We compared analogue and digital 24-h electrocardiogram (ECG) recordings in the assessment of heart rate variability (HRV) in 41 patients with suspected coronary artery disease. The patients underwent ambulatory ECG recordings simultaneously with analogue (A) (Marquette 8500) and digital (D) (Oxford Medilog FD-3) recorders. Digital ECG recordings were analysed with the Excel Medilog II system (DE) and analogue recordings were analysed with both Marquette (AM) and Excel Medilog II (AE) systems. SDNN, SDANN, rMSSD and pNN50 were calculated for the assessment of time domain measures of HRV, and powers of low (0.04-0.15 Hz) and high (0.15-0.40 Hz) spectral components of HRV for the assessment of frequency domain measures of HRV. Correlations between time domain measures of HRV derived from AM, AE and DE recording and analysis techniques were high (r = 0.803-0.999, P < 0.001). SDANN and pNN50 assessed with AM and DE were equal, whereas SDNN and rMSSD values differed slightly but significantly from each other (P < 0.01 for both). All time domain measures of HRV assessed with AE differed significantly from AM (P < 0.001) and from DE (P < 0.01). Correlations between frequency domain measures of HRV derived from AM, AE and DE were also high (r = 0.973-0.992, P < 0.001). Despite this, frequency domain measures of HRV assessed with AM, AE and DE differed significantly from each other (P < 0.001). In conclusion, when analysed with their own analysis systems, analogue and digital recordings provided almost identical time domain, but not frequency domain, measures of HRV. In addition, analysis of an ECG recording with an analysis system of different manufacturer should be avoided.