Verena Gotta (1), Sara Bachmann (2), Marc Pfister (1), Birgit Donner (3)
(1) Pediatric Pharmacology and Pharmacometrics, (2) Pediatric Endocrinology and Diabetology, (3) Pediatric Cardiology, University of Basel Children's Hospital, Basel, Switzerland
Objectives: The heart-rate corrected QT (QTc) interval is subject to significant inter- and intra-individual variability. QTc lengthening is associated with increased risk of ventricular arrhythmia and sudden cardiac death. There are several known physiologic factors explaining QTc variability. If they are not correctly accounted for, quantification of pathologic or drug-induced QTc lengthening can be biased, the risk of false-positive associations increased, and sensitivity to detect small effects decreased. Pharmacometric modeling has been advocated as a tool to account for complex sources of variability of QTc-prolongation and increase confidence in quantitative associations. In children with type 1 diabetes (T1D) an association between nocturnal hypoglycemia (<3.7mmol/l) and QTc lengthening (QTcB: heart-rate correction using Bazett equation) of 7 ms compared to euglycemia has previously been observed[1]. The objective of this extended model-based analysis was to increase the quantitative understanding of this association, and to quantify other sources of QTc variability in the pediatric diabetic population.
Methods: Data originated from a prospective observational study in 25 cardiac healthy children with T1D in the home setting (age range: 8.1-17.6 years, duration of T1D: 0.5-13.3 years)[1]. In each patient continuous subcutaneous glucose and Holter electrocardiogram measurements were obtained for 5 consecutive nights, averaged over 5 minutes for analysis. Linear and non-linear mixed effect modelling was used to compare Bazett with individual heart-rate correction (QTcB/QTcI) and to account for different levels of random QTc variability (between-subject, inter-night, residual intra-individual variability: BSV, INV, RV) (Base model). Subsequently, extended models accounting for circadian variation, age and gender on the QTc interval were developed, followed by investigation of the glucose-QTc relationship (univariable and combined adjusted analysis). Finally, five factors that may modify sensitivity to QTc-lengthening (i.e. steepness of the glucose-QTc relationship) were investigated: duration of diabetes, HbA1c, time in hypoglycemia, glucose variability and pre-study magnesium serum levels (all stratified by median).
Results: Random variability was reduced in the QTcI compared to the QTcB Base model (BSV: ±12.6 vs 14.1 ms, INV: ±7.5 vs 7.8 ms, RV: ±8.0 vs 8.5 ms), and was further reduced in the adjusted covariate model (BSV: ±9.7, INV: ±6.5, RV: ±7.4 ms). This model accounted for significantly (P<0.01) shortened QTc in adolescent boys (-14.6 ms), circadian variation (cosine function: amplitude=19.2 ms, shift=2.9h), and a linear glucose-QTc relationship (strongest association assuming a delayed effect, e.g. through sympatho-adrenergic stimulation: slope: -0.76 ms per 1 mmol/L increase in glucose level, delay rate constant: 0.56-h ≈ equilibration half-life of 1.2h). A linear glucose-QTc relationship was visually and statistically (lower AIC) preferred over alternative relationships, including categorical stratification (estimating 1.75 ms longer QTc in hypo- versus euglycemia; relative standard error, RSE: 27%). Adding BSV and INV on the slope parameter quantified relevant random variability (±1-2 ms per 1 mmol/L), however with the typical mean slope estimate approximating 0 (large RSE >100%), i.e. predicting both positive and negative glucose-QTcI associations, even within the same patient; no significant correlation with patient characteristics were identified. When not including BSV/INV on the slope, significantly different sensitivity was estimated according to HbA1c (>/<7.5%), duration of diabetes (>/<5.2 years) and time in hypoglycaemia (>/<3.6%/nighttime), but not glucose variability (>/<36%) or magnesium levels (>/<0.8 mmol/L).
Conclusions: A clinically mild association of nocturnal hypoglycemia with longer QTc intervals could be confirmed in this model-based analysis. Relevant circadian variation quantified in this study, with longest QTc around 03:00 AM, may contribute to higher risk of sudden unexpected overnight death in the pediatric diabetic population. In analogy to a thorough QT study, conduct of a controlled study design comparing several experimental controlled glucose levels is warranted to further characterize underlying relationships between hypoglycemic events and QTc interval in children with T1D.
References:
[1] Bachmann et al. Pediatr Diabetes (2021) 22(7):1023-1030
Reference: PAGE 30 (2022) Abstr 10078 [www.page-meeting.org/?abstract=10078]
Poster: Drug/Disease Modelling - Paediatrics