HIV surveillance programs do not typically collect comprehensive data on antiretroviral therapy (ART). We validated a population-based measure of ART initiation that uses HIV viral load (VL) results in the absence of data on ART.Methods:
We used CD4/VL data reported to NYC HIV Surveillance for persons aged ≥13 years and diagnosed with HIV from 2006 to 2012 to validate estimates of ART initiation date based on 3 ART initiation definitions: (1) ≥1-log decline in copies per milliliter between 2 VLs over 3 months; (2) ≥2-log decline in copies per milliliter between 2 VLs over 3 months; and (3) the earliest of either a ≥1-log decline in VL over 3 months, or a change from detectable VL to undetectable VL (<400 copies/mL) over any interval. We plotted median CD4 counts by quarter before and after ART initiation to compare estimated initiation date with nadir of the CD4 trajectory.Results:
A total of 24,348 persons were diagnosed with HIV in NYC from 2006 to 2012. In all, 12,123 persons had probable ART initiation based on ≥2-log decline, 12,719 based on ≥1-log decline, and 14,311 based on ≥1-log decline or detectable–undetectable change. Lowest median CD4 count occurred at the estimated ART initiation date for all 3 definitions. The definition based on a ≥1-log VL decline or a change from detectable to undetectable VL captured more ART initiations and identified earlier initiation dates.Conclusions:
Serial VL measures are a valid source for estimating ART initiation. A definition that includes a ≥1-log VL decline or a change from detectable to undetectable VL performed best.