Background: We wanted to determine whether immunologic markers predict later discordance between CD4+ T-cell and HIV-1 RNA changes during antiretroviral therapy.

Methods: This analysis included 10 prior antiretroviral studies, and 1,126 evaluable treatment-naïve and -experienced subjects. 4 intervals were studied: from study baseline to weeks 16 and 24, and time of HIV-1 RNA nadir to 16 and 24 weeks thereafter. 2 novel methods defined discordance, and were based on CD4+ T-cell (I = immunologic) and HIV-1 RNA (V = virologic) changes being better (+) or worse (-) than expected. Subject data were split into exploratory and confirmatory data sets. Recursive partitioning (RP) was applied to the exploratory data set to identify a subset at increased risk for discordance (either I-/V+ or I+/V-). The predictive value of being in the subset was tested in the confirmatory data set. Adjustments included protease inhibitor in the study regimen. Based on data availability, factors analyzed included activated (CD38+HLA-DR+) and naïve (CD45RA+CD62L+) CD4+ and CD8+ T cells, serum neopterin and β-2 microglobulin.

Results: At baseline and nadir, later discordance was most consistently predicted by CD4+ T cells and HIV-1 RNA. Other factors that helped predict discordance in some analyses were absolute naïve CD4+ T cells and absolute CD8+ T cells. In subjects with baseline HIV-1 RNA <4.53 log₁₀ copies/mL and receiving a PI, <178 naïve CD4⁺ T cells/mm³ at baseline predicted I-/V+ discordance at week 16 (p<0.00001). Among subjects with >271 CD4+ T cells at RNA nadir, having >800 CD8+ T cells at RNA nadir predicted I-/V+ discordance 16 weeks hence (p<0.00001).

Conclusions: Subsequent I+/V- and I-/V+ discordance were best predicted by CD4+ T-cell and HIV-1 RNA values. Risk of I-/V+ discordance was increased in subsets of patients (defined by RP) with fewer naïve CD4⁺ T cells and more CD8+ T cells. Targeted studies may better define the role of these surrogate markers in practice.