2.1. Study Population
Health-related data from 1905 patients referred for diagnostic coronary angiography to the First Department of Cardiology of the Medical University of Gdansk between 2003 and 2006 were used to compile a study population (Wirtwein et al. 2017)(Wirtwein et al. 2018). In this project, as in our previous publication (Racis et al. 2020b), the same group with coronary atherosclerosis confirmed in coronary angiography was enrolled, however, the final analysis was conducted on 1020 individuals since 79 patients lacked follow-up data; detailed inclusion and exclusion criteria are presented in Fig 1.
2.2. Follow-Up Study
The prospective data were obtained from the National Polish Health Service by means of the patients’ names and Polish residence identification numbers. All patients were observed from the date of coronary angiography until 31 December 2011. The data used to determine the evaluated end points were collected in a 7-year follow-up (mean, 96 months).
The end points, also defined as CV events, were as follows: (1) CV death, (2) nonfatal myocardial infarction (MI) and (3) a combined end point of CV death, nonfatal MI or nonfatal cerebral stroke (CV death/MI/stroke). The International Statistical Classification of Diseases and Related Health Problems defined the term CV death. The term MI was applied to non–ST-elevation MI and ST-elevation MI. MI and stroke diagnoses were performed according to the European Society of Cardiology and the European Stroke Organization guidelines, respectively.
2.3. Genetic Analyses
The analysis included eight SNPs: PON1 c.575A>G, MPO c.-463G>A, SOD2 c.47T>C, GCLM c.-590C>T, NOS3 c.894G>T, NOS3 c.-786T>C, CYBA c.214C>T and CYBA c.-932A>G; these SNPs constituted the same set of SNPs analysed in our previous study. Primers, probe sequences, concentrations of reagents, and genotyping conditions are listed in Table S1. The risk alleles were defined according to their potential to increase the extent of atherosclerosis, as presented in our previous publication (Racis et al. 2020b). After each SNP was investigated individually, the additive effect of the eight SNPs was analysed as a GRS that reflected the total impact of genetic variants.
To construct the GRS model the number of risk alleles of every patient was established (0: no risk allele; 1: one risk allele, 2: two risk alleles) and multiplied by eight (the number of SNPs). Although seventeen groups of patients could be created (from 0 [having no risk alleles] to 16 [having risk alleles exclusively]), only 12 groups were selected—from the group with one to the group with 12 risk alleles—because no patients had 0, 13, 14, 15 or 16 risk alleles present. Thus, patients were divided amongst the following groups: 1 (n = 16), 2 (n = 50), 3 (n = 113), 4 (n = 178), 5 (n = 185), 6 (n = 210), 7 (n = 134), 8 (n = 79), 9 (n = 40), 10 (n = 9), 11 (n = 5) and 12 (n = 1). Then, these 12 groups were merged into four GRS groups according to the number of patients: the first three groups (1, 2 and 3 [n = 179]); the last five groups (8-12 [n = 136]) and the middle groups in sets of two (4 and 5 [n = 363]; 6 and 7 [n = 344]). Ultimately, according to the analysis of the Kaplan-Meyer curves, which reflected the different behaviour of groups 1-3, based on the likelihood adaptive fusing model selection, the whole population was divided into two final groups—GRS < 4 (n = 179) and GRS ≥ 4 (n = 843), which were named the low GRS group and the high GRS group, respectively (Fig 2).
2.4. Statistical Analysis
Continuous variables were expressed as means ± standard deviations (SDs). Categorical (dichotomous) variables were expressed as frequencies (%). For all SNPs, the risk allele frequencies were calculated. Hazard ratios (HRs) and 95% CIs were determined for each event using multivariate Cox proportional hazards models, in which one key independent variable (risk allele, GRS) was adjusted for age and sex. The clinical characteristics of the patients were presented as means ± SDs and as medians for continuous variables (i.e., body mass index, triglyceride level) or percentages for categorical variables (i.e., smoking history, hypertension, diabetes). Smoking status was self-reported. Deviations from the Hardy-Weinberg equilibrium for the genotypes were assessed using a chi-square test. The chi-square tests compared the frequencies of categorical variables between groups, and the Wilcoxon test compared levels of continuous variables between groups. The follow-up events were presented using the Kaplan-Meier curves, and the odds ratio (OR) and the log-rank test assessed the differences in survival (cumulative incidence of events) among different groups. The model with two groups according to the number of risk alleles (GRS < 4 and GRS ≥ 4) was chosen after adaptive fusing based on the likelihood ratio method as the one with the highest log-likelihood (Sitko and Biecek 2017). An alpha value of 0.05 was considered significant. All statistical analyses were performed using R version 4.0.2 (https://www.R-project.org).