The Dallas Heart Study (DHS) is a multi-ethnic probability-based, population study 
, with 1,506 non-Hispanic Blacks, 942 non-Hispanic Whites, and 501 Hispanics available for the analysis presented here. Eight SNPs chosen to tag the linkage disequilibrium (LD) block previously associated with QT interval (based on LD in the HapMap CEU samples), including the SNP associated with SCD 
, were genotyped in all individuals. Thirteen additional individuals were excluded due to excessive missing genotype data (<50% complete); all SNPs showed minor missing data (≥99% complete) with no strong deviation from Hardy-Weinberg equilibrium (P<0.005) in any of the populations. The non-Hispanic White and Hispanic participants had similar patterns of LD and similar allele frequencies, whereas the non-Hispanic Black participants demonstrated significantly less LD, with complementary minor alleles for 6 of 8 SNPs (, ).
Comparison of linkage disequilibrium (LD) patterns across ethnic groups.
Summary of results for association of NOS1AP SNPs with QT interval stratified by ethnicity.
The QT interval is highly correlated with heart rate (RR interval), age, and sex at the population level, and thus we performed multiple linear regression analyses separately in each ethnic population to adjust for these factors, and analyzed the residual QT under an additive genetic model (). In non-Hispanic Whites, the results mirrored our previous findings, with the strongest association observed for rs16847548, with each allele associated with an increase in the QT interval of 2.57 ms (P
0.0004). While no nominally significant SNPs were identified in Hispanics, all SNPs exhibited the same direction of effect as in non-Hispanic Whites. Strikingly, and in contrast to previous reports, rs16847548 was strongly associated with QT interval in non-Hispanic Blacks, with each allele associated with a prolongation of the QT interval by 3.22 ms (P
). Another novel finding was that rs16856785, which is uncorrelated with rs16847548 (r2
<0.01), was also associated with QT interval in non-Hispanic Blacks (+1.60 ms, P
0.01). While not significant in Hispanics or non-Hispanic Whites, the direction of effect observed for rs16856785 was the same (+0.58 ms and +0.95 ms, respectively). To formally test for independent effects of these two SNPs, we performed a forward stepwise regression in each ethnic group, first including the most strongly associated SNP and then sequentially adding SNPs in order of strength of association, only retaining them in the model if the P-values were <0.05. Only rs16847548 was retained in Hispanics and non-Hispanic Whites, but both rs16847548 and rs16856785 were significant in non-Hispanic Blacks indicating that these 2 SNPs independently influence QT interval in non-Hispanic Blacks (). To determine whether population stratification was influencing our results, we adjusted for both global and local ancestries as inferred by 2,270 ancestry informative markers using ANCESTRYMAP 
, and our findings were unchanged. Comparing non-Hispanic Blacks homozygous at both SNPs for the QT lengthening allele to non-Hispanic Blacks homozygous for the complementary alleles revealed a 13.9±4.5 ms difference in QT interval.
Demonstration of independent effects for rs16847548 and rs16856785 on QT interval stratified by ethnicity.
When stratified by sex, we observed stronger effects for NOS1AP
variants in women (), with a consistent effect across all 3 ethnic groups, and an average addition of +2.18 ms in the effect estimate relative to men for rs16847548. To increase our power to test for a potential sex interaction, we leveraged the similarity in effect sizes and allele frequencies for rs16847548 across all three ethnic groups, and performed a joint analysis while adjusting for ethnicity. Across all samples, we observed genome-wide significance for rs16847548 alone (P
), while incorporating sex and an interaction term for rs16847548 and sex into the regression model indicated a statistically significant stronger effect in women (one-sided P
0.027 for the interaction term). No significant interaction was observed for rs16856785.
Sex-stratified analysis of NOS1AP SNPs for association with QT interval.
To confirm both the sex interaction and independence of the effects of rs1684758 and rs16856785, we reanalyzed data from the combined Atherosclerotic Risk in Communities (ARIC) 
and Cardiovascular Health Study (CHS) 
cohorts. QT interval was corrected for age, sex, RR interval, and study using linear regression. We confined our analyses to the white individuals (N
14,107), as no association between NOS1AP
variants and QT interval was previously observed in the black individuals in those samples 
. Incorporating both rs16847548 and rs16856785 into a linear regression model revealed that both were significantly associated with QT interval (P<2×10−16
, respectively), with similar effect sizes as observed for the DHS (+2.22 ms and +1.74 ms, respectively), confirming that these two SNPs are independently associated with QT interval (). The interaction between rs16847548 and sex in Whites was also highly significant (+2.00 ms, P
for the interaction term), whereas no sex interaction was observed for rs16856785 (P