Computational Study Finds Looser Roller Pump Occlusion Does Not Always Reduce Hemolysis

Extracorporeal life support systems provide critical intervention during severe cardiopulmonary failure, yet the mechanical hardware required to circulate blood introduces profound hemolytic risks. In modern critical care, roller pumps and centrifugal pumps represent the primary mechanical drivers utilized across cardiopulmonary bypass, hemodialysis, and veno-arterial extracorporeal membrane oxygenation. As red blood cells traverse these pumps, they encounter extreme mechanical shear stresses that disrupt erythrocyte membranes, precipitating intracellular hemoglobin leakage into circulating plasma. In prolonged therapies such as veno-arterial extracorporeal membrane oxygenation, which frequently extend across days or weeks, continuous hemolysis progressively depletes circulating haptoglobin, the physiological protein tasked with scavenging free hemoglobin. Clinical registry data across 9,111 pediatric patients indicates that centrifugal pumps cause significantly higher rates of clinically significant hemolysis than roller pumps, making blood trauma mitigation essential.

Within roller pumps, the mechanical occlusion setting established between the rotating roller and the stator housing acts as the primary determinant of hemolysis. Despite laboratory evidence demonstrating that operating roller pumps under slightly loose, under-occlusive configurations substantially lowers hemolysis compared to full occlusion, clinical perfusionists almost universally operate roller pumps in fully occlusive states. This entrenched practice persists because clinicians fear that retrograde backflow escaping across an under-occlusive gap introduces flow measurement inaccuracies. Consequently, circuits remain fully occluded even though experimental bench tests show minor under-occlusion curbs cellular damage while diminishing pump flow by less than ten percent. Furthermore, computational engineering research aimed at optimizing roller pump mechanics has lagged far behind centrifugal pump development, where design improvements have largely reached saturation. Computational roller pump studies remain scarce and methodologically incomplete, leaving clinicians without rigorous engineering data to systematically balance flow delivery against blood cell trauma across varying degrees of under-occlusion.

Existing computational fluid dynamics studies evaluating roller pumps have relied on two distinct numerical approaches, each characterized by severe physical limitations. The first approach utilizes three-dimensional fluid-structure interaction simulations that capture the complex deformation of peristaltic tubing during roller compression. However, these two-way coupled simulations demand extreme computational expense and suffer from chronic divergence issues, forcing researchers to evaluate clearance gaps far exceeding clinically relevant dimensions without incorporating hemolysis models. The second approach employs computationally efficient two-dimensional dynamic mesh simulations that prescribe wall deformation through analytical user-defined functions. Crucially, all prior computational studies applied oversimplified boundary conditions, imposing zero pressure gradients or idealized no-flux parameters across the pump inlet and outlet. In clinical practice, particularly veno-arterial extracorporeal membrane oxygenation, circuits generate substantial outlet discharge pressures reaching 300 millimeters of mercury against negative inlet suction. By omitting realistic pressure gradients, previous simulations failed to replicate authentic backflow mechanics, producing data detached from genuine clinical operating environments.

To resolve these longstanding computational shortcomings, researchers formulated an integrated evaluation framework combining high-fidelity two-dimensional computational fluid dynamics in Ansys Fluent 2024 R2 with experimental hydrodynamic measurements. The experimental bench loop comprised a fluid reservoir, a magnetic-inductive flow meter, inlet and outlet pressure sensors sampling at 200 hertz, a custom two-roller pump with micrometer-scale gap adjustment, and a downstream throttle control valve. To reproduce the hydraulic resistance imposed by clinical oxygenators and vascular cannulas, the downstream valve was calibrated under fully occlusive pumping at thirty revolutions per minute. This baseline calibration established an operating flow rate of 6.0 liters per minute against an outlet pressure of 350 millimeters of mercury, directly matching hemodynamic conditions observed in extracorporeal membrane oxygenation therapy. Once calibrated, the throttle valve position remained strictly locked throughout all subsequent measurements, ensuring that every investigated under-occlusion setting operated against an authentic, experimentally derived afterload that informed computational boundary conditions.

The experimental physical apparatus featured an aluminum stator housing with an inner diameter of 280 millimeters encasing a flexible Tygon tube having an internal diameter of 19 millimeters and a wall thickness of 3.2 millimeters. The central rotor mechanism incorporated a leading roller and a trailing roller with a 180-degree occlusion angle, supported by four guiding rollers that prevented vertical tube displacement and provided a 100-milliliter displacement volume. To achieve micrometer-level gap calibration, two pairs of linear stepper motors actuated four trapezoidal lead screws to position the roller carriers along dual linear guides. As the working circulating medium, the researchers utilized a calibrated water-glycerol solution matching blood dynamic viscosity of 3.5 centipoise and density of 998.2 kilograms per cubic meter. Because shear-induced red blood cell destruction occurs at extreme shear stresses exceeding 400 pascals, the Newtonian fluid analog avoided anticoagulation handling challenges while maintaining rigorous hydrodynamic equivalence for comparative evaluation.

The research team systematically investigated five distinct under-occlusion configurations, designated uo1 through uo5, by progressively retracting the roller carriers to enlarge the clearance gap. Each setting was defined by its relative flow reduction compared to full occlusion, alongside corresponding measured circuit pressures. Setting uo2 produced a 25 percent flow reduction to 4.5 liters per minute at 174 millimeters of mercury, corresponding to a 0.3-millimeter gap in the CFD simulation. Setting uo3 yielded a 37 percent flow reduction to 3.8 liters per minute at 139 millimeters of mercury across a 0.41-millimeter gap. Setting uo4 reached a 50 percent flow drop to 3.0 liters per minute across a 0.618-millimeter gap, while setting uo5 yielded a 63 percent flow loss to 2.2 liters per minute across a 0.9-millimeter gap.

To resolve these complex flow dynamics, the computational fluid domain was discretized into approximately 314,000 computational cells partitioned across three specialized meshing zones. Rigid inlet and outlet tubing segments were structured with quadrilateral elements, while the dynamic peristaltic core utilized unstructured triangular elements with a 0.3-millimeter characteristic size governed by remeshing and smoothing algorithms. Near-wall prism layers along the deforming inner tube were preserved using smoothing techniques, maintaining a first-layer cell height of 0.007 millimeters to satisfy near-wall turbulence resolution criteria. Dynamic mesh motion was stabilized using a boundary distance diffusion function, with diffusion parameters scaled from 1.7 down to 0.5 as gap dimensions contracted. Transient flow simulations ran with a conservative numerical time step of 0.0005 seconds, corresponding to 0.1 degrees of roller rotation per step. Periodic stability was formally attained by the fifth pumping cycle, where consecutive cycles showed identical values, allowing an Eulerian shear-induced hemolysis model to track cumulative blood damage across impermeable walls.

Analysis of computational velocity fields revealed that every under-occlusive configuration generated an intense reverse fluid jet passing backward through the clearance gap toward the negative inlet pressure zone. This retrograde jet generated extreme von Mises shear stresses concentrated almost entirely within the narrow constriction and its immediate downstream wake, persisting throughout the compression phase. In the tightest configuration, uo1, the narrow 0.125-millimeter gap restricted backflow leakage, maintaining a stable, moderate jet without widespread turbulence. However, as the gap expanded in uo2 and uo4, the stronger reverse flow triggered flow instability and formed energetic swirling vortex structures behind the advancing roller. Crucially, the simulations demonstrated that the majority of cellular trauma originates directly within the high-velocity clearance jet where local shear stresses far surpass physiological tolerance. The downstream swirling vortices behind the roller do not significantly generate new hemolysis; instead, they serve as convective transport mechanisms that distribute damaged cells throughout the pump domain before downstream discharge.

To evaluate cumulative blood trauma, the researchers calculated the velocity-weighted average hemolysis index near the pump outlet, followed by time-integration across full pumping periods. The resulting data revealed a surprising, non-linear relationship between occlusion clearance and cellular destruction. Rather than producing a steady decline in hemolysis, the three tightest settings formed an unexpected damage plateau. Setting uo2 generated the highest total hemolysis of all investigated configurations, exhibiting approximately seven percent greater hemolysis than the tightest setting, uo1, despite suffering an additional fifteen percent reduction in delivered volumetric flow. Setting uo3, which experienced a 37 percent flow reduction, generated total hemolysis levels virtually indistinguishable from uo1. This counterintuitive behavior highlights the competing fluid dynamics governing roller pump performance: while widening the physical gap theoretically lowers localized geometric shear, it simultaneously permits a larger volume of blood to leak backward under the trans-pump pressure gradient, intensifying reverse jet velocity and counteracting geometric relief.

Significant reductions in blood damage occurred only when clearance gaps expanded to extreme dimensions in settings uo4 and uo5, where maximum local hemolysis dropped from 1.8 percent in uo1 and 0.8 percent in uo2 down to roughly 0.12 percent in uo4. However, this late hemolytic improvement came at the severe expense of clinical utility, cutting flow output by half or more and crippling circulatory support. The authors noted several physical limitations, including idealized two-dimensional planar tube compression that neglects three-dimensional corner leakage channels, the use of empirical Eulerian hemolysis equations, and the application of a non-cellular Newtonian fluid analog. Additionally, while the experimental pump featured 19-millimeter tubing compared to clinical 12.7-millimeter circuits, hydrodynamic Reynolds numbers remained closely matched between 1,800 and 2,700 within the transitional regime, confirming shared physical flow behavior.

Subject of Research: Computational assessment of hemolysis and hemodynamics in under-occlusive roller pumps under clinically relevant operating conditions

Article Title: Methodology for computational assessment of hemolysis in under‐occlusive roller pumps operated at clinically relevant conditions

Article References: Kohút, J., Jagoš, J., Suk, P., Formánek, M., & Burša, J. (2026). Methodology for computational assessment of hemolysis in under‐occlusive roller pumps operated at clinically relevant conditions. Bioengineering & Translational Medicine, Article e70191. https://doi.org/10.1002/btm2.70191

Image Credits: AI Generated

DOI: 10.1002/btm2.70191

Keywords: roller pumps, hemolysis, computational fluid dynamics, under-occlusion, extracorporeal membrane oxygenation, shear stress, cardiopulmonary bypass, peristaltic tubing, plasma free hemoglobin, boundary conditions, dynamic meshing, backflow jet, fluid-structure interaction, haptoglobin depletion, Reynolds number, Ansys Fluent