Evaluation of the false recent classification rates of multiassay algorithms in estimating HIV type 1 subtype C incidence
AIDS Research and Human Retroviruses
Botswana-Harvard AIDS Institute Partnership, Gaborone, Botswana; Division of Medical Virology, University of Stellenbosch, Stellenbosch, South Africa; Washington State University College of Veterinary Medicine, Pullman, WA, United States; Department of Biostatistics, Harvard School of Public Health, Boston, MA, United States; Division of Sleep Medicine, Brigham and Women's Hospital, Boston, MA, United States; Department of Immunology and Infectious Diseases, Botswana-Harvard School of Public Health AIDS Initiative, Harvard School of Public Health, FXB 402, 651 Huntington Avenue, Boston, MA 02115, United States; National Health Laboratory and Botswana-Harvard HIV Reference Laboratory, Gaborone, Botswana; National Health Laboratory Service, Cape Town, South Africa; R. Stempel College of Public Health and Social Work, Florida International University, University Park, Miami, FL, United States
Laboratory cross-sectional assays are useful for the estimation of HIV incidence, but are known to misclassify individuals with long-standing infection as recently infected. The false recent rate (FRR) varies widely across geographic areas; therefore, accurate estimates of HIV incidence require a locally defined FRR. We determined FRR for Botswana, where HIV-1 subtype C infection is predominant, using the BED capture enzyme immunoassay (BED), a Bio-Rad Avidity Index (BAI) assay (a modification of the Bio-Rad HIV1/2+O EIA), and two multiassay algorithms (MAA) that included clinical data. To estimate FRR, stored blood samples from 512 antiretroviral (ARV)-naive HIV-1 subtype C-infected individuals from a prospective cohort in Botswana were tested at 18-24 months postenrollment. The following FRR mean (95% CI) values were obtained: BED 6.05% (4.15-8.48), BAI 5.57% (3.70-8.0), BED-BAI 2.25% (1.13-4.0), and a combination of BED-BAI with CD4 (>200) and viral load (>400) threshold 1.43% (0.58-2.93). The interassay agreement between BED and BAI was 92.8% (95% CI, 90.1-94.5) for recent/long-term classification. Misclassification was associated with viral suppression for BED [adjusted OR (aOR) 10.31; p=0.008], BAI [aOR 9.72; p=0.019], and MAA1 [aOR 16.6; p=0.006]. Employing MAA can reduce FRR to <2%. A local FRR can improve cross-sectional HIV incidence estimates. © 2014, Mary Ann Liebert, Inc.
adult; age; article; Botswana; CD4 lymphocyte count; classification algorithm; cohort analysis; enzyme immunoassay; female; follow up; gender; human; Human immunodeficiency virus 1 (strain C); Human immunodeficiency virus 1 infection; Human immunodeficiency virus infected patient; major clinical study; male; middle aged; opportunistic infection; priority journal; virus load; young adult; algorithm; blood; classification; geography; HIV Infections; HIV Seropositivity; Human immunodeficiency virus 1; incidence; pregnancy; prospective study; Human immunodeficiency virus antibody; Adult; Algorithms; Botswana; CD4 Lymphocyte Count; Female; Geography; HIV Antibodies; HIV Infections; HIV Seropositivity; HIV-1; Humans; Immunoenzyme Techniques; Incidence; Male; Pregnancy; Prospective Studies; Viral Load