Elodie Valade 1, Martva Valle 1
1 Parexel International (, Spain)
Introduction: The International Council for Harmonisation (ICH) guideline allows for modeling of concentration-QTc (C-QTc) data from Phase I dose-escalation studies to be used as primary analysis for QT prolongation risk assessment of new drugs.[1] The White Paper on C-QTc recommends to detail the procedure for handling concentration data below the limit of quantification (BLQ) in a modeling analysis plan.[2] However, there is no clear recommendation on how to handle these BLQ concentrations. Therefore, the aim of this work is to assess and compare, through simulations, commonly used methods to handle BLQ concentrations in the context of a C-QTc analysis and evaluate the bias/error associated with each BLQ handling method.
Methods: Two sets of 100 phase I studies with single ascending dose parallel design, including placebo and active doses of 1, 3,10, 30, 100, 200, 400 and 800 mg with a total sample size of 64 subjects (6 active; 2 placebo per dose level) were simulated.[3] A 5 ms effect on the baseline-corrected QTc interval corrected for placebo (ΔΔQTc) at the identified maximum plasma concentration (Cmax) of interest was simulated. The first set included a low treatment-specific intercept (0.5 ms, slope of 0.00238 ms per ng/mL) and the second set included a high treatment-specific intercept (2.5 ms, slope of 0.00132 ms per ng/mL). For each set, 6 methods to handle BLQ concentrations were created: BLQ1: exclude all BLQ concentrations (pre- and post-dose); BLQ2: include all BLQ as 0 (pre- and post-dose); BLQ3: include all BLQ as LLOQ/2 (pre- and post-dose); BLQ4: exclude BLQ pre-dose and include BLQ post-dose as 0; BLQ5: exclude BLQ pre-dose and include BLQ post-dose as LLOQ/2; BLQ6: include BLQ pre-dose as 0 and include BLQ post-dose as LLOQ/2. Two scenarios for the proportion of BLQ were explored either 10% or 40% of the observations post-dose. Relative mean standard error (RMSE) and bias were calculated for each set of simulations, BLQ method and proportion of data BLQ for ΔΔQTc mean effect and upper bound of the 90% confidence interval (CI) at the Cmax of interest.
Results: A total of 624 matching plasma concentrations and baseline-corrected QTc interval (ΔQTc) were obtained for each simulation scenario. The RMSE and bias for ΔΔQTc mean effect were higher for BLQ1 among all simulation scenarios. These metrics for BLQ1 were slightly higher when 40% BLQ post-dose compared to 10% BLQ post-dose. For high treatment-specific intercept scenario, RMSE and absolute bias were 88% and 29% for 10% BLQ; and 190% and 51% for 40% BLQ, respectively.
For BLQ4 and BLQ5, RMSE and bias for ΔΔQTc mean effect were similar and comparable or lower than for BLQ1, with RMSE 96-97% and absolute bias 31% for high treatment-specific intercept. These metrics were not affected by the proportion of BLQ post-dose.
Lowest RMSE and bias for ΔΔQTc mean effect were obtained for BLQ2, BLQ3 and BLQ6 with comparable results between these 3 methodologies. These metrics were slightly higher when 40% BLQ post-dose compared to 10% BLQ post-dose. RMSE and bias for ΔΔQTc mean effect were <1% when 10%BLQ whereas RMSE was <4.7% and absolute bias <1.3% for 40% BLQ. BLQ6 had the lowest RMSE and absolute bias among simulated scenarios: for 10% BLQ, low treatment-specific intercept RMSE and absolute bias were 0.03% and 0.01%, respectively and were 0.2% and 0.04% for high treatment-specific intercept. For 40% BLQ, low treatment-specific intercept RMSE and absolute bias were 0.9% and 0.4%, respectively and were 2.2% and 0.6% for high treatment-specific intercept.
Similar trends for RMSE and absolute bias on upper bound of 90% CI were observed, with more precision obtained for BLQ6.
Conclusions: Among the 6 tested methods to handle BLQ concentration in the context of a C-QTc analysis, the bias/error was more pronounced for BLQ1 (excluding all BLQ) and was the lowest for BLQ6 (include BLQ pre-dose as 0 and include BLQ post-dose as LLOQ/2).
References:
[1] International Council for Harmonisation. E 14 The Clinical Evaluation of QT/QTc Interval Prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs. https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e-14-clinical-evaluation-qtqts-interval-prolongation-and-proarrhythmic-potential-non-antiarrhythmic-drugs-step-5_en.pdf
[2] Garnett C, Bonate PL, Dang Q, Ferber G, Huang D, Liu J, Mehrotra D, Riley S, Sager P, Tornoe C, Wang Y. Scientific white paper on concentration-QTc modeling. J Pharmacokinet Pharmacodyn. 2018 Jun;45(3):383-397. Erratum in: J Pharmacokinet Pharmacodyn. 2018 Jun;45(3):399.
[3] Tsamandouras N, Duvvuri S, Riley S. Impact of Phase 1 study design on estimation of QT interval prolongation risk using exposure-response analysis. J Pharmacokinet Pharmacodyn. 2019 Dec;46(6):605-616.
Reference: PAGE 34 (2026) Abstr 12245 [www.page-meeting.org/?abstract=12245]
Poster: Methodology - Other topics