OBJECTIVE: The objective of this study is to investigate the influence of the day-to-day variability of the measures of heart rate variability (HRV) on the sample size calculation for the study of cardiac autonomic neuropathy. MATERIAL AND METHODS: We analyzed HRV in the frequency domain [very low (VLF), low (LF), and high frequency (HF) bands] and in the time domain [the root mean squared of successive RR intervals differences (RMSSD); the mean RR intervals (RRNN); the standard deviation of RR intervals (SDNN) and the coefficient of variation (CV)] during a 5-min electrocardiogram record. We also analyzed the heart rate response to deep breathing [expiration:inspiration ratio], to the Valsalva maneuver and to standing [maximum:minimum ratio] and the blood pressure response to standing. The day-to-day variability was assessed by calculating the within-subject standard deviations (WSSD), limits of agreement, typical errors and the ratio of the WSSD to the mean values obtained on days 1 and 2 (WSSD/GM). RESULTS: Sixty-seven healthy subjects (45 females), aged 27 (19-39) years, were recruited. The RMSSD, CV, VLF, LF, HF and blood pressure response to standing showed marked variability (WSDD/GM (%)=237.7, 455.1, 69.9, 126.5, 81.3 and 380.5, respectively), while the RRNN, SDNN, Valsalva, expiration:inspiration and maximum:minimum ratio showed less variability (WSSD/GM (%)=6.4, 24.5, 18.6, 11.0 and 14.1, respectively). The minimum differences expected to be statistically significant for the autonomic measurements were calculated. CONCLUSION: Some tests that assess HRV showed adequate reproducibility. This study allows the determination of a sample size calculation for longitudinal or drug-testing studies.