Statistical Analysis

The incidence, mortality, and clinical features of COVID-19 in this cohort were described as either frequencies or means and standard deviations for categorical and quantitative variables, respectively. Logistic and Poisson regression models were used when appropriate, depending on the nature of the dependent variable, to (a) compare the demographics, anthropometric measures, comorbidities, life habits, frailty surrogates, and APOE genotype between individuals with and without symptoms of COVID-19 to identify those variables associated with the disease; (b) compare the same data between the subsample of symptomatic volunteers with a clinical ascertainment of the diagnosis of COVID-19 and those without symptoms trying to reduce the noise of other eventual conditions; and (c) compare the individuals with mild and those with moderate/severe symptoms to identify variables associated with disease severity. Since the incidence and severity of symptoms of COVID-19 showed a trend to be associated with older age and male sex, both variables were included in regression analyses as covariates. To assess the impact of the model upon data, we reported odds ratios (OR) and their corresponding 95% confidence intervals (CI).

Since this is an exploratory study, correction for multiple comparisons was not performed. R (version 3.4.2) was used for statistical analyses; bilateral p values were reported.