Psychotic disorders are prevalent, with an estimated 23.6 million cases worldwide, (1) and are associated with substantial clinical, personal, economic and societal burden (2). Under standard care, treatment is not initiated until the first episode of psychosis (FEP), which is associated with suboptimal clinical outcomes (2). Preventive approaches, such as intervening during the clinical high risk for psychosis (CHR-P) state (3), can alter the course of the disorder, improve long-term outcomes and reduce burden (2, 4).

To target interventions for CHR-P individuals at the highest risk of developing psychosis, accurate prediction of FEP onset is essential. Clinical interviews can be used to determine whether individuals meet CHR-P criteria, but while these have high specificity (i.e., adept at ruling out psychosis risk) their sensitivity is sub-optimal (i.e., many of those meeting CHR-P criteria will not develop psychosis) (5–7). For those who do meet CHR-P criteria, there is some refinement of prognostic risk prediction through the three CHR-P subgroups [attenuated psychotic symptoms (APS), brief limited intermittent psychotic symptoms (BLIPS) and genetic risk and deterioration (GRD)]. BLIPS is a relatively uncommon subgroup (10% of all CHR-P individuals) (8), where individuals experience a psychotic episode that spontaneously resolves within a week without antipsychotic treatment (9). BLIPS carry the highest level of psychosis risk (38% transition to psychosis within 2 years)(8) compared to APS (85% of CHR-P individuals; of whom 24% transition within 2 years) (8) and GRD (5% of CHR-P individuals; of whom 8% transition within 2 years) (8). However, it is currently not possible to stratify using clinical criteria beyond this, limiting the ability to accurately predict outcomes.

For more effective risk stratification, information is needed from additional sources, particularly from those directly relating to underlying neuropathology (i.e., biomarkers) as these will likely capture different elements of disorder progression. Alterations in frontal, hippocampal and striatal functioning reliably appear, persist and progress from the CHR-P state (10–15) to FEP (15–19) and chronic schizophrenia (13, 15, 20–22). Capturing these neurobiological changes could be informative in refining estimates of psychosis risk and enhancing risk stratification. Given the associated increase in psychosis risk estimates (23), we hypothesize that BLIPS may display more pronounced neurobiological differences compared to APS and GRD. However, there are currently no published studies investigating neurobiological differences between CHR-P subgroups due to the high prevalence of APS individuals in CHR-P samples.

Pseudo-continuous arterial spin labelling (ASL) allows for indirect measurement of regional, resting neuronal activity through regional cerebral blood flow (rCBF) (24), a potential biomarker for psychosis onset (10–12, 15). This approach exploits the intimate relationship between neuronal activity and its regional blood supply as a result of the phenomenon of neurovascular coupling. Modern pCASL sequences such as the one used in this study, are even more sensitive to neurovascular coupling as they can be tailored to reflect arterial blood flow from the vascular domain (arterioles and capillaries), where this phenomenon occurs.

CHR-P individuals display elevated dopamine signaling in the midbrain and striatum (25–28), as well as an altered relationship between hippocampal activation and striatal dopamine functioning (29, 30). This may in part stem from hippocampal gray matter abnormalities that have been consistently shown in CHR-P and FEP (31–34). CHR-P individuals display increased striatal rCBF and reduced frontal rCBF compared to clinical controls (15), with striatal rCBF correlating with attenuated positive symptom scores in CHR-P individuals (15). Similarly, CHR-P individuals have elevated rCBF in hippocampus and basal ganglia (10, 11), with hippocampal hyper-metabolism thought to drive downstream striatal dopamine dysfunction—a core mechanism underlying psychotic symptoms (13, 35, 36). There is evidence of hippocampal rCBF decreasing over time in CHR-P individuals who subsequently remitted from the CHR state (10). Additionally, there is a strong correlation between hippocampal rCBF and prefrontal GABA levels in CHR-P individuals who transition to psychosis, with no correlation in those who do not transition (12).

There are also distinct neurobiological differences between medicated and unmedicated FEP patients, with medicated patients showing increased rCBF in the striatum and reduced rCBF in the frontal cortex compared to clinical controls (15), while unmedicated FEP show reduced rCBF in the frontal cortex and no change in striatal rCBF compared to healthy controls (37).

Due to the low proportion of CHR-P individuals who meet BLIPS criteria in CHR-P samples, there has been difficulty in recruiting samples that are large enough to be sufficiently powered to investigate differences in rCBF between APS and BLIPS. Therefore, in this study, we combined data from four studies to investigate differences in global and regional rCBF between CHR-P subgroups. Due to the very small proportion of individuals meeting GRD criteria and this subgroup having psychosis risk that is not significantly different from HCs (8), analyses were restricted to APS and BLIPS subgroups. As rCBF in frontal, hippocampal and striatal regions appear to positively correlate with psychosis risk (10, 14, 24, 25), we expected rCBF in these regions in BLIPS individuals to be higher than in APS. As secondary hypotheses, we expected healthy controls (HCs) to have reduced rCBF in these regions compared to APS and BLIPS, with FEP patients having reduced rCBF compared to HCs, APS and BLIPS.

To our knowledge, this is the largest neuroimaging dataset of CHR-P individuals meeting BLIPS criteria to date, and was achieved by combining data from four independently conducted studies. As such, this is the first study to compare rCBF between APS and BLIPS in order to investigate the potential of rCBF for risk stratification of CHR-P. However, our results do not provide evidence that rCBF, whether based on global gray matter, within ROIs strongly implicated in psychosis pathophysiology or within whole-brain voxel-wise analyses, is able to discriminate between BLIPS and APS at a group level. There are also no differences between these groups and HCs and FEP, in contrast to previous findings (10, 11, 37). In fact, these findings present moderate-to-weak evidence for no rCBF differences between APS and BLIPS across all ROIs.

Our results suggest that the APS and BLIPS CHR-P subgroups may not be neurobiologically distinct in terms of rCBF. However, the evidence for the null hypothesis ranges between ROIs from weak (e.g., left hippocampus) to moderate (e.g., right frontal cortex) so our findings may not be conclusive. The weakest evidence was for rCBF in the left hippocampus, which has been most strongly implicated in neurobiological contrasts between CHR-P and HCs (10–12). Previous research investigating neurobiological differences between subgroups of the CHR-P population have been mixed. For example longitudinal hippocampal rCBF changes are associated with remission from the CHR-P state (10) but baseline hippocampal rCBF does not appear to be associated with transition to psychosis (12). This may indicate that trajectories of rCBF may be more predictive of clinical outcomes, rather than baseline values alone. Similarly, assessing rCBF alone may not be sufficient for distinguishing CHR-P subgroups. For example, the interaction between prefrontal GABA levels and hippocampal rCBF is significantly different in CHR-P individuals who transition and those who do not (12). Data from additional time points or imaging modalities could allow for greater discrimination between CHR-P sub-populations, including APS and BLIPS.

Sociodemographic differences between our sample and those of previous studies may account for some differences in the results. As BLIPS is a less common CHR-P subgroup compared to APS, all matching between subgroups was in relation to the BLIPS sample. This means that the characteristics of HC, APS and FEP included in this study were selected to match the BLIPS cohort and are therefore not necessarily representative of their underlying populations. For example, despite the proportion of males historically being generally similar across APS and BLIPS (8, 23, 42, 64), due to one study (contributing n = 6 BLIPS; 30% of the total sample) only recruiting males, 90% of our BLIPS sample was male, which is substantially higher compared to previous studies (closer to 50%) (10–12, 15). While greater proportion of males is associated with greater psychosis risk in BLIPS samples (65), this is also the case in CHR-P more broadly (66). Enrichment of psychosis risk through other sociodemographic factors outside of CHR-P subgroup (e.g., migrant status, childhood trauma, urbanicity, parental severe mental illness, etc) (67) could have reduced the likelihood of finding evidence of neurobiological between-group differences and may reduce generalisability more broadly. Sex can similarly impact rCBF, with rCBF generally higher across all brain structures in females compared to males (68). By matching groups for age and sex, our analyses reduced the potential impact of sex, as well as age effects (68). While explicit matching has not been used in previous CHR-P research, age and sex have consistently been included in sensitivity analyses as covariates and significant differences between CHR-P and HCs in rCBF were retained (10–12, 15). Moreover, differences in rCBF remained non-significant when sex was included in the model, emphasizing that sex likely did not impact our results.

Two of the most robust predictors of psychosis onset in CHR-P individuals are attenuated positive psychotic symptom severity and global functioning at baseline, but the pattern of presentation varies between APS and BLIPS (66). The acute onset and spontaneous resolution of BLIPS frank psychotic symptoms (42) could have impacted rCBF levels, reducing the likelihood of finding differences between BLIPS and APS. While BLIPS experience positive psychotic symptoms at higher severity than the attenuated positive psychotic symptoms experienced by APS individuals, these are experienced for 1 week or less before spontaneously resolving (42). The BLIPS individuals were scanned after this period of frank psychotic symptoms had resolved, as indicated by the similarity of CAARMS scores between APS and BLIPS subjects. This is further emphasized by the significantly higher levels of functioning, as measured by the GAF, seen in BLIPS compared to APS subjects in this study. The dynamic nature of rCBF combined with its associations with symptom severity (69) and medication effects (44), which can also fluctuate, mean that scans performed at different times may result in altered perfusion. For example, greater differences may be seen in rCBF if participants were scanned at the peak of symptom severity, however, this would be ethically and logistically challenging. It also could be argued that despite remission from full psychosis that BLIPS still have a greater risk of transition compared to APS so finding neurobiological differences, such as rCBF, would still be plausible. In addition, many BLIPS individuals also experience attenuated psychotic symptoms, meeting criteria for both BLIPS and APS. Such individuals were considered to be solely BLIPS in these analyses. These factors may contribute to the observed similarities between APS and BLIPS in terms of rCBF.

Low rates of transition to FEP could also dilute potential differences in perfusion between APS and BLIPS subgroups. While expected transition rates are 24 and 38% in APS and BLIPS, respectively, (23) only 15% of APS individuals and 10% of BLIPS individuals in our sample developed a psychotic disorder over follow-up. Due to the demands of the study procedures, it may be that the recruited CHR-P individuals were less severely unwell than the CHR-P population average. The low transition rate may have reduced potential differences in rCBF between APS and BLIPS and limited our ability to explore any putative relationship between baseline rCBF and subsequent transition.

Despite the large overall sample, the power was still limited, particularly for the BLIPS subgroup. The literature base exploring rCBF in CHR-P individuals is not yet extensive with only a few papers published (10–12, 15). Due to this and the weak-to-moderate evidence for the null hypothesis and fluctuations in neurobiology over the timecourse of the CHR-P state, future research should aim to investigate larger samples of BLIPS and APS with multiple imaging modalities acquired at multiple time points, which can be achieved through collaboration and harmonization of data through large scale international consortia, e.g., HARMONY incorporating NAPLS (70), PRONIA (71), PSYSCAN (72), and ENIGMA (73). Similarly, recruitment of BLIPS could be improved through expanding CHR-P detection efforts to other brief psychotic conditions (e.g., acute and transient psychotic disorders and brief psychotic episodes), which all have significant overlap with BLIPS (65, 74–77). More consistent changes in cerebral perfusion in psychosis have been seen in studies measuring cerebral blood volume (CBV) compared to rCBF; (78) this may be a promising avenue to explore in future research. A potential limitation of our study may arise from the use of slightly different 3D pCASL acquisition protocols (see Supplementary Methods S1–S3). Since the data originate from multiple studies spanning several years, improvements made to the 3D pCASL pulse sequence allowed us to use a longer post-labelling delay in a relatively small number of the datasets acquired at a later stage (n = 27; 18%). Thus, there may be slightly different sensitivity to tissue perfusion in those data, although we used ComBat to mitigate against this. Moreover, due to a lack of harmonization between study protocols, we were limited in terms of available covariates (e.g., differences in instruments to assess anxiety meant we were unable to include this in statistical models, despite anxiety showing effects on rCBF in past research) (79, 80). However, the current analysis is focused on clinical prediction which needs methods that are robust to between-site differences in scanners, local methodologies, participant populations and other factors. Further to this, our groups were defined on the basis of investigating potential differences between BLIPS and the other groups. We therefore cannot conclude that there are no differences between all groups, only that we were unable to find a statistically significant difference between BLIPS and the other groups.

In conclusion, we have found weak-to-moderate evidence for an absence of difference in global GM, frontal, hippocampal and striatal rCBF between APS and BLIPS. Moreover, we did not find differences in rCBF between BLIPS and HCs or FEP either in whole-brain voxel-wise analysis or ROIs. These results suggest that rCBF alone may not be suitable for risk stratification in CHR-P individuals.

The datasets presented in this article are not readily available because at the time the data was collected it was not routine for participants to be asked for their consent to share data publicly, so this permission was not obtained. Whilst we are in favor of data being open access, no supporting data is available for this study. Requests to access the datasets should be directed to DO (dominic.a.oliver@kcl.ac.uk).

The studies involving human participants were reviewed and approved by National Health Service UK Research Ethics Committee. The patients/participants provided their written informed consent to participate in this study.

PA, AE, PM, and PF-P: conceptualization. DO, CD, FZ, PS, AD, AC, HB, MA, and NC: data processing. DO, FZ, and PS: data analysis. DO: writing–original draft. All authors: interpretation and writing–review and editing. All authors contributed to the article and approved the submitted version.

This study was supported by Wellcome Trust grants to PF-P (215793/Z/19/Z) and OH (094849/Z/10/Z), Medical Research Council (MRC) grants to PM (G0700995) and OH (MC_A656_5QD30_2135), a Maudsley Charity grant (No. 666) to OH, and the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) at South London and Maudsley NHS Foundation Trust and King’s College London (PM, OH, and PF-P).

SJ has received honoraria for educational talks given for Lundbeck, Janssen, and Sunovion. OH has received investigator-initiated research funding from and/or participated in advisory/speaker meetings organized by Angelini, Autifony, Biogen, Boehringer-Ingelheim, Eli Lilly, Heptares, Global Medical Education, Invicro, Janssen, Lundbeck, Neurocrine, Otsuka, Sunovion, Recordati, Roche and Viatris/Mylan. OH has a patent for the use of dopaminergic imaging. OH is a part-time employee and stockholder of Lundbeck A/S. PF-P has received research funds or personal fees from Lundbeck, Angelini, Menarini, Sunovion, Boehringer Ingelheim, Mindstrong, Proxymm Science, outside the current study.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.