Genotyping Factors
The business off lymphoblastoid telephone traces, quality-control away from genomic DNA, purchase of hereditary studies, and you may genotyping quality control metrics was did centered on basic measures. Delight comprehend the on line-simply Research Supplement for those information.
Consensus unmarried-nucleotide polymorphisms (SNPs) that enacted quality control both in stages (genome-wider relationship and loved ones-dependent phase) was in fact combined for everybody available sibships (2239 SNPs was in fact imputed on probands). Sibships was basically confirmed whenever pairwise pi_hat thinking were anywhere between 0.thirty-five and you can 0.65; trials had been taken out of good sibship in the event that siti incontri gratis scambisti estimated pi_hat value wasn’t in this variety. That it dataset of the combined genotyping phases means the last dataset for all then described analyses. New disperse away from people in the research is actually revealed in the Figure step 1.
Genetic Analysis Data
All family-based analyses were conducted with PLINK 1.07 software. 8 The dFam utility within PLINK implements a siblings-based transmission-disequilibrium test and was used to conduct these analyses. The dFam option is a powerful test for sibling-only datasets, incorporating data across sibships as well as using data from estimated parental genotypes to calculate expected allele frequencies for comparison with observed allele frequencies. The association test is based on the Cochran-Mantel-Haenszel test. Bonferroni correction for the number of tested SNPs corresponds to a minimum probability value for a genome-wide significance of P<8.91?10 ?6 .
Extra Analytical Analyses
Frequencies of stroke risk factors (hypertension, hyperlipidemia, and diabetes) between affected and unaffected participants were compared by using ? 2 tests. The correlation between affected sibling age at stroke was estimated by using the Pearson test of correlation. These analyses were conducted across all TOAST subtypes as well as after stratification by concordant and discordant subtypes among affected sibling pairs. Linear regression was used to determine the confidence intervals and linear fit of the age association, as shown in Figure 2. Kappa statistics were calculated to quantify concordance of phenotypes of interest within sibling pairs for all ages and stroke subtypes as well as models stratified by age (<65-year proband as defining age strata) and stroke subtype. All analyses that did not include genetic data were conducted by using scripts written in R (R Development Core Team, 2008). 9
Figure 2. Correlation between proband and sibling age at stroke. Correlation coefficient=0.83. P<0.0001. Pairs are points, the blue line is the linear model, and gray shading is the 95% confidence interval.
Performance
A total of 312 affected sibling pairs (312 probands) were enrolled at 70 centers across the United States and Canada. After quality control filtering, the final study population consisted of 223 probands, 248 stroke-affected siblings, and 84 stroke-unaffected siblings (total sample size, 555). Ischemic stroke–affected individuals had expected high rates of conventional atherosclerotic risk factors (Table 1). Stroke-affected individuals (probands and affected siblings) were significantly more likely to have hypertension (P<0.0001), hyperlipidemia (P=0.002), and diabetes (P=0.008) than were stroke-unaffected individuals. Stroke-affected siblings were somewhat older than the probands. This difference of 2 years (P=0.057) is expected, as an older sibling of the proband would be more likely to have a stroke than a younger sibling.
Sibling age at the time of stroke was strongly correlated with proband age at the time of stroke, despite the sibling’s being older. As shown in Figure 2 for all sibling pairs, the correlation coefficient was r=0.83 (95% CI, 0.78–0.86; P<2.2?10 ?16 ). For affected sibling pairs who had the same stroke subtype, the correlation coefficient was not different from all pairs, r=0.83 (95% CI, 0.75–0.89; P<2.2?10 ?16 ). This was the same for sibling pairs in which the affected siblings had different stroke subtypes, r=0.83 (95% CI, 0.77–0.87; P<2.2?10 ?16 ). More than 50% of the variance in age at stroke onset in siblings could be predicted by the age of the proband at the time of stroke. As shown in Table 2, there was significant concordance with affected siblings for TOAST subtype (kappa=0.13, P=5.06?10 ?4 ); this relation remained significant for sibling pairs in which the proband was <65 years old at the time of stroke and for sibling pairs in which the proband was 65 years or older.