Revisiting the hERG safety margin after 20 years of routine hERG screening

Introduction: It has been two decades since screening new molecules and potential clinical drug candidates against the hERG potassium channel became a routine part of safety pharmacology. The earliest heuristic for what was an adequate safety margin to separate molecules with a potential liability to cause the arrhythmia torsade de pointes (TdP) from those with no such liability emerged in 2002 and was determined to be a hERG IC50 value 30-fold above the therapeutic free plasma concentration (Webster, Leishman, & Walker, 2002). In the intervening years nonclinical and clinical ICH guidance has been introduced and intense scrutiny has been applied to the QT interval of the electrocardiogram in animals and man. Has the 30-fold heuristic stood the test of time? Methods: The hERG margins between the IC50 value and the therapeutic unbound plasma concentrations were examined for 367 compounds. These margins were compared against the categories used by www.CredibleMeds. com to classify a drug's TdP risk. A subset of 336 of these drugs were compared against their US product labels with respect to black box warnings on QTc prolongation or TdP, warnings and precautions on QTc or TdP, and QTc language in the clinical pharmacology section. Results: Against the CredibleMeds classification the means of the margins for Known, Conditional, or Possible Risk of TdP, and Not Listed (presumably no TdP liability) were 4.8, 28, 71 and 339, respectively. Against the US label language the means of margins for black boxes and warnings were 3.1 and 26, respectively. The average margins associated with, positive QTc outcome, negative QTc outcome and no QTc language were 16, 479 and 204, respectively. Based on ROC curves the optimal hERG margin thresholds to separate Known risk of TdP from Not Listed and, QTc prolongation positive from QTc negative were 37- and 50-fold, respectively. Conclusions: The observed optimal margin of 50-fold in the current study is not appreciably different from a previously reported 45-fold optimal margin (Gintant, 2011). The margin falls between the margins for negative (QTc outcome or no QTc language) and positive (positive QTc outcome, warnings or black boxes) compounds. The observed optimal margin of 37-fold in the current study is not appreciably different from the commonly used 30-fold optimal margin (Webster et al., 2002). This margin falls between those for drugs with a known or conditional TdP risk and those where it is at best a possible risk, and from the 240 drugs not listed on www. CredibleMeds.com. It is expected that there would be a small numerical difference (e.g. 37 vs. 50, or as previously published 30 vs. 45) between optimal cut-offs for the TdP liability and QTc prolongation predictions since some QTc positive drugs are described on CredibleMeds.com as having only a Possible Risk of TdP as they are not associated with TdP when used as directed. The fact that the margins in each category form distributions is also expected given biologic variability. However, we argue that a more consistent manner of assessing hERG potency and evaluating relevant exposures would be likely to reduce the spread in these distributions and make margins even more useful as a decision-making data point.