GABAergic dysfunction in Parkinson’s disease: insights from in vivo proton magnetic resonance spectroscopy

Abstract

Parkinson’s disease (PD) pathogenesis extends beyond dopamine with evidence of serotonergic, noradrenergic, cholinergic and involvement of γ-aminobutyric acid (GABA). This article aims to systematically review the existing literature on GABAergic alterations in PD as quantified by in vivo proton MRS with a focus on ascertaining the influence of GABA on motor and non-motor symptoms, motor subtypes, effect of medication status and methodological variations across studies. A systematic search of PubMed and Scopus was carried out in April 2025 with a relevant Boolean phrase and 22 studies met the inclusion criteria. Considerable methodological variability was observed across studies in magnetic field strengths, pulse sequences, voxel location and size, quantification reference and processing methods. Review of reported GABA levels revealed a heterogeneous pattern of alterations across studies, regions of interest and clinical phenotypes, precluding meta-analysis. These observations suggest a multifaceted, region-specific dysregulation of GABAergic neurotransmission extending beyond the canonical dopaminergic framework. GABA plays a definite modulatory role in the pathogenesis of PD, and complex, distinct regional alterations contribute to the development of motor and non-motor symptoms in PD. Standardisation of GABA spectroscopy methods and appropriate patient stratification is crucial to identify definite patterns of alterations.

Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder characterised by bradykinesia, rigidity, rest tremor and postural instability¹ as well as non-motor symptoms (NMS) which can commonly include sleep disturbances, olfactory dysfunction, mood disturbances (depression, anxiety), cognitive deficits, autonomic dysfunction, and pain^2,3. Dopaminergic loss in the substantia nigra pars compacta (SNpc) and widespread accumulation of α-synuclein is considered pathognomonic in PD⁴, with dopamine replacement therapy being the therapeutic mainstay. However, this dopaminergic-centric view has been challenged by accumulating evidence highlighting the role of other neurotransmitter system^5,6. Primary evidence of this arises from the fact that unlike bradykinesia and rigidity, tremor does not correlate with striatal dopaminergic depletion⁵; rather serotonergic^6,7, noradrenergic⁸ abnormalities, and pallidal rather than striatal dopaminergic depletion^9,10 are suggested. The role of other neurotransmitter systems is further exemplified by NMS in PD, for instance serotonergic dysfunction in depressive symptoms¹¹, noradrenergic alterations which contribute to autonomic dysfunction¹², and cholinergic deficits in cognitive impairment¹³.

γ-aminobutyric acid (GABA) is the most abundant inhibitory neurotransmitter which intricately regulates excitatory-inhibitory balance across the basal ganglia thalamocortical (BGTC) and cerebello thalamocortical (CTC) networks¹⁴. Dopaminergic neurons from the SNpc exert modulatory effects in the BGTC, via direct and indirect pathways using GABA from the inhibitory projection neurons^15,16, with disruptions contributing to motor and non-motor in PD¹⁷. Non-dopaminergic systems interact with GABAergic and dopaminergic pathways to create a complex excitatory-inhibitory balance that is likely to form the basis for the diverse symptom profile observed in PD¹⁸. Understanding how GABAergic alterations arise from and contribute to the pathophysiology of PD is essential to explain the full spectrum of symptomatology to develop targeted therapeutic strategies. Attempts have been made to indirectly evaluate synaptic GABAergic activity in PD using transcranial magnetic stimulation (TMS), particularly measures of short-interval intracortical inhibition (SICI), long-interval intracortical inhibition (LICI) and cortical silent period (CSP)¹⁹. Studies using TMS have reported reduced LICI (mediated by GABA_B), and altered SICI-LICI interactions in PD, supporting the concept of GABAergic dysfunction in PD^20,21, though findings related to SICI have been inconsistent across reports.

Magnetic resonance spectroscopy (MRS) is a non-invasive imaging modality that permits in vivo quantification of brain metabolite concentrations. Although proton MRS (¹H MRS) is widely applied in research settings and for selected clinical applications such as brain tumour characterisation and metabolic disorders performed, GABA spectroscopy is challenging owing to its low concentrations and spectral overlaps with more abundant metabolites such as creatine²². To mitigate this, spectral editing techniques need to be used. For instance, MEscher-GArwood (MEGA) editing employs frequency-selective editing pulses to isolate the GABA signal at 3 ppm and can be combined with different localisation sequences such as Point RESolved Spectroscopy (PRESS)²³, Localized Adiabatic SElective Refocusing (LASER)²⁴ or semi-LASER. Alternatively, short echo-time non-edited STimulated Echo Acquisition Mode (STEAM) sequence at ultra-high field (>3 T) can be used. It is important to note that MEGA-PRESS editing, at standard echo time (TE = 68 ms), results in co-editing of macromolecules (MM) at 3 ppm alongside GABA. This combined signal is referred to as GABA + ²⁵. Macromolecule-suppressed GABA editing, which uses modified pulse parameters to minimise macromolecule contributions, can be used as an alternative, though it remains less common in PD literature. These distinctions are important for interpreting spectroscopy findings and will be noted throughout this review. Several studies have attempted to quantify GABA in PD and ascertain correlations between symptoms, drug responsiveness and neurochemical profiles. However, significant variability across studies in patient selection, magnetic field strength, MRS localization sequence, voxel location, voxel size and quantification method, makes a comprehensive understanding of GABAergic dysfunction in PD difficult.

The current study aims to systematically review the existing literature on GABAergic alterations in PD as quantified by ¹H MRS with a focus on ascertaining the influence of GABA on motor and non-motor symptoms, motor subtypes, effect of medication status and methodological variations across studies.

Results

The initial search returned 114 studies (Fig. 1). After removal of 48 duplicates and preliminary screening, 66 studies were eligible for abstract screening. Forty-four studies were excluded based on pre-specified exclusion criteria: animal studies (n = 17), other movement disorders (not PD) (n = 12), review articles (n = 8), non-MRS modalities such as NMR-saliva studies (n = 2), MRSI (multi-voxel, not single-voxel) (n = 1), methods paper on motion correction (n = 1), PD study where GABA was not reported (n = 1), and miscellaneous (editorial, unrelated)(n = 2). Finally, a total of 22 articles were included in this systematic review^26–47.

Fig. 1. PRISMA flow diagram illustrating the systematic search and selection process.

Records were identified from PubMed (n = 50) and Scopus (n = 64). After removal of 48 duplicate records, 66 abstracts were screened. Forty-four were excluded based on pre-specified criteria (animal studies, other movement disorders, review articles, non-single-voxel MRS, and other). Twenty-two full-text articles were assessed for eligibility, all of which met inclusion criteria, yielding a final sample of 22 studies. Created in BioRender. Prasad, S. (2026) https://BioRender.com/y5qlbap.

Demographic and clinical characteristics

All except one study⁴³ included patients with PD and healthy controls (HC), with the sample sizes ranging from 10 to 60 subjects per group, and most subjects were between the ages of 50 to 70 (Table 1). With the exception of Pesch et al.³⁴, which recruited only male participants in all groups, all other studies had a relatively equal sex distribution.

Table 1.

Demographic and clinical details of subjects in studies evaluating GABA in patients with Parkinson’s disease

Authors, year	Sample size	Sex (M:F)	Age	Duration of illness	MDS-UPDRS-III/ UPDRS-III (OFF)	MDS-UPDRS-III/ UPDRS-III (ON)	LEDD	Medication state while scanning
Oz et al.26	PD: 10; HC: 11	PD: 4:6; HC: 4:7	PD: 59 ± 10; HC: 59 ± 8	2.3 ± 1.5	NR	NR	NR	OFF
Emir et al.27	PD: 13; HC:12	PD: 6:7; HC: 7:5	PD: 56 ± 10; HC: 54 ± 8	7.5 ± 14	30.6 ± 9.2	NR	Pons: 380.4 ± 298.9 (n = 11); Putamen: 342.6 ± 298.2 (n = 11); SN: 350.0 ± 395.2 (n = 5)	OFF
Dharmadhikari et al.28	PD: 19; HC: 18	PD: 10:9; HC: 11:7	PD: 63.38 ± 9.12; HC: 59.63 ± 10.24	0.75-11* No mean, SD, median	33.34 ± 10.90	NR	NR	OFF (3, drug naïve)
O’Gormon Tuura et al.29	PD:20; HC: 17	PD: 16:4; HC: 13:4	PD: 63 ± 6.8; HC: 62 ± 12.8	9.25 ± 4.16* calculated from table	NR	27.95 ± 8.18*	1058.75 ± 360.66*	OFF
Elmaki et al.30	PD:21; HC:15	PD: 13:8; HC: 11:4	PD: 61.24 ± 9.31; HC: 57.13 ± 7.15	3.5 ± 2.7	31.2 ± 7.9	NR	NR	OFF
Gong et al.31	TDPD: 9; PIGD: 13; HC: 16	TDPD: 3:6; PIGD: 5:8; HC: 7:9	TDPD: 58.8 ± 7.7; PIGD: 59.8 ± 8.4; HC: 66 ± 9.9	TDPD: 3.8 ± 3.4; PIGD: 3.3 ± 1.7	TDPD: 32.1 ± 9.9; PIGD: 30.2 ± 7.7	NR	NR	OFF
O’Gormon Tuura et al.32	PD: 16; HC:16	PD 13:3; HC: 12:4	PD 65; HC: 62 *No SD	9.68 ± 3.94	NR	Pre-DBS:27.62 ± 8.15 (n = 16); Post-DBS: 16.92 ± 6.46 (n = 14)	Pre-DBS: 1053.12 ± 371 (n = 16); Post-DBS: 354.28 ± 307.07 (n = 14)	OFF
Firbank et al.33	PD + VH: 19; PD-VH: 17; HC: 20	PD + VH: 13:4; PD-VH: 13:4; HC: 14:6	PD + VH: 75.5 ± 4.5; PD-VH: 72.3 ± 5.1; HC: 75.4 ± 5	PD + VH: 11 ± 7.4; PD-VH: 9.6 ± 6.5	PD + VH: 34.7 ± 18.8; PD-VH: 55.9 ± 19.3	NR	PD + VH: 717.3 ± 421.7; PD-VH: 673.5 ± 428.2* ?LEDD	ON
Pesch et al.34	PD: 35 -AR: 19; -MPD: 16; HC: 35	PD: 35:0 -AR: 19:0; -MPD: 16:0; HC: 35:0	PD: 59 -AR: 60; -MPD: 59; HC: 55 *Median	PD: 4.7(2.5-7.7 -AR: 6(3.6-8.8); -MPD: 3.5(1.5-5.4)) *Median, IQR	PD: 34(23-43) -AR: 34(24-43); -MPD: 34.5(25.5-45.5)) *Median, IQR	NR	NR	ON
Piras et al.35	PD: 20; HC: 20	PD: 10:10; HC: 12:8	PD: 58.8 ± 9.6; HC: 54.25 ± 16.62	3.51 ± 1.78	NR	12.30 ± 6.01 *?ON	335.0 ± 260.99	NR ?ON
van Nuland et al.36	Dopa res tremor: 17; Dopa resp tremor: 23; No tremor: 20; HC: 22	Dopa res tremor: 13:4; Dopa resp tremor: 9:14; No tremor: 11:9; HC: 12:10	Dopa res tremor: 60.9 ± 9.6; Dopa resp tremor: 61.3 ± 12.8; No tremor: 60.2 ± 9.2; HC: 63.1 ± 10.5	Dopa res tremor: 2.6 ± 1.8; Dopa resp tremor: 5.7 ± 5.4; No tremor: 5.1 ± 2.1	NR *sub-scores reported: axial, brady + rig, rest tremor, tremor, brady+rig	NR	Dopa res tremor: 381 ± 329; Dopa resp tremor: 518 ± 289; No tremor: 645 ± 503	OFF and ON (200/50)
Delli Pizzi et al.37	PD + SSD: 23; PD-SSD: 19; SSD: 14; HC: 19	PD + SSD: 14:9; PD-SSD: 14:5; SSD: 7:7; HC: 11:8	PD + SSD: 66.3 ± 7.9; PD-SSD: 66.7 ± 7; SSD: 64.8 ± 7.3; HC: 64.8 ± 10.3	PD + SSD: 4.6 ± 2.1; PD-SSD: 3.5 ± 2.3	NR	PD + SSD: 14.3 ± 6.1; PD-SSD: 13.8 ± 6.0 *? ON state	NR	OFF
Parkin et al.38	PD:17; Young HC:20; Older HC: 18	PD: 8:9; Young HC: 10:10; Older HC: 8:10	PD:61.3 ± 7.2: Young HC: 16.7 ± 5; Older HC: 61.3 ± 6.8	5 ± 3.1	30.47 ± 12.10	NR	475.48 ± 351.57	OFF
Song et al.39	PD: 11; HC: 11	PD: 6:5; HC: 7:4	PD: 60.7 ± 8.7; HC: 53.7 ± 16.1	3.4 ± 1.8	30.1 ± 7.9	15.5 ± 7.9	NR	OFF and ON
Song et al.40	PD: 18; HC: 18	PD: 8:10; HC: 11:7	PD: 60.7 ± 7.6; HC: 56.2 ± 11	2.9 ± 1.5	32.1 ± 7.6	NR	NR	OFF
Seger et al.41	PD: 19; HC: 11	PD: 14:5; HC: 9:4	PD: 64.9 ± 8.7; HC: 68.9 ± 8	5.7 ± 4.2	36.9 ± 16.4	25.5 ± 11	535 ± 448	ON
Delli Pizzi et al.42	PD + SSD: 18; PD-SSD: 18; SSD: 13; HC: 17	PD + SSD: 10:8; PD-SSD: 14:4; SSD: 7:6; HC: 9:8	PD + SSD: 65.9 ± 8.4; PD-SSD: 66.7 ± 7.2; SSD: 63.8 ± 9.1; HC: 64.6 ± 10.9	PD + SSD: 4.7 ± 2.1; PD-SSD: 3.7 ± 2.5	PD + SSD: 13.6 ± 6.3; PD-SSD: 13.8 ± 6.2	NR	NR	OFF
Trujillo et al.43	PD + ICD: 19; PD-ICD: 14	PD + ICD: 10:9; PD-ICD: 9:5	PD + ICD: 61.6 ± 6.5; PD-ICD: 67.2 ± 7.4	PD + ICD: 4.3 ± 3.3; PD-ICD: 7.2 ± 4.2	PD + ICD: 29.5 ± 14.2; PD-ICD: 30.7 ± 10.6	PD + ICD: 20.0 ± 9.9; PD-ICD: 27.8 ± 10.6	PD + ICD: 642.7 ± 366.9; PD-ICD: 792.2 ± 436.1	OFF and ON (only DA)
Shukla et al.44	PD: 38; HC:30	PD: 24:14; HC: 20:10	PD: 69.07 ± 7.08; HC: 66.16 ± 7.79	NR	NR	NR	NR	NR
Liu et al.45	PD + D: 19; PD-D: 19; HC: 24	PD + D: 7:12; PD-D: 31:33; HC: 12:12	PD + D: 62.8 ± 8.7; PD-D: 64.2 ± 7.9; HC: 60.8 ± 8.4	PD + D: 8.9 ± 5.5; PD-D: 7.5 ± 4.8	PD + D: 30.6 ± 9.8; PD-D: 35.1 ± 16.1	PD + D: 18.5 ± 7.4; PD-D: 17.9 ± 10.1	NR	ON, OFF and ON (n = 15)
Tian et al.46	EOPD: 10; LOPD: 40; HC: 52	EOPD: 7:3; LOPD: 10:30; Young HC: 8:4; Old HC: 10:30	EOPD: 44.7 ± 5.5; LOPD: 66.3 ± 6.1; Young HC: 45.7 ± 3.1; Old HC: 65.6 ± 6.6	NR	NR	NR	NR	OFF
Yan et al.47	PD: 60; HC: 47	PD: 24:36; HC: 18:29	PD: 59.43 ± 7.37; HC: 57.11 ± 6.85	3 (1-4) *Median and IQR	32 (18-45) *	NR	475 (350-700)*	OFF

AR Akinetic rigid Parkinson’s disease, brady: bradykinesia, DA Dopamine agonist, DBS Deep brain stimulation, Dopa res Dopamine resistant, Dopa resp Dopamine responsive, EOPD Early onset Parkinson’s disease, F Female, HC Healthy Controls, IQR Interquartile range, LEDD Levodopa Equivalent Daily Dose, LOPD Late onset Parkinson’s disease, M Male, MDS-UPDRS Movement Disorders Society-Unified Parkinson’s Disease Rating Scale, MPD Mixed variant of Parkinson’s disease, NR Not reported, PD Parkinson’s disease, PD-ICD Parkinson’s disease without impulse control disorder, PD-SSD Parkinson’s disease without somatic symptom disorder, PD-VH Parkinson’s disease without visual hallucinations, PD + ICD Parkinson’s disease with impulse control disorder, PD + SSD Parkinson’s disease with somatic symptom disorder, PD + VH Parkinson’s disease with visual hallucinations, PIGD Postural Instability and gait disturbance variant of PD, rig rigidity, SD Standard deviation, SSD Somatic symptom disorders, TDPD Tremor dominant Parkinson’s disease, UPDRS Unified Parkinson’s Disease Rating Scale.

Duration of illness ranged from early PD ( < 1 year) to advanced PD ( > 10 years). Disease severity was reported by 19 studies, using with the UPDRS-III(n = 12)^{27,30,31,33,35,37–40,42,48} and MDS-UPDRS III(n = 7)^{29,32,34,36,41,43,47} (total UPDRS-III values not reported by van Nuland et al.³⁶). Of this only OFF state scores were reported in nine studies^{27,28,31,33,34,38,40,42,47}, ON state scores by four studies^29,32,35,37, and both OFF and ON scores in four studies (n = 4)^39,41,43,45. While three of these performed the ON state assessment with levodopa/carbidopa, the ON state assessment by Trujillo et al.⁴³ was with only dopamine agonists (DA). The mean of OFF state scores ranged from 13.6 to 55.9^33,37, and ON state scores from 13.8 to 27.95^29,37. Direct comparison of these ranges is not meaningful as they derive from different patient cohorts. Levodopa equivalent daily dose (LEDD) was reported by ten studies.

Nine studies included subgroups based on motor subtypes^31,34, additional NMS^{33,37,42,43,45}, altered levodopa responsiveness to tremor³⁶ or surgical intervention-deep brain stimulation (DBS)³². Gong et al.³¹, included tremor dominant PD (TDPD), and the postural instability and gait disturbance (PIGD) variant, while Pesch et al.³⁴ included akinetic-rigid and the mixed subtypes. PD with somatic symptom disorder (SSD) was evaluated in two studies^37,42, PD with impulse control disorder (ICD)⁴³ in one and PD with depression in one⁴⁵. Dopamine resistant tremor was evaluated by van Nuland et al.³⁶, and patients pre and post DBS were evaluated by O’Gormon Tuura et al.³².

Acquisition

Magnetic field strengths included - 3 Tesla (T) (n = 19), 4 T (n = 1)²⁶, and 7 T (n = 2)^27,41 (Table 2). Majority used a standard 32 channel receive coil or 8 channel radiofrequency (RF) coil, while 16 channel and dual tuned ¹H/³¹P transmit and receive coils were used in single studies each. An early study by Oz et al., used a transverse electromagnetic (TEM) volume coil²⁶. MEGA-PRESS was the most common sequence used to measure GABA, followed by STEAM in 3 studies^26,27,41. All MEGA-PRESS studies used a TE of 68 ms, except two with TE=69ms^29,32, and one with TE=120ms⁴⁴. The study using TE = 120 ms⁴⁴ and those employing STEAM^26,27,41 used non-standard parameters relative to the conventional MEGA-PRESS GABA+ experiment, and this is relevant to the interpretation of their GABA measures. Acquisition transients ranged from 96-320 for MEGA-PRESS.

Table 2.

Details of hardware, acquisition details of studies evaluating GABA in patients with Parkinson’s disease

Authors, year	Field strength	Manufacturer, Model	RF coil	Pulse sequence	Editing pulses (ppm)	Duration	Bandwidth	VOI location	VOI size	TR/TE	Number of excitations/ acquisitions	Water suppression method	Quality metrics
Oz et al.26	4 T	Oxford Magnet Technology with INOVA console (Varian)	TEM volume coil (T)(R)	STEAM	NA	NA	NA	Bilateral SN	2.2 mL	4500/5 ms, TM:42 ms	NEX = 400	VAPOR	LW(PD/HC): 13.5 ± 2.2/13.3 ± 2.1; SNR(PD/HC): 6 ± 1.2/6.5 ± 0.8; CRLB: GABA < 40
Emir et al.27	7 T	Magnex with Siemens console	16 channel (T)(R)	STEAM	NA	NA	NA	Pons; Putamen; SN	Pons: 30x10x15mm³; Putamen: 12x8x18mm³; SN: 6x13x13mm³	5000/8 ms; TM: 32 ms	Pons, putamen:128; SN: 384	VAPOR	LW (PD/HC)- Pons: 13.2 ± 2.3/12.9 ± 3; Putamen: 19.7 ± 5.9/20.5 ± 7.1; SN: 24.6 ± 3.9/20.8 ± 4.9 SNR (PD/HC)- Pons: 15.1 ± 1.9/16.2 ± 1.7; Putamen: 11.0 ± 2.8/10.5 ± 2.5; SN: 5.8 ± 1.6/7.5 ± 1.2
Dharmadhikari et al.28	3 T	Siemens Magnetom Tim Trio	NR	MEGA-PRESS	1.9/7.5	NR	44 Hz	Right thalamus; Right striatum	25x30x25mm3	2000/68 ms	256	NR	LW(PD/HC)- Thalamus: 21.35 ± 4.1/18.14 ± 3.2 Striatum: 23.91 ± 2.8/ 24.09 ± 2.2; SNR (PD/HC)- Thalamus:18.64 ± 4.7/23.06 ± 5.4) Striatum: 17.91 ± 2.8/ 18.71 ± 4.1; CRLB(PD/HC)- Thalamus:12.44 ± 2.1/12.87 ± 1.8 Striatum: 10.81 ± 1.9/12.14 ± 3.2
O’Gormon Tuura et al.29	3 T	GE MR750	8 channel (R)	MEGA-PRESS	1.9/7.5	16 ms	NR	Left basal ganglia; Left PFC	30 mL	1800/69 ms	320	NR	NR
Elmaki et al.30	3 T	Philips Achieva TX	8 channel phased-array (T)(R)	MEGA-PRESS	NR/NR	NR	NR	Left basal ganglia	30x30x 20mm3	2000/68 ms	320	NR	GABA+ fit error (PD/HC): 5.73 ± 2.18%/5.51 ± 1.20%
Gong et al.31	3 T	Philips Achieva TX	8 channel phased-array (R)	MEGA-PRESS	1.89/NR	14 ms	88 Hz	Left basal ganglia	30x30x 20mm3	2000/68 ms	256	CHESS	NR
O’Gormon Tuura et al.32	3 T	GE MR750	8 channel (R)	MEGA-PRESS	1.9/7.5	16 ms	NR	Left basal ganglia	30 mL	1800/69 ms	320	NR	NR
Firbank et al.33	3 T	Philips Achieva	8 channel (R)	MEGA-PRESS	1.9/7.5	NR	NR	Occipital lobe (midline)	45x32x 20mm3	2000/68 ms	320	VAPOR	NR
Pesch et al.34	3 T	Philips Achieva X series	32 channel (R)	MEGA-PRESS	NR/NR	NR	NR	Thalamus; Striatum (head of caudate, putamen, part of GPi)	30x30x 25mm3	2000/68 ms	256	NR	NR
Piras et al.35	3 T	Philips Achieva	32 channel (R)	MEGA-PRESS	1.9/7.5	14 ms	NR	Left and right cerebellar hemisphere	30x30x 30mm3	2000/68 ms	256	NR	NR
van Nuland et al.36	3 T	Siemens Prisma	32 channel (R)	MEGA-PRESS	1.9/NR	17.2 ms	44 Hz	Thalamus, MC, VC *contralateral to side with more prominent motor symptoms	18x24x 18mm3	1500/68 ms	96-Thalamus, 128-MC	CHESS	FWHM (PD/HC)- Thalamus: 9.2 ± 4.4/9.6 ± 4.5; MC: 7.3 ± 2.1/7.4 ± 2.8; VC:7.4 ± 3.0/7.7 ± 3.2 CRLB (PD/HC)- Thalamus: 24.1 ± 13.9/22.6 ± 13.4; MC: 15.3 ± 5.5/14.5 ± 5.8; VC:17.9 ± 10.0/ 16.8 ± 11
Delli Pizzi et al.37	3 T	Philips Achieva	8 channel phased-array (R)	MEGA-PRESS	1.9/7.46	14 ms	NR	mPFC	20x30x30mm3	2000/68 ms	320	CHESS	CRLB < 20
Parkin et al.38	3 T	Siemens Magnetom Verio	32 channel (R)	MEGA-PRESS	NR/NR	NR	NR	Mid-DLPFC	20x20x20mm3	2000/68 ms	128	VAPOR	FWHM: 8.9-17.9
Song et al.39	3 T	Philips Achieva TX	8 channel phased-array (R)	MEGA-PRESS	1.9/7.46	14 ms	NR	Upper brainstem	10x25x20mm3	2000/68 ms	320	MOIST	NR
Song et al.40	3 T	Philips Achieva TX	8 channel phased-array (R)	MEGA-PRESS	1.9/7.46	14 ms	NR	Upper brainstem	10x25x30mm3 (7.5 mL)	2000/68 ms	320	MOIST	Values NR. Based on Figure 3 in manuscript FWHM approx. PD:9, HC:8
Seger et al.41	7 T	Siemens Terra	NR	STEAM	NA/NA	NA	NA	Left putamen	14x32x17mm3	8000/4 ms; TM:28 ms	72	NR	LW (PD/HC): 0.06 ± 0.01ppm/0.06 ± 0.01ppm; SNR (PD/HC): 28.4 ± 6.6; 32.9 ± 9; CRLB (PD/HC): 10.8 ± 2.2/11.5 ± 2.5
Delli Pizzi et al.42	3 T	Philips Achieva	NR	MEGA-PRESS	NR/NR	NR	NR	mPFC	20x30x 30mm3	2000/68 ms	320	NR	NR; SNR: Compared to “Big GABA”
Trujillo et al.43	3 T	Philips Achieva	32 channel (R)	MEGA-PRESS	1.9/8	14 ms	140 Hz	Right thalamus, right MC	Thalamus: 30x22x 28mm3; MC: 40x25x 25mm3	3000/68 ms	320	NR	NR
Shukla et al.44	3 T	Philips Achieva	Dual tuned 1H/31P (T)(R)	MEGA-PRESS	NR/NR	NR	NR	SN, Left hippocampus (GSH)	30x35x 25mm3	2500/120 ms	320	CHESS	FWHM, values NR
Liu et al.45	3 T	Philips Ingenia Elition	32 channel (R)	MEGA-PRESS	1.89/7.46	14 ms	NR	Left thalamus, Left MFC	30x22x 28mm3	2000/68 ms	288	VAPOR	FWHM (PD/HC)- Thalamus: 10.93 ± 1.20/11.24 ± 1.34; MC: 10.53 ± 1.66/10.57 ± 2.13 SNR (GABA-PD/HC)- MFC: 17.04 ± 2.42/ 18.86 ± 1.63; MC: 16.44 ± 2.78/ 15.93 ± 0.85
Tian et al.46	3 T	Philips Ingenia Elition	32 channel (R)	MEGA-PRESS	1.9/7.5	NR	100 Hz	Left sensori-motor cortex	30x30x 30mm3	2000/68 ms	320	VAPOR	FWHM < 25; CRLB < 15
Yan et al.57	3 T	Philips Elition X	32 channel (R)	MEGA-PRESS	1.89/7.46	NR	NR	Midbrain	15x25x 30mm3	2000/68 ms	128	VAPOR	FWHM (PD/HC)- 9.19 ± 1.26/ 9.39 ± 0.91

¹H Hydrogen, ³¹P Phosphorous, CHESS Chemical shift selective saturation, DLPFC Dorsolateral prefrontal cortex, FWHM Full-width at half maximum, GPi Globus pallidus interna, GSH Glutathione, LW Line width, MC Motor cortex, MEGA-PRESS MEscher-GArwood point-resolved spectroscopy, MFC Medial frontal cortex, MOIST Multiply Optimized Insensitive Suppression Train, mPFC Medial prefrontal cortex, NA Not applicable, NEX Number of excitations, NR Not reported, PFC Prefrontal cortex, (R) Receive only, RF Radiofrequency, SN Substantia nigra, STEAM STimulated echo acquisition mode, (T) Transmit, TE Echo time, TEM Transverse electromagnetic, TM Mixing time, TR Repetition time, VAPOR Variable power radiofrequency pulses with optimized relaxation delays, VC Visual cortex, VOI Voxel of interest. Quality metrics: Signal to noise ratio, line-width, full-width at half maximum.

Voxels were placed in a total of 14 regions (Table 2) -thalamus(n = 5), basal ganglia/striatum(n = 5), SN (n = 3), occipital cortex (n = 2), motor cortex (MC) (n = 2), putamen(n = 2), medial prefrontal cortex (mPFC) (n = 2),and dorsolateral prefrontal cortex (DLPFC), sensory motor cortex (SMC), prefrontal cortex (PFC), midbrain, brainstem, pons and cerebellum- one study each. Voxel sizes varied between location, and only 5 studies^{29,32,33,35,46} reported volumes at or 27 mL. While larger voxel are generally recommended for MEGA-PRESS²⁵ GABA editing to optimise SNR, smaller voxels may be appropriate in situations where spatial specificity is prioritised or where brain atrophy is a confound, and can be offset by increasing the number of acquisitions.

Details pertaining to water suppression were reported in 13 studies, and VAriable Power radiofrequency pulses with Optimized Relaxation (VAPOR)(n = 7) was the most common method followed by chemical shift selective saturation (CHESS)(n = 4) and Multiply Optimized insensitive suppression train (MOIST)(n = 2). Quality metrics in the form of LW, FWHM, SNR or CRLB were reported in 14 studies.

Medication state during acquisition

Medication state during acquisition of spectroscopy data was reported in all except one study⁴⁴. Most acquired in the OFF state (n = 13), followed by ON state(n = 4), and both OFF and ON state data(n = 4).

Data analysis

Edited spectral data was processed using Gannet(n = 12), followed by LCModel which uses linear combination modelling(n = 4), Gannet + Advanced method for accurate, robust, and efficient spectral editing (AMARES)⁴⁹ (n = 1), FID-A + LCModel(n = 1), and KALPANA (n = 1). Both FID-A⁵⁰ and KALPANA⁵¹ are MATLAB based processing toolboxes.

GABA + , representing the combined signal contributions of GABA, macromolecules and homocarnisine at 3ppm, was reported by most studies (n = 19), while GABA without macromolecule contamination was reported by three studies who used STEAM for acquisition (Table 2). The quantification reference for GABA varied: most reported GABA normalised to water (n = 11), followed by total creatine (tCr) (n = 7). One study reported both methods tCr and water⁴⁵, and only one study reported absolute GABA concentrations⁴⁴.

Results of GABA spectroscopy

There were numerous domains of variability between studies, ranging from motor subtypes, NMS, medication states and voxel locations (Table 3, Figs. 2 and 3). The results are reported based on each of the above domains. First variability was observed based on location of the voxel, wherein:

Table 3.

Data analysis method and results of studies evaluating GABA in patients with Parkinson’s disease

Authors, year	Software	Output measures	Quantification reference	GABA/ GABA +	Key results
Oz et al.26	LCModel	GABA	NAA	Values NR. No difference between PD and HC.	-Unique neurochemical profile of the SN can be measured using STEAM spectroscopy at 4 T. -Trends observed between PD and HC.
Emir et al.27	LCModel	GABA	Water	Pons- PD: 1.66 ± 0.4, HC: 1.06 ± 0.2; Putamen- PD: 2.16 ± 0.4, HC: 1.66 ± 0.2 SN- Values NR. No difference between PD and HC	-Significantly ↑ GABA in the pons and putamen in PD.
Dharmadhikari et al.28	LCModel 6.2-0 R	GABA+	Water	Thalamus - PD: 2.06 ± 0.09, HC: 1.73 ± 0.09; Striatum: Values NR. No difference between PD and HC.	-PD has ↑ thalamic GABA compared to HC. -This may contribute to their different processing of proprioceptive information. -Dissociable roles of striatal and thalamic GABA, with striatal GABA affecting response speed and thalamic GABA affecting response conflict.
O’Gormon Tuura et al.29	LCModel 6.31-H	GABA+	Water	Left BG- PD: 4.61 ± 0.57, HC: 3.93 ± 0.75; Left PFC- PD: 3.48 ± 0.90, HC: 2.96 ± 0.96	-PD had significantly ↑ GABA in the BG which correlated with gait disturbances. -PFC GABA negatively correlated with postural stability in AR PD and PFC Glx negatively correlated with difficulties turning in bed. -Associations were more pronounced in AR vs TDPD
Elmaki et al.30	Gannet	GABA+	Water	PD: 1.31 ± 0.21, HC: 1.62 ± 0.26	-GABA+ levels were significantly ↓ in PD -Suggesting a role for GABAergic dysfunction in PD pathogenesis.
Gong et al.31	Gannet 2.0	GABA+	Water	TDPD: 1.15 ± 0.16, PIGD: 1.36 ± 0.18, HC: 1.56 ± 0.23	-GABA+ levels were significantly ↓in PD BG compared to HC. -TDPD subtype had ↓ GABA+ levels than the PIGD subtype. -Significant negative correlation between GABA+ levels and UPDRS scores in the PIGD group, but not in the TDPD.
O’Gormon Tuura et al.32	LCModel	GABA+	Water	Values NR. BG: PD > HC	-BG GABA levels were significantly ↑ and pontine Glx levels were significantly ↓ in PD compared to HC. -BG Glx levels were a significant predictor of DBS outcome, indicating a role for glutamatergic neurotransmission in therapeutic mechanisms. -Pontine Gln negatively correlated with motor outcomes post-DBS.
Firbank et al.33	Gannet AMARES (jMRUI)	GABA+	tCr	PD + VH: 0.091 ± 0.01; PD-VH: 0.101 ± 0.01; HC: 0.099 ± 0.01	PD + VH has lower occipital GABA compared to PD-VH and HC.
Pesch et al.34	LCModel 6.2-0R	GABA(+)	Water	Striatum- PD: 2.0(1.8-2.2), HC: 1.9 (1.8-2.1); Thalamus- PD: 2.1 (1.9-2.4), HC: 1.9 (1.7-2.4)	-Striatal GABA correlated with MDS-UPDRS III. -AR PD patients had ↓ Glx but ↑ thalamic GABA. -Thalamic Glx was linked to larger tremor amplitudes.
Piras et al.35	Gannet 3.0	GABA(+)	Water	Mean cerebellar- PD: 3.59 ± 0.61, HC: 3.48 ± 0.46; Left cerebellar – PD: 3.55 ± 0.96, HC: 3.53 ± 0.56; Right cerebellar – PD: 3.62 ± 0.82 HC: 3.44 ± 0.54	-↑ cerebellar GABA levels are associated with ↑ cognitive deficits in PD. -Non-dopaminergic basis for cognitive deficits in PD and highlights the role of cerebellum in cognitive processes.
van Nuland et al.36	LCModel	GABA(+)	tCr	Values NR. No significant differences between PD and HC.	-GABA levels were not significantly altered by PD, clinical phenotype, or medication. -MC GABA levels inversely correlated with disease severity, particularly rigidity and tremor, in OFF and ON medication states. -Potential modulatory or neuroprotective role of GABA in Parkinson’s disease.
Delli Pizzi et al.37	Gannet 3.0	GABA+	tCr	PD + SSD: 0.10 ± 0.02, PD-SSD: 0.08 ± 0.02; SSD: 0.11 ± 0.02; HC: 0.08 ± 0.02	-PD + SSD and SSD have ↑ GABA in mPFC compared to PD-SSD and HC. -May be a functional signature of SSD, independent of PD pathology.
Parkin et al.38	Gannet 3.0	GABA+	tCr	Values NR. Reduced GABA in PD vs HC.	Instruction-based learning linked to ↓ GABA levels in the mid-DLPFC.
Song et al.39	Gannet 3.0	GABA+	Water	OFF state:2.40 ± 0.77, ON state: 3.29 ± 0.61, HC: 3.50 ± 0.79	-Upper brainstem OFF state GABA+ levels were significantly ↓ than HC. -ON state GABA+ levels significantly ↑ in PD. -Levodopa may help normalize GABA+ levels and contribute to improved motor function in PD.
Song et al.40	Gannet	GABA+	Water	PD: 4.57 ± 0.94, HC: 5.89 ± 1.16	-GABA + levels are significantly ↓ in the upper brainstem of PD compared to HC. -This reduction in GABA + levels could serve as an early biomarker for PD.
Seger et al.41	FID-A, LCModel 6.3-0I	GABA	Water	PD: 1.50 ± 0.26; HC: 1.26 ± 0.31	-Putaminal GABA levels are significantly ↑in PD compared to HC. -↑ GABA levels are associated with a ↓ dopaminergic response. - Independent predictors of both absolute and relative treatment responses, with an inverse correlation observed between GABA levels and treatment efficacy. -Suggests significant role of nondopaminergic neurotransmitters in motor response in PD.
Delli Pizzi et al.42	Gannet	GABA+	tCr	PD + SSD:0.113 ± 0.028; PD: 0.084 ± 0.020; SSD: 0.116 ± 0.026; HC: 0.083 ± 0.020	-PD + SSD, SSD exhibited ↑ GABA levels in the mPFC in comparison to HC or PD-SSD.
Trujillo et al.43	Gannet 3.0	GABA+	Water	Values NR. Thalamic response to DA reduced in PD + ICD. No difference in MC.	-PD + ICD has a significantly ↓ thalamic GABA response to DA and correlated with ↑ impulsivity scores. -Findings highlight the role of dopamine therapy in modulating thalamic GABA concentrations and its potential impact on ICD.
Shukla et al.44		GABA+	Absolute	SN(Absolute)- PD: 3.2 ± 0.86 HC: 3.51 ± 0.81; SN(PVC)- PD: 4.58 ± 1.21, HC: 4.84 ± 1.15	-Significant ↓ of GSH in SN and LH of PD patients compared HC. -No significant change in GABA.
Liu et al.45	Gannet 3.1	GABA+	tCr, Water	Values NR. -GABA+ levels ↑ in the MFC of PD + D vs PD-D. -No significant differences in GABA+ levels between PD and HC, no difference in OFF vs ON states	-GABA+ and Glx levels were not significantly different between PD and HC. -In PD + D, GABA+ levels were ↑ in the MFC compared to PD-D. -Link between ↑ GABA+ levels and depressive symptoms.
Tian et al.46	Gannet	GABA(+)	tCr	EOPD: 0.08 ± 0.02, LOPD: 0.05 ± 0.02, Young HC: 0.11 ± 0.02, Old HC: 0.09 ± 0.02	-GABA levels were ↓ in PD, indicating role of GABAergic dysfunction -Suggest neurotransmitter and morphological changes as potential markers for early diagnosis and progression of PD, especially early-onset PD.
Yan et al.47	Gannet 3.0	GABA+	tCr	PD: 0.12 ± 0.03, HC: 0.14 ± 0.02	PD has ↓ midbrain GABA levels, indicating neurotransmitter imbalance.

AMARES Advanced method for accurate, robust, and efficient spectral fitting, AR Akinetic rigid, BG Basal ganglia, CSF Cerebrospinal fluid, EOPD Early onset Parkinson’s disease, GABA Gamma amino butyric acid, Gln Glutamine, Glu Glutamate, Glx Glutamate + glutamine, GSH Glutathione, HC Healthy controls, ICD Impulse control disorder, jMRUI Java based Magnetic Resonance User Interface, LOPD Late onset Parkinson’s disease, MC Motor cortex, mPFC Medial Prefrontal cortex, NR Not reported, PD Parkinson’s disease, PD-D Parkinson’s disease without depression, PD-ICD Parkinson’s disease without impulse control disorder, PD-SSD Parkinson’s disease without somatic symptom disorder, PD-VH Parkinson’s disease without visual hallucinations, PD + D Parkinson’s disease with depression, PD + ICD Parkinson’s disease with impulse control disorder, PD + SSD Parkinson’s disease with somatic symptom disorder, PD + VH Parkinson’s disease with visual hallucinations, PFC Prefrontal cortex, PIGD Postural instability and gait disturbance, PVC Partial volume correction, SSD Somatic symptom disorder, STEAM Stimulated echo acquisition mode, tCr Total creatine(creatine + phosphocreatine), TDPD Tremor dominant Parkinson’s disease, UPDRS Unified Parkinson’s disease ratings scale

GABA(+): Studies which report “GABA” as the output measure but have actually measured GABA+.

Fig. 2. Brain regions evaluated across included studies and key GABA findings by voxel location.

Dashed boxes indicate the approximate voxel locations; these are generalised to optimise visual inclusion and may not be fully anatomically precise. Panel A: axial view showing thalamus, basal ganglia/striatum, putamen, and medial prefrontal cortex. Panel B: lateral view showing motor cortex, dorsolateral prefrontal cortex, prefrontal cortex, and sensorimotor cortex. Panel C: lateral view showing occipital cortex, cerebellar hemispheres, brainstem, pons, midbrain, and substantia nigra. Results are expressed as comparisons between groups (e.g., PD-OFF medication vs. HC; PD-ON medication vs. HC). Abbreviations: PD: Parkinson’s disease; HC: Healthy controls; OFF: medication OFF state; ON: medication ON state; ICD: impulse control disorder; SSD: somatic symptom disorder; VH: visual hallucinations; D: depression; TDPD: tremor dominant PD; PIGD: postural instability and gait disturbance; DBS: deep brain stimulation; GPe: globus pallidus externa; GPi: globus pallidus interna; BGTC: basal ganglia thalamocortical; CTC: cerebello-thalamocortical. Created in BioRender. Prasad, S. (2026) https://BioRender.com/s92her0.

Fig. 3. Summary matrix of GABA findings across studies and brain regions.

Rows represent comparison groups and columns represent individual studies organised by brain region. Colour coding: purple = GABA higher in the first-named group; teal = GABA equal between groups; yellow = GABA lower in the first-named group; hatched = comparison not evaluated in that study. Brain regions (x-axis): Thalamus, Basal ganglia/Striatum (BG/Str), Putamen, Prefrontal cortex (PFC), Medial PFC (mPFC), Dorsolateral PFC (DLPFC), Motor cortex (MC), Sensorimotor cortex (SMC), Occipital cortex (Occ.C), Cerebellum, Midbrain, Brainstem, Pons, Substantia nigra. Comparison groups (y-axis) include PD-OFF/HC, PD-ON/HC, PD-OFF/PD-ON, TDPD/PIGD, Dopa-resistant/Dopa-responsive, PD + D/PD-D, PD + VH/HC, PD + SSD/PD-SSD, PD + SSD/HC, PD + ICD/PD-ICD, and PD(?state)/HC. Created in BioRender. Prasad, S. (2026) https://BioRender.com/p08k459.

Thalamus: Thalamic alterations were evaluated by 4 studies^28,34,36,45. Dharmadhikari et al.²⁸ and van Nuland et al.³⁶ evaluated PD in the OFF state but reported contradictory results – higher GABA in PD, and no difference, respectively. No difference was reported between patients with PD in the ON medication state and HC by all three studies that performed this comparison^34,36,45.

Basal ganglia, striatum: Significant variations were observed in results reported for variations of GABA between PD OFF state and HC. Of the 5 studies, 2 reported higher GABA in PD^29,32, while 2 reported lower GABA^30,31 and 1 reported no difference³⁴.

Putamen: Emir et al.²⁷ and Seger et al.⁴¹ reported GABA in PD-OFF medication and PD-ON medication states respectively. In comparison to HC, both reported higher levels of GABA in PD.

Prefrontal cortex: O’Gormon Tuura et al. found no difference between PD-OFF state and HC²⁹.

Medial prefrontal cortex: Delli Pizzi et al., reported no difference between PD-OFF and HC^37,42

Dorsolateral prefrontal cortex: PD-OFF state was found to have a lower GABA level in comparison to HC³⁸.

Motor cortex: In both the OFF and ON medication states, van Nuland et al., reported no difference between patients with PD and HC³⁶.

Sensory motor cortex: Tian et al., reported lower GABA in PD-OFF state compared to HC⁴⁶.

Occipital cortex: No difference was found between patients PD-OFF medication³⁶ or PD-ON medication^33,36 compared to HC.

Cerebellum: Piras et al., found no difference between PD-ON state and HC for either right cerebellar hemisphere, left cerebellar hemisphere or the mean cerebellar GABA³⁵.

Midbrain: No difference was reported between PD-OFF state and HC⁴⁷.

Brainstem: Song et al. consistently reported a lower brainstem GABA in PD-Off state compared to HC^39,40, with no difference between PD-ON state and HC³⁹.

Pons: Higher pontine GABA was reported by Emir et al., in PD-OFF state²⁷.

Substantia nigra: Both Oz et al. and Emir et al. reported no difference between PD-ON state and HC^26,27. Shukla et al. also reported a similar observation, however, the medication state during acquisition was not clearly reported in their study⁴⁴.

Significant variations were also observed on the basis of the motor symptoms of different patient groups and these symptoms included the motor subtype, presence or absence of axial symptoms and responsiveness of tremor to dopamine.

Motor subtypes: Gong et al. evaluated the TDPD and PIGD variants and reported distinct GABAergic profiles - TDPD had lower basal ganglia GABA levels than PIGD³¹. GABA and UPDRS scores were inversely correlated in the complete PD group, and the PIGD subtype. No correlation was observed for TDPD.

Axial symptoms: Basal ganglia GABA positively correlated with gait disturbance, gait summary scores, and difficulty standing up from a chair in akinetic-rigid PD²⁹. PFC GABA negatively correlated with postural stability in akinetic-rigid PD.

Dopamine-resistant tremor: Van Nuland et al. evaluated thalamic, MC, and occipital cortex GABA in dopamine-resistant and dopamine-responsive PD tremor and found no influence of clinical phenotype³⁶. MC GABA inversely correlated with OFF and ON rigidity and tremor scores. Negative correlations were reported between disease severity and GABA in TDPD.

As stated earlier, NMS form a major component of the symptom complex in PD and GABA related alterations have been reported based on the type of NMS.

Depression: Liu et al. reported higher mPFC GABA and equal thalamic GABA in PD with depression compared to PD without depression⁴⁵. However, no correlations were observed between GABA levels and Hamilton depression rating scale (HAM-D) and Hamilton Anxiety Rating Scale (HAM-A).

Somatic symptom disorder: A higher mPFC GABA was consistently reported in PD with SSD (PD + SSD) in comparison to PD without SSD (PD-SSD) and HC^37,42. Similar observations were reported in subjects with SSD alone, suggestive that higher GABA is an SSD trait independent of PD.

Impulse control disorders: Patients with PD + ICD were had lower thalamic GABA and equal MC GABA in comparison to PD-ICD⁴³. The change in thalamic GABA between OFF and ON medication (dopamine agonist) states (∆GABA) significantly correlated with impulsivity scores as measured by Questionnaire for Impulsive-Compulsive Disorders in Parkinson’s Disease-Rating Scale.

Visual hallucinations: Firbank et al., reported lower occipital cortex GABA in PD with visual hallucinations (VH) compared to both PD without VH and HC³³. No correlations were observed between GABA and neuropsychiatric inventory hallucination score but positive correlations were observed between GABA and visual acuity.

Cognition: Piras et al. reported positive correlations between mean cerebellar GABA and Stroop Word-Color Test short form error interference effect (SWCT- IE-E) and Stroop Word-Color Test short form time interference effect (SWCT- IE-T)³⁵. Hemispheric differences were noted with the left hemisphere GABA positively correlating with the SWCT-IE-E, and right hemisphere GABA negatively correlating with SWCT-IE-T.

NMS severity score: Song et al. reported no correlations between brainstem GABA and total NMS severity questionnaire scores, or individual domains⁴⁰.

Finally, the role of medical or surgical intervention on GABA levels has also been evaluated. Wherein, three studies reported a direct comparison of OFF and ON state GABA levels^36,39,45. Liu et al. reported no difference between mPFC GABA in the OFF and ON states⁴⁵. Song et al. reported significantly higher ON state brainstem GABA³⁹ compared to the OFF state and HC. Van Nuland et al. reported no differences between thalamic, MC and occipital cortex OFF and ON state GABA levels³⁶. No correlations were reported by Gong et al. between levodopa responsiveness and ∆GABA³⁹. Seger et al. reported inverse correlations between ON state putaminal GABA with response to dopaminergic treatment.

O’Gormon Tuura et al. attempted to evaluate if pre-DBS basal ganglia and striatal GABA could predict response to DBS³². GABA was initially identified as a significant predictor of outcome however; this correlation was lost after removing an outlier, limiting the strength of this conclusion.

Quality assessment

Quality assessment was based on MRS-Q⁵² (Fig. 4). High overall quality, defined by satisfaction of Domain 2 (sequence) and Domain 3 (scan parameters), was met by only two studies - Firbank et al. and Tian et al.^33,46. A high-quality sequence (Domain 2) was used in 86.3%(n = 19), 63.6%(n = 14) had the required number of averages (Domain 3-A), and only 22.7%(n = 5) had an adequate voxel size (Domain 3-B). 77.27%(n = 17) had an appropriate TE (Domain 3-C), however, 4 studies that did not satisfy this criterion used non-standard sequences – specifically, one used MEGA-PRESS with a non-standard TE (120 ms) and three used STEAM for which a TE of 68 ms is not applicable. It should be noted that these MRS-Q criteria were primarily developed for MEGA-PRESS at 3 T. For Seger et al.⁴¹., who scanned at 7 T using STEAM, MRS-Q criteria for sequence and scan parameters are not directly applicable and results for this study should be interpreted accordingly. Similarly, Oz et al.²⁶. used STEAM at 4 T and clearly reported acquisition parameters. Hence, a “No” rating for these domains reflects non-satisfaction of MEGA-PRESS specific criteria rather than absence of reporting. These criteria were recently established and several of the included studies are nearly two decades old. Therefore, lack of reporting does not imply poor quality.

Fig. 4. MRS-Q quality assessment of included studies across 11 domains (D1–D11).

Dark blue = Yes (criterion met); light blue = Partial; white = No (criterion not met); hatched = Not reported. D1: Scanner field strength; D2: Sequence; D3-A: Number of averages; D3-B: Voxel size/volume; D3-C: Echo time; D4: Data points; D5: Quality measures; D6: Data visualisation; D7: Scanner drift; D8: Power calculation; D9: Partial volume correction; D10: Frequency and phase correction; D11: Analysis method. Note: MRS-Q criteria were developed primarily for MEGA-PRESS at 3 T. For studies using STEAM (Oz et al., 2006; Emir et al., 2012; Seger et al., 2021) and ultra-high-field acquisition (Seger et al., 2021 at 7 T), D2 and D3 criteria are not directly applicable and should be interpreted with caution. Created in BioRender. Prasad, S. (2026) https://BioRender.com/0nt5bv6.

Discussion

The current systematic review examined the role of GABA in PD as measured by single-voxel GABA spectroscopy. Across 22 included studies, a range of patterns of GABAergic alterations were noted in PD suggesting a multifaceted, region specific dysregulation of GABA related neurotransmission that extends the concepts of PD pathogenesis beyond the traditionally implicated canonical dopaminergic framework. Alterations in levels of GABA (or GABA + ) levels were reported across numerous cortical and subcortical regions, with either increases, decreases, or no change in comparison to HC. This heterogeneity reflects the complex neurochemical landscape in PD. Although methodological inconsistencies limit the possibility of carrying out a meta-analysis, patterns do emerge. Collectively, the observations provide evidence of GABAergic dysfunction in PD that is intricately associated with motor and non-motor symptoms, and highlight the need for standardisation from both clinical and technical aspects of spectroscopy acquisition.

To contextualise these observations, it is necessary to first consider the GABAergic architecture of the motor circuits most relevant to PD pathogenesis. Direct and indirect pathways, are the primary networks related to motor functioning which are tightly linked to PD pathogenesis, relying on GABA as the key modulatory neurotransmitter (Fig. 5). Dopamine is responsible for modulation of these pathways via excitatory D1 receptors - direct(facilitatory) pathway, and inhibitory D2 receptors -indirect(inhibitory) pathway¹⁵. Degeneration of dopaminergic neurons and subsequent dopaminergic deficit leads to an imbalance, producing excessive inhibition of the ventral lateral thalamic nucleus and in turn reduced facilitation of the MC. This model explains rigidity and bradykinesia; however, it does not necessarily explain tremor⁵³. Rest tremor is suggested to arise from a tremor network comprised of the basal ganglia-thalamocortical, and cerebello-thalamocortical circuit^9,16,54. The ‘dimmer-switch’ hypothesis suggests the GPi triggers tremor activity, and the cerebello-thalamocortical circuit modulates tremor amplitude⁵⁴. Thalamocortical communication is further modulated by dopamine-dependent thalamic self-inhibition of the ventral intermediate nucleus (VIM) by the thalamic reticulate nucleus via D4 receptors⁵⁵. Dyskinesia is also associated with GABA wherein, loss of dopaminergic neurons leads to a reduction in postsynaptic activation of GABA receptors resulting in hyperexcitability of striatal neurons and subsequent dyskinesia⁵⁶.

Fig. 5. Schematic representation of the role of GABA in the basal ganglia thalamocortical (BGTC) and cerebello-thalamocortical (CTC) networks.

Panel A: normal connectivity. Panel B: alterations in Parkinson’s disease. Red lines: inhibitory GABAergic pathways; green lines: excitatory projections; orange/yellow lines: dopaminergic projections. Thickened lines indicate increased activity or connectivity relative to normal; dashed lines indicate reduced activity or connectivity. GPe: Globus pallidus externa; GPi: Globus pallidus interna; VIM: Ventral intermediate nucleus. Note: findings depicted in Panel B reflect a synthesis of established PD pathophysiology and theoretical models. Findings derived directly from the studies in this systematic review are those pertaining to GABA levels in specific regions (e.g., basal ganglia, thalamus, motor cortex, brainstem, mPFC); proposed mechanisms (e.g., GABA collapse hypothesis, calcium neurotoxicity) represent theoretical frameworks drawn from the wider PD literature and are not directly measured by MRS. Created in BioRender. Prasad, S. (2026) https://BioRender.com/a4skbi8.

Beyond its role in circuit-level modulation of motor function, GABA also exerts neuroprotective and metabolic influences that are directly relevant to the dopaminergic neuronal loss that defines PD. The GABA collapse hypothesis suggests that neurons are protected from calcium-based neurotoxicity by GABAergic inputs⁵⁷. Dopaminergic neurons are particularly vulnerable owing to their high energy requirements and dependence on slow calcium-based pacemaker activity⁵⁸. A reduction in GABA is known to dysregulate calcium voltage gated channels leading to calcium related excitotoxicity and oxidative stress which may promote ɑ-synuclein aggregation and further injury to dopaminergic neurons⁵⁹. In the context of Fig. 5, the GABA collapse hypothesis implies that region-specific GABA reductions may not only reflect circuit-level dysfunction but may actively contribute to neuronal loss and disease progression. Alterations in GABA metabolism, which include synthesis, reuptake, or changes in GABA receptor expression can also contribute to dysfunction. For instance, glutamic acid decarboxylase (GAD) is the principal enzyme for GABA synthesis (converting glutamate to GABA), and a reduction of GAD expression in the PFC has been reported in PD⁶⁰.

It is imperative to understand that GABA is not restricted to only the above-described networks, rather GABAergic neurons are widespread including other critical cortical and subcortical regions, and alterations in these areas may significantly contribute toward the development of NMS in PD¹⁷. Furthermore, it is important to appreciate that MRS measures total tissue GABA — encompassing all pools including synaptic, neurotransmitter, and metabolic compartments — and cannot distinguish between these pools. The precise compartmental contribution to the MRS GABA signal remains debated and likely depends on acquisition parameters. Extracellular GABA, involved in tonic inhibition^61,62, is thought to be a component of the MRS signal, but the relative contributions of intracellular and extracellular pools are not definitively established^63–66. MRS also has reduced sensitivity in smaller subcortical structures, which is relevant when interpreting results from the thalamus, putamen, and SN.

Despite this well-established physiological significance, translating these mechanistic concepts into consistent in vivo spectroscopic findings has proven difficult, as the existing literature presents a highly heterogeneous picture. It is evident that GABA plays a crucial physiological role and that alterations may contribute to PD symptomatology. However, the existing literature presents a highly heterogenous picture, with gross variations in results across studies, regions of interest and clinical phenotypes (Figs. 2 and 3). This heterogeneity reflects not only methodological differences but also genuine biological complexity: the included studies varied substantially in PD subgroups examined (e.g., TDPD vs. PIGD vs. unspecified), disease stage and duration (from <1 year to >10 years), the anatomical regions targeted, and the GABA-related metabolic pathways that may be differentially affected. The multiple functional roles of GABA ranging from motor circuits, limbic networks, and metabolic neuroprotection, suggests that a single, consistent directional change across regions and phenotypes would not be expected from the underlying pathophysiology of PD. Hence, characterising a definitive direction of GABA alteration across all regions and all patient groups is not consistent with the complexity of PD biology.

However, some patterns do emerge. With the exception of mPFC^37,42, brainstem^39,40 and SN^26,27, where no difference was observed, no other voxel with multiple studies reported consistent results. Within the basal ganglia which is the primary area of interest to PD, significant heterogeneity exists – with equal observations of high^29,32, equal^28,34 and low levels^30,31 of GABA in PD, particularly in the OFF state.

The thalamus warrants particular attention. Unlike single-function relay nuclei, the thalamus is a hub of varied afferent and efferent connections. This is relevant in context of evaluating motor symptoms in PD, wherein the ventral oralis anterior, i.e., ventral lateral nucleus is linked with the BGTC, whereas the VIM is linked with the CTC. It is plausible voxels encompassing multiple nuclei may produce averaged results that obscure nucleus-specific changes, limiting interpretability. In the same vein, the large voxel sizes required for GABA editing mean that spatial specificity is limited. For example, a basal ganglia voxel encompasses the putamen, globus pallidus externa (GPe), and GPi, each of which likely harbours specific, distinct GABAergic changes.

One notable exception to this heterogeneity, however, emerges when the effect of dopaminergic medication is considered. In contrast, to the heterogeneity of PD-OFF versus HC comparisons, almost all studies comparing PD-ON state to HC consistently reported equal GABA levels, irrespective of brain region. Only a single study reported higher putaminal ON state GABA compared to HC⁴¹. This is a crucial observation as it highlights levodopa induced normalisation of neurotransmitters in PD. The elevated putaminal GABA in that study may be due to striatal GABAergic neurons, rather than GABAergic afferents, as primary striatal afferents are dopaminergic and glutamatergic, whereas medium sized GABAergic spiny neurons are the predominant cells in the striatum. Animal models suggest that nigrostriatal pathway lesions lead to elevated striatal GABA, and this may be observed in PD⁶⁷.

Very few studies evaluated medication induced variations in GABA levels, and majority reported no differences between OFF and ON medication states. While intriguing, it is plausible and implores further evaluation of the effect of medication on the directionality of change, i.e., toward or away from normal. van Nuland et al.³⁶, evaluated this in the thalamus, motor and occipital cortex, however, they reported no influence of disease subtype or medication state. In contrast, Song et al.³⁹, reported lower brain stem GABA in PD-OFF state compared to HC and PD-ON state, suggesting levodopa induced normalisation of GABA. The absence of consistent observations related to medication effects may reflect the multiple confounders involved. These can include dose of levodopa, time point in the ON state when data was acquired, individual variations in pharmacodynamics and pharmacokinetics of levodopa, stage of disease, dopamine reserve, etc.

Alongside medication state, motor subtype represents another clinically important source of GABAergic variability that has been insufficiently explored in the literature. The exploration of GABAergic influences is inadequately and inconsistently explored. However, there are relevant observations within the limited studies. For instance, Gong et al, compared TDPD and PIGD PD³¹. Their observation of higher basal ganglia GABA in PIGD supports the classical model of PD pathogenesis, as it indicates the possibility of higher basal ganglia inhibition producing reduced excitation of the MC and subsequent reduction in movement. Axial symptoms of PD are often poorly responsive to dopaminergic treatment, and indicates the possibility of non-dopaminergic influences. The basal ganglia is a crucial hub for sensorimotor integration and descending projections toward the midbrain play a role in postural control and gait¹⁵. O’Gormon Tuura et al. reported a positive association between basal ganglia GABA and gait disturbance scores in akinetic-rigid PD, and PFC GABA and postural stability²⁹. However, they did not directly compare akinetic-rigid PD and TDPD. Only van Nuland et al.³⁶, focused on PD tremor and implications of dopaminergic medication. They reported an inverse relationship between MC GABA and tremor, suggesting that a reduction in cortical GABA produces higher tremor severity which is in line with current concepts and role of the CTC in tremor modulation. An attempt was also made to evaluate if basal ganglia GABA could predict response to subthalamic nucleus DBS. The study had limitations in sample size and moreover, the observed results of GABA being a good predictor were nullified after removal of a significant outlier³². However, this is an interesting concept and deserves to be evaluated by future studies.

The influence of GABAergic dysfunction is not confined to motor circuitry; several studies have specifically examined its contribution to the NMS burden of PD. Several studies have focused on evaluating NMS in PD as these extend beyond the conventional understanding of PD pathogenesis. The motor network is one of many roles of the basal ganglia as it is well established that limbic and associative/cognitive loops also exist^68,69. This implies that dopaminergic deficit or even excess can strongly influence the development of NMS. Almost all studies evaluating GABA and NMS have focused on neuropsychiatric domain of NMS–depression, SSD, ICD, hallucinations, and cognition. Perhaps this is supported by the corpus of previous studies evaluating spectroscopic alterations in psychiatric disorders^70,71. The presence of existing information helps form parallels between pure psychiatric disorders and the psychiatric NMS seen in PD, aiding in ascertaining the basis for NMS.

Depression in the context of PD may arise from an imbalance in excitation and inhibition in the PFC secondary to altered inhibitory transmission to key glutamatergic regions⁷² leading to suppressed excitatory neurotransmission. This observation was also made by Liu et al.⁴⁵ who reported higher mPFC GABA in PD with depression. The study designs by Delli Pizzi et al.^37,42 to evaluate PD with SSD is ideal to dissect the exact changes. They included four groups – PD with SSD, PD without SSD, pure SSD and HC. These comparisons revealed that higher mPFC GABA was seen in both PD with SSD and pure SSD, indicating that higher GABA was a SSD trait rather than associated with PD. Such a design would also be helpful to understand other crucial NMS like ICD and psychosis which often tend to be medication related. Trujillo et al.⁴³, reported a reduction in thalamic GABA in PD with ICD, further substantiating the complex regional interplay and impaired inhibitory mechanisms in thalamocortical circuits. Psychosis in PD is an important NMS, and Firbank et al. suggested that low GABA levels may predispose development of hallucinations, however, the occurrence of visual hallucinations is controlled by other factors such as visual environment and attention³³. These observations open therapeutic avenues for attempting to use other groups of medication. Interestingly, cognitive alterations were evaluated in relation to cerebellar GABA levels by Piras et al.³⁵, and significant role of GABA in producing decreased efficiency in filtering task-irrelevant information. Although the cerebellum is not a conventional target for cognition research in PD, this above observation significantly expands the network perspective of cognitive NMS.

Interpreting these clinical observations, however, requires careful consideration of the substantial methodological variability that characterises the included studies. There are several reasons to explain the heterogeneity of results across the observed studies. From a methodological perspective, there was variability in field strengths, pulse sequences, voxel location, voxel size, quantification reference, and methods of processing. Majority of the evaluated studies used MEGA-PRESS, however, this sequence although robust has limitations due to sensitivity to motion, frequency shifts and a large chemical shift displacement error (CSDE)²². This may impede the specificity of data acquired from the voxel, particularly in crowded subcortical area of the basal ganglia and thalamus.

Several studies have reported GABA/tCr ratios however there is lack of consensus about the stability of tCr concentrations and if it is truly unaltered in patients with PD for use as an internal reference. Altered tCr levels have been reported in the PD basal ganglia but not in other regions, and some studies report altered level following administration of levodopa^31,73. Hence, it may be inappropriate to use tCr as a reference especially for the basal ganglia. Future studies should aim to report GABA concentrations in absolute units (mmol/L), using water referencing with explicit corrections for partial volume effects (CSF, grey matter, and white matter fractions), T₁/T₂ relaxation, and macromolecule baseline contamination, and should clearly state which of these corrections were applied. The term ‘absolute values’ alone does not guarantee comparability across studies unless these methodological choices are standardised and transparently reported. This is emphasised by the observation that within voxels similar observations of GABA alterations were often reported by the same group who probably followed relatively uniform methods. For e.g. both observations of high GABA in the basal ganglia were by the same group^29,32, or even the mPFC^37,42, brainstem^39,40 and SN^26,27.

An important finding of this review is that the terms “GABA” and “GABA + “ have been used interchangeably across the included studies despite referring to the same experimental measure in the majority of cases. Based on the editing pulse parameters reported in each study’s methods (Table 2), all 18 MEGA-PRESS studies applied editing pulses at approximately 1.9 ppm (edit-ON) — consistent with the traditional GABA+ experiment — and therefore measured the combined signal of GABA, macromolecules, and homocarnosine at 3 ppm, regardless of the notation applied in the original publication. No study in this review used MM-suppressed GABA editing. The three STEAM studies^26,27,41 represent a genuinely distinct measure, as GABA is resolved from overlapping resonances by LCModel fitting of short-echo unedited spectra rather than by J-difference editing, and is not subject to macromolecule co-editing. Future studies should clearly state the editing pulse parameters (ON/OFF editing pulse frequency, duration, bandwidth) to allow independent verification of the GABA measure and should consistently apply the GABA+ notation when standard MEGA-PRESS editing parameters are used.

Considering the fact that motor subtypes, NMS and medication can contribute to specific changes in GABA levels, clinical phenotyping is critical to outcomes. Numerous studies reported no difference between PD and HC, however it is possible that the admixture of multiple phenotypes in a single PD group led to nullifying of specific alterations in comparison to HC. These may have become evident if patient selection and groups were more specific. The dosage of levodopa used to induce an ON state for scanning was also not clearly described in most studies. It is evident that levodopa may have a strong modulatory influence and variations in dosing across subjects even within the same study will lead to variations. In addition, use of the other medication, caffeine, other stimulants, was not consistently reported and this can often confound results. Perhaps similar to the MRSinMRS guidelines, clinical reporting guidelines are also required to ensure transparency and increase interpretability.

These methodological considerations apply equally to the present review, which has several limitations that should be acknowledged. Several limitations of this review should be acknowledged. First, all screening and data extraction were performed by a single reviewer (SP), which increases the risk of missed studies and extraction errors; this process was not verified by a second independent reviewer. Second, the literature search was restricted to two databases (PubMed and Scopus), and grey literature and preprints were not systematically searched, which may have led to incomplete identification of the evidence base. Third, restriction to English-language publications may have excluded relevant studies published in other languages. Fourth, the heterogeneity of included studies — in terms of patient populations, acquisition parameters, and analysis methods — precluded formal meta-analysis and limits the strength of pooled conclusions. Fifth, MRS-Q was the only quality assessment tool applied; standard risk-of-bias tools used in systematic reviews of observational studies (e.g., Newcastle-Ottawa Scale) were not applied. Sixth, the applicability of MRS-Q criteria to studies using STEAM or ultra-high-field acquisitions is limited, as those criteria were developed primarily for MEGA-PRESS at 3 T. Finally, the cross-sectional nature of most included studies precludes causal inference about the relationship between GABA alterations and PD pathophysiology or symptom development.

In conclusion, the current systematic review provides a comprehensive synthesis of in vivo single-voxel GABA spectroscopy studies in PD. GABA plays a definite modulatory role in the pathogenesis of PD, and complex, distinct regional alterations contribute to the development of motor and non-motor symptoms. Thalamic, basal ganglia, brainstem, and cortical GABAergic changes are differentially associated with motor subtypes, dopaminergic medication state, and neuropsychiatric NMS, though substantial inter-study heterogeneity precludes definitive conclusions about directional changes across regions. The consistent finding that PD-ON medication state GABA is comparable to HC across most regions highlights the modulatory influence of levodopa on GABAergic neurotransmission.

This area of research has significant potential. Methodological uniformity — including adoption of standardised acquisition protocols (e.g., MEGA-semiLASER to reduce CSDE), clear specification of GABA vs. GABA+ measures, and consistent use of water-referenced quantification with appropriate corrections — is essential for progress. Detailed clinical phenotyping, longitudinal study designs, and assessment of broader NMS domains (including sleep dysfunction) are required to establish the role of GABA across the PD disease spectrum. Integration of spectroscopy with other methods of evaluating GABAergic function including ¹¹C-Flumazenil PET imaging and TMS, and simultaneous assessment of excitatory neurotransmitters to characterise the excitatory-inhibitory balance would substantially advance understanding. Evaluation of prodromal PD, at-risk populations, and genetic mutation carriers may aid in developing GABA-related imaging biomarkers for early detection and monitoring of PD.

Methods

Search strategy

This review was conducted in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-analysis statement (PRISMA)⁷⁴. A systematic search of two databases – PubMed (via NCBI) and Scopus (via Elsevier) was carried out in April 2025, through to the date of search. The following Boolean phrase was used (“magnetic resonance” AND “spectroscopy”) AND (“parkinson’s disease” OR “parkinsons disease” OR “Parkinson disease”) AND (“GABA” OR “gamma-aminobutyric acid” OR “gamma-amino-butyric acid” OR” γ-Amino-butyric acid” OR “γ-aminobutyric acid”). No additional databases (e.g., Web of Science, EMBASE, Cochrane) were searched, and this is acknowledged as a potential limitation. Grey literature and preprints were not systematically searched. No date limits were applied. Reference lists of included studies were also manually searched to ensure key studies were not overlooked.

Criteria for inclusion and exclusion

Inclusion criteria: Studies were included if they met all of the following criteria (1) Population: Human subjects (2) Disease: PD (3) MRS method: Field strength of 3 T or higher, pulse sequences established to measure GABA (MEGA-PRESS, STEAM, etc), single-voxel MRS. (4) Language: English

Exclusion criteria: Studies were excluded based on the following criteria: (1) Review article (2) Animal models of PD (3) Magnetic resonance spectroscopic imaging (MRSI; i.e., multi-voxel, non-single-voxel approaches). This was followed as the focus of this review was on single-voxel studies to allow more targeted region-specific interpretation (4) Non-MRS studies (5) Studies where GABA was not reported as an outcome.

Study selection and data extraction

All articles obtained from PubMed and Scopus were pooled and duplicates were removed based on title and author matching. Following this, titles and abstracts were assessed for eligibility based on the inclusion and exclusion criteria. When titles and abstracts were inconclusive, full texts were reviewed. Full texts of all eligible articles were obtained and data was extracted.

All screening and data extraction was carried out by the first author (SP). This single-reviewer process is acknowledged as a limitation as it may affect the completeness of the evidence base.

Data was extracted based on the Minimum Reporting Standards for in vivo Magnetic Resonance Spectroscopy (MRSinMRS) guidelines⁷⁵. This was used as a data extraction template and not as an exclusion criteria. These standards guided documentation of the following:

Bibliometric data

Authors, year of publication

Demographic and clinical details

Sample size, biological sex, age at assessment, duration of illness, Movement Disorder Society-Unified Parkinson’s Disease Rating Scale part III (MDS-UPDRS-III) or UPDRS-III OFF and ON state scores (OFF: without medication, ON: after consumption of dopaminergic medication), and levodopa equivalent daily dose (LEDD);

Hardware and acquisition details

Field strength, manufacturer, model, radiofrequency (RF) coil, pulse sequence, location of voxel-of-interest (VOI), VOI size, repetition time (TR), echo time (TE), water suppression method;

Medication state while scanning

Medication OFF or ON state of the subject

Data quality metrics

Signal-to-noise ratio (SNR), line width (LW), quantification precision (Cramer-Rao minimum lower bounds (CRLB), GABA fit error).

Data analysis

Software, output measures, quantification reference, GABA/GABA+ levels.

In terms of the GABA measure reported, MEGA-PRESS at a TE of 68 ms with editing pulses at 1.9 ppm results in a co-edited signal from GABA and macromolecules (MM) at 3 ppm; this combined signal is conventionally referred to as GABA + ²⁵. Some publications do not use this notation explicitly but have nonetheless applied editing parameters consistent with a traditional GABA+ experiment. Macromolecule-suppressed GABA editing, which uses modified pulse parameters (e.g., editing at alternative ppm offsets) to minimise macromolecule contributions, yields a more specific GABA measure. In this review, both GABA and GABA+ have been included and are interpreted in context; studies using STEAM at ultra-high field provide a distinct GABA measure not subject to the same macromolecule contamination as standard MEGA-PRESS.

Quality assessment

Quality of included studies was evaluated using MRS-Q - a quality appraisal tool specifically designed for the systematic review of MRS studies⁵². It is important to note that MRS-Q criteria were primarily developed for MEGA-PRESS acquisitions at 3 T, and their application to studies using STEAM or ultra-high-field (7 T) acquisitions requires caution, as some criteria (e.g., TE, sequence type) are not directly applicable. Where MRS-Q criteria were not appropriate for a given study’s acquisition method, this is noted in the quality assessment results. No additional risk-of-bias assessment tools beyond MRS-Q were applied, which is acknowledged as a limitation.

Author contributions

S.P.—Conceptualization, Project Administration, Investigation, Methodology, Formal Analysis, Writing: Original Draft Preparation. D.K.D.—Validation, Writing: Review & Editing. M.K.—Validation, Writing: Review & Editing. R.Y.—Validation, Writing: Review & Editing. P.K.P.—Validation, Writing: Review & Editing. J.S.—Validation, Writing: Review & Editing.

Data availability

The datasets generated and/or analyzed during the current study (including extracted summary data from included studies) are not publicly available in a dedicated repository but are available from the corresponding author on reasonable request. All primary data are available in the original published studies cited in this review.

Competing interests

The authors declare no competing interests.