Hippocampal volume in early psychosis: a 2-year longitudinal study

Smaller hippocampal volume is one of the most robust neuroanatomical findings in schizophrenia, with meta-analytic effect sizes indicating substantial reductions and a gradient across illness stages. Cross-sectional studies report the largest hippocampal volume deficits in chronic schizophrenia, moderate deficits in the first two years of schizophrenia, smaller differences in schizophreniform disorder, and near-normal volumes in high-risk individuals, suggesting potential progression. However, longitudinal studies and meta-analyses generally indicate stable hippocampal volume during the first 2–5 years after onset, creating a discrepancy between cross-sectional and longitudinal evidence. The authors propose two explanations: (1) only specific hippocampal subregions (especially anterior regions and CA subfields) may change in early psychosis and could be obscured when analyzing total hippocampal volume; and (2) early psychosis is diagnostically heterogeneous, and hippocampal volume reductions may be confined to individuals whose illness progresses from schizophreniform disorder to schizophrenia. Prior reports suggest anterior hippocampus and CA subfields are preferentially affected in early psychosis and high-risk converters, whereas chronic schizophrenia shows broader anterior and posterior involvement. This study tests the primary hypothesis that anterior hippocampal volume, particularly CA subfields, decreases over two years following illness onset. A secondary hypothesis is that longitudinal hippocampal volume changes are most pronounced in individuals whose diagnosis progresses to schizophrenia, compared to those who remain diagnosed with schizophreniform disorder.

Design: Prospective 2-year longitudinal MRI study with up to four scans per participant (~8-month intervals; median follow-up times: ~8, 16, and 23 months). Participants: N=137 enrolled (72 early psychosis [EP], 65 healthy controls [HC]); baseline MRI passing QC available for 63 EP and 63 HC; study completers: 56 EP (89%) and 52 HC (83%). EP participants were in the first 2 years of a non-affective psychotic disorder; average duration of psychosis <7 months at recruitment. Groups were matched on age, gender, race, and parental education. Diagnostics and clinical measures: Diagnoses via SCID (DSM-IV-TR) with consensus by a psychiatrist; PANSS used to assess symptom severity at scanning; onset of psychosis determined with the Symptom Onset in Schizophrenia Inventory (SOS); duration of psychosis calculated from SOS onset to enrollment; duration of untreated psychosis from SOS onset to first antipsychotic; antipsychotic exposure quantified as chlorpromazine equivalents (no first-generation antipsychotics); premorbid IQ via WTAR; cognition via SCIP at baseline and 2-year follow-up. Imaging acquisition: 3D T1-weighted MRI on one of two identical 3T Philips Intera Achieva scanners with 32-channel head coil (voxel size 1 mm^3; FOV 256 mm^2; 170 slices; gap 0 mm; TE 3.7 ms; TR 8.0 ms). Images were visually checked for motion/artifacts. Imaging processing: Freesurfer 6 longitudinal pipeline for hippocampal subfield segmentation using a probabilistic atlas; Bayesian within-subject template to reduce error and enhance sensitivity to change. Composite subfield definitions created for CA (cornu ammonis), DG (dentate gyrus), and Sub (subiculum) to enhance generalizability across segmentation protocols. Volumes measured separately for anterior (head) and posterior (body) hippocampal regions; tail not analyzed due to limited subfield segmentation reliability and prior evidence of anterior-limited differences. Quality control removed all data from 5 EP and 1 HC; additionally excluded some regional volumes (4 EP and 2 HC anterior; 1 EP and 1 HC posterior). Volumes averaged across hemispheres for region/subfield analyses to reduce model parameters. Statistical analysis: Linear mixed-effects models (R: lme4, emmeans, car). Covariates: estimated intracranial volume, scanner, baseline age, sex; random effect: participant. Pre-registered power analysis indicated 80% power to detect d=0.46 in primary longitudinal analysis. Model 1 tested overall hippocampal volume difference and stability over time with fixed effects Group (EP vs HC), Hemisphere (left/right), and Time (months from baseline). Model 2 tested regional/subfield effects with fixed effects Group, Region (anterior/posterior), Subfield (CA/DG/Sub), and Time, including interactions; multiple-comparison adjustments via Bonferroni; follow-up effect sizes as Cohen’s d. Exploratory associations with clinical/cognitive measures described in supplement. Trajectory analysis: Defined EP subgroups by diagnostic course among completers—(1) SZ stable (schizophrenia at entry; N=16), (2) SZF progression (baseline schizophreniform or bipolar with psychotic features progressing to schizophrenia at 2 years; N=28 total), and (3) SZF stable (maintained schizophreniform; N=14). HC served as a separate group. A trajectory × region × subfield interaction remained significant with or without inclusion of four bipolar-to-schizophrenia converters (F6,2043 = 2.17, p = 0.04).

- Overall hippocampal volume: Early psychosis participants had significantly smaller hippocampal volumes than healthy controls (main effect of Group: F1,155 = 13.08, p < 0.001), with no evidence that this difference changed over the 2-year follow-up (Group × Time: F1,484 = 0.26, p = 0.61); no strong lateralization (Group × Hemisphere: F1,482 = 2.26, p = 0.13). - Regional and subfield specificity: Volume deficits were more pronounced in the anterior hippocampus than the posterior region. Among subfields, the cornu ammonis (CA) showed the largest and most consistent deficit relative to DG and Sub; these subfield-specific deficits were stable over time (no progressive decline over 2 years). - Prognostic value and illness trajectory: Baseline anterior CA volume was smaller in individuals who were diagnosed with schizophrenia at 2-year follow-up, but not reduced in those who maintained a diagnosis of schizophreniform disorder. A trajectory × region × subfield interaction was significant (F6,2043 = 2.17, p = 0.04), supporting differential subregional involvement by illness course. Overall, smaller hippocampal volume—particularly anterior CA—appears prognostic of progression to schizophrenia rather than diagnostic of psychosis per se. - Retention and sampling: High retention (EP 89%, HC 83%) with up to four time points per participant strengthened longitudinal inferences.

The findings reconcile discrepancies between cross-sectional and longitudinal literatures by demonstrating that hippocampal abnormalities in early psychosis are regionally specific and linked to illness trajectory rather than reflecting rapid progressive atrophy across the entire structure. Although cross-sectional work suggests a gradient of hippocampal volume loss with illness chronicity, this study shows that within the first two years after onset, total hippocampal volume is stable. Critically, the deficits localize to the anterior hippocampus—especially the CA subfields—aligning with emerging models of anterior hippocampal dysfunction in early psychosis. The prognostic association—smaller baseline anterior CA volume in those who progress to schizophrenia but not in those who remain schizophreniform—indicates that hippocampal subfield volumes may index underlying pathophysiology relevant to illness persistence and conversion. Thus, hippocampal volume is not a general diagnostic marker for psychosis but may help predict clinical outcomes and guide individualized early interventions. These results underscore the importance of examining hippocampal subregions and accounting for diagnostic trajectories when interpreting structural brain changes in early psychosis.

This 2-year longitudinal study demonstrates that hippocampal volume deficits in early psychosis are stable overall but regionally specific, with the anterior hippocampus—particularly CA subfields—most affected. Baseline anterior CA reductions are associated with progression to schizophrenia, suggesting hippocampal subfield measures have prognostic value for illness course. The work advances understanding of early psychosis by clarifying that rapid progressive global hippocampal atrophy is unlikely in the initial years post-onset and highlights the need to focus on subregional pathology and clinical trajectories. Future research should validate prognostic utility in larger, multi-site cohorts, integrate multimodal biomarkers (e.g., functional and molecular imaging), and extend follow-up beyond two years to capture later-stage progression.

- Hippocampal tail was not analyzed due to segmentation limitations, potentially missing posterior changes. - Automated subfield segmentation, while using a validated longitudinal pipeline, may still be sensitive to atlas assumptions and image resolution; composite subfields were used to enhance comparability but may obscure finer-grained subfield effects. - Hemispheric volumes were averaged for region/subfield analyses to reduce parameters, which may reduce sensitivity to lateralized effects. - Follow-up duration was limited to approximately two years; later progressive changes may not be captured. - Although scanner model was identical and included as a covariate, use of two scanners may introduce subtle site-related variance. - Medication exposure was quantified and first-generation antipsychotics were not used; however, potential medication effects on hippocampal volume cannot be fully excluded.