ADRD Biomarker Heterogeneity: The Zetterberg Framework
Overview
At [AD/PD 2026](/conferences/adpd-2026), Henrik Zetterberg (University of Gothenburg, Sahlgrenska Academy) delivered a keynote framing disease heterogeneity as the central, unresolved challenge in [Alzheimer Disease and Related Dementias (ADRD) biomarker research](/biomarkers/blood-based-biomarkers-neurodegeneration)[@zetterberg2026]. His core message: neurodegeneration is not a single disease but a spectrum of distinct biological processes that manifest clinically as similar syndromes. Effective biomarkers must account for this diversity at genetic, molecular, cellular, and clinical levels.
Zetterberg's framework builds on a decade of fluid biomarker work from his group and others, and aligns with emerging consensus that the field must move beyond simple amyloid-positive vs. amyloid-negative dichotomies toward multidimensional, patient-specific biological profiles.
The Heterogeneity Problem
Why Standard Biomarkers Fail
Traditional biomarker approaches assume a relatively homogeneous disease process: amyloid accumulation drives tau pathology, which drives neurodegeneration, which drives cognitive decline. This linear cascade model underpins the [amyloid cascade hypothesis](/mechanisms/amyloid-cascade-hypothesis)[@hardy1992][@karran2011].
However, multiple independent cohorts have demonstrated that:
...ADRD Biomarker Heterogeneity: The Zetterberg Framework
Overview
At [AD/PD 2026](/conferences/adpd-2026), Henrik Zetterberg (University of Gothenburg, Sahlgrenska Academy) delivered a keynote framing disease heterogeneity as the central, unresolved challenge in [Alzheimer Disease and Related Dementias (ADRD) biomarker research](/biomarkers/blood-based-biomarkers-neurodegeneration)[@zetterberg2026]. His core message: neurodegeneration is not a single disease but a spectrum of distinct biological processes that manifest clinically as similar syndromes. Effective biomarkers must account for this diversity at genetic, molecular, cellular, and clinical levels.
Zetterberg's framework builds on a decade of fluid biomarker work from his group and others, and aligns with emerging consensus that the field must move beyond simple amyloid-positive vs. amyloid-negative dichotomies toward multidimensional, patient-specific biological profiles.
The Heterogeneity Problem
Why Standard Biomarkers Fail
Traditional biomarker approaches assume a relatively homogeneous disease process: amyloid accumulation drives tau pathology, which drives neurodegeneration, which drives cognitive decline. This linear cascade model underpins the [amyloid cascade hypothesis](/mechanisms/amyloid-cascade-hypothesis)[@hardy1992][@karran2011].
However, multiple independent cohorts have demonstrated that:
- Not all amyloid-positive individuals progress at the same rate — some remain cognitively stable for years
- Non-amyloid pathologies (alpha-synuclein, TDP-43, vascular injury, neuroinflammation) co-occur in many patients, driving clinical outcomes independently of amyloid
- Genetic heterogeneity — APOE ε4 carriers, LRRK2 G2019S carriers, and sporadic cases show distinct biomarker trajectories
- Demographic heterogeneity — Age, sex, education, and vascular risk profiles alter both biomarker levels and clinical expression
Evidence for Multiple Parallel Pathways
Key evidence for heterogeneity in ADRD comes from several converging lines:
Post-mortem studies — Up to 50% of clinically diagnosed Alzheimer's patients have mixed pathology at autopsy (co-pathology of amyloid, tau, alpha-synuclein, vascular, or TDP-43)[@petersen2024]
Fluid biomarker studies — plasma p-tau217 and p-tau231 distinguish amyloid-positive patients, but their prognostic value diminishes when non-AD pathologies are present[@hansson2024]
Clinical trial failures — Monotargeting amyloid has yielded modest results in only early-stage, carefully selected populations[@van2023]
Genetic studies — GWAS has identified >40 loci affecting AD risk, pointing to diverse biological pathways (immune, lipid metabolism, endosomal sorting, synaptic function) beyond amyloid processing[@cummings2024]The Zetterberg Heterogeneity Framework
Zetterberg's framework identifies four layers of heterogeneity that biomarker development must address:
Mermaid diagram (expand to render)
Layer 1: Genetic Heterogeneity
ADRD risk is driven by hundreds of genetic variants with varying effect sizes. The framework distinguishes:
- Amyloid-directed genetics: APOE ε4 (strong risk), CLU, PICALM, BIN1
- Immune/microglial genetics: TREM2, PLCG2, INPP5D, CSF1R
- Synucleinopathy genetics: LRRK2 G2019S, SNCA multiplications, GBA variants
- Tau/FTD genetics: MAPT, GRN, C9orf72
- Polygenic risk scores (PRS): Composite scores aggregate hundreds of variants, but different PRS components predict different biomarker trajectories
Implication for biomarkers: Genetic background determines which fluid biomarkers are most informative. A TREM2 variant carrier may show biomarker changes driven by microglial activation that precede amyloid changes by years.
Layer 2: Molecular Heterogeneity
Multiple distinct proteinopathies can drive neurodegeneration independently or in combination:
- Amyloid-beta (Aβ) — plaques, soluble oligomers
- Phosphorylated tau (p-tau181, p-tau217, p-tau231) — spreading pathology
- Alpha-synuclein (α-syn) — Lewy bodies, neuronal inclusions
- TDP-43 (TAR DNA-binding protein 43) — limbic-predominant age-related TDP-43 encephalopathy (LATE)
- Vascular pathology — white matter lesions, microinfarcts
- Neuroinflammation — elevated GFAP, IL-6, sTREM2
Each pathological species has its own optimal fluid biomarker (e.g., plasma Aβ42/40 for amyloid, plasma p-tau217 for tau, CSF α-syn RT-QuIC for synuclein). When multiple pathologies co-occur, biomarker interpretation requires deconvolution.
Layer 3: Cellular Heterogeneity
Even within a single molecular pathology, different cell types respond differently:
- Microglial states: homeostatic (TMEM119+, P2RY12+) → disease-associated (DAM/MGND, IRM), transitioning through intermediate states[@zetterberg2026]
- Astrocyte reactivity: A1 (neurotoxic) vs. A2 (neuroprotective) phenotypes
- Synaptic loss: Pre-synaptic (NPTX1, NPTXR) vs. post-synaptic (GAP-43) markers[@schweigert2024]
This layer explains why biomarkers reflecting a single cell type may not capture the full disease process. Neurodegeneration results from complex cross-talk between glia, neurons, and vascular cells.
Layer 4: Clinical Heterogeneity
The same molecular pathology produces different clinical syndromes:
- Amnestic vs. non-amnestic presentation in Alzheimer's
- Parkinsonism vs. dementia-first in synucleinopathies
- Rapid vs. slow progression even with similar biomarker profiles
- Atypical variants: posterior cortical atrophy, logopenic aphasia, behavioral variant FTD
Clinical heterogeneity reflects not only the distribution of pathology but also individual cognitive reserve, education, comorbidities, and medication effects.
Implications for Fluid Biomarker Development
Blood-Based Biomarker Strategies
The framework has direct implications for the ongoing development of blood-based biomarkers:
Multiplex panels over single markers — No single biomarker captures the full heterogeneity. Panels of Aβ42/40, p-tau181, p-tau217, NfL, GFAP, and α-synuclein together provide a more complete profile[@blennow2024]
Subtype-specific cutoffs — Different reference ranges may be needed for carriers vs. non-carriers of risk variants (e.g., lower p-tau217 thresholds for APOE ε4 homozygotes)
Temporal dynamics — Biomarkers change at different rates depending on disease stage. Aβ changes precede tau changes by years; NfL elevations track neurodegeneration more closely
Cross-disease discrimination — The framework enables more accurate differential diagnosis by recognizing that overlapping molecular pathologies create distinct biomarker signaturesClinical Trial Design
The heterogeneity framework reshapes clinical trial design:
- Enrichment strategies: Selecting patients based on multiple biomarkers (e.g., amyloid-positive + specific tau strain + absence of α-synuclein co-pathology) rather than single criterion
- Outcome measure selection: Cognitive endpoints may not capture drug effects if the treated pathology is not the primary driver of clinical decline in that patient subset
- Stratified medicine: Ultimately, ADRD drug development may need to follow oncology's model of molecularly defined subtypes with targeted therapies
Research Priorities
Based on this framework, Zetterberg outlined key research priorities for the field:
Longitudinal multimodal biomarker profiling — Following the same individuals with repeated fluid, imaging, and clinical assessments over decades to understand biomarker trajectories in context
Pathology-specific biomarker validation — Correlating fluid biomarkers with post-mortem neuropathology to establish ground truth for each molecular species
Subtype discovery — Using unsupervised clustering (machine learning) on multimodal biomarker data to identify disease subtypes that transcend clinical diagnostic categories
Therapeutic target linkage — Matching biomarker subtypes to specific therapeutic mechanisms (anti-amyloid, anti-tau, anti-inflammatory, anti-synuclein) to enable precision medicineCross-Links
- [Blood-based biomarkers for neurodegeneration](/biomarkers/blood-based-biomarkers-neurodegeneration) — General overview of fluid biomarkers
- [Plasma p-tau217 biomarker](/biomarkers/plasma-ptau217-alzheimers) — Key tau biomarker in Zetterberg's framework
- [Alpha-synuclein PET imaging advances](/biomarkers/alpha-synuclein-pet-imaging-advances) — Synuclein-specific imaging
- [Amyloid cascade hypothesis](/mechanisms/amyloid-cascade-hypothesis) — The linear model this framework challenges
- [Neuroinflammation biomarkers](/biomarkers/neuroinflammation-biomarkers) — Glial activation markers
- [Henrik Zetterberg researcher page](/researchers/henrik-zetterberg) — Lead researcher in ADRD biomarkers
References