Ventricular tachycardia (VT) provokes sudden cardiac death (SCD), which is a major cause of mortality in developed countries. Implantable cardioverter-defibrillators (ICDs) are an efficient therapy for SCD prevention. In this study we analyze heart rate variability (HRV) in data stored by ICDs. In 29 patients exhibiting VT episodes, the last 1000 normal beat-to-beat intervals are analyzed and compared to an individually acquired control time series (CON). HRV analysis is performed with standard parameters of time and frequency domain as suggested by the HRV Task Force. For scaling analyses of heart rate time series, the fractal dimension is analysed, applying Higuchi\'s algorithm (HFD). Furthermore, detrended fluctuation analysis (DFA) is performed. None of the standard HRV parameters shows significant differences between CON and VT. Before the onset of VT, the scaling characteristics by means of HFD and DFA are significantly changed. In conclusion, scaling analysis reveals changes in autonomic heart rate modulation preceding VT.