Translating psychiatric imaging into clinical utility

The lack of clinical relevance for psychiatric imaging is particularly concerning in the early phases of schizophrenia, because of the severe clinical, functional, social and economic long-term consequences of the illness. In this sense the finding of Bodnar et al that alterations in the cingulate cortex during a first episode of schizophrenia are related to psychopathology and outcomes are of great interest. ^{Reference Bodnar, Achim, Malla, Joober, Benoit and Lepage1} Structural alterations in the cingulate cortex have been confirmed at a meta-analytical level in participants presenting with a first episode of schizophrenia. ^{Reference Bora, Fornito, Radua, Walterfang, Seal and Wood5} However, cingulate function and structure has been reported to be especially sensitive to remedial antipsychotic treatment in schizophrenia ^{Reference Lahti, Weiler, Holcomb, Tamminga and Cropsey6} and there is evidence indicating that a few weeks of antipsychotic treatment modulate the functional response in this region. ^{Reference Snitz, Macdonald, Cohen, Cho, Becker and Carter7} As the participants in Bodnar et al's study were receiving antipsychotics, the clinical significance of their findings is questioned. Antipsychotic exposure can play a prominent confounding role in second-generation psychiatric imaging, militating against its clinical application. ^{Reference Smieskova, Fusar-Poli, Allen, Bendfeldt, Stieglitz and Drewe8} One possible approach to circumvent this problem would be to selectively analyse participants with a first episode who are drug naive. In a recent meta-analysis including participants with untreated first-episode schizophrenia we confirmed that structural alterations in the cingulate cortex are present before the initiation of antipsychotic treatment. ^{Reference Fusar-Poli, Radua, McGuire and Borgwardt9}

An alternative option would be to endorse ‘close in’ clinical high-risk approaches to identify a group of individuals with higher transition rates (18% after 6 months of follow-up, 22% after 1 year, 29% after 2 years and 36% after 3 years ^{Reference Fusar-Poli, Bonoldi, Yung, Borgwardt, Kempton and Barale10} ) than those observed in the general population. This clinical strategy aims at identifying neural changes occurring prior to the onset of schizophrenia and may improve the ability of neuroimaging to predict clinical outcomes in schizophrenia (for a review of structural and functional findings see Fusar-Poli et al ^{Reference Fusar-Poli, Borgwardt, Crescini, Deste, Kempton and Lawrie11} and Smieskova et al ^{Reference Smieskova, Fusar-Poli, Allen, Bendfeldt, Stieglitz and Drewe12} ). Overall, these studies have shown that several abnormalities in brain anatomy and neurophysiology that are fundamental to schizophrenia are also present in people at high risk of schizophrenia, and may therefore represent vulnerability markers. ^{Reference Fusar-Poli, Borgwardt, Crescini, Deste, Kempton and Lawrie11} Interestingly, structural brain abnormalities in the insula and temporal lobe are also found to covary with levels of schizotypy in healthy individuals. ^{Reference Ettinger, Williams, Meisenzahl, Moller, Kumari and Koutsouleris13} A recent meta-analysis of whole brain structural studies comparing 896 participants at high risk with 701 controls confirmed cingulate alterations in the high-risk group when compared with the control group, and additional abnormalities in temporal, prefrontal, parahippocampal/hippocampal regions. ^{Reference Fusar-Poli, Borgwardt, Crescini, Deste, Kempton and Lawrie11} Volumetric reductions in cingulate as well as in temporal, insular, prefrontal cortex and in cerebellum have also been associated with longitudinal development of schizophrenia over follow-up. ^{Reference Smieskova, Fusar-Poli, Allen, Bendfeldt, Stieglitz and Drewe12} However, because of the paucity of structural imaging studies specifically linking brain changes and clinical outcomes, second-generation imaging studies are not able to definitively ascertain which structural abnormalities are specific to vulnerability as opposed to later transition to schizophrenia.

Linking imaging findings to clinical status

In this sense, Bodnar et al's study linking imaging findings with remission status points to a crucial gap in the high-risk literature. ^{Reference Bodnar, Achim, Malla, Joober, Benoit and Lepage1} In fact, a recent systematic review showed a literature bias in that nearly half of the high-risk studies provided no characteristics of those participants who did not develop schizophrenia. ^{Reference Simon, Velthorst, Nieman, Linszen, Umbricht and de Haan14} The largest study published to date showed the non-converting high-risk group demonstrated significant improvement in attenuated positive symptoms, negative symptoms, and social and role functioning with more than 50% of this non-converting sample no longer presenting with any high-risk symptoms. ^{Reference Addington, Cornblatt, Cadenhead, Cannon, McGlashan and Perkins15} However, this group remained on average at a lower level of functioning than non-psychiatric comparison participants, suggesting that initial high-risk categorisation is associated with persistent disability for a significant proportion. ^{Reference Addington, Cornblatt, Cadenhead, Cannon, McGlashan and Perkins15} In line with Bodnar and colleagues' approach it would be very useful to address brain changes associated with remission status within the high-risk cohort to identify protective neurobiological markers of later development of illness. In terms of predicting clinical outcome and strengthening clinical applications for psychiatric imaging, there is evidence from functional and neurochemical high-risk studies that the extent of abnormality at baseline is predictive of subsequent conversion to psychosis. ^{Reference Smieskova, Fusar-Poli, Allen, Bendfeldt, Stieglitz and Drewe12} These neurofunctional abnormalities were not only related to different duration of high risk but also to grey matter reductions. ^{Reference Smieskova, Allen, Simon, Aston, Bendfeldt and Drewe16} Furthermore, structural abnormalities were positively correlated with clinical outcomes such as global functioning, negative symptomatology and hallucinations. Additional MRS studies in high-risk individuals have linked abnormal neuronal density and membrane turnover in cingulate as well as in frontal and insular lobes with later development of psychosis. ^{Reference Stone, Day, Tsagaraki, Valli, McLean and Lythgoe17} Positron emission tomography studies addressing dopaminergic neurotransmission before and after the onset of psychosis found an increased striatal presynaptic dopamine synthesis capacity that predicts the onset of illness, ^{Reference Howes, Bose, Turkheimer, Valli, Egerton and Stahl18} in line with consistent evidence pointing to early striatal presynaptic dopaminergic alterations in schizophrenia. ^{Reference Fusar-Poli and Meyer-Lindenberg19} Overall, second-generation imaging research into the high-risk state for psychosis has exponentially progressed, sustaining preventive interventions in clinical psychiatry. ^{Reference Ruhrmann, Schultze-Lutter, Bechdolf and Klosterkotter20} However, despite the potential, the validity of high-risk criteria is still greatly debated and the problem of the high number of false positives severely undermines the benefits of preventive interventions.

Third-generation psychiatric imaging

There is, therefore, an urgent need for psychiatric imaging to move towards third-generation paradigms in line with Bodnar et al's study. Studies need to move away from simple investigations of the neurobiology underlying the early phases of schizophrenia, towards imaging that translates into clinically useful information, targeting longitudinal outcomes including transition, remission and response to preventive interventions. Third-generation psychiatric imaging in early psychosis will benefit from utilising even more complex techniques including multimodal approaches, ^{Reference Fusar-Poli, Howes, Allen, Broome, Valli and Asselin21} multicenter analyses ^{Reference Mechelli, Riecher-Rossler, Meisenzahl, Tognin, Wood and Borgwardt22} or automated diagnostic methods (support vector machines (SVMs)). ^{Reference Orrù, Pettersson-Yeo, Marquand, Sartori and Mechelli23} In particular, multivariate pattern recognition methods such as SVM are able to predict progression through different disease stages and categorise individual brain scans by separation of images from different groups, taking into account the interregional dependencies of different pathologies. Support vector machines use information from all voxels to reflect differences between groups in order to create models that allow predictions of clinical outcomes in individual patients (i.e. prediction of subsequent conversion to psychosis) with an accuracy of 82%. ^{Reference Koutsouleris, Borgwardt, Meisenzahl, Bottlender, Moller and Riecher-Rossler24}

Future third-generation imaging studies in early schizophrenia will also benefit from the incorporation of new sources of neurobiological information such as whole genome sequencing, proteomic, lipidomic and expression profiles and cellular models derived from recent research on induced pluripotent stem cells. ^{Reference Meyer-Lindenberg25} For psychiatric imaging to be something more than basic neuroscience more studies such as the one by Bodnar et al are urgently required in the high-risk and early schizophrenia literature, to selectively link basic research and clinical outcomes. ^{Reference Fusar-Poli, Borgwardt and McGuire26} These new third-generation neuroimaging approaches will sustain a research enterprise that it is hoped will improve and create therapeutic options for early schizophrenia and ultimately help in the treatment of our patients.

Funding

S.B. was supported by the Swiss National Science Foundation (No. ).