The baseline demographics are summarized in , the electrophysiologic features at baseline in , and medications in .
Demographics of Cases and Controls at Time of ICD Implantation.
Electrophysiological characteristics of cases and controls.
Baseline use of key medications.
Cases and controls differed by age, height, weight, location of myocardial infarction, and age of first myocardial infarction (). In addition, there were 704 males and 200 females in the study. Gender specific differences were observed in mean left ventricular ejection fraction (LVEF), weight, height, smoking, and approximate age for the first MI.
The electrophysiologic characteristics at baseline also differed between the groups with respect to incidence of atrioventricular block, left bundle branch block, and a history of spontaneous ventricular arrhythmia (). For the 102 patients with qualifying tachycardias, the mean cycle length was 297 msec (S.D.
56.5). The histogram of average ventricular cycle length for these 102 patients is shown in Figure S1
. Finally, there were also some imbalances in baseline medications, particularly anti-arrhythmic drugs ().
A genome-wide association was performed between individuals that had an LTA (cases) and those that did not (controls) over the course of at least three years (). Gender and genetic background were tested for association with case-control status, were not found to be significant, and were not included in the model. We did not detect any regions associated at P<5×10−8
and report top regions at P<1×10−5
(). The strongest association at rs11856574 (P
), located in the hypothetical gene KIAA0574
(hypothetical protein LOC23359). We detected no evidence of population stratification (genomic inflation factor
1.0035, Figure S2
). On a subset of individuals for which we had length of follow-up, we performed survival analysis GWA (Figure S3
). We did not detect any associations at P<5×10−8
, but do note a region at 13q14.2 (rs2854357 P
7) with a consistent signal across multiple SNPs.
Genome-wide association results of Cases with LTA vs. Controls without a LTA.
Top regions associated at P<10−5.
We looked specifically at P-values for 42 SNPs previously reported to be implicated in prolonged QT duration or SCD 
. None were associated in the study dataset at Bonferroni corrected P-values (P<0.05/42
0.0012). Table S1
shows the P-values from directly genotyped or imputed genotypes for these 42 SNPs, indicating that these SNPs may be of limited prognostic value in identifying individuals likely to have an LTA among those that are candidates for an ICD with prior myocardial infarction.
We calculated our power to detect effects at an alpha of 10−7
for the sample sizes observed here. We were well powered to detect effects of large effect size (OR>3.78) in common variants (minor allele frequency (MAF)>0.05) given our sample size (Figure S4
). For very common variants (MAF>0.25), we had 80% power to detect effects of OR>2.15. Thus, it is unlikely that there is a common variant with a large effect strongly associated with LTA in individuals of European ancestry. Using genome-wide genotyping chips such as the Ilumina Human660 BeadChip, there is good coverage of common variation in European ancestry populations 
, however not all regions of the genome, in particular regions of low linkage disequilibrium, may be covered well.
We conducted a copy number variation analysis, using QuantiSNP and PennCNV to estimate copy number variation in each individual. CNV regions covering at least 10 probes and 100 kb from both methods were combined and tested for association with case/control status in PLINK. There were no regions that were associated at a multiple testing corrected P-value less than 0.05 when all CNVs were combined. However, one region on chromosome 16 (position 33,395,681–33,506,617) was associated (minimum P
0.0097) when deletions alone were tested (Figure S5
). The region is flanked by a target of p53 (TP53TG3
) and a creatine transporter (SLC6A8
). As this region is a duplication of a region on chromosome X 
, we tested whether having a CNV in this region was associated with gender and saw no association. These CNVs require validation using quantitative PCR and the association will require verification in a replication study.