at SSI can be found in the supplementary eMethods.Contrasts in cytokine levels between patients and controls were analyzed censoring values below LLOD (using the LLOD determined by the laboratory at SSI), and in a sensitivity analyses also censoring values below lower level of quantification (LLOQ).Results from both analyses are reported, and effects are reported as Relative Mean Difference (MD) between groups with 95 % confidence intervals (95 %CI).All secondary analyses were adjusted for multiple testing using the Bonferroni method.As sensitivity analyses, analyses of primary and significant outcomes including adjustment for blood contamination, time from collection to freezing and storage time were performed.See Supplementary Table 4 for information on LLOD and LLOQ.To perform pooled cytokine analyses, cytokine levels were standardized on log-scale using mean and SD from censored log-normal models.We then as a second step combined the standardized cytokine values in each category into a single variable.Contrast in levels between patients and healthy controls were then analyzed, adjusting for sex, age, and plate.To account for samples being differentially censored across cytokines, the second and third steps were performed jointly in a censored log-normal model using weighting giving equal weight to each cytokine in a category.Two-sided tests with p < 0.05 was considered significant.Statistical analyses were performed in R, (R Core Team, 2018) with the package "survival" for censored Gaussian model (Therneau, 2021).

Post hoc analyses
We pooled the cytokines into groups based on their function-pro-or anti-inflammatory, and Th1 or Th2 response related (Supplementary Table 3)-and investigated contrasts in these pooled groups between patients and control.
Subgroup analyses were performed for the primary outcomes with patients divided into groups with PANSS total </≥ 65, PANSS negative </≥ 18, GAF-S >/≤ 39, BACS z-score of >/≤ − 1.0, inpatients/outpatients, smokers/non-smokers, and treated/not treated with antipsychotic drugs.We decided post hoc to perform these subgroup analyses also on outcomes where contrasts between patients and controls were found.
We additionally post hoc decided to visually investigate whether there were apparent clinical distinctions in participants with "extreme" cytokine levels.This was done by focusing on all individuals with at least 5 cytokines with levels in the top 10th percentile of the entire cohort.
Possible associations between cytokine levels (IL-6, IL-8 and cytokines with indication of difference between patients and controls) and routine blood and CSF measurements (hs-CRP, white blood cell count (WBC) in blood and CSF and CSF:serum albumin ratio (Jeppesen et al., 2022) were estimated in censored log-normal models adjusted for sex, age, group and plate for all subjects, with both cytokines and routine markers log-transformed before analysis.

Results
Table 1 provides an overview of demographic data for participants consisting of the 104 patients with recent-onset psychotic disorders and the 104 individually matched healthy controls.Both groups included 64 males (61.5 %) and groups were comparable in age (26.1 vs 26.6 years, p = 0.634).Patients and controls did not differ significantly in BMI (p = 0.603).However, a larger proportion of patients smoked compared to controls (38.5 % vs 13.5 %, p < 0.001), whereas patients had a lower weekly intake of alcohol units (3.3 vs 5.3, p = 0.035).Patients had been diagnosed with their first psychotic disorder diagnosis on average 3.1 months prior to inclusion.At time of intervention, 36 patients (34.6 %) were diagnosed with schizophrenia (ICD-10: F20) and the remaining with other non-organic, non-affective psychotic disorders, 50.0 % (N = 52) were inpatients, 77.9 % (N = 81) were taking antipsychotic medication, and their mean total PANSS score was 62.5.Details on type of antipsychotics used can be found in Supplementary Table 5.
Median and range of observed cytokine levels can be found in Table 2 and Fig. 1. Results from primary and secondary analyses censoring values below LLOD are listed in Table 3 and illustrated in Supplementary Fig. 1.See Supplementary Table 6 for results from analyses censoring values below LLOQ.
Adjusting analyses for time from collection to freezing, storage time, blood contamination, or all of the above did not significantly change results (Supplementary Table 7A).

Secondary outcomes
We found significantly increased CSF levels in patients compared to controls of IL-4 when censoring values < LLOD (MD 1.30, 95 %CI 1.04-1.61,p = 0.018).We found a trend towards an increase of IFN-γ when censoring values below LLOD (MD 1.26, 95 %CI 0.99-1.59,p = 0.056).However, analyses of IL-4 and IFN-γ with censoring of values < LLOQ was not possible due to too many undetectable values.A decrease in CSF levels of IL-16 in patients compared to controls was observed, both in analyses censoring values < LLOD (MD 0.83, 95 %CI 0.74-0.94,p = 0.004) and < LLOQ, where measurements from 89.4 % of patients and 82.7 % of controls were censored.Adjusting analyses for time from collection to freezing, storage time, blood contamination, or all the above did not change results significantly (Supplementary Table 7B).After adjusting for multiple testing, no results remained significant.No significant differences were found between the two groups for the rest of the analyzed cytokines.
Nine of the 26 cytokines were undetectable in more than 50 % of  patients and controls.For 17 of 26 cytokines levels were < LLOQ in more than 90 % of both patients and controls (Fig. 2, supplementary Figs. 2 and 3).

Cytokines grouped by function
Pooling cytokines by function revealed no significant differences between patients and healthy controls in levels of pro-inflammatory, anti-inflammatory, Th1-related or Th2-related cytokines (all p-values above 0.80, Supplementary Table 8).

Subgroup analyses
Most of the analyses of effect heterogeneity between subgroups on IL-6, IL-8, IL-4, IL-16, or IFN-γ did not yield significant results, including analyses of medicated versus non-medicated participants.For some cytokines and subgroups, a significant difference was found between the respective patients and controls, with all results being in line with the overall findings.Supplementary Table 9A-F provides an overview.CSF IL-6 displayed effect heterogeneity between subgroups for BACS and GAF scores (p = 0.014 and p = 0.024, respectively) among patients, but  the subgroups were not significantly different from healthy controls.Moreover, when grouping healthy controls using the same BACS cutoff, there were no longer evidence for effect heterogeneity (p = 0.847).We found significant effect heterogeneity of IL-4, with patients with lower cognitive scores showing greater effect (MD of 2.83 (95 %CI: 1.32-6.08)vs MD of 1.27 (95 %CI: 0.97-1.66),p = 0.034) in analyses where healthy controls were subdivided based on their BACS scores.The exploratory investigation of individual participants with very high cytokine levels can be found in Supplementary Table 10.No distinctions from the total cohort were clearly apparent in individuals with more than 5 cytokines in the top 10th percentile.

Association analysis between CSF cytokines and routine CSF and blood markers
Analyses of associations between routine markers (previously published (Jeppesen et al., 2022) and relevant cytokines (IL-6, IL-8, IL-4, IL-16 and IFN-γ) is presented in Supplementary Table 11 and Supplementary Figs.4A-B.Significant positive associations were found between IL and 8 and both CSF WBC and CSF:serum albumin ratio, as well as IL-16 and CSF WBC and CSF:serum albumin ratio when including both patients and controls.When including only patients, aforementioned associations remained significant, but only before adjusting for multiple testing.

Discussion
This study is the largest to date on CSF levels of cytokines in patients with psychotic disorders compared to individually matched healthy controls.Our patient group represents a broad spectrum of psychotic disorders, without neurological symptoms or any signs of inflammatory disorders.We found no significant contrast in neither CSF IL-6 nor IL-8 in this cohort with identical in-and exclusion criteria among the participants except for the psychosis diagnosis.Secondary analyses found higher levels of IL-4 in patients, as well as a trend towards an increase in IFN-γ, whereas levels of IL-16 were found to be lower in patients as compared to controls, though none of the results remained significant after correction for multiple testing.Significant associations between IL-8 and IL-16 and both CSF WBC and CSF/serum albumin ratio were found.
CSF levels of IL-6 have previously been found elevated in patients with psychotic disorders, as recently summarized in a meta-analysis (Orlovska-Waast et al., 2018) including seven of the ten studies published to date.Of the remaining three studies not included herein, (Gallego et al., 2018;Coughlin et al., 2016;Kegel et al., 2017) only one (Coughlin et al., 2016) found significant differences between cases and controls of IL-6 in the CSF.Likewise, CSF levels of IL-8 were found to be increased in the recent meta-analysis, (Orlovska-Waast et al., 2018) which included three of the six studies that have reported hereon to date.Two of the remaining three studies (Hidese et al., 2021;Gallego et al., 2018) agreeingly found IL-8 to be significantly increased in cases compared to controls.When we in contrast hereto find no significant differences, there are many possible explanations.The comprehensiveness of our intervention may result in the most severely ill patients either not being able to participate at all, or at least not being able to participate during their most acute phase of illness.With most changes in levels of cytokines being fleeting and short-lasting, possible increases appearing in acute phases of psychotic disorders might not be detectable even days or weeks hereafter.It is therefore possible that a shorter timeframe from illness debut or a criterium of a minimum score on PANSS, as opposed to our strategy of a broader screening of patients, would have resulted in more detectable cytokine alterations among patients with psychosis, though subgroup analyses did not indicate differences in cytokine levels between higher and lower psychopathological scores.Previous meta-analyses have found several CSF cytokines to differ between patients and controls with small to medium effect sizes; however, levels of cytokines in the CSF are overall low both in cases and controls.Though the multiplex panel used herein has high sensitivity compared to others, current available analyses methods have their limitations regarding sensitivity for CSF measurements, which results in a high proportion of unquantifiable cytokine levels.Many previous studies on CSF cytokines in patients with psychotic disorders have either not reported on their statistical handling hereof or have handled it with simpler methods such as substitution of censored values with LLOD/2, which has been found to increase bias (Lynn, 2001).Thus, when thoroughly establishing LLOD values as per suggested standards to ensure that the quality of the multiplex assay is not overestimated (Singh et al., 2023), and when more appropriately handling these undetectable values as left-censored, the possibility of incorrectly finding small differences between cases and controls is less likely.Lastly, most of the previous studies with positive findings have included populations that were markedly older and with longer duration of disease than the one in our study.(Hidese et al., 2021 Feb;Gallego et al., 2018;Sasayama et al., 2013;Schwieler et al., 2015) Of note, proportions of patients treated with antipsychotic vary among previous studies, but are generally in line with that of our population.
Our findings of increased IL-4, a trend towards increased IFN-γ, and decreased IL-16 levels in the CSF of patients with psychotic disorders are novel.Notably though, none of the three were significantly different after correction for multiple testing.To the best of our knowledge, no other studies to date have reported on the levels of these markers in CSF; some studies have included IL-4 and/or IFN-γ as analytes but reported only that they were not sufficiently detected, (Hidese et al., 2021;Gallego et al., 2018;Coughlin et al., 2016;Schwieler et al., 2015;Soderlund et al., 2009) and no studies have yet investigated CSF IL-16 levels in patients with psychotic disorders.IL-16 is a cytokine that modulates Tcell activation and has been implicated to play a role in multiple autoimmune diseases (Glass et al., 2006).Investigations of peripheral levels of IL-16 in patients with psychotic disorders are sparse, but one study did find increased levels as well as a positive correlation with negative symptoms (Yan et al., 2022).IL-4, a cytokine involved in Th2 responses (Berger, 2000;Müller et al., 1999); has been found to be significantly decreased in blood from patients in one meta-analysis including patients with FEP (Goldsmith et al., 2016), while another including only drugnaïve FEP found that IL-4 correlated positively with negative symptoms (Dunleavy et al., 2022).IFN-γ mainly have pro-inflammatory properties (Turner et al., 2014); and is, as opposed to IL-4, involved in Th1 responses (Müller et al., 1999).Peripheral levels of IFN-γ has been found to be increased in patients across both treatment status and stage of illness (Miller et al., 2011;Goldsmith et al., 2016;Dunleavy et al., 2022) and have been found to decrease with antipsychotic treatment (Marcinowicz et al., 2021).
The finding of positive association between IL-8 with CSF WBC and CSF/serum albumin ratio indicates that increased cytokine levels are associated with increased neuroinflammation and dysfunction of the blood-brain barrier, which is in line with the role of IL-8 as a mainly proinflammatory cytokine.This is of interest, as previous analyses of this population found increased blood-brain barrier permeability in patients as compared to controls (Jeppesen et al., 2022).However, the fact that the same positive association was found for IL-16 which is not commonly association with pro-inflammatory actions and which was found to be decreased in patients as compared to controls, warrants caution in the interpretation of the impact of the associations.
Our findings in subgroup analyses should also be interpreted with caution; no significant difference was found regarding IL-6 in patients grouped by cognitive scores as compared to controls, and no effect heterogeneity was observed when also subdividing controls based on BACS scores.Subdividing patients and controls based on clinical variables resulted in small sample sizes, as for example regarding IL-4, where only one measurement above LLOD was available for the group of healthy controls with lower BACS scores.

Strengths and limitations
This study is the largest study to date on cytokines in the CSF of patients with psychotic disorders compared to healthy controls.The methodology of our study strengthens our findings, as our study is a high-quality study both when rating it using the Newcastle-Ottawa Scale (Wells et al., 2000) or the newer BIOCROSS evaluation tool for crosssectional biomarker studies (Wirsching et al., 2018).We have published a thorough study protocol prior to analyses, and with thorough clinical and psychopathological evaluation of both cases and controls, including validation of diagnoses given by referring doctors and the exclusion of psychiatric disorders in our controls, our case/control definition is of high quality.This also allowed analyses of a broad spectrum of clinical variables, such as cognition and level of functioning.In recruitment of both cases and controls, we included strict exclusion criteria, thus limiting other aspects influencing the immune system and strengthening comparability of participants on other factors than psychiatric status.Patients and controls were recruited from the same community, and patients were recruited both from in-and outpatient clinics to ensure representativeness.Regarding biological sampling, we narrowed the time frame for sampling of CSF to minimize effect of diurnal variation, all laboratory personnel were blinded to case/control status, and we used a multiplex panel validated for use with CSF.Additionally, we thoroughly accounted for our methodology regarding establishment of LLOD and handling of samples with levels below, as per previous recommendations (Lynn, 2001;Singh et al., 2023).
An important limitation of our study is the limitations embedded in the sensitivity of multiplex cytokine measurements.Although the panel from MesoScale is one of the most sensitive multiplex assays currently available, the concentration of many cytokines in CSF are very low.Thus, a large proportion of analytes were detectable in less than 50 % of samples, and even more measurements fell below LLOQ.This impacts power and statistical analyses.However, as opposed to most previously published studies, we report our results in a way that highlights this and makes the limitations transparent.Additionally, it should be noted that participants were not required to fast prior to the lumbar puncture procedure, nor were they required to remain physically inactive, and both food intake and physical exercise might possibly impact levels of some of the measured cytokines.

Conclusions and perspectives
We found no significant difference in CSF IL-6 or IL-8.Prior to correction for multiple testing, we did find increased IL-4 and decreased IL-16, as well as a tendency towards increased IFN-γ levels in patients, which are all novel findings, as CSF levels of these three cytokines have never been reported on in patients with psychotic disorders.Additional studies on these cytokines are needed to evaluate our findings.

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.Relative mean difference of levels of selected cytokines in patients versus healthy controls Analyses of contrasts between patients and controls, with values below lower limit of detection (determined by SSI) censored, adjusted for age, sex, and plate.Error bars represent 95 % confidence intervals.IL: interleukin, IFN; interferon.* p-value < 0.05, ~ p-value = 0.056.

Table 1
Study population characteristics of the 104 patients with recent-onset psychotic disorders compared with 104 individually matched healthy controls.
HC: healthy controls, n: number of subjects with data.a Pearson's Chi-squared test.b Welch two-sample t-test.c HC used as reference for z-score.PANSS: Positive and Negatives Symptom Scale, BACS: Brief Assessment of Cognition in Schizophrenia, GAF-S: Global Assessment of Functioning (Symptoms).

Table 2
Median and range of observed cytokine data.
a Geometric mean and SD.b Number of participants with detectable values.

Table 3
Relative mean difference between patients and healthy controls for all cytokines, censoring values below LLOD.
All analyses are adjusted for age, sex, and plate.HC: Healthy controls, MD: Relative mean difference, LLOD: Lower limit of detection, IFN: interferon, IL: interleukin, IP: interferon gamma-induced protein, MCP: monocyte chemoattractant protein, MDC: macrophage-derived chemokine, MIP: macrophage inflammatory protein, TARC: thymus-and activation-regulated chemokine, TNF: tumor necrosis factor.aseeTable2fornumbers on patients and controls separately.* not significant after correction for multiple testing.