Clinical characteristics of ARIC and CHS participants at the baseline examinations
Baseline demographic and cardiovascular risk factors are shown by SCD status and by self-reported race in . As expected, many of the well established cardiovascular risk factors were significantly associated with SCD risk. Among both whites and blacks, individuals who died from SCD were significantly older, had higher systolic blood pressure, and fibrinogen and lower HDL cholesterol. They were also more likely to be male, smokers, less well educated, and have a history of diabetes, hypertension, and myocardial infarction at the baseline examination. Individuals who experienced SCD had significantly longer mean QT intervals at baseline, prior to the event. In whites, the age-, sex-, heart-rate, and study-adjusted mean QT duration was 403 ms for those who did not ultimately have SCD and 411 for those who did (P<0.001); and the corresponding values were 402 and 410 ms, respectively, in blacks (P<0.001).
Baseline characteristics of 14,737 white and 4,558 black ARIC and CHS participants by SCD status and race
While the two studies are largely comparable, there were modest differences in the significance and magnitude of associations between cardiovascular risk factors and SCD risk between the two studies. For example, smoking, BMI, and total cholesterol were not significantly associated with SCD in CHS but were in ARIC ().
Baseline characteristics of 14,309 ARIC and 4,986 CHS participants by SCD status
Association between QT interval and SCD
Over a median follow up of 14.1 years in ARIC and 12.2 years in CHS, 334 whites and 164 blacks experienced sudden cardiac death (cumulative incidence rate per 1,000 person-years: 2.0 overall; 1.4 in ARIC and 4.5 in CHS; 1.8 in whites and 2.9 in blacks). Given the older age of the CHS cohort, 222 of the 498 SCD events (45%) occurred in CHS while 276 events occurred in ARIC. Of all CHD mortality adjudicated (N = 985), 40.2% were classified as definite SCD and 10.4% as possible SCD. The majority of cases of SCD occurred out of hospital (90%).
Among whites, longer QT interval was associated with the development of SCD after adjust for age, sex, heart rate, and study. Compared to whites in the first quintile of QT interval, those in the 2nd, 3rd, 4th, and 5th quintiles were 1.43 (95% CI 0.97 to 2.13), 1.77 (95% CI 1.16 to 2.69), 2.33 (95% CI 1.49 to 3.64), and 3.58 (95% CI 2.20 to 5.81) times more likely to have suffered SCD, respectively (P for trend <0.0001). Similar dose response relationship was observed when the analysis was repeated using QT deciles (P for trend <0.0001). Among blacks, longer QT interval was also associated with SCD risk; however, the dose response relationship was less apparent, possibly due to the smaller number of cases. Compared to blacks in the first quintile of QT interval, those in the 2nd, 3rd, 4th, and 5th quintiles were about 1.75 (95% CI 1.07 to 2.88), 2.04 (95% CI 1.19 to 3.49), 1.62 (95% CI 0.86 to 3.05), and 2.54 (95% CI 1.34 to 4.82) times more likely to have suffered SCD (P for trend =0.02).
Association between 19 SNPs in NOS1AP and QT interval
Allele frequencies of the 19 SNPs examined are shown in by ethnicity. The LD pattern of these 19 SNPs were quite different in blacks as compared with whites, with considerably less LD among SNPs in blacks (). Eleven of 19 SNPs examined were significantly associated with age-, sex-, and heart rate-adjusted QT interval in whites with P≤0.003. However, no SNPs were significantly associated with QT interval in blacks.
Genotypic association between 19 SNPs in NOS1AP and QT interval and SCD by self-reported race
Figure 1 Plots showing the linkage disequilibrium (LD) pattern and association results for both QT interval and SCD in whites (A) and blacks (B) for 19 SNPs genotyped to tag the NOS1AP locus and surrounding region that exhibited the strongest association with (more ...)
In whites, the most significant SNP in the present study was rs16847548 (P
=2.2 × 10−18
), which is in LD with rs4657139 but was not typed in previous studies11–15, 17
. The frequency of the C allele of rs16847548 was 0.22 in whites. After adjusting for age, sex, and heart rate, the mean QT interval of individuals with TT, TC, and CC genotypes at rs16847548 were 399, 401, and 403 ms respectively in ARIC; 411, 414, and 416 ms respectively in CHS; and 402, 404, and 407 ms respectively in the combined dataset. This effect size of approximately 5 ms difference between the two homozygous groups is consistent with our previous observations. In whites, the percent variation (R2
) in the QT interval distribution (uncorrected QT interval) that was explained by rs16847548 was 0.2% in both the individual studies and the combined dataset. In comparison, the R2
associated with other variables was: 0.5% for age, 0.1% for sex, 0.2% for diabetes, 0.4% for history of MI at baseline, and 67% for heart rate in white ARIC participants. Among white CHS participants, the R2
associated with age, sex, diabetes, history of MI at baseline, and heart rate were 0.03%, 1.1%, 0.4%, 0.5%, and 63%, respectively.
Associations between NOS1AP genotypes and SCD
Consistent with the observation of longer mean QT interval associated with the C allele of rs16847548, this allele was also significantly associated with increased risk of SCD in whites. Indeed, only 179 of the 8,905 individuals (2%) carrying the TT genotype at rs16847548 suffered from SCD (), whereas, 25 of the 706 (3.5%) individuals with the CC genotype experienced SCD. In whites, the crude relative hazards that were estimated using a co-dominant model suggested a dose-response relationship between copies of the C allele at rs16847548 and SCD. Using an additive model, the age-, sex-, and study-adjusted RH for each C allele was 1.31, 95% CI 1.10 to 1.56; P=0.002 ().
Unadjusted and adjusted relative hazard (RH) of SCD by rs16847548 and rs12567209 genotypes in whites from ARIC and CHS
In addition, a downstream neighboring SNP, rs12567209, not correlated with rs16847548 (r2 = 0.02), was also associated with SCD in whites (age-, sex-, and study-adjusted RH for each A allele =0.57 assuming an additive model, 95% CI 0.39 to 0.83; P=0.003). Due to the low frequency of the A allele (MAF=0.07), a dominant model was also used for the analysis of rs12567209. The age-, sex-, and study-adjusted relative hazard of SCD comparing those with at least one copy of the A allele to those with the GG genotype was 0.53 (95% CI 0.36 to 0.79; P=0.002), thus both the additive and the dominant models were consistent with the data. The present study is not able to distinguish whether one model was a better fit than the other (additive model shown in ). Surprisingly, rs12567209 was only modestly associated with QT interval (), suggesting that the effect on risk for SCD was not necessarily conveyed through modulation of QT interval. The mean age-, sex-, heart-rate, and study-adjusted QT interval for GG, AG, and AA genotypes were 403, 403, and 401 ms respectively (P for additive =0.05; P for dominant model =0.08).
To demonstrate the independent effect on risk of SCD for rs16847548 and rs12567209, we included both SNPs in the same model and found that both SNPs remained associated with SCD. Moreover, there was no significant interaction between these two SNPs (P for interaction=0.68 adjusted for age, sex, and study). Assuming an additive model for both SNPs and after adjusting for age, sex, and study, the RH for each copy of the C allele of rs16847548 was 1.27 (95% CI 1.06 to 1.51; P=0.008) and 0.62 (95% CI 0.42 to 0.90; P=0.012) for each copy of the A allele of rs12567209 (). The associations between both SNPs and SCD risk were relatively consistent across the two studies although the study-specific p values did not reach the Bonferroni-corrected significance level due to their smaller sample sizes. The corresponding RH for rs16847548 was 1.32 (95% CI 1.01 to 1.70; P=0.04) for ARIC and 1.24 (95% CI 0.98 to 1.57; P=0.08) for CHS. The corresponding RH for rs12567209 was 0.84 (95% CI 0.52 to 1.39; P=0.51) for ARIC and 0.44 (95% CI 0.25 to 0.79; P=0.006).
No SNPs were significantly associated with SCD in blacks at the α=0.003 level (). The RH for each C allele of rs16847548 was 1.07 (95% CI 0.81 to 1.41; P
=0.64) after adjusting for age, sex, and study. On the other hand, the RH for each A allele of rs12567209 was 1.40 (opposite direction as the association in whites; 95% CI 0.97 to 2.03; P
=0.07) after adjusting for age, sex, and study (Supplementary Table 1
Multivariate genotype association analyses with SCD
To explore whether the effect of NOS1AP SNPs on the risk of SCD is entirely mediated through modulation of the QT interval, we added both QT interval and heart rate as variables into the Cox proportional hazards model in whites, after adjusting for age, sex, and study. The relative hazard of SCD associated with each additional copy of the C allele at rs16847548 decreased from 1.27 (model 2 in ) to 1.22 (model 3 in ), and the relative hazard for each additional A allele at rs12567209 changed from 0.62 to 0.63. Further adjustment for existing cardiovascular risk factors that were associated with SCD modestly attenuated the significance of the associations for rs16847548 (RH=1.17; 95% CI 0.97 to 1.42).
Additional analyses were performed to examine the impact of either co-morbidities or other SNPs on the robustness of the association between rs16847548 and rs12567209 and SCD risk in whites. First, exclusion of whites with a previous history of MI strengthened both associations (for each C allele of rs16847548 RH=1.39 for model 1, 95% CI 1.13 to 1.69; P
=0.002; for each A allele of rs12567209 RH=0.49 for model 1; 95% CI 0.30 to 0.78; P
=0.003). Second, exclusion of 623 individuals with electrocardiographic QRS complex >120 ms, which is indicative of a bundle branch block or other conduction defect, also resulted in a stronger association between rs16847548 and SCD risk (RH=1.40 for model 1; 95% CI 1.16 to 1.68; P
<0.001). However, when QRS duration was included in the fully adjusted model (model 3), the relative hazard changed minimally from 1.23 to 1.25 (95% CI 1.04 to 1.51). The age-, sex-, and study-adjusted RH for rs12567209 changed from 0.57 to 0.62 (95% CI 0.42 to 0.92; P
=0.02) upon exclusion of QRS complex >120 ms. Finally, among whites, none of the tests of interaction between either SNP, SCD, and known cardiovascular risk factors (study, history of MI, sex, diabetes, age at last follow up, diabetes, hypertension, family history of CVD, obesity, dyslipdemia, and smoking) was statistically significant (Supplemental Tables S2 and S3
Lack of association between rs16847548 and rs12567209 and non-sudden cardiac mortality in whites
It is possible that the association between rs16847548 and/or rs12567209 with SCD is due to an association with overall CHD mortality. Therefore, survival analyses were also conducted for CHD mortality that was not coded as SCD (non-SCD CHD mortality) and all other mortality that were neither SCD nor CHD (non-SCD & non-CHD) mortality. shows the cumulative incidences of SCD, non-SCD CHD, and non-SCD & non-CHD mortality, while accounting for each other as competing case of death, by rs16847548 or rs12567209 in whites. The cumulative incidence of SCD per 1,000 person-years of whites with the TT, TC, and CC genotypes at rs16847548 were 1.5, 2.0, and 2.8, respectively (). On the other hand, rs16847548 was not associated with non- sudden CHD mortality (age-, sex-, and study-adjusted RH=0.98, 95% CI 0.83 to 1.17; P=0.86; ) nor with non-SCD & non-CHD mortality (age-, sex-, and study-adjusted RH=1.00, 95% CI 0.94 to 1.07, P=0.94; ). For rs12567209, the cumulative incidence of SCD per 1,000 person-years of whites with the GG, AG, and AA genotypes were 1.9, 0.9, and 2.2, respectively (). As for rs16847548, no association was observed with non- sudden CHD mortality (age-, sex-, and study-adjusted RH=0.83, 95% CI 0.62 to 1.11); P=0.21; ) nor with non-SCD & non-CHD mortality (age-, sex-, and study-adjusted RH=99 95% CI 0.89 to 1.10, P=0.86; ).
Figure 2 Cumulative incidence curves, accounting for competing cause of death, in whites. Kaplan-Meier survival curves in whites by rs16847548 (A,C,E) and by rs1267209 (B,D,F). (A,B) SCD, (C,D) non-SCD CHD mortality, and (E,F) non-SCD & non-CHD mortality. (more ...)