Table 2.

Parameter estimates for viral rebound models assuming different time-dependent recrudescence rates. ΔAIC is computed relative to the best-fit stylized distribution, the log-logistic, with parameter estimates in electronic supplementary material, table S1.

	recrudescence rate (r(t)) model	parameters	estimate (95% CI)	ΔAIC
(1)	r(t) = r0 e−k(t−τ)	r0	0.058 (0.049,0.069) per day	77.7
	single-phase decay, with	k	0.011 (0.009,0.014) per day
	r(t) → 0 as t → ∞	τ	4.04 (3.28,4.94) days
(2)	r(t) = r∞ + (r0 − r∞)e−k(t−τ)	r0	0.088 (0.073,0.106) per day	−10.7
	single-phase decay, with	r∞	0.002 (0.001, 0.004) per day
	r(t) → r∞ ≠ 0 as t → ∞	k	0.029 (0.022,0.040) per day
		τ	4.98 (4.58,5.41) days
(3)	$r(t)=r_{\mathrm{\infty }}\hspace{0.17em}{\mathrm{e}}^{-k_2(t-\tau )}+(r_0-r_{\mathrm{\infty }})\hspace{0.17em}{\mathrm{e}}^{-k_1(t-\tau )}$,	r0	0.088 (0.068,0.113) per day	−8.7
	where k2 < k1	r∞	0.002 (0.001,0.004) per day
	biphasic decay, with	k1	0.029 (0.019,0.053) per day
	r(t) → 0 as t → ∞	k2	5 × e−07 (1 × e−10, 0.002) per day
		τ	4.98 (4.57,5.42) days