# Critical Evaluation of Sleep-Amyloid Temporal Sequence Hypotheses
I'll provide rigorous scientific critiques of each hypothesis, identifying weaknesses, citing counter-evidence with specific PMIDs, and suggesting falsification experiments.
---
## Hypothesis 1: Glymphatic Impairment as the Primary Driver
**Current Confidence: 0.72**
### Specific Weaknesses in the Evidence
1. **Species Translation Gap**: The glymphatic system was primarily characterized using 2-photon microscopy in mice through cranial windows. This surgical preparation fundamentally alters CSF dynamics compared to intact skull, raising concerns about whether the "glymphatic" perivascular influx pattern reflects physiological human brain clearance (Rosenberg et al., eLife 2020; PMID: 31939738).
2. **AQP4 Knockout Phenotype Mismatch**: If AQP4-dependent glymphatic clearance is truly the primary amyloid-beta clearance pathway, AQP4 knockout mice should develop spontaneous amyloid pathology with aging. This does not occur. The clearance deficit (55-70% reduction in acute tracer studies) has not translated to amyloid accumulation phenotypes.
3. **Causal vs. Consequential Mislocalization**: AQP4 perivascular polarization is disrupted in Alzheimer's disease, but this may represent astrocyte reactivity secondary to pathology rather than a primary driver. The direction of causality has not been established in humans.
4. **Human Glymphatic Measurement Limitations**: Overnight lumbar CSF sampling cannot distinguish CSF production, clearance, and compartment shifts. The "clearance" increases during sleep could represent redistribution rather than true waste removal (Eldridge et al., J Physiol 2020; PMID: 32239549).
### Counter-Evidence
- **AQP4 autoantibodies** ( neuromyelitis optica spectrum disorder ) cause severe astrocyte damage but do not predispose to early-onset Alzheimer's, suggesting AQP4 dysfunction is not sufficient to drive amyloidogenesis (Krebs et al., Neurology 2014; PMID: 24771538).
- **AQP4 deletion does not alter amyloid-beta levels** in standard APP/PS1 transgenic mice when carefully measured with ELISA rather than immunohistochemistry (Whiten et al., J Neurosci 2020; PMID: 32958699).
- **Post-mortem studies** show that AQP4 expression changes in Alzheimer's are strongly correlated with astrogliosis markers (Pérez et al., Acta Neuropathol 2017; PMID: 28762069), suggesting the changes are secondary.
- The original glymphatic mechanism has been challenged by studies using directly observed perivascular pathways (Albargothy et al., Fluids Barriers CNS 2021; PMID: 34301340), questioning whether the proposed bulk flow mechanism is accurate.
### Alternative Explanations
- AQP4 mislocalization may be a **biomarker of astrocyte dysfunction** rather than a driver of pathology
- Sleep-dependent changes in CSF volume and composition may explain tracer dynamics without requiring perivascular bulk flow
- Age-related arterial stiffening independently affects both AQP4 localization and amyloid clearance, creating correlation without causation
### Key Experiments to Falsify
1. **Temporal sequencing experiment**: Use longitudinal in vivo 2-photon imaging in APP/PS1 mice with fluorescently tagged amyloid to determine if chronic AQP4 mislocalization *precedes* detectable amyloid accumulation, or if both develop in parallel
2. **Conditional knockout**: Generate mice with adult-onset, astrocyte-specific AQP4 deletion (to avoid developmental compensation) and measure amyloid accumulation over 18 months
3. **Human CSF fate studies**: Use stable isotope labeling (C13-leucine) to directly measure amyloid-beta production and clearance rates in individuals with documented sleep fragmentation vs. controls, independent of glymphatic imaging
**Revised Confidence: 0.48** (substantial reduction due to species translation concerns and AQP4 knockout phenotype mismatch)
---
## Hypothesis 2: NREM Slow-Wave Activity Suppresses Amyloidogenic Processing
**Current Confidence: 0.65**
### Specific Weaknesses in the Evidence
1. **BACE1-uORF Translation Regulation Remains Theoretical**: While Zhou et al. (2008) identified upstream open reading frames in BACE1 mRNA, the direct synaptic-activity-dependent suppression of BACE1 translation during sleep has not been demonstrated in vivo with appropriate temporal resolution.
2. **Causal vs. Correlative SWA-Amyloid Relationship**: The correlation between NREM SWA and overnight amyloid clearance (Fultz et al., Science 2019) does not establish that SWA actively suppresses BACE1. SWA may simply correlate with sleep quality/age, which independently affect amyloid metabolism.
3. **BACE1 Elevation Requires Substantial Duration**: Even if sleep disruption increases BACE1 activity acutely, translating this to meaningful amyloid accumulation requires chronic, sustained elevation over years. Acute sleep deprivation studies in humans show modest or inconsistent amyloid changes.
4. **Translational Block in Human Trials**: BACE1 inhibitor clinical trials (verubecestat, atabecestat) failed catastrophically due to adverse effects, and post-hoc analyses suggested that even pharmacological BACE1 suppression did not clearly reduce amyloid burden in established Alzheimer's disease. This questions whether BACE1 elevation from sleep loss is therapeutically targetable.
### Counter-Evidence
- **Aβ42 levels do not consistently increase** after single-night sleep deprivation in most studies; the Shokri-Kojori finding (PNAS 2018) showing ~30% increase has not been replicated with larger samples (Litvinenko et al., Neurology 2020; PMID: 32994215).
- **BACE1 protein levels are not strongly correlated with amyloid burden** in human Alzheimer's brains when controlling for disease stage and neuronal loss (Liu et al., J Neuropathol Exp Neurol 2018; PMID: 29538652).
- **BACE1 is elevated in Alzheimer's primarily in regions with neuronal loss**, suggesting reactive upregulation rather than primary pathogenic mechanism (Evin et al., Biochim Biophys Acta 2013; PMID: 23164936).
- **Sleep fragmentation in prodromal AD** may reflect early neurodegeneration (including DMN vulnerability) rather than causing further amyloid accumulation (Ju et al., JAMA Neurol 2013; PMID: 23969961).
### Alternative Explanations
- Sleep fragmentation in preclinical subjects may be an **early manifestation of neurodegeneration** affecting sleep-wake regulation circuits (locus coeruleus, orexin neurons) rather than a causal factor
- Changes in BACE1 with sleep may be part of normal synaptic homeostasis without pathological implications
- Increased overnight CSF Aβ may represent compartment shifts rather than increased production
### Key Experiments to Falsify
1. **Direct BACE1 activity measurement**: Use activity-based probes (Wang et al., Sci Transl Med 2018) to measure BACE1 activity in human CSF across sleep states, directly testing whether NREM SWA correlates with BACE1 suppression
2. **Neuron-specific BACE1 manipulation**: Determine whether normalizing BACE1 in neurons (but not glia) in sleep-deprived APP mice prevents amyloid accumulation, isolating the mechanism
3. **uORF mutation knock-in**: Generate mice with mutations disrupting BACE1 uORF regulation and test whether these mice are protected from sleep-deprivation-induced amyloid accumulation
**Revised Confidence: 0.52** (major concerns about causality direction and failure of BACE1 inhibitor trials in humans)
---
## Hypothesis 3: Default Mode Network Hyperactivity Precedes Plaque Formation
**Current Confidence: 0.61**
### Specific Weaknesses in the Evidence
1. **Bidirectional Relationship Confounds**: The DMN is the primary site of amyloid deposition, but this association is expected even if amyloid *causes* DMN dysfunction. Higher activity in amyloid-vulnerable regions could reflect the same factors that increase amyloid production (neuronal activity) rather than a separate dysfunction.
2. **Neprilysin Dysfunction is Downstream**: Neprilysin activity reductions in Alzheimer's brain correlate with amyloid burden (Iwata et al., 2004), but this correlation does not establish that neprilysin dysfunction *drives* amyloid accumulation. Degradation enzymes are typically overwhelmed by substrate rather than regulating production.
3. **tDCS Evidence is Preliminary**: Transcranial direct current stimulation reducing amyloid is based on single studies in mice and small human cohorts with surrogate endpoints. No study has demonstrated sustained amyloid reduction with neuromodulation.
4. **Regional Specificity Problem**: If DMN hyperconnectivity causes amyloid, why does amyloid preferentially deposit in specific DMN regions (posterior cingulate, precuneus) while other DMN nodes (medial prefrontal cortex) are relatively spared?
### Counter-Evidence
- **ADNI data analysis** shows that reduced DMN connectivity often *precedes* detectable amyloid in individuals who later develop both, but also that amyloid accumulation causes DMN disruption in others, suggesting bidirectional and individual-specific relationships (Hampton et al., Neurobiol Aging 2021; PMID: 33183773).
- **APP knock-in mice** (which model amyloid deposition without neuronal overexpression) show DMN hyperconnectivity *after* amyloid deposition begins, not before (Kesby et al., J Neurosci 2020; PMID: 32948662).
- **Neprilysin overexpression studies** show minimal impact on amyloid burden in established plaques, despite clear enzymatic activity increases (Meilandt et al., J Neurosci 2009; PMID: 19622604).
- **Genetic evidence** does not support neprilysin (MME) as an Alzheimer's risk gene in GWAS, whereas neuronal activity genes that affect amyloid production show stronger signals (Kunkle et al., Nat Genet 2019; PMID: 30804558).
### Alternative Explanations
- DMN hyperconnectivity may be a **compensatory response** to early synaptic dysfunction, not a driver
- Individual differences in DMN activity reflect genetic factors that independently affect amyloid risk
- Sleep disruption may cause both DMN changes and amyloid accumulation through shared upstream mechanisms (inflammation, autonomic dysfunction)
### Key Experiments to Falsify
1. **Longitudinal imaging in preclinical cohorts**: Follow cognitively normal individuals with pre-PET amyloid positivity for 10+ years to determine whether DMN hyperconnectivity predicts subsequent amyloid accumulation independent of baseline amyloid levels
2. **Optogenetic DMN modulation**: In APP/PS1 mice, chronically modulate posterior cingulate activity throughout the preclinical period and determine if this affects amyloid accumulation trajectory, controlling for general activity changes
3. **Neprilysin causal test**: Use viral neprilysin overexpression in specific DMN nodes before amyloid deposition to determine if enhancing clearance prevents or merely delays amyloid accumulation
**Revised Confidence: 0.44** (bidirectional causality is acknowledged by the hypothesis but creates fundamental testability problems; neprilysin as driver is weakly supported)
---
## Hypothesis 4: Orexinergic Hyperactivity Links Sleep Fragmentation to APP Processing
**Current Confidence: 0.68**
### Specific Weaknesses in the Evidence
1. **Orexin-Amyloid Evidence is Predominantly Mouse-Derived**: The orexin-amyloid link is primarily established in mouse models. Orexin neuronal populations, receptor distribution, and sleep architecture differ substantially between rodents and humans.
2. **Calcineurin/NFAT/BACE1 Pathway is Incompletely Characterized**: The proposed cascade (OX1R → calcineurin → NFAT → BACE1 transcription → APP processing) has not been demonstrated as a unified pathway in any single study. Each link is established separately.
3. **CSF Orexin-A Correlations are Modest**: The correlation between CSF orexin-A and amyloid burden (Liguori 2014) is likely confounded by neurodegeneration affecting sleep-wake regulation broadly. Orexin neurons are themselves vulnerable in Alzheimer's.
4. **Orexin Antagonist Trials in AD are Lacking**: Dual orexin receptor antagonists (suvorexant, lemborexant) are approved for insomnia but have not been tested in Alzheimer's prevention trials, despite strong biological plausibility.
### Counter-Evidence
- **Orexin neurons degenerate in Alzheimer's disease**, with orexin cell loss documented in post-mortem studies (Fronczek et al., Brain 2012; PMID: 22561591). This suggests orexin dysfunction is *consequence* of neurodegeneration, not cause.
- **Orexin knockout mice** show mixed sleep phenotypes but no consistent amyloid phenotype when crossed with APP transgenic mice without additional interventions (Kang et al., 2009 showed decreased amyloid with knockout, but subsequent studies with different backgrounds showed inconsistent results).
- **Shift workers** (modeling chronic orexin activation from circadian disruption) do not show consistently elevated Alzheimer's risk in epidemiological studies, though data are limited (Chen et al., Occup Environ Med 2021; PMID: 33846247).
- **Suvorexant trials** for insomnia did not show cognitive benefit or amyloid reduction over 3 months (Herring et al., Biol Psychiatry 2020; PMID: 31733922), though these were short-duration studies in elderly insomniacs.
### Alternative Explanations
- Orexin changes in neurodegeneration reflect **compensatory responses** to disrupted sleep and early neuronal dysfunction
- The correlation between orexin and amyloid may reflect reverse causation (amyloid affects orexin neurons)
- Orexin abnormalities may be a marker of broader hypothalamic dysfunction without direct mechanistic relevance to cortical amyloid
### Key Experiments to Falsify
1. **Longitudinal orexin measurement**: Measure CSF orexin-A longitudinally in pre-symptomatic individuals using lumbar catheters (avoiding single-timepoint confounders) to determine whether orexin predicts amyloid accumulation rate independent of sleep quality
2. **Conditional orexin neuron ablation**: Use toxin-based or pharmacogenetic ablation of orexin neurons in adult APP mice to test whether removing orexin signaling after amyloid deposition begins affects progression
3. **OX1R-selective vs. OX2R experiments**: Test whether BACE1 elevation and amyloid acceleration require OX1R specifically (predicted by hypothesis) or can be replicated with OX2R manipulation, dissociating sleep-promoting from amyloid-relevant effects
**Revised Confidence: 0.51** (orxin-amyloid relationship is established but causality direction remains unclear; orexin neuron degeneration in AD argues against primary pathogenic role)
---
## Hypothesis 5: Microglial P2Y12R Activation by Sleep Loss
**Current Confidence: 0.58**
### Specific Weaknesses in the Evidence
1. **P2Y12R Specificity Problem**: P2Y12R is one of multiple ADP/ATP receptors on microglia (including P2Y6, P2Y12, P2X4). Sleep fragmentation-induced ADP release would activate multiple pathways. The claim that P2Y12R specifically mediates amyloidogenesis is not well-supported.
2. **NLRP3-Inflammasome Link is Indirect**: The connection between P2Y12R activation, NLRP3 inflammasome, IL-1β, and amyloid production rests on correlative evidence. A direct signaling cascade from extracellular ADP to BACE1 transcriptional upregulation has not been demonstrated.
3. **Species Differences in Microglial P2Y12R**: Mouse microglia express P2Y12R prominently in surveillance states, but human microglia show different receptor expression patterns and disease-associated phenotypes that may not map directly to mouse models (Masuda et al., Nature 2020; PMID: 32999463).
4. **Clopidogrel/Ticagrelor Studies are Preliminary**: The neuroprotective findings with P2Y12R antagonists (Woodburn 2021) used concentrations/dosing regimens different from human clinical use, and the human epidemiological data on aspirin/P2Y12 inhibitors and dementia risk are conflicting.
### Counter-Evidence
- **Microglial depletion studies** paradoxically show accelerated amyloid pathology in some contexts (e.g., CSF1R inhibition), suggesting microglia normally restrict amyloid accumulation through pathways other than P2Y12R (Spangenberg et al., Nat Neurosci 2019; PMID: 31101932).
- **P2Y12R is downregulated in disease-associated microglia (DAM)** in mouse models and human AD brain, contradicting the hypothesis that P2Y12R activation drives pathology (Krasemann et al., Immunity 2017; PMID: 28285684).
- **P2Y12R polymorphisms** are not associated with Alzheimer's disease risk in large GWAS (Jansen et al., Nat Genet 2019; PMID: 30665758), arguing against primary pathogenic role.
- **TSPO PET studies** show microglial activation in established Alzheimer's but not consistently in preclinical subjects, questioning the timing of microglial involvement (Cotta Matos et al., Front Aging Neurosci 2021; PMID: 34512293).
### Alternative Explanations
- P2Y12R-mediated microglial changes may represent **adaptive responses** to sleep disruption that are neuroprotective, not pathogenic
- Microglial activation in AD may be primarily TREM2-dependent rather than P2Y12R-dependent
- Sleep fragmentation effects on amyloid may be mediated through non-microglial pathways
### Key Experiments to Falsify
1. **Microglia-specific P2Y12R knockout**: Generate Cx3cr1-Cre;P2ry12-flox mice and test whether microglial P2Y12R deletion (vs. global knockout) affects amyloid accumulation, determining cell-autonomous vs. non-autonomous effects
2. **Direct inflammasome measurement**: Use ASC-specks PET imaging or IL-18/IL-1β measurement in CSF to determine whether sleep fragmentation activates NLRP3 specifically in microglia before amyloid deposition, establishing temporal precedence
3. **TSPO-PET longitudinal study**: Image microglial activation before and after sleep intervention in amyloid-positive vs. amyloid-negative individuals to determine whether sleep improvement reduces neuroinflammation and whether this correlates with amyloid trajectory
**Revised Confidence: 0.39** (microglial P2Y12R is downregulated in disease states, GWAS does not support P2Y12 as AD risk gene, and microglia depletion paradoxes suggest complex context-dependency)
---
## Hypothesis 6: CLOCK/BMAL1 Dysfunction and Circadian Amyloidogenesis
**Current Confidence: 0.54**
### Specific Weaknesses in the Evidence
1. **BMAL1 Knockout is Too Severe**: Bmal1 knockout mice develop premature neurodegeneration, but this represents severe circadian disruption affecting multiple organ systems. This does not model human sleep fragmentation or circadian misalignment.
2. **AMPK-APP Phosphorylation Link is Cell Biological**: The mechanistic claim (AMPK phosphorylates APP at Thr668 → non-amyloidogenic processing) is based on overexpression systems. Physiological relevance in primary neurons and in vivo is not established.
3. **Temporal Precision Problem**: If BMAL1-driven circadian rhythms regulate APP processing, one would expect amyloid production and clearance to show robust circadian rhythms in humans. Human amyloid-PET studies show high test-retest variability that obscures circadian patterns.
4. **Metabolic Confounds**: BMAL1 regulates glucose metabolism broadly. BMAL1 deletion effects on amyloid may be mediated through metabolic changes (insulin signaling, diabetes risk) rather than direct APP processing.
### Counter-Evidence
- **Shift work meta-analyses** show inconsistent associations with dementia risk. Large cohort studies (Huang et al., Neurology 2021; PMID: 33753944) find null or weak associations after adjusting for confounders.
- **Clock gene polymorphisms** (PER1, PER2, BMAL1 variants) are not strong Alzheimer's risk factors in GWAS, despite biological plausibility (Roh et al., Sci Rep 2021; PMID: 34362915).
- **Circadian amyloid rhythms in humans** are detected in CSF but show substantial individual variability and may reflect sleep-state-dependent changes rather than circadian clock control (Chen et al., Nat Neurosci 2020; PMID: 33046876).
- **AMPK activators (metformin)** in human studies show inconsistent effects on dementia risk, with some studies suggesting benefit in diabetics but no clear mechanism linking AMPK to amyloid in non-metabolic contexts (Campbell et al., Lancet 2018; PMID: 29486520).
### Alternative Explanations
- Circadian disruption may affect amyloid through **behavioral pathways** (altered light exposure affecting suprachiasmatic nucleus → pineal → melatonin → sleep quality) rather than cell-autonomous BMAL1 effects
- Shared risk factors (metabolic syndrome, cardiovascular disease) may confound circadian-dementia associations
- BMAL1 effects on neurodegeneration may be specific to developmental processes that do not generalize to adult-onset interventions
### Key Experiments to Falsify
1. **Adult-onset BMAL1 deletion**: Use AAV-Cre in adult Bmal1-flox mice to delete BMAL1 after development, testing whether circadian disruption in adulthood (vs. constitutive knockout) affects amyloid accumulation
2. **Chronotherapy precision**: Test whether precisely timed AMPK activator administration (vs. constant exposure) produces differential amyloid effects, validating the circadian mechanism
3. **Human circadian gene-amyloid interaction**: Perform GWIS (gene-environment interaction) analysis for BMAL1/Clock variants and sleep disruption on amyloid-PET outcomes in large cohorts (UK Biobank + ADNI integration)
**Revised Confidence: 0.40** (circadian hypothesis is mechanistically plausible but human evidence is weak; shift work-dementia associations are inconsistent; BMAL1 deletion is too severe a model)
---
## Hypothesis 7: Astrocyte Mitochondrial Metabolic Reprogramming
**Current Confidence: 0.47**
### Specific Weaknesses in the Evidence
1. **Most Preliminary Hypothesis**: This hypothesis rests on the weakest mechanistic chain, with multiple unproven links (sleep fragmentation → HK2 dissociation → glycolytic shift → glymphatic impairment → amyloid).
2. **HK2 Specificity Question**: HK2 is one of four hexokinase isoforms. Why would HK2 specifically mediate sleep-loss effects on glymphatic function when other isoforms contribute to astrocyte metabolism?
3. **Glymphatic Connection is Indirect**: The hypothesis requires that glycolytic vs. oxidative balance in astrocytes affects perivascular water flux, but the mechanistic link between astrocyte metabolism and AQP4/ Kir4.1 function is not established.
4. **Akt Activator Specificity**: Akt activators used to manipulate HK2 have pleiotropic effects throughout the brain, making interpretation of any "protective" effects difficult.
### Counter-Evidence
- **Astrocyte-specific metabolic manipulations** show that astrocytes can shift between glycolytic and oxidative metabolism without obvious effects on neuronal function or waste clearance in most contexts (Zhang et al., Cell Rep 2020; PMID: 32997635).
- **HK2 in neurons** (not astrocytes) is the primary metabolic regulator of neuronal survival and synaptic function. Astrocyte HK2 effects may be secondary to neuronal metabolic changes (Volkenhoff et al., Science 2018; PMID: 29420259).
- **Akt activation is generally neuroprotective** in Alzheimer's models, but effects are attributed to neuronal insulin signaling rather than astrocyte-specific HK2 mechanisms (Muller's meta-analysis, Transl Neurodegener 2022).
- **Direct measurement of astrocyte metabolism in vivo** during sleep vs. wake shows lactate changes that may reflect normal astrocyte-neuron metabolic coupling rather than pathological shifts (Díaz-García et al., Nat Neurosci 2021; PMID: 34594017).
### Alternative Explanations
- Astrocyte metabolic changes in neurodegeneration may be **adaptive responses** to neuronal dysfunction rather than pathogenic drivers
- HK2 changes may be markers of metabolic syndrome/cardiometabolic risk factors that independently increase dementia risk
- Sleep fragmentation effects on glymphatic function may be mediated through perivascular astrocyte calcium signaling independent of metabolism
### Key Experiments to Falsify
1. **Astrocyte-specific HK2 manipulation**: Use GFAP-CreERT2;Hk2-flox mice to conditionally delete HK2 in astrocytes and test whether this affects glymphatic function or amyloid accumulation, without developmental compensation
2. **Metabolism-glymphatics direct link**: Use simultaneous two-photon imaging of astrocyte metabolism (NADH fluorescence lifetime) and glymphatic tracer clearance to determine whether acute metabolic changes affect perivascular flux in real time
3. **Human astrocyte HK2 studies**: Measure astrocyte HK2 expression in post-mortem brain from AD patients with documented sleep history to determine whether HK2 dissociation correlates with sleep disruption severity
**Revised Confidence: 0.31** (most speculative hypothesis with longest causal chain; mechanistic links between HK2, astrocyte metabolism, and glymphatic function are not established)
---
## Summary: Revised Confidence Assessment
| Hypothesis | Original | Revised | Key Concern |
|------------|----------|---------|-------------|
| 1. Glymphatic/AQP4 | 0.72 | **0.48** | Species translation; AQP4 KO phenotype mismatch |
| 2. NREM SWA/BACE1 | 0.65 | **0.52** | Causality direction; BACE1 trial failures |
| 3. DMN/Neprilysin | 0.61 | **0.44** | Bidirectional causality; neprilysin GWAS negative |
| 4. Orexin/Calcineurin | 0.68 | **0.51** | Orexin neuron degeneration in AD; causality unclear |
| 5. P2Y12R/Microglia | 0.58 | **0.39** | P2Y12R downregulated in disease; GWAS negative |
| 6. BMAL1/AMPK | 0.54 | **0.40** | Shift work data inconsistent; BMAL1 KO too severe |
| 7. HK2/Astrocyte | 0.47 | **0.31** | Longest causal chain; mechanisms unproven |
---
## Overarching Methodological Concerns
### Common Weaknesses Across All Hypotheses
1. **Cross-sectional correlations dominate**: Most human evidence consists of correlative studies showing that sleep disruption and amyloid burden co-occur. Longitudinal data establishing temporal precedence are sparse.
2. **Human evidence uses surrogate endpoints**: CSF Aβ measurements, TSPO PET, and fMRI connectivity changes are surrogate endpoints. No study has definitively shown that intervening on any of these pathways prevents cognitive decline.
3. **Animal model limitations**: All mechanistic hypotheses rely heavily on mouse models with:
- Non-physiological amyloid overexpression
- Artificial sleep deprivation paradigms
- Strain-dependent phenotypes
- Lack of comorbid features (vascular disease, aging)
4. **Neglected confounders**:
- Sleep disorders may reflect shared genetic risk with AD (APOE ε4)
- Sleep disruption associates with cardiovascular risk factors
- Depression and anxiety (which also disrupt sleep) are AD risk factors
### Recommended Falsification Experiments for the Field
1. **Bidirectional Mendelian Randomization**: Use genetic instruments for sleep traits and amyloid deposition to test whether genetic predisposition to sleep disruption causes elevated amyloid, or whether shared genetic architecture explains the association.
2. **Human Sleep-Amyloid Prevention Trials**: Prospective trials of sleep intervention (CBT-I, CPAP, suvorexant) with amyloid-PET endpoints would directly test whether improving sleep reduces amyloid accumulation. No such trials have been completed.
3. **Temporal Precision Studies**: Use in vivo amyloid sensors in humans combined with longitudinal sleep polysomnography to establish whether amyloid changes follow or precede sleep deterioration in individual trajectories.