Bovine blood samples were obtained from a local distributor (Tokyo Shibaura-Zoki, Tokyo, Japan). We adjusted it to attain a hematocrit of 32% and 6.5 g/dL of total protein concentration using saline (Otsuka, Tokyo, Japan), plasma, and whole blood. The blood contained 20 mM of sodium citrate, which was used as an anticoagulant.

Polysulfone (PS) (ABH-22PA; Asahi-Kasei, Tokyo, Japan) and cellulose triacetate (CTA) (FIX-210Seco; NIPRO, Osaka, Japan) membranes were used. The specifications of both membranes are listed in Table 1. The number of fibers was counted individually.

Using 2 L of bovine blood, we conducted HD, pre-dilution online HDF, and I-HDF experiments for 123 min. The blood (QB) and dialysate (QD) flow rates were set at 200 mL/min and 500 mL/min, respectively. The bovine blood temperature was set to 37 °C using a hot stirrer. Heparin (NP; NIPRO) was continuously infused at 3,000 U/h. Filtration was performed after circulating the blood flow pump for 1 min, and the pressures of the blood inlet (PBi), blood outlet (PBo), dialysate inlet (PDi), and dialysate outlet (PDo) sides of the HDF membrane were continuously recorded. A personal dialysis machine (DBG-03; Nikkiso, Tokyo, Japan) and dialysis fluid concentrate (AF-2; FUSO, Kawasaki, Japan) were used.

The QB, QD, and ultrafiltration flow rate (QF) were set to 200, 500, and 30 mL/min, respectively. The replacement fluid was delivered from the dialysis fluid tank at the same flow rate as the QF to the bovine blood. The total filtration volume was 3.69 L (Fig. 1a).

Schematic diagram of the experiments. Experiments are conducted ex vivo using 2 L of bovine blood at 37 °C. a Hemodialysis (HD) mode, b pre-dilution online hemodiafiltration (HDF) mode, and c intermittent infusion HDF (I-HDF) mode

The QB, QD, QF, and supplemental flow rate (QS) were set at 200, 500, 30, and 208 mL/min, respectively. The same volume flow rate of QF was added from the dialysis fluid tank to the bovine blood (Fig. 1b).

The QB, QD, and QF were set to 200, 500, and 30 mL/min, respectively. Supplemental fluid (24 mL/min) was added to the dialysis tanks. The back filtration flow was caused by the dialysis fluid tank, and back filtration was injected at 200 mL/min for 20 s at 10 min intervals (Fig. 1c).

A pressure gauge was created using an analog-to-digital converter, pressure transducer, and direct current adapter, and it was measured at 5 s intervals. Pressure was measured continuously at four points: PBi, PBo, PDi, and PDo of the HDF module. The pressure differences between the ΔPA of the module, ΔPV, TMP, and Lp were calculated as the average values over 5 s using the following formula:

$$TMP =\frac_+_\right)}-\frac_+_\right)}$$

where ΔPA denotes pressure differences between the inlet portion (mmHg), ΔPV denotes pressure differences between the outlet portion (mmHg), Lp denotes the filtration coefficient (mL m−2 h−1 mmHg−1), TMP denotes transmembrane pressure (mmHg), QF denotes the filtration flow rate (mL/h), and A denotes the membrane surface area (m2).

The initial fouling characteristics were obtained using the Lp, which was expressed on a logarithmic scale.

This was not required for this study.