Aims: Potassium disturbances are common and associated with increased morbidity and mortality, even in patients without prior cardiovascular disease. We examined six electrocardiographic (ECG) measures and their association to serum potassium levels. Methods and results: From the Copenhagen General Practitioners' Laboratory, we identified 163,547 individuals aged >= 16 years with a first available ECG and a concomitant serum potassium measurement during 2001-2011. Restricted cubic splines curves showed a non-linear relationship between potassium and the Fridericia corrected QT (QTcF) interval, T-wave amplitude, morphology combination score (MCS), PR interval, P-wave amplitude and duration. Therefore, potassium was stratified in two intervals K: 2.0-4.1 mmol/L and 4.2-6.0 mmol/L for further analyses. Within the low potassium range, we observed: QTcF was 12.8 ms longer for each mmol/L decrease in potassium (p < 0.0001); T-wave amplitude was 43.1 mu V lower for each mmol/L decrease in potassium (p < 0.0001); and MCS was 0.13 higher per mmol/L decrease in potassium (p < 0.001). Moreover, P-wave duration and PR interval were prolonged by 2.7 and 4.6 ms for each mmol/L decrease in potassium (p < 0.0001), respectively. Within the lowest potassium range (2.0-4.1 mmol/L) P-wave amplitude was 3.5 mu V higher for each mmol/L decrease in potassium (p < 0.0001). Within the high potassium range associations with the above-mentioned ECG parameters were much weaker. (C) 2018 Elsevier Inc. All rights reserved.