Influence of Meals and Glycemic Changes on QT Interval Dynamics

Abstract

Thorough QT/QTc studies have become an integral part of early drug development programs, with major clinical and regulatory implications. This analysis expands on existing pharmacodynamic models of QT interval analysis by incorporating the influence of glycemic changes on the QT interval in a semimechanistic manner. A total of 21 healthy subjects enrolled in an open-label phase 1 pilot study and provided continuous electrocardiogram monitoring and plasma glucose and insulin concentrations associated with a 24-hour baseline assessment. The data revealed a transient decrease in QTc, with peak suppression occurring approximately 3 hours after the meal. A semimechanistic modeling approach was applied to evaluate temporal delays between meals and subsequent changes that might influence QT measurements. The food effect was incorporated into a model of heart rate dynamics, and additional delayed effects of the meal on QT were incorporated using a glucose-dependent hypothetical transit compartment. The final model helps to provide a foundation for the future design and analysis of QT studies that may be confounded by meals. This study has significant implications for QT study assessment following a meal or when a cohort is receiving a medication that influences postprandial glucose concentrations.

Keywords: QT interval; glucose; heart rate; mathematical modeling.

Figure 1

Model diagram of food and glycemic effects on the QT interval. Symbols and model equations are defined in Pharmacodynamic Modeling in the Methods section.

Figure 2

Temporal profiles of median pharmacodynamic end points. Column labels (cohorts A, B, and C) indicate the study group. Symbols indicate data points from the measured response end points: glucose (solid squares), insulin (open diamonds), HR (open circles), QT (open stars), and QTcF (solid triangles). Each row is served by dual y axes to facilitate comparison of data across study groups. The breakfast meal was provided at 7 am for cohort A and at 6 am for cohorts B and C. HR, heart rate; QTcF, QT corrected by Fridericia's method.

Figure 3

Comparison of observed measurements with population and individual fitted curves for 2 representative subjects. Column labels (glucose, HR, and QT) indicate the measured pharmacodynamic end point. Row labels indicate both the time window of interest (breakfast window [top 2 rows], full 24‐hour [bottom 2 rows]) and the identifier for the 2 representative subjects (A [rows 1 and 3], B [rows 2 and 4]). Observed measurements are represented by open circles. Population‐fitted curves are represented by solid lines, and individual‐fitted curves are represented by dashed lines. HR, heart rate.

Figure 4

VPC for the breakfast window and full 24‐hour models. Column labels (glucose, HR, and QT) indicate the measured response variable. The top row shows the VPCs for the breakfast window period, and the bottom row corresponds with the full 24‐hour period. Symbols represent individual observed data, and dashed lines are the 5th, 50th, and 95th percentiles of the observed data. Solid lines are medians of the simulations, and shaded areas define the 5th–95th percentiles of 1000 simulations. VPC, visual predictive check.

Figure 5

Model‐predicted mean (SE) profiles of (A) heart rate, (B) uncorrected QT for a 24‐hour day, (C) change in heart rate and glucose (QT) for the breakfast window, and (D) change in uncorrected and corrected QT. For A and B, open circles represent fasting conditions, and open triangles represent fed conditions. Symbols and error bars represent the mean and SE of 500 simulated profiles. chGL, change in glucose; mQTcG, heart rate and glucose‐corrected QT; QTc, heart rate–corrected QT; SE, standard error.