Disruption of dopaminergic and glutamatergic neurotransmission has been proposed to be central to the pathophysiology of schizophrenia1-4. Single photon computed emission tomography (SPECT) and positron emission tomography (PET) allow the dopamine system to be studied in vivo, while in vivo quantification of glutamate levels is possible using proton magnetic resonance spectroscopy (1H-MRS).

Meta-analyses of available studies have found consistent evidence of higher striatal dopamine synthesis and release capacity in schizophrenia, and shown that this is greatest in the associative region of the striatum5, 6. In contrast, meta-analyses of studies investigating dopamine D2/D3 receptor availability have not shown significant patient-control differences in schizophrenia, although reporting increased variability in receptor availability6-9.

Meta-analyses of studies examining glutamate function have shown that, in individuals with psychosis, glutamate levels are higher in the basal ganglia, the glutamate metabolite glutamine is higher in the thalamus, while glutamate in combination with glutamine (Glx) is higher in the hippocampus1. In the frontal cortex, a recent meta-analysis of 7-Tesla studies reported lower glutamate in patients10.

These findings indicate that dopamine and glutamate dysfunction occurs in schizophrenia, but raise the question of whether it predates the onset of the disorder. It is possible to investigate neurochemical changes prior to the onset of schizophrenia by studying people at increased risk for developing the disorder.

The presence of sub-clinical symptoms prior to the development of psychosis has long been recognized11. People with schizotypal disorder experience sub-clinical psychotic symptoms, and are at increased risk of developing psychotic disorders, predominantly schizophrenia, with a risk of 25-48% over long-term follow-up12-14. The introduction of structured clinical assessments has also allowed the identification of individuals at clinical high risk (CHR) for psychosis, in whom the risk of transition to psychosis is around 20-30% over two years15. To meet criteria for CHR, a person is required to show one or more of the following at or above threshold levels: schizotypal disorder plus recent onset functional impairment, and/or brief intermittent psychotic symptoms, and/or attenuated psychotic symptoms16.

In addition to studying individuals at increased clinical risk, research has also been undertaken to quantify neurochemical functioning in individuals at genetic high risk (GHR) for schizophrenia. These studies have either investigated non-psychotic relatives of individuals with schizophrenia, or individuals with copy number variants, such as the copy number deletion of 1.5-5 megabases at 22q11.2, which is associated with a ~45% lifetime risk of developing psychosis and ~35% lifetime risk of developing schizophrenia17, 18.

There is some evidence that neurochemical dysfunction may primarily exist in a subgroup of high-risk individuals who subsequently develop psychosis19, 20. If neurochemical alterations occur only in a subgroup of high-risk individuals, this would be expected to lead to increased variability of the parameter in question in the high-risk group21. Novel meta-analytic techniques now allow for the quantification of variability across studies22-24. It is therefore possible to test meta-analytically the hypothesis that greater variability of dopamine and glutamate measures exists in high-risk individuals compared to controls.

A number of 1H-MRS, PET and SPECT studies have investigated dopamine and glutamate functioning in CHR and GHR groups25-28, but to our knowledge no meta-analyses of the dopamine findings has been undertaken, and an earlier meta-analysis of the glutamate findings29 is now outdated, since six new studies have been published after it was conducted30-35, increasing the sample size by 574 subjects. Moreover, variability has never been investigated for either dopamine or glutamate studies.

In the present paper, we meta-analyze neuroimaging studies of the dopamine and glutamate systems in individuals at high clinical or genetic risk for psychosis to provide the best estimate of the magnitude and variability of group differences across samples and settings.

METHODS Search strategy and study selection

EMBASE, PsycINFO and Medline were searched from January 1, 1960 to November 26, 2020. Titles and abstracts were searched for the words (“schizophrenia” OR “psychosis” OR “schizophreniform” OR “prodrom*” OR “at risk mental state” OR “high risk” OR “22q” OR 16p OR “vcfs” OR “velocardiofacial”) AND (“positron emission tomography” OR “PET” OR “single photon emission tomography” OR “SPET” OR “single photon emission computed tomography” OR “SPECT” OR “MRS” OR “spectroscopy”) AND (“dopamine” OR “glutamate”).

We included studies of: a) subjects meeting established research criteria for having an at risk mental state for psychosis determined using a structured assessment instrument (the Comprehensive Assessment of At-Risk Mental States36 or the Structured Interview for Prodromal Symptoms37); b) subjects meeting DSM or ICD criteria for a diagnosis of schizotypal personality disorder/schizotypal disorder; and c) non-psychotic people at increased genetic risk for schizophrenia (for example, relatives of individuals with schizophrenia, or non-psychotic individuals with a diagnosis of 22q11.2 deletion syndrome or 16p11.2 duplication syndrome). These studies had to report one or more imaging measures of striatal presynaptic dopaminergic function, striatal D2/D3 receptor availability, glutamate or Glx concentrations, for patient and control groups. As in previous meta-analyses5, 6, studies of striatal presynaptic dopamine function included those of dopamine synthesis capacity, dopamine release capacity, and synaptic dopamine levels. Furthermore, studies had to provide data enabling the estimation of standardized mean differences between patient and control groups for the relevant parameter.

We excluded data in individuals with comorbid substance dependence, as this may have significant effects on the dopamine system38-40.

Data extraction

The primary outcome of interest was the imaging parameter reported for patient and control groups. In addition, first author, year of study, number of participants, participant age, participant gender, antipsychotic treatment, transitions to psychosis observed over clinical follow-up, and symptom scores were extracted.

Where dopamine measures for the whole striatum were not provided, but data for the caudate and putamen were reported, whole striatum values were calculated by weighting these values by their volumes as reported in the Oxford-GSK-Imanova Structural-Anatomical Striatal Atlas (43% and 57% respectively). If data for ventral striatum were reported, the following weightings were used to derive a summary outcome for the whole striatum: 36% for caudate, 48% for putamen, and 16% for ventral striatum41. If only functional subdivisions were reported, the following weightings – based on templates used in previous imaging studies25, 42 – were used to derive a summary outcome for the whole striatum: 12.1% for limbic striatum, 61.9% for associative striatum, and 26.0% for sensorimotor striatum.

Data analysis

For the meta-analysis of mean differences, standard effect sizes (Hedges’ g) for individual studies were estimated.

The relative variability of imaging measures in high-risk individuals compared to controls can be quantified using the variability ratio (VR), where ln is natural logarithm; and are the unbiased estimates of the population standard deviation for the high-risk and control groups; Sh and Sc are the reported standard deviations, and nh and nc are the sample sizes. In biological systems, however, variance often scales with mean22, 23, and we therefore used the log coefficient of variation ratio (CVR) as our primary outcome measure in this analysis, where and are the mean symptom scores of high-risk and control groups.

All statistical analyses were carried out using the ‘metafor’ package (version 2.0.0) in the statistical programming language R (version 3.3.1). Separate meta-analyses were conducted for GHR and CHR individuals. For dopamine studies, a distinction was made between studies of presynaptic dopaminergic function and those of D2/D3 receptor availability. Glutamate studies were analyzed separately both on the basis of the region studied and on whether they assessed glutamate or Glx. Meta-analysis was only performed if at least three eligible studies were available. Egger's test, funnel plots and trim and fill analyses were conducted to test for publication bias, and the I2 statistic was used to quantify study inconsistency.

In both the meta-analysis of standardized mean differences and that of CVR, individual study effect sizes were entered into a random effects meta-analytic model using restricted maximum likelihood estimation.

The time period of risk is longer in people with schizotypal disorder compared to individuals meeting criteria for an at-risk mental state. Sensitivity analyses were therefore conducted to determine the effect of excluding the studies of schizotypal disorder on the findings.

Meta-regressions were undertaken to investigate potential associations between study effect sizes and age, gender composition and publication year. These analyses were performed in all instances where there were at least five eligible studies.

A significance level of p<0.05 (two-tailed) was used for all analyses.

RESULTS

A total of 5,454 papers were identified. Forty-eight of these met inclusion criteria, reporting data on 1,288 high-risk individuals and 1,187 controls (Figure 1). The average age of study participants was 26.5 years, and 52.6% of participants were male.

PRISMA flow chart. CHR – clinical high risk, GHR – genetic high risk

Striatal presynaptic dopaminergic function in clinical high-risk subjects

Eight studies of CHR individuals met inclusion criteria18, 42-48 (see Table 1). The studies included a total of 188 CHR individuals and 151 controls. The two groups did not differ significantly in terms of striatal presynaptic dopaminergic function (Hedges’ g=0.28, 95% CI: –0.03 to 0.59, p=0.07) (see Figure 2). The I2 value was 46%, indicating moderate between-study inconsistency. Neither Egger's test (p=0.75) nor trim and fill analysis suggested publication bias.

Table 1. Studies investigating striatal dopamine in individuals at clinical or genetic high risk for psychosis Study Probands Controls PET tracer N Age (yrs., mean) At-risk group Antipsychotic treatment N Age (yrs., mean) Presynaptic dopaminergic function Huttunen et al49 17 34.1 FDR All naïve 17 33.0 18F-DOPA Brunelin et al28 8 28.5 FDR All naïve 10 27.7 11C-raclopride + metabolic stress Shotbolt et al27 7 43.0 1 MZ, 6 DZ All naïve 20 39.0 18F-DOPA Kasanova et al50 16 42.4 FDR All naïve 16 38.1 18F-fallypride + reward task van Duin et al51 12 33.1 22q All naïve 16 38.1 18F-fallypride + reward task Rogdaki et al52 21 26.1 22q All naïve 26 26.1 18F-DOPA Abi-Dargham et al43 13 36.0 SPD Free for ≥21 days 13 34.0 [123I] IBZM + AMPH Howes et al18 30 24.2 CHR All naïve 29 25.6 18F-DOPA Egerton et al44 26 22.7 CHR 24 free/naïve, 2 medicated 20 24.5 18F-DOPA Bloemen et al45 14 22.0 CHR All free and less than 1 week lifetime use 15 22.2 [123I]IBZM +AMPT Tseng et al46 24 23.6 CHR All naïve 25 25.1 [11C]-(+)-PHNO + MIST Howes et al42 51 23.0 CHR All naïve 19 25.1 18F-DOPA Girgis et al47 14 22.4 CHR All free 14 22.7 [11C]-(+)-PHNO + AMPH Thompson et al48 16 37.4 SPD All naïve 16 37.0 11C-raclopride + AMPH D2/D3 receptor availability Hirvonen et al54 11 50.2 6 MZ, 5 DZ All naïve 13 51.5 11C-raclopride Lee et al55 11 25.1 2 MZ, 9 FDR All naïve 11 25.5 11C-raclopride Brunelin et al28 8 27.7 FDR All naïve 10 28.5 11C-raclopride van Duin et al51 12 33.1 22q All naïve 16 38.1 18F-fallypride Vingerhoets et al53 15 28.2 22q All naïve 11 26.6 [123I]IBZM Abi-Dargham et al43 13 36.0 SPD Free for ≥21 days 13 34.0 [123I]IBZM Tseng et al46 24 23.6 CHR All naïve 25 25.1 [11C]-(+)-PHNO Vingerhoets et al53 16 23.1 CHR All naïve 11 26.6 [123I]IBZM Girgis et al47 14 22.4 CHR All free 14 22.7 [11C]-(+)-PHNO Thompson et al48 16 37.4 SPD All naïve 16 37.0 11C-raclopride CHR – clinical high risk, FDR – first-degree relatives, MZ – monozygotic twins, DZ – dizygotic twins, 22q – 22q11 deletion syndrome, SPD – schizotypal disorder, PET – positron emission tomography, AMPH – dextroamphetamine, AMPT – alpha-methyl-paratyrosine depletion, MIST – Montreal Imaging Stress Test, IBZM – I-(S)-2-hydroxy-3-iodo-6-methoxy-N-[1-ethyl-2-pyrrodinyl)-methyl]benzamide

Forest plots of studies investigating standardized mean differences of measures of dopamine and glutamate in individuals at clinical or genetic high risk for psychosis. SMD – standardized mean difference (Hedges’ g), CHR – clinical high risk, GHR – genetic high risk, Da – dopamine, Glu – glutamate, Glx – glutamate + glutamine; K – number of studies

A sensitivity analysis excluding the two studies of schizotypal disorder was conducted, and provided similar results (Hedges’ g=0.25, 95% CI: –0.10 to 0.60, p=0.17). When the six studies reporting functional subdivisions were analyzed on a by-subdivision basis, there was no evidence for differences in striatal presynaptic dopaminergic function for any subdivision (associative: g=0.20, p=0.20; sensorimotor: g=0.20, p=0.12; limbic: g=0.21, p=0.26).

The meta-analysis of variability did not show differences in variability for CHR individuals compared to controls (CVR=0.13, 95% CI: –0.01 to 0.27, p=0.06) (see Figure 3).

Forest plots of studies investigating variability differences of measures of dopamine and glutamate in individuals at clinical or genetic high risk for psychosis. CVR – coefficient of variation, CHR – clinical high risk, GHR – genetic high risk, Da – dopamine, Glu – glutamate, Glx – glutamate + glutamine, K – number of studies

Striatal presynaptic dopaminergic function in genetic high-risk subjects

Six studies reported findings in individuals at increased genetic risk for schizophrenia, four of which examined relatives of individuals with schizophrenia27, 28, 49, 50, and two reported findings in individuals with 22q11 deletion syndrome51, 52 (see Table 1). These studies reported data on 81 GHR individuals and 105 controls. There was no significant difference in striatal presynaptic dopaminergic function between the two groups (Hedges’ g=0.24, 95% CI: –0.40 to 0.88, p=0.46) (see Figure 2). The I2 statistic was 77%, indicating substantial between-study inconsistency. Egger's test was significant (p=0.02), although a trim and fill analysis did not suggest any potentially missing studies.

The meta-analysis of variability did not show differences in variability for GHR individuals compared to controls (CVR= –0.04, 95% CI: –0.25 to 0.17, p=0.72) (see Figure 3).

Striatal D2/D3 receptor availability in clinical high-risk subjects

Five studies43, 46-48, 53 examined striatal D2/D3 receptor availability in 83 CHR individuals and 79 controls (see Table 1). There were no significant differences between the two groups (Hedges’ g=–0.08, 95% CI: –0.48 to 0.33, p=0.70) (see Figure 2). The I2 value was 39%, indicating moderate between-study inconsistency. Neither Egger's test (p=0.9) nor trim and fill analysis suggested publication bias.

The meta-analysis of variability did not show differences in variability for CHR individuals compared to controls (CVR=0.11, 95% CI: –0.17 to 0.39, p=0.43) (see Figure 3).

Striatal D2/D3 receptor availability in genetic high-risk subjects

Five studies28, 51, 53-55 examined striatal D2/D3 receptor availability in 57 GHR individuals and 61 controls. There was no significant difference between the two groups (Hedges’ g=–0.03, 95% CI: –0.39 to 0.34, p=0.88) (see Figure 2). The I2 value was 0%, indicating low between-study inconsistency. Neither Egger's test (p=0.9) nor trim and fill analysis suggested publication bias.

The meta-analysis of variability showed significantly reduced variability for GHR individuals compared to controls (CVR=–0.24, 95% CI: –0.46 to –0.02, p=0.03) (see Figure 3).

Glutamate function in clinical high-risk subjects

Three studies35, 56, 57 measured glutamate (215 CHR individuals, 133 controls), and ten studies33, 35, 56-63 measured Glx (375 CHR individuals, 306 controls) in the prefrontal cortex (see Table 2). Neither set of studies found any significant differences between CHR individuals and controls (glutamate: g=0.01, 95% CI: –0.21 to 0.22, p=0.96; Glx: g=0.01, 95% CI: –0.15 to 0.16, p=0.92) (see Figure 2). Both glutamate and Glx studies showed low between-study inconsistency (I2=0%). Neither set of studies showed evidence of publication bias as examined using Egger's test (glutamate: p=0.63; Glx: p=0.93) and trim and fill analysis.

Table 2. Studies investigating glutamate function in individuals at clinical or genetic high risk for psychosis Study Probands Controls Substance measured N Age (yrs., mean) At-risk group Antipsychotic (AP) treatment N Age (yrs., mean) Prefrontal cortex Byun et al58 20 21.8 CHR N=8 low-dose AP 20 22.0 Glx Natsubori et al59 24 21.7 CHR N=10 taking AP 26 22.3 Glx Egerton et al56 75 23.3 CHR N=3 taking AP 55 24.6 Glu, Glx de la Fuente-Sandoval et al60 23 21.4 CHR All naïve 24 20.7 Glx