INTRODUCTION: Several noninvasive measures of cardiac risk such as heart rate variability (HRV) cannot be used in patients with atrial fibrillation (AF). One promising exception is the measure of ventricular cycle length entropy (VCLE) where initial data suggest that a reduction in VCLE portends an increased risk of cardiac death in patients with chronic AF. In this study, we hypothesized that measures of short-term HRV during sinus rhythm would correlate with measures of cycle length entropy during paroxysms of AF. METHODS: We tested 25 Holter recordings of paroxysmal AF from the Physionet AF Prediction Database. We calculated HRV parameters including standard deviation of all NN intervals (SDNN), the root mean square root of the differences between adjacent NN intervals (RMSSD), standard deviation of 5-minute averages of NN intervals (SDANN), percentage of adjacent NN interval differences >50 ms (pNN50), and interbeat correlation coefficient (ICC) from 30 minutes of normal sinus rhythm, and entropy measures (the Shannon Informational Entropy [ShEn] and Average of Approximate Entropy [ApEn]) from 5 minutes of AF that occurred during the same 24-hour monitor. Pairwise correlations were used to assess associations, as regression residuals were normally distributed. RESULTS: The mean entropy measures during AF were: ShEn: 4.78 +/- 0.82, ApEn: 0.198 +/- 0.21. When assessed during the 30 minutes immediately preceding AF onset, ICC showed a significant negative correlation with both ShEn (r =-0.65, P < 0.001) and ApEn (r =-0.60, P < 0.01). RMSSD also correlated with both ShEn (r = 0.41, P = 0.04) and ApEn (r = 0.39, P = 0.05), but other HRV measures showed no correlation with VCLE during AF. CONCLUSION: Reductions in RMSSD or increases in ICC, two short-term HRV measures that are known to reflect parasympathetic function in sinus rhythm, are correlated with reductions in the entropy of ventricular response intervals during AF. Our findings suggest that entropy during AF may be modulated, in part, by vagal innervation.