Incidence of clinically defined infections. Negative binomial regression analysis with and without adjustment for time treated with hydroxyurea. The bars and point estimates for the CIs correspond to crude incidence rate ratios. ∗The adjusted result is from a Poisson mixed-effects regression model, as the negative binomial counterpart did not converge or had a non-positive variance estimated for the random intercept. †No convergence (no event in the placebo group).