Design and In-Vitro Evaluation of an Artificial Placenta made from Hollow Fiber Membranes

Abstract

For infants born at the border of viability, care practices and morbimortality rates vary widely between centers. Trends show significant improvement, however, with increasing gestational age and weight. For periviable infants, the goal of critical care is to bridge patients to improved outcomes. Current practice involves ventilator therapy, resulting in chronic lung injuries. Research has turned to artificial uterine environments, where infants are submerged in an artificial amniotic fluid bath and provided respiratory assistance via an artificial placenta. We have developed the Preemie-Ox, a hollow fiber membrane bundle that provides pumpless respiratory support via umbilical cord cannulation. Computational fluid dynamics was used to design an oxygenator that could achieve a carbon dioxide removal rate of 12.2 mL/min, an outlet hemoglobin saturation of 100%, and a resistance of less than 71 mmHg/L/min at a blood flow rate of 165 mL/min. A prototype was utilized to evaluate in-vitro gas exchange, resistance, and plasma free hemoglobin generation. In-vitro gas exchange was 4% higher than predicted results and no quantifiable plasma free hemoglobin was produced.

Introduction:

Complications of premature birth, birth before 37 weeks gestational age (GA), are the leading cause of global neonatal mortality.¹ Pre-term infants born between 22-28 weeks GA (extremely premature infants, EPI)² are at the highest risk for morbimortality.¹ Pre-term infants born at least 25 weeks into gestation are termed ‘viable,’ as the majority, 67-75%,¹ survive with intensive medical care. Those infants born prior to 22 weeks of gestation or who weigh less than 500g are not considered viable as intensive medical care has shown to be futile.³ EPIs that are born between the 22–25-week period are considered ‘periviable’. The estimated effectiveness of intensive care at the periviable stage requires an individual assessment of the developmental maturity, birth weight, clinical condition, patient response to care, and parental judgement. If periviable EPIs are considered candidates for support, the goal of care is to bridge them to the viability milestone.

A fetus of 22-23 weeks GA has not yet begun to produce surfactant or undergo alveolar differentiation.⁴ As a result, clinical care of periviable EPIs is focused on pulmonary support. The fetus is provided with steroids to stimulate lung development, surfactant to decrease the work of breathing, and mechanical ventilation (MV) to increase oxygenation and CO₂ removal rates (vCO₂).⁴ While MV is necessary to prevent respiratory failure, the forceful expansion of the lungs and oxygen rich atmosphere disrupts vascular endothelial growth factor signaling in the lung. This results in bronchopulmonary dysplasia (BPD), a condition characterized as simplified alveoli, variable increases in interstitial cellularity and/or fibroproliferation, and subnormal capillary counts.^2,4 All EPIs treated with MV develop some degree of BPD and are susceptible to neurodevelopmental delay, suppressed respiratory function, long term cardiovascular impairments, and growth failure.²

The use of extracorporeal membrane oxygenation (ECMO) would provide the oxygenation and CO₂ removal support that EPIs require while avoiding MV related iatrogenesis. Only infants who are 34 weeks of GA or who weigh at least 2 Kg are considered candidates for ECMO.^5,6 Younger and lighter infants have underdeveloped cardiovascular systems that are unable to successfully amalgamate with currently available hollow fiber membrane oxygenators (HFMOs). The relatively large priming volume and vascular return pressures of currently marketed HFMOs induce circulatory overload while large surface areas require high levels of anticoagulation and may produce excessive hemolysis.⁵ This exacerbates the pre-existing risk for thrombosis and hemorrhage of an EPI due to underdeveloped vasculature⁷ and clotting factors.^5,8 In addition, the environment in which ECMO therapy is practiced is not ideal for EPIs. Premature pulmonary exposure to a gaseous environment and the resulting fetal shunt occlusion is not prevented. A superior therapy for the EPI is a system that supplies oxygen and removes CO₂ independent of the lungs while allowing the infant to subsist in a fluidic environment. The submersion of a fetus within an artificial amniotic fluid bath, the artificial uterus, with an externally accessible HFMO, the artificial placenta achieves this goal. The Children’s Hospital of Philadelphia, a collaborative group between the University of Australia and Tohoku University Hospital, and a Canadian group with The Hospital for Sick Children have employed this strategy in premature lamb^6,9-19 or pig models.²⁰ An alternative method to fetal submersion, practiced by the University of Michigan,^6,21-29 is fetal intubation with a fluid filled endotracheal tube. All three groups access fetal circulation by placing drainage cannula(s) in the umbilical arteries (UAs) and a return cannula in either the umbilical vein (UV)^15,19,20 or interior jugular vein.²⁷ The limiting factor for acute and chronic studies performed with artificial uterine environments has been iatrogenesis and study complications from the HFMO and its integration into the fetal circulation, primarily due to HFMO size and geometry.^{5,15,19,20,25,27}

We are developing a highly efficient HFMO that is smaller than any oxygenator used in the field. The size of the oxygenator typically accounts for most of the surface area present in an ECMO circuit. Decreasing device surface area also mitigates the likelihood of circulatory overload, required anticoagulation to achieve therapeutic levels, and exposure of blood to foreign surface area. In addition, bundle geometry was configured to minimize device resistance. This allows circuit flow to be driven by the fetal heart and reduces heart afterload, curtailing chances of heart failure due to hypertension. The requirements for this device include a priming volume less than 30 mL, a resistance less than 71 mmHg/L/min, a vCO₂ of 12.2 mL/min and an outlet hemoglobin saturation of 100% at a blood flow rate of 165 mL/min, and a normalized index of hemolysis (NIH) less than 0.05 g/100L. A HFMO geometry that met these requirements was modeled, manufactured, and characterized in-vitro with gas exchange and hemolysis studies.

Materials and Methods:

Computational Design and Flow Evaluation:

Bundles of varying geometries were computationally evaluated for gas exchange efficiency, priming volume, and pressure drop prior to prototype manufacturing. Previously published oxygenation and CO₂ removal mass transfer correlations slightly modified to incorporate the Haldane Effect were used to predict the gas exchange capabilities of 52 potential geometries. Of the 11 geometries that met both gas exchange and resistance requirements, the device with the lowest surface area was selected to be manufactured. The HFMO is composed of polymethylpentene (PMP) fibers (OXYPLUS^™, 3M^™ MEMBRANA^™) and has a diameter of 2.5 cm, a length of 3.2 cm, a surface area of ~0.1 m² (Fig. 1), and a porosity of 0.48. Computational fluid dynamic analysis of the flow distribution in the device was completed using the Free and Porous Media Flow physics of COMSOL Multiphysics (COMSOL INC., Stockholm, Sweden). Blood was modeled as an incompressible fluid with a density of 1050 kg/m³ with a dynamic viscosity of 2.9 cP. The bundle was modeled as a single porous medium with the permeability quantified using a modified Blake-Kozeny equation.³⁰

Figure 1:

Preemie-Ox prototype. The HFMO is assembled into a custom machined test housing shown in the image of the prototype.

Resistance apparatus and measurement:

The resistance apparatus consisted of a 3.5 cm inner diameter acrylic tube with height increments marked at 6, 10, 12, 14, and 16 cm above the outlet of the device. The bottom of the tube was sealed with an acrylic disk and a 3/16” port was introduced tangent to the bottom of the column. A piece of connection tubing directed flow from the column to the device. The column and device were connected such that there was no vertical or horizontal gap between the outlet connector of the column and the inlet connector of the device. A 3 cm length of 3/16” tubing was placed on the outlet connector of the device to control fluid flow via a tubing clamp (Fig. 2). The fluid was a fetal blood analogue made with carboxymethycellulose sodium salt (Sigma Aldrich, St. Louis, MO) at a dynamic viscosity of 2.9 ± 0.1 cP. The device was primed with the solution prior to column connection. After connection, the resistance fixture was filled with the blood analogue above the height mark of 16 cm. The tubing clamp was removed from the device outlet and the passage of fluid through the column and the device was video recorded. The time from an in-frame stopwatch (Traceable Stopwatch, Thomas Scientific, Swedesboro, NJ) was used to calculate the elapsed time for the blood analogue to pass from each height increment to the final height. The protocol was performed in triplicate.

Figure 2:

Resistance apparatus configuration. Fluid from the apparatus drains into a collection reservoir (not pictured). Scissor lab jacks were used to keep the reservoir and device level.

In this experimental design the device will account for the majority of the resistance the fluid will encounter as it drains from the column. Flow resistance, R, integrated over the device in the vertical direction yields:

$$R=\frac{P_0}{Q}$$ [1]

where Q is the volumetric flow rate, given by Q = V₀/A (superficial velocity, V₀, and column cross sectional area, A). The gage pressure, P₀, depends on the height of the fluid, h(t), in the column as a function of time:

$$P_0=\rho gh(t)$$ [2]

where g is gravitational acceleration and ρ is the fluid density. The conservation of mass, V₀ = −dh/dt, relates the two equations and can be integrated to yield:

$$\text{ln}\left(\frac{h_i}{h_f}\right)=\frac{\rho g}{AR}\Delta t$$ [3]

where Δt is the time span it takes the fluid to drop from an initial (h_i = 16, 14, 12, or 10 cm) to final height (h_f = 6 cm).

In Vitro Gas Exchange:

Gas exchange was performed using abbatoir adult bovine blood (8200811, Lampire Biological Laboratories, Pipersville, PA) heparinized upon collection (1000 IU/mL). The blood was treated with gentamicin (0.1 mg/mL) upon arrival in the laboratory and was used within 24 hours of the bleed date. A hemoglobin of 11 ± 1 g/dL was achieved by diluting blood with phosphate buffered saline.

The test circuit (Fig. 3) was composed of the device, two compliant blood reservoirs (MVR 800, Medtronic Minneapolis, MN), a Pedimag (Abbott Laboratories, Chicago, IL), and an Affinity oxygenator (Medtronic, Minneapolis, M). Both reservoirs were submersed in a heated water bath maintained at a temperature of 37 ± 1°C. The Pedimag recirculated blood through a single reservoir while the Affinity conditioned blood to EPI venous conditions (PCO₂ = 42.5 ± 5 mmHg and PO₂ = 26.5 ± 5 mmHg).³¹ Blood flow rate was maintained at 165 mL/min and was ultrasonically monitored using a Transonic flow probe (Transonic Systems Inc., Ithaca, NY). Once conditioned, blood flow was directed to follow a single pass through the test circuit to the second empty reservoir. Three sweep gas flow rates, 750, 900, and 1000 mL/min, were randomly cycled through for a total of 3 samples per sweep gas flow rate. A WMA-4 CO₂ analyzer (PP Systems, Amesbury, MA) monitored the CO₂ concentration in the exhausted sweep gas from the device. Once gas exchange reached steady state, blood samples were taken from the device inlet and outlet for blood gas analysis (Rapidpoint 405 blood gas analyzer, Siemens, Deerfield, IL). Steady state was achieved when the concentration of CO₂ in the exhaust sweep gas changed by less than 10 ppm. Oxygenation and vCO₂ were calculated according to previously published equations^32,33 and normalized to targeted venous conditions.

Figure 3:

Schematic of gas exchange circuit. Clamps direct the recirculation of blood through the venous reservoir during conditioning. Once conditioning is complete blood flow is directed through the circuit to the empty reservoir. During this single pass flow gas flow via the de-oxygenator is stopped and pure oxygen sweep gas is flowed through the artificial placenta. Once the conditioned blood is depleted, the circuit is converted back to a recirculation circuit so blood can be conditioned to venous values.

In Vitro Hemolysis:

Hemolysis testing followed ATSM standard F1841-19. Blood was acquired and treated in the same manner as in-vitro gas exchange experiments. The control circuit consisted of a Pedimag and a Paragon Neonatal (Chalice Medical, Nottinghamshire, UK) modified to exclude the heat exchanger. The experimental circuit consisted of the Preemie-Ox and a Pedimag. Pedimags were set to 1000 ± 100 RPM resulting in both circuits having statistically equivalent flow rates (227 ± 6 mL/min, p=0.1). Testing was performed for 6 hours with samples of hemoglobin (Rapidpoint 405 blood gas analyzer, Siemens, Deerfield IL), hematocrit (IEC Mb, Microcentrifuge, International Equiptment CO.), protein concentration (Reichert TS 400 Refractometer, Reichert, Inc., Depew, NY), and plasma free hemoglobin (pfHb) evaluated hourly. To measure pfHb, whole blood was centrifuged (accuSpin Micro 17, Fisher Scientific, Hampton, NH) according to F1841-19 section 8.8 and spectrophotmetrically analyzed (Genesys 10 UV-Vis Spectrophotometer, Thermo Fisher Scientific, Waltham, MA) at 540 nm. Absorbance values were correlated to pfHb concentration using a calibration curve generated from samples of known pfHb. A therapeutic index of hemolysis (TIH) and normalized index of hemolysis (NIH) were calculated according to standard formulas for both the control and experimental circuit.

Results:

Computational Design and Flow Evaluation:

Computations predicted a vCO₂ of 12.2 mL/min at 165 mL/min blood flow rate and an outlet saturation of 100%. CFD results showed uniform flow distribution throughout the bundle with axial flow having a coefficient of variation equal to 3.3% (Fig. 4(a)). Maximum shear was predicted to be 2.4 Pa and was located at the elbow of the outlet plenum. Minimum shear was predicted to be 0.02 Pa and was located in the main volume of the outlet plenum. Flow patterns showed no recirculation or separation in the plenums or bundle (Fig. 4(b)).

Figure 4:

CFD analysis results for 165 mL/min flow showing (a) fluid velocity through the fiber bundle (m/s) and (b) streamlines of velocity magnitude through the device (m/s).

Resistance apparatus and measurement:

The resistance of the device was 51 ± 0 mmHg/L/min across all tested fluid column height changes (Fig. 5). Total priming volume of the device was 15 mL.

Figure 5:

Graph of elapsed time vs. ln(h_i/h_f). The slope of the linear relationship was used to determine the device resistance according to Eq. 3.

In Vitro Gas Exchange:

A one-way ANOVA showed that the vCO₂ was statistically equivalent between the three sweep gas flow rates tested (data not shown, p = 0.22). Mean vCO₂ averaged over all three sweep gas flow rates was 12.7 ± 0.9 mL/min at a blood flow rate of 163 ± 2 mL/min (n=9). This is within 4% of computational predictions. Hemoglobin was completely oxygen saturated in all three cases (SO₂ = 99.4 ± 0.4).

In Vitro Hemolysis:

Fig. 6 shows pfHb over time for the control and experimental circuit. Levels of pfHb did not change over time according to a repeated measures ANOVA (p = 0.24). As there was no pfHb generated over time, sensible values of NIH and TIH could not be calculated.

Figure 6:

Plasma free hemoglobin (pfHb) concentration over time for experimental and control circuits. Levels of pfHb did not statistically increase over time for either circuit (repeated measures ANOVA, p=0.24).

Discussion:

The goal of the research presented in this publication was to design and test a HFMO that mimics the resistance and gas exchange capabilities of a placenta from an EPI. This goal was achieved using a mathematical model of mass transfer and computational fluid dynamics verified with benchtop testing. Our laboratory has developed an HFMO that has a surface area that is 33-88% smaller and a priming volume that is up to 81% smaller than the devices utilized in other published studies (Table 1).

Table 1:

Hollow Fiber Membrane Oxygenators used in Artificial Placenta Research

Research Group	Device Name	Publications	Device Surface Area [m2]	Device Priming Volume [mL]
University of Michigan	Capiox Baby RX	7, 22-30	0.50	43
	Medos HiLite 800 LT		0.32	55
Children’s Hospital of Philadelphia	Quadrox iD Pediatric	17-20	0.80	81
	Quadrox iD Neonatal		0.38	38
University of Australia / Tohoku University Hospital Collaboration	EVE System	10-16	0.15	18
The Hospital for Sick Children	Quadrox iD Pediatric	21	0.80	81
	Quadrox iD Neonatal		0.38	38
	Rabbit Oxygenator		-	15
University of Pittsburgh	Preemie-Ox	-	≈ 0.1	15

Ensuring that the artificial placenta circuit should have the same, or lower, resistance as the native placenta minimizes the potential for afterload induced congestive heart failure. The minimum placental resistance that a fetus experiences from mid-gestation to term is 100 mmHg/L/min.^31,34 The artificial placenta circuit is composed of three cannulas, two for the UAs and one for the UV, tubing, and the HFMO. To account for the resistance of the cannulas and tubing, the Hagen-Poiseuille equation was used to predict the pressure drop across two 9 Fr arterial cannulas with a length of 5 cm in parallel, a 16 Fr venous cannula with a length of 5 cm, and 36” of 3/16” tubing. The inner diameters of the UAs and UV cannulas were estimated to be the same diameter of the respective arteries at 22 weeks GA.³⁵ Total expected resistance of a circuit with this composition is 29 mmHg/L/min, leaving 71 mmHg/L/min available for the HFMO. The measured resistance of the Preemie-Ox is 33% less than this requirement, minimizing the chance of heart failure due to supraphysiologic resistance of the artificial placenta circuit.

Three sweep gas flow rates were tested to evaluate the point at which sweep gas independent vCO₂ is achieved. When the sweep gas flow rate to blood flow rate ratio is too low, the accumulation of CO₂ in the sweep gas reduces the gradient driving gas exchange, thereby reducing device efficiency. The regime in which this accumulation limits device efficiency is known as sweep gas dependent gas exchange. Federspiel et al showed that sweep gas independent vCO₂ is achieved when the gas flowrate is 40-60 times greater than the vCO₂.³⁶ This benchmark was achieved within the Preemie-Ox at the lowest tested sweep gas flow rate of 750 mL/min (60 ± 3). Increasing sweep gas flowrate, once in the sweep gas independent regime, will not increase device efficiency. The statistical equivalence of vCO₂ at all three sweep gas flow rates supports the evidence that sweep gas independence is achieved in the Preemie-Ox at a sweep gas flow rate of 750 mL/min. In clinical practice targeted arterial conditions will be achieved using a N₂/O₂ sweep gas mixture.

In this study hemolysis was evaluated at a flow rate higher than the intended use of the Preemie-Ox. This was done to ensure that both pump rpm and flow rate were similar between the two circuits. Having statistically different flow rates or rpms between the two circuits would not support a 1-to-1 comparison between the experimental and control device. Despite running the circuits at a blood flow rate higher than the maximum intended use of the Preemie-Ox, the generation of pfHb was not detected in either circuit. This result was expected as the root-mean-square shear stress within the device at this flow rate (0.8 Pa) is 35% smaller than the value that has shown to significantly increase cellular and molecular mediators of coagulation.³⁷ In addition, no material induced trauma was expected based on the relatively low surface area of the HFM bundle.³⁷ Logic follows that at the lower blood flow rate of 165 mL/min the experimental and control circuits would not generate any measurable quantity of pfHb as Shear stress within the circuit would decrease with blood flow rate.

A limitation of the in-vitro analysis of the Preemie-Ox is the use of adult bovine blood for hemolysis and gas exchange experiments. Pediatric patients treated with ECMO show trends of increasing hemolysis as patients decrease in age and weight.³⁸ Fetal red blood cells (RBCs) are known to have a shorter lifespan, a larger size, less deformability, and are more fragile than adult RBCs.³⁹ EPIs also typically suffer from anemia of prematurity resulting from a lack of maternal iron, a delayed EPO response, and phlebotomy losses from clinical testing.⁸ As a result of these factors, patients receive an increase in the number of adult blood transfusions correlating with a decrease in GA.⁴⁰ The quantification of the effect that the composition of fetal blood may have on device generated hemolysis is therefore problematic. The lack of plasma free hemoglobin generation resulting from benchtop studies, however, provides assurance that the use of the Preemie-Ox should not result in irreparable levels of hemolysis.

Adult Hb (HbA) contains two alpha and two beta chains each associated with a heme group. Fetal hemoglobin (fHb) structurally differs from HbA as it contains two alpha and two gamma chains. Gamma chains have an increased affinity for oxygen compared to the Beta chains present in HbA. In-utero this adaptation allows fetuses to achieve adequate gas exchange despite the relatively low, compared to atmospheric, oxygen content of placental blood.⁴ A HFMO placed into fetal circulation could achieve higher rates of oxygenation, up until complete Hb saturation, than when used at the same blood flow rate in an adult. The oxygenation rate of the Preemie-Ox is not expected to increase in the fetal setting as it achieves 100% oxygen saturation in adult blood. CO₂ removal, however, may increase in the fetal setting as the increased affinity of fHb for oxygen results in an increased offloading of Hb bound CO₂ into plasma. In this manner the CO₂ concentration gradient between the sweep gas and blood would increase, resulting in increased vCO₂. In this manner, testing the Preemie-Ox in adult blood represents a hypothetical worst-case scenario where the entire circulating fHb volume of the fetus has been replaced with HbA.

In conclusion, our laboratory modeled, manufactured, and benchtop tested a HFMO designed to be used as an artificial placenta within an artificial uterine environment. The HFMO met the 12.2 mL/min vCO₂ target while completely saturating hemoglobin at 165 mL/min. The resistance of the Preemie-Ox was 51 mmHg/L/min, 33% smaller than the targeted value of 71 mmHg/L/min. A low resistance device minimizes fetal heart afterload and decreases the likelihood of hypertensive heart failure in an in-vivo setting. No hemolysis was detected in either the control or experimental circuit over 6 hours, therefore values of NIH and TIH could not be calculated. Future work involves utilizing a phosphorylcholine coating to minimize the immune reaction to hydrophobic PMP fibers. The device will be reassessed post coating and tested in in-vivo fetal lamb studies.