Henrik Zetterberg's heterogeneity framework, presented at the AD/PD 2026 conference in Copenhagen, articulates a fundamental reorientation in how the field should interpret biomarker data across Alzheimer Disease and Related Dementias (ADRD). His central thesis is that disease heterogeneity — not a single pathophysiological cascade — is the primary challenge facing biomarker validation, diagnostic classification, and clinical trial design[1].
The framework moves beyond the traditional view of Alzheimer's disease as a linear amyloid → tau → neurodegeneration cascade toward a model in which multiple independent biological processes (amyloid, tau, TDP-43, vascular injury, neuroinflammation, alpha-synuclein, aging-related resilience mechanisms) interact in varying combinations across individuals to produce clinically similar phenotypes. This heterogeneity explains why blood biomarkers that perform well in one cohort may underperform in others, why anti-amyloid therapies have shown heterogeneous response rates, and why single-marker diagnostic thresholds fail to capture the full spectrum of AD pathology.
The A/T/N classification framework[2] was a major advance for standardizing biomarker reporting, but its binary positive/negative framework masks substantial within-group heterogeneity:
Zetterberg argues that the field must move from categorical biomarker reporting (positive/negative) to continuous, multi-dimensional biomarker profiles that capture an individual's specific constellation of pathological processes[3].
The three major axes of heterogeneity that Zetterberg's framework identifies are:
The phosphorylated tau biomarker family (p-tau181, p-tau217, p-tau231) has emerged as the most specific blood-based readout for AD-type pathology, but their performance varies across heterogeneity axes:
| Biomarker | AD-specificity | Early sensitivity | Cross-disease signal | Key references |
|---|---|---|---|---|
| p-tau217 | Very high | Moderate (appears after Aβ PET+) | Low | [@thijssen2022; @mattsson2019] |
| p-tau231 | High | High (earliest tau marker) | Low | [5] |
| p-tau181 | High | Moderate | Low | [3:1] |
P-tau217 shows the strongest correlation with amyloid burden and clinical progression, making it the most widely adopted clinical marker[5:1]. However, its performance is modulated by:
Neurofilament light chain (NfL) reflects the rate of neuronal injury regardless of cause, making it powerful for tracking progression but limited for differential diagnosis. Zetterberg's framework emphasizes NfL as an ensemble-level monitoring tool — useful for clinical trials to track global neurodegeneration, but requiring complementary disease-specific markers to determine the underlying cause of injury.
Glial fibrillary acidic protein (GFAP) provides information about astrocytic reactivity that neither p-tau nor NfL captures. Elevated GFAP may indicate an earlier "pre-injury" state in the amyloid → tau → neurodegeneration cascade, useful for identifying individuals in a window where neuroprotective intervention might be most effective.
The framework advocates for multi-analyte panels that simultaneously measure:
Such panels allow individual-level biomarker "signatures" that can reveal heterogeneity not apparent from single-marker analysis[6].
Traditional AD clinical trial enrichment relied on amyloid PET positivity as the primary inclusion criterion. Zetterberg's framework argues that biomarker-based enrichment must go further to account for heterogeneity:
The framework distinguishes between:
For trials targeting disease modification, biomarker endpoints should carry primary weight. For symptomatic or prevention trials, composite cognitive endpoints may be more appropriate, with biomarker data used for mechanistic verification.
The heterogeneous clinical responses to anti-amyloid monoclonal antibodies (lecanemab, donanemab, gantenerumab) illustrate why the heterogeneity framework matters:
This framework builds on and cross-links to the following pages:
Zetterberg H, et al. Heterogeneity, heterogeneity, heterogeneity: A framework for understanding ADRD biomarker variation. AD/PD 2026 Conference Proceedings. 2026. ↩︎
Jack CR Jr, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease. Alzheimer's & Dementia. 2018. ↩︎
Blennow K, et al. Blood biomarkers for Alzheimer's disease — from research to clinical practice. Lancet Neurol. 2024. ↩︎ ↩︎
Hajjar I, et al. APOE4 genotype modifies blood biomarker performance in diverse populations. Neurology. 2024. ↩︎ ↩︎
Hansson O, et al. Blood biomarkers for Alzheimer's disease — implementation and clinical utility. Nat Aging. 2024. ↩︎ ↩︎
Chen SD, et al. Blood biomarkers for Lewy body diseases — heterogeneity and overlap with Alzheimer's disease. Nat Rev Neurol. 2024. ↩︎
Cummings J, et al. Alzheimer's disease drug development pipeline: 2024. Alzheimer's & Dementia. 2024. ↩︎