950
Toward More Plausible Estimates of HIV-1 Incidence in Cross-sectional Serologic Surveys in Africa: Application of a HIV-1 Incidence Assay with Post-assay Adjustments
Andrea Kim*1, S McDougal1, J Hargrove2,3, J Hargrove2,3, M Morgan1, M Nolan4, L Marum5, A Abdullahi6, J Humphrey3,7, J Humphrey3,7, K Mutasa3, and B Parekh1
1CDC, Atlanta, GA, US; 2DST-NRF Ctr of Excellence in Epidemiological Modelling and Analysis, South Africa; 3ZVITAMBO Project, Harare, Zimbabwe; 4CDC Côte d'Ivoire, Abidjan; 5CDC Kenya, Nairobi; 6Natl HIV/AIDS and STD Control Prgm, Kenya Ministry of Hlth; and 7Johns Hopkins Univ Bloomberg Sch of Publ Hlth, Baltimore, MD, US
Background: The President’s Emergency Plan for AIDS Relief
aims to prevent 2 million HIV infections in resource-limited countries by 2010.
Monitoring HIV incidence trends is vital to assessing the initiative’s effect.
Currently, HIV incidence is mathematically modeled from antenatal clinic and
population-based cross-sectional serologic surveys using a UNAIDS supported Estimation
and Projection Package and Spectrum software. The BED capture enzyme
immunoassay (BED-CEIA) has been used to estimate HIV incidence but may
overestimate incidence in cross-sectional settings with long-term HIV
infections. Post-assay adjustments that correct for long-term infections
misclassified as recent infections on the BED-CEIA have been developed and are recommended
for use in estimating HIV incidence.
Methods: The BED-CEIA was applied to sera from Cote d’Ivoire’s
1998-2004 antenatal clinic
surveys and Kenya’s
2003 Demographic Health Survey. Annualized BED HIV incidence estimates
and 95% confidence intervals were calculated. Estimates were adjusted using 2
formulae that corrected for specificity and sensitivity, and specificity of the
assay. Unadjusted and adjusted estimates were compared to the UNAIDS model assuming
a 9-year post-infection survival.
Results: In Cote d’Ivoire, annualized BED HIV
incidence was 3.9% (95%CI 3.3 to 4.5) and modeled HIV incidence was 1.2%. BED
HIV incidence peaked in women aged 25 to 29 years. Adjusted BED HIV incidence
was 3.1% (95%CI 2.3 to 4.0) and 2.9% (95%CI 2.1 to 3.7), respectively. Adjusted
estimates were twice that of modeled incidence and peaked in youngest women
aged 15 to 19 years. In Kenya,
annualized BED HIV incidence was 3.0% (95%CI 2.3 to 3.7) and modeled HIV
incidence was 0.3%. BED HIV incidence was highest in persons aged ≥40
years. Adjusted BED HIV incidence was 2.4% (95%CI 1.9 to 2.9) and 2.2% (95%CI
1.7 to 2.7), respectively. Adjusted estimates were 8 times higher than modeled
incidence and remained highest in persons aged ≥40 years.
Conclusions: Although adjusted BED HIV incidence provided
plausible estimates of HIV incidence in Cote
d’Ivoire, the implausibility of Kenya’s adjusted
estimates suggest that the adjustments do not account for all
misclassification. Further validations of the adjustment formulae, including
application in different populations with comparison to directly observed
incidence, refinement of the imputed constants of the formulae, evaluation of
new laboratory methods, and reappraisal of the UNAIDS model are underway.
|
