Administration of caffeine in the evening produces poor sleep. Patients with insomnia have characteristic electrocardiogram (ECG) changes, including increased heart rate (HR), increased sympathetic activity, and decreased parasympathetic activity. Fifteen young adult normal subjects slept in the laboratory for several nights prior to randomization into a caffeine protocol where subjects received caffeine 400 mg 30 min prior to one night of sleep and placebo randomly prior to another night. ECG was sampled at a rate of 500 Hz and recorded onto a PC. Data samples of 256-s periods of the ECG trace were taken from wake (before sleep), stage II, and REM for placebo and caffeine conditions. A peak detection algorithm was used to identify the R-R intervals (in milliseconds) from the ECG. A common QT variability algorithm was used to find the QT interval for each beat using the time-stretch model. The powers for HR and QT series were integrated in the bands of LF (low frequency: 0.04-0.15 Hz) and HF (high frequency: 0.15-0.5 Hz) bands. There was a significant caffeine by sleep stage interaction for LF/HF ratios (F = 4.0; df = 2, 18; P = .04). LF/HF ratios were significantly higher during REM following caffeine administration. There was also a significant caffeine by sleep stage interaction for QTvi (QT variability normalized for mean QT interval divided by HR variability normalized for mean HR; F = 5.6; df = 2, 12; P = .02). QTvi was also significantly higher during REM following caffeine administration. The higher LF/HF ratios and QTvi during REM are most likely due to the sympathetic effects of caffeine. These findings suggest that excessive caffeine intake may result in adverse cardiovascular events in vulnerable subjects.