Sequential Blood Filtration for Extracorporeal Circulation: Initial Results from a Proof-of-Concept Prototype

Abstract:

Micropore filters are used during extracorporeal circulation to prevent gaseous and solid particles from entering the patient’s systemic circulation. Although these devices improve patient safety, limitations in current designs have prompted the development of a new concept in micropore filtration. A prototype of the new design was made using 40-μm filter screens and compared against four commercially available filters for performance in pressure loss and gross air handling. Pre- and postfilter bubble counts for 5- and 10-mL bolus injections in an ex vivo test circuit were recorded using a Doppler ultrasound bubble counter. Statistical analysis of results for bubble volume reduction between test filters was performed with one-way repeated-measures analysis of variance using Bonferroni post hoc tests. Changes in filter performance with changes in microbubble load were also assessed with dependent t tests using the 5- and 10-mL bolus injections as the paired sample for each filter. Significance was set at p < .05. All filters in the test group were comparable in pressure loss performance, showing a range of 26–33 mmHg at a flow rate of 6 L/min. In gross air-handling studies, the prototype showed improved bubble volume reduction, reaching statistical significance with three of the four commercial filters. All test filters showed decreased performance in bubble volume reduction when the microbubble load was increased. Findings from this research support the underpinning theories of a sequential arterial-line filter design and suggest that improvements in microbubble filtration may be possible using this technique.

Arterial-line filters for extracorporeal circulation (ECC) are used to prevent gas and other foreign particles from entering the patient’s circulatory system. Although these devices improve patient safety (1), the problems faced in fully meeting their intended purpose are linked to the understanding that gas bubbles contained in a fluid flow can undergo changes in both quality and quantity (2). It is proposed that the surface active forces governing the use of micropore filters are less effective against bubbles that can change shape, be compressed into smaller volumes, or fragmented into smaller particles at the micron-pore interface (2). Consequently, present arterial-line filters used for ECC are unable to remove all free gas infiltrates from the extracorporeal blood supply returning to the patient (3–6). Although limitations in current filter technology contribute to problems associated with ECC, the push to advance surgical treatment of cardiovascular disease will likely continue to increase demand on perfusion system performance as a means of achieving better patient outcomes. Moreover, numerous investigators have emphasized the need for further improvements in component design and the air handling performance of ECC systems (7–12). This has prompted the development of a new concept in micropore filtration aimed at addressing current deficiencies as a step toward improved patient outcomes.

The new filter, as illustrated in Figure 1, includes multiple nonpleated micropore screens placed sequentially and adapted so that fluid passes from the inlet through the screens to the outlet following a unique up-and-down flow path to improve the depth of filtration while making more efficient use of a smaller screen surface area. Efforts to establish performance of the new design were carried out in two phases. After phase one, which included numerical modeling and analysis using computer-based software (13), prototype manufacturing and testing were completed in phase two. This article details the outcomes of phase two research used to analyze and evaluate a new concept in micropore filtration for ECC and compares a prototype of the new design with four commercially available arterialline filters.

Figure 1.

Inset graphic showing exterior view of the prototype filter. Cutaway showing interior view with sequential filter placement.

METHODS

Prototype Assembly

Astereolithography apparatus was selected as themethod for prototype manufacturing for its ability to produce transparent models capable of withstanding fluid flows and pressures commonly used during ECC. The unassembled model consisted of five parts, which required sanding and reworking for fit before assembly. In preparation for model assembly, a 40-μm medical-grade micropore screen fabric (Sefar AG, Heiden, Switzerland) was cut to measure and mounted on two filter support frames using a water-resistant epoxy. Mounting blocks were used to stabilize the support frames while attention was given to ensure that the screen fabric remained taut against all contact surfaces. Once both filter screens were fixed, the five parts were then assembled and bonded with additional water-resistant epoxy. After assembly, the finished prototype was used in wet laboratory simulations and compared against the Medtronic Affinity, Dideco D734, PALL AL8, and Maquet Quart arterial-line filters for performance in pressure loss and gross air handling. Specifications of the filters included in the test group are shown in Table 1.

Table 1.

Test filter specifications.

Filter	Filter Type	Prime Volume (mL)	Filter Screens	Total Surface Area (cm2)	Filter Density (cm2/mL)	Pore Size (μm)
Affinity	Concentric nonpleated	212	1	545	2.57	38
AL8	Radial pleated	170	1	630	3.71	40
D734	Radial pleated	195	1	655	3.36	40
Prototype	Sequential nonpleated	192	2	294	1.53	40
Quart	Planar nonpleated	180	12	570	3.17	40

Circuit Design and Setup

A modified version of the test circuit previously described by Herbst was used to conduct all studies (2). Modifications to the circuit included placing an arterial filter with purge line in the 180-cm length of 3/8-inch polyvinyl chloride tubing connecting the pump raceway outlet to reservoir R2 (Figure 2). The injection port for bolus delivery was placed 90 cm from the filter inlet, and the BCC200 bubble counter was set to monitor and record microbubbles between 10 μm and 500 μm in size.

Figure 2.

Circuit design showing flow direction and test filter placement.

The circuit without a test filter was first primed with normal saline .9% solution (9.0 L) at room temperature (23.6°C) for initial debubbling. To limit any effects that may result from differences in surface coatings between the test filters, 20% albumin (600 mL) was added to the debubbled circuit to standardize surface tension of the priming fluid (14,15). With the bypass loop attached, each filter was CO₂-flushed for 3 minutes and then clamped at the inlet and outlet ports before being added to the circuit for priming. All test filters were handled until completely debubbled with the purge port open. For the Maquet Quart filter, the three-way tap at the top of the filter housing was manipulated to ensure complete debubbling of both pre- and postfilter chambers and then kept open in the prefilter position during testing. The integrated filter bypass port at the bottom of the Quart housing was kept closed during testing. To verify that the priming phase was complete, each filter was run at 6 L/min while using the BCC200 bubble counter to confirm the absence of circulating microbubbles. To maintain a constant circuit volume, each filter type was deprimed back into the circuit after testing in preparation for the next filter to be primed and analyzed.

Performance Testing

Pressure Loss:

Once priming was completed, pump flow rate was varied while pressure loss data were taken for comparative analysis. With the priming fluid maintained at room temperature (23.6°C), pre- and postpressure monitoring ports were used to record pressure drops across each filter at incremental flow rates between 2 and 6 L/min. All pressure drop measurements were taken with the filter purge port open.

Gross Air Handling:

To compare gross air handling, the Gampt bubble counter was used to record differences in microbubble number and volume between the inlet and outlet of each filter after intermittent bolus air injections. During testing, the purge port was kept open and pump flow rate was maintained at 5 L/min while intermittent bolus injections consisting of either 5 mL or 10 mL of room air were injected proximal to the filter inlet using a syringe pump and 22-gauge needle. The syringe pump rate was set at 200 mL/h for the 5-mL tests and 400 mL/h for the 10-mL tests to complete bolus injections in 80 seconds. Data collection was started before testing to confirm a zero baseline and then continued while microbubble counts between 10 μm and 500 μm were recorded pre- and postfilter for comparative analysis. After each bolus injection and data collection, flow rate was increased to 6 L/min while the circuit and test filter were handled to completely debubble the system before the start of the next trial. The process of bolus injection, data collection, and debubbling was repeated six times for each filter and for each level of intervention to establish average test values for statistical analysis.

Statistical Analysis

PASW Statistics Version 18 (SPSS Inc., Chicago, IL) was used to perform statistical analysis on the test data. Descriptive statistics summarize the mean and standard deviation (SD) for bubble number and volume measured proximal to the filter inlet. Inferential statistics comparing the percent of bubble volume reduction (BVR) was carried out with one-way repeated-measures analysis of variance using Bonferroni post hoc tests. Dependent t tests were also performed to compare changes in filter performance with changes in embolic stress load using the 5-mL and 10-mL bolus injections as the paired sample for each filter. Results were considered significant at p < .05.

RESULTS

Table 2 details the outcomes for pressure loss performance testing, showing a range in pressure drop between 26 and 33 mmHg at 6 L/min. Prototype performance was better than expected, demonstrating a flow-through resistance less than the value predicted in phase one of the research (13). Table 3 summarizes the grand means and SDs for bubble number and volume measured prefilter.

Table 2.

Pressure loss measured in mmHg using 0.9% saline/20% HA prime at 23.6°C with the filter purge port open.

Trial	Filter Type	6LPM	5LPM	4LPM	3LPM	2LPM
1	Affinity	26	17	11	6	3
2	D734	31	22	15	9	5
3	AL8	33	24	17	12	6
4	Prototype	30	21	14	9	5
5	Quart	31	23	16	9	5

HA, human albumin.

Table 3.

Descriptive statistics showing grand mean for bubble number and volume measured at test filter inlet.

5-mL Injection Test (n = 30)	Minimum	Maximum	Mean	Standard Deviation
Bubble number	160	437	259.77	80.942
Bubble volume (μL)	12.7	27.9	17.673	2.9533
Over-range count	170	387	239.87	44.512
Bolus volume (μL)	1434	3623	2457.73	558.238
10-mL injection test (n = 30)
Bubble number	220	486	348.70	60.802
Bubble volume (μL)	17.6	34.3	27.990	3.9702
Over-range count	240	464	382.77	51.183
Bolus volume (μL)	3032	7925	4534.87	1125.148

In terms of filter performance in BVR, the prototype had a higher percent removal rate for both bolus injection tests in comparison to all other filters in the study group, whereas the Quart filter scored lowest (Figures 3 and 4). Outcomes of the post hoc analysis provided in Table 4 show that mean percent BVR for the prototype was significantly higher than the Affinity, AL8, and Quart filters (p < .05), but not the D734 filter (p = .230). Also reported, mean percent BVR for the Affinity, D734, and AL8 filters were all significantly higher than the Quart (p < .05). In this study, all test filters showed a decrease in BVR performance for the 10-mL bolus injection compared with the 5-mL test. Results for the dependent t tests provided in Table 5 show that differences in mean percent BVR between the two interventions reached statistical significance for all filters except the Affinity (p = .299). Figures 5 through 9 show pre- and postfilter mean bubble counts for the 5-mL and 10-mL bolus injection tests.

Figure 3.

Means plot for percent bubble volume reduction (BVR) (5-mL test).

Figure 4.

Means plot for percent bubble volume reduction (BVR) (10-mL test).

Table 4.

Test filter performance in percent BVR with one-way repeated-measures analysis of variance using Bonferroni post hoc tests.

(a) Filter Type	(b) Filter Type	Mean Difference (a – b)	Significance	95% Confidence Interval
Affinity	D734	−.004183	.193	−.009331	.000964
	AL8	−.002550	1.000	−.007697	.002597
	Prototype	−.008233*	.000	−.013381	−.003086
	Quart	.011208*	.000	.006061	.016356
D734	Affinity	.004183	.193	−.000964	.009331
	AL8	.001633	1.000	−.003514	.006781
	Prototype	−.004050	.230	−.009197	.001097
	Quart	.015392*	.000	.010244	.020539
AL8	Affinity	.002550	1.000	−.002597	.007697
	D734	−.001633	1.000	−.006781	.003514
	Prototype	−.005683*	.023	−.010831	−.000536
	Quart	.013758*	.000	.008611	.018906
Prototype	Affinity	.008233*	.000	.003086	.013381
	D734	.004050	.230	−.001097	.009197
	AL8	.005683*	.023	.000536	.010831
	Quart	.019442*	.000	.014294	.024589
Quart	Affinity	−.011208*	.000	−.016356	−.006061
	D734	−.015392*	.000	−.020539	−.010244
	AL8	−.013758*	.000	−.018906	−.008611
	Prototype	−.019442*	.000	−.024589	−.014294

^*Significant at the .05 level.

BVR, bubble volume reduction.

Table 5.

Dependent t test results showing changes in percent BVR with changes in embolic stress load.

Filter Type	Mean 5-mL BVR	Mean 10-mL BVR	Significance (two-tailed)
Affinity	.985717 ± .0058533	.981917 ± .0068221	.299
D734	.991950 ± .0018641	.984050 ± .0021934	.004
AL8	.990417 ± .0015664	.982317 ± .0034516	.001
Prototype	.994067 ± .0017818	.990033 ± .0025256	.002
Quart	.976150 ± .0052789	.969067 ± .0021639	.027

BVR, bubble volume reduction.

Figure 5.

Affinity filter showing pre- and postfilter mean bubble counts for 5-mL and 10-mL bolus injection tests.

Figure 6.

D734 filter showing pre- and postfilter mean bubble counts for 5-mL and 10-mL bolus injection tests.

Figure 7.

AL8 filter showing pre- and postfilter mean bubble counts for 5-mL and 10-mL bolus injection tests.

Figure 8.

Prototype filter showing pre- and postfilter mean bubble counts for 5-mL and 10-mL bolus injection tests.

Figure 9.

Quart filter showing pre- and postfilter mean bubble counts for 5-mL and 10-mL bolus injection tests.

DISCUSSION

This work combines numerous efforts aimed at developing a novel concept in arterial-line filtration for ECC. Results from this study support the underpinning theories of the new design and suggest that further improvements in microbubble removal using this technique may be possible. Advancing development of the new filter concept has been an arduous journey and underscores the difficulty to researching progress in this area. Much of the difficulty comes from the variable nature that gas in solution can assume and the way its behavior is influenced by surrounding conditions. Adding to this was the technical challenge of designing a model that could help better understand this behavior in practice (2). These difficulties were further compounded by the complexity of understanding differences in functionality between multiple filter types and the impact individual design characteristics have on performance outcomes. Although performance can be affected by factors such as the filter housing shape, volume, or its fluid path, the combined effect makes interpreting the interplay between factors and how each influences performance more difficult. In example, a change in filter housing size not only affects the internal volume and blood flow velocity through the filter, but also affects the ratio between priming volume and filter screen surface area, which is seen as a velocity gradient between the housing and micropore screen. Changes in internal volume also affect the drag forces exerted by the fluid flow and whether bubbles of a given size will escape through the purge port, make contact with the filter screen, or pass through toward the filter outlet. This example demonstrates the specific balance created between factors that are inherent to each filter design and how changing one factor can affect the other design features and overall performance of the device. To further illustrate functional benefits of the new design and how this can be used to better understand the clinical application of arterial-line filters, the following discussion explores the various design features that combine to affect performance of the arterialline filters tested.

In review of Tables 1 and 2, it might be expected that the D734 and AL8 filters, having a larger screen surface area, would have the lowest pressure drop because they offer the largest total pore diameter available for flow. Despite this, the D734 and AL8 displayed the highest pressure drop in this test group. Differences in how the filter screen is placed within the housing can help explain this outcome. The D734 and AL8 filters increase surface area by using a pleated screen design, which affects pressure drop in at least three ways. First, pleated filters commonly use a coarse mesh material as a support structure to help maintain shape of the pleats and prevent the soft micropore fabric from moving within the fluid flow. As the support mesh overlaps and comes in direct contact with the filter screen, it reduces the number of functional pores available for flow, causing an increase in resistance. Additionally, pleat dimensions affect density of the filter element and increase resistance further as larger volumes of filter material and support mesh are placed within the confined space of the housing. For the D734, the height and width of each pleat was approximately 5.5 cm × 1 cm, increasing the amount of screen fabric by 3.7 times compared with the area of a nonpleated filter element of the same size. In the AL8, size of the open area of pleating measured approximately 4.5 cm × 1 cm and increases surface area by 5.7 times. Finally, distance between each pleat determines how tightly the filter element is packed and contributes to the total amount of resistance offered. Although both filters had the same number of pleats (n = 54), the filter element of the D734 had an approximate base circumference of 20.4 cm or 2.6 pleats/cm compared with a base circumference of 15.7 cm or 3.4 pleats/cm in the AL8. It is reasonable that the smaller internal volume and more denser tightly packed AL8 filter yields a higher flow-through resistance in comparison to the larger, less densely packed D734. It makes sense that this would also contrast the pressure drop measured when the flow path is less restricted as seen in some nonpleated filter designs.

Pleated filters benefit from the denser, more compact filter element as it augments the barrier for microbubble filtration. By pleating the screen, the fluid path through the filter is lengthened and made more tortuous, improving its ability to trap microbubbles. This approach appears to have a positive effect because performance in gross air handling show the two pleated filters to be slightly better than their nonpleated commercially available counterparts. As a single design feature, however, pleating would account for only part of the device’s total effect in microbubble removal. Variations in filter housing shape and the amount of screen surface area can help further explain differences in performance between the D734 and AL8 filters. As shown in Table 1 and Figure 10, the D734 is bigger by volume, includes a raised surface above the filter element, and contains a larger screen surface area compared with the AL8. Although the larger volume aids air separation by providing a greater drop in fluid velocity, the raised surface creates a chamber above the filter element to trap larger bubbles and helps direct them toward a conveniently located purge port. More importantly, the larger screen surface area of the D734 increases the number of pores and reduces pressure drop, allowing greater accumulation of microbubbles to occur on the screen without exceeding its bubble point pressure. In view of the results from this study, the combination of design features in the D734 shows improved performance over those of the AL8, suggesting that there are limits to the filtering effects of pleating and that these limits can be exceeded or extended by the interplay of other design features.

Figure 10.

Scaled comparison of filter-housing shape and size between the Dideco D734 and PALL AL8 filters.

Pleating also has its drawbacks, because both the D734 and AL8 were more difficult to debubble and contain a much larger blood contact surface area. In the same way that pleating acts as a trap to improve bubble retention, wetting out a pleated screen requires more effort because the air it contains must be worked through a longer more tortuous fluid path. To achieve improvements in bubble retention, pleated filters use an excessive amount of filter material, and although the required support mesh adds to an already larger blood contact surface, it affects the number of functional pores making the design less efficient. This differs from the prototype, which showed better performance in microbubble removal using a smaller screen surface area. Although the amount of contact surfaces contained in ECC devices remains important, it highlights only one of the proposed benefits of the prototype filter. The following focuses on the blood flow path as a design feature and how improvements made in the prototype contribute to help increase performance.

Although each filter is unique, the fluid paths can be classified according to similarities in orientation of the inlet and outlet ports and by the flow pattern through the filter element. The Affinity, D734, and AL8 all have top inlet and bottom outlet ports, whereas the prototype and Quart both use a bottom-in bottom-out design. All filters in the test group use a circular flow pattern that surrounds and passes through the filter element except for the Quart, which divides the blood flow among six individual filter cassettes enclosed in a square-shaped housing. Because the Affinity has a similar fluid path but with a smaller screen surface area in comparison to both the D734 and AL8, it could be reasoned that the lower flow-through resistance measured in the Affinity results mainly from its larger priming volume and nonpleated design. The opposite is true when comparing the Affinity to the Quart filter. Although it had the smaller screen area with a nominal reduction in micron-pore size, the Affinity still showed a lower pressure drop over the Quart, which may be explained by differences in priming volume and the blood flow path. This contrasts a comparison with the prototype, and although the fluid paths according to orientation of the inlet and outlet ports differ between the two, the larger priming volume and screen surface area of the Affinity produced an expected decrease in flow-through resistance in favor of the commercial filter. Although this hints at the relationship that exists between fluid resistance and the blood path, it leaves unanswered questions concerning how the prototype achieves an acceptable pressure drop with such a significant reduction in screen surface area. A closer look at the fluid paths of both the prototype and Quart filters, two seemingly similar designs, will highlight important differences to address this query.

As a result of its planar construction, length of the fluid path in the Quart between the outlet and each of its 12 filter screens varies, whereas the radial design of the prototype keeps this aspect of filter dimensions constant. Variations in length affect flow distribution and may therefore alter the pattern of use between the different screen positions in the Quart. With no easy way of knowing the real affect if any that these design features have on the Maquet filter, it is worth reciting that length affects flow and flow affects resistance. The Quart filter also has several points along its fluid path where the blood flow makes nearly a 180° change in direction (16). Acute changes in flow direction affect shear stress and increase the amount of energy loss, contributing to the total pressure drop across the filter. Figure 11 is used to portray the fluid paths of both filters between the inlet and outlet. For convenience, the nonscaled tracing representing the Quart depicts the fluid path through only one side of a single filter cassette, whereas the tracing representing the prototype shows only one filter screen. As illustrated, the prototype improves on flow through the filter by straightening the blood path and eliminating any sharp angles. Additionally, unlike the Quart in which flow through each of its filter cassettes is directed down along a guide plate toward the filter outlet (16), the prototype is designed to improve lift in the fluid entering the filter to help augment microbubble separation from the blood supply returning to the patient (13). As flow across the filter screens in the Quart moves downward, a rising inner wall in the prototype is positioned between the inlet and outlet so that flow across its filter screens is directed upward (13).

Figure 11.

Nonscaled comparison of the blood flow path between the Maquet Quart and prototype filters showing approximate angle and direction of flow in relation to the rising inner wall or guide plate found in each design.

Shortcomings in this study include the bolus delivery system coupled with limitations in the Doppler ultrasound monitor used. As shown in Table 3, the large number of over-range and bolus volume events indicates that the majority of gas injected was outside the measuring capabilities of the BCC200 monitor. Refinements in the method of bolus delivery are needed to improve the reliability and validity of measurements taken for both bubble number and volume. Other concerns may involve the use of a single sample from each filter type. However, in considering the amount of regulatory control over the manufacturing process and performance of medical devices, it was felt that large imperfections in the handmade model would make any small performance variations that may exist within each commercial filter type acceptable for this proof-of-concept study. Lastly, although the rationale for using a whole blood or blood analog for studies concerned with creating near clinical conditions is clear (3,16–18), the use of a nonblood analog in this work was required to confirm phase one predictions of the research and maintain the development link between the numerical model and physical prototype (13). Although the priming solution used reduces viscous drag forces and the amount of free gas pulled across each test filter, the results remain comparable and allow the commercial filters to serve as a suitable benchmark to compare the prototype with. Testing to evaluate the blood handling capabilities of a sequential filter and its performance under conditions of normalized fluid viscosity are still needed.

Moving forward, this research raises several appropriately timed questions. Is integrating a dedicated safety component of the perfusion circuit into the oxygenator moving in the right direction? Do improvements that make better use of the gas-absorbing capabilities of hollowfiber oxygenators require that safety components in the ECC system be eliminated? From a standpoint of reducing blood contact surface area alone, the answer to these questions might be yes, but it leaves unanswered concerns related to patient safety in extreme or emergent situations. How much blood contact surface area relative to the total contact surface of the circuit is being reduced by integrating the arterial-line filter and at what cost? Results from this study suggest that surface area reductions achievable by eliminating the filter could be less than originally thought. If positioning the arterial-line filter as a dedicated component away from the pump–oxygenator remains important to overall safety, is its integration degrading or removing an important safety net for the patients we treat? With no attempt at humor, it might be analogous to sewing the seatbelts into the car seats to save on the material needed to cover them.