Background and aim: Chronic kidney disease (CKD) affects more than 10% of the world population. CKD patients present impaired potassium homeostasis, which increases the risk of ventricular arrhythmias. Non-invasive estimation of serum potassium before the patient experiences serious effects is of major importance. Current methods for potassium estimation are limited. We investigated changes in maximum Lyapunov exponent (MLE) of T-waves in the electrocardiogram (ECG) and assessed the relationship with potassium. Methods: ECGs of twelve CKD patients undergoing hemodialysis (HD) were processed and T-waves in one-minute windows were extracted for each hour during the HD session. ECGs were additionally calculated from simulated transmural ventricular fibers. The MLE was calculated based on Rosenstein algorithm. Potassium was measured at different time points during HD and simulated from 3 to 7 mmol/l in the modeled ventricular fibers. Results: In CKD patients, MLE took higher values at the beginning and end of the HD session, corresponding to the lowest and highest potassium values. However, the pattern of such relationships depended highly on the characteristics of each patient. The MLE-potassium relationship and its variability was reproduced in the simulations (Fig. 1). The high inter-individual variability in T-wave morphology could be explained by differences in transmural heterogeneities, with 10% variations in the proportion of midmyocardial cells leading to changes larger than 45% in MLE. Conclusions: Changes in MLE calculated from T-waves of the ECG have the potential to be used as indicators of potassium variations in CKD patients, but the associated inter-individual variability should be taken into account, especially under hyper- and hypokalemic conditions.