To assess reduction of infarct size by therapeutic intervention, a high predictive accuracy is mandatory. The CK release in the circulation (CKr) was studied in 12 consecutive patients after uncomplicated myocardial infarction, admitted within 5 h after onset of symptoms. Despite improvement of existing methods, such as a more frequent sampling, CK-MB determination instead of total CK determination and use of a gamma-exponential instead of a log-normal curvefitting technique, the correlation between CK, predicted from measurements within 7 h after the start of CK rise and CKr calculated after completion of the CK curve remained poor. Computer simulations were done to investigate measurement errors as a cause of this failure. Normally distributed noise, with standard deviations ranging from 0.2% to 8.0% of peak CK-MB, was added to the first points of an ideal gamma-exponential CK-MB curve and predictions were made from these "noisy" points. A small noise already produced a great variation in prediction: 0.8% noise resulted in a deviation of predicted CKr from calculated CKrranging from -20 to + 6%. It is concluded that adequate prediction of infarct size from serial CK determinations in the first 7 h after onset of the CK rise must fail if the precision of the biochemical determination is not less than 0.4%.