Does Restrictive Anorexia Nervosa Impact Brain Aging? A Machine Learning Approach to Estimate Age Based on Brain Structure

Abstract

Background: Anorexia nervosa (AN), a severe eating disorder characterized by extreme weight loss and malnutrition, leads to significant brain structure alterations. This study uses machine learning (ML) to estimate brain age from structural MRI scans and explores brain predicted age difference (brain-PAD) as a potential biomarker in AN.

Methods: Structural MRI scans from female patients aged 10–40 years were collected from two institutes (Boston, USA and Jena, Germany), including acute AN (acAN; n=113), weight-restored AN (wrAN; n=35), and age-matched healthy controls (HC; n=90). An ML model trained on 3487 HC T1-weighted MRI scans (ages 5–45 years) across ten datasets used 377 neuroanatomical features. The model’s performance was evaluated based on mean absolute error (MAE) and correlation with actual age, and it was applied to estimate brain-PAD in the test groups.

Findings: The model achieved an MAE of 2.37 years and a correlation (r=0.88) in HCs. In acAN patients, brain age was overestimated by 2.25 years, indicating advanced brain aging. wrAN patients showed significantly lower brain-PAD than acAN (0.26 years, p=0.0026) but no difference from HC (p=0.98), suggesting that weight restoration largely normalizes brain structure. A significant correlation was found between brain-PAD and BMI in the acAN group (r=-0.291, p=0.005), indicating that lower BMI is linked to greater brain aging.

Interpretation: Our findings indicate that AN patients exhibit older brain ages compared to age-matched controls. However, this difference does not reflect an accelerated aging process, as the discrepancy diminishes in older patients. Notably, weight restoration effectively normalizes brain age, bringing it close to expected levels.

Funding: The study is supported by the Interdisciplinary Center for Clinical Research, Jena University Hospital (AMSP17, MSP19), the German Research Foundation (DFG, SCHU3432/2-1), and the National Institute of Mental Health (NIMH). Similarly, from Harvard Catalyst, The Harvard Clinical and Translational Science Center (NIH Award #UL1 RR025758), and the National Institute of Health (NIH) (grant number: R01MH103402 and K24MH135189).

Declaration of Interest: KTE: Royalties or licenses: Cambridge University Press. Consulting fees: Equip Health, Inc. Participation on a Data Safety Monitoring Board or Advisory Board: Stanford University/University of California, San Francisco, University of Pittsburgh, Cincinnati Children’s Hospital. MM: Grants or contracts from any entity: NIH (grant number: 5R01DK124223-04 (NIDDK/NIH), 1R01HD114914-01 (NICHD/NIH), 5R01MH116205-05 (NIMH/NIH), 5R01DK122581 (NIDDK/NIH), 5R01MH103402 (NIMH/NIH), AR220116 (DoD). Royalties or licenses: UpToDate and Medscape. Consulting fees: Regeneron, Kyss Therapeutics, and Lumos Pharmaceuticals. Payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events: University of Michigan. Support for attending meetings and/or travel: Pediatric Endocrine Society, Endocrine Society, International Meeting of Pediatric Endocrinology, European Society of Pediatric Endocrinology, Human Growth Foundation, PESTOLA. Receipt of equipment, materials, drugs, medical writing, gifts or other services: Tonix Pharmaceuticals. JJT: Royalties or licenses: Harvard Health Publications, Cambridge University Press, and Oxford University Press. Consulting fees: Equip Health. Leadership or fiduciary role in other board, society, committee or advocacy group, paid or unpaid: President of the Eating Disorders Research Society and Past-President of the Academy for Eating Disorders. KKM: Grants or contracts from any entity: NIH, Massachusetts General Hospital, Amgen (investigator-initiated grant). Stock or stock options: Bristol-Myers Squibb (BMS), General Electric, Boston Scientific, and Becton Dickinson. Receipt of equipment, materials, drugs, medical writing, gifts or other services: Medication at no cost from Amgen. All others have nothing to declare.

Ethical Approval: All participants (and legal guardians for participants under 18) gave informed written consent. The study was approved by the local Ethics Committee of JUH and conducted according to the Declaration of Helsinki (2013).

Keywords: Eating Disorder, Acute anorexia nervosa (acAN), Weight-restored AN (wrAN), Brain aging, Structural MRI, Machine learning (ML), Brain-PAD