Changes in Total Lymphocyte Count as a Surrogate for Changes in CD4 Count Following Initiation of HAART: Implications for Monitoring in Resource-Limited Settings

Abstract

A major obstacle to the administration of highly active antiretroviral therapy (HAART) in resource-limited settings is the high cost of CD4 count testing. The total lymphocyte count (TLC) has been proposed as a surrogate marker to monitor immune response to therapy. To assess, in a developed country setting, the capability and clinical utility of TLC change as a surrogate marker for CD4 count change in monitoring patients on HAART. Longitudinal co-variation between changes in TLC and concomitant changes in CD4 count following the initiation of HAART was examined using a retrospective cohort study of 126 HIV-positive patients attending The Miriam Hospital, Brown University, Providence, RI. Analyses included evaluation of the direction of TLC change as a marker for direction of CD4 change, using sensitivity and specificity; evaluation of absolute change in TLC as a marker for benchmark changes in CD4 (≥50 over 6 months, ≥100 over 12 months), using receiver-operator characteristic (ROC) curves; and a regression model of change in TLC as a function of change in CD4, to understand within-individual variation of longitudinal TLC measures. In the first 24 months of HAART, the sensitivity of a TLC increase as a marker for CD4 count increase over the same time period ranged from 86–94%, and the specificity ranged from 80–85%. The median change in TLC among patients with a CD4 count rise of ≥100 cells/mm3 at 1 year of HAART was +766 cells/mm3 while that of patients with a CD4 count rise of 3 Within the first 2 years of HAART, the direction of change in TLC appears to be a strong marker for direction of concomitant change in CD4 count (sensitivity 86–94% and specificity 80–85%, depending on length of interval). Positive and negative predictive values depend on the proportion of CD4 changes that are positive. In this cohort, that proportion is 87.9%, which yields high positive predictive value (96–98%) but lower negative predictive value (43–63%). Findings from the regression model suggest that taking multiple measurements of TLC at more frequent intervals may reduce variability and potentially improve predictive accuracy.