A Meta-Analysis of Neural Correlates of Reward Anticipation in Individuals at Clinical Risk for Schizophrenia

Abstract

Background

Aberrant striatal responses to reward anticipation have been observed in schizophrenia. However, it is unclear whether these dysfunctions predate the onset of psychosis and whether reward anticipation is impaired in individuals at clinical high risk for schizophrenia (CHR).

Methods

To examine the neural correlates of monetary anticipation in the prodromal phase of schizophrenia, we performed a whole-brain meta-analysis of 13 functional neuroimaging studies that compared reward anticipation signals between CHR individuals and healthy controls (HC). Three databases (PubMed, Web of Science, and ScienceDirect) were systematically searched from January 1, 2000, to May 1, 2022.

Results

Thirteen whole-brain functional magnetic resonance imaging studies including 318 CHR individuals and 426 HC were identified through comprehensive literature searches. Relative to HC, CHR individuals showed increased brain responses in the medial prefrontal cortex and anterior cingulate cortex and decreased activation in the mesolimbic circuit, including the putamen, parahippocampal gyrus, insula, cerebellum, and supramarginal gyrus, during reward anticipation.

Conclusions

Our findings in the CHR group confirmed the existence of abnormal motivational-related activation during reward anticipation, thus demonstrating the pathophysiological characteristics of the risk populations. These results have the potential to lead to the early identification and more accurate prediction of subsequent psychosis as well as a deeper understanding of the neurobiology of high-risk state of psychotic disorder.

Significance Statement.

Although reward-related brain activity abnormalities have been observed in schizophrenia, findings of reward anticipation in individuals at clinical high risk for schizophrenia (CHR) are still unclear. We performed a whole-brain meta-analysis based on functional magnetic resonance imaging (fMRI) studies to elucidate the neurobiological basis underlying reward anticipation between CHR individuals and healthy controls. In our meta-analysis, we found that CHR individuals showed increased anticipation-related activation in the medial prefrontal cortex and anterior cingulate cortex as well as decreased activation in the mesolimbic circuit, including the putamen, parahippocampal gyrus, insula, cerebellum, and supramarginal gyrus. These results provide further support for the aberrant salience hypothesis in at-risk populations, help us understand the neurobiology of the high risk of psychotic disorder, and could have implications for early identification and early intervention.

INTRODUCTION

Schizophrenia (SZ) is a developmental disorder that begins with a prodromal period of altered functioning or symptomatology before the onset of threshold psychosis (McCutcheon et al., 2020). It has been recognized that the behavioral and neuropsychological manifestations in the early phases of SZ are different from those of late symptoms and chronic illness stage (Insel, 2009), thus highlighting the importance of exploring the neural correlates of reward processing in the prodromal phase (Lieberman et al., 2019). Moreover, the recruitment and evaluation of individuals at clinical high risk for SZ (CHR) could overcome confounding factors that are encountered when studying patients with SZ, such as antipsychotic medications and illness duration (Li et al., 2016a). It is therefore important to investigate the psychopathological mechanisms among CHR individuals that could highlight early vulnerability factors.

“Clinical high risk” individuals have been defined as individuals with attenuated psychotic symptoms, brief limited intermittent psychotic symptoms, or a recent decline in functioning (Yung and McGorry, 1996; Yung et al., 2003). These prodromal syndromes can generally be assessed by several structured instruments [e.g., the Structured Interview for Prodromal Syndromes (Miller et al., 2003) and the Comprehensive Assessment of At-Risk Mental States (Yung et al., 2005)]. However, detecting psychosis risk solely based on the symptom characteristics of CHR individuals yields low efficiency. For example, relying on subtle symptoms in a CHR group led to a correct 24-month disease prediction in only 29% (Fusar-Poli et al., 2012). Fortunately, there are early neurobiological abnormalities apparent long before behavioral and subclinical symptoms emerge, such as abnormal laminar cortical organization, genetic alterations, and structural and functional magnetic resonance imaging (fMRI) abnormalities (Borgwardt et al., 2007; Fusar-Poli, et al., 2011a). A recent study used a support vector machine approach to analyze structural and fMRI data and effectively identified CHR individuals with a sensitivity of 68% and a specificity of 76%, resulting in an accuracy of 72% (Valli et al., 2016). Therefore, it is essential to examine potential neurobiological markers in CHR individuals that can shed light on early altered neural processes prior to significant psychopathologies, as found in patients with psychiatric disorders.

Impaired anticipation of future rewards might play an important role in understanding psychotic symptoms in CHR individuals. Studies measuring monetarily incentivized responses have found that CHR individuals are more likely to attend to irrelevant stimulus features, which is interpreted as an aberrant motivational salience attribution (Schmidt et al., 2017). It is hypothesized that chaotic firing of dopaminergic neurons projecting to limbic and cortical areas mediates the inadequate attribution of salience to irrelevant events, which in turn results in positive symptoms (Kapur et al., 2005). Wotruba and colleagues reported a significant correlation between positive symptoms and the anticipation signal in the ventral striatum (VS) and the insula in CHR individuals (Wotruba et al., 2014). In addition, negative symptoms are thought to be associated with the transition to psychosis (Piskulic et al., 2012). In a study that examined reward processing among the CHR population, individuals with greater negative symptoms showed greater BOLD (blood oxygen level dependent) signals in the ventromedial prefrontal cortex (Millman et al., 2020). These studies provide some evidence that the CHR state may be associated with alterations in reward anticipation. Therefore, studying reward anticipation in the at-risk period may help us understand aberrant salience in CHR individuals.

Although previous studies have identified several likely neural substrates for monetary anticipation in CHR individuals to date, the results are inconclusive. Several fMRI experiments have demonstrated abnormal brain activity in mesocorticolimbic circuitry during anticipation of reward-capturing “motivational salience” in the CHR population (Smieskova et al., 2015; Michielse et al., 2019). Alterations in motivational salience processing in the VS and cortical regions (such as the prefrontal cortex and cingulate cortex) have been described in prepsychotic individuals (Bourque et al., 2017; Schmidt et al., 2017). However, findings from these studies have been relatively mixed. Some studies have found that CHR individuals exhibit greater activations in the cingulate cortex (Li et al., 2016b; Bourque et al., 2017), insula (Wilson et al., 2019), and superior frontal gyrus (Wotruba et al., 2014). In contrast, other studies have reported that CHR individuals show deactivations in the posterior cingulate cortex and postcentral gyrus (Yan et al., 2016) during anticipation. Such inconsistencies may be due to multiple factors, including the heterogeneity of the employed reward paradigm, the difference in clinical features, and the data processing strategy.

Consequently, a recent meta-analysis attempted to address the question of the neural bases of reward signaling in SZ from the perspective of distinct stages of reward processing (Radua et al., 2015). However, the sample of this meta-analysis included both CHR and SZ, resulting in unclear neurobiological mechanisms underlying CHR. In addition, this meta-analysis focused only on the VS instead of the whole brain. To map the potentially distinctive neural profiles associated with CHR and identify whether anticipation-related dysfunction predates the onset of SZ, it is necessary to investigate the biological mechanisms underpinning CHR during reward anticipation.

Therefore, to address these doubts and inconsistencies and to improve current knowledge of reward impairments in CHR individuals, the current meta-analysis systematically synthesized the available evidence regarding whole-brain neural correlates of reward anticipation in CHR individuals and identified potential neuroimaging markers of SZ prodrome. The monetary incentive delay (MID) task (Knutson et al., 2001) and salience attribution test (SAT) (Roiser et al., 2009) are operant conditioning paradigms to study reward anticipation in which participants learn to respond to a discriminative stimulus by pushing a button, thereby creating a distinct state of reward anticipation (Rademacher et al., 2010). We further examined the influence of demographic and clinical factors that contribute to reward-related brain responses with meta-regression analysis. Based on the literature reviewed above, we expected to find abnormal brain activation in the mesocorticolimbic circuitry in CHR individuals compared with healthy controls (HC) during anticipation.

METHODS

Search Strategy

A comprehensive literature search of whole-brain fMRI studies comparing CHR individuals and HC was carried out in the PubMed (www.pubmed.org), Web of Science (www.webofknowledge.com), and ScienceDirect (www.sciencedirect.com) databases from January 1, 2000, to May 1, 2022. The search terms used were (1) “schizophrenia” OR “schizophrenic” OR “schizoaffective” OR “Schizotypy” OR “psychosis” OR “psychotic”; (2) “high risk” OR “ultra-high risk” OR “at-risk mental state” OR “clinical risk” OR “basic symptoms” OR “psychotic-like experiences” OR “prodromal phase”; (3) “functional magnetic resonance imaging” OR “fMRI” OR “neuroimaging”; and (4) “monetary incentive delay task” OR “MID” OR “Salience Attribution Task” OR “SAT” OR “reward anticipation.” To increase the chances of identifying additional, relevant studies, the reference lists of the included articles and relevant review papers were manually searched. For studies without available coordinates at the whole-brain level, we asked the authors whether they could provide such information. If the authors did not respond, we excluded their studies if we could not obtain coordinate information and included their studies if we could not obtain some clinical data (e.g., psychiatric symptom scale score) but could obtain coordinate information. The flow diagram of the inclusion and exclusion process of the selected articles is described in Figure 1.

Figure 1.

Flow diagram of the inclusion and exclusion process of selected articles. Of the 439 articles initially identified, a total of 13 studies were enrolled for the reward anticipation meta-analysis. CHR, clinical high risk for schizophrenia; fMRI, functional magnetic resonance imaging; HC, healthy controls; MID, monetary incentive delay; ROI, region of interest; SAT, salience attribution test.

Studies were eligible if they met the following criteria: (1) included CHR individuals with attenuated psychotic symptoms, brief limited intermittent psychotic symptoms, or a recent decline in functioning (Yung and McGorry, 1996; Yung et al., 2003); (2) reported neuroimaging findings allowing a comparison of CHR and HC groups; (3) investigated fMRI brain responses during reward anticipation; and (4) reported significant results as 3D coordinates in either the Talairach or the Montreal Neurological Institute space based on a whole­brain analysis.

The exclusion criteria included the following: (1) publications were case reports, book chapters, reviews, or meta-analyses; (2) individuals were clinically diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders or International Statistical Classification of Diseases and Related Health Problems; (3) absence of a control group; and (4) studies reported only region of interest findings. Four fMRI studies including unaffected relatives/siblings of SZ patients were thus not included in our meta-analysis (Grimm et al., 2014; de Leeuw et al., 2015; Hanssen et al., 2015; Li et al., 2018). For 1 study where multiple independent CHR samples were compared with HC, the appropriate coordinates were included as separate datasets (Yan et al., 2016).

Data Extraction

The following data were recorded from each article: sample sizes, mean age, percentage of males, ethnicity, education level, income, collection site, severity of positive and negative symptoms, task performance (including the reaction time [RT], RT acceleration [mean RT differences between gain/loss and neutral conditions], hit rate, and total amount gain), and several MRI methodological parameters (including the MRI field strengths, imaging data analysis software, and the statistical threshold).

To facilitate the comparison of different scales used in different studies, a common variable of positive/negative symptoms was created. First, we extracted the mean and SD of the positive/negative symptom dimension from the included articles as measured by the following rating scales: the Community Assessment of Psychic Experience, Chapman Social Anhedonia Scale, Revised Chapman Social Anhedonia Scale, Revised Chapman Physical Anhedonia Scale, Structured Interview for Psychosis-risk Syndromes, Positive and Negative Syndrome Scale, Comprehensive Assessment of At-Risk Mental States interview, and Structured Interview for Prodromal Syndromes. For each measure of symptoms, we then rescaled each sample mean score according to the highest obtainable score on each particular instrument. For samples with more than 1 measure reported, the average value was calculated and recorded.

This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Moher et al., 2009). A 9-point checklist was used to evaluate study rigorousness via an objective description of the individual studies (supplementary Table 3 for the full criteria). The literature search, quality evaluation, and data extraction were independently performed by 2 investigators. Disagreements were resolved by discussion and, when necessary, by consulting a third member of the meta-analysis team. Kappa statistics showed a high agreement between reviewers for study selection (k = 0.78, P < .01). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist and assessments of study quality are provided in the supplementary Materials.

Meta-Analysis of Relevant Studies

Reward-related activation differences during reward anticipation between CHR individuals and HC were analyzed using seed-based d mapping (SDM) software (version 5.15, https://www.sdmproject.com), which combined peak coordinates and statistical parametric maps, and used standard effect size and variance-based meta-analytic calculations. This method has been widely applied in previous meta-analyses of structural and fMRI studies (Radua et al., 2010; Zeng et al., 2021; Yang et al., 2022). In the SDM analysis, studies reporting no significant group differences can be included, and both positive and negative differences can be reconstructed in the same map. Therefore, it is an optimal method for comparing CHR individuals with HC without biasing the results (Ferreira and Busatto, 2010; Radua et al., 2012a).

The SDM methods and steps have been described in detail elsewhere (Radua and Mataix-Cols, 2009). First, peak coordinates and corresponding statistics of task-evoked brain activation differences between CHR individuals and HC were extracted from the included articles, and maps of statistics other than t values, such as Z scores and P values, were converted in advance into t values by online converters. Second, standard Montreal Neurological Institute maps of the differences in anticipation-related activation were created by means of a Gaussian kernel. Third, the mean map was calculated and weighted by the squared root of the sample size of each study. The default kernel size and statistical thresholds were as follows: full width at half maximum = 20 mm, P = .005, peak height threshold = 1, and extent threshold = 10 (Radua et al., 2012a, 2012b).

In addition, to examine the effect of potential confounding variables and moderators on abnormal task-evoked activation as well as to assess the robustness of the findings, complementary analyses, including jackknife sensitivity analyses, heterogeneity analyses, subgroup analyses, and meta-regression analyses, were performed. We performed meta-regression analyses with a strict threshold of P < .0005 and considered only the brain regions identified in the main analysis of brain activation differences between the CHR and HC groups during reward anticipation (Radua et al., 2012a, 2015).

RESULTS

Included Studies and Sample Characteristics

Fourteen datasets from 13 whole-brain fMRI studies with 318 CHR individuals (mean age = 23.30 years; percentage of males = 55.00%) and 426 HC (mean age = 23.24 years; percentage of males = 59.00%) were included during the reward anticipation phase (Juckel et al., 2012; Roiser et al., 2013; Wotruba et al., 2014; Smieskova et al., 2015; Li et al., 2016b; Yan et al., 2016; Bourque et al., 2017; Schmidt et al., 2017; Winton-Brown et al., 2017; Michielse et al., 2019; Wilson et al., 2019; Millman et al., 2020; Uldall et al., 2020). No significant difference was found for age (t = 0.025, P = .659) or sex (χ² = 1.626, P = .202) between the CHR and HC groups. The participant demographics and imaging characteristics of the included articles are provided in Table 1.

Table 1.

Demographic and Clinical Characteristics of the Studies Included in the Meta-Analysis^a

Study	Clinical high risk for schizophrenia		Healthy controls		Imaging characteristics
	No. (male)	Mean age, y	No. (male)	Mean age, y	Tesla	SPM	Threshold
Anticipation
Bourque et al., 2017	27 (9)	14.26	135(47)	14.35	3.0 T	Y	Corrected
Juckel et al., 2012	13 (11)	25.46	13 (11)	25.69	1.5 T	Y	Uncorrected
Li et al., 2016b	15 (9)	20.13	19 (11)	37.70	3.0 T	Y	Uncorrected
Michielse et al., 2019	47 (7)	21.50	40 (7)	37.10	3.0 T	N	Corrected
Millman et al., 2020	22 (10)	17.26	19 (12)	18.03	3.0 T	N	Corrected
Roiser et al., 2013	18 (7)	25.70	18 (10)	26.25	3.0 T	Y	Uncorrected
Schmidt et al., 2017	23 (12)	24.38	13 (10)	24.38	3.0 T	Y	Uncorrected
Smieskova et al., 2015	34 (26)	24.35	19 (10)	26.42	3.0 T	Y	Corrected
Uldall et al., 2020	19 (0)	47.00	31 (0)	38.00	3.0 T	N	Corrected
Wilson et al., 2019	17 (7)	24.10	19 (11)	23.90	3.0 T	Y	Corrected
Winton et al., 2017	29 (13)	21.27	32 (18)	23.69	3.0 T	Y	Uncorrected
Wotruba et al., 2014	21 (15)	25.10	24 (13)	30.10	3.0 T	Y	Uncorrected
Yan et al., 2016 dataset1	15 (8)	19.33	22 (11)	19.78	3.0 T	Y	Corrected
Yan et al., 2016 dataset2	18 (9)	19.28	22 (11)	19.78	3.0 T	Y	Corrected

^aAbbreviations: N, No; No., number; SPM, statistical parametric mapping; Y, Yes.

Main Meta-Analysis

In the pooled meta-analysis of reward anticipation, relative to the HC, the CHR individuals exhibited higher activations in the medial prefrontal cortex (mPFC) and anterior cingulate cortex (ACC) and lower activations in the left cerebellum (extending to the parahippocampal gyrus), right supramarginal gyrus (SMG), and right insula (extending to the putamen) in response to monetary stimuli (Table 2; Figure 2).

Table 2.

Results of the Anticipation Meta-Analysis for Brain Activation Difference Between CHR Individuals and HC^a

Brain regions	MNI coordinates x, y, z	SDM value	P value	No. of voxels	Breakdown	Egger’s test (P value)	Heterogeneity
CHR > HC
mPFC and ACC	4, 40, 28	1.330	.000464499	1019	Left superior frontal gyrus, medial, BA 32	.111	No
					Right anterior cingulate/paracingulate gyri, BA 32
					Right superior frontal gyrus, medial, BA 9
					Left anterior cingulate/paracingulate gyri, BA 32
					Left superior frontal gyrus, medial
					Right median cingulate/paracingulate gyri, BA 32
					Left superior frontal gyrus, medial, BA 9
					Left anterior cingulate/paracingulate gyri, BA 24
					Right superior frontal gyrus, medial, BA 10
					Right superior frontal gyrus, medial, BA 32
					Left superior frontal gyrus, medial, BA 8
					Left anterior cingulate/paracingulate gyri
					Right anterior cingulate/paracingulate gyri, BA 24
					Left median cingulate/paracingulate gyri, BA 24
					Right median cingulate/paracingulate gyri, BA 24
					Right superior frontal gyrus, medial, BA 8
					Left superior frontal gyrus, medial, BA 10
					Right superior frontal gyrus, medial
					Left superior frontal gyrus, medial, BA 24
					Right anterior cingulate/paracingulate gyri
					Left median cingulate/paracingulate gyri, BA 32
					Left median cingulate/paracingulate gyri
					Right median cingulate/paracingulate gyri, BA 9
					Right median cingulate/paracingulate gyri
CHR < HC
Left cerebellum and parahippocampal gyrus	−22, −40, −18	−2.051	.000005186	1633	Left fusiform gyrus, BA 37	.428	No
					Left cerebellum, hemispheric lobule VI, BA 37
					Left cerebellum, hemispheric lobule IV/V, BA 37
					Left fusiform gyrus, BA 30
					Left cerebellum, hemispheric lobule IV/V, BA 30
					Left median network, cingulum
					Middle cerebellar peduncles
					Left parahippocampal gyrus, BA 30
					Left inferior network, inferior longitudinal fasciculus
					Left cerebellum, hemispheric lobule VI
					Left fusiform gyrus, BA 20
					Left cerebellum, crus I
					Left parahippocampal gyrus, BA 37
					Left lingual gyrus, BA 30
					Left lingual gyrus, BA 37
					Left cerebellum, crus I, BA 37
					Left cerebellum, hemispheric lobule IV/V
					Left lingual gyrus, BA 27
					Left lingual gyrus
					Left parahippocampal gyrus
					Left cerebellum, hemispheric lobule IV/V, BA 20
					Left fusiform gyrus
					Left cerebellum, hemispheric lobule VI, BA 20
					Left parahippocampal gyrus, BA 20
					Left lingual gyrus, BA 19
					Left cerebellum, hemispheric lobule IV/V, BA 27
					Left cerebellum, hemispheric lobule VIIB
Right SMG	42, −40, 42	−1.775	.000159979	363	Right supramarginal gyrus, BA 40	.175	No
					Right inferior parietal (excluding supramarginal and angular) gyri, BA 40
					Right supramarginal gyrus, BA 2
					Right inferior parietal (excluding supramarginal and angular) gyri, BA 2
					Right postcentral gyrus, BA 2
					Right postcentral gyrus, BA 3
Right insula and putamen	36, 20, 0	−1.485	.001398563	246	Right insula, BA 48	.198	No
					Right insula, BA 47
					Right lenticular nucleus, putamen, BA 48
					Right inferior network, inferior fronto-occipital fasciculus
					Right insula
					Right lenticular nucleus, putamen, BA 47
					Right inferior frontal gyrus, opercular part, BA 48

^aAbbreviations: ACC, anterior cingulate cortex; BA, Brodmann area; CHR, clinical high risk for schizophrenia; HC, healthy controls; MNI, Montreal Neurological Institute; mPFC, medial prefrontal cortex; SDM, seed-based d mapping; SMG, supramarginal gyrus. Results were threshold at P = .005, peak height threshold of 1, extent threshold of 10.

Figure 2.

Task-evoked activation differences during reward anticipation between clinical high risk for schizophrenia (CHR) individuals and healthy controls (HC) in the meta-analysis. Brain regions that showed significant differences in anticipation-evoked activity in CHR individuals relative to HC. Red indicates regions that showed hyperactivity in CHR individuals relative to HC, and blue indicates regions that showed hypoactivity in patients relative to HC. The color scale represents probability values from statistical permutation testing (Z values). For the main analysis of the anticipation stage, CHR individuals showed increased activations in the medial prefrontal cortex (mPFC) and anterior cingulate cortex (ACC) but decreased activations in the left cerebellum (extending to the parahippocampal gyrus), right supramarginal gyrus (SMG), and right insula (extending to the putamen).

Quality Assessment and Sensitivity Analysis

The mean score for quality assessment was 8.5 out of 9 points (range = 8–9 points), as shown in supplementary Table 4. In the whole-brain jackknife sensitivity analysis, the decreased activation in the left cerebellum was preserved in all included datasets. The hyperactivations in the mPFC and ACC remained significant in 13 datasets and failed to emerge in 1 dataset (Bourque et al., 2017). The hypoactivity in the right SMG and right insula remained significant in 13 datasets and failed to emerge in 1 dataset (Michielse et al., 2019) (Table 3).

Table 3.

Subgroup and Jackknife Sensitivity Analyses in Brain Activation Difference Between CHR Individuals and HC During Reward Anticipation Stage^a

Datasets	Activation	Deactivation
	mPFC and ACC	Left cerebellum	Right SMG	Right insula
Subgroup analysis
Datasets including only adult CHR (n = 12)	N	Y	Y	Y
Datasets including CHR without comorbidity (n = 12)	Y	Y	Y	Y
Datasets using the MID paradigm (n = 11)	Y	Y	Y	Y
Datasets using a 3-T MRI scanner (n = 13)	Y	Y	Y	Y
Sensitivity analysis
Bourque et al., 2017	N	Y	Y	Y
Juckel et al., 2012	Y	Y	Y	Y
Li et al., 2016b	Y	Y	Y	Y
Michielse et al., 2019	Y	Y	N	N
Millman et al., 2020	Y	Y	Y	Y
Roiser et al., 2013	Y	Y	Y	Y
Schmidt et al., 2017	Y	Y	Y	Y
Smieskova et al., 2015	Y	Y	Y	Y
Uldall et al., 2020	Y	Y	Y	Y
Wilson et al., 2019	Y	Y	Y	Y
Winton et al., 2017	Y	Y	Y	Y
Wotruba et al., 2014	Y	Y	Y	Y
Yan et al., 2016 dataset1	Y	Y	Y	Y
Yan et al., 2016 dataset2	Y	Y	Y	Y

^aAbbreviations: ACC, anterior cingulate cortex; CHR, clinical high risk for schizophrenia; MID, monetary incentive delay; mPFC, medial prefrontal cortex; MRI, magnetic resonance imaging; N, brain region is no longer significantly increased/decreased in the subgroup analysis or in the Jackknife analysis; SMG, supramarginal gyrus; Y, brain region remains significantly increased/decreased in the subgroup analysis or in the Jackknife analysis.

Heterogeneity and Bias Analyses

As shown in Table 2, there was no significant between-study heterogeneity in any of the peak coordinates. Additionally, Egger’s test showed no “small study” bias (P > .05) for any of the clusters reported as significant (Sterne et al., 2000). See Table 2 for detailed results.

Subgroup Analyses

To control for any possible differences observed between studies, the subgroup analyses were repeated to include only those studies that were methodologically homogenous. Therefore, we conducted subgroup analyses for those datasets including only adult CHR individuals, for those including CHR individuals without comorbidities, for those using the MID paradigm, and for those using a 3-T MRI scanner. The main results of the meta-analysis remained largely unchanged when the analyses were repeated in all subgroup analyses, except for the adult subgroup analysis. For the adult subgroup analysis, the hyperactivation in the mPFC and ACC in the pooled main analysis did not remain significant. The hypoactivation in the left cerebellum, right SMG, and right insula in the adult subgroup analysis was consistent with that in the pooled main analysis. The detailed results are shown in Table 3.

Meta-Regression Analyses

Meta-regression analysis revealed that the mean age, percentage of males, positive and negative symptom severity, RT, RT acceleration, and hit rate were not linearly associated with task-related activation during reward anticipation. The effect of other variables, including ethnicity, education level, income, and total amount gain on brain activation, could not be explored due to insufficient data.

DISCUSSION

In the current whole-brain meta-analysis of brain activation from published reward-related fMRI studies, we examined the neural basis of monetary anticipation in CHR individuals. Our whole-brain meta-analysis provided evidence that CHR individuals showed increased anticipation-related activations in the mPFC and ACC cortical regions but decreased activation in the mesolimbic circuit, including the right putamen, left parahippocampal gyrus, left cerebellum, right SMG, and right insula. These findings in the CHR group confirmed brain reward-related activation during reward anticipation while performing the reward task, which might indicate a broader vulnerability to motivational-related symptoms in persons at risk for psychosis.

In the present study, we observed hyperactivity in response to monetary reward in the mPFC and ACC during reward anticipation in CHR individuals. Consistent with our findings, previous studies have also identified exaggerated prefrontal activity during reward processing in populations at high familial risk for psychosis (Myin-Germeys and van Os, 2007; van Buuren et al., 2011; Li et al., 2012; de Leeuw et al., 2015), ultra-high risk populations (Seiferth et al., 2009; Wotruba et al., 2014), and healthy people with SZ risk–associated genes (Radulescu et al., 2013). In particular, the PFC dysfunction found in individuals at ultra-high risk for psychosis is associated with an increased risk for transition to psychosis (Allen et al., 2012). The frontal lobe regions, including the mPFC and ACC, have been reported to play a role in learning reward associations, selecting reward goals, choosing actions to obtain reward, and monitoring the potential value of switching to alternative courses of action (Rushworth et al., 2011). In detail, the ACC is important for value integration to guide effort-based decisions in humans and reflect both anticipated effort and anticipated reward (Croxson et al., 2009; Klein-Flugge et al., 2016). The frontal medial regions of the prefrontal cortex, having reciprocal connectivity with several neuromodulatory systems, are key components of neural circuits involved in detecting the motivational significance of external stimuli (Tzschentke, 2000). These results support the “aberrant salience” notion that psychotic symptoms may arise from inappropriately attributing salience to irrelevant stimuli. The increased responses of the mPFC/ACC in CHR individuals may be due to oversensitivity to the salience of external stimuli and reward monitoring (Weissman et al., 2006). Another possible explanation is that the prefrontal cortex exerts top-down control over the interaction between midbrain dopaminergic systems and the striatum, and increased activity in the PFC may inhibit reward-related behavior (Ferenczi et al., 2016). The hyperactivation in these regions was considered to be a compensatory mechanism for structural and functional deficits of the mesolimbic system (Quintana et al., 2003; Harvey et al., 2007). Thus, our findings of greater responses in the mPFC and cingulate cortex may suggest increased reward expectation, salience processing, and cognitive responses to the task or could be a compensatory mechanism to counteract the changes in the mesolimbic regions.

In contrast, we found that CHR individuals exhibited decreased activation in the mesolimbic system, including the putamen, parahippocampal gyrus, insula, SMG, and cerebellum, during reward anticipation. A considerable number of fMRI studies have revealed a correlation between activity in the ventral and dorsal striatum and reward prediction error in regard to both primary rewards (McClure et al., 2003) and monetary rewards (Breiter et al., 2001). Individuals with CHR also exhibited reduced striatal activity during reward anticipation compared with controls, which has been widely reported in previous studies (Papanastasiou et al., 2018). Volumes in the putamen, whose output is modulated by dopaminergic inputs, are known to be abnormal in SZ (Kumakura et al., 2007; Perez-Costas et al., 2010) and in individuals at high risk for psychosis (Bloemen et al., 2013; Egerton et al., 2013). Given the key role of dorsal striatal regions in anticipatory or motivational components of reward-seeking behavior, the smaller volume and reduced activity in the putamen may be a neural substrate for anticipatory anhedonia and an indication of an earlier high-risk stage (Wu et al., 2019). The insula is widely connected to many areas of the cortex and limbic system involved in reward and decision-making, such as the frontal pole, dorsolateral prefrontal cortex, ACC, and striatum (Butti and Hof, 2010). Human and animal studies have implicated the insula in the assessment of stimuli and the allocation of appropriate arousal for motor preparation and selective attention, 2 processes that play key roles in the MID task (Eckert et al., 2009). These processes may be inhibited in CHR individuals, thus leading to decreased activation in the insula during the anticipation of reward (Michielse et al., 2019). Additionally, the hippocampal-parahippocampal region showed both reduced activity and reduced coupling during declarative memory with fMRI in high-risk populations, thus reflecting abnormal memory function in at-risk individuals (Rasetti et al., 2014). The current findings highlight mesolimbic circuit dysfunction in the pursuit of reward, which contributes to motivational impairment in the risk state of psychosis.

Our meta-analysis also revealed that blunted brain activation was found in the cerebellum in CHR individuals compared with HC. By sending direct excitatory projections to the ventral tegmental area, the cerebellum can modulate brain-wide reward circuitry (Carta et al., 2019). Recent evidence has indicated that granule cells in the cerebellum encode salient predictive information about the future and signal reward expectations (Wagner and Luo, 2020). An accumulating body of work has also highlighted cerebellar dysfunction in psychosis-risk individuals (Bernard et al., 2014, 2017). Abnormal cerebellar activation in CHR may imply difficulties in learning and achieving automaticity in task execution. Therefore, our finding of reduced cerebellar activation could be a neural substrate for reward anticipation-related deficits in CHR individuals.

The findings of decreased activation in the right SMG in the CHR group with respect to the controls have previously been reported in reward anticipation (Michielse et al., 2019). Anatomically, the SMG is one of the largest areas of the inferior parietal lobule, a key region that receives and recognizes visual information activated by visual cortices (Stoeckel et al., 2009). SMG abnormalities may reflect deficits in recognizing reward expected visual cues among CHR individuals. Similarly, an fMRI study on CHR individuals also reported decreased activation of the SMG during the N-Back task (Fusar-Poli, et al., 2011b). Compared with the prefrontal cortex or cingulate, the SMG appears to be relatively overlooked, although it has been widely considered a crucial component of the impaired frontal–limbic–temporal–parietal network involved in the SZ development process (Torrey, 2007). Here, we provided evidence for this finding by showing decreased SMG activity in CHR individuals compared with that in controls.

Notably, the subgroup analyses showed that the increased brain activity in the mPFC/ACC did not survive in the adult CHR subgroup analyses. The critical role of the mPFC and ACC in reward prediction, reinforcement learning, and prompting effortful behavior has been widely documented (Vassena et al., 2014). In line with previous studies, we found that hyperactivity in these regions appeared to be more pronounced in younger participants (Otsuka et al., 2006; Gee et al., 2012). Uniquely adolescent characteristics such as risk-taking behavior and impulsivity are typical of this age and might substantially contribute to both psychological and neurobiological vulnerability in those groups (Arnett, 1992; Laviola et al., 2003). Some evidence supported that adolescent risk-taking behavior might be driven by disproportionately increased activation of the motivational circuit by potential gain cues relative to the influence of inhibitory circuits (Chambers et al., 2003; Bjork et al., 2004). Together, our subgroup results suggested that adolescents exhibited more recruitment of the mPFC/ACC region while anticipating responses for gains than adults. Future studies could conduct longitudinal research on the at-risk individuals or directly compare adult CHR individuals with young CHR individuals to better observe the potential effect of neurodevelopmental mechanisms associated with reward processing.

Considering the significance of brain-behavior association, it is important to investigate relationships between brain measures and behavioral performance. However, in our current study, correlations between task performance (RT, RT acceleration, and hit rate) and monetary-induced activities were not significant. Many previous studies and reviews have disentangled the relationships between MID variables and symptom deficits, and there is tentative but mixed evidence for the association. At the behavioral level, the HC group appeared to win a higher cumulative total of money (Michielse et al., 2019) and show a faster response to reward (Uldall et al., 2020) or a higher hit rate (Winton-Brown et al., 2017) compared with high-risk individuals. Moreover, most studies have suggested an association between monetary reward deficits and psychotic symptoms (Wotruba et al., 2014; Kirschner et al., 2018) such that poor MID is associated with more pronounced psychotic symptoms. However, several studies have also reported that RT (Demidenko et al., 2020), monetary win (Grimm et al., 2014), and hit rate (van Leeuwen et al., 2019) did not differ in high-risk individuals. At the neural level, 1 study revealed that insular activation was negatively correlated with RT acceleration (as indexed by the mean RT difference between neutral and salience conditions) during the presentation of salient stimuli compared with that during the presentation of neutral stimuli (Wilson et al., 2019). This may imply that insular activation in CHR individuals was associated with impaired discrimination of salience and could be linked to aberrant motivational salience processing in CHR individuals. Another study found that the risk population showed increased ACC activity with a gain in reaction speed but at the expense of correctness during a choice reaction task (Mulert et al., 2003). Considering the function of the PFC/ACC in mediating the cognitive control of response selection (Badre and Wagner, 2004), enhanced neural activation may reflect more focus on, more effort required for, or improved efficiency in modulating, coordinating, and integrating goal-relevant information within the task demand. The null results in the current meta-regression analysis can be attributed to multiple sources, such as clinical features, cue features of the employed paradigm, and data processing strategy. Clearly, whether incentive salience–induced brain activity directly correlates with actual behavioral responses will be an important question for future study.

Focusing on common and distinct dysfunctional patterns of brain activation between CHR groups and SZ groups is of great significance to delineate neurobiological characteristics in different stages of psychosis. Although we did not make a direct comparison of SZ and CHR individuals in the current meta-analysis, compared with the results of recent meta-analyses focusing on SZ, we found that SZ and CHR individuals showed similar brain hypoactivation patterns in the mesolimbic circuit but different activation profiles in the PFC/ACC during reward anticipation (Leroy et al., 2020; Zeng et al., 2022). Importantly, SZ exhibited reduced anticipation-related activity in both ventral and dorsal striatal regions (Zeng et al., 2022), whereas the CHR group showed reduced activation only in the DS region. Hypoactivation in the dorsal and VS during the anticipation of monetary incentives was also observed in first-episode SZ patients (Schlagenhauf et al., 2009; Esslinger et al., 2012; Nielsen et al., 2012; Hanssen et al., 2015), chronic SZ patients (Juckel, et al., 2006a, 2006b), and their unaffected first-degree relatives (Grimm et al., 2014; de Leeuw et al., 2015). Our finding of only DS hypoactivation in CHR individuals is in contrast to findings of decreased activation in the VS in patients (Simon et al., 2015) and siblings of patients (Grimm et al., 2014) but is in line with the findings of 1 study in people with subclinical psychotic experiences (Michielse et al., 2019). DS dysfunction in the CHR group may suggest impairment in the construction of associations between monetary rewards and cues, thereby impeding salience attribution (Li et al., 2018). Contrary to the finding of decreased activations in the PFC/ACC in SZ patients (Zhang et al., 2016; Leroy et al., 2020), CHR individuals exhibited increased mPFC/ACC activation in our meta-analysis. This is consistent with results from a previous study that the high-risk sample exhibited enhanced activation in the superior frontal gyrus and the cingulate cortex (Wotruba et al., 2014). These results may provide evidence that PFC/ACC activity could be disturbed before the transition to psychosis, and this pattern is different from that of SZ patients. However, shared and unique functional brain activity patterns between psychotic patients and high-risk individuals need to be clarified in future studies.

Exploring the pathophysiology in at-risk individuals prior to psychosis onset is potentially important for early diagnosis and intervention. SZ is now considered to be a brain disorder and developmental disorder rather than only a psychological conflict or stable disease (Seidman and Mirsky, 2017). The onset of mental illness generally occurs early, but its diagnosis is delayed (Wang et al., 2005). One of the possible interpretations of this phenomenon is that clinical and neurobiological profiles in the early phases are unlike those in the later phases (Insel, 2009). In our current meta-analysis, our findings of distinct patterns of enhanced responses in the mPFC and ACC and lower reactivity in the mesolimbic circuit in CHR individuals may represent a putative biomarker in individuals at high risk for psychosis. These patterns may contribute to our understanding of abnormalities in the CHR population, thus offering great potential to identify reliable biomarkers of psychosis prodrome and diagnostic criteria for presyndromal SZ. Furthermore, by studying individuals with a risk of developing this pathology, CHR research will help guide the development of an evidence-based early intervention for reducing symptoms or preventing the transition to psychosis. Such research may also help to improve the prediction of subsequent psychosis. However, translating these findings into clinical practice is a challenge, and further research is needed to produce clinically validated biomarkers and proven therapeutic strategies for psychotic disorders.

Limitations

Some limitations should be considered when assessing the impact of these findings. First, instead of focusing on cross-sectional studies, longitudinally following individuals to identify functional alterations would have been ideal. A longitudinal study could be performed to determine whether the altered trajectories of brain activation during reward tasks precede psychosis onset and predict the transition to psychosis in a high-risk group. Second, the correlations between reward-related activation and task performance variable (including total amount gain) could not be investigated because of insufficient data. Third, this meta-analysis is unable to indicate whether the reported monetary response is reflective of psychosis-specific risk or general vulnerability for psychopathology because all the included studies relied on healthy volunteers as a reference point and there were too few studies that recorded help-seeking information for controls. Future research will be needed to recruit help-seeking controls to elucidate psychosis-specific vulnerability mechanisms. Fourth, we included only studies that used monetary stimuli to elicit brain responses. Because clinical and CHR individuals have extensive cognitive impairment, it is not clear whether other forms of reinforcing stimuli (e.g., social and emotional stimuli) would lead to the same results. Furthermore, some reward paradigms, such as the MID task, are relatively simple, so it is difficult to identify the reinforcement learning deficits of individuals.

CONCLUSION

In summary, we demonstrated that CHR individuals showed increased activations in the prefrontal regions involving the mPFC and ACC and decreased activation in the mesolimbic circuitry during reward anticipation. Our results further support the notion that the CHR state is associated with abnormalities in reward-related brain function, thus enhancing the understanding of the pathophysiological characteristics in the CHR stage.

Supplementary Materials

Supplementary data are available at International Journal of Neuropsychopharmacology (IJNPPY) online.

Author Contributions

J. Zeng and X. Yang contributed to the study conception and design and supervised the study. J. Yan and L. You significantly contributed to the analysis and manuscript preparation. T. Liao, Y. Luo, and B. Chen helped perform the analysis with constructive discussions. X. Yang, J. Yan, and J. Zeng wrote the manuscript, which was reviewed by all authors and approved for publication.

Conflict of Interests

The authors declare no financial and non-financial conflicts of interest.

Data Availability

No new data were generated in support of this research.