Effects of blood flow restriction combined with visual feedback training on lower limb function and the serum levels of BDNF and VEGF in patients with stroke

Clinical Question (PICO)

In medically stable adults aged 40–70 with a first ischemic or hemorrhagic stroke within six months and Functional Ambulation Category (FAC) ≥2, does adding affected-thigh blood flow restriction training (BFRT) to visual feedback training (VFT), alongside the same conventional rehabilitation, improve balance and lower-limb function over four weeks compared with VFT and conventional rehabilitation alone?

The principal balance measurements came from the Pro-Kin platform. Mobility, activities of daily living, strength, and circulating brain-derived neurotrophic factor (BDNF) and vascular endothelial growth factor (VEGF) were additional outcomes. No prespecified treatment-effect subgroups or interaction tests are reported; ischemic and hemorrhagic stroke participants were pooled.

Bottom Line

Adding BFRT produced greater short-term gains in balance, motor function, walking ability, and activities of daily living than VFT alone. Pro-Kin limit of stability reached 73.67 versus 56.56 after treatment; the reported adjusted difference, oriented control minus experimental, was −15.65 (95% CI −22.54 to −8.75; P<0.001; Table 4).

This is a promising, small, single-center rehabilitation trial—not proof of durable independence, fewer falls, or cerebral neurovascular repair. The abstract attributes the biomarker increases to neurovascular plasticity, but the full-text conclusion explicitly states that serum BDNF and VEGF do not directly demonstrate central plasticity or neural repair.

Design

Trial type: Parallel-group randomized trial; participants and therapists unblinded, outcome evaluators blinded. No sham cuff.
N: 40 randomized; all 40 completed treatment and the four-week assessment.
- VFT plus conventional rehabilitation: 20.
- BFRT plus VFT plus conventional rehabilitation: 20.
Randomization: 1:1, using a random-number table after assigning participant identification numbers. Allocation-concealment procedures and stratification are not described.
Setting: Rehabilitation Hospital of Chongqing Medical University, China; one inpatient center.
Enrollment: July 2025–May 2026. Of 52 assessed, nine were ineligible and three declined.
Follow-up: Four weeks; no longer-term assessment reported.
Analysis: The statistical section describes independent-samples t tests or Mann–Whitney tests for between-group comparisons, and Wilcoxon tests for within-group ordinal comparisons. Results additionally report adjusted continuous differences and ordinal-regression adjusted odds ratios, but the adjustment model, covariates, and multiplicity procedure are not adequately specified. No explicitly labeled intention-to-treat analysis is given; all randomized participants completed assessment.
Primary outcome: Pro-Kin balance assessment: limit of stability and center-of-pressure trajectory area and length with eyes open and closed. Methods also refer broadly to balance and gait-related function; sample-size planning centers on Pro-Kin and a repeated-measures group-by-time interaction.
Sample-size planning: Target 40 after a minimum estimate of 34, assuming effect size f=0.25, two-sided α=0.05, and 80% power.

Population

Inclusion Criteria

First-ever stroke within six months, confirmed by CT or MRI; ischemic and hemorrhagic stroke both eligible.
Age 40–70 years.
Stable vital signs, clear consciousness, and sufficient understanding, communication, and cooperation for training.
FAC ≥2 and cardiopulmonary capacity adequate for rehabilitation.
Written informed consent.

Exclusion Criteria

Bone or joint disease, including fracture-surgery history or osteoarthritis.
Malignant hypertension, malignant tumor, or thrombosis.
History of psychiatric disorders.

Baseline Characteristics (control / experimental)

Age: 58.05±8.13 / 56.35±8.71 years.
Male: 12/20 (60%) / 14/20 (70%).
Time since stroke: 38.90±9.41 / 39.20±11.84 days.
Ischemic stroke: 11/20 (55%) / 12/20 (60%); hemorrhagic stroke: 9/20 (45%) / 8/20 (40%).
Fugl-Meyer lower-extremity assessment (FMA-LE): 12.10±3.09 / 12.05±3.03, on a 34-point scale.
Timed Up and Go Test (TUGT): 26.40±5.84 / 25.70±4.52 seconds.
Modified Barthel Index (MBI): 50.35±10.90 / 50.65±9.78, on a 100-point scale.
FAC: median 2 (IQR 2–3) in both groups.

Interventions

Both groups: Conventional rehabilitation five days weekly for four weeks: physical therapy and occupational therapy each 30 minutes per session; physical-factor therapy and traditional rehabilitation each 20 minutes per session.
Control: Additional Pro-Kin PK-252 visual feedback balance training for 20 minutes daily, five days weekly, four weeks. Standing weight shifts controlled game-based tasks selected by the therapist.
Experimental: The same VFT with a 10-cm-wide cuff on the proximal affected thigh, inflated to 160–200 mmHg throughout the 20-minute session. No additional cuff-associated resistance exercise was performed. Blood pressure, oxygen saturation, and cuff pressure were checked every five minutes, with continuous observation for pain, numbness, fatigue, skin changes, or intolerance.

Important implementation boundary: Individual arterial limb occlusion pressure was not measured. The fixed pressure range is a description of the research intervention, not a transferable prescription for unsupervised use or for all stroke survivors.

Outcomes

Primary Outcome — Pro-Kin balance at four weeks:

The following are the authors’ adjusted differences in Table 4, with the reported control-minus-experimental orientation. Higher limit of stability and smaller trajectory measurements indicate better performance; measurement units are not specified in the table.

Measurement	Control after treatment, mean±SD	Experimental after treatment, mean±SD	Reported adjusted difference (95% CI)	P
Limit of stability	56.56±12.41	73.67±10.62	−15.65 (−22.54 to −8.75)	<0.001
Trajectory area, eyes open	328.60±85.49	281.35±51.02	57.54 (25.12 to 89.96)	<0.001
Trajectory area, eyes closed	492.30±200.11	390.40±42.80	103.30 (27.12 to 179.49)	0.009
Trajectory length, eyes open	436.90±101.92	373.35±63.72	70.30 (28.03 to 112.56)	0.002
Trajectory length, eyes closed	581.35±175.75	485.20±68.88	93.04 (8.68 to 177.39)	0.032

No stroke-subtype, age, or sex-specific treatment estimate with a confidence interval is presented. These pooled results cannot establish equal benefit across those groups. The report does not supply a clinically important difference threshold for these platform measurements or observed fall rates.

Secondary Outcomes — Table 3 unless otherwise specified:

FMA-LE: 19.85±2.76 versus 22.20±1.96 after treatment; reported adjusted difference −2.371 points (95% CI −3.681 to −1.061; P<0.001; reported adjusted P<0.001), favoring BFRT.
TUGT: 19.15±3.80 versus 15.40±4.44 seconds after treatment. The table reports an adjusted difference of 7.817 seconds (95% CI 4.441–11.194; P<0.001; adjusted P<0.001). Reporting caution: the simple post-treatment gap is only 3.75 seconds, calculated from the displayed means. The larger adjusted estimate is insufficiently explained and should not be presented as an established 7.8-second causal benefit.
MBI: 70.65±11.53 versus 80.25±8.25; reported adjusted difference −9.454 points (95% CI −15.084 to −3.824; P=0.002; adjusted P=0.010).
FAC, Table 2: Post-treatment medians 3 (IQR 3–4) versus 4 (IQR 3.25–4); reported adjusted OR 0.148 (95% CI 0.037–0.587; P=0.007). The modeled category/reference direction is not explained sufficiently to translate this OR into a probability of independent walking.
Strength, Table 2: Seven manual muscle-testing comparisons favored the experimental arm, with reported adjusted ORs ranging 0.050–0.230. For example, hip-extension OR 0.050 (95% CI 0.008–0.316; P=0.001) and ankle-dorsiflexion OR 0.230 (95% CI 0.056–0.938; P=0.040). These are ordinal-model estimates, not relative reductions in disability.
Exploratory BDNF: Post-treatment serum concentrations 13.99±2.87 versus 16.63±3.38 ng/mL; adjusted difference −2.629 (95% CI −4.502 to −0.757; P=0.007; adjusted P=0.035).
Exploratory VEGF: 105.06±8.03 versus 112.21±8.10, reported in ng/mL; adjusted difference −7.479 (95% CI −12.714 to −2.244; P=0.006; adjusted P=0.030).

Blood was collected immediately after the final training session. Peripheral biomarker differences are not evidence of brain tissue repair, and this design does not distinguish an acute exercise response from a sustained biological change.

Adverse Events / Safety:

The authors report no adverse events, serious adverse events, treatment-related complications, or withdrawal-requiring protocol deviations during four weeks. Only 20 participants received BFRT, with thrombosis excluded and close therapist monitoring. This does not establish safety in high-risk patients or detect uncommon harm. Symptomatic intracranial hemorrhage, recurrent stroke, mortality, and falls are not separately tabulated as clinical endpoints.

Figures

Study enrollment and allocation flowchart — *Figure 1.* Flowchart of the study.

Lower-limb muscle-strength outcomes — *Figure 2.* Box plot showing between-group comparisons of lower limb muscle strength before and after treatment, including hip flexion, hip extension, hip abduction, knee extension, knee flexion, ankle dorsiflexion, and ankle plantar flexion.

Motor function, mobility, walking, and daily-living outcomes — *Figure 3.* Between-group comparisons of FMA-LE, TUGT, FAC, and MBI scores before and after treatment. * P < 0.05 compared with before treatment; # P < 0.05 compared with the control group.

Primary balance-platform outcomes — *Figure 4.* Between-group comparisons of Pro-Kin balance test parameters before and after treatment, including limit of stability, movement trajectory area (eyes open and eyes closed), and movement trajectory length (eyes open and eyes closed). * P < 0.05 compared with before treatment; # P < 0.05 compared with the control group.

Circulating BDNF and VEGF outcomes — *Figure 5.* Between-group comparisons of serum BDNF and VEGF levels before and after treatment. * P < 0.05 compared with before treatment; # P < 0.05 compared with the control group.

Criticisms

Small, selected sample: One center, 20 exposed patients, relatively young participants who could already walk with assistance, and approximately five to six weeks since stroke. Applicability to nonambulatory patients, older adults, unstable vascular disease, or chronic stroke is uncertain.
No sham intervention: Assessor blinding helps, but visible cuff treatment leaves expectations, attention, and therapist behavior as competing explanations for part of the incremental benefit.
Training-to-test overlap: The platform used for VFT also produced the primary balance measurements. Better test performance is useful, but generalization to everyday walking, falls, and durable participation requires separate confirmation.
Analysis transparency: Adjusted differences, ordinal-regression effects, and adjusted P values appear in Results without sufficiently detailed methods. Allocation concealment and a clearly prespecified hierarchy among multiple outcomes are not established by the report. No attrition occurred, so missing follow-up data are not the main concern.
Numerical inconsistency requiring clarification: The TUGT adjusted difference of 7.817 seconds is not readily reconciled with the displayed 3.75-second post-treatment difference. The paper also misdirects one muscle-strength cross-reference to Table 4, which actually contains the primary balance results.
Dose and safety uncertainty: Fixed cuff pressure without individual occlusion measurement leaves actual vascular restriction unknown. Absence of events in 20 treated participants over four weeks cannot exclude clinically important uncommon complications.
Mechanism overstated in the abstract: Serum BDNF/VEGF increases support a peripheral biological response, not the asserted cerebral neurovascular mechanism. The more cautious full-text conclusion deserves priority.
Subtypes pooled: Including both ischemic and hemorrhagic stroke does not substitute for powered, prespecified subtype analyses.

Funding

The Rehabilitation Hospital of Chongqing Medical University in-hospital Scientific Research Fund supported the study (RHCQMU2025006), as did the Shizhu Tujia Autonomous County Science and Technology Bureau (Skj2025006). The funding statement says funders had no role in study design, data collection or analysis, publication decisions, or manuscript preparation. The authors declared no competing interests.

Registration: ChiCTR2500104181; this is a Chinese Clinical Trial Registry identifier, not an NCT identifier. Ethics approval: Rehabilitation Hospital of Chongqing Medical University, 2025 Ethics Approval No. 10.

The paper

Authors. Hai Jiang et al.
Title. Effects of blood flow restriction combined with visual feedback training on lower limb function and the serum levels of BDNF and VEGF in patients with stroke.
Journal. PLOS One. 21(9): e0358740.
Year. 2026; published September 28, 2026.
DOI. 10.1371/journal.pone.0358740
PMCID. PMC13618907

Deep Dive — click to expand

What this is

A look-back selection published four days before this review, this trial asks whether adding a peripheral training stimulus to an existing balance-learning task produces more recovery without adding another therapy session. Forty participants received four weeks of rehabilitation; half also wore an inflated thigh cuff during visual feedback training. The strongest result is better measured balance, accompanied by favorable motor and daily-living scores. The weakest leap is treating circulating growth-factor changes as evidence that the brain repaired itself. The numbers justify replication, not a universal cuff protocol.

1. Shadow Audit

Three quiet details matter more than the abstract’s broad conclusion. First, 17 of the 40 participants had hemorrhagic rather than ischemic stroke; the pooled result contains no subtype-specific efficacy or safety estimate. Second, the intervention was a fixed 160–200-mmHg cuff exposure without measured individual occlusion pressure, so a nominally identical treatment may have produced different physiological doses. Third, the TUGT means imply a 3.75-second post-treatment separation while the adjusted model reports 7.817 seconds. This discrepancy does not erase the balance signal, but it makes confident quoting of the mobility effect premature. Finally, the authors’ own full-text conclusion retreats from the abstract’s mechanistic claim: peripheral BDNF and VEGF are not direct demonstrations of cerebral plasticity.

2. Inversion Engine

What would need to be true for the favorable interpretation to fail? On the reported adjusted limit-of-stability scale, the experimental advantage is 15.65 units, with a lower confidence bound of 8.75 units after reversing the control-minus-experimental sign. A sensitivity scenario that removed an 8.75-unit advantage would move that interval’s lower boundary to zero; removing the point advantage would require 15.65 units. These are editorial sensitivity thresholds, not bias-adjusted trial estimates. A sham-controlled replication could test whether expectancy and practice on the assessment platform explain that much of the observed difference. More importantly, the practical conclusion could fail even if the platform result replicates: equal falls, equal community participation, or a disappearing advantage after treatment would mean better short-term test performance did not buy durable independence.

3. Second-Order Catalyst

If the effect is real, the first credible adopters are specialist inpatient rehabilitation teams already using visual feedback platforms—not emergency stroke pathways or unsupervised home programs. The population nearest the evidence is a stable, cooperative patient in early recovery who can stand and walk with assistance. The fastest useful change is institutional: develop a supervised research pathway with explicit vascular-risk exclusions, measured occlusion dose, and independent functional follow-up. Adoption speed should be governed by safety and replication rather than the convenience of adding a cuff to an existing session. This is an implementation hypothesis, not a guideline recommendation.

4. Asymmetric Leverage

The operational attraction is concurrent treatment: 20 sessions of VFT over four weeks, with BFRT added during the same 20-minute blocks. That is 400 minutes of cuff exposure per treated participant, calculated from the schedule, without an additional exercise session. If the effect translates into daily life, a modest improvement across an existing rehabilitation service could be valuable. But “no extra session” is not “no extra cost”: cuff fitting, monitoring, staff training, and eligibility screening remain necessary. The observed MBI post-treatment gap is 9.60 points; the reported adjusted advantage is 9.454 points. Neither is a demonstrated reduction in caregiver hours or institutional placement. With zero reported events among 20 exposed patients, the approximate rule-of-three upper bound is 15% per patient over this short observation period—an illustration of imprecision, not an estimated complication rate.

5. Paradigm Destroyer

The habit this paper challenges is assuming that visual feedback alone exhausts the useful stimulus available during a low-load rehabilitation session. It does not destroy the need for conventional rehabilitation, individual vascular safety assessment, or patient-important endpoints.

A defensible two-sentence protocol update: “Do not use serum BDNF or VEGF increases as evidence that a rehabilitation treatment repaired the brain. Consider BFRT plus visual feedback only through an appropriately supervised evaluation pathway, with individualized safety assessment and outcomes beyond the training platform.”

MVP — Minimum Viable Proof

The bedside-changing proof would be a reproducible improvement in meaningful mobility or independence when measured by blinded evaluators, sustained after the intervention, without an unacceptable increase in vascular or other complications. This trial supplies a candidate signal: reported adjusted FMA-LE advantage 2.371 points (95% CI 1.061–3.681) and MBI advantage 9.454 points (95% CI 3.824–15.084), with signs reoriented to experimental benefit. It does not establish a patient-important threshold, durability, or adequately precise safety. The minimum next trial should clarify the TUGT model, use a credible sham cuff, specify one primary functional endpoint, and measure the actual restriction dose.

Best Combination

A relevant earlier randomized study by Li, Liu, and Xiong in 2025 enrolled 34 stroke patients and compared cycle-ergometer training with higher versus low sham-like cuff pressure. Its full text describes Doppler assessment of arterial occlusion pressure, intervention pressure at 40–80% of that value, and a control pressure of 10%. That study reported favorable muscle and motor outcomes, but it used a different exercise mode, schedule, and selected population (primary study, DOI 10.1002/brb3.70683). It therefore supports the plausibility of a peripheral adjunct, not direct replication of this fixed-pressure VFT protocol.

The useful combination is methodological: retain Jiang et al.’s concurrent balance-learning intervention, borrow the earlier study’s measured-dose and sham-comparator approach, and add longer-term independent mobility and safety outcomes. These small, heterogeneous studies should not be casually pooled into one universal effect size. Their common message is that the question deserves a better trial, not that any cuff regimen is interchangeable with another.

Overvalue Warning

Do not promote a surrogate into a mechanism. A post-session rise in circulating BDNF or VEGF does not establish brain angiogenesis, neural repair, or mediation of functional recovery. The study lacks the central imaging or electrophysiological measurements needed to examine that claim.
Do not promote a selected short trial into a universal safe treatment. Twenty BFRT recipients, fixed pressure, no sham condition, pooled stroke subtypes, and four-week follow-up cannot justify applying the regimen to every stroke survivor. Clarify the analysis and establish durable patient-important benefit before equating statistical significance with a standard-of-care change.

Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.