Session S71.5

Risk Stratification in Congestive Heart Failure Patients Using a Model-Based Approach to Heart Rate Turbulence Characterization

JP Martínez, D Smith, A Bayés de Luna, P Laguna*, L Sörnmo

Universidad de Zaragoza
Zaragoza, Spain

Heart rate turbulence (HRT) is the typical pattern of response of the heart rate to a ventricular premature beat (VPB), consisting of an early heart-rate acceleration followed by a deceleration. HRT is currently assessed by two parameters: turbulence onset (TO) and turbulence slope (TS), both obtained by averaging RR tachograms following a VPB. We have shown recently that a model-based detection-theoretical approach to HRT results in detection statistics outperforming TO/TS in identifying the presence or absence of HRT.
The aim of this work is to evaluate this approach for risk stratification in a population of patients with congestive heart failure (CHF). A population of 58 patients from the MUSIC2 study with sinus rhythm and ischemic dilated cardiomyopathy has been studied. Ten of the patients suffered cardiac death during the 3-year follow-up period. After identifying the eligible VPBs in each patient, the tachograms around each VPB and the post-VPB IPFM modulating signal were obtained and averaged for each patient. For each patient we computed the TO and TS indices, as well as a Neyman-Pearson detection statistic (NP) on the subspace defined by the 3 first KL functions of the IPFM input modulating signal [Martinez et al. EMBC 2009].
We found significant differences (Mann-Withney U test) between survivors and cardiac death groups for TS (ms/beat): 2.6 vs 1.6, p=0.035, and NP: -0.10 vs -1.38, p=0.0048, but not for TO (%), -0.20 vs 0.20, NS. Similarly, we found significant differences between NYHA classes II and III in both parameters, TS: 3.0 vs 1.5, p=0.0033, NP: -0.10 vs -1.63, p=0.0026. To evaluate risk stratification, we selected a threshold for NP of -1.0 with the help of ROC curves. The usual thresholds (TS<2.5 TO>0) were used for TS and TO. Only the parameters TS and NP presented a Relative Risk significantly greater than 1. TS: Se: 90.0%, Sp: 54.2%, Relative Risk: 7.8 (95% Confidence Interval:1.13-54.4). NP: Se: 90.0%, Sp: 66.7%, Relative Risk: 11.9 (95% CI:1.68-83.7). These results showed improved performance in risk stratification of this new description of HRT based on the KLT of the IPFM input signal with respect to classical parameters.

(Abstract Control Number: 43)