Prediction of hypertension based on the genetic analysis of longitudinal phenotypes: a comparison of different modeling approaches for the binary trait of hypertension

Abstract

For the analysis of the longitudinal hypertension family data, we focused on modeling binary traits of hypertension measured repeatedly over time. Our primary objective is to examine predictive abilities of longitudinal models for genetic associations. We first identified single-nucleotide polymorphisms (SNPs) associated with any occurrence of hypertension over the study period to set up covariates for the longitudinal analysis. Then, we proceeded to the longitudinal analysis of the repeated measures of binary hypertension with covariates including SNPs by accounting for correlations arising from repeated outcomes and among family members.We examined two popular models for longitudinal binary outcomes: (a) a marginal model based on the generalized estimating equations, and (b) a conditional model based on the logistic random effect model. The effects of risk factors associated with repeated hypertensions were compared for these two models and their prediction abilities were assessed with and without genetic information.Based on both approaches, we found a significant interaction effect between age and gender where males were at higher risk of hypertension before age 35 years, but after age 35 years, women were at higher risk. Moreover, the SNPs were significantly associated with hypertension after adjusting for age, gender, and smoking status. The SNPs contributed more to predict hypertension in the marginal model than in the conditional model. There was substantial correlation among repeated measures of hypertension, implying that hypertension was considerably correlated with previous experience of hypertension. The conditional model performed better for predicting the future hypertension status of individuals.

Longitudinal Data Analysis in Genome‐Wide Association Studies

“…Contributions also varied with respect to the main outcome variable used for analysis. Two groups focused on the binary outcome hypertension [Choi et al., ; Xia and Lin, ], and several groups used quantitative blood pressure phenotypes. Some analyzed only systolic blood pressure (SBP) [Musolf et al., ; Tan et al., ; Vaitsiakhovich et al., ], and a few others included both SBP and diastolic blood pressure (DBP) [Eu‐ahsunthornwattana et al., ; Hossain and Beyene, ; Wang et al., ].…”

“…Hossain and Beyene [] compared different linear mixed models, and Tan et al [] adopted a two‐level hierarchical model that captured relationships within and across individuals. Choi et al [] fitted a generalized linear mixed model (GLMM) and compared this with marginal (population average) estimation and prediction, using longitudinally measured hypertension status as the outcome of interest. One group, Xia and Lin [], used a partial GLMM strategy.…”

“…For the analysis based on real phenotype data, most of the studies reported no significantly associated SNPs with hypertension or blood pressure measurements. A few of the studies reported some associations, including rs3093642, rs13157168, and rs956918 [Vaitsiakhovich et al., ] and rs10510257 and rs1047115 [Choi et al., ]. There were no common findings between studies, which can be attributed to the different outcomes considered (hypertension, SBP, or DBP) and the different ways the outcomes were summarized (e.g., average or slope).…”

“…Others wanted to take advantage of the multiple measurements from each individual, the focus being not necessarily change over time [e.g., Hossain and Beyene, 2014]. Two groups were interested in using the mean phenotype [Eu-ahsunthornwattana et al, 2014;Wang et al, 2014], and one group was interested in both binary and time-to-event phenotypes for identifying candidate SNPs that were further investigated using longitudinal data in a logistic regression framework [Choi et al, 2014].…”

“…Two groups focused on the binary outcome hypertension [Choi et al, 2014;Xia and Lin, 2014], and several groups used quantitative blood pressure phenotypes. Some analyzed only systolic blood pressure (SBP) [Musolf et al, 2014;Tan et al, 2014;Vaitsiakhovich et al, 2014], and a few others included both SBP and diastolic blood pressure ( Choi et al [2014], analyzed time to first incidence of hypertension as an outcome and performed a genome-wide association study to select SNPs that were later included in a longitudinal logistic regression model with the outcome of hypertension (yes/no) and SNPs as independent variables. Choi et al also considered logistic regression as an alternative to time-to-event analysis, and they used any incidence of hypertension as an outcome.…”