Associate Professor Virginia Tech Blacksburg, Virginia, United States
Introduction:: More than 1 in 7 US adults and 5.4 million people worldwide are estimated to have end stage renal disease, for which hemodialysis is a primary therapy. The median life expectancy for hemodialysis patients is only 5 years due to the poor performance of hemodialyzers compared to healthy kidneys. Remarkably, the design of hemodialyzers hasn’t changed significantly since the hollow fiber dialyzer was introduced in the 1960s. A growing body of evidence demonstrates that current hemodialyzer designs and operating regimens are not optimized for the clearance of uremic toxins. But efforts to improve the performance of hemodialyzers are hindered by the lack of high-fidelity computational models that capture the complex geometry and solute clearance physics of hemodialyzers. Here, we aim to fill this gap by developing the first multi-fiber hemodialyzer models that predict the complex, multiscale, three-dimensional blood and dialysate flow fields and clearance physics through and around the hundreds to thousands of parallel hollow fibers that hemodialyzers contain.
Materials and Methods:: We developed families of three-dimensional multi-fiber finite element (FEM) models of several previously commercially available hemodialyzers. The models ranged in complexity from 1-fiber to 127-fiber models. We performed three in silico studies using the models, validating our results against published in vitro studies. The studies investigated: i) the effect of hemodialyzer fiber packing density (φ) on the clearance (CL) of creatinine, a primary uremic toxin; ii) the effect of total hollow fiber dialyzer membrane surface area (A) on the CL of creatinine; and iii) the effect of the dialysate flow rate on the hemodialyzer’s creatinine mass transfer coefficient (Ko). In the first study, we varied blood flow rate (Qblood) in the clinically relevant range of 100 to 400 mL/min while keeping the dialysate flow rate (Qdialysate) constant at 500 mL/min in a single-fiber FEM model of the Nikkiso FLX-12GW hemodialyzer with φ = 0.296, 0.353, 0.441, and 0.531. In the second study, we used Qblood = 300 mL/min and Qdialysate = 500 mL/min and employed N-fiber FEM models with successively higher numbers of hollow fibers, (N = 1, 7, 19, 27, 61, 91, and 127), to predict the CL of creatinine in three Asahi-Kasei hemodialyzer models (APS-08SA, APS-15SA, and APS-25SA ) that vary in A (0.8, 1.5, and 2.5 m2). In the third study, we held Qblood constant at 200 mL/min and varied Qdialysate from 200 to 800 mL/min in a high packing density hemodialyzer (APS-15SA, φ = 0.57) using several N-fiber models (N varied from 7 to 127).
Results, Conclusions, and Discussions:: The first study confirmed known inaccuracies in clearance predictions using single-fiber models for high packing density hemodialyzers.1 Figure 1ashows the CL of creatinine versus φ. For the lowest-φ model, the calculated CL values matched the published in vitro values closely.2 But the error in predicted CL increased with φ, to almost 10%, for φ = 0.53. The second study demonstrated that the multi-fiber FEM hemodialyzer models we are developing can capture complex phenomena previously captured only in in vitro studies. Figure 1b shows the predicted Ko for creatinine versus A using N-fiber FEM hemodialyzer models. N >=7 fiber model predictions were closer to the in vitro fouled Ko values.3 This observation is in line with the fact that Ko decreases as A increases.4 The third study provides further model validation and a method for identifying the number of hollow fibers needed for flow physics accuracy for a given hemodialyzer model. Figure 1c shows the inverse mass transfer coefficient (1/ Ko) for Creatinine plotted against the inverse nondimensional mean dialysate flow speed, 1/ReD, for 7- to 127-fiber models of the APS-15SA. These data allow one to determine the exponent, x, in the power-law relationship 1/Ko = A + B*(1/ReD)x , where A and B are constants, for the different N-fiber models. The 19-fiber model was found to provide the closest exponent value (0.5373) to the in vitro value of 0.47.5 Exponent values greater than 0.33 signify the presence of local turbulence effects due to inter-fiber interactions.6
Acknowledgements (Optional): : The authors thank Michael Rocco and Pirouz Daeihagh for their contributions and acknowledge support from the Wake Forest Innovations Catalyst Fund and the Virginia Tech Regenerative Medicine Interdisciplinary Graduate Education Program.
References (Optional): : 1. Happel, J. Viscous flow relative to arrays of cylinders. AIChE J.5, 174–177 (1959).
2. Yamashita, A. C., Fujita, R., Tomisawa, N., Jinbo, Y. & Yamamura, M. Effect of packing density of hollow fibers on solute removal performances of dialyzers. Hemodial. Int.13, 2–7 (2009).
3. Kiguchi, T., Ito, H. & Yamashita, A. C. Article Effect of Membrane Surface Area on Solute Removal Performance of Dialyzers with Fouling. Membranes (Basel).12, (2022).
4. Leypoldt, J. K. et al. Hemodialyzer mass transfer-area coefficients for urea increase at high dialysate flow rates. Kidney Int.51, 2013–2017 (1997).
5. Kunikata, S. et al. Technical characterization of dialysis fluid flow of newly developed dialyzers using mass transfer correlation equations. ASAIO J.55, 231–235 (2009).
6. Costello, M. J., Fane, A. G., Hogan, P. A. & Schofield, R. W. The effect of shell side hydrodynamics on the performance of axial flow hollow fibre modules. J. Memb. Sci.80, 1–11 (1993).
