Burden-dependent bias of cerebellar reference regions in amyloid PET monitoring of anti-amyloid therapy: implications for reference-region choice
Background The introduction of anti-amyloid therapies has made longitudinal monitoring of amyloid-β burden with PET a clinical necessity, and the reference region used for quantification should behave consistently during rapid amyloid clearance. We evaluated the longitudinal behavior and clinical validity of candidate reference regions in a real-world treated cohort. Methods We analyzed 138 patients treated with lecanemab or aducanumab with paired amyloid PET; 123 also completed the Seoul Cognitive Status Test (SCST). Four reference regions were compared: cerebellar gray matter (CG), whole cerebellum (WC), pons and eroded white matter (WM). Evaluation criteria were burden-dependent net bias between references, change in the reference-region signal itself during treatment (within-scan ratios to pons in the 85 patients imaged with the same tracer at both visits), and the association between Centiloid change (ΔCL) and change in SCST Total, with formal comparison of the reference-specific slopes. Between-reference p values are nominal. Results CG and WC showed burden-dependent net bias relative to WM (quadratic p = 0.021 and 0.013), reproduced with pons as the comparator, so that cerebellar-anchored estimates of clearance diverged most in high-burden patients. No reference region was unchanged during treatment, but the burden dependence of the change was weakest for WM and strongest for the cerebellar references (Spearman ρ − 0.22 vs. − 0.31 to − 0.33), although the slopes did not differ significantly. Only WM-anchored ΔCL was significantly associated with change in SCST Total (β = +0.0067 z per CL, p = 0.029), with the largest slope in every burden subset; the difference from pons was nominally significant ( p = 0.035), that from the cerebellar references was not ( p = 0.06–0.07), and the association was not confirmed in the same-tracer subset. Conclusions Cerebellar references carry a burden-dependent bias that is most consequential in the high-burden patients in whom monitoring matters most. Eroded WM, with the least burden-dependent bias and the closest alignment with cognitive change, appears better suited to tracking longitudinal amyloid change during anti-amyloid therapy, pending validation in larger, treatment-specific cohorts.
Authors
- SEUNG HWAN MOON
- Seongbeom Park (ORCID: https://orcid.org/0000-0002-0759-6826)
- Sang Won Seo (ORCID: https://orcid.org/0000-0002-8747-0122)
- Young Ju Kim (ORCID: https://orcid.org/0000-0003-2927-6724)
- Soo Hyun Cho (ORCID: https://orcid.org/0000-0002-4262-1468)
- Seongmi Kim
- Kichang Kwak
- Duk L. Na
- Jun Pyo Kim
- Hee Jin Kim
Institutions
- Chonnam National University (KR)
- Samsung Medical Center (KR)
- Chonnam National University Hospital (KR)
- Sungkyunkwan University (KR)
Publication Details
- Journal
- Alzheimer s Research & Therapy
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1186/s13195-026-02256-5
- Primary Topic
- Alzheimer's disease research and treatments
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Korea Dementia Research Center
- Korea National Institute of Health