Es would most likely not have been affected. The measures of immune activation were not assessed at enrollment into the study cohort, and we therefore cannot make temporal inferences of a causal nature nor could we control for them in the multivariable Cox models. We used prevalence rate ratios (PRRs) to estimate the associations of interest as risk was our parameter of interest. We chose PRRs as
Es would most likely not have been affected. The measures of immune activation were not assessed at enrollment into the study cohort, and we therefore cannot make temporal inferences of a causal nature nor could we control for them in the multivariable Cox models. We used prevalence rate ratios (PRRs) to estimate the associations of interest as risk was our parameter of interest. We chose PRRs as