Contact
A river delta from above. Photograph by Cosmin Andrei Buzamat.
01v1.0 · October 2026 · Arcas DynamicsNeurovascular biologyGlymphatic clearance
HypothesisNeuroscience

Perivascular fibrosis as a mechanical bottleneck.

Scar-forming cells deposit collagen inside the channels that carry fluid through the brain and clear its waste. Because flow through a narrow channel scales steeply with its open volume, a small deposit should cost a disproportionate share of clearance, and the signals it fails to clear should drive further deposition. The result is a feed-forward loop with a threshold: below it, scarring resolves; above it, clearance decline sustains itself.

None of the individual links is new. Closing them into a loop with a critical point is.

The argument.

Four links, each supported separately in the literature. The claim that they close into a loop is new.

  1. 01DepositionInjury, inflammation or chronic vascular stress activates perivascular fibroblasts, which deposit collagen I, fibronectin and basement membrane into the perivascular space.
  2. 02ObstructionThe matrix shrinks the open volume of the channel and stiffens the vessel wall, lowering both the channel’s conductance and the arterial motion that drives flow.
  3. 03RetentionSlower exchange concentrates what would otherwise be cleared: TGF-beta, damage-associated molecules and protein aggregates.
  4. 04Re-activationRetained TGF-beta and the stiffer matrix activate the same fibroblasts, and cells pulling on stiff matrix release active TGF-beta already stored in it.

Step 04 feeds step 01. The loop closes, and it has a threshold.

Live model.

Two parts of the model, running live in your browser from the paper’s equations. Drag the controls to change the inputs.

Fig. 1a Matrix narrows the channelLive model
Filled20%
Flow lost, cube law48.8%
Flow lost, fitted exponent 2.7545.9%
Amplification2.44×

With 20% of the channel filled, the cube law gives 48.8% of the flow lost.

Fig. 1b Two stable states, one thresholdSimulated
OutcomeSelf-sustaining
Settles at F0.601
Clearance left6.3%

Starting from F₀ = 0.32, the scar sustains itself, settling at F = 0.601 with 6.3% of clearance left.

Model estimates under illustrative parameters: a = 0.05, KS = 1, b = 0.6, n = 4, θ = 0.3, d = 1, P = 0.5, kother = 0.1, kpvs = 1.

The equations.

01 Thin-gap lubrication
Q∝h3 ΔPμL  ⇒  QQ0=(1−f)3Q \propto \frac{h^{3}\,\Delta P}{\mu L} \;\Rightarrow\; \frac{Q}{Q_{0}} = (1 - f)^{\htmlClass{hi}{3}}

Filling 10, 20 and 30% of the gap costs 27, 49 and 66% of the flow.

02 Fit to realistic cross-sections
QQ0≈(1−f)2.75\frac{Q}{Q_{0}} \approx (1 - f)^{\htmlClass{hi}{2.75}}

From resistance falling about 1,700-fold as area rises 15-fold (Tithof et al., 2019). The exponent is a fit to their figures, not a value they state.

03 Porous medium, Kozeny–Carman
k=d2 φ3180 (1−φ)2k = \frac{d^{2}\,\varphi^{3}}{180\,(1 - \varphi)^{2}}

Porosity falling from 0.40 to 0.30, a 25% loss of open volume, cuts permeability by about 69%.

04 The loop
dFdt=a SKS+S+b FnFn+θn−d F\frac{\mathrm{d}F}{\mathrm{d}t} = a\,\frac{S}{K_{S} + S} + b\,\frac{F^{n}}{F^{n} + \theta^{n}} - d\,F

Solute-driven activation, stiffness-driven self-activation and matrix turnover. Clearance falls as C = (1 − F)³, and the solute sits at the steady state that clearance allows.

What it builds on.

Published findings the model rests on, as the paper cites them.

About 40%
fall in amyloid-beta clearance in aged mice, alongside a 27% drop in arteriolar pulsatilityKress et al., 2014, Annals of Neurology
About 60%
fall in glymphatic exchange after traumatic brain injury, lasting at least a monthIliff et al., 2014, Journal of Neuroscience
24%
higher adjusted risk of dementia after traumatic brain injury, in a cohort of 2,794,852 peopleFann et al., 2018, The Lancet Psychiatry
1,700×
fall in hydraulic resistance as a perivascular channel’s cross-section grows 15-foldTithof et al., 2019, Fluids and Barriers of the CNS

Predictions.

Each one names the result that would refute it.

  1. P1Collagen predicts clearanceRefuted if: perivascular collagen adds no independent variance to regional clearance, or AQP4 polarisation dominates it.
  2. P2Losses are amplifiedRefuted if: fractional clearance loss is at or below the fractional loss of open volume.
  3. P3A threshold, with hysteresisRefuted if: graded injury gives a graded, fully reversible relation, with no range where outcomes split into two groups.
  4. P4Early treatment works, late does notRefuted if: an intervention long after the threshold restores clearance as well as an early one.
  5. P5The sleeve stiffens as tissue softensRefuted if: the perivascular sleeve is no stiffer than the surrounding parenchyma in injured or aged tissue.
  6. P6Dilation feeds narrowingRefuted if: dilated and collagen-rich segments of the same vessel are spatially unrelated.

How to test it.

First testStiffness around the vessel wall

Atomic force or Brillouin microscopy on acute slices, registered to vessel position, measuring stiffness across the first 20 micrometres from the wall alongside YAP and TAZ in perivascular fibroblasts.

P5
Experiment 2Collagen against AQP4

Tracer clearance in young, aged and injured mice, with perivascular collagen, vessel density and AQP4 polarisation measured in the same animals, using optical clearing to avoid fixation artefacts. About 10 animals per group (model estimate).

P1P6
Experiment 3Graded injury, timed treatment

Cortical impact at graded depths, with clearance imaged at one week, one month and three months, looking for two separate outcome groups; then partial anti-fibrotic treatment given early or late.

P3P4

Where it could be wrong.

The stiffness direction may be wrong.

Bulk brain tissue softens with injury, Alzheimer’s disease and age. The model needs a locally stiffer sleeve around the vessel, which has not been measured.

Perivascular spaces enlarge in disease.

Enlarged spaces look like the opposite of obstruction. The model places dilation upstream of a restriction on the same vessel, which Prediction 6 tests.

The strongest evidence is meningeal.

Fibrosis is known to obstruct fluid flow in the meninges. Extending that into the brain’s own perivascular channels is the new and unproven part.

Both feedback arms are fragile.

If the sleeve does not stiffen and retention is weak for a short-range signal, nothing holds the loop closed, and the model reduces to one-way obstruction with no threshold.

About this paper.

Authors
Arcas Dynamics, Toronto, Canada
Version
1.0, October 2026
Predictions
6, none tested yet
DOI
Not yet assigned
Preprint
Not yet posted
Licence
CC BY 4.0. Share and adapt with credit.
Download PDF

Changelog

  1. v1.0 · October 2026First published.

How to cite

@misc{arcas2026perivascular,
  author       = {Arcas Dynamics},
  title        = {Perivascular fibrosis as a mechanical bottleneck},
  year         = {2026},
  month        = oct,
  note         = {Hypothesis paper 01, version 1.0, CC BY 4.0},
  howpublished = {\url{https://arcasdynamics.com/research/perivascular-fibrosis}}
}