Abstract Background and Aims Kt/V urea is conventionally used hemodialysis (HD) adequacy marker, whereas many other uremic toxins with differing removal kinetics affect the mortality of end stage kidney disease (ESKD) patients. Among them, uric acid (UA) has been shown to cause cardiovascular mortality of ESKD patients. Recently, determination of UA concentration in blood with optical sensors during HD has been attempted by Lin et al in 2021 and Żyłka et al in 2023, but the accuracy has been relatively low. Moreover, Żyłka et al measured light absorption of spent dialysate at 573 nm to determine UA concentration, whereas previous research has shown that dialysate and UA have negligible light absorption in that region. The aim of this observational study was to predict concentration of UA with higher accuracy in ESKD patients’ blood during HD with various treatment settings by monitoring spent dialysate noninvasively with an optical sensor. Method Twenty-two ESKD patients on chronic HD were included into the study. Each patient received four 4 h long midweek HD sessions: one standard HD and three haemodiafiltration (HDF) sessions with different settings. Median (interquartile range) settings of the 1 HD session were: effective blood flow (Qb) = 199 (198-199) mL/min, dialysate flow (Qd) = 297 (297-298) mL/min, ultrafiltration rate (UF) = 13 (8-16) mL/min, dialyzer area 1.5 m2 and for the 3 HDF sessions: Qb = 298 (296-356) mL/min, Qd = 788 (489-792) mL/min, substitution fluid flow (Qsubs) = 96 (66-106) mL/min, UF = 11 (8-15) mL/min, dialyzer areas 1.8 m2 and 2.2 m2. Blood samples were collected before the start and at the end of each HD session from the arterial blood line and spent dialysate samples were collected 7, 60, 120, 180 and 240 min after the start from the outlet of the HD machine. Concentration of UA in spent dialysate and serum samples were determined with HPLC. An optical sensor (Optofluid Technologies OÜ, Tallinn, Estonia) was connected to the spent dialysate outlet of the HD machine and ultraviolet (UV) light absorption of spent dialysate was measured online in real time for each HD session. A linear interactions regression model was created using light absorbance at four different UV wavelengths to predict UA concentrations in collected spent dialysate samples. A curve of UA concentration for the whole HD session was determined based on the model and UV absorbance data measured with the optical sensor. Thereafter, UA concentration in spent dialysate at the start and at the end of each session (UAsensor hereinafter) was found from the concentration curve. The data of 88 HD sessions were equally divided into calibration and validation sets. Linear regression model was created using data of calibration set to predict concentration of UA (UApredicted) in blood as follows: UApredicted = a • UAsensor • $\frac{{{\rm{Qd}} + {\rm{\ Qsubs}} + {\rm{UF\ }}}}{{{\rm{Qb}}}}$, where “a” is the slope, UAsensor is the UA concentration in spent dialysate found from the concentration curve of the optical sensor, Qb is effective blood flow and Qd+Qsubs+UF is the total flow of spent dialysate. Subsequently, the accuracy of the model was evaluated on the calibration and validation data using Bland Altman analysis. Results UA concentrations in spent dialysate determined with HPLC and estimated with the optical sensor were strongly correlated (R2 = 0.992) and with comparable accuracy (BIAS±SE) (0.1 ± 3.84) µmol/L. Coefficient of determination 0.951 with the accuracy of 0.00 ± 31.43 µmol/L was achieved for calibration group, and coefficient of determination 0.962 with the accuracy of −7.64 ± 32.06 µmol/L was achieved for validation group between the UA concentration in blood determined with HPLC and corresponding values predicted by the optical sensor, respectively (Fig. 1). Conclusion Concentration of UA in dialysate and blood can be estimated noninvasively and accurately with the optical sensor in real time. This can be used to estimate removal efficiency and levels of UA during HD. In future, removal of other uremic toxins could be additionally monitored leading to a more personalized HD treatment.
Read full abstract