Longitudinal hippocampal subfield development associated with psychotic experiences in young people

Psychotic experiences (PE) occur in up to 8.1% of young people without a formal psychotic disorder but are linked to higher risk of later psychosis, other psychopathology, and subtle neurocognitive impairments. Prior work, including longitudinal analyses in the same cohort, has shown hippocampal volume reductions associated with PE and schizotypy, paralleling findings in individuals at clinical high risk and early psychosis. As imaging resolution improved, interest shifted from global hippocampal measures to cytoarchitecturally distinct subfields (CA1–4, dentate gyrus, subiculum complex), with meta-analyses suggesting subfield analyses may reveal subtle abnormalities missed by whole-structure approaches. Reviews indicate CA1 deficits are prominent in prodromal psychosis and may progress across subfields with illness development. The present study aims to characterize longitudinal developmental trajectories of hippocampal subfields over approximately six years in adolescents reporting PE, hypothesizing significant deficits in CA1 volumetric development in the PE group compared to controls, thereby providing insight into neural mechanisms of psychosis vulnerability absent confounds of diagnosis or antipsychotic exposure.

The literature implicates hippocampal abnormalities across the psychosis spectrum, with CA1 reductions frequently reported in ultra-high-risk and early psychosis samples and thought to be more severe in converters. Some models propose an initial CA1 dysfunction with later spread across subfields as illness progresses. Meta-analytic and longitudinal studies suggest that reliance on whole-hippocampal measures may obscure subfield-specific changes. Adolescents with schizotypy or subclinical symptoms have shown hippocampal size and growth deficits, aligning with prodromal-stage findings. However, prior work largely involves cross-sectional, treatment-seeking cohorts, emphasizing the need for longitudinal subfield analyses in community youth with PE to clarify developmental trajectories and localization of hippocampal changes.

Design and participants: Community sample from the Adolescent Brain Development (ABD) study. Initially 211 youths aged 11–13 years recruited from schools in Dublin and Kildare, Ireland; the initial 20% of recruits were enriched 2:1 for elevated Adolescent Psychotic Symptom Screener scores (≥2). All were invited to MRI at baseline (1–3 years later; mean 2). MRI completers: 100 at baseline, 69 at follow-up 1 (~2 years later), 55 at follow-up 2 (~4 years later). Final analytic sample included 78 participants with at least one MRI: 33 who met consensus criteria for definite PE at least once (PE group) and 45 with no PE at any time (controls). None met criteria for psychotic disorder or were taking antipsychotics; three PE participants reported antidepressant use by final follow-up. Ethics approval obtained; assent/consent procedures followed. Clinical assessment: Baseline and follow-up 1 used K-SADS interviews with adolescent and parent; follow-up 2 used SCID-5 with participants. The psychosis sections were supplemented with SOCRATES items to assess PE. PE classifications were determined by expert consensus (two psychiatrists, one psychologist) at each timepoint. SIPS total scores collected at baseline and follow-up 2. Childhood adversity (baseline) and lifetime DSM-5 diagnoses (excluding simple phobias; collected at follow-up 2 only) were treated as dichotomous covariates. MRI acquisition: 3T Philips Achieva scanner (TCIN, Dublin). T1-weighted fast field echo 3D transverse: TE/TR = 8.4/3.9 ms, flip angle 8°, 256×256 matrix, 180 slices at 0.9 mm thickness, FoV 230 mm, scan time 5:44 min. Image processing: FreeSurfer v6.0 longitudinal stream with developmental toolbox for hippocampal reconstruction, volumetric segmentation, and parcellation; hippocampal subfields extracted: subiculum, presubiculum, CA1, CA2/3 (combined due to small CA2), and CA4/DG (CA4 functionally included in DG). Quality control followed ENIGMA protocols; outliers defined as values outside Q1–1.5 IQR or Q3+1.5 IQR. Three participants removed (one PE with structural abnormality; one PE and one control with corrupted data). Statistics: Linear mixed-effects (LME) models (R 4.0.2, lme4 v1.1.23) per ROI, accommodating missing data and variable intervals. Fixed effects: group (PE vs control), time since baseline (months), Group×Time interaction (when retained), intracranial volume (ICV), baseline age (months), gender, handedness; lifetime DSM-5 diagnosis included due to group difference; adversity not included (no group difference). Random effects: subject-specific random intercepts (random slopes not retained after model selection). Model selection used likelihood ratio tests and AICC in a top-down approach. Multiple comparisons controlled via FDR across ROIs. For ROIs with uncorrected group effects, associations with SIPS total scores over time in the PE group were tested with LME using the same covariates. Primary outcomes: Longitudinal volumes of bilateral hippocampal subfields (CA1, CA2/3, CA4/DG, subiculum, presubiculum).

Demographics: No between-group differences in number of scans, ages at timepoints, handedness, socioeconomic status, or baseline adversity. PE group had higher proportion male (66.7% vs 42.2%, p = 0.033) and more lifetime DSM-5 diagnoses among those assessed at follow-up 2 (76% vs 30.4%, p = 0.001). All participants were antipsychotic-naïve. Model selection: Random intercept models fit best; random slopes were not retained. Full fixed-effects model (including Group×Time) fit best for right CA1 and right CA2/3; reduced model (no interaction) for left presubiculum and right subiculum; null model for remaining ROIs. Longitudinal Group×Time interactions (uncorrected): - Right CA1: Group×Time B = −0.37 (SE 0.18), t = −2.082, p = 0.043; FDR p = 0.21; Cohen’s d = −0.61. Pattern: control volumes increased over time; PE volumes remained relatively stable/decreased. - Right CA2/3: Group×Time B = −0.22 (SE 0.085), t = −2.58, p = 0.013; FDR p = 0.065; Cohen’s d = −0.79. Pattern: control volumes increased; PE volumes decreased. Approached significance after FDR with a large effect size. Between-group main effects (time-independent; uncorrected): - Right subiculum smaller in PE vs controls: B = −24.07 (SE 11.43), t = −2.1, p = 0.039; FDR p = 0.19; d = −0.52. - Left presubiculum smaller in PE vs controls: B = −23.69 (SE 10), t = −2.37, p = 0.021; FDR p = 0.10; d = −0.59. Multiple comparison correction: None of the above effects survived FDR correction. Other subfields: No significant effects in right presubiculum, left subiculum, or bilateral CA4/DG. Symptoms: No significant associations between subfield volumes and SIPS total scores over time within the PE group (note SIPS available only at baseline and follow-up 2).

The study investigated whether adolescents with psychotic experiences show divergent developmental trajectories in hippocampal subfields compared to controls. Before multiple-comparison correction, PE was associated with decreasing or static volumes over time in right CA1 and right CA2/3, contrasting with increases in controls. Additionally, lower time-independent volumes in right subiculum and left presubiculum were observed in the PE group. Although none of these effects survived FDR correction, moderate-to-large effect sizes suggest potentially meaningful, subtle differences that may reach statistical significance with larger samples. The strongest longitudinal effect in right CA2/3 partially diverges from the hypothesis emphasizing CA1, yet aligns with literature in clinical psychosis indicating involvement of CA2/3 and the subiculum complex. The pattern may reflect diffuse hippocampal alterations rather than a single subfield-specific deficit, consistent with prior findings of global right hippocampal differences in this cohort. The observed trajectories in PE could indicate developmental delays in hippocampal maturation, potentially related to atypical synaptic pruning and grey matter maturation during adolescence. Lack of association with total symptom severity may reflect subtle neuroanatomical differences not captured by coarse symptom measures or limited timepoints. Strengths include antipsychotic-naïve, population-based youth sample, consensus-based PE classification, consistent scanner, and longitudinal FreeSurfer processing; however, segmentation accuracy concerns and limited covariate data warrant caution. Overall, results support hippocampal involvement—especially CA1 and CA2/3—in psychosis vulnerability mechanisms in adolescents with PE.

This first longitudinal study of hippocampal subfields in adolescents with psychotic experiences suggests subtle, developmentally divergent trajectories in right CA1 and CA2/3 and reduced volumes in the subiculum complex, though effects did not survive FDR correction. Findings reinforce the centrality of hippocampal circuitry in psychosis vulnerability and point toward potentially diffuse subfield involvement rather than a single focal abnormality. Future research should employ larger samples to increase power, refine subfield segmentation methods, incorporate richer longitudinal symptom and environmental data, and examine differences between transient and persistent PE, as well as spectrum-based approaches to psychosis phenotyping. Integrating measures of stress biology and cognitive mechanisms may clarify how hippocampal subfield development contributes to psychosis risk and resilience.

- Multiple-comparison correction: No subfield findings survived FDR correction, indicating risk of false positives in uncorrected results. - Limited symptom and diagnostic covariate data: SIPS scores available only at baseline and follow-up 2; lifetime DSM-5 diagnoses collected only at follow-up 2; childhood adversity measured only at baseline, potentially missing dynamic influences. - Sample size and power: Modest longitudinal sample with attrition; insufficient numbers to compare transient versus persistent PE or transition to psychosis. - Phenotypic categorization: Exclusion of “weak” PE limits spectrum-based analyses and may reduce sensitivity to gradations in PE severity. - Segmentation concerns: Accuracy limitations of FreeSurfer v6.0 hippocampal subfield segmentation have been noted in the literature. - Demographic imbalance: Higher proportion of males and more lifetime DSM-5 diagnoses in the PE group could confound results despite covariate adjustment.