The study protocol was approved by the Institutional Review Board of the Academic Medical Center, University of Amsterdam, and was conducted according to the principles of the Declaration of Helsinki. All participants gave written informed consent.
The study population consisted of participants of the A
enetics in the NE
(AGNES) case-control study which has been described in detail previously 
. In brief, the AGNES study enrols patients with a first acute STEMI. Cases are defined as patients who were successfully resuscitated after documented VF that occurred between the onset of symptoms and coronary intervention. Controls are defined as acute STEMI patients who did not develop VF. Patients with a previous MI or major co-morbidity were excluded. All individuals studied were of self-declared European ancestry.
Measurement of ECG indices
The first recorded ECG that was acquired during STEMI and before reperfusion treatment was used for analysis. For cases, both ECGs acquired pre-VF as well as post-VF were included. ECGs of insufficient quality (due to e.g. baseline drift or missing leads) and ECGs with rhythms other than sinus rhythm or atrial fibrillation (AF) were excluded. Patients with AF (n
23) were excluded for the analyses of PR interval, but included for the other ECG indices. All ECGs were digitized at 400 dpi (giving a spatial resolution 0.064 mm or 1.6 ms / pixel on an ECG traced at 25 mm / sec). Calibrated measurements were performed on-screen after 4 times enlargement of the digitized ECGs in ImageJ (National Institutes of Health, Bethesda, Maryland, http://rsb.info.nih.gov/ij/
). RR-interval (heart rate) was measured from the ECG. PR interval was measured from the onset of the P-wave to the onset of ventricular depolarization. QRS duration was measured from the onset of ventricular depolarization to the J point. QTc interval was measured using the tangent method 
. Lead II was used when the T wave was of sufficient amplitude to warrant QT measurement, otherwise leads V5 or V2 were used. QT was corrected for heart rate (QTc) using Fridericia's formula (QTc
QT / (cube root (RR)). ST deviation was calculated as the sum of all ST deviation from baseline at 60 ms after the J point in all 12 leads. ST deviation was not reported if individual leads were disconnected or in patients with left bundle branch block. The mean of three consecutive beats wherever possible was measured for all parameters and used in subsequent analyses. Patients with AV block, PR interval ≥ 200 ms or QRS duration ≥ 120 ms were excluded from the ECG analyses involving the continuous endpoints PR interval, QRS duration and QTc interval. For PR interval and QRS duration, additional dichotomous endpoints were assessed i.e. PR interval ≥ 200 ms and QRS duration ≥ 120 ms. Patients with AV block were included in the dichotomized PR interval endpoint as PR > 200 ms.
Selection of SNPs and Genotyping
We inspected the published GWAS concerning heart rate and ECG indices of conduction and repolarization and identified SNPs reported to be associated with these parameters at genome-wide significant P
. In case of high linkage disequilibrium (r2≥0.75) between identified SNPs, only a single SNP, capturing the maximum amount of variation present in the correlated SNPs, was selected for our analysis. This was the case for SNPs at the SCN10A
, CDKN1A, GJA1
loci. A total of 65 SNPs were identified in this way. Genotypes for these SNPs were either obtained by direct genotyping (Illumina610 Quad genotyping array) or were estimated by imputation using HapMap build 36 as the reference panel. Details on genotyping and imputation in the AGNES sample have been described previously 
We tested differences between cases and controls in continuous variables with an independent sample t-test, or the Mann-Whitney U test when the data was not normally distributed. We compared differences in the percentages of categorical variables with the Pearson χ2
test. We used linear regression modelling in association analyses of continuous endpoints and logistic regression modelling for association analyses of dichotomous endpoints (VF, PR≥200 ms and QRS≥120 ms). In all models, we assumed an additive genetic model and corrected for age, sex and the culprit artery (harbouring the occluding lesion) 
as well as the possible interaction between culprit artery and the SNPs. The occurrence of the latter was first tested using a Wald test.
Our analysis plan had a two-stage design. In the first stage, we tested for association of the selected SNPs with the corresponding ECG parameter during STEMI. In the second stage, we selected those SNPs with a (suggestive) significant effect on the ECG parameter and analyzed their effect on the occurrence of VF.
The Bonferroni thresholds for statistical significance were P≤0.0002 for the first stage (corrected for two tests per SNP and six outcomes: HR, QTc, PQ, PQ≥200, QRS and QRS≥120, resulting in a total of 210 tests) and P≤(0.05/number of SNPs carried over from stage 1) for the second stage. For both stages, P-values between the Bonferroni threshold and 0.05 were considered as a suggestive trend.
Power and detectable effects
Given the observed standard deviations in our study population for heart rate and the ECG indices, the reported effects of the identified SNPs would result in effect sizes ≤ 0.2 SD. With the present range in sample sizes (417 – 515), the power to detect an effect size of 0.2 of a SNP with a minor allele frequency (MAF) ranging from 0.05 to 0.5 would range from 1 to 31% for a two-sided p-value of 0.0002 (Bonferroni threshold) and from 24 – 90% for a two-sided p-value of 0.05 (suggestive trend). The present study, therefore, lacks the power to significantly detect effect sizes as found in the general population (from the GWAs). However, given the fact that our heart rate and ECG indices were measured in patients experiencing a STEMI, we hypothesized that the SNP effects might be markedly increased in this sensitized population.
Given the present range in sample sizes, the study had 80% power to detect effect sizes of 0.7 to 0.3 (± 5% explained variance) for the quantitative traits at MAFs ranging from 0.05 to 0.5 assuming an additive genetic model and a two-sided significance threshold of 0.0002 (Stage 1). For PR interval with an overall standard deviation of 20 ms, for example, this translates into detectable allele effects (beta) ranging from 14 to 6 ms. For the dichotomized endpoints, we were able to detect odds ratio's ranging from 2.3 to 1.6 (Stage 1, α