Seventy patients undergoing adenosine myocardial perfusion scintigraphy were studied. All patients reported abstention from products containing caffeine in the 12 h prior to the test. Blood samples were drawn prior to initiation of the stress test, and serum caffeine levels were determined using high-performance liquid chromatography. All patients were also asked about their coffee and tea drinking habits. Seventy-four percent of patients had measurable serum caffeine levels (n = 52) ranging from 0.1 to 8.8 mg.l-1. Results were correlated with maximum pulse rate, systolic and diastolic blood pressure changes and clinical symptoms during the test. There was no correlation between coffee or tea drinking habits and serum caffeine levels. A serum caffeine level of 2.9 mg.l-1 was considered a cut-off point for comparing patients. No significant difference was seen in mean maximum change of pulse rate, systolic and diastolic blood pressure between patients with serum caffeine levels > or = 2.9 mg.l-1 and those with lower serum caffeine levels. Of eight patients with serum caffeine levels > or = 2.9 mg.l-1, six had no symptoms (75%). When patients were classified as patients with no symptoms or patients with symptoms (mild, moderate or severe), a significant difference was demonstrated between patients with serum caffeine levels > or = 2.9 mg.l-1 and those with lower levels. This suggests 12 h abstention from caffeine may be insufficient. Whether this translates into false-negative perfusion scans should be the subject of a larger study.