2.4. Statistical analysis plan

When applicable, descriptive statistics will be summarized as mean ± SD or median and range for continuous variables such as age, BMI, and other lab values, or frequency and percentages for categorical variables such as sex, cardiometabolic disorders, and other health determinants.

Age- and sex-adjusted and unadjusted overall prevalence and confidence intervals (CI) of hypertension, dyslipidemia, T2D, metabolic syndrome, HU, and gout will be reported. We will estimate HU prevalence and 95% CI defined as SUA ≥ 6 mg/dL in females and SUA ≥ 7 mg/dL males. We will also estimate kidney function using the MDRD equations. Using the case definition of chronic kidney disease (CKD) as an estimated glomerular filtration rate < 60 mL/min/1.73m², we will estimate the prevalence of CKD in FA and assess its association with cardiometabolic disorders using 2-sample Chi-Square or Fisher’s exact tests, as appropriate. Prevalence estimates of cardiometabolic disorders, HU, and gout will be compared to the US using publicly available data from NHANES or other studies using Chi-Square tests.

The prevalence and 95% CI of HU, gout, and cardiometabolic disorders risk alleles, and pharmacogenetics variants will be estimated in FA. Further, these estimates will be compared with other populations from the 1000 Genomes Project Phase III using Chi-Square or Fisher’s exact tests. Following CPIC guidance on phenotyping, estimated prevalence of normal, intermediate, and poor drug metabolizers in FA will be calculated for each pharmacogene and compared with other populations as above. Associations between HU or gout and the development of cardiometabolic disorders (hypertension, T2DM, dyslipidemia, central adiposity, metabolic syndrome) and CKD will be assessed using 2-sample Chi-Square or Fisher’s exact tests. Differences in continuous outcomes such as blood biochemistry test results (uric acid, lipids, glucose levels, etc.) among the three genotypes of each genetic loci will be analyzed using Kruskal-Wallis test or analysis of variance (ANOVA). To estimate the association between variants and phenotypic traits, we will use mixed effects models having age and sex as covariates. The p-values for variants will be adjusted for multiple testing using a False Discovery Rate. Adjusted associations with HU or gout will be assessed using generalized linear models by adding additional features as covariates.

To account for the polygenic nature of developing cardiometabolic disorders, individual weighted genotype risk score (wGRS) will be constructed with a weight of 0 when no significant association exists between the risk allele and phenotype of interest (e.g., HU, gout, T2DM, dyslipidemia, etc.). A wGRS of 1, 2 and 3 will be based on odds ratio (OR) values, 1 < OR ≤ 2, 2 < OR ≤ 3, and OR ≥ 3, respectively. The cumulative effect of wGRS on phenotype of interest along with other clinical variables will be examined using logistic regression models. Multivariable log-binomial or log-Poisson regression models will be used to estimate relative risk and predict the effect of genetic and non-genetic factors in developing cardiometabolic disorders using the individual’s cumulative wGRS, comorbidities, and select social determinants of health. Receiver Operating characteristic (ROC) curve analysis, including sensitivity, specificity and area under the ROC curve will be used to assess the ability of prespecified health determinants to discriminate between those with and without cardiometabolic disorders.