\textbf {Background and aim:} Non-invasive estimation of serum potassium, \kaco, and calcium, \caco, are of major importance to prevent ventricular arrhythmias and sudden cardiac death, but current ambulatory estimation methods, mostly based on T~wave properties, are limited. We present an analysis based on QRS slopes to detect and quantify electrolyte abnormalities in end-stage renal disease (ESRD) patients.
\textbf{Methods:} We applied principal component analysis to 12-lead electrocardiograms (ECGs) of 29 ESRD patients undergoing hemodialysis (HD). In the first principal component, we analyzed two-minute segments at the end of each HD hour and computed a mean warped QRS complex (MWQRS) representing an optimal average of QRS complexes. We calculated the upward (\QRSus) and downward (\QRSds) slope of the MWQRS and we quantified the relative slope changes with respect to the end of the HD session ($\Delta$\QRSus\ and $\Delta$\QRSds).
\textbf{Results:} $\Delta$\QRSus\ and $\Delta$\QRSds\ significantly increased and decreased, respectively, during HD in association with decreasing \kaco{}\ and increasing \caco{}. The median correlation coefficients of $\Delta$\QRSus\ and $\Delta$\QRSds\ with the change in \kaco\ were $-$0.84 and 0.88, respectively. Corresponding correlation coefficients with the change in \caco\ were 0.78 and $-$0.91. The quantified QRS~slope changes were related to both amplitude and duration changes in the QRS~complex during HD.
\textbf{Conclusions:} Changes in QRS~slopes are strongly related to variations in \kaco{} and \caco{} during HD and could be used to monitor electrolyte concentrations in ESRD patients.