Dynamics of HIV-1 viral load rebound among patients with previous suppression of viral replication

Objective: To model the dynamics of HIV-1 rebound in patients receiving suboptimal therapy after suppression of plasma viremia to < 200 copies/ml by triple combination therapy. Design: Mathematical modeling of data from 23 patients switched to indinavir maintenance therapy after viral replication was suppressed with a combination of indinavir, zidovudine and lamivudine. Modeling of HIV-I rebound among 23 patients on zidovudine/lamivudine maintenance was also performed for comparison. Methods: Evaluation of slopes of rebound and of their heterogeneity; calculation of the basic reproductive number (R-o, the number of newly infected cells arising from each productively infected cell); regression analyses for predictors of the slope of rebound. Results: Rebound of plasma HIV RNA followed a sigmoid curve with an initial exponential phase. There was significant heterogeneity in the slopes of rebound for individual patients (P < 0.001). In the indinavir maintenance rebounds, the average initial slope was estimated to be 0.587/day (doubling time 1.2 days). The slopes of rebound in patients on zidovudine/lamivudine maintenance tended to be less steep on average (P = 0.025). Among patients taking indinavir maintenance, the average R-o for the initial rebound of viremia was 4.3; in multivariate regressions, the slope of rebound was steeper during early rebound and in patients with higher viral load at the start of triple therapy or a higher CD4 cell count when indinavir monotherapy was initiated. The slope was less steep in patients with a greater increase in the number of CD4 cells during triple therapy. Conclusions: The rates of viral load increase among patients with viral rebound while receiving less than triple therapy are similar to those reported in patients interrupting therapy. Variability among patients may depend on viral fitness, target cell availability and extent of immune reconstitution. (C) 2000 Lippincott Williams & Wilkins.