Concentration‐QT modelling in early clinical oncology settings: Simulation evaluation of performance

AIMS: Concentration-QT modelling (C-QTc) of first-in-human data has been rapidly adopted as the primary evaluation of QTc interval prolongation risk. Here, we evaluate the performance of C-QTc in early oncology settings (i.e., patients, no placebo or supratherapeutic dose, 3 + 3 designs). METHODS: C-QTc performance was evaluated across three oncology scenarios using a simulation-estimation approach: (scen1) typical dose-escalation testing six dose levels (n = 21); (scen2) small dose-escalation testing two dose levels (n = 9); (scen3) expansion cohorts at one dose level (n = 6-140). True ΔΔQTc effects ranged from 3 ms ("no effect") to 20 ms ("large effect"). Performance was assessed based on the upper limit of the ΔQTc two-sided 90% CI against a threshold of 10 or 20 ms. RESULTS: The performance against the 10 ms threshold was limited based on C-QTc data from typical dose escalation (scen1) and acceptable performance was observed only for relatively large expansions (n ≥ 45; scen3). Performance against the 20 ms threshold was acceptable based on C-QTc data from a typical dose escalation (scen1) or dose expansion cohort n > 10 (scen3). In general, pooling C-QTc data from dose escalation and expansion cohorts substantially improved the performance and reduced the ΔQTc 90% CI width. CONCLUSION: C-QTc performance appeared limited using a 10 ms threshold, but acceptable against a 20 ms threshold. Selection of threshold may be informed by the benefit-risk balance in a specific disease area. Acceptable precision (i.e., confidence intervals) of the estimated ΔQTc, regardless of its magnitude, can be facilitated by pooling data from dose escalation and expansion cohorts.