1. Key Considerations for Demonstrating Utility of Non-Clinical Models Pipeline issues Drugs are often being discarded early if they have a less than a several hundred-fold hERG IC50/Clinical plasma level ratio High confidence in a non-clinical arrhythmia battery will permit drugs that might be discontinued to be further developed Examples of drugs that would not have been developed include verapamil, amiodarone, and sodium pentobarbital

2. Against What Should We Validate Non-clinical Assays; QT Prolongation in Humans, Clinical Proarrhythmic Events, or Other Clinical Events, Such As All-Cause Mortality? What should non-clinical models focus on? TdP- consensus for this approach All arrhythmias also considered Clinical Endpoint QT prolongation that is clinically meaningful? Not necessarily the real issue Validity has not been established for TdP- issue of specificity for individual drugs Approach not supported by group Examine high risk/sensitive patients- long QT syndrome pts or other risk factors for TdP Clinical outcomes Hard endpoint TdP hard to capture in post- marketing analyses What will give confidence that a preclinical “battery” will actually predict the risk of drugs with TdP and small QT effects? Focus on drugs with TdP and small QT effects vs those with large QT effects

3. Which Drugs to Validate Against? Modest QT Effects With and without TdP Role of SCD in identifying these drugs? Important to rule out other possible causes of SCD besides TdP since this could obscure sensitivity How to define drugs Must have high confidence that drugs are accurately defined with respect to TdP risk Readily accepted drugs- here there is robust data Some reports of TdP- need to be carefully ascertained Use a formal blinded adjudication approach Isolated cases- problematic- unlikely to be confident in association with TdP Further need identified for epidemiologic expertise ?Use epidemiology approaches Concentration relationships can help Some drugs should be studies that have multiple channel effects in order to generalize results

4. Assay Issues Should be prospective Need to test a group of assays using the same drugs Difficult to prospectively test a large # of assays- logistics Define the models based on available data and scientific mechanistic understanding Concentrations should be in the high therapeutic range Take into account known risk factors Model should be one that can be used/validated widely at multiple sites Some regulators have an emphasis on in vivo assays Well controlled Positive controls

5. Additional Issue Autonomic perturbations may result in increases in the QTc without lengthening ventricular repolarization because of artifacts associated with imperfect QT correction methodologies An issue that warrants careful examination