Revised acute phase estimates.

Our estimates (black) of the excess hazard-months attributable to the acute phase (EHMacute) based on (A) viral load trajectories (Fig. 2) and (B) our fit of a couples transmission model to the Rakai retrospective cohort. We compare these estimates with previous Rakai-based estimates that did not adjust for these biases (gray). These include (C) Wawer et al.’s adjusted and (D) unadjusted Poisson regressions [17], (E) Hollingsworth et al.’s variable hazard survival analysis [18], and (F) Powers et al.’s estimates that used a Bayesian framework to combine estimates from (E) and a mathematical modeling fit to an epidemic curve [19].