A substantial body of evidence implicates dopaminergic dysfunction in the etiology of psychotic disorders (1–4). In particular, converging in vivo evidence from positron emission tomography (PET) and other molecular imaging studies indicates that psychotic disorders are associated with dysregulated presynaptic striatal dopaminergic function (for reviews, see references 1, 5, 6). It has thus been proposed that striatal hyperdopaminergia underlies the onset of psychosis, preceding illness onset and increasing further to lead to the development of frank psychosis (4, 7).

The onset of psychotic disorders is typically preceded by a prodromal phase, characterized by functional decline and attenuated psychotic symptoms—perceptual disturbances and paranoid ideas that are less severe than in frank psychosis (8, 9). Structured assessments and operationalized criteria have been developed for identifying the clinical features (termed an “at-risk mental state”) that are associated with an ultra-high risk of developing psychosis. Patients who meet these criteria are said to be at ultra-high risk for psychosis, as about one-third will develop a psychotic illness within 2 years (9, 10).

We previously reported that dopamine synthesis capacity is higher on average in ultra-high-risk patients than in healthy comparison subjects (11). However, at the time of that study, the clinical outcome of the study participants was yet to be determined. Because only a proportion of those at ultra-high risk will subsequently develop a psychotic disorder, the relationship between dopamine dysfunction and the subsequent onset of illness is still unclear. In the present study, we tested the hypothesis that dopamine synthesis capacity would be elevated in the group of ultra-high-risk patients who went on to develop psychosis relative to the healthy comparison group. A related prediction was that within this group, the severity of presenting symptoms would be directly related to dopamine synthesis capacity. Because schizotypal personality disorder is independently associated with presynaptic dopaminergic dysfunction and psychotic-like symptoms (12–15), we excluded ultra-high-risk patients with schizotypal personality disorder from the main analysis, but we present their data for comparison. As altered dopamine synthesis capacity may be localized to the associative subdivision of the striatum in psychotic disorders (11), we examined striatal subdivisions in addition to the striatum as a whole.

Participants

The study was approved by the participating hospitals' research ethics committees, and all participants gave written informed consent. Patients who met the Comprehensive Assessment of At-Risk Mental States (8) criteria for ultra-high risk of psychosis on the basis of attenuated psychotic symptoms (N=30; mean age=25.0 years [SD=4.1]; 17 of them [57%] male) were studied using [¹⁸F]6-fluoro-l-dopa ([¹⁸F]DOPA) PET at clinical presentation. We previously reported baseline data in a subgroup (N=24) of this ultra-high-risk sample (11), but at that early stage their subsequent clinical outcome was unknown. The present study includes follow-up data on all ultra-high-risk patients included in our original baseline study (11) plus follow-up data on six additional ultra-high-risk patients recruited after that study was published. A minimum of 3 years of follow-up was used, as the great majority of transitions to psychosis in ultra-high-risk patients occur within this interval (9, 10). Six of the ultra-high-risk patients met DSM-IV criteria for comorbid schizotypal personality disorder and were excluded from the analysis because this disorder has been independently associated with elevated striatal dopamine release (13). However, we present their data separately for comparison. The analysis thus includes the remaining 24 ultra-high-risk patients (15 of them [63%] male) and the matched healthy comparison subjects. All of these participants were antipsychotic naive.

Healthy comparison subjects (N=29; 20 of them [69%] male) were recruited contemporaneously from the same geographical area and sociodemographic background by local advertisement and by approaching the social contacts of ultra-high-risk patients (after first receiving written permission to do so). They were matched to the ultra-high-risk group on the basis of age (within 5 years). Parental socioeconomic status and ethnicity were determined using U.K. standard criteria (the five-class categorization; see www.ons.gov.uk/about-statistics/classifications/current/soc2010/soc2010-volume-3-ns-sec--rebased-on-soc2010--user-manual/index.html). Exclusion criteria for all participants were a history of a neurological or medical disorder (other than past minor self-limiting illnesses) or head injury; illicit drug or alcohol abuse or dependence; pregnancy; or contraindication to scanning. All participants were instructed to fast and to abstain from psychoactive substances, including caffeine, tobacco, and alcohol, for at least 12 hours before scanning. All participants received a urine drug screen 1 hour before scanning and clinical assessments to confirm that they had not recently taken any illicit substances. Additional exclusion criteria for comparison subjects were a family history of a psychotic disorder or a personal history of any axis I or II psychiatric disorder, assessed using the Structured Clinical Interview for DSM-IV (SCID; 16).

Clinical outcomes at follow-up were determined using the SCID. This assessment was conducted by an experienced psychiatrist who was independent of the study, was blind to the imaging data, and was trained in the use of the interview. All patients recruited into the study completed follow-up. Nine of the ultra-high-risk patients (38%) developed a psychotic disorder (the psychotic transition group) within the follow-up period. These disorders included schizophrenia (N=7), schizophreniform disorder (N=1), and bipolar I disorder with a psychotic manic episode (N=1). The remaining ultra-high-risk patients (N=15) have not developed a psychotic disorder (the nontransition group). None of the patients in the schizotypal personality disorder group have developed a psychotic disorder to date.

Clinical Measures

The following instruments were used to assess subjects at the time of the scan: the Comprehensive Assessment of At-Risk Mental States (8), the Positive and Negative Syndrome Scale (PANSS) (17), the Global Assessment of Functioning scale (GAF), and a structured questionnaire for assessing exposure to cigarettes and alcohol (adapted from the Cannabis Experiences Questionnaire [18]; available on request).

PET Scanning

Participants underwent [¹⁸F]DOPA PET imaging using an ECAT/EXACT3D PET scanner (Siemens/CTI, Knoxville, Tenn.). Participants received 150 mg of carbidopa and 400 mg of entacapone orally 1 hour before scanning to reduce the formation of radiolabeled metabolites. Participants were positioned with the orbitomeatal line parallel to the transaxial plane of the tomograph, and head position was marked and monitored throughout the scan via laser crosshairs and a camera. The 95-minute emission scan was preceded by a short transmission scan using a 150-MBq cesium-137 rotating point source to correct for attenuation and scatter. Thirty seconds after the start of the emission scan, a bolus intravenous injection of 150 MBq of [¹⁸F]DOPA was administered. Structural MRI was performed to exclude intracranial abnormalities.

Image Analysis

Region-of-interest analysis.

The region-of-interest analysis was carried out blind to group status (by O.D.H.) and comprised the whole striatum and the limbic, associative, and sensorimotor subdivisions of the striatum (illustrated in Figure S1 in the data supplement that accompanies the online edition of this article). The whole striatum and its subdivisions were defined using previously described criteria and combining the left and right sides (11, 19). The striatal subdivisions reflect the topographical arrangement of corticostriatal projections and the functional organization of the striatum (19, 20) (see the online data supplement).

Regions of interest were automatically placed on individual [¹⁸F]DOPA PET dynamic images without observer bias by using statistical parametric mapping software (SPM5; Wellcome Department of Cognitive Neurology, London) to normalize the region-of-interest map to each individual PET space (using the PET summation image). A Patlak graphical analysis was used to calculate influx constants (k_i^cer values, where k_i indicates it is the influx constant and ^cer denotes the reference region used). k_i^cer values were determined for the regions of interest relative to uptake in the reference region (21). The reference region was the cerebellum, as defined using the HamNet probabilistic brain atlas (22). Striatal [¹⁸F]DOPA k_i^cer values (denoted as Ki in some previous publications [11]) reflect the presynaptic synthesis and storage of striatal dopamine for release, respond to experimental manipulation of brain dopaminergic systems, and correlate with the striatal release of dopamine (11, 23).

Voxel-based analysis.

The region-of-interest analysis was complemented by an independent voxel-based analysis. Voxel-based k_i^cer parametric images of the brain were constructed from movement-corrected images using a wavelet-based Patlak approach that provides a higher signal-to-noise ratio than the original graphical procedure (24). The parametric image for each participant was then normalized into standard space using the participant's PET summation image and the [¹⁸F]DOPA template. Statistical parametric mapping was conducted using SPM5 and a striatal mask defined according to previously described criteria (19) to compare groups. Results are presented corrected for multiple comparisons using random field theory as applied in SPM5 (p<0.05, corrected at the family-wise error rate).

Statistical Analysis

The data were normally distributed as assessed using the Kolmogorov-Smirnov test apart from the substance use data. After confirming homogeneity of variance with Levene's test, analysis of variance was used to determine whether there was an effect of group (psychotic transition, nontransition, and healthy comparison groups) on striatal k_i^cer value and demographic and clinical variables for parametric variables, and the Kruskal-Wallis test was used for nonparametric variables. Where there were significant group effects, independent t tests using Bonferroni correction for multiple comparisons (four striatal regions across three groups) were used to determine whether mean k_i^cer values were significantly elevated in the psychotic transition group relative to the healthy comparison group in line with the main hypothesis, and to determine whether there were significant differences in mean k_i^cer values between the nontransition and comparison groups and between the psychotic transition and nontransition groups. The relationship between k_i^cer values and symptom scores was explored using Pearson's correlation coefficient, and the influence of individual data points was assessed using Cook's distance-centered leverage plots. Mean k_i^cer values in the schizotypal personality disorder group were compared with those in the psychotic transition and nontransition groups using independent t tests, but this analysis should be considered exploratory as we had no a priori hypotheses for this comparison. A two-tailed significance threshold of 0.05 was used throughout.

Demographic and Clinical Characteristics

There was no significant effect of group on mean age, parental socioeconomic status, cigarette or alcohol use, or radioactivity administered (psychotic transition group, mean=148.2 MBq [SD=5.2]; nontransition group, mean=148.8 MBq [SD=3.4]; healthy comparison group, mean=142.9 MBq [SD=14.0]). In ethnic composition, the psychotic transition group was 56% white, 33% black, and 11% other; the nontransition group was 53% white, 27% black, and 20% other; and the healthy comparison group was 52% white, 41% black, and 7% other. There was no significant difference in the clinical characteristics between the psychotic transition and nontransition groups at the time of imaging (Table 1).

Striatal Dopamine Synthesis Capacity

The region-of-interest analysis revealed that there was a significant effect of group on mean k_i^cer value in the whole striatum (Figure 1; F=5.8, df=2, 50, p=0.006) and in its associative subdivision (F=5.8, df=2, 50, p=0.006), but not in the limbic or sensorimotor subdivisions. The group effect remained significant after excluding the one patient in the psychotic transition group who had developed a non-schizophreniform psychosis (whole striatum: F=5.1, df=2, 49, p=0.01; associative striatum: F=4.8, df=2, 49, p=0.013).

Comparison Between the Psychotic Transition and Healthy Comparison Groups

The comparison between the psychotic transition and healthy comparison groups showed that after adjustment for multiple comparisons, mean k_i^cer values were significantly elevated in the psychotic transition group in the whole striatum (psychotic transition group, mean=0.0153/min [SD=0.0012]; comparison group, mean=0.0140/min [SD=0.0010]; p=0.004, corrected) and in its associative subdivision (psychotic transition group, mean=0.0149/min [SD=0.0011]; comparison group, mean=0.0136/min [SD=0.0010]; p=0.015, corrected). The effect size (Cohen's d) of the elevation in k_i^cer was 1.18 in the whole striatum and 1.24 in the associative striatum. There was no significant difference in the limbic (psychotic transition group, mean=0.0153/min [SD=0.0013]; comparison group, mean=0.0140/min [0.0019]; p=0.2) or sensorimotor (psychotic transition group, mean=0.0164/min [SD=0.0015]; comparison group, mean=0.0152/min [SD=0.0015]; p=0.095) striatal subdivisions. The voxel-based analysis also identified a greater k_i^cer in the psychotic transition group relative to the comparison group in a voxel cluster with its focus in the left head of the caudate, which lies within the associative subdivision (Figure 2). This difference was significant at p<0.05, corrected for multiple comparisons using the family-wise error rate, and it remained significant after excluding the psychotic transition patient with a non-schizophreniform psychosis. The healthy comparison group > psychotic transition group contrast revealed no significant difference, even at an uncorrected statistical threshold (p<0.05).

Comparison Between the Psychotic Transition and Nontransition Groups

After adjustment for multiple comparisons, the k_i^cer value was significantly elevated in the psychotic transition compared to the nontransition group in the whole striatum (Figure 1; nontransition group, mean=0.0142/min [SD=0.0011], p=0.036) and its associative subdivision (nontransition group, mean=0.0136/min [SD=0.0012], p=0.015), but not in the limbic (nontransition group, mean=0.0149/min [SD=0.0015], p>0.990), or sensorimotor (nontransition group, mean=0.0153/min [SD=0.0013], p=0.222) striatal subdivisions. The voxel-based analysis also identified a greater striatal k_i^cer in the psychotic transition group than in the nontransition group, with a peak in the caudate (Figure 3; p=0.036, corrected at the family-wise error rate). The contrast of the nontransition group > psychotic transition group showed no significant differences, even at an uncorrected statistical threshold (p<0.05).

Comparison Between the Nontransition and Healthy Comparison Groups

The region-of-interest analysis indicated that there was no significant difference in mean k_i^cer value in the nontransition group relative to the healthy comparison group in the whole striatum or its subdivisions, and there were no significant differences in the corresponding voxel-based analysis for the contrast of nontransition group > comparison group, even at an uncorrected threshold (p<0.05).

The Relationship Between Striatal Dopamine Synthesis Capacity and Symptoms

Within the psychotic transition group, there was a significant positive relationship between whole striatal k_i^cer values and total score on both the Comprehensive Assessment of At-Risk Mental States (Figure 4; r=0.67, p=0.049) and the PANSS rating scales (r=0.71, p=0.032). These findings were not driven by outlying or high-influence data points, as assessed using Cook's distance-centered leverage plots. These correlations were not evident in the nontransition group.

Striatal Dopamine Synthesis Capacity in the Schizotypal Personality Disorder Group

There was no significant difference in the k_i^cer values between the schizotypal personality disorder and psychotic transition groups (the k_i^cer values for the schizotypal personality disorder group and the statistical comparisons are presented in the online data supplement). However, k_i^cer values were significantly elevated in the schizotypal personality disorder group relative to the healthy comparison group in the whole striatum (p=0.006) and its associative (p=0.003) and sensorimotor (p=0.047) but not limbic subdivisions, and relative to the nontransition group in the associative (p=0.023) subdivision, but this difference fell short of statistical significance in the whole striatum (p=0.053) and was not significant in the sensorimotor or limbic subdivisions.

We found that dopamine synthesis capacity was elevated with a large effect size (>1) in the patients who presented with prodromal signs of psychosis and went on to a first episode of a psychotic disorder. This elevation was evident relative to healthy comparison subjects and to patients who presented with similar clinical features but did not go on to develop psychosis. In the psychotic transition group, there was a direct relationship between the magnitude of dopaminergic dysfunction and the severity of prodromal symptoms at presentation. Dopamine synthesis capacity in the ultra-high-risk patients who did not subsequently develop psychosis was not significantly different from that in healthy comparison subjects and was not correlated with symptom severity at presentation.

These findings support the dopamine hypothesis (4, 7) by providing the first evidence, to our knowledge, that dopaminergic dysfunction predates the onset of frank psychotic illness in people with symptoms that are truly prodromal to a psychotic disorder, and extend our finding of a longitudinal increase in dopamine synthesis capacity in this group (25). These findings do not preclude the involvement of other neurotransmitter systems that may act upstream to alter dopaminergic function (26, 27). For example, animal studies show that cortical damage can alter striatal dopaminergic function (28, 29), and human neuroimaging data indicate that cortical dysfunction is related to striatal dopaminergic function in both ultra-high-risk patients (30) and patients with schizophrenia (31, 32). Furthermore, we recently found an altered relationship between striatal dopaminergic and cortical glutamatergic indices in psychotic transition patients (33), suggesting that both may be involved in the development of psychosis.

Findings in the Schizotypal Personality Disorder Group

The finding that dopamine synthesis capacity was elevated in the schizotypal personality disorder group relative to the healthy comparison group is consistent with evidence that individuals with schizotypal personality disorder show elevated dopamine release, at a level similar to that in patients with remitted schizophrenia and intermediate between levels in acute schizophrenia and in healthy comparison subjects (13). Because schizotypal personality disorder is independently associated with a greater lifetime risk of schizophrenia relative to the general population (14), the patients in the schizotypal personality disorder group probably continue to have an elevated risk of developing schizophrenia despite not having developed it during the follow-up period. Our data, along with previous findings in schizotypal personality disorder, thus suggest that dopaminergic dysfunction may be related to vulnerability to schizophrenia spectrum disorders, rather than to frank psychosis per se.

However, because a proportion of ultra-high-risk patients in any given sample are likely also to have comorbid schizotypal personality disorder (8), the lack of differences between the psychotic transition and schizotypal personality disorder groups limits the utility of this PET measure as a marker for impending psychosis, although greater sensitivity may be achieved using novel image analytic approaches (34).

Methodological Considerations

A critical consideration in all longitudinal studies in ultra-high-risk samples is that some subjects who have not developed psychosis might still do so after the end of the follow-up period. However, follow-up of ultra-high-risk patients indicates that the great majority of transitions occur within the first 24 months, after which the transition rate sharply declines (9, 10), although occasional transitions may still occur several years after presentation. In the present study, all the patients were followed up for at least 36 months. The development of psychosis has also been associated with reductions in gray matter volume, but in the cerebral cortex rather than in the striatum (35). Moreover, the PET normalization procedure we employed is accurate even when there is structural change (36), and striatal volume loss would, if anything, reduce rather than increase k_i^cer values (37). It thus seems unlikely that our neurochemical findings were secondary to structural alterations in the striata of the psychotic transition group. However, given the evidence of altered corticostriatal interactions in ultra-high-risk patients, future work should investigate the relationship between cortical changes and striatal dopaminergic function. While we could not detect a significant difference in dopamine synthesis capacity between the nontransition group and the healthy comparison group, we cannot exclude the possibility that this reflected a lack of statistical power. Studies using larger samples are needed to clarify whether dopamine synthesis capacity in the nontransition group is intermediate between that in psychotic transition patients and healthy volunteers.

Specificity of the Findings

Our key clinical outcome measure was the development of a first episode of a psychotic disorder, and thus the findings support a link between dopamine dysfunction and psychosis in general rather than as a manifestation of a particular psychotic disorder. Nevertheless, all but one of the patients who developed psychosis met diagnostic criteria for a schizophreniform disorder, and exclusion of the one patient who met criteria for bipolar disorder did not alter the results. It is possible that the findings may vary with type of psychotic disorder, and this issue may be addressed in future studies.

Our finding that the dopaminergic abnormality in the psychotic transition group was localized to the associative striatum extends previous findings in schizophrenia that elevations in dopamine synthesis capacity (11) and synaptic dopamine levels (38) are localized to the associative striatum, by indicating for the first time that this is also the case in the prodrome that precedes the first episode of psychosis. The associative striatum is functionally linked to the dorsolateral prefrontal cortex (see Figure S1 in the online data supplement), a cortical area that has been shown to be the site of functional impairments in ultra-high-risk patients (39). Our finding in the associative striatum thus suggests that frontal-striatal interactions may be important in the development of psychosis. While [¹⁸F]DOPA PET measurements in the striatum show high reliability (40), reliability is lower in smaller structures. However, the voxel-based analysis localized the peak difference to the left head of the caudate, which is within the associative striatum, and thus supports the associative localization of dopaminergic abnormalities and suggests that there may be a lateralization effect that warrants further investigation. Finally, our findings are unlikely to be an effect of antipsychotic medication, as all the ultra-high-risk patients were antipsychotic naive.

The authors are grateful to the volunteers and staff of the OASIS (Outreach and Support in South London) and LEO (Lambeth Early Onset) teams (South London and Maudsley NHS Foundation Trust) and of GE Imanet.