Differential longitudinal changes of hippocampal subfields in patients with anorexia nervosa

Abstract

Background: Anorexia nervosa (AN) is a mental disorder characterized by dietary restriction, fear of gaining weight, and distorted body image. Recent studies indicate that the hippocampus, crucial for learning and memory, may be affected in AN, yet subfield-specific effects remain unclear. We investigated hippocampal subfield alterations in acute AN, changes following weight restoration, and their associations with leptin levels.

Methods: T1-weighted magnetic resonance imaging scans were processed using FreeSurfer. We compared 22 left and right hemispheric hippocampal subfield volumes cross-sectionally and longitudinally in females with acute AN (n = 165 at baseline, n = 110 after partial weight restoration), healthy female controls (HCs; n = 271), and females after long-term recovery from AN (n = 79) using linear models.

Results: We found that most hippocampal subfield volumes were significantly reduced in patients with AN compared with HCs (~-3.9%). Certain areas such as the subiculum exhibited no significant reduction in the acute state of AN, while other areas, such as the hippocampal tail, showed strong decreases (~-9%). Following short-term weight recovery, most subfields increased in volume. Comparisons between participants after long-term weight-recovery and HC yielded no differences. The hippocampal tail volume was positively associated with leptin levels in AN independent of body mass index.

Conclusions: Our study provides evidence of differential volumetric differences in hippocampal subfields between individuals with AN and HC and almost complete normalization after weight rehabilitation. These alterations are spatially inhomogeneous and more pronounced compared with other major mental disorders (e.g. major depressive disorder and schizophrenia). We provide novel insights linking hypoleptinemia to hippocampal subfield alterations hinting towards clinical relevance of leptin normalization in AN recovery.

Keywords: FreeSurfer; anorexia nervosa; hippocampus; leptin; longitudinal study.

Fig. 1

Hippocampal (subfield) volumes in patients with acute anorexia nervosa (AN) vs healthy controls (HCs). Bar graphs with error bars for study groups AN (n = 165) and HCs (n = 271) displaying adjusted means (estimated marginal means [EMMs], mm³) ± SEM of individual whole hippocampus and hippocampal subfield volumes in separate brain hemispheres, covarying for age at the date of research (linear and quadratic effects) and estimated total intracranial volume. FDR‐q: P‐values were multiple testing–adjusted using false discovery rate (FDR) across all hippocampal subfields (whole hippocampus adjusted separately using FDR). Significance levels for volume differences between study groups are stated as: ***q < 0.001; **q < 0.01; *q < 0.05; ns, nonsignificant. Effect size statistics are provided as partial η². lh, left brain hemisphere; rh, right brain hemisphere; GLM, general linear model.

Fig. 2

Associations between significantly altered hippocampal subfield volumes and leptin concentrations in anorexia nervosa (AN). Scatter plots with individual data points, linear regression lines, and 95% confidence intervals around the regression line (gray band) in the AN study group (plasma leptin measurement available in 142 of 165 patients) displaying associations between individual hippocampal subfield volumes that were significantly altered in patients with AN vs healthy controls (HCs) and logarithmically transformed (base 10) (log₁₀) leptin concentration (leptin values lower than the limit of detection were multiply imputed using censored likelihood multiple imputation, associations were examined via robust linear models [RLMs] applying M estimation and Huber weighting for fitting via iterated reweighted least squares). Standardized residuals of hippocampal subfield volumes are plotted after adjustment of raw volume measures for age at the date of research (linear and quadratic effects) and estimated total intracranial volume using robust multiple linear regression. RLM statistics are provided as t values (unstandardized β divided by its standard error), unadjusted P‐value (computed via robust Wald F test), and effect size estimate partial η². lh, left brain hemisphere; rh, right brain hemisphere.

Fig. 3

Hippocampal subfield comparison between patients with acute, short‐ and long‐term treated anorexia nervosa (AN) and healthy controls (HCs). Bar graph with error bars for study groups with acute AN at time point 1 (AcAN‐TP1) and acute AN at time point 2 (AcAN‐TP2) (n = 110), long‐term recovered from AN (RecAN) (n = 79), and HCs (n = 271) displaying estimated marginal mean (EMM, mm³) ± SEM of individual whole hippocampus and hippocampal subfield volumes in separate brain hemispheres, covarying for age at the date of research (linear and quadratic effects) and estimated total intracranial volume. Variance and covariance parameters estimated via restricted maximum likelihood, and Satterthwaite method applied for degrees of freedom., FDR‐q: P‐values were multiple testing–adjusted using false discovery rate (FDR) across all assessed pairwise contrasts (AcAN‐TP2–AcAN‐TP1, AcAN‐TP2–HCs, RecAN–HCs, and AcAN‐TP2–RecAN) per hippocampal subfield and across all hippocampal subfields. Significance levels are stated as: ***q < 0.001; **q < 0.01; *q < 0.05; ns, nonsignificant. Effect size statistics for linear mixed effects (LME) model contrasts are provided as Cohen d. CA1, cornu ammonis 1; CA3, cornu ammonis 3; CA4, cornu ammonis 4; DG, dentate gyru; GC, granule cell; HATA, hippocampus amygdala transition area; lh, left brain hemisphere; ML, molecular layer; rh, right brain hemisphere.