Three Vascular Patterns Emerge After Chronic Stroke Despite Similar Functional Performance
A small but carefully phenotyped stroke cohort showed that vascular injury after stroke does not follow a single pattern. In 45 adults with chronic stroke, unsupervised k-means clustering separated participants into three distinct vascular profiles using age plus four noninvasive measures: flow-mediated dilation, carotid intima-media thickness, pulse wave velocity, and augmentation index. The profiles diverged across every vascular measure assessed, with all between-cluster comparisons reaching statistical significance at p < 0.01.
The most important signal was heterogeneity. One cluster showed elevated arterial wave reflection despite relatively preserved endothelial function and arterial structure, while another showed impairment across all vascular variables. Age increased across clusters, but age alone did not explain the spread in vascular biology. In practical terms, two people of similar stroke history may arrive at very different vascular phenotypes, even before traditional bedside measures would reveal much separation.
Endothelial Function, Arterial Thickness, and Stiffness Do Not Fail in Lockstep
The physiologic mix in this cohort matters because the measures capture different parts of the vascular system. Flow-mediated dilation reflects endothelial nitric oxide-mediated vasodilatory capacity, making it a useful readout of vascular responsiveness. Carotid intima-media thickness tracks structural remodeling in the arterial wall, while pulse wave velocity and augmentation index capture large-artery stiffness and wave reflection. When these variables move together, vascular aging often looks orderly. After stroke, that order appears to fracture.
One cluster in this analysis had relatively preserved endothelial and structural measures but a more adverse wave reflection profile. Another cluster carried the heaviest vascular burden across the board. The biology is plausible: endothelial dysfunction, arterial wall remodeling, and central stiffness are related but not interchangeable processes, and post-stroke patients often carry a mosaic of vascular injury shaped by pre-existing risk, the stroke event itself, and long-term deconditioning. A single summary label such as “vascular dysfunction” can obscure those differences.
The study design strengthened that point by including age in the clustering model. Even after accounting for vascular aging, the profiles did not collapse into a simple age gradient. The pattern argues for multidimensional phenotyping rather than assuming that older age and more time since stroke will necessarily map to a uniformly worse vascular state.
Functional Measures Stayed Flat Across the Vascular Clusters
The vascular groups did not separate by everyday performance. Neither predicted VO2peak nor gait speed differed by cluster, with p values of 0.31 and 0.65, respectively. That finding runs against the intuitive expectation that worse vascular health should automatically translate into slower walking or lower cardiorespiratory fitness in a chronic stroke population.
There are several biologically reasonable explanations. Functional performance after stroke depends on more than vascular health: motor impairment, balance, spasticity, cardiorespiratory conditioning, fatigue, compensatory gait strategies, and rehabilitation exposure all shape what a person can do on a given day. A vascular phenotype may be important for long-term cardiovascular risk without being tightly mirrored by a comfortable walking speed or a submaximal fitness estimate in a relatively small sample.
The null functional association also limits how far the current data can be pushed. These results do not show that vascular health is unimportant after stroke. They show something narrower: in this baseline cohort, noninvasive vascular phenotype did not neatly track the two functional outcomes that were tested. The signal is in the biology, not yet in the performance endpoints.
Baseline Phenotyping From the FAST Trial Provides a Template for Risk Stratification
The analysis drew from baseline data in the Fitness after Stroke trial, conducted between March 7, 2023 and August 12, 2025, and registered as NCT05936008. Participants were 20 to 85 years old and 6 months to 15 years post-stroke. All were able to walk overground without continuous physical assistance and were stable enough for a submaximal exercise protocol and vascular testing. Many were already taking blood pressure and statin medications, which were withheld until after vascular assessments to reduce medication-related distortion of the measurements.
That testing approach matters clinically. FMD was measured in the morning under standardized fasting and activity restrictions; carotid intima-media thickness was assessed with bilateral ultrasound; and arterial stiffness and wave reflection were captured with pulse wave velocity and augmentation index. The investigators used semi-automated edge-detection software and average values from bilateral measures where side-to-side differences were not significant. The result is a physiologic profile built from multiple vascular compartments rather than a single proxy marker.
This kind of profiling is still exploratory, but it is already useful conceptually. Stroke survivors are often treated as if vascular risk is homogeneous once the event has occurred. These data argue the opposite. Some patients may carry dominant endothelial impairment, some may show prominent stiffness or wave reflection, and others may have diffuse abnormalities across the vascular tree.
What the Three Clusters Suggest at the Bedside
For rehabilitation clinicians and vascular specialists, the immediate takeaway is not that clustering should enter routine practice tomorrow. It is that chronic stroke populations may contain biologically distinct subgroups that conventional clinical history does not fully expose. A patient with preserved walking speed can still have an adverse vascular profile. Another with a similar stroke timeline may show much less vascular disturbance.
This has implications for future risk stratification and for intervention design. Exercise training, blood pressure control, lipid management, and other vascular therapies may not exert identical effects across these phenotypes. A person whose dominant abnormality is wave reflection may not be the same person who needs the most aggressive attention to endothelial dysfunction or arterial remodeling. The clinical question becomes more precise: which vascular pathway is driving risk in which patient?
Small Sample, Exploratory Clusters, and an Unfinished Clinical Question
The cohort included only 45 participants, split into three clusters of 13, 11, and 21 individuals. That size is appropriate for exploratory phenotyping, but it is not large enough to define clinical cut points or to claim prognostic value. The study is also baseline-only, so it cannot answer whether these vascular profiles predict recurrent stroke, myocardial infarction, hospitalization, or mortality over time.
Still, the data establish a foundation. Chronic stroke is not a single vascular state, and age did not fully explain the heterogeneity. Future work will need to test whether these cluster-derived profiles are reproducible in larger cohorts, whether they remain stable over time, and whether they identify patients who benefit from targeted vascular intervention. Until then, the profiles are best viewed as a biologic map of post-stroke diversity rather than a finished bedside tool.