Schizophrenia affects about 1% of the population, 1 and is a major cause of global disease burden. Reference Lopez2 The investigation of the biological underpinnings of the disease are crucial steps to the rational development of more effective treatments for the illness. Reference McGuire, Howes, Stone and Fusar-Poli3,Reference Lewis and Gonzalez-Burgos4 The dopamine hypothesis of schizophrenia has been an enduring theory of the neurobiology underlying the disorder. Reference Meltzer and Stahl5,Reference Van Rossum6 In its earliest instance it was a biochemical hypothesis (increased dopamine transmission) without a precise molecular or anatomical specificity. In the past two decades the hypothesis has been refined to enhance its molecular and anatomical specificity. The newer conceptualisation of the hypothesis postulates an increase in dopamine release and synthesis capacity in the striatum and links this to positive symptoms. Reference Davis, Kahn, Ko and Davidson7,Reference Howes and Kapur8 The meta-analytic evidence for this is robust, with an effect size of 0.8, and has been recently reviewed. Reference Howes, Kambeitz, Kim, Stahl, Slifstein and Abi-Dargham9 The hypothesis has also evolved to link low dopamine activity in cortical regions to negative symptoms and cognitive deficits. Reference Davis, Kahn, Ko and Davidson7,Reference Howes and Kapur8 This has been highly influential - for example, the paper Reference Davis, Kahn, Ko and Davidson7 proposing that low cortical dopamine activity is primary and causes deficit symptoms has attracted nearly 1000 citations to date (see http://apps.webofknowledge.com/). However, to our knowledge the direct evidence for dopaminergic alterations in cortical and other extrastriatal regions has not been previously systematically reviewed and this thus forms the focus of our meta-analysis.
Developments in molecular imaging in the past two decades have enabled the regional specificity and nature of dopaminergic dysfunction to be studied in vivo in patients with schizophrenia. Reference Davis, Kahn, Ko and Davidson7,Reference Howes and Kapur8,Reference Abi-Dargham10 The initial studies focused on the striatum Reference Wong, Wagner, Tune, Dannals, Pearlson and Links11-Reference Reith, Benkelfat, Sherwin, Yasuhara, Kuwabara and Andermann13 and there have now been over 50 studies that have investigated striatal dopamine function in schizophrenia. Reference Howes, Kambeitz, Kim, Stahl, Slifstein and Abi-Dargham9 The in vivo investigation of extrastriatal dopamine function has been technically more complicated, not least because dopaminergic projections to other regions are sparse compared with the striatum. However, the development of high-affinity selective tracers such as [11C]-FLB 457, Reference Halldin, Farde, Högberg, Mohell, Hall and Suhara14 [18F]-fallypride, Reference Mukherjee, Yang, Brown, Lew, Wernick and Ouyang15 [123I]-epidepride Reference Kessler, Mason, Votaw, De Paulis, Clanton and Ansari16 and [11C]-(+)-4-propyl-9-hydroxynaphthoxazine (PHNO) Reference Martin, Williams, Pettibone, Zrada, Lotti and Taylor17 has enabled the in vivo investigation of D2/D3 receptors in extrastriatal regions, and the tracers [11C]-NNC and [11C]-SCH23390 have enabled this for D1 receptors. Reference Elsinga, Hatano and Ishiwata18,Reference Sedvall, Pauli, Karlsson, Farde, Nordström and Nyberg19 Besides dopamine receptors, tracers with specific binding to the dopamine transporter or that use radiolabelled-dihydroxyphenylalanine (DOPA) to index dopamine synthesis capacity have been used to investigate different aspects of dopaminergic neurotransmission in extrastriatal regions. Reference Elsinga, Hatano and Ishiwata18 We recently conducted a meta-analysis of striatal dopaminergic function in schizophrenia. Reference Howes, Kambeitz, Kim, Stahl, Slifstein and Abi-Dargham9 This found that presynaptic dopamine function - specifically dopamine synthesis capacity, dopamine release and baseline synaptic dopamine levels - were highly significantly elevated in schizophrenia with a large effect size (Cohen’s d = 0.8). There was also an elevation in D2/D3 receptor availability but this was relatively small (Cohen’s d = 0.24) and inconsistent. In contrast there was no evidence of an alteration in dopamine transporter or D1 receptor availability. We now aim to systematically review the extrastriatal molecular imaging findings in schizophrenia and, where there are sufficient studies, conduct a meta-analysis to evaluate the evidence for dopaminergic dysfunction outside of the striatum.
Method
Search and selection strategy
The entire PubMed, PsycINFO and MEDLINE electronic databases were searched from 1 January 1950 up to 31 December 2012. Initially, studies were screened based on a search using a comprehensive search term [(“Positron Emission Tomography” OR “PET” OR “Single photon emission tomography” OR “SPET” OR “Single Photon Emission Computed Tomography” OR “SPECT”) AND (“dopamine”) AND (“schizophrenia” OR “psychosis” OR “psychotic” OR “schizophreniform”) AND (“thalamus” OR “thalamic” OR “cingulate” OR “cortex” OR “frontal” OR “prefrontal” OR “temporal” OR “parietal” OR “midbrain” OR “substantia nigra” OR “hippocampus” OR “amygdala”) AND (1950:2012/12/31[DP])]. Only publications in peer-reviewed journals in English language were considered. To be included in the meta-analysis a paper needed to report in vivo positron emission tomography (PET) or single photon emission computed tomography (SPECT)/single photon emission tomography (SPET) imaging findings of extrastriatal dopaminergic function in patients with schizophrenia and a control group with sufficient data to enable the mean and standard deviations for both groups to be calculated. Studies were excluded if healthy controls had any neurological or psychiatric disease or if patients had any neurological condition or psychiatric disease other than a psychotic disorder. Current antipsychotic treatment was an exclusion criterion for the studies of D2/D3 receptors, because it is clear this affects dopamine receptor binding potential. Reference Farde, Wiesel, Halldin and Sedvall20
Where the studies reported overlapping samples, the paper reporting the largest sample size was used and the other paper excluded to prevent double counting. For the region of the brain sampled, if two papers reported different definitions of regions applied to the same data, the paper reporting the regional definition closest to that used by other papers in the meta-analysis was used to ensure greatest comparability. As prior antipsychotic treatment may influence dopaminergic indices, Reference Abi-Dargham, Xu, Thompson, Gil, Kegeles and Urban21 data were presented combined and, where available, separately for patients who had previously received antipsychotic treatment and for patients who were antipsychotic-naive to enable findings to be compared.
Data extraction
The main outcome measure was the difference in the dopaminergic imaging parameter between healthy controls and patients with schizophrenia. The following additional information was extracted from all the studies: names of the authors, year of publication, population characteristics of the healthy control and patient groups (group size, age, gender, antipsychotic use, diagnosis, symptom ratings), characteristics of the imaging (radiotracer, other methodological factors reported), scanner characteristics (scanner type and resolution), and modelling method. The data were extracted by one author (J.K.) and checked twice additionally to assure accuracy. In case of uncertainties data were checked by another author (O.D.H.) and consensus reached. As there are no established criteria for assessing the quality of molecular imaging studies, we have summarised methodological aspects of each study to enable individual judgements to be made (see online Tables DS1 and DS2).
Data analysis
A minimum of five studies was required to proceed with the meta-analysis as findings become less reliable with fewer studies. Reference Ioannidis and Lau22 The R statistical programming language 23 version 2.10.1 on Mac OS-X (version 10.6.8) with the package ‘metafor’ Reference Viechtbauer24 was used to conduct meta-analyses as well as power calculations. Reference Champely25 A significance level of P<0.05 (two-tailed) was used for all analyses.
The standardised effect sizes of the individual studies were entered in a random-effects meta-analytic model, Reference Hedges and Olkin26,Reference Hedges and Vevea27 which does not assume homogeneity among studies. The summary effect sizes (cohen’s d) were computed using a restricted maximum-likelihood estimator. Reference Raudenbusch, Cooper, Hedges and Valentine28 Heterogeneity was assessed in the studies by calculating the I 2 value, which is a sample size independent measure that describes the percentage of total variation across studies that is as a result of heterogeneity rather than chance. Reference Higgins, Thompson, Deeks and Altman29 As a guideline, I 2 values of 0-40% indicate heterogeneity that is not important, 30-60% moderate heterogeneity, 50-90% substantial heterogeneity and 75-100% considerable heterogeneity Reference Higgins, Thompson, Deeks and Altman29 (for further details see Higgins & Green Reference Higgings and Green30 ). Pre-specified analyses to evaluate potential sources of bias and sensitivity analyses were conducted as follows. To examine the potential effect of prior antipsychotic treatment, the analysis was repeated for subgroups defined by treatment history (antipsychotic-naive or previously treated) to determine whether this influenced the findings. The potential effects of publication year, gender and the age of participants was evaluated using meta-regression. Reference Hedges and Olkin26 To account for differences in precision, studies were weighted by their sample size before being entered in a regression model with the variable of interest as a predictor for the standardised effect sizes.
Publication bias was evaluated by inspection of the funnel plot (a plot of effect sizes on the x-axis against standard error (1/precision)) for evidence of asymmetry. Publication bias is suggested if studies with small precision and small effect size are absent. Publication bias was further evaluated using Egger’s test. Reference Egger, Smith, Schneider and Minder31 To assess the influence of individual studies on the estimated summary effect size, a post hoc leave-one-out approach was applied by re-running the meta-analysis after leaving out one different individual study at successive iterations.
Results
After initial identification of 242 articles, 219 articles were excluded (see online Fig. DS1 for reasons for excluding studies). This resulted in a sample of 23 studies of extrastriatal dopamine in schizophrenia including 13 studies of D2/D3 receptors, 4 studies of D1 receptors, 5 studies of dopamine synthesis capacity and 1 study of dopamine transporter availability. The most commonly used outcome measure was the binding potential relative to the non-displaceable compartment (BPND). The BPND is the equilibrium ratio of the concentration of specifically bound radioligand relative to the sum of the free and non-specifically bound radioligand, estimated from activity in a reference region. Alternatively the equilibrium ratio relative to either the total or free (unbound) concentration of radioligand in plasma can be used, denoted as BPP or BPF respectively. Reference Innis, Cunningham, Delforge, Fujita, Gjedde and Gunn32 The BPND, BPP and BPF are all proportional to the concentration of receptors available to be bound in the tissue of interest. Reference Innis, Cunningham, Delforge, Fujita, Gjedde and Gunn32 As all the studies reported BPND and only two studies reported BPP and BPF, we used BPND for the meta-analyses. For the meta-analysis of thalamic D2/D3 receptors some studies reported results of overlapping samples, in which case the study with the largest sample was selected for inclusion. Thus, Buchsbaum et al Reference Buchsbaum, Christian, Lehrer, Narayanan, Shi and Mantil33 was excluded as there was sample overlap with Lehrer et al. Reference Lehrer, Christian, Kirbas, Chiang, Sidhu and Short34 Also there was sample overlap between Yasuno et al, Reference Yasuno, Suhara, Okubo, Sudo, Inoue and Ichimiya35 Yasuno et al Reference Yasuno, Suhara, Okubo, Ichimiya, Takano and Sudo36 and Suhara et al. Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37 Only Suhara et al Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37 was included as the thalamic region used was closest to that of other studies in the available literature. For Talvik et al Reference Talvik, Nordström, Olsson, Halldin and Farde38 and Talvik et al Reference Talvik, Nordström, Okubo, Olsson, Borg and Halldin39 there appeared to be sample overlap. As attempts to contact the authors to verify this were unsuccessful, we included only one of both studies in each meta-analysis. We included Talvik et al Reference Talvik, Nordström, Olsson, Halldin and Farde38 in the meta-analysis of the temporal cortex and Talvik et al Reference Talvik, Nordström, Okubo, Olsson, Borg and Halldin39 for the meta-analysis of the thalamus to maximise sample size and to avoid potential sample overlap. Kegeles et al Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 applied the simplified reference tissue model (SRTM) and a two-tissue compartmental model to the imaging data and reported D2/D3 BPND for both. As most other studies in this analysis applied an SRTM approach, we selected these data from Kegeles et al Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 to ensure the modelling approach was the same across studies. Kegeles et al Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 reported D2/D3 BPND values corrected for partial volume effects. These were entered in the analysis rather than uncorrected values.
The characteristics of the participants for the included studies are shown in Table DS1 and the imaging methods are summarised in Table DS2. There were sufficient studies to conduct separate meta-analyses for the D2/D3 receptor availability in the thalamus, the substantia nigra and for the temporal cortex. There were too few studies to enable reliable meta-analysis of findings for other regions or for other measures of dopaminergic function - for these studies we have plotted the individual effect sizes to aid comparison and evaluation of trends and summarised the findings below.
D2/D3 receptor availability
Thalamus
Eight studies comprising 138 patients with schizophrenia and 126 healthy controls met inclusion criteria for the meta-analysis. The summary effect size for thalamic D2/D3 BPND between patients with schizophrenia and healthy controls was d = –0.32 but did not reach significance (95% CI –0.68 to 0.03, z = –1.787, P = 0.07, I 2 = 48.79%, 95% CI 0-84.25%, see Fig. 1(a)). Inspection of the funnel plot (Fig. 1(b)) and Egger’s test (z = –0.6727, P = 0.5011) did not show evidence for publication bias. The sensitivity analysis for the thalamus showed a significant summary effect size for two out of eight iterations. There was a significant summary effect size when Kegeles et al Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 or Glenthoj et al Reference Glenthoj, Mackeprang, Svarer, Rasmussen, Pinborg and Friberg41 were excluded from the meta-analysis. Meta-regression analysis did not show evidence for an effect of year of publication (β = 0.0275, F(1,6) = 0.1591, P = 0.7038, see Fig.1(c)), gender (β = 1.257, F(1,4) = 0.1675, P = 0.7033) or age (β = –0.0695, F(1,5) = 0.3863, P = 0.5615) on the estimated summary effect size. Restricting the meta-analysis to the five studies that included participants with previous exposure to antipsychotic medication showed a non-significant effect size of d = –0.34 (95% CI –0.78 to 0.1, z = –1.5058, P = 0.1, I 2 = 42.83%, 95% CI 0-89.38%). The studies of medication-naive participants showed effect sizes of 0.35, Reference Glenthoj, Mackeprang, Svarer, Rasmussen, Pinborg and Friberg41 –0.62 Reference Talvik, Nordström, Okubo, Olsson, Borg and Halldin39 and –0.77. Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37 However, there were too few studies of medication-naive participants to permit meta-analysis.
Temporal cortex
We included six studies in the meta-analysis comprising 84 patients with schizophrenia and 86 healthy controls. There was no evidence for a reduction in D2/D3 availability in the temporal cortex in patients with schizophrenia (d = –0.23, 95% CI –0.54 to 0.07, z = –1.5097, P = 0.1, I 2 = 0%, 95% CI 0-67.2%, see Fig. 2(a)). Inspection of the funnel plot (Fig. 2(b)) and Egger’s test (z = 0.6917, P = 0.4891) did not show evidence for publication bias. The sensitivity analysis for the temporal cortex did not show a significant summary effect size in any of the six iterations. Meta-regression analysis did not show evidence for an effect of year of publication (β = –0.0297, F(1,4) = 0.7138, P = 0.4458, see Fig. 2(c)), gender (β = –1.1818, F(1,3) = 2.2341, P = 0.2319) or age (β = 0.0535, F(1,4) = 1.4834, P = 0.2902) on the estimated summary effect size. The meta-analysis of the temporal cortex included three studies in drug-free patients that reported effect sizes of d = –0.42, Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 d = –0.33 Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 and d = 0.0. Reference Tuppurainen, Kuikka, Viinamäki, Husso-Saastamoinen, Bergström and Tiihonen45 The studies in drug-naive participants reported comparable effect sizes of d = 0.49, Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37 d = –0.12 Reference Glenthoj, Mackeprang, Svarer, Rasmussen, Pinborg and Friberg41 and d = 0.27. Reference Talvik, Nordström, Olsson, Halldin and Farde38 The small number of studies did not allow further analysis of the moderating effect of medication in this region.
Substantia nigra
We included five studies in the meta-analysis comprising 61 patients with schizophrenia and 72 healthy controls. There was no significant difference between patients with schizophrenia and healthy controls (d = 0.04, 95% CI –0.92 to 0.99, z = 0.075, P = 0.9, I 2 = 84.8%, 95% CI 52.89-98.49%, see Fig. 3(a)). Inspection of the funnel plot (Fig. 3(b)) and Egger’s test (z = –0.79, P = 0.43) did not show evidence for a publication bias. The sensitivity analysis for the substantia nigra did not show a significant summary effect size in any of the five iterations. Meta-regression did not show evidence for an effect of year of publication (β = 0.1649, F(1,3) = 0.3381, P = 0.6018, see Fig. 3(c)), gender (β = –0.6768, F(1,3) = 0.0375, P = 0.8587) or age (β = 0.0085, F(1,3) = 0.0008, P = 0.9793) on the estimated summary effect size. Also after excluding the one study of drug-naive patients there was still no significant effect (d = –0.04, 95% CI –1.31 to 1.22, z = –0.0628, P = 0.9, I 2 = 88.59%, 95% CI 60.53-99.28%).
Findings in the other regions
There were seven studies with a total of 109 patients and 120 controls investigating D2/D3 BPND in 11 other extrastriatal regions. The effect sizes for these regions are shown by study in Fig. 4. Out of three studies of the anterior cingulate, one reported a significant decrease of D2/D3 BPND in patients with schizophrenia Reference Yasuno, Suhara, Okubo, Ichimiya, Takano and Sudo36 and two reported no significant change. Reference Talvik, Nordström, Olsson, Halldin and Farde38,Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 One study reported a significant decrease of D2/D3 BPND in the uncus. Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 There were no significant changes reported in the entorhinal cortex, Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 the hippocampus, Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37,Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40,Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 the amygdala, Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40,Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 the prefrontal cortex, Reference Yasuno, Suhara, Okubo, Ichimiya, Takano and Sudo36,Reference Talvik, Nordström, Olsson, Halldin and Farde38,Reference Glenthoj, Mackeprang, Svarer, Rasmussen, Pinborg and Friberg41 the occipital cortex, Reference Suhara, Okubo, Yasuno, Sudo, Inoue and Ichimiya37 the parietal cortex, Reference Yasuno, Suhara, Okubo, Ichimiya, Takano and Sudo36 the insula Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 or the globus pallidum. Reference Graff-Guerrero, Mizrahi, Agid, Marcon, Barsoum and Rusjan43
Dopamine synthesis capacity
Four studies investigated presynaptic dopamine synthesis capacity outside the striatum. Reference Nozaki, Kato, Takano, Ito, Takahashi and Arakawa46-Reference Kumakura, Cumming, Vernaleken, Buchholz, Siessmeier and Heinz49 These studies report data from 58 healthy controls and 53 patients in 12 different regions (Fig. 4(a)). There were significant changes in two regions in one study. This study reported a significant decrease in the middle frontal cortex and a significant increase in the posterior cingulate. Reference Elkashef, Doudet, Bryant, Cohen, Li and Wyatt48
D1 receptor availability
There were four studies of D1 receptor availability including a total of 55 patients and 68 controls and reporting results for 15 different regions Reference Abi-Dargham, Xu, Thompson, Gil, Kegeles and Urban21,Reference Okubo, Suhara, Suzuki, Kobayashi, Inoue and Terasaki50-Reference Abi-Dargham, Mawlawi, Lombardo, Gil, Martinez and Huang52 (summarised in Fig. 4(b)). There were three reports of significant changes in the prefrontal cortex, and one report of no significant change. Abi-Dargham et al Reference Abi-Dargham, Mawlawi, Lombardo, Gil, Martinez and Huang52 reported a significant increase in prefrontal D1 BPND in patients with schizophrenia. However, Okubo et al Reference Okubo, Suhara, Suzuki, Kobayashi, Inoue and Terasaki50 report a significant decrease. Both of these studies included drug-naive and previously drug-treated patients. The most recent study reported a significant elevation in D1 receptor availability in antipsychotic-naive patients, but no difference from controls in previously anti-psychotic-treated patients. Reference Abi-Dargham, Xu, Thompson, Gil, Kegeles and Urban21 No other extrastriatal regions showed significant changes in D1 BPND in schizophrenia in any of these studies.
Fig. 1 (a) Forrest plot, (b) funnel plot and (c) meta-regression with year of publication of meta-analysis of thalamic D2/D3 BPND.
Error bars represent 95% confidence intervals; RE model, random-effects model.
Dopamine transporter availability
One study investigated dopamine transporter availability outside the striatum in 12 healthy controls and 8 patients for two regions Reference Arakawa, Ichimiya, Ito, Takano, Okumura and Takahashi53 (Fig. 4(d)). This study reports a significant increase in the thalamus but no significant change in the substantia nigra.
Discussion
D2/D3 receptor availability
Our meta-analyses showed small, non-significant reductions in D2/D3 receptor availability in the thalamus (d = –0.32, P = 0.07) and temporal cortex (d = –0.23, P = 0.1) in schizophrenia and no significant difference between patients and controls in the substantia nigra (d = 0.04, P = 0.9). Although these were not statistically significant, it is important to note that the summary estimates were relatively imprecise and the confidence intervals for the thalamus and temporal cortex included moderate-large reductions as well as very small elevations. Furthermore, the leave-one-out sensitivity analysis for the thalamus found that the reduction was significant on two iterations, indicating that the meta-analysis was sensitive to the inclusion of two individual studies, and that the lack of significance should not be considered as conclusively excluding a reduction. This sensitivity could reflect differences in methodology or sample characteristics (see later) and indicates the need for further large studies to conclusively address the issue. Although there were too few studies to permit meta-analysis in other regions, the majority of studies found no significant differences in patients, and where there were significant decreases this was either a sole finding or not replicated in other studies. Thus, overall there do not appear to be marked alterations in D2/D3 receptor availability in extrastriatal regions but there may be a small reduction in D2/D3 availability in the thalamus. The clinical significance of a small reduction in D2, if it is present, is not clear, and, although one study does report a significant correlation between lower D2/D3 receptor availability in subregions of the thalamus and higher symptom severity, Reference Yasuno, Suhara, Okubo, Sudo, Inoue and Ichimiya35 other studies report the opposite relationship Reference Talvik, Nordström, Olsson, Halldin and Farde38,Reference Kessler, Woodward, Riccardi, Li, Ansari and Anderson44 or no relationship in these regions. Reference Glenthoj, Mackeprang, Svarer, Rasmussen, Pinborg and Friberg41
Fig. 2 (a) Forrest plot, (b) funnel plot and (c) meta-regression with year of publication of meta-analysis of temporal D2/D3 BPND.
Error bars represent 95% confidence intervals; RE model, random-effects model.
D1 receptor availability
There were too few studies to permit meta-analysis and findings were inconsistent across studies. A factor that could underlie the inconsistency is suggested by a study in rats comparing the effects of dopamine depletion on the binding of the two tracers used. Reference Guo, Hwang, Lo, Huang, Laruelle and Abi-Dargham54 This found that although dopamine depletion increased, as expected, the binding of [11C]-NNC, it paradoxically decreased the binding of [11C]-SCH23390. Thus, if this also occurs in humans, low cortical dopamine levels in schizophrenia would be associated with opposite effects on the binding of these tracers. Additionally, antipsychotic treatment has been found to reduce D1 receptor density, Reference Lidow and Goldman-Rakic55 indicating that prior antipsychotic treatment could be confounded in some studies. A further complicating issue is that both the tracers used in the studies show appreciable binding to 5-hydroxytrptamine (5HT2A) receptors. Reference Ekelund, Slifstein, Narendran, Guillin, Belani and Guo56 In the absence of more selective D1 tracers, blocking studies using selective 5HT2A or D1 compounds would enable the D1 specific signal to be evaluated. In summary, the available data of frontal D1 binding are limited and further studies in drug-naive patients are warranted.
Dopamine transporter availability
There has been remarkably little investigation of dopamine transporter availability in vivo outside of the striatum. The two findings were of a significant increase in dopamine transporter availability in the thalamus and no difference in the substantia nigra. Also, it is noteworthy that the tracer used shows good test-retest characteristics in both regions, Reference Hirvonen, Johansson, Teräs, Oikonen, Lumme and Virsu57 indicating that measurement is reliable and suggesting that further investigation is warranted, particularly in the thalamus given the potential reduction in D2/D3 receptor availability here.
Fig. 3 (a) Forrest plot, (b) funnel plot and (c) meta-regression with year of publication of meta-analysis of D2/D3 BPND in the substantia nigra.
Error bars represent 95% confidence intervals; RE model, random-effects model.
Dopamine synthesis capacity
The majority of findings indicate there was no significant alteration in cortical dopamine synthesis capacity in schizophrenia, although there was both a large effect size reduction in one frontal cortical region and a large effect elevation in the posterior cingulate in the same study. However, the findings in the frontal cortex, and the negative findings in the temporal cortex and amygdala need to be considered in the context of the relatively low reliability of imaging dopamine synthesis capacity in these regions Reference Egerton, Demjaha, McGuire, Mehta and Howes58 and evidence that radiolabelled DOPA does not give a specific PET signal in cortical regions. Reference Cropley, Fujita, Bara-Jimenez, Brown, Zhang and Sangare59 Importantly, Cropley et al Reference Cropley, Fujita, Bara-Jimenez, Brown, Zhang and Sangare59 found that there was greater uptake of [18F]-DOPA in white matter than grey matter in frontal cortex, and that partial volume correction for white matter reduced the [18F]-DOPA uptake rate constant while increasing its variability. Thus, the cortical findings, both positive and null, need to be considered with these major caveats in mind. Further investigation of cortical dopamine synthesis would benefit from the development of more specific radiotracers.
General methodological considerations
Variation in the quality of studies is a potential source of bias in the meta-analyses. One important potential source of bias, particularly for small structures such as the substantia nigra, is the influence of partial volume effects. These become important for structures two to three times smaller than the resolution of the scanner. Reference Rousset, Ma and Evans60,Reference Kim, Shidahara, Tsoumpas, McGinnity, Kwon and Howes61 These effects include loss of signal in affected regions as well as spill-over from neighbouring structures. Where there is a loss of volume in a region, as there is in cortical regions in schizophrenia, Reference Fornito, Yücel, Patti, Wood and Pantelis62 this could bias group comparisons. Six of the studies examined whether there were differences in the volume of structures examined between the schizophrenia and controls group - four found no difference, Reference Viechtbauer24-Reference Hedges and Vevea27,Reference Karlsson, Farde, Halldin and Sedvall51 and two found a significant difference. Reference Raudenbusch, Cooper, Hedges and Valentine28,Reference Graff-Guerrero, Mizrahi, Agid, Marcon, Barsoum and Rusjan43 However, only one study Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 made use of partial volume correction methods to reduce the impact of partial volume effects. Reference Kim, Shidahara, Tsoumpas, McGinnity, Kwon and Howes61,Reference Rousset, Collins, Rahmim and Wong63 When this was applied in the uncus, for example, values changed markedly: there was a 109% increase in D2/D3 BPND in controls and an 88% increase in patients with schizophrenia. Reference Kegeles, Slifstein, Xu, Urban, Thompson and Moadel40 Type of illness may also be important, Reference Demjaha, Murray, McGuire, Kapur and Howes64 with evidence that patients whose condition is refractory do not show the same dopaminergic changes, for example, and this was not evaluated in the studies.
Fig. 4 Overview of the effect sizes found by (a) studies of presynaptic dopamine synthesis, (b) D1 receptor studies, (c) D2 receptor studies and (d) dopamine transporter (DAT) studies for different extrastriatal regions.
Error bars represent 95% confidence intervals. Significant changes are indicated by * for P<0.05, ** for P<0.01 and *** for P<0.001. Previous treatment of the patient sample is indicated by (DF) for drug-free and (DN) for drug-naive. Please note that the significance level plotted in Fig. 4 has not been corrected for multiple comparisons.
It is important to note that binding potential, the outcome measure in the receptor studies, depends on the density of receptors, affinity of the receptor for the radiotracer and occupancy by endogenous dopamine. Alterations in any one of these parameters could contribute to differences or, conceivably, if more than one parameter is affected in a condition, the net result could either be that they cancel each other out or are additive. Reference Egerton, Mehta, Montgomery, Lappin, Howes and Reeves65 Studies using techniques that enable separate measurements of receptor density (B max), affinity (1/K d; where K d is the dissociation constant for the radiotracer from the receptor) and intrasynaptic dopamine levels Reference Abi-Dargham, Rodenhiser, Printz, Zea-Ponce, Gil and Kegeles66-Reference Hietala, Någren, Lehikoinen, Ruotsalainen and Syvälahti68 are needed to disentangle these issues. Although most D2/D3 receptor radiotracers we included in the meta-analyses have similar affinity for D2 and D3 receptors, PHNO, as used in Graff-Guerrero et al, Reference Graff-Guerrero, Mizrahi, Agid, Marcon, Barsoum and Rusjan43 has a higher affinity for D3 than D2 receptors, Reference Narendran, Slifstein, Guillin, Hwang, Hwang and Scher69 which means it is relatively less sensitive to D2 alterations.
For the meta-analyses it is important to note that the sample sizes were modest and, as meta-analytic results are less stable with smaller sample sizes, Reference Trikalinos, Churchill, Ferri, Leucht, Tuunainen and Wahlbeck70 consequently may change significantly with the addition of future studies. Furthermore, it is advisable not to solely rely on significance tests to interpret results, but to also consider the summary effect size estimate and associated confidence interval. Reference Stang71 This is particularly the case as the meta-analyses lacked power to detect small differences. Finally, we found evidence of moderate to large heterogeneity in our meta-analyses (indicated by I 2 values greater than 50%). However, the random-effects approach to meta-analysis we used is relatively robust to heterogeneity. Reference Lehrer, Christian, Kirbas, Chiang, Sidhu and Short34
Implications and future directions
An updated version of the dopamine hypothesis proposes that negative and cognitive symptoms of schizophrenia are secondary to cortical hypodopaminergia, whereas psychotic symptoms are secondary to subcortical hyperdopaminergia. Reference Davis, Kahn, Ko and Davidson7,Reference Howes and Kapur8 In the past two decades consistent in vivo evidence has accrued for subcortical hyperdopaminergia, with a large effect size elevation in striatal dopamine synthesis and release in schizophrenia Reference Howes, Kambeitz, Kim, Stahl, Slifstein and Abi-Dargham9 and potentially sufficient specificity to be a biomarker. Reference Bose, Turkheimer, Howes, Mehta, Cunliffe and Stokes72,Reference Howes, Montgomery, Asselin, Murray, Grasby and McGuire73 In contrast, our meta-analysis of dopamine studies in cortical regions highlights the relative paucity of in vivo studies to support the hypothesis of reduced cortical dopaminergic function. This is surprising given how influential the hypothesis of cortical hypodopaminergia has been and the central role of dopamine-blocking drugs in the treatment of schizophrenia. Reference Howes, Egerton, Allan, McGuire, Stokes and Kapur74
Our finding of a small but not statistically significant reduction in D2/D3 availability in the thalamus contrasts with our previous findings in the striatum, where there was a small but inconsistent elevation. Reference Howes, Kambeitz, Kim, Stahl, Slifstein and Abi-Dargham9 One potential explanation for this inconsistency is that there are group difference in the volumes of these structures which, if they were smaller in schizophrenia, would lead to lower values due to partial volume effects (see above).
Our results indicate that the initial report of a large reduction in thalamic D2/D3 receptor availability Reference Yasuno, Suhara, Okubo, Sudo, Inoue and Ichimiya35 appears to have moderated into a non-significant reduction with subsequent reports. However, as the confidence interval includes -0.6, it remains possible that there is a moderate-large effect size reduction in schizophrenia. Furthermore, given the finding of elevated thalamic dopamine transporter availability and the role of the thalamus in relaying sensory information, Reference Sánchez-González, García-Cabezas, Rico and Cavada75 dopaminergic dysfunction here could plausibly play a role in the development of hallucinations and other psychotic symptoms. Thus, further investigation of thalamic dopamine function is warranted.
The substantia nigra is the location of the cell bodies of the dopamine neurons projecting to the striatum. As such the lack of alterations in dopamine transporter or D2/D3 availability in the nigra suggests there is no increase in dopamine neuron density or altered D2/D3 autoreceptor function underlying the striatal dopaminergic dysfunction seen in schizophrenia Reference Abi-Dargham, Rodenhiser, Printz, Zea-Ponce, Gil and Kegeles66,Reference Laruelle, Abi-Dargham, van Dyck, Gil, D'Souza and Erdos76-Reference Howes, Montgomery, Asselin, Murray, Valli and Tabraham78 and its prodrome. Reference Howes, Bose, Turkheimer, Valli, Egerton and Stahl79,Reference Mizrahi, Addington, Rusjan, Suridjan, Ng and Boileau80 Taken with evidence for altered dopamine synthesis capacity in the nigra, Reference Howes, Williams, Ibrahim, Leung, Egerton and McGuire81 this suggests there is increased nigral dopamine function but no compensatory change in nigral D2/D3 receptors.
A fundamental issue that remains to be established is whether there is reduced cortical dopamine release in schizophrenia. Although the relatively lower density of dopamine receptors in cortical regions has made this more challenging to study than in the striatum, Reference Egerton, Mehta, Montgomery, Lappin, Howes and Reeves65 evidence from some, Reference Narendran, Slifstein, Guillin, Hwang, Hwang and Scher69,Reference Narendran, Mason, May, Chen, Kendro and Ridler82-Reference Chou, Halldin and Farde85 although not all, Reference Mukherjee, Christian, Narayanan, Shi and Collins86-Reference Frankle, Mason, Rabiner, Ridler, May and Asmonga92 of the studies with high affinity D2/D3 selective radiotracers indicates that this is possible with sufficient reliability for group comparisons. There is, thus, the potential to test this unresolved aspect of the dopamine hypothesis in vivo in future studies. Finally, there has been relatively little in vivo investigation of the upstream and downstream regulation of dopaminergic signalling in schizophrenia; both of which are needed to understand the nature of dopaminergic dysfunction in the disorder. Reference Stone, Howes, Egerton, Kambeitz, Allen and Lythgoe93,Reference Blasi, Napolitano, Ursini, Taurisano, Romano and Caforio94
Implications
Despite the wide influence of the cortical dopamine hypofunction hypothesis and the central role of dopamine blockade in the treatment of schizophrenia, there is relatively limited direct evidence of altered dopaminergic function in cortical and other extrastriatal regions in schizophrenia. In this context it is worth remembering that secondary indices (such as reduced blood flow or altered frontal activation) are not a substitute for direct evidence. The available data are inconclusive and further investigation is warranted to determine whether there are alterations in thalamic D2/D3 receptors, and in D1 receptor availability, dopamine synthesis capacity and dopamine transporter availability.
Funding
This study was funded by a Medical Research Council (UK) grant to O.D.H. (grant number: MC-A656-5QD30) and the National Institute of Health Research Biomedical Research Council.
eLetters
No eLetters have been published for this article.