Comparison of data reduction methods for family based association analysis of blood pressure intermediate phenotypes during physiological testing

¹J Sandoval, ¹I Arenas, ¹P Hamet, ²E Merlo
¹CR-CHUM - Technopôle Angus, 2901, Rachel Street East – Room 402 Montreal, Canada, ²École Polytechnique de Montréal, Case postale 6079, succursale Centre-ville Montréal, Canada

Blood pressure (BP) is a complex phenotype resulting from interactions among many physiological variables. Evaluation of BP main determinants during physiological testing may increase the power to detect associated loci. However, there are no studies comparing the usefulness of data reduction techniques for association studies on BP phenotypes. In this study we compared the average method, principal components (PC), and factor analysis (FA) to generate univariate traits for FBAT association studies during physiological testing. 183 individuals were subjected to orthostatic manoeuvres, and systolic (SBP), diastolic (DBP) and mean arterial BP (MAP) and intermediate phenotypes (total peripheral resistance (TPR), heart rate (HR) and stroke volume (SV) were measured every 5 and 2 minutes during supine (30min) and after adopting the standing position (10 min) using cardiac impedance. 200 single-nucleotide-polymorphic-markers (SNPs) selected from previous linkage studies of candidate genes for vascular and cardiac function were tested for family based association studies with FBAT using the mean of the measurements during supine or standing periods (Average method). For PC and FA with (FA-rot) and without (FA-nonrot) orthogonal rotation, the three first factors explaining either the phenotypic variance or the correlation between all phenotypes were retained and used as a trait for FBAT. Standing was associated with drastic changes in all phenotypic means and an increased variance compared with the supine period. BP was highly correlated with TPR (r=0.49) and moderately correlated with SV (r=-0.25) and HR (r=0.27), whereas TPR and SV were also highly correlated (r=-0.73). The overall P-values obtained from FBAT with each method were compared using the Wilcoxon test. Although for both the supine and standing periods FA-rot and PC provided better overall FBAT P-values compared with the average method, estimates were more significant for standing (P=0.03 and 0.04 for supine; 0.007 and 0.008 for standing). Analysis by phenotypes revealed that the best FBAT P-values using FA-rot and PC were obtained for heart rate. However, during standing better estimates of association for DBP and MAP were obtained with PC (P=0.04 and 0.05), whereas FA-rot gave better P value estimates for SV and TPR (P=0.05 and 0.001). These observations suggest that reduction techniques which reflect the common variance or the correlation between blood pressure intermediate phenotypes may improve the results of association tests.