SummaryThe objective of this study was to compare linear and non-linear analysis of heart rate variability (HRV) in terms of correlation with haemodynamic fluctuation during induction of general anaesthesia. Pre-operatively, HRV was estimated by the MemCalc method in 114 patients scheduled for general anaesthesia. After anaesthesia was induced with propofol, fentanyl and vecuronium bromide, tracheal intubation was performed. Haemodynamic fluctuations during induction of anaesthesia were recorded and the correlation between pre-operative HRV and haemodynamic fluctuation was examined using logistic regression analysis. As an index of non-linear analysis of HRV, ultra short-term entropy (UsEn) correlated better with blood pressure fluctuation than did the ratio of the power of low frequency component of HRV to that of high frequency component (LF ⁄ HF). In contrast, although LF ⁄ HF significantly correlated with heart rate increase caused by tracheal intubation, the correlation between UsEn and heart rate fluctuation was not significant. Recently, some investigators have found linear analysis of heart rate variability (HRV) to predict the incidence of hypotension caused by induction of general [7] or spinal anaesthesia [8,9]. While frequency-domain measures of HRV, such as spectral analysis, focus on the status of cardiac autonomic activity, it is proposed that non-linear indices may provide insight into the overall structure of the heart rate regulating system. Although it was reported that pre-operative point dimension correlation (PD2), as a non-linear index of HRV, predicted hypotension after spinal anaesthesia [10], there is no previous report examining the correlation between non-linear index of HRV and circulatory response to induction of general anaesthesia.The MemCalc method, which is a combination of the maximum entropy method for spectral analysis and the non-linear least squares method for fitting analysis (Tarawa, Suwa Trust, Tokyo, Japan), has recently been developed [11,12]. This method enables us to estimate reliable HRV from a series of RR intervals over 30 s and recognises the abnormal RR interval of premature beats or artefacts, including noise, and removes it automatically. The MemCalc method does not cause distortion of the power calculation even if the underlying variation is changed, e.g. a change in respiration. When compared to