After assessing the confounder model, lags of 5 min-averages of personally measured air pollutants concurrent and up to 30 min preceding the respective 5 min-ECG intervals were added to the model separately and effects were estimated linearly. As a sensitivity analysis we performed polynomial distributedlag models (PDL, degree = 3) which included all 5 min-averages ofan air pollution variable up to 1 h preceding the 5 min ECG-intervalssimultaneously. In order to avoid multicollinearity, we used a distributed lag model which forces estimates to a polynomial shape.Results are presented as percent change from the outcome mean per increase in interquartile range (IQR) of the respective air pol-lutant