The median relationship (IC50 for hERG divided by Ce5%) was (for dog and anaesthetised guinea pigs, respectively), 5/34 for dofetilide, 7/28 for “X”, 6/13 for “Y”, 336/94 for “Z” and 279/114 for “Q”. Thus, as a rule of thumb Ce5% may be crudely estimated as hERG IC50 divided by 10–30 for putatively clean hERG inhibitors. However, other mechanisms beyond Cav3.2, Nav1.5 and Ito inhibition can blur this relationship. Concomitant hERG/IKs inhibition may be such a mechanism. doi:10.1016/j.vascn.2012.08.044 A new automated arrhythmia detector in preclinical safety studies Philippe Michon, David Labarre, Philippe Zitoun, Florence Koeppel NOTOCORD Systems, Croissy-sur-Seine, France Cardiac arrhythmias are a major drug adverse effect and have caused drug withdrawal in a number of occasions. For this reason, it is important to identify such cardiac liability early on in the drug development process. However, direct arrhythmia assessment in ECG signals is a time consuming effort and is rarely achieved, due to lack of a suitable automated tool. This study presents the performance evaluation of a new software application called ARR30a for automated arrhythmia detection in preclinical studies. The five major arrhythmia types are detected, namely sinus pauses, atrial beats, junctional beats, ventricular beats and AV-blocks II (aka. isolated P-waves). The algorithm is based on rhythm analysis, QRS morphology and P-wave analysis. The software has been tested on a database of representative ECG signals of over 80 files acquired from various animal species and experimental protocols. Performances were measured in terms of predictivity and sensitivity by comparing arrhythmia manual marking with automat- ed detection. In large animals (dogs, nonhuman primate and pigs), ARR30a sensitivities reached 78.0%, 90.6%, 82.2% for junctional beats, ventricular beats and AV-blocks II respectively, with predictivities of 93.5%, 83.4% and 94.4% respectively. Significantly lower sensitivity was observed in small animals for junctional beats. This is mainly due to challenging detection of low amplitude P-waves combined with noise in rat ECG signals. Robustness to noise was assessed by adding increasing noise levels to ECG, and showed no significant impact on arrhythmia detection at moderate noise levels. Processing time for a 24-hour signal was about 4 and 6 min in dogs and rats respectively, on a 3 GHz processor. Overall, ARR30a is a powerful tool to quantify the incidence of arrhythmias in animal ECG signals. doi:10.1016/j.vascn.2012.08.045 Decreased QT variability with DSI ECG pattern recognition: Comparison to algorithm-based analysis in 60 human subjects Olivier Meyer a , Henry H. Holzgrefe b a Institute of Clinical Pharmacology-Roche, Strasbourg, France b Charles River Laboratories, Reno, NV, United States Introduction: Technological advancements have enabled the con- tinuous acquisition of high fidelity ECGs where full evaluation is only feasible via computer programs employing various T-wave detection algorithms (ALG). Recently, pattern recognition software (PRO) has been developed which matches ECG waveforms to user defined templates. Methods: We compared the baseline raw QT and RR values obtained in 60 human subjects (24 h, 1000 Hz, Mortara Instruments, H12+) with ALG and PRO analysis (ECG PRO, ver. 5.0, Data Sciences, Intl). Data were analyzed in beat mode, and the intra-individual SD derived for successive 2 min intervals for 24 h. Results: The mean QT was longer (371 ± 18 vs. 396 ± 20 ms, and the SD reduced (12.6 ± 5.9 vs. 4.9 ± 1.1 ms), ALG vs. PRO, respective- ly. The QT SD ranges were 5.4–29.8 ms (ALG) and 3.4–7.2 ms (PRO). RR was similar for both methods (883 ± 71 vs. 900 ± 74 ms), ALG vs. PRO, respectively. Conclusions: With PRO, raw QT intervals were significantly longer and demonstrated markedly decreased intra-individual variability. There were no differences in R-wave detection. Manual inspection of ALG and PRO QT measurements demonstrated that PRO analysis effectively corrected for algorithm-based inaccuracies in T-end determinations due to variable waveform morphology. PRO analysis resulted in uniform QT variance over time, in conjunction with a sig- nificant SD reduction. PRO dramatically improved the precision and reduced the variance of raw QT measurements, resulting in improved statistical power with a concomitant reduction in subject number. doi:10.1016/j.vascn.2012.08.046 Efficacy of ECG pattern recognition analysis in human subjects with variable T-Wave morphology: Impact on study power Olivier Meyer a , Henry H. Holzgrefe b a Institute of Clinical Pharmacology-Roche, Strasbourg, France b Charles River Laboratories, Reno, NV, United States Introduction: Between-subject standard deviation (SD) and number of subjects (n) define power. While n must be adequate to accommodate unexpected increases in SD, the inclusion of excess subjects needlessly consumes experimental resources. Recently, new methodologies have been described which improve precision in raw QT interval measurement. Methods: Continuous Holter ECGs (24 h, 1000 Hz, Mortara In- struments, H12+) were obtained from a human TQT study (n = 60). Automated ECG interval measurements were performed by algorithm (ALG) and pattern recognition (PRO) analyses (ECG PRO, ver. 4.9, Data Sciences, Intl). For ALG and PRO, raw QT intervals were grouped in ascending 10 ms RR bins and modeled by log–log-linear regression. Individual QT rate-correction factors (β) and r 2 were derived for ALG and PRO. Results: β was similar for both methods (0.296 ± 0.08 vs. 0.289 ± 0.06, ALG vs. PRO, respectively). With ALG 3/58 subjects exhibited an unacceptable (b 0.9) r 2 of 0.77 ± 0.19. With PRO r 2 was improved to 0.98 ± 0.01 (P = 0.01). Conclusions: PRO improved r 2 by providing precise and repro- ducible raw QT interval measurements in subjects with variable T-wave morphology which could not be accurately analyzed by ALG. PRO allowed for the recovery and inclusion of 3 subjects that would have been excluded from the final study analysis. PRO has the potential to recover subjects which are not amenable to ALG analysis, increasing the available n, improving study power, and eliminating the need for additional subjects. This is particularly important in preclinical studies where n is limited by design. doi:10.1016/j.vascn.2012.08.047 Integration into preclinical toxicity studies of functional endpoints captured with telemetry: To implant or not? Jean-michel Guillon, Mohamed Slaoui, Chrystelle Battut, Bernadette Hamon, Michel Doubovetzky, Xavier Palazzi Abstracts 172