Predicting thorough QT safety profile using hERG IC₁₀ versus IC₅₀

The human ether-à-go-go-related gene (hERG) K⁺ channel is the principal off-target mediator of drug-induced QT prolongation, and hERG IC₅₀ remains the only required in vitro cardiac safety assay. However, clinically relevant QT prolongation (ΔΔQTc ≥10 ms) typically occurs at much lower hERG inhibition (5-20%), and requires that IC₅₀ be transformed into a safety margin to assess risk at therapeutic exposures. Thus, we examined the relationship between hERG inhibition and clinical ΔΔQTc to determine whether inhibition metrics closer to clinical concentrations improve prediction. Across 44 drug/dose pairs, 13.1% hERG inhibition corresponded to mean 10-ms ΔΔQTc, while 10.1% inhibition elicited 10-ms ΔΔQTc at the upper confidence interval (UCI). As a predictor, raw IC₁₀ demonstrated comparable sensitivity (mean ΔΔQTc: 95.5% vs. 100%; UCI: 85.7% vs. 92.9%) and specificity (mean: 77.3% vs. 76.2%; UCI: 87.5% vs. 53.3%) to FDA safety margins calculated from IC₅₀. Logistic regression using IC₁₀ improved discrimination of ΔΔQTc UCI ≥ 10 ms over IC₅₀ (AUC 0.95 vs. 0.90, p = 0.076). In contrast, performance for mean ΔΔQTc ≥10 ms was similar (AUC ∼0.96 for both). The improvement in ΔΔQTc UCI prediction by IC₁₀ was explained by differences in Hill slope between drugs. Those with shallower Hill slope were more readily predicted unsafe by IC₁₀ and those with steeper Hill slope were more readily predicted safe. These findings suggest that hERG IC₁₀ provides a direct index of QT liability, which takes advantage of differences in Hill slope to improve prediction of ΔΔQTc liabilities near critical hERG inhibition of ∼10%.