Scalable synthesis of asymmetric hemodialysis membranes to enhance performance and biocompatibility in flat sheet and hollow fiber configurations

Abstract

Hemodialysis is indispensable for patients with end-stage renal disease (ESRD). Yet, the performance of conventional polymeric membranes is restricted by protein adsorption, poor hemocompatibility, and thrombo-inflammatory responses. We hypothesized that functional modification of a cellulose acetate (CA) matrix with selected additives could overcome these limitations. By enhancing hydrophilicity, permeability, and anticoagulant behavior, such membranes could provide improved therapeutic potential. To evaluate this, CA-based flat sheet membranes (FSMs) were fabricated through non-solvent-induced phase separation and scaled into hollow fiber membranes (HFMs) by dry-wet jet spinning. Polyethyleneimine (PEI) and polyethylene glycol (PEG) were incorporated to adjust the pore structure, surface chemistry, and transport properties. Citric acid and gelatin were introduced as anticoagulant agents to assess their impact on blood compatibility. A comprehensive characterization was carried out, including SEM, FESEM, AFM, FTIR, tensile testing, porosity measurements, and contact angle analysis. Membrane performance was evaluated through pure water flux and dialysis simulations with urea, creatinine, lysozyme, and bovine serum albumin (BSA). Among the FSMs, CA-4 achieved a water flux of 54.40 L m⁻² h⁻¹ at 2 bar, with 78% urea clearance, 31% creatinine clearance, and 94% BSA retention. Transition to a hollow fiber geometry enhanced scalability and clinical relevance. HF-2 displayed a flux of 83.34 L m⁻² h⁻¹ at 2 bar, ∼66.5% urea clearance, and 90.3% protein retention. These values indicate a clinically significant balance between permeability and selectivity. Biocompatibility testing showed that citric acid-modified membranes reduced platelet adhesion and thrombus formation, while maintaining hemolysis ratios below the ASTM F-756-08 threshold of 5.5%. Gelatin-modified membranes lowered hemolysis up to 2.4% but promoted protein adsorption and platelet adhesion. This makes them more suited for regenerative applications than for dialysis. Overall, the results validate the hypothesis that the integration of PEI, PEG, citric acid, and gelatin into CA membranes enhances both physicochemical and biological performance. The scalable fabrication approach presented here provides a framework for next-generation hemodialysis membranes. These membranes improve solute clearance, minimize blood incompatibility, and support safer renal replacement therapy.

---

Introduction

Hemodialysis is a life-sustaining treatment that filters out accumulated metabolic toxic wastes for patients with acute and chronic kidney disease (CKD). CKD progresses to end-stage renal disease (ESRD) as the final phase of kidney failure. 10–15% of renal function has declined at this stage, and a patient is completely dependent on renal replacement therapies (hemodialysis or transplants) for survival. Globally ≈2 million people rely on this procedure, and predictive model projections estimated that in 2030 this figure could escalate up to 4 million.¹ Hemodialysis, the first of its kind since the 1960s, has opened the pathway to successful ex vivo organ substitution therapies to reduce morbidity and mortality rates. Due to its extracorporeal nature, side reactions are rare.² However, in the long run, ESRD patients experience a significantly reduced quality of life due to a high risk of cardiovascular complications, hypertension, diabetes, infections, and premature mortality, which results in a higher hospitalization rate than in other chronic diseases.³ Despite the limited availability of donor organs (restricting the possibility of kidney transplantation), the high mortality rate,⁴ and the cost,⁵ patients continue to rely on hemodialysis as the most widely practiced treatment method.⁶ Hence, there is a continuous need to upgrade the dialysis procedure to enhance treatment efficacy and cost-effectiveness.

A hemodialyzer, i.e., artificial kidney, contains a semi-permeable polymeric membrane that mimics the function of a healthy kidney by screening out uremic metabolites from blood, excluding excess bodily fluids, and balancing electrolytes in the body. Generally, membranes are fabricated using natural and synthetic polymers with optimal selectivity and permeability. Their primary function is to remove endogenous toxins with specific molecular weights while retaining functional proteins in the bloodstream. Also, they maintain a good blood-membrane compatibility to prevent any inflammatory response in the human body.⁷ Climate change and sustainability challenges have motivated researchers to use natural polymers instead of synthetic ones because of their biodegradability, eco-friendliness, non-toxicity, and unique properties suitable for advanced membrane engineering. Among natural polymers, cellulose acetate (CA) is a favourable option as it has excellent hydrophilic/hydrophobic properties, high chemical and mechanical stability, biocompatibility, broad-spectrum solubility in organic solvents, and affordability.^8–11 Its membranes show superior transport properties with minimal protein accumulation on the surface. However, the application of CA-based materials is constrained by multiple challenges. These include excessive hydrophobicity caused by extensive acetylation, which renders the membrane more susceptible to protein adsorption and could potentially lead to critical, life-threatening complications due to the activation of the complement alternative pathway in the body. The agglomeration of proteins on the membrane interface during dialysis affects its permeability and rejection efficiency. Furthermore, the hydrophobic nature of CA typically leads to poor hemocompatibility, thus necessitating the use of additional anticoagulants during the treatment to prevent clotting. Polymer blending, surface grafting, antifouling coatings, cross-linking, infusion of nanoparticles, surface functionalization and bioartificial kidneys^12,13(incorporated with renal cells) are some modifications typically made to CA membranes to address these challenges.

Polymer blending is an extensively used technique that involves blending the CA with different additives to enhance the stability, hydrophilicity, and biocompatibility of the polymer matrix.¹⁴ This technique incorporates the unique properties of each additive into the polymer matrix to improve its mechanical, thermal, and chemical properties, thereby fine-tuning its intrinsic characteristics for specific applications. Polyethylene glycol (PEG), polyether imide, polyethyleneimine (PEI), hyperbranched polyether glycol (HPG), and polyvinylpyrrolidone (PVP) are some of the additives that have been incorporated into polymeric membranes to enhance their physicochemical properties.^15–18 PEI is a polymer with an abundance of amine groups that can interact strongly with metal cations through the lone pairs on the nitrogen atoms, or with anions when protonated. Due to its strong buffering capacity to condense biomolecules like DNA or proteins, it is extensively employed as an effective carrier for genes or drugs in both in vitro and in vivo environments. Waheed et al.¹⁹ created dialysis membranes using CA, formic acid, and PEI that showed high urea clearance with albumin protein retention of up to 90%. While PEIs exhibit high transfection activity and loading capacity, they carry the risk of significant cytotoxic effects²⁰ because of non-specific interactions with negatively charged proteoglycans on cell surfaces.²¹

To mitigate the cytotoxicity of PEI, PEG is added due to its wetting ability, non-toxicity, high mobility, and antifouling properties against plasma proteins and platelets. The flexible linear chains of PEG comprise multiple ether groups that enhance the hydrophilicity of CA and allow it to act as a porogen. However, PEG can induce thrombogenicity and membrane fouling, which potentially compromises the efficacy of hemodialysis and patients’ long-term safety. However, when combined with PEI, PEG can enhance its hemocompatibility and biocompatibility. In this study, both PEG and PEI are incorporated into CA membranes to utilize PEG's hydrophilicity and pore-forming abilities along with PEI's strong ionic interactions and buffering capacity. Zhang et al.²² have shown that blending PEG and PEI in membranes enhanced mechanical stability and membrane permeability, making them more suitable for medical use. The addition of additives transformed the finger-like structure into a spongy one, which improved the hydrophilicity and antifouling properties of the membranes. Similarly, Senthilkumar et al.²³ fabricated a polysulfone hemodialysis membrane in which PEG was incorporated as a pore former and PEI as a modifier. This combination generated a defect-free thin skin and spongy sub-layer membrane that has increased hydrophilicity, water content, porosity, and permeation as compared to the neat membrane.

Similarly, hemocompatibility is a key parameter in evaluating the performance of dialysis membranes. A hemocompatible surface limits coagulation, platelet adhesion, thrombosis formation, and inflammation when interacting with human blood, thus resembling a healthy endothelium in hemodynamic conditions, taking inspiration from biotechnological concepts to help in performing most of the renal functions.^24,25 Most polymeric surfaces are not hemocompatible and require anticoagulants, high flow rates, and physio-chemical modification to become suitable for interaction with the blood. The formation of clots in the fibers of the dialyzer results in low elimination of uremic toxins, and typically 150–300 mL of blood is lost when the extracorporeal circuit is completely clotted.²⁶ Additives such as heparin, vitamin E, chitosan, graphene oxide,²⁷ citric acid, hyaluronic acid, titanium dioxide, and zwitterionic²⁸ polymers attract research interest to be extensively incorporated in polymer-based membranes to enhance their hemocompatibility and ensure patient safety. However, some of them lack performance during critical clinical trials. Citric acid (CC) is a carboxylic acid found in citrus fruits. It is a vital intermediate in the Krebs (tricarboxylic acid) cycle, a crucial stage in cellular respiration. Additionally, it is recognized for its non-toxic and biocompatible properties.²⁹ Pre-clinical studies have shown that dialyzer membranes coated with citric acid reduce the need to use systemic anticoagulants.³⁰ Calcium in the blood is complexed with citric acid to play a pivotal role in regulating clotting factors. In addition, various inorganic molecules capable of binding to calcium have been identified to diminish blood clot formation. The critical study conducted by Zarbock et al.³¹ showed that the use of regional citrate as an anticoagulant resulted in fewer cases of bleeding and gastrointestinal complications in patients. Besides citric acid, gelatin (GL, dried collagen) has been considered an effective anti-coagulant due to its exceptional biocompatibility in hydrogels, microspheres, sealants, tissue adhesives, and carriers used in drug-delivery systems. The functional groups present in these additives, e.g. –NH₂ in PEI, –OH in PEG, and –COOH in CA allow membranes to interact effectively with biomolecules. To enhance the biocompatibility of the CA/PEG/PEI membrane, both citric acid and gelatin are blended into the membranes to assess their chemical interaction with blood.

To assess the feasibility of upscaling this technology, a system was designed to transform CA/PEG/PEI FSM into HFM. The ease of controlling the experimental conditions under which FSM's were fabricated allowed their precise characterization and efficient material screening. Transformation of membranes would facilitate their practical application and maximize dialysis efficiency. The potential for improvement is in the design of hemodialysis membranes to stimulate innovation, reduce costs, and improve patient outcomes. Herein, we demonstrate how the synthesis of cellulose acetate membranes via NIPS is impacted by incorporating PEG and PEI as pore-forming agents, optimizing their physicochemical properties, permeability, hydrophilicity, and solute clearance. We investigate the role of citric acid and gelatin as anticoagulants in enhancing the biocompatibility of CA/PEI/PEG membranes. Furthermore, we compare the efficacy of both membranes for toxin removal and protein retention, evaluating their performance in biomedical applications. Our findings provide insights into membrane design for improved hemocompatibility and separation efficiency.

Materials and methods

Materials

CA/PEI/PEG polymers of different molecular weights (30, 0.2, and 25 kDa) were obtained from Sigma-Aldrich, Germany. Formic acid (FA) with purity ≥ 98% was sourced from Scharlab, Spain. Urea (60.06 Da, ≥ 98%), lysozymes (14.3 kDa > 97%, 92.09 molecular weight), and bovine serum albumin (BSA, 66 kDa) for dialysis simulation testing were procured from Sigma Aldrich. Sodium dodecyl sulphate, a BCA protein assay kit, fibrinogen (human plasma), and glutaraldehyde (25%) for biocompatibility testing were also sourced through Sigma Aldrich. Furthermore, gelatin was procured from Daejung Korea, whereas citric acid and creatinine (113.54 Da, ≥ 98%) were obtained from Merck, Germany. Deionized water was produced in the laboratory.

Synthesis of the flat sheet membrane

CA polymer was dried for 8 hours in a vacuum oven at 120 °C to remove moisture prior to fabrication. The casting solution of CA/PEI/PEG membrane was prepared by blending varying percentages of PEI/PEG with 15.5% of CA and formic acid (solvent), as shown in Table 1. For PEI, it must first be dissolved in 1 mL of hot water before being added to the solution. To prepare a homogenous mixture, blended polymers were constantly stirred for 24 hours at room temperature and sonicated for 2 hours to remove any trapped air bubbles. Subsequently, the casting solution was poured slowly onto a dust-free glass plate and fabricated using a doctor blade with a thickness of 200 μm and evaporated for 30 seconds. The cast membrane was then immersed in a deionized water-filled coagulation bath to initiate phase inversion. After thorough washing, the membranes were soaked in deionized water for 24 hours to remove any residual solvent content before further testing, as shown in Fig. 1. The CA/PEG/PEI membranes were prepared with citric acid and gelatin, following the same procedure described earlier.

Table 1 Compositions of CA casting solution with varying concentrations of PEI and PEG employed in the synthesis of hemodialysis membranes

Sample	CA (%)	PEI (%)	PEG (%)	Water (%)	FA (%)
CA-1	15.5	1	0	1	83.5
CA-2	15.5	1.5	0	1	83
CA-3	15.5	1	1.5	1	82
CA-4	15.5	1	2	1	81.5
CA-5	15.5	1	2.5	1	81

---

Fig. 1 Schematic representation of the flat sheet membrane (NIPS method) and hollow fiber membrane (dry-wet spinning) fabrication process.

Synthesis of the hollow fiber membrane

CA powder was dried in the oven to remove moisture for 24 hours before developing the dope solution of HFMs. The dope solution consisted of 18% CA, 78% FA, and additives (PEI, PEG, and water, each at 1%), which were stirred for 24 hours to make a homogenous solution. The dry-wet spinning technique was applied to fabricate HFMs. The spinneret specifications include a 0.3 mm internal diameter and a 0.6 mm outer diameter. Distilled water was employed as the bore fluid, while tap water was used in the coagulation tank at ambient conditions, as shown in Fig. 1. In dry inversion, HFMs were evaporated in air, while in wet inversion (solution solidification), they were soaked in a water bath. In the spinning process: the air gap was variable, while the dope extrusion rate (DR), membrane collection speed (CS), and bore fluid flow rate (B2FR) remained constant, as shown in Table 2. Following spinning, the hollow fiber membranes underwent a treatment that involved soaking in tap water for three days to remove any solvent and then aeration for drying.

Table 2 Summary of spinning parameters for the synthesis of hollow fiber membranes

Sample	Air gap (cm)	DR (cm^3 min^−1)	B2FR (cm^3 min^−1)	CS (Hz)
HF-1	6.5	1	0.5	3
HF-2	10	1	0.5	3
HF-3	13.8	1	0.5	3
HF-4	15.5	1	0.5	3

---

Membrane characterization

The characteristic absorption bands of the blended polymer were analyzed through FT-IR (PerkinElmer Spectrum 100 spectrometer), operating within the spectral range of 4000–400 cm⁻¹ at the resolution of 4 cm⁻¹. Flat sheet membrane samples were cut and positioned in the pallet holder, while hollow fiber samples were transformed into KBr pellets to examine the presence of diverse functional groups.³² Detailed surface and cross-sectional images of FSMs were examined using the SEM technique (JOEL JSM-6490LA), while HFMs were analyzed with the FESEM technique (Hitachi SU8020). Before analysis, the samples were dried to eliminate moisture content. Gold sputter coating was applied for the SEM samples, while platinum coating was utilized for the FESEM samples. Membrane surface morphology and cross-section images were captured at 5 kV at various spanning magnifications for comprehensive analysis. Membrane topology was analyzed by AFM analysis (NanoWizard 3 system) using a silicon probe (Tap300-G) on a 5.0 μm × 5.0 μm area of the membrane's surface.³³ All membrane images were quantitatively analyzed and processed using JPK Image Processing software. Next, membrane diffusive transport properties were assessed by porosity. This was measured by calculating the volume of voids and then describing the total volume in 0–100% or between the ranges of 0–1 to quantify its permeability.³⁴ Similarly, gravimetric porosity was determined using the wet and dry weight method. Membranes were cut into 1 × 1 cm² pieces, oven-dried, immersed in water for 24 hours, and reweighed. It was calculated using eqn (1).W = weight of the membrane in the dry and wet state respectively, (g), ρ_w = density of water, g cm⁻³, ρ_p = density of polymer, g cm⁻³.

HFM pore size and its distribution within the membranes were assessed by applying high pressure to non-wetted liquid mercury within the pores, known as mercury intrusion porosimetry (Micromeritics Auto Pore IV 9500). The membranes were cut into around 1 cm pieces and weighed as approximately 0.25 g. This estimated the total pore volume rather than individual pore diameters and emphasized the connections between adjacent pores.³⁵ The degree of swelling in the membrane samples was calculated by cutting them into dimensions of 1 cm × 1 cm. The membranes were pre-dried at 60 °C for 12 hours and weighed (W_dry). Then the samples were immersed in deionized water for 24 hours, removed, and dried with filter paper before being weighed again to determine wet weight (W_wet) as shown in eqn (2)

Stress–strain behavior of the membrane was investigated by employing a tensile strength test (SHIMADZU AGS-X) at a capacity of fifty kN. ASTM-standard D 8802-02 was followed, and the strain rate was adjusted at 0.5 mm min⁻¹.³⁶ The static contact angle on the surface of the membrane was measured using the sessile drop method by employing the contact angle system OCA (Data Physics, USA). The ultra-pure water dosing rate was adjusted at 0.1 μL s⁻¹, with a constant dosing rate of 0.2 μL. The angle was recorded on the outer surface of the membrane after 3 seconds, and three readings were taken for each sample.³⁷

Membrane performance

The pure water flux was calculated by a dead-end filtration cell module (Hp 4750-Sterlitech) with an active membrane area of 7.085 cm.³⁸ Initially, run for 10 minutes to fill the module volume and stabilize flow and pressure. The pressure was held at 2 bars using nitrogen gas and permeates were collected at intervals of 30 minutes. Multiple iterations of the experiment were conducted for reliability. Eqn (3) calculated the flux whereas eqn (4) calculated permeance.J = flux, L m⁻² h⁻¹, T = time, h, A = total area of membrane, m², V = volume of the water permeated, L.The hemodialysis performance of the fabricated membranes was evaluated using a custom-built dialysis setup designed to simulate clinical conditions, as shown in Fig. 2. The system consisted of two chambers separated by the test membrane, with one side containing a blood-mimicking solution and the other side filled with deionized water as dialysate. The blood-mimicking solution comprised urea (150 mg dL⁻¹), creatinine (10 mg dL⁻¹), lysozyme (50 mg dL⁻¹, 14.3 kDa), and bovine serum albumin (BSA, 100 mg dL⁻¹, 67 kDa)³⁹ to represent representative uremic toxins and essential plasma proteins. The effective membrane area was fixed for FSMs, and the lumen of the HFMs served as the blood-contact surface during counter-current flow. The experiment lasted 4 hours under ambient lab conditions, with readings taken hourly. BSA and lysozyme were detected using a UV-spectrophotometer at 278 and 280 nm, respectively.⁴⁰ Creatinine was detected using Jaffe's method,⁴¹ and urea by the p-dimethyl amino benzaldehyde method.⁴² Protein retention (BSA) was calculated as the percentage of initial protein concentration maintained in the feed solution, while toxin clearance was expressed as the reduction in solute concentration in the blood-mimicking chamber relative to initial values.A_t = final concentration, A_i = initial concentration.

Fig. 2 Schematic representation of the laboratory dialysis experiments, (left) FSM setup with blood-mimicking fluid in contact with the dialysate, and (right) HFM configuration showing the setup for the dialysis experiment.

The urea, creatinine, and lysozyme clearances were calculated by eqn (6)

C_t = final concentration, C_i = initial concentration

Biocompatible testing

The biocompatibility performance of the CA/PEG/PEI membranes modified with citric acid and gelatin, as shown in Table 3 is tested against blood to investigate their behavior and reaction.

Table 3 Different ratios of citric acid and gelatin in the CA casting solution for biocompatibility testing

Sample	CA (%)	PEI (%)	FA (%)	Gelatin (%)	Citric acid (%)	Water (%)	PEG (%)
CA-4	15.5	1	80.5	0	0	1	2
CA-G1	15.5	1	79.5	1	0	1	2
CA-G2	15.5	1	79	1.5	0	1	2
CA-G3	15.5	1	78.5	2	0	1	2
CA-CC1	15.5	1	79.5	0	1	1	2
CA-CC2	15.5	1	79.5	0	1.5	1	2
CA-CC3	15.5	1	78.5	0	2	1	2

---

Protein adsorption on membrane samples (1 × 1 cm²) was studied by immersing them in a phosphate buffer solution (pH = 7.4) for 2 hours to neutralize the membranes. Albumin and fibrinogen (0.45 and 0.03 g dL⁻¹) protein solutions in phosphate-buffered saline (PBS) were prepared. The samples were then incubated in solution for two hours at 37 °C and shaken at 1000 rpm for 2 hours. The samples were gently rinsed with PBS and dissolved in 1 wt% aqueous solution of sodium dodecyl sulfate (SDS) for 1 hour. The quantification of protein was done by using a bicinchoninic acid assay kit in SDS solution.⁴³ For thrombus formation, membrane samples (1 × 1 cm²) were soaked in 1.5 mL of whole human blood and subjected to a 2-hour incubation period in a 5% CO₂ environment at 37 °C. Post-incubation, the samples underwent rigorous washing with PBS to remove non-adherent components. Assessment of in vitro thrombus formation on the membrane surfaces was conducted using a graded ethanol treatment combined with critical point drying.⁴⁴ The degree of thrombus formation was calculated as:

where W_t = weight of blood coagulated membrane (g), and W_d = weight of dry membrane (g).To evaluate the platelet adhesion, 10 mL of anti-coagulated whole blood was obtained from a volunteer and placed in a centrifuge bottle. Centrifugation at 1000 rpm for ten minutes facilitated the separation of the supernatant, yielding plasma-rich plasma (PRP). Membrane samples (1 cm × 1 cm) were pre-washed with PBS (pH = 7.4) and placed in 24-well plates. Each membrane sample was treated with 100 μL of PRP using a pipette and subsequently incubated at 37 °C for two hours. Following incubation, thorough washing was repeated three times to eliminate unstable platelets. The adsorbed proteins were immobilized onto the membrane surface by overnight immersion in a 2.5 wt% glutaraldehyde solution. Subsequent drying steps involved sequential immersion in ethanol–water solutions with concentrations of 50%, 75%, 85%, 95%, and 100% (v/v) for 10 minutes each. The platelet adhesion on the CA membranes was observed using scanning electron microscopy (SEM) after freeze-drying with liquid nitrogen.³⁹ For hemolysis ratio experimentation, membrane samples (1 cm × 1 cm) were washed thrice with deionized water and 0.9 wt% aqueous solution of NaCl for ten minutes in sequence. Then, the samples were soaked in the NaCl aqueous solution for 30 minutes at a temperature of 37 °C in a water bath. Whole blood of 200 μL was added to the NaCl solution, in which the samples were kept for one hour at 37 °C. The blood was centrifuged at 1500 rpm for 10 minutes, and the top layer absorbance was measured at 545 nm by an ultraviolet spectrophotometer (Analytic Jane). Pure water was taken as a positive reference while 0.9 wt% aqueous solution was taken as a negative reference.⁴⁰ The ratio was calculated using eqn (8)where HS = absorption value of membrane samples, HN= absorption value of the negative reference, and HP = absorption value of the positive reference.

Results and discussion

FSM and HFM characterization

Surface morphology. The impact of PEI and PEG additives on the CA membrane matrix can be observed in the SEM micrographs shown in Fig. 3. The presence of PEI alone in the CA-1 and CA-2 membranes creates a porous skin layer that indicates instantaneous demixing and enhanced phase inversion, as corroborated by Waheed et al.⁸ On the other hand, the addition of PEG increases the solution viscosities of the CA-3, CA-4, and CA-5 membranes due to their high molecular weight (rich in the –OH group), resulting in a dense skin layer and the rapid diffusion speed of the casting solution (formic acid).^45,46 This generates smaller pore diameters, closer pore cells, and increased pore density.⁴⁷ The surface nodules and aggregations observed in CA-3, the transition membrane, are likely caused by material separation at the top surface. The slower diffusion of PEG in CA-4 and CA-5 happens because higher concentration leads to the formation of a denser surface. The diffusivities of the additives and solvents during phase inversion control pore formation in the membrane. According to the theory of pore formation, a slow and controlled outflow of an additive leads to the formation of interconnected spongy channels, as seen in the images, rather than macro void-type tunnels.⁴⁸ It is also possible that PEG inclusion slows down the nucleation process due to its effect on solution viscosity and solubility characteristics.

Fig. 3 SEM images of the surface and cross-section of CA/PEI/PEG flat sheet membranes.

Fig. 4 shows the cross-sectional and surface morphologies of the fabricated HFM's. The appearance of a spongy support layer and a dense surface layer with no or few pores facilitates the elective transport and barrier properties. These images further validated the asymmetric structure observed in the scanning electron micrographs. The cross-section SEM of HF-1 shows a void near the lumen, and the cross-sections of HF-2/3/4 reveal a dense cross-section. The inner spongy structure disintegrates in HF-4 as the air gap increases. The HFMs undergo two distinct solidification processes during dry-wet jet spinning. The first process occurs upon contact with the internal solidifier, while the second takes place later when exposed to the external solidifier in the air. In this zone, humidity-induced phase separation starts on the outer surface, and condensed moisture spreads evenly by diffusion. The internal layer of the HFM's experiences a rapid solvent exchange, while the outer layer experiences a slow exchange, resulting in varying pore sizes. Stress inconsistencies caused by surface tension, solvent or coagulant clumping, rapid coagulant flow, and volume changes can result in defects. These issues lead to the formation of small voids due to spinodal decomposition.⁴⁹ Previous studies reported that slower solidification forms sponge-like structures⁵⁰ as shown in these membranes. The solvent in the coagulation bath (water) counters the effect of the solvent in the casting solution (formic acid) and creates the conditions for delayed demixing, thereby suppressing the coagulation process. This suppresses the formation of macro-voids and instead leads to the development of honeycomb-like voids. Rather than forming an equilibrium pattern, this interaction process creates an unstable flow. In this flow, a less viscous fluid displaces a highly viscous one. This phenomenon is a fundamental part of the phase inversion process. According to Kesting,⁵¹ it is also possible that when there's a low concentration of PEG in the polymer solution, it rapidly forms separate regions (nuclei) before non-solvents can diffuse into the solution and displace the solvent. These nuclei contain high concentrations of solvents. When the membrane solidifies, it forms a pore structure that resembles a honeycomb. The uneven solvent distribution due to slow non-solvent diffusion creates this distinctive pore architecture.

Fig. 4 FESEM images of HFMs showing its (1) lumen, (2) surface, and (3) cross-sections with different air gap ratios: HF-1, HF-2, HF-3, and HF-4.

Functional group analysis

The FTIR spectra of the CA/PEI/PEG blended membranes are displayed in Fig. 5. The strong absorption peaks of –OH (3766–3336 cm⁻¹), –CH (2960 and 2860 cm⁻¹), and C=O (1750 cm⁻¹) in the spectrum in Fig. 5(a) confirm the presence of these functional groups in the CA/PEI blended membrane, indicating that no chemical reaction occurred during the blending process. The absorption peak in the 1240 and 1040 cm⁻¹ range is associated with the stretching vibration of the C–O bond in alcohols, ethers, or esters. The presence of PEI in the CA matrix is verified by the absorption in the range of 1621 to 1460 cm⁻¹, which can be attributed to the stretching vibration of the =N–H groups in the PEI molecules. Additionally, minor peak shifts between 2920 and 2850 cm⁻¹ in the C–H stretching region are attributed to methylene groups in CA and PEI.⁵² The PEI concentration might have affected the C=O group in CA with a sharp peak at 1750 cm⁻¹, while the =N–H peak at 1621 cm⁻¹ broadens, indicating a higher concentration of primary amine groups.⁵³Fig. 5(b) shows the effects of increasing the PEG concentration on the spectra to enhance membrane porosity and performance while keeping the PEI concentration constant. The spectra reveal a broad peak in the 3376–3377 cm⁻¹ range due to the stretching vibration of the –O–H bonds in the alcohol groups of PEG. The characteristic absorption peaks due to –CH₂ at 2869 cm⁻¹, C–H at 1374 cm⁻¹, C–O at 1044.90 and 1233 cm⁻¹, and C=O at 1641 cm⁻¹ confirm the presence of PEG. Moreover, the intensity of the C–H peak at 1433 cm⁻¹ in CA-1 decreases with PEG addition, denoting increased hydrogen bonding from the alcohol groups. The C–O–C (ether group) stretching vibration at 1158 cm⁻¹ indicates the presence of PEG, and the adsorption peak at 619.33 cm⁻¹ arises due to the torsional vibration of the O–H bond. These peaks confirm the coexistence of hydrophobic and hydrophilic moieties within PEG as well as the polymer blend of CA, PEI, and PEG.⁵⁴

Fig. 5 FTIR spectra of flat sheet membranes prepared with (a) CA/PEI and (b) CA/PEG/PEI blends.

Similarly, the FTIR spectra of HFM's are illustrated in Fig. 6. The absorption observed in the range 3500–3200 cm⁻¹ corresponds to OH-stretching, while stretching vibrations of aliphatic chains (–CH) are responsible for the peak in the range 3000–2800 cm⁻¹. The characteristic absorption at 1750–1659 cm⁻¹ represents the stretching vibration of the carbonyl group (–C=O). In addition, the presence of an amine group (N–H₂) is indicated by absorption within the 1600–1500 cm⁻¹ range. The decrease in transmittance intensity from HF-1 to HF-4 reflects the reduced copolymers on the membrane surface. More intense and sharper peaks are detected for smaller air gaps, while larger ones produce broader peaks. The spectra confirm that the chemical composition of the HFMs remains the same, but thickness decrease with larger air gap distances.

Fig. 6 FTIR spectra of hollow fiber membranes with different air gaps (HF-1, HF-2, HF-3, HF-4).

Surface roughness analysis

The surface roughness (R_a) and AFM images of the CA/PEG/PEI membranes are shown in Fig. 7. CA-5 has the highest surface roughness of R_a = 48.1 ± 0.42 nm. The transformation in the texture of the skin layers of the other membranes from porous to dense explains why their surface roughness is lower than that of CA-1. The CA-3 membrane exhibits the lowest surface roughness of R_a = 27.9 ± 0.22 nm. This reduced roughness is advantageous for biocompatibility, as it reduces protein adsorption, inflammation, and clotting.⁹ For CA-5 membranes, the high concentration of PEG can cause the membrane to become denser. At the same time, the high viscosity of the doping solution also increases the number of surface irregularities (nodules and ridges).

Fig. 7 AFM topographic maps of the flat sheet cellulose acetate membranes prepared by varying the PEI and PEG concentrations in the dope solution.

Likewise, the topographic images of the active layer of each HFM visualize the surface roughness to determine their biocompatibility as shown in Fig. 8. As can be observed, an increase in the air gap from 6.5 cm to 15.5 cm during the spinning process results in an increase in surface roughness from 21.55 ± 0.36 nm to 31.86 ± 0.22 nm. HF-4 spun at an air gap of 15.5 cm has the maximum surface roughness because its CA matrix disintegrated, resulting in diminished stability.

Fig. 8 AFM height maps of the hollow fiber membranes with different air gaps (HF-1, HF-2, HF-3, HF-4).

This potential deterioration of the pore structure is why CA fibers are not usually spun at exceedingly high air gaps due to limitations in mechanical, structural, and phase inversion processes.³⁷ As observed in the FESEM image in Fig. 4, the increase in air gap from 6.5 cm to 15.5 cm causes the honeycomb-like structure to stretch out and the parallel nodular structure to become more pronounced. From the literature, it is evident that CA membranes have lower surface roughness and a refined top layer due to the formation of fewer pores.⁵⁵ In contrast, adding PEG and PEI to the CA matrix causes surface roughness to increase due to the complex interplay of non-uniform pore size, surface layer instabilities, and instantaneous demixing.

Mechanical properties

A good dialysis membrane must be mechanically strong. The key mechanical properties (tensile strength and percentage elongation) were measured and are presented in Fig. 9. The CA-1 membrane exhibits the highest stress-bearing capacity, while CA-2 and CA-5 have the lowest stress-bearing capacity. This can be explained in terms of the inverse relation between mechanical strength and porosity. When the porosity increases the links between the membrane layers weaken due to which the membrane's capacity to endure applied force decreases. On the other hand, a different trend is observed for percentage elongation. CA is a brittle polymer whereas PEG and PEI are flexible. The incorporation of both additives increases the elongation limit due to the plasticizing nature of PEG and the elasticity of PEI.³⁹ However, an increase in PEG concentration lowers the solution stability and therefore leads to lower overall mechanical strength.⁵⁶

Fig. 9 (a) Stress–strain curves and (b) strain-rate response of flat sheet cellulose acetate membranes prepared with varying PEI/PEG content.

FSM and HFM transport properties

The functional characterization of the membrane is essential for analyzing its selectivity, permeance, and biocompatibility.

FSM hydrophilicity

The water content of the membrane demonstrates its water-uptake capacity and degree of swelling, thereby indicating how fluid is transported through the spongy structure of the CA/PEI/PEG membranes. As can be seen in Fig. 10, CA-2 has the highest water content of up to 55%, indicating macro-void formation on the skin layer, which was also revealed by the SEM in Fig. 3. Using only PEI as the additive to CA-1 and CA-2 resulted in the conversion of their porous surface into macro voids and thus increased their water content. This can be attributed to the low interfacial adhesion between the CA and PEI molecular chains due to the decrease in the miscibility of the doping solution caused by the high concentration of PEI. The adhesion might have induced supramolecular aggregation in the dope solution, thereby increasing the membrane porosity and density. On the other hand, adding PEG to the CA-3 membrane improves its hydrophilicity (due to the abundance of –OH groups) but results in lower water content. The molecular interaction of both additives generates a thicker layer due to a difference in their diffusivity of leaching out in non-solvent (water), which suddenly causes a drop in water content. In the CA-4 and CA-5 membranes, the water content steadily increased with PEG concentration, signifying increased sponginess. The increase in the concentration of PEG in the casting solution leached out during the gelation process results in forming pores, which eventually become the domain for water molecules to reside in. The difference in the molecular weight and densities of PEG and PEI allows for diverse pore sizes; instantaneous demixing with PEG leads to the formation of a honeycomb-like homogenous pore structure.⁵⁷ Another reason could be that the hydrophilic nature of PEG has an affinity for attracting water molecules inside the matrix of the membrane. Even after phase inversion, some of the PEG remains within the membrane matrix, resulting in its hydrophilic nature facilitating the attraction and retention of water molecules. This steadily increased the water content.⁵⁸

Fig. 10 Hydrophilicity and water absorption capacity of the CA-modified membranes.

The contact angle is a measure of the hydrophilicity/hydrophobicity of hemodialysis membranes and the permeability of fluids and solutes through their porous structure. Materials contacting human blood ought to have a balance between hydrophobic and hydrophilic properties. All the membranes fabricated in this study have shown contact angles less than 90°, as illustrated in Fig. 10. This indicates that CA/PEG/PEI membranes are hydrophilic. The CA-4 membrane exhibits the lowest contact angle (58°), and CA-1 has the highest contact angle (90°). In the CA-1 and CA-2 membranes, PEI enhances hydrophilicity due to the non-covalent interaction between ether, imide, and acetate groups within the CA/PEI chain, which competes with water molecules for hydrogen bonds on the membrane surface. This implies that these groups are more likely to bond with each other than with water, which reduces the surface energy and results in a lower contact angle.⁵⁹ The CA-2, CA-3, and CA-4 membranes are rendered more hydrophilic than CA-1 due to a higher proportion of polar additives (PEG and PEI). The contact angle of all the membranes decreases slightly with drop age, even as the concentration of PEG increases. In the case of CA-5, the presence of hydroxyl, carboxyl, and amine groups increases the contact angle, reducing its optimality for dialysis operations.⁴⁰ In general, the measurement of the contact angle of a polymeric surface at thermodynamic equilibrium indicates its wettability, which is dictated by the chemical composition and porosity of the membrane. The contact angle is significantly affected by porosity since water penetrates the membrane by capillary action through the pores. Ultra micropores, with a diameter less than 2 nm, are suitable for dialysis to filter out uremic toxins while retaining bovine serum albumin (BSA).

FSM porosity

In this study, the membrane porosity was improved from ∼10% to ∼60% as displayed in Fig. 11. CA-5 showed the lowest percentage porosity. The high porosity of CA-2 in comparison to CA-1 confirms the formation of macro-voids upon increasing the concentration of PEI alone. On the other hand, adding PEG creates an asymmetric structure with a thick surface and a porous support layer. However, the increase in PEG concentration causes a decline in membrane porosity because the matrix of CA becomes thicker due to more intense polymer chain entanglement and increased solution viscosity. This results in a reduced membrane porosity because the denser matrix hinders the formation of effective pores. CA-4 has porosities of approximately 50%, which fall within the optimal range of 30% to 60%, making them suitable for dialysis applications.⁶⁰ Maintaining this porosity range is important for achieving the balance between permeability and selectivity required for waste removal and fluid control. Low interfacial free energy between PEI and PEG and blood components is assumed to limit platelet adsorption and adhesion on the dialysis membrane.³⁹

Fig. 11 Structural characteristics of the CA-modified membranes.

FSM pure water flux

The transport properties of membranes determine how efficiently uremic toxins can be removed from the blood while retaining proteins during hemodialysis. A pure-water flux experiment was conducted to evaluate the effect of additives on CA membranes, with the results illustrated in Fig. 12. CA-1 has the highest flux of 118.8 L m⁻² h⁻¹, and CA-2 has 84.7 L m⁻² h⁻¹, which confirms the presence of macropores and significant internal spaces. On the other hand, CA-3, CA-4, and CA-5 give similar values of 65.56, 54.40, and 49.02 L m⁻² h⁻¹, respectively. High flux membranes have a large pore size that allows greater removal of uremic toxins and middle-sized molecules (β2-macroglobulin) during dialysis. Water was used as a preliminary solvent to examine the behaviour of CA/PEI/PEG membranes. The presence of additives in the CA matrix increases pure water flux by facilitating pore formation. However, the increase in flux is not linear, which may be attributed to changes in the solvent concentration. The PEG in the casting solution acts as a pore-forming agent, facilitating nucleus formation after the cast membrane is immersed in the coagulation bath. Increasing the quantity of additives generally enhances flux. However, excessive amounts of PEG and PEI can detach from the polymeric solution during coagulation. PEI forms macro-voids in CA-2, leading to low water flux because voids result in a non-uniform pore size distribution and weaken the structural integrity of the membrane. Similarly, the addition of PEG instigates thermodynamic instability in the membrane, leading to instantaneous demixing.³⁹ When the PEG concentration exceeds 1%, the casting solution becomes more viscous. This increase in viscosity is suggested to have delayed the rate of diffusion between the solvent and the non-solvent in the coagulation bath. It causes a dense layer to form on the top and suppresses pore formation.⁶¹ According to Darcy's law, the flux of a porous membrane depends on the viscosity and the surface tension of the dope solution. The hydroxyl, ether, and imide polar groups in the CA/PEG/PEI membranes efficiently interact with water molecules, disrupting water clusters and facilitating their transport through the membrane. In contrast, pure CA membranes lack this ability resulting in lower flux.⁶² These results represent improved porosity and flux, but care must be taken to avoid excessive concentrations of the additives since that could adversely impact membrane structure and function.

Fig. 12 Pure water flux of different composition of flat sheet membranes.

HFM pure water flux

HFM's pure water flux results are shown in Fig. 13 where it is increased linearly over time. Of the four HFMs, HF-1 has the lowest of 77.7 L m⁻² h⁻¹, whereas HF-4 has the highest of 91.7 L m⁻² h⁻¹. This indicates that a thinner skin layer is formed with an increased air gap. Moreover, the porosity also improved since it is directly linked to the flow of fluid. Mulder⁵⁰ reported that sponge-like structures have smaller and tortuous pores that cause more resistance to fluid flow compared to finger-like structures but show higher selectivity for smaller-size molecules and particle retention in filtration systems.

Fig. 13 Pure water flux for hollow fiber membranes spun at different air gaps.

FSM simulated dialysis analysis

Hemodialysis membranes should function as a barrier to the transfer of functional proteins into the dialysate while permitting the passage of small-to-middle-sized molecules and uremic toxins. To assess the selectivity of the CA/PEI/PEG membrane, an experiment simulating hemodialysis was conducted. A mixture of BSA, urea, and creatinine solution was used to mimic the patient's blood, while deionized water served as the dialysate. The solute rejection percentages were recorded and shown in Table 4. CA-1 and CA-2 retained less than 86% of the proteins but rejected a high percentage of creatinine likely due to the presence of macro-voids. CA-3 and CA-4 gave the best results, with CA-4 having a maximum urea clearance of 78% due to its spongy infrastructure. In CA-5, the maximum amount of protein was retained, but only a meagre amount of toxins was removed.

Table 4 Permeability and solute clearance percentages of flat sheet membranes

Membrane	Thickness (mm)	Permeance L m^−2 h^−1 bar^−1	Blood mimic fluid			SD
			BSA (%)	Urea (%)	Creatinine (%)
CA-1	0.03	59.39	86	69	33	±0.00035
CA-2	0.13	42.34	84	57	22	±0.00028
CA-3	0.04	32.78	95	70	20	±0.00262
CA-4	0.04	27.2	94	78	31	±0.00403
CA-5	0.09	24.51	99	10	12	±0.00113

---

The normal range of molecular weights for protein retention during dialysis is 64–66 kDa. BSA (67 kDa) with a hydraulic diameter of 3.6 nm was used as a marker to determine solute rejection percentages. A higher percentage of BSA retention indicates a greater ability to block protein transfer from the blood to the dialysate, such that more proteins are retained. Despite solute rejection and BSA retention rates exceeding 75%, some membranes showed poor water flux, underscoring the common trade-off between membrane performance and morphology.³⁷ As the PEG content in the membrane increased, the porous structure of the membrane caused a decrease in BSA retention, indicating that the maximum pore size exceeded 66 kDa. BSA rejection depends on its configuration and interactions with various polymers. As an elliptical molecule,⁶³ BSA experiences different steric hindrances when entering membrane pores from different directions. It may pass through pores in the long axial direction but block them in the short axial direction. Similarly, BSA molecules form aggregates and clusters in the solution, another cause for steric hindrance. The smaller pore size on the dense top layer than the bulk average pore size ensures that the membrane retains the BSA molecules during dialysis.⁶⁴ A lower ratio of PEG in the membrane increased urea clearance percentages, aligning with previous studies on ultrafiltration membranes.^65–67 Reducing hydrophilic additives promotes urea permeability and solute flux through the membrane. However, PEI alone in the CA-1 and CA-2 membranes did not show good clearance percentages. PEG plays a crucial role in altering protein permeability in CA membranes. In this study, PEG increased urea clearance while reducing protein adsorption whereas the amino group in PEI contributed to BSA retention.⁵¹ The CA-4 membrane due to its good solute rejection percentages serves as a foundation for designing HFMS for the next step to assess the practical applicability and efficiency of dialysis membranes. In controlled laboratory settings, FSMs help design membranes with precise characterization and permeability. However, their structure is not fully compatible with commercial operations. In contrast, HFMs have a higher surface area-to-volume ratio. This configuration is more suitable for clinical use as it improves mass transfer and operational efficiency.

HFM simulated dialysis experiment

During hemodialysis, impure blood flows through the lumen of the hollow fibers, whereas the dialysate flows in the opposite direction outside the fibers. The inner surface of the membrane is in direct contact with the plasma and facilitates the transfer of uremic toxins into the dialysate. The dense surface layer retains the essential proteins, such as albumin, while allowing the smaller and middle-size molecules to pass through. To study this process, a blood-mimicking fluid was passed through the HFM's to evaluate their performance, and solute retention data is displayed in Table 5.

Table 5 Solute clearance percentages of hollow fiber membranes for the hemodialysis simulation experiment

Membrane	Thickness (mm)	Permeance (L m^−2 h^−1 bar^−1)	BSA (%)	Lysosome (%)	Urea (%)	Creatinine (%)
HF-1	0.01	38.75	91.7 ± 0.072	13.2 ± 0.087	56.8 ± 0.056	38.5 ± 0.043
HF-2	0.021	41.67	90.3 ± 0.086	14.8 ± 0.076	66.5 ± 0.067	39.1 ± 0.052
HF-3	0.011	43.05	87.6 ± 0.0022	11.9 ± 0.044	71.7 ± 0.0223	27.2 ± 0.077
HF-4	0.031	45.87	83.5 ± 0.034	10.6 ± 0.054	52.9 ± 0.0324	29.5 ± 0.0578

---

Although the HFMs showed high flux rates, their protein retention percentage declined. This trade-off is common in membrane fabrication and is linked to its morphology. HF-1 has the highest protein rejection percentage but the lowest urea removal percentage and flux rate. The membrane fibers spun at a 3 cm air gap showed the highest percentage of protein rejection (91%) but the lowest water flux (77.75 L m⁻² h⁻¹). These results illustrate that the flux through HFMs depends on the porosity, while the percentage of protein rejection is influenced by the pore size, as reported by Mansur et al.³⁷ The flux rate of HF-2 and HF-3, which had higher pore sizes, was nearly identical, but the protein rejection percentages started to decline in HF-3 as the pores began to disintegrate, as shown in Fig. 11. The spongy structure of the membrane helps in the removal process and prevents backflow of the blood. For HF-4, the protein retention was below par. The flux and rejection of proteins by the ultrafiltration membrane can be explained by the concept of protein adsorption and the consequent pore constriction due to the electrostatic and hydrophobic interactions between the membrane surface and the protein molecules, because both the top and sub-layer play crucial roles in protein retention.⁶⁸

HFM MIP analysis

After the performance tests, HF-2 and HF-3 were selected for mercury porosimetry to precisely measure the sizes and distribution of pores in the membranes. It is important to understand that MIP is usually used to assess the support layers of membranes containing large pores, but not an ideal method to analyze the skin layer of hemodialysis membranes, with a typical pore range of 5 nm. The skin layer of the membranes effectively removes uremic toxins and retains essential proteins in the blood, while the support layer primarily contributes to the mechanical structure and flow dynamics. Therefore, the MIP data presented here focuses on the support layer's porosity and pore distribution and does not directly correlate with the membrane's selective filtration properties, which are determined by the skin layer. The plot of cumulative intrusion against pore diameter in Fig. 14 shows that for small pore sizes in HF-2, the cumulative intrusion started at 0.6 mL g⁻¹ and gradually decreased to near zero at a pore diameter of approximately 200 μm. This gradual decline indicates a wide distribution of pore sizes, suggesting a heterogeneous pore structure with a significant number of medium-sized pores. On the other hand, for HF-3, the cumulative intrusion started at 0.9 mL g⁻¹ for small pore sizes. It showed a steep initial decrease followed by a more gradual decline. The intrusion reached near zero at a pore diameter of approximately 350 μm. The sharp decline indicates a more uniform pore structure with a significant number of smaller pores, suggesting a higher density of initial pores. The structural differences between HF-2 and HF-3 explain the difference in their selectivity and permeability. HF-2 has a higher initial porosity, relatively uniform pore size, and higher permeability. This ensures the efficient removal of uremic toxins and small waste molecules, but essential proteins may be lost. In contrast, HF-1 has a broad pore size distribution. The gradual reduction in its cumulative intrusion demonstrates a balance between permeability and selectivity. This balance allows it to retain a higher percentage of essential proteins while filtering out small waste molecules. As the air gap increased, the gravity-induced effect became more evident in the membrane's porosity and led to an increase in pore size.

Fig. 14 Mercury porosimetry graph of the HF-2 and HF-3 hollow fiber membrane.

Comparative analysis

The results shown in Fig. 15(a) and (b) represent a combination of lysozyme rejection and urea clearance (%) for reported membranes in the literature. Most membranes are in the range of 30–50% for lysozyme rejection and 80–90% for urea. HF-2 shows a moderate performance with approximately 66.5% urea clearance and relatively low lysozyme rejection (around 14.8%). Similarly, in Fig. 15(c), the correlation between water flux (L m⁻² h⁻¹) and protein retention (%) is illustrated. Membranes with higher protein retention generally show a broad range of water fluxes, with most values above 200 L m⁻² h⁻¹. HF-2 shows ∼90% protein retention and 83.34 L m⁻² h⁻¹ water flux, indicating its potential as a hemodialysis membrane.

Fig. 15 (a) Comparison of water flux and protein retention with other reported membranes, (b) comparison of lysozyme rejection and urea clearance with other reported membranes, and (c) comparison between flat sheet membrane and hollow fiber membrane.

The comparison of the HFMs and FSMs is done on four parameters: water flux, urea, creatinine, and BSA rejection, as shown in Fig. 13(c). FSM shows much higher water flux than HFM and allows more water to pass through the membrane. HFM filters more protein molecules in terms of BSA rejection. Both membranes perform similarly in urea clearance, but HFM has a slight edge. FSM achieves higher creatinine rejection and is more effective at filtering creatinine. Overall, FSM excels in water flux and creatinine filtration, while HFM performs better in BSA and urea clearance.

Biocompatibility analysis

To assess biocompatibility, the CA-4 membrane was blended with different concentrations of citric acid and gelatin to evaluate their performance as anticoagulants. The CA-4 membrane demonstrated the best results in the FSM set, removing 78% of the urea and 31% of the creatinine while 94% of the BSA was retained.

FTIR analysis

The FTIR spectra of the CA-4 membrane blended with gelatin for improved biocompatibility are presented in Fig. 16(a). The –OH stretching band in the 3376–3377 cm⁻¹ region overlaps with the stretching vibration of the –N–H bonds in gelatin in the 3270–3370 cm⁻¹ region. The band at 1328 cm⁻¹ is attributed to the wagging vibration of proline side chains in the gelatin molecules. The C–H stretching vibration at 2947 cm⁻¹ is consistently observed in all membranes. Moreover, the abundance of amine groups in PEI and gelatin is evident from the amide-I peak at 1650 cm⁻¹ and the amide-II peak at 1550 cm⁻¹ which is a characteristic of protein structures.³⁸ The spectra reveal that G-3, G-2, and G-1 contain high amounts of gelatin and PEG, with prominent –OH, C=O, and amide I/II peaks. In contrast, the CA-4 membrane shows weaker amide peaks but stronger C–O–C vibrations indicating a higher content of cellulose acetate.

Fig. 16 FTIR spectra of the cellulose acetate (CA-4) membrane modified with (a) gelatin and (b) citric acid.

The FTIR spectra of CA-4 mixed with different concentrations of citric acid to enhance membrane biocompatibility are shown in Fig. 16(b). The prominent absorption peak at 3374 cm⁻¹ indicates the presence of –O–H groups, overlapping with the peak due to the N–H bonds in PEI. The hydroxyl group is present in PEG, CA, and citric acid, with the latter containing three carboxylic groups in its structure. The spectra of CA-CC3 display the highest transmittance, showing an increase in the intensity of absorption due to the OH group with the increased concentration of citric acid. The –CH stretching vibration in the range of 3000–2800 cm⁻¹ further corroborates the presence of CA and PEG. The carbonyl (C=O) stretching vibration appears as a broad band from 1760–1690 cm⁻¹. The C=C stretching vibration appears in the blended membrane at a wave number of 1500 cm⁻¹,⁶⁹ with the exact carboxylic acid position varying depending on saturation, internal hydrogen bonding, and dimerization. The absence of –NH and –NH₂ in citric acid explains the lack of absorption peaks in the 2500–2000 cm⁻¹ region.⁷⁰ The C=C stretching vibration appears in the blended membrane at a wavelength of 1500 cm⁻¹.⁷

Assessment of protein adhesion

Platelets play a crucial role in hemostasis but their aggregation in the blood can induce thrombosis formation. Fibrinogen adsorbs onto the platelet cell surface membrane and binds with its integrins, thereby causing platelet activation and adhesion. SEM images were taken to evaluate platelet adhesion by examining platelets attached to the CA-4, CA/gelatin membranes, and CA/citric acid membranes. Fig. 17 and 18 reveal that many platelets had adhered and aggregated on the surface of the pristine membrane where their shape appeared flattened and deformed.⁷¹ In contrast, platelets on the CA-G1 and CA-CC1 membranes exhibited smooth edges with little deformation and pseudopodia. When blood contacts an artificial membrane, plasma proteins adsorb onto the surface and form a layer that strongly affects platelet adhesion and activation. Additives play an important role in membrane biocompatibility by reducing the hydrophobic interactions between CA and plasma proteins.

Fig. 17 SEM images of protein adsorption taking place on the surface of a CA-4 modified membrane with gelatin.

Fig. 18 SEM images of protein adsorption taking place on the surface of the CA-4 modified membrane with citric acid.

Protein adsorption and platelet adhesion are closely linked. The highly polar hydroxyl and carboxyl groups in citric acid form hydrogen bonds with water molecules, creating a hydration layer that prevents the adsorption and aggregation of biocomponents. On the other hand, increasing the gelatin concentration enhances plasma protein adsorption, which would be more favourable for bone regeneration and tissue engineering rather than dialysis.

Platelet adsorption study

Platelet adsorption is assessed by evaluating the adsorption of BSA and fibrinogen on the membrane surface. Fibrinogen is particularly important in blood plasma as it combines with platelet GP IIb/IIIa, thus necessitating its quantification.⁷² Chen et al.⁷³ reported that membranes modified with hydrophilic polymers have shown reduced protein adsorption and enhanced biocompatibility. Hence, biomaterials with low protein adsorption are the primary targets.⁷⁴ The extent of protein adsorption on a polymeric surface depends on various membrane-protein interactions such as hydrophobic and hydrophilic interactions, hydrogen bonding from –OH groups, van der Waals forces, and electrostatic effects. The hydrophobic interactions are minimal in the CA-4 membrane, as shown in Fig. 19, due to the presence of reactive functional groups that reduce interfacial energy. Consequently, the protein adsorption properties of the blended membranes are controlled by imine, hydroxyl, and amide functional groups. As the concentration of gelatin in the membrane increases, protein adhesion and proliferation also improve. Therefore, gelatin should be used in minimal amounts. Citric acid contains numerous polar groups with a strong affinity for water, thus enriching the membrane surface with reactive functional groups. When the concentration of citric acid increases in CA-CC3, more BSA attaches to the membrane and makes it unsuitable for dialysis operations.

Fig. 19 Platelet adsorption on the CA-modified hemodialysis membranes.

Evaluation of hemolysis ratio and thrombus formation

The hemolysis ratio is another important aspect of hematology. It is done to quantify the erythrocyte damage caused by artificial surfaces.⁴⁰ Hemodialysis is a continuous process that lasts for 4–6 hours, in which erythrocytes may rupture on the membrane surface and release hemoglobin; this is known as hemolysis. According to the ASTM F-756-08 standard, a safe biomaterial has a hemolysis ratio below 5%.⁷⁵ Hemolysis tests were performed in vitro on the CA-4 membrane and its variants. Fig. 20 illustrates that the pristine CA-4 membrane has the highest HR of 9.8%, whereas CA-G1 and CA-G3 have shown a minimum HR of 2.39% and 4.2%, respectively. The hemolysis ratios of CA-CC1 (6%) and CA-CC2 (5.5%) were nearly identical, while CA-CC3 had a slightly lower HR (4.75%). The amino group in gelatin played a more active role than the carboxyl group of citric acid in restricting protein adsorption, as evident from the fewer proteins bound to the gelatin-blended membranes.

Fig. 20 Hemolysis ratio and thrombus formation for the CA-modified membranes.

On the same pattern, the degree of thrombus formation was determined by testing with human blood without any coagulant. As can be seen in Fig. 20, CA-CC1 showed minimal thrombus formation while CA-4 showed the highest frequency of thrombus formation, as it contained fewer polar functional groups than gelatin and citric acid. Generally, thrombosis is initiated by the adsorption of plasma proteins onto the membrane, resulting in the activation of the platelets. Using polar additives such as gelatin and citric acid enhances the hydrophilicity of the membranes, thus enhancing their biocompatibility.⁴⁴ Leypoldt et al.⁷⁶ reported that the citric acid grafted membrane has shown good antithrombogenicity results. This can be attributed to the flexible hydrophilic molecules on the membrane surface forming a diffusive layer and resisting protein adsorption and platelet adhesion. The hydroxyl group on the cellulosic membrane is responsible for complement activation. The masking of the hydroxyl group can be done by using the amino group of polyethyleneimine. PEG chains carry carboxyl groups at their terminals. This enhances the hydrophilicity of the membrane due to the esterification reaction taking place between the carboxyl and the hydroxyl groups. This property prevents complement activation from taking place on the surface of the membranes.

Gelatin-containing membranes exhibit increased platelet adhesion, which can promote intraluminal thrombosis within hollow fibers and leads to flow obstruction, reduced effective blood flow, abbreviated dialysis duration, and diminished solute clearance; consequently, these surfaces are unsuitable for hemodialysis unless modified to mitigate thrombogenicity.⁷⁷ In contrast, regional citrate anticoagulation acts by chelating ionized calcium (iCa²⁺),⁷⁸ transiently lowering free-calcium activity required for multiple steps of the coagulation cascade and thereby limiting clot formation in the extracorporeal circuit. Together, these considerations indicate that optimizing hemodialysis performance requires membrane surfaces engineered to suppress platelet adhesion and, where appropriate, the adjunct use of regional citrate anticoagulation to maintain circuit patency and preserve solute clearance.

Statistical analysis

Biocompatibility analysis demonstrated clear differences between pristine CA-4 and the modified membranes. Protein adsorption assays showed that gelatin incorporation led to significantly higher fibrinogen adsorption compared to CA-4, particularly for CA-G2 and CA-G3 (p < 0.01), while CA-G1 exhibited reduced BSA adsorption relative to CA-4 (p < 0.01). By contrast, citric acid-modified membranes (CA-CC2 and CA-CC3) displayed elevated BSA adsorption (p < 0.01), whereas CA-CC1 remained statistically comparable to CA-4. Hemolysis testing confirmed that CA-4 induced the highest red blood cell lysis (∼9.9%), while all modified membranes showed significantly lower hemolysis (p < 0.001), with CA-G1 exhibiting the lowest value (∼2.4%). Thrombus formation was also greatest for CA-4 (∼8.4) and was markedly reduced across all modified membranes (p < 0.01), with CA-CC1 and CA-CC2 showing the lowest thrombus scores. These outcomes, summarized in Table 6, confirm that gelatin and citric acid modifications substantially improved the hemocompatibility of CA membranes.

Table 6 Biocompatibility outcomes of pristine and modified membranes. Values are mean ± SD (n = 3). Pairwise comparisons versus CA-4 were evaluated using Welch's t-test. Significance: *p* < 0.05 (*), *p* < 0.01 (**), *p* < 0.001 (***)

Membrane	Fibrinogen (a.u.)	p vs. CA-4	BSA (a.u.)	p vs. CA-4	Hemolysis (%)	p vs. CA-4	Thrombus (a.u.)	p vs. CA-4
CA-4	0.090 ± 0.010	—	0.34 ± 0.02	—	9.86 ± 0.06	—	8.36 ± 0.06	—
CA-G1	0.200 ± 0.020	0.0037 (**)	0.26 ± 0.01	0.009 (**)	2.39 ± 0.02	<0.001 (***)	3.80 ± 0.02	<0.001 (***)
CA-G2	0.375 ± 0.010	0.000004 (***)	0.33 ± 0.03	0.66 (ns)	4.49 ± 0.03	<0.001 (***)	3.73 ± 0.03	<0.001 (***)
CA-G3	0.450 ± 0.020	0.0001 (***)	0.38 ± 0.02	0.07 (ns)	5.20 ± 0.03	<0.001 (***)	5.08 ± 0.03	<0.001 (***)
CA-CC1	0.127 ± 0.030	0.16 (ns)	0.32 ± 0.15	0.84 (ns)	4.17 ± 0.05	<0.001 (***)	1.16 ± 0.05	<0.001 (***)
CA-CC2	0.175 ± 0.030	0.029 (*)	0.42 ± 0.02	0.008 (**)	5.23 ± 0.51	<0.001 (***)	1.93 ± 0.51	0.0018 (**)
CA-CC3	0.254 ± 0.020	0.0007 (***)	0.46 ± 0.10	0.0069 (**)	7.44 ± 0.03	<0.001 (***)	2.45 ± 0.03	<0.001 (***)

---

The improvement in hemocompatibility upon gelatin incorporation may be attributed to the hydrophilic functional groups of gelatin, which enhance water uptake and modulate protein adsorption profiles. Although fibrinogen binding increased on gelatin-containing membranes, the overall reduction in hemolysis and thrombus formation suggests that the adsorbed layer may be less thrombogenic and more biocompatible. Citric acid crosslinking, on the other hand, is known to alter surface charge and introduce additional carboxyl groups, which can reduce platelet activation and clot formation. The strong reduction in thrombus formation observed for CA-CC1 and CA-CC2 aligns with this mechanism. Taken together, these findings demonstrate that the incorporation of both gelatin and citric acid contributes to a more favorable blood-material interaction, albeit through different molecular pathways.

The performance of the developed hemodialysis membranes can be directly attributed to the synergistic role of their individual constituents. Cellulose acetate (CA), employed as the base polymer, provided the structural stability and inherent biocompatibility required for dialysis membranes. However, its excessive hydrophobicity and tendency for protein adsorption limited performance, necessitating chemical modification. This baseline limitation explains the comparatively modest permeability and higher hemolysis ratio observed for pristine CA membranes, establishing the importance of incorporating functional additives.

The incorporation of polyethyleneimine (PEI) played a critical role in improving toxin clearance and protein retention. The amine-rich backbone of PEI imparted strong ionic interactions that enhanced the rejection of macromolecules such as bovine serum albumin (BSA), while maintaining high urea permeability (Table 4). This effect is evident in CA-1 and CA-2 membranes, where PEI alone generated macro-voids that supported solute transport but also reduced uniformity of pore size. The buffering capacity of PEI thus contributed significantly to selective protein retention but, in isolation, was insufficient to optimize flux and hemocompatibility.

Polyethylene glycol (PEG) served as a pore-forming agent and hydrophilicity enhancer, introducing spongy interconnected pores that reduced fouling and facilitated water transport. Increasing PEG concentration reduced the contact angle to as low as 58° (Fig. 10) and improved the balance between water uptake and protein retention. The CA-4 membrane exemplifies this effect, achieving the best overall trade-off between permeability and selectivity, with 78% urea clearance and 94% protein retention. Nevertheless, excessive PEG loading (CA-5) induced dense structures that compromised solute clearance, underscoring the importance of maintaining optimal ratios of PEG and PEI.

The biocompatibility of the membranes was further enhanced by introducing citric acid and gelatin as anticoagulant modifiers. Citric acid contributed through its carboxyl groups, which chelate calcium ions and disrupt the coagulation cascade, thereby reducing platelet adhesion and thrombus formation. Membranes such as CA-CC1 and CA-CC3 demonstrated hemolysis ratios below 5% and minimal thrombus formation (Fig. 18), confirming citric acid's role in improving hemocompatibility. In contrast, gelatin provided amino groups that reduced hemolysis but simultaneously promoted protein adsorption and platelet adhesion. While such behavior may be advantageous for tissue regeneration applications, it is less desirable for long-term dialysis operations where anti-thrombogenic properties are prioritized.

Taken together, these findings highlight how the combined use of PEG and PEI mitigates the individual shortcomings of each additive-PEG countering the cytotoxicity and non-specific interactions of PEI, and PEI reducing PEG-induced fouling. The addition of citric acid further strengthened anti-thrombogenic performance, whereas gelatin introduced application-specific trade-offs. By aligning structural (SEM, AFM), transport (flux, porosity), and biocompatibility (hemolysis, platelet adhesion) results with constituent roles, this study demonstrates a coherent strategy to tune dialysis membrane performance for clinical safety and efficiency.

Conclusions

In this study, three types of hemodialysis membranes: flat sheet, hollow fibre, and biocompatible flat sheet, were developed and optimized using cellulose acetate (CA) as the base polymer. Varying concentrations of polyethyleneimine and polyethylene glycol were blended using the NIPS method. These membranes were designed to improve morphology, hydrophilicity, and filtration efficiency for uremic toxins while retaining functional proteins. The flat sheet membrane (CA-4) showed the best overall performance. It achieved 74% urea clearance, 31% creatinine clearance, and 94% protein retention. Its hydrophilicity, with a contact angle of 58°, water uptake capacity of 40%, and surface roughness of 6.212 nm, made it highly suitable for hemodialysis. Transforming CA-4 into hollow fiber membranes further optimized its clinical performance. Among the hollow fiber membranes, HF-2 emerged as the most promising candidate. It achieved a urea clearance of 66.5%, creatinine clearance of 39.1%, and lysozyme clearance of 14.8% while retaining 90.3% of proteins. HF-2's performance was supported by its uniform pore size distribution and cumulative intrusion starting at 0.9 mL g⁻¹, which ensured high filtration efficiency and selectivity. Its densely uniform pore structure enabled effective mass transfer while minimizing protein loss. Additionally, HF-4 demonstrated the highest water flux, reaching 47 mL m⁻² h⁻¹ mmHg. However, its protein retention was slightly reduced to 83.5% due to changes in morphology caused by larger air gap distances during spinning.

To enhance biocompatibility, citric acid and gelatin were incorporated into the membranes. Citric acid-modified membranes, such as CA-CC3, showed better results with reduced protein adsorption, minimal thrombus formation, and hemolysis ratios below 5% (CA-CC1: 6%, CA-CC2: 5.5%, CA-CC3: 4.75%). In contrast, gelatin-modified membranes like CA-G3 promoted higher protein adsorption and adhesion, making them more appropriate for tissue regeneration rather than dialysis. In conclusion, HF-2 stands out as the best membrane due to its superior balance of toxin clearance, protein retention, and biocompatibility. Overall, the CA/PEG/PEI membranes exhibited increased hydrophilicity, permeability, and in vitro hemocompatibility relative to the base material, supporting their translational potential. Before clinical deployment, however, confirmatory studies are needed to establish durability and thrombogenicity under flow. Recommended evaluations include ISO 10993 hemocompatibility and extractables/leachables testing, assessment of sterilization compatibility, dynamic blood-loop experiments under physiologic shear, and validation in large-animal models. If these milestones are achieved, the membranes will enhance dialyzer performance by reducing protein adsorption and platelet activation, with regional citrate anticoagulation available as an adjunct where appropriate.

Conflicts of interest

There are no conflicts to declare.

Data availability

All relevant data supporting the findings of this study are included within the article.

Acknowledgements

The authors would like to acknowledge the Lahore University of Management Sciences (LUMS) for their grant under STG-180.

References

1. C. P. Kovesdy, Kidney Int. Suppl., 2011,(12), 7–11.
2. A. S. Levey, W. B. Schwartz and J. Coresh, Lancet, 2012, 379, 165–180.
3. S. M. Jalal, M. R. M. Beth and Z. M. Bo Khamseen, Int. J. Environ. Res. Public Health, 2022, 19, 15.
4. H. D. Humes, W. H. Fissell and K. Tiranathanagul, Kidney Int., 2006, 69, 1115–1119.
5. A. Thakuria, B. Kataria and D. Gupta, J. Nanopart. Res., 2021, 23, 1–30.
6. T. Resch, B. Cardini, R. Oberhuber, A. Weissenbacher, J. Dumfarth, C. Krapf, C. Boesmueller, D. Oefner, M. Grimm and S. Schneeberger, Front. Immunol., 2020, 11, 631.
7. A. Almakki, M. E. S. Mirghani and N. A. Kabbashi, Chem. Nat. Resour. Eng. J., 2019, 2, 47–55.
8. H. Waheed and A. Hussain, Int. J. Health Med., 2017, 5–9.
9. S. Tahazadeh, T. Mohammadi, M. Ahmadzadeh Tofighy, S. Khanlari, H. Karimi and H. Banna Motejadded Emrooz, J. Membr. Sci., 2021, 638, 119692.
10. M. Oprea and S. I. Voicu, Ind. Crops Prod., 2023, 206, 117716.
11. O. Azhar, Z. Jahan, F. Sher, M. B. Khan Niazi and S. J. Kakar, Mater. Sci. Eng. C, 2021, 126, 112127.
12. A. Modi, S. K. Verma and J. Bellare, J. Colloid Interface Sci., 2018, 514, 750–759.
13. N. Pandey and J. Bellare, J. Mater. Chem. B, 2025, 13, 9392–9406.
14. D. Zhong, Y. Xu, X. Li, J. Zhou, Y. Wang, H. Li and C. Li, J. Membr. Sci., 2021, 618, 118690.
15. A. Idris and L. K. Yet, J. Membr. Sci., 2006, 280, 920–927.
16. P. S. Parsamehr, M. Zahed, M. A. Tofighy, T. Mohammadi and M. Rezakazemi, Desalination, 2019, 468, 238–248.
17. M. Sivakumar, L. Susithra, D. R. Mohan and R. Rangarajan, J. Macromol. Sci., Part A, 2006, 43, 1541–1551.
18. Y. H. Zhao, B. K. Zhu, X. T. Ma and Y. Y. Xu, J. Membr. Sci., 2007, 290, 222–229.
19. H. Waheed, A. Hussain and S. Farrukh, Desalination Water Treat., 2016, 57, 24799–24806.
20. Y. He, G. Cheng, L. Xie, Y. Nie, B. He and Z. Gu, Biomaterials, 2013, 34, 1235–1245.
21. Z. Chen, Z. Lv, Y. Sun, Z. Chi and G. Qing, J. Mater. Chem. B, 2020, 8, 2951–2973.
22. Q. Zhang, X. Lu, J. Liu and L. Zhao, Membranes, 2015, 5, 120–135.
23. S. Senthilkumar, S. Rajesh, A. Jayalakshmi, G. Aishwarya and D. R. Mohan, J. Polym. Res., 2012, 19, 1–11.
24. D. Labarre, Trends Biomater. Artif. Organs, 2001, 15, 1–3.
25. A. Modi, S. K. Verma and J. Bellare, Colloids Surf., B, 2018, 167, 457–467.
26. M. M. Engelen, P. Verhamme and T. Vanassche, Semin. Nephrol., 2023, 43, 151473.
27. A. Modi, S. K. Verma and J. Bellare, ACS Appl. Bio Mater., 2020, 3, 1589–1597.
28. A. Mollahosseini and A. Abdelrasoul, Chem. Eng. Res. Des., 2023, 195, 466–479.
29. R. T. Tran, J. Yang and G. A. Ameer, Annu. Rev. Mater. Res., 2015, 45, 277–310.
30. L. Li, C. Cheng, T. Xiang, M. Tang, W. Zhao, S. Sun and C. Zhao, J. Membr. Sci., 2012, 405–406, 261–274.
31. A. Zarbock, M. Küllmar, D. Kindgen-Milles, C. Wempe, J. Gerss, T. Brandenburger, T. Dimski, B. Tyczynski, M. Jahn, N. Mülling, M. Mehrländer, P. Rosenberger, G. Marx, T. P. Simon, U. Jaschinski, P. Deetjen, C. Putensen, J. C. Schewe, S. Kluge, D. Jarczak, T. Slowinski, M. Bodenstein, P. Meybohm, S. Wirtz, O. Moerer, A. Kortgen, P. Simon, S. M. Bagshaw, J. A. Kellum and M. Meersch, JAMA, 2020, 324, 1629–1639.
32. Z. Salahuddin, S. Farrukh and A. Hussain, Int. J. Polym. Anal. Charact., 2018, 23, 483–492.
33. N. S. Aseri, W. J. Lau, P. S. Goh, H. Hasbullah, N. H. Othman and A. F. Ismail, J. Water Process Eng., 2019, 32, 100912.
34. Z. Jia and C. Tian, Desalination, 2009, 247, 423–429.
35. H. S. Zakria, M. H. D. Othman, R. Kamaludin and A. Jilani, IOP Conf. Ser. Mater. Sci. Eng., 2021, 1142, 012014.
36. W. Renpu, Advanced Well Completion Engineering, Elsevier, 2011, pp. 221–294.
37. S. Mansur, M. H. D. Othman, A. F. Ismail, S. H. Sheikh Abdul Kadir, F. Kamal, P. S. Goh, H. Hasbullah, B. C. Ng and M. S. Abdullah, J. Appl. Polym. Sci., 2016, 133, 43633.
38. J. Ran, J. Hu, L. Chen, X. Shen and H. Tong, Polym. Compos., 2017, 38, 1579–1590.
39. S. Senthilkumar, S. Rajesh, D. Mohan and P. Soundararajan, Sep. Sci. Technol., 2013, 48, 66–75.
40. A. Gao, F. Liu and L. Xue, J. Membr. Sci., 2014, 452, 390–399.
41. H. Husdan and A. Rapoport, Clin. Chem., 1968, 14, 222–238.
42. H. H. Brown, Anal. Chem., 2002, 31, 1844–1846.
43. K. Ishihara, K. Fukumoto, Y. Iwasaki and N. Nakabayashi, Biomaterials, 1999, 20, 1545–1551.
44. G. Arthanareeswaran, D. Mohan and M. Raajenthiren, J. Membr. Sci., 2010, 350, 130–138.
45. B. Chakrabarty, A. K. Ghoshal and M. K. Purkait, J. Membr. Sci., 2008, 309, 209–221.
46. Y. Ma, F. Shi, J. Ma, M. Wu, J. Zhang and C. Gao, Desalination, 2011, 272, 51–58.
47. R. Guan, H. Zou, D. Lu, C. Gong and Y. Liu, Eur. Polym. J., 2005, 41, 1554–1560.
48. I. C. Kim and K. H. Lee, J. Membr. Sci., 2004, 230, 183–188.
49. W. L. Chou, D. G. Yu and M. C. Yang, Polym. Adv. Technol., 2005, 16, 600–607.
50. J. M. M. Mulder, Basic Principles of Membrane Technology, 1st edn, 1991.
51. ed. K. K. S. W. S. Winston, Membrane Handbook, SpringerNew York, NY, 1992, vol. 1.
52. H. Waheed and A. Hussain, Bionanoscience, 2019, 9, 256–265.
53. S. J. Park, R. K. Cheedrala, M. S. Diallo, C. Kim, I. S. Kim and W. A. Goddard, J. Nanopart. Res., 2012, 14, 1–14.
54. H. S. Mansur, R. L. Oréfice and A. A. P. Mansur, Polymer, 2004, 45, 7193–7202.
55. Q. Wang, Y. Fu, X. Yan, Y. Chang, L. Ren and J. Zhou, Appl. Surf. Sci., 2017, 412, 10–18.
56. Q. Zhang, X. Lu, J. Liu and L. Zhao, Membranes, 2015, 5, 120–135.
57. A. Pagidi, R. Saranya, G. Arthanareeswaran, A. F. Ismail and T. Matsuura, Desalination, 2014, 344, 280–288.
58. G. Arthanareeswaran, P. Thanikaivelan, K. Srinivasn, D. Mohan and M. Rajendran, Eur. Polym. J., 2004, 40, 2153–2159.
59. Y. C. Ko, B. D. Ratner and A. S. Hoffman, J. Colloid Interface Sci., 1981, 82, 25–37.
60. N. Said, W. J. Lau, Y.-C. Ho, S. K. Lim, M. N. Zainol Abidin and A. F. Ismail, Membranes, 2021, 11, 767.
61. S. Mulyati, S. Aprilia, S. Syawaliah, M. A. Armando and H. Mawardi, IOP Conference Series: Materials Science and Engineering, Institute of Physics Publishing, 2018, vol. 352.
62. H. Waheed, F. T. Minhas and A. Hussain, Polym. Bull., 2018, 75, 3935–3950.
63. T. Peters Jr. and A. J. Stewart Jr., Biochim. Biophys. Acta, 2013, 1830, 5351–5353.
64. Z. Xiong, F. Liu, H. Lin, J. Li and Y. Wang, ACS Biomater. Sci. Eng., 2016, 2, 2207–2216.
65. I. C. Kim and K. H. Lee, J. Membr. Sci., 2004, 230, 183–188.
66. M. J. Han and S. T. Nam, J. Membr. Sci., 2002, 202, 55–61.
67. B. Torrestiana-Sanchez, R. I. Ortiz-Basurto and E. Brito-De La Fuente, J. Membr. Sci., 1999, 152, 19–28.
68. B. Chakrabarty, A. K. Ghoshal and M. K. Purkait, J. Membr. Sci., 2008, 309, 209–221.
69. N. Dhenadhayalan, K. C. Lin, R. Suresh and P. Ramamurthy, J. Phys. Chem. C, 2016, 120, 1252–1261.
70. M. N. Z. Abidin, P. S. Goh, A. F. Ismail, M. H. D. Othman, H. Hasbullah, N. Said, S. H. S. A. Kadir, F. Kamal, M. S. Abdullah and B. C. Ng, Mater. Sci. Eng., C, 2016, 68, 540–550.
71. W. Zou, H. Qin, W. Shi, S. Sun and C. Zhao, Langmuir, 2014, 30, 13622–13630.
72. W. B. Tsai, J. M. Grunkemeier, C. D. McFarland and T. A. Horbett, J. Biomed. Mater. Res., 2002, 60, 348–359.
73. H. Chen, L. Yuan, W. Song, Z. Wu and D. Li, Prog. Polym. Sci., 2008, 33, 1059–1087.
74. Z. Yin, C. Cheng, H. Qin, C. Nie, C. He and C. Zhao, J. Biomed. Mater. Res. B: Appl. Biomater., 2015, 103, 97–105.
75. D. Zhong, Z. Wang, J. Zhou and Y. Wang, J. Membr. Sci., 2021, 618, 118690.
76. J. K. Leypoldt, A. K. Cheung, T. Chirananthavat, J. F. Gilson, C. D. Kamerath and R. Barry Deeter, ASAIO J., 2003, 49, 81–87.
77. T. P. Lemmens, Q. Luo, S. J. H. Wielders, J. L. J. M. Scheijen, S. Al-Nasiry, R. R. Koenen, P. Wenzel and J. M. E. M. Cosemans, Thromb. Res., 2024, 244, 109201.
78. V. Pistolesi, S. Morabito, V. Pota, F. Valente, F. Di Mario, E. Fiaccadori, G. Grasselli, N. Brienza, V. Cantaluppi, S. De Rosa, V. Fanelli, M. Fiorentino, M. Marengo and S. Romagnoli, J. Anesth. Analg. Crit. Care, 2023, 3, 1–14.

---