Assessing linear CD4 decline quantifying diagnosis delay after HIV seroconversion: assessing the linearity assumption of CD4 decline

Abstract

Purpose: To estimate time from seroconversion to diagnosis, researchers have modeled time based on CD4 decline, assuming the square root of the CD4 count decreases linearly over time before antiretroviral treatment (ART) initiation. If true, utilizing CD4 counts reported anytime in the pre-ART period would result in estimates of diagnosis delay that are not appreciably different.

Methods: We applied CD4 depletion model parameters from seroconverter cohorts to New York City residents diagnosed from 2006 to 2015, having two or more pre-ART CD4 counts.

Results: Median diagnosis delays based on first or second pre-ART CD4 counts were similar (n = 12,849; 2.8 years, interquartile range [IQR]: 0-7.7, and 2.8 years, IQR: 0-7.6, respectively; P = .09, Wilcoxon signed-rank test). Among people whose second pre-ART CD4 count was measured more than 6 months after diagnosis (n = 2761), the average diagnosis delay based on first pre-ART CD4 count was shorter (1.5 years, IQR: 0-5.4) than the second pre-ART CD4 count (1.7 years, IQR: 0-6.0) but not significantly (P = .12).

Conclusions: Results are consistent with the linearity assumption of the CD4 depletion model. To estimate population-level diagnosis delay, researchers may use pre-ART CD4 counts reported more than 6 months post-diagnosis.

Keywords: CD4 depletion model; Diagnosis delay; HIV; Incidence.