The analysis of heart rate variability (HRV) requires a sufficiently high ECG sampling frequency. As to spectral analysis, the 1996 Guidelines on HRV defined the optimal range to be between 250 and 500Hz, suggesting parabolic interpolations to refine the R peak for sampling rates between 100 and 250Hz. Multiscale Entropy (MSE) is a popular complexity-based HRV method for which the effects of quantization errors by low sampling rates have never been evaluated systematically. Our aim is therefore to evaluate the effects on MSE of low ECG sampling rates and of the parabolic interpolation. We considered the 21 ECG recordings sampled at 500Hz for 10-12 minutes in supine participants of the public EuroBavar dataset. We decimated the ECG simulating decreasing sampling frequencies from 250 to 50Hz. We extracted the R-R intervals (RRI) without and with peak reconstruction by parabolic interpolation. For each RRI series, we estimated the MSE with embedding dimension m=2 deriving SampEn and MSE at high-frequency (MSE-HF) and low-frequency (MSE-LF) scales as in [1]. The 500Hz sampling rate was the reference. The estimates at each subsampling were expressed as the percentage of the reference and the associated error was quantified by the interquartile range (IQR) of their distribution. SampEn showed high sensitivity to the sampling rate: IQR was >10% at 250Hz and >30% at 167Hz. The errors for SampEn dramatically decreased with the parabolic interpolation being IQR<2% up to 71Hz subsampling. The MSE-HF and MSE-LF estimates were less sensitive to the sampling rate, the errors being not larger than 2% even at 50Hz and lower after parabolic interpolation. Thus the sampling rate is much more critical for SampEn than for MSE at larger scales, and when the sampling rate is lower than 250Hz, interpolation procedures should refine the R peak. [1] Entropy, 2019 doi:10.3390/e21060550