The Impact of Delayed Switch to Second-Line Antiretroviral Therapy on Mortality, Depending on Definition of Failure Time and CD4 Count at Failure

Abstract

Little is known about the functional relationship of delaying second-line treatment initiation for human immunodeficiency virus-positive patients and mortality, given a patient's immune status. We included 7,255 patients starting antiretroviral therapy during 2004-2017, from 9 South African cohorts, with virological failure and complete baseline data. We estimated the impact of switch time on the hazard of death using inverse probability of treatment weighting of marginal structural models. The nonlinear relationship between month of switch and the 5-year survival probability, stratified by CD4 count at failure, was estimated with targeted maximum likelihood estimation. We adjusted for measured time-varying confounding by CD4 count, viral load, and visit frequency. Five-year mortality was estimated to be 10.5% (95% CI: 2.2, 18.8) for immediate switch and to be 26.6% (95% CI: 20.9, 32.3) for no switch (51.1% if CD4 count was <100 cells/mm3). The hazard of death was estimated to be 0.37 (95% CI: 0.30, 0.46) times lower if everyone had been switched immediately compared with never. The shorter the delay in switching, the lower the hazard of death-delaying 30-59 days reduced the hazard by 0.53 (95% CI: 0.43, 0.65) times and 60-119 days by 0.58 (95% CI: 0.49, 0.69) times, compared with no switch. Early treatment switch is particularly important for patients with low CD4 counts at failure.

Keywords: HIV; causal inference; second-line ART; targeted learning; treatment switching.

Figure 1

Flow diagram for inclusion of patients in our analysis, International Epidemiology Databases to Evaluate AIDS Collaboration, Southern Africa (IeDEA-SA), 2004–2017. ART, antiretroviral treatment; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; VL, viral load; WHO,World Health Organization.

Figure 2

Hazard ratio for death of each switching-delay duration subgroup, International Epidemiology Databases to Evaluate AIDS Collaboration, Southern Africa, 2004–2017. ”No switch” is the referent, and the analysis used inverse probability of treatment weighting of marginal structural models. A) Main analysis, baseline: confirmed failure (second viral load (VL) > 1,000). B) Secondary analysis, baseline: first VL > 1,000. Delay strategy refers to the duration of switching delay: strategy 0, no switch (reference category); strategy 1, <30 days; strategy 2, 30–59 days; strategy 3, 60–119 days; strategy 4, 120–179 days; strategy 5, 180–359 days; strategy 6, ≥360 days.

Figure 3

Probability of death 5 years after virologic failure for different CD4 count categories at time of failure and depending on month of switch (i.e., extent of delay), International Epidemiology Databases to Evaluate AIDS Collaboration, Southern Africa, 2004–2017. Estimates are based on working marginal models estimated with longitudinal targeted maximum likelihood estimation, as specified in Web Appendix 1. Note that the causal quantity of interest is defined as a projection of the true causal dose-response curve (i.e., the true relationship between time/switch time and mortality) onto the specified working model.