A polygenic predictor of baseline QTc is associated with sotalol-induced QT prolongation

Long-QT intervals predispose to the morphologically distinctive ventricular tachycardia, Torsade-de-Pointes (TdP).¹ Exaggerated QT-prolongation can occur after exposure to numerous commonly-used drugs, most of which result in block of the rapid delayed rectifier potassium channel(I_Kr).¹ Following drug exposure, an increase in the QT corrected for heart rate (QTc) to >500 msec or of >60 msec versus baseline has been used as a marker for TdP risk and threshold for liable drug discontinuation (e.g. E14 FDA guidance).¹ However, interindividual QT response to a drug is highly variable, and risk stratification of low- versus high-risk for extreme QT prolongation remains a challenge. Large genome-wide studies have identified sets of single nucleotide variants (SNV) influencing baseline QTc, and polygenic risk scores (PRS) integrating their effects have been developed to capture this influence.² Here, in a prospective European ancestry cohort of almost 1000 subjects, we examine the relationship between a QTc-PRS (PRS for QTc prior to drug) and QTc prolongation in response to a standardized challenge with the I_Kr-blocking drug sotalol.

Using the GENEREPOL³ cohort (NCT00773201, institutional review board approved) of 990 healthy individuals (informed written consent obtained) with a normal baseline ECG who were prospectively challenged with a single oral dose of 80mg sotalol, we measured the QTc change between baseline and 3hours post-dosing (ΔQTc; Fridericia’s correction) as previously described.³ Data will be made available upon reasonable request to the corresponding author, while General Data Protection Regulation (GDPR) compliance need to be guaranteed by the recipient. The plasma concentrations of sotalol 3hours post-dosing and of potassium were also measured. Subjects were genotyped using the MEGA^exarray, with 978 samples (99%) passing standard quality control and 12 excluded for the following: heterozygosity F>0.19(n=1), patient duplication(n=1), missing sample data(n=1), and unconfirmed genetic/clinical record sex match(n=9). We constructed a QTc-PRS consisting of the 465,399 SNV from a previously generated 1,110,495 SNV QTc-PRS² that were present in our imputed genomes at major allele frequency>0.01 and with imputation r²>0.7. We generated a multivariable regression model (ANCOVA, best model approach selected by Bayesian Schwarz information criterion) studying the association of ΔQTc with QTc-PRS, while testing adjustment for sex, age, baseline QTc, plasma potassium, plasma sotalol concentration, and the first ten principal components of genetic ancestry as independent variables. P-value<0.05 was considered significant.

The analyses included 978 healthy subjects (607, 62% female), with median age 23.6years (interquartile range(IQR)=21.0–32.7). Median baseline QTc was 388msec(IQR=376–399). Three hours post-sotalol intake, the median ΔQTc was 20.7msec (IQR=12.0–30.3). Ten subjects (1.02%) had ΔQTc≥60msec (“extreme responders”). None had a post-sotalol QTc above 500msec. In the extreme responder group, the QTc-PRS was significantly greater than in those with a smaller drug response, or “normal responders” (mean z-score=0.75 vs. −0.0077, p=0.008, Mann Whitney test, Figure 1A-B). We found that QTc-PRS(not-adjusted) was able to discriminate normal responders vs. extreme-responders with an area under the receiver operating curve of 0.74 (95% Confidence Interval (CI) 0.62–0.87). In this cohort, a QTc-PRS z-score threshold of −0.44 had a 100% sensitivity and 100% negative predictive value to detect extreme responders, leaving 32.1% (n=314) below this threshold (Figure 1C for the detailed diagnostic properties of the main QTc-PRS thresholds). Furthermore, in the multivariable regression model (adjusted r²=0.35, p<0.0001), ΔQTc was associated with QTc-PRS (β_{standardized-coefficient}=0.095±0.026, p=0.0002) while including sex (β=−0.42±0.026 for male vs. female, p<0.0001), age (β=−0.080±0.026, p=0.002), plasma sotalol level (0.36±0.026, p<0.0001), and plasma potassium (0.082±0.026, p=0.002). In this latter model, the baseline QTc and the first ten principal components of genetic ancestry were not significant and not retained in the final model.

Figure 1:

(A) Interindividual variation in the QTc response to 80 mg of sotalol per os (ΔQTc, QTc 3 hours post dosing vs. baseline) versus the polygenic risk score for the baseline QTc (QTc-PRS, N=978). The QTc-PRS is higher in extreme-responders (ΔQTc ≥60 msec, red, n=10) than in normal responders (blue, n=968) by the two-sided Mann-Whitney test (P=0.008) (B) The observed ΔQTc (msec) with its 95% confidence intervals as a function of the QTc-PRS (z-score) in females and males. Specificity (C), sensitivity (C), negative predictive values (D) in identifying extreme-responders as a function of QTc-PRS (z-score). PRS = polygenic risk score, QTc-PRS = polygenic risk score for the baseline QT interval, corrected for heart rate (QTc).

Statistics : To calculate the QTc-PRS (z-score), for each subject (j), we sum, for each (i) of n risk alleles (n=465,399), the subject’s allele dosage X_ij ( X_ij = 0, 1, or 2) multiplied by that allele’s effect size β_i (as previously defined²). The sum is normalized to the mean (μ) and standard deviation (σ) of the pre-normalization QTc-PRS distribution of the cohort, such that a QTc-PRS of 0 equals the cohort mean, and the QTc-PRS unit equals 1 cohort standard deviation.

Our results show that the common genetic variants associated with QTc prior to drug exposure also capture part of the repolarization response (ΔQTc) dynamics following exposure to an I_Kr blocking-drug in a cohort of approximatively a thousand participants, with ECG acquisition and QTc measurement quality standards similar to those applied in thorough QT-studies, which are usually performed on approximately 20–40 patients. This finding allows for first estimates of the diagnostic properties of QTc-PRS to identify extreme drug-induced QTc-prolongation (herein, 1% of subjects) and confirms a previous pilot report that a simplified QTc-PRS (61 SNV) was associated with the extent of QT-prolongation after drug challenge in 22 healthy subjects, and was also associated with TdP risk in 216 TdP cases and 771 drug-exposed controls.⁴ The idea that a baseline PRS predicts drug response has also been reported for a baseline PRS for coronary artery disease and outcomes in PCSK9 inhibitor therapy, and a baseline PRS for depression and response to antidepressants.⁵ Thus, the data we report not only define how QTc-PRS predicts ΔQTc but also add to the emerging concept that the genetic architecture of common disease or traits can be used to predict drug response. Additionally, the potential of a very high QTc-PRS negative predictive value to rule-out extreme responders in this proof of concept work is promising but few patients had extreme QTc prolongation and our findings deserves further validation to confirm that QTc-PRS could be used to avoid QTc monitoring in a predicted very low risk population.

Abbreviations:

FDA

Food and Drug administration

I_Kr

rapid delayed rectifier potassium channel

IQR

inter-quartile range

PRS

polygenic risk scores

QTc

QT corrected for heart rate

SNV

single nucleotide variants

TdP

Torsade-de-Pointes