1 Introduction

Glutathione (GSH), a key antioxidant, has been increasingly studied in relation to neurocognition and clinical severity in psychosis. Previous evidence suggests that alterations in GSH levels may contribute to cognitive deficits, symptom severity, and overall functional outcomes. This meta-analysis aims to synthesize available correlations between GSH measures—derived from both neuroimaging (MRS) and peripheral blood samples—and a range of cognitive domains as well as global clinical impressions in individuals with psychosis.

We focus on the following cognitive domains and related outcomes:

Imaging-Based Domains: - Ideational Fluency - Processing Speed - Verbal Memory - Functioning

Total GSH (GSHt) Domains: - Executive Function - Global Cognitive Score - Processing Speed - Cognitive Flexibility - Working Memory - Clinical Global Impression (CGI) - Functioning

Reduced GSH (GSHr) Domains: - Working Memory - Global Cognitive Score - Executive Function - Verbal Memory - Processing Speed - Clinical Global Impression (CGI)

By integrating data from multiple studies, we aim to identify which domains show consistent associations with GSH levels and to what extent these associations may inform our understanding of underlying oxidative stress mechanisms in psychosis.

2 Methods

2.1 Data Acquisition

Studies reporting correlation coefficients between GSH (imaging- or blood-derived) and cognitive or clinical outcomes in psychosis populations were included. Correlation coefficients (r) and sample sizes (n) were extracted. Fisher’s z-transformation was used to normalize correlation coefficients for meta-analysis.

2.2 Statistical Analysis

  • Random-effects meta-analyses were conducted using the metafor package in R.
  • Each domain was analyzed separately.
  • Pooled effects were back-transformed to r for interpretability.
  • Heterogeneity (I²) and p-values from Wald-type tests were reported.
  • Significance was set at p < 0.05.

2.3 Software

  • R (version ≥4.0)
  • tidyverse for data manipulation.
  • metafor for meta-analysis computations.

3 Load Libraries

library(tidyverse)
library(metafor)
library(kableExtra)

4 Functions

fisherz <- function(r) {
  0.5 * log((1+r)/(1-r))
}

fisherz2r <- function(z) {
  (exp(2*z)-1)/(exp(2*z)+1)
}

prep_data <- function(df) {
  df %>%
    mutate(zi = fisherz(r),
           vi = 1/(n-3))
}

run_meta <- function(df, domain_name) {
  df_prep <- prep_data(df)
  if(nrow(df_prep) < 2) {
    warning(paste("Only one study or no data for", domain_name))
    return(NULL)
  }
  res <- rma(yi = zi, vi = vi, data = df_prep, method = "REML")
  pred <- predict(res, transf = fisherz2r)
  I2 <- max(0, 100 * (res$tau2 / (res$tau2 + median(1/df_prep$vi))))

  tibble(
    Domain = domain_name,
    k = res$k,
    Pooled_r = round(pred$pred, 3),
    CI_LB = round(pred$ci.lb, 3),
    CI_UB = round(pred$ci.ub, 3),
    p_value = round(res$pval,4),
    I2 = paste0(round(I2,1), "%")
  )
}

5 Data Input

(Ensure the data below is accurate and uses periods instead of commas. The data comes from the previously determined tables.)

# Imaging-Based Domains
imaging_ideational <- data.frame(
  authors = c("Matsuzawa et al. 2008","Coughlin et al. 2021"),
  r = c(0.21,0.61),
  n = c(36,16)
)

imaging_processing <- data.frame(
  authors = c("Matsuzawa et al. 2008","Coughlin et al. 2021"),
  r = c(-0.14,0.26),
  n = c(36,16)
)

imaging_verbal <- data.frame(
  authors = c("Matsuzawa et al. 2008","Coughlin et al. 2021"),
  r = c(0.18,0.12),
  n = c(36,16)
)

imaging_functioning <- data.frame(
  authors = c("Lesh et al. 2021","Mackinley et al. 2022","Ravanfar et al. 2022"),
  r = c(0.185,0.04,0.452),
  n = c(33,53,12)
)

# GSHt (Total)
gsht_executive <- data.frame(
  authors = c("Martinez-Cengotitobengoa et al. 2012","Gonzalez-Liencres et al. 2014","Coughlin et al. 2021","Gares-Caballer et al. 2022"),
  r = c(0.072,0.171,0.45,0.40),
  n = c(28,41,24,30)
)

gsht_global <- data.frame(
  authors = c("Nucifora et al. 2017","Coughlin et al. 2021","Gares-Caballer et al. 2022"),
  r = c(0.245,0.57,0.34),
  n = c(51,24,30)
)

gsht_processing <- data.frame(
  authors = c("Gonzalez-Liencres et al. 2014","Coughlin et al. 2021","Gares-Caballer et al. 2022","Lin et al. 2023","Lin et al. 2023"),
  r = c(-0.119,0.41,0.21,-0.172,0.045),
  n = c(41,24,30,92,219)
)

gsht_flexibility <- data.frame(
  authors = c("Gonzalez-Liencres et al. 2014","Coughlin et al. 2021","Lin et al. 2023","Lin et al. 2023"),
  r = c(0.2,-0.06,-0.059,0.054),
  n = c(41,24,92,219)
)

gsht_workingmem <- data.frame(
  authors = c("Lin et al. 2023","Lin et al. 2023"),
  r = c(-0.094,-0.027),
  n = c(92,219)
)

gsht_cgi <- data.frame(
  authors = c("Raffa et al. 2009","Gares-Caballer et al. 2022"),
  r = c(-0.28,-0.08),
  n = c(88,30)
)

gsht_functioning <- data.frame(
  authors = c("Lin et al. 2023","Lin et al. 2023"),
  r = c(-0.107,0.076),
  n = c(92,219)
)

# GSHr (Reduced)
gshr_workingmem <- data.frame(
  authors = c("Cruz et al. 2021","Piatoikina et al. 2021"),
  r = c(-0.041,-0.003),
  n = c(85,125)
)

gshr_global <- data.frame(
  authors = c("Guidara et al. 2020","Cruz et al. 2021"),
  r = c(0.118,-0.092),
  n = c(66,85)
)

gshr_executive <- data.frame(
  authors = c("Cruz et al. 2021","Piatoikina et al. 2021"),
  r = c(-0.114,0.043),
  n = c(85,125)
)

gshr_verbal <- data.frame(
  authors = c("Cruz et al. 2021","Piatoikina et al. 2021"),
  r = c(0,0.02),
  n = c(85,125)
)

gshr_processing <- data.frame(
  authors = c("Cruz et al. 2021","Piatoikina et al. 2021"),
  r = c(0.038,0.03),
  n = c(85,125)
)

gshr_cgi <- data.frame(
  authors = c("Raffa et al. 2009","Ballesteros et al. 2013"),
  r = c(-0.32,0.208), # Assumed corrected correlation
  n = c(88,54)
)

6 Running All Meta-Analyses

domain_list <- list(
  "Imaging-Ideational Fluency" = imaging_ideational,
  "Imaging-Processing Speed" = imaging_processing,
  "Imaging-Verbal Memory" = imaging_verbal,
  "Imaging-Functioning" = imaging_functioning,

  "GSHt-Executive" = gsht_executive,
  "GSHt-Global Cognitive Score" = gsht_global,
  "GSHt-Processing Speed" = gsht_processing,
  "GSHt-Cognitive Flexibility" = gsht_flexibility,
  "GSHt-Working Memory" = gsht_workingmem,
  "GSHt-CGI" = gsht_cgi,
  "GSHt-Functioning" = gsht_functioning,

  "GSHr-Working Memory" = gshr_workingmem,
  "GSHr-Global Cognitive Score" = gshr_global,
  "GSHr-Executive Function" = gshr_executive,
  "GSHr-Verbal Memory" = gshr_verbal,
  "GSHr-Processing Speed" = gshr_processing,
  "GSHr-CGI" = gshr_cgi
)

results <- bind_rows(lapply(names(domain_list), function(dn) {
  df <- domain_list[[dn]]
  run_meta(df, dn)
}))

results %>%
  kable("html", caption="Meta-Analysis Results for All Cognitive Domains") %>%
  kable_styling(full_width = FALSE, bootstrap_options = c("striped","hover"))
Meta-Analysis Results for All Cognitive Domains
Domain k Pooled_r CI_LB CI_UB p_value I2
Imaging-Ideational Fluency 2 0.392 -0.063 0.712 0.0889 0.3%
Imaging-Processing Speed 2 0.005 -0.360 0.370 0.9783 0.1%
Imaging-Verbal Memory 2 0.163 -0.124 0.425 0.2642 0%
Imaging-Functioning 3 0.134 -0.073 0.330 0.2034 0%
GSHt-Executive 4 0.264 0.084 0.427 0.0044 0%
GSHt-Global Cognitive Score 3 0.355 0.157 0.526 0.0006 0%
GSHt-Processing Speed 5 0.032 -0.145 0.206 0.7267 0.1%
GSHt-Cognitive Flexibility 4 0.035 -0.067 0.137 0.5003 0%
GSHt-Working Memory 2 -0.047 -0.158 0.065 0.4154 0%
GSHt-CGI 2 -0.233 -0.399 -0.052 0.0119 0%
GSHt-Functioning 2 0.002 -0.172 0.177 0.9791 0%
GSHr-Working Memory 2 -0.018 -0.154 0.118 0.7940 0%
GSHr-Global Cognitive Score 2 0.004 -0.199 0.207 0.9664 0%
GSHr-Executive Function 2 -0.023 -0.174 0.129 0.7669 0%
GSHr-Verbal Memory 2 0.012 -0.125 0.148 0.8643 0%
GSHr-Processing Speed 2 0.033 -0.104 0.169 0.6351 0%
GSHr-CGI 2 -0.067 -0.536 0.433 0.8035 0.2%

7 Results

Selected Results:

  • GSHt-Executive (k=4): Pooled r=0.264, CI[0.084,0.427], p=0.0044
    Suggests a small-to-moderate positive association between total GSH and executive functioning.

  • GSHt-Global Cognitive Score (k=3): Pooled r=0.355, CI[0.157,0.526], p=0.0006
    Indicates a moderate positive correlation between total GSH and overall cognitive performance.

  • GSHt-CGI (k=2): Pooled r=-0.233, CI[-0.399,-0.052], p=0.0119
    Higher GSHt appears modestly related to lower illness severity as measured by CGI.

All other domains show non-significant results, with confidence intervals overlapping zero and p-values > 0.05. Heterogeneity (I²) is minimal across all analyses, indicating relatively consistent effects within each domain’s included studies.

8 Interpretation

The findings from this meta-analysis highlight a few key points:

  1. Significant Positive Correlations with Cognitive Outcomes (GSHt):
    Total GSH (GSHt) levels show a notable positive association with executive function and global cognitive performance. This suggests that individuals with higher GSHt may exhibit better cognitive functioning, potentially reflecting the role of oxidative balance in cognition for those with psychosis.

  2. Association with Clinical Severity (GSHt-CGI):
    The negative correlation with CGI indicates that higher GSHt might be related to slightly lower overall clinical severity. While the effect is modest, it points towards the possibility that improving systemic antioxidant levels could have some beneficial effects on clinical status.

  3. Non-Significant Relationships for Other Domains:
    Most domains, including imaging-based cognitive outcomes and measures linked to reduced GSH (GSHr), did not demonstrate statistically significant correlations. This could indicate that the relationship between GSH and cognition/clinical outcomes is more specific to certain GSH measures (e.g., total GSH) or certain cognitive domains (e.g., executive function, global cognition).

  4. Minimal Heterogeneity:
    The low I² values suggest little variability among studies within each domain’s analysis. This consistency increases confidence in the observed patterns, though the small number of studies in some domains reduces statistical power.

9 Conclusion

These results provide preliminary evidence that total GSH levels (GSHt) correlate positively with executive functioning and overall cognitive performance and are modestly associated with lower clinical severity (CGI) in psychosis. In contrast, no strong evidence emerged for other domains or for imaging-based GSH measures. Further research with larger sample sizes and standardized methodologies is warranted to clarify these relationships and potentially guide novel therapeutic strategies targeting oxidative stress in psychosis.