Table 1.

In vivo MRS of GSH for mental health: summary of findings

Condition	Brain region	MR‐Sq	Study and main finding	#Part.	d min	Other GSH‐specific correlations
BD	ACC	3 T PRESS	(Lagopoulos et al., 2013) No difference in GSH between patients and controls: (t(115) = 1.253, P = .213).	53 (BD) 51 (HC)	0.6	GSH vs YMRS: (ρ = −0.198; P = .214; N = 41) GSH vs HDRS: (ρ = 0.127; P = .385; N = 49) GSH vs age of onset: (ρ = −0.09; P = .522; N = 53)
						GSH vs duration (ρ = −0.125; P = .374; N = 53).
		3 T PRESS	(Chitty, Lagopoulos, et al., 2013) Risky drinking in patients showed less GSH than non‐risky: (t(48) = 2.44, P = .015).	33 (BD) 17 (HC)	0.9	High alcohol use disorders identification test score negatively correlated with GSH in BD subjects (r = −0.478, P = .005).
		3 T PRESS	(Chitty, Lagopoulos, Hickie, & Hermens, 2015a, 2015b) Change in alcohol use, smoking and age predict changes in GSH (at 15 months): (F(3, 26) = 3.69, P < .05). Not controlled as comorbid.	30 (BD)	0.5	GSH vs alcohol frequency: r = −0.381, P < .05 GSH vs smoking frequency: r = −0.367, P < .05
		3 T J‐PRESS	(Soeiro‐de‐Souza et al., 2016) No difference in GSH between euthymic BD and controls. Lac ∝ GSH in BD patients	50 (BD) 38 (HC)	0.6	Lac vs GSH Patients: (B = 0.20, t = 3.2, P = .003 [0.07, 0.33]) Controls:(B = 0.17, t = 0.64, P = .11 [0.04, 0.39])
	ACC + Hip	3 T PRESS	(Chitty, Lagopoulos, Hickie, & Hermens, 2014) No difference in GSH in either region between BD and Controls. Differences mediated by drinking and smoking.	64 (BD) 49 (HC)	0.5	GSHHip vs risky drinking (BD): (r = 0.489, P < .021) GSHACC vs smoking (BD): (t(53) = 4.162, P < .001)
	l‐Hip	3 T PRESS	(Chitty et al., 2015b) Controls: mismatch negativity correlated with GSH decrease in both temporal sites. r left = −0.542 [−0.8, −0.06] r right = −0.374 [−0.67, 0.07] No association in BD patients.	28 (BD) 22 (HC)	0.8	GSH vs left‐MMN (r = 0.068, P = .74, 95% [−0.36, 0.69])
						GSH vs right‐MMN (r = −0.057, P = .78, 95% [−0.52, 0.73]).
	OCC + mPFC	3 T SPECIAL	(Godlewska, Yip, Near, Goodwin, & Cowen, 2014) No difference between BP and control in either region (for GSH or other metabolites).	13 (BD)11 (HC)	1.2
	ACC + OCC	7 T STEAM	(Masaki et al., 2016) After treatment: No change in GSHOCC Decrease in GSHACC (P = .033) GSHACC plc. =1.31 土 0.043 GSHACC Ebselen = 1.17土0.07	20 (HC)	0.6
Schiz. (SZ)	ACC	4 T STEAM	(Terpstra et al., 2005) STEAM was within uncertainty of edited spectra in in vivo tests (P = .4). GSH levels of patients not different from controls (P = .4, differences >10%).	13 (SZ)9 (HC)	1.3	GSHpat = 1.6土0.2 GSHcont. = 1.5土0.3
		MEGA‐ PRESS
		7 T STEAM	(Brandt et al., 2016) GSH differences between patients and controls under the age of 40: [t(25) = −2.47, P = .021]	24 (SZ) 24 (HC)	0.8	GSH not correlated with age Overall no GSH difference between patients and controls.
	ACC + LI + VC	7 T STEAM	(Kumar et al., 2018) GSH lower in patients vs healthy controls‐ only in ACC voxel ACC P = .008 LI P = .784 VC P = .464	28 (SCH) 45 (HC)	0.7	GSH and glutamine correlated in all three voxels GSH vs ACC: r = 0.56 GSH vs LI: r = 0.80 GSH vs VC: r = 0.65
	mPFC	1.5 T PRESS	(Do, Trabesinger et al.) Cerebrospinal fluid GSH sample showed 27% decrease in patients (P < .05). MRS showed 52% decrease in patients (P < .0012).	14 (SZ) 14 (HC)	1.1
	pMFC	3 T MEGA‐ PRESS	(Matsuzawa et al., 2008) No significant difference between patients and controls.	20 (SZ) 16 (HC)	1.0	For patients GSH correlated to negative symptoms SANS and BPRS (r = −0.68, P < .001) and related to trail making test A (P < .05).
	Imag.	4 T proton echo planar spectroscopic imaging	(Bustillo et al., 2011) No GSH‐specific hypothesis tested.	30 (SZ) 28 (HC)	0.8
Major Depression (MD)	OCC, bilat.	3 T MEGA‐ PRESS	(Lapidus et al., 2014) GSH negatively correlated to anhedonia severity: (r = −0.55, P = .01).	11 (MD) 10 (HC)	1.3	MDD sample in isolation showed associations between anhedonia and GSH: (r = −0.53, P = .09). No associations between fatigue severity and GSH
	OCC	3 T SPECIAL	(Godlewska, Near, & Cowen, 2015) GSH was decreased in depressed patients F = 5.10, P = .028 F = 4.28, P = .042 (con. Age/sex)	39 (MD) 31 (HC)	0.7
		3 T PRESS	(Freed et al., 2017) GSH decreased in MD patients' vs HCs P = .04	19 (MD) 8 (HC)	1.3	No correlation between GSH and anhedonia, MD severity, or onset
	Imag.	3 T MRSI	(Li et al., 2016) In left putamen, GSH decreased in patients (P = .044) Patient increase post therapy not significant.	16 (MD) 10 (HC)	1.2	GSH/tCrpat. = 0.23士0.06 GSH/tCrcont. = 0.28士0.05
Early Psych. (FEP/EP)	Temp	3 T PRESS	(Berger et al., 2008) Bilateral GSH increase in treatment group response (F 1,12 = 6.1, P = .03) No longer significant when affective psychotic patients removed.	24 (FEP)	0.6	PANSS negative symptom change negatively correlated with GSH (r = −0.57, P = .041). Percent change in GSH and Glutamate/Glutamine correlated: (r = 0.64, P = .01)
		3 T PRESS	(Wood et al., 2009) GSH 22% higher in patients than controls: (F 1,46 = 4.7, P = .035). No difference in other tests: hemisphere (P = .137), group‐by‐hemisphere (P = .513).	30(FEP) 18(HC)	0.9	Patients not responding to topical niacin show 35% higher GSH than responders (F 1,28 = 5.1, P = .007).
	mPFC	3 T SPECIAL	(Monin et al., 2015) Potential dependence between GSH levels and white matter integrity during PFC developments.	30 (EP) 40 (HC)	0.7	Controls: GSH correlated to general FA (r = 0.34, P = .03) and functional connectivity (r = 0.40, P = .01). Patients controlled for medication and duration: GSH correlated to general FA (0.31, P = .01).
		3 T SPECIAL	(Xin et al., 2016) GSH decrease (P = .006) in glutamate‐cysteine ligase catalytic high‐risk (1.15 土 0.17) compared to low‐risk (1.34土0.8).	25 (EP) 33 (HC)	0.8	GSHmPFC correlated to GSHblood in controls (P = .021) but not in patients (P = .39).
CHR for psychosis	mPFC	3 T PRESS	(Hafizi et al., 2018) GSH and TSPO radioligand significant negative association in healthy volunteers, but not clinical high‐risk group—indicative of an abnormal interaction TSPO expression and redox status in CHR group	27 (CHR) 21 (HC)	0.9	mPFC GSH and [18F]FEPPA VT (radioligand of TSPO) not sig. Different between groups.
		3 T PRESS	(Da Silva, Hafizi et al. 2018a) No sig difference between mPFC GSH in drug‐naïve patients vs healthy controls	30 (CHR) 27 (HC)	0.8	No sig correlations between cerebral GSH and clinical and neuropsychological measures No sig difference between GPx activity and CHR vs HC (F = 0.15, P = .70)
						Significant effect lifetime cannabis use in GPx activity (F = 7.41, P = .01)
		3 T PRESS	(Da Silva et al., 2018) No sig difference between mPFC GSH in drug‐naïve patients vs healthy controls	27 (CHR) 16 (HC)	0.9	No differences between microglial activation and GSH between groups
	ACC + Striat.	3 T PRESS	(Demro et al., 2017) GSHACC was shown to correlate significantly with P3 (Grandiosity, negative) and P5 (Disorganized Communication, positive) as measured by the SIPS GSHSTR was shown to (negatively) correlate significantly only to Grandiosity (P3)	12 (CHR)	0.7	GSH correlation with SIPS: P1: r ACC = −0.578 (0.062), r STR = −0.566 (0.088) P2: r ACC = −0.074 (0.828), r STR = −0.474 (0.167) P3 r ACC = −0.673 (0.023), r STR = −0.775 (0.009) P4: r ACC = −0.259 (0.441), r STR = −0.409 (0.241) P5: r ACC = 0.645 (0.032), r STR = 0.138 (0.704) Positive symptom sum: r ACC = −0.134 (0.695), r ACC = −0.550 (0.099)
Age‐related (AD, Deprs‐ at.risk, sleep‐ apnea, and MCI)	Th	3 T PRESS	(Duffy et al., 2015) Older patients at risk of DEP taking placebo had larger GSH/Cr (t = 2.0, P = .049).	51 (DEP) (28 treat+ 23 plac.)	0.8	Increased GSHTh associated with worsening symptoms (r = 0.43, P = .043)
	ACC	3 T PRESS	(Duffy et al., 2015) Older patients at risk of DEP: increased GSH in the ACC (t = 2.7, P = .012)	58 (DEP) 12 (HC)	0.9	Depressed patients showed a correlation between HADS symptoms and GSH/Cr (r = 0.28, P = .035). Depressed patients showed a negative correlation between verbal learning and GSH/Cr (r = −0.28, P = .04)
		3 T PRESS	(Duffy et al., 2016) GSH/Cr correlated with decreased executive function.	24 (ARD)	0.6	GSHACC vs Oxygen desat: r = −0.54, P = .007 GSHACC vs apnea‐hypopnea: r = .42, P = .050 GSHACC vs response inhib: r = −.49, P = .015. GSHACC vs set shifting: r = −0.43, P = .37.
	ACC + PCC	3 T PRESS	(Duffy et al., 2014) MCI patients showed increased levels of GSH in the cingulate: GSHACC (t = −2.2, P = .03) GSHPCC (t = −2.9, P = .05)	54 (MCI) 41 (HC)	0.6	MCI GSHACC: 0.47 土 0.15 MCI GSHPCC: 0.37土 0.07
						Control GSHACC: 0.41土 0.10
						Control GSHPCC: 0.29土 0.05
	Vari.	3 T MEGA‐ PRESS	(Mandal, Tripathi, & Sugunan, 2012) Female AD showed decreased GSH in RFC (P = .003) compared to young controls. Male AD showed decreased GSH in the LFC	25 (ym) 20 (yf) 9 (om) 6 (of)	~1.3	GSHLFC different from GSHRFC in young female (P = .02) and male (P = .001) subjects. GSHLFC vs GSHRFC: young females (r = 0.641, P = .004)
			(P = .05) compared to young controls. Gender differences in GSH distribution evident.	5 (mMCI) 6 (fMCI) 7 (mAD) 7 (fAD)		GSHLPC vs GSHRpC: young females (r = 0.797, P = .000) GSHLFC vs GSHLPC: young males (r = 0.481, P = .032) (Healthy young males/females (ym/yf); healthy older males/females (om/of); males/females with mild cognitive impairment and Alzheimer's disease.)
Mult.	ACC	3 T PRESS	(Hermens, Lagopoulos, Naismith, Tobias‐Webb, & Hickie, 2012) Clustering of patients based on metabolites: 3 subgroups. GSH responsible for cluster 2.	37 (DD) 29 (BP) 22 (PD) 25 (HC)	N/A	Discriminant function 2 (40% variance) characterized by GSH/Cr (r = −0.753).
Suicidal behaviour	dPFC	3 T SPECIAL	(Jollant, Near, Turecki, & Richard‐Devantoy, 2016) No association was found between GSH and suicidal behaviour (F = 0.5, P = .6), but associations with other metabolites identified.	15 (SA) 10 (PC) 33 (HC)	~1.0	GSHdPFC Suicide Attempters: 0.24土0.03 GSHdPFC Patient Controls: 0.23 土 0.02 GSHdPFC Healthy Controls: 0.23土0.03
Chronic fatigue	OCC + Imag.	3 T MEGA‐ PRESS/ MRSI	(Shungu et al., 2012) GSH deficits in patients with CFS and MDD relative to controls: (F 2,40 = 15.93; P < .001)	15 (CFS) 15 (MD) 13 (HC)	1.0	GSH was inversely correlated with ventricular lactate (r = −0.545, P = .001) and a range of key indices of physical health.
Autism	Basal ganglia + dPFC	3 T PRESS	(Durieux et al., 2016) No GSH difference observed between patients and controls.	21 (ASD) 29 (HC)	0.8	Correlation between GSH and Autism spectrum disorder was observed in either region.
PTSD	ACC + dLPFC	3 T MEGA‐ PRESS	(Michels et al., 2014) Observed GSH levels 22.73% higher in patients than controls (F = 5.757, P = .025)	12 (PTS) 17 (HC)	1.1	GSHACC: PTSD = 0.15 ± 0.03, Non = 0.11 ± 0.03 (d = 1.33) GSHdLPFC: = 0.14 ± 0.03, Non = 0.11 ± 0.03 (d = 1.00)
Emerging Unipolar/Bi polar	ACC	3 T PRESS	(Naismith et al., 2014) GSH not associated with unipolar/bipolar differentiation (t = 1.15, P = .255).	53 (EBD)	0.4	GSH not associated with sleep midpoint (r = 0.211, P = .151)
Mood	ACC + HIPP	3 T PRESS	(Hermens et al., 2018) Positive correlations between Fractional Anisotropy in the stria terminalis and GSHACC were found across groups (r = 0.215, P < .01). DEP or BD in combination with decreased GSHACC was associated with reduced FA.	94 (DEP) 76 (BD) 59 (HC)	0.2	Decreased white matter integrity was associated with decreased GSHHIPP.
Anorexia Nervosa	ACC + OC + PUT	7 T STEAM	(Godlewska et al., 2017) GSH levels unchanged between groups.	13 (AN) 12 (HC)	1.2	AN (SEM) HC (SEM) p‐value GSHACC 1.19 (0.07) 1.27 (0.10) 0.38 CRLB 10.2 (2.9) 8.9 (2.8) GSHOCC 0.95 (0.03) 0.94 (0.04) 0.85 CRLB 10.5 (3.0) 10.67 (3.08) GSHPUT 1.51 (0.45) 1.10 (0.05) 0.43 CRLB 15.0 (4.5) 15.4 (4.6)
Prader‐Willi Synd.	ACC + P‐OCC	‐ MEGA‐ PRESS	(Rice, Lagopoulos, Brammer, & Einfeld, 2016) No between‐group differences were observed for GSH.	15(PWS) 15 (HC)	1.0
.Validation studies	mPFC	7 T PRESS	(Choi et al., 2010)			Optimized PRESS with sub‐TE pair showed improved selectability of coupled metabolites (eg, Glu, Gln, GSH).
	mPFC + rPFC	7 T J‐PRESS	(An et al., 2015)			TE‐optimized J‐PRESS was shown to minimize NAA signals while retaining GSH peak resolution.
	Hipp	3 T semi‐LASER	(Bednařík et al., 2015)			Using a short‐echo sequence with 5 minutes averaging the GSH CRLB was kept below 30% in a 4 mL voxel at 3 T.
	ACC	7 T PRESS	(Lally et al., 2016)			Intra Class Correlation (ICC) using TE‐optimized PRESS both within sessions (ICC > 0.7) and between sessions (ICC > 0.6) showed good repeatability. GSH negatively associated with age (r = −0.37, P < .05).
	ACC + PCC	3 T STEAM	(Wijtenburg et al., 2014)			Short‐TE phase rotation STEAM at 3 T showed excellent reproducibility for GSH: absolute reliability: SEM < 9.9%, relative reliability: ICCs 0.42‐0.51
	Midline parietal	3 T HERMES	(Saleh et al., 2016)			HERMES scanning protocol showed excellent separation of GABA and GSH. Results agree with MEGA‐PRESS, achieving similar signal‐to‐noise ratio in half the time.
	mPFC	7 T PRESS	(Choi et al., 2010)			Optimized PRESS sequence showed lower CRLBs of Gln and GSH than with STEAM.
		3 T SPECIAL	(Schubert, Kühn, Gallinat, Mekle, & Ittermann, 2017)			Short‐TE SPECIAL sequence was used to measure MRS spectra at 3 T in 21 healthy adults. GSH was detected with low uncertainty (CRLB < 30%) in only 16 cases.
		3 T J‐PRESS	(Jensen, Auerbach, Pisoni, & Pizzagalli, 2017)			Test–retest reliability of metabolite quantification was assessed in a 3 T shortened J‐resolved MRS sequence in healthy adolecents. GSH demonstrated satisfactory reliability with a score of 8.8–4.1%.

Abbreviations: ACC, anterior cingulate cortex; AD, Alzheimer's disease; AN, anorexia nervosa; ARD, age‐related disorder; ASD, autism spectrum disorder;BD, bipolar disorder; BP, bipolar disorder; BPRS, brief psychiatric rating scale; CAT, catalase; CFS, chronic fatigue syndrome; CRLB, cramer‐rao lower bound; CHR, clinical high risk; CRLB, Cramer Rao lower bound; DD, depressive disorder; DEP, depression; dLPFC, dorsal left prefrontal cortex; dPFC, dorsolateral prefrontal cortex; EBD, emerging bipolar disorder; EP, early psychosis; EPI, echo‐planar imaging; FA, fractional anisotropy; fAD, females with Alzheimer's disease; FEPPA, tracer; fMCI, females with mild cognitive impairment; FSL, FMRIB software library; GSH, glutathione; GCL, glutamate cysteine ligase; GABA, gamma‐aminobutyric acid; GSSG, glutathione disulphide; HADS, hospital anxiety and depression scale; HC, healthy control; HDRS, hamilton depression rating scale; HERMES, Hadamard encoding and reconstruction of mega‐edited spectroscopy; HIPP, hippocampus; ICC, inferior colliculus; J‐PRESS, J‐resolved spectroscopy; LFC, left frontal cortex; LI, left insular; mAD, males with Alzheimer's disease; MCI, mild cognitive impairment; MD, major depression; MDA, malondialdehyde; MDD, major depressive disorder; mMCI, males with mild cognitive impairment; MMN, mismatch negativity; MRS, magnetic resonance spectroscopy; MRSI, magnetic resonance spectroscopy imaging; mPFC, medial prefrontal cortex; NAA, n‐acetyl aspartate; OCC, occipital cortex; OCD, obsessive compulsive disorder; PANSS, positive and negative symptom scale; PC, patient controls; PD, psychotic disorder; PFC, prefrontal cortex; pMFC, posterior medial frontal cortex; PTS, post traumatic stress; PTSD, post traumatic stress disorder; PRESS, point‐resolved spectroscopy; PUT, putamen; PWS, prader‐willi syndrome; RFC, right frontal cortex; SA, suicide attempters; SANS, scale for assessment of negative symptoms; SOD, superoxide dismutase; SEM, standard error of the mean; SIPS, structured interview for psychosis‐risk syndromes; STEAM, stimulated echo acquisition model; SZ, schizophrenia; TBARS, thiobarbituric acid; TE, echo time; THC, tetrahydrocannabinol; TM, mixing time; TSPO, translocator protein; VC, visual cortex; VT, total distribution volume; YMRS, young mania rating scale.

Summary of findings, describing the clinical group, brain region of interest, magnetic resonance sequence used, study reference and main findings, number of control vs clinical participants, dmin (the minimum effect size that could be founds significant given the cohorts in each study), and other GSH‐specific correlations that are not directly related to the outcomes of the paper, but important nonetheless.