Short Communication: Assessing Estimates of HIV Incidence with a Recent Infection Testing Algorithm That Includes Viral Load Testing and Exposure to Antiretroviral Therapy

Abstract

A recent infection testing algorithm (RITA) that includes a test for recent HIV infection and a viral load (VL) test is the recommended strategy to estimate national HIV incidence, reducing false-recent misclassification to <1%. The inclusion of information on exposure to antiretroviral therapy (ART), as a supplement to VL testing, could improve RITA performance by further lowering false-recent misclassification of true long-term infection. In 2012, Kenya and South Africa conducted national population-based surveys that collected information on HIV recency (i.e., HIV antibody seroconversion, on average, in the past 130 days) using the Limiting Antigen avidity (LAg-Avidity) enzyme immunoassay, HIV RNA levels, and ART exposure among HIV-infected respondents aged 15–49 years. In Kenya, ART exposure was defined as testing positive for one or more antiretroviral (ARV) drugs using high-performance liquid chromatography coupled with tandem mass spectrometry, and, if not ARV-positive, self-reporting a history of ART exposure. In South Africa, ART exposure was defined as testing ARV-positive. Two RITA strategies were compared: RITA #1 defined recent infection as testing LAg-Avidity-recent with unsuppressed VL (HIV RNA ≥1,000 copies/ml), and RITA #2 defined recent infection as testing LAg-Avidity-recent with unsuppressed VL and, if unsuppressed, having no ART exposure. RITA-derived incidence among persons aged 15–49 years in Kenya was 0.9% on RITA #1 and 0.8% on RITA #2. In South Africa, RITA-derived incidence was 2.2% on RITA #1 and 1.7% on RITA #2. Among specimens testing recent with unsuppressed VL in Kenya and South Africa, 16.0% and 19.7% had evidence of ART exposure, respectively. Although the performance of a VL- and ART-based RITA was encouraging, additional research is needed across HIV-1 subtypes and subpopulations to calibrate and validate this algorithm.

Background

Immunological tests for recent infection (TRI) that distinguish recent from long-term HIV infection provide information to estimate population-level HIV incidence and can be used to evaluate interventions aimed to reduce HIV transmission. On its own, TRI performance varies substantially across HIV-1 subtypes, with assay false-recent ratios (FRRs) ranging from 1% to 80%, resulting in overestimated HIV incidence in some settings.¹ Recent infection testing algorithms (RITAs) that combine results from a TRI with a viral load (VL) biomarker can be used to improve the predictive value of a RITA by reducing misclassification of treated individuals as recently infected on the algorithm.²

A RITA with a VL biomarker can reduce assay FRRs to <1%, and since 2015, it has been recommended by the World Health Organization (WHO) for cross-sectional HIV incidence estimation.¹ With improved predictive power, RITAs have the potential to enhance interventions to improve HIV case detection through accelerated partner notification and prevent onward transmission to HIV-negative contacts.^3,4

In addition to VL status, including information on antiretroviral therapy (ART) exposure in a RITA may further reduce false-recent classification, leading to additional improvement in the predictive value of the RITA. Since 2015, population-based HIV impact assessments in sub-Saharan Africa have used the WHO-recommended RITA, namely the Limiting Antigen avidity (LAg-Avidity) enzyme immunoassay combined with a VL biomarker, to estimate population-level HIV incidence. National HIV serosurveys that include VL testing and measurement of ART exposure provide the opportunity to evaluate whether information on ART exposure is a useful component to a RITA. Using nationally representative data from Kenya and South Africa, we compared a LAg-Avidity-based RITA with VL and ART exposure status with the WHO-recommended RITA for estimating national HIV incidence.

Materials and Methods

In 2012, South Africa and Kenya conducted national population-based household surveys to monitor trends in HIV prevalence and incidence, and evaluate the impact of the national HIV response.^5,6 Information on demographics, behavior, and use of HIV services, including ART for respondents who self-reported prior HIV diagnosis (Kenya only), was collected through interviews. Respondents provided dried blood spot samples for centralized laboratory testing, which included HIV antibody, HIV recency, VL, and antiretroviral (ARV) drug testing for specimens that tested HIV positive. Respondents included individuals aged ≥2 years in South Africa and 18 months–64 years in Kenya.

The LAg-Avidity assay (Maxim Biomedical, Inc., Rockville) was applied to HIV-positive samples. A normalized optical density (ODn) cutoff of 1.5 was used to classify specimens as testing recent (ODn ≤1.5) or nonrecent (ODn >1.5).⁷ Qualitative detection of ARV drugs was performed on HIV-positive samples using high-performance liquid chromatography coupled with tandem mass spectrometry and tested for ARV drugs representative of national treatment regimens in 2012. These included Lamivudine (3TC), Nevirapine (NVP), Efavirenz (EFV), and Lopinavir (LPV) in Kenya, and 3TC, NVP, EFV, LPV, Zidovudine (AZT), Atazanavir, and Darunavir in South Africa. Samples above the lower limit of detection (<0.02 μg/ml) for the ARV drugs tested were classified as ARV-positive. HIV RNA testing was conducted (Abbott Molecular, Inc., Des Plaines, IL), and viral suppression was defined as HIV RNA <1,000 copies/ml.

Two RITA strategies were compared: (1) testing LAg-Avidity-recent with unsuppressed VL (RITA #1) and (2) testing LAg-Avidity-recent with unsuppressed VL and, if unsuppressed, having no exposure to ART (RITA #2). In South Africa, ART exposure was defined as testing ARV-positive for at least one ARV drug; in Kenya, it was defined as testing ARV-positive for at least one ARV drug, and if not ARV-positive, having self-reported a history of ART exposure.

HIV incidence was calculated using a standardized formula for RITA-derived incidence, and a mean duration of recent infection (MDRI) of 130 days (95% CI 118–142) was applied to annualize HIV incidence.⁷ Incidence estimates were weighted using individual sampling weights to account for complex survey design and corrected to account for HIV-positive specimens with missing RITA results (14.6% Kenya; 1.6% South Africa). We found no statistically significant differences in demographics, CD4+ T-cell count, and HIV RNA concentration among those with complete and missing RITA results and assumed results were missing at random. Owing to unavailability of a locally derived residual FRR for the two RITA strategies in both countries, a residual FRR of 0% was assumed in the incidence calculation. For comparison across the two countries, we restricted our analysis to persons aged 15–49 years.

Results

Table 1 compares annualized HIV incidence among persons aged 15–49 years by RITA strategy. In Kenya, RITA #1 produced an annualized incidence estimate of 0.92%. With the addition of ART exposure status in RITA #2, HIV incidence reduced to 0.82%, reflecting a 10.9% reduction in HIV incidence. In South Africa, the annualized HIV incidence was 2.24% on RITA #1, and with the addition of ART exposure status in RITA #2, the annualized HIV incidence decreased to 1.72%, resulting in a total incidence reduction of 23.2%.

Table 1.

Annualized HIV Incidence Among Persons Aged 15–49 Years by Recent Infection Testing Algorithm, Kenya and South Africa, 2012

	Kenya ^a			South Africa ^b
Algorithm	Number testing recent	Incidence % (95% CI) ^c	% Reduction	Number testing recent	Incidence % (95% CI)^c	% Reduction
RITA #1: LAg+VL	25	0.92 (0.56–1.28)	—	91	2.24 (2.02–2.46)	—
RITA #2: LAg+VL+ART	21	0.82 (0.48–1.15)	10.9	73	1.72 (1.38–2.06)	23.2

In Kenya, 9,951 participants aged 15–49 years provided a blood sample for HIV testing, 548 of these tested HIV-positive, and 470 of these had results available for RITA analysis.

In South Africa, 14,721 participants aged 15–49 years provided a blood sample for HIV testing, 2,238 of these tested HIV-positive, and 2,202 of these had results available for RITA analysis.

Incidence estimation assumed a residual assay false-recent ratio of 0%, and weighted to account for sampling design and survey nonresponse.

ART, antiretroviral therapy; LAg-Avidity, Limiting Antigen avidity enzyme immunoassay; RITA, recent infection testing algorithm; VL, viral load.

Further analysis of LAg-recent samples by VL and ART status revealed that of 25 LAg-recent specimens with unsuppressed VL in Kenya, 4 (16.0%) had evidence of ART exposure (2 testing ARV-positive and 2 self-reporting ART exposure). In South Africa, 13 (19.7%) of 66 LAg-recent specimens with unsuppressed VL had evidence of ART exposure (data not shown).

Discussion

Using nationally representative data from Kenya and South Africa, we confirmed that nearly 20% of recent infection identified on the WHO-recommended RITA had evidence of ART exposure and, by extension, true long-term infections from individuals who were likely to be failing treatment. The addition of ART exposure status, as a supplement to VL testing in a RITA reduced HIV incidence by nearly one-quarter in South Africa and one-tenth in Kenya between RITA #1 and RITA #2, although the change observed in Kenya was not statistically significant. Although methods differ, our data suggest that a revised RITA with VL and ART exposure status is consistent with modeled estimates of HIV incidence in 2012 generated by the Joint United Nations Programme on HIV/AIDS spectrum model (Kenya: modeled estimate = 0.45%^*; South Africa: modeled estimate = 1.52%).^8,9

Caution should be exercised when interpreting RITA estimates given that variations in the parameters of MDRI and FRR overall and across HIV-1 subtypes have not been assessed, nor have validations been conducted against directly observed incidence estimates.² As rapid ART initiatives are scaled, further field evaluation of RITA performance will also be needed in this context. Notably, differences in the survey settings and methodology in the two countries, including level of HIV prevalence, HIV subtype distribution, and measurement of ART exposure, may have impacted the accuracy of the RITA estimates. Although testing for ARV drugs provides direct evidence of ART exposure, a positive result reflects recent and not lifetime exposure to ART. Self-reported ART exposure is simpler to collect but may under- or overestimate ART exposure due to social desirability and recall bias. Finally, missing RITA results may have impacted incidence estimates in Kenya where ∼15% of results were not available.

This report provides preliminary evidence that a LAg-based RITA combining VL and ART exposure status may improve estimates of population-level HIV incidence. In both countries, the revised RITA identified nearly 20% of long-term infections that were on treatment but misclassified as recently infected due to unsuppressed VL. Correction resulted in a greater reduction of HIV incidence in South Africa, a country with lower treatment coverage but higher burden of HIV than Kenya in 2012.^5,6 ARV drug testing is the preferred method for confirming ART exposure in surveys but is costly and may miss persons who are nonadherent to ART. Given the high predictive value of self-reported ART use, this approach combined with verification through health records in settings with robust health information system for HIV patients may be considered as a supplement to ARV drug testing.¹⁰

Footnotes

Acknowledgments

The authors would like to thank the survey participants, field teams, and stakeholders who made the Kenya and South Africa national surveys possible. The organizations who supported the survey in Kenya were: the National AIDS and STI Control Programme (NASCOP), Kenya National Bureau of Statistics, (KNBS), National Public Health Laboratory Services (NPHLS), National AIDS Control Council (NACC), National Council for Population and Development (NCPD), Kenya Medical Research Institute (KEMRI), U.S. Centers for Disease Control and Prevention (CDC/Kenya, CDC/Atlanta), United States Agency for International Development (U.S.AID/Kenya), University of California, San Francisco (UCSF), Joint United Nations Team on HIV/AIDS, Japan International Cooperation Agency (JICA), the Elizabeth Glaser Pediatric AIDS Foundation (EGPAF), Liverpool Voluntary Counselling and Testing (LVCT), the African Medical and Research Foundation (AMREF), the World Bank, and the Global Fund. The organizations in South Africa that supported the survey were: Human Sciences Research Council in partnership with the U.S. CDC, the National Institute for Communicable Diseases, the University of Cape Town, the South African Medical Research Council, and Global Clinical and Viral Laboratories. The Kenya survey was made possible by support from the U.S. President's Emergency Plan for AIDS Relief (PEPFAR) through cooperative agreements (#PS001805, GH000069, and PS001814) through the U.S. Centers for Disease Control and Prevention, Division of Global HIV & TB (DGHT). This work was also funded in part by support from the Global Fund, World Bank, and the Joint United Nations Team for HIV/AIDS. The South Africa survey was funded by PEPFAR through CDC under the terms of Cooperative Agreement Number 3U2GGH000570 with additional financial support from the United Nations Children's Fund, the South African National AIDS Council, and the Bill and Melinda Gates Foundation.

Disclaimer

A.A.K. and T.R. are members of the WHO HIV Incidence Assay Working Group. The findings and conclusions in this report are those of the authors, and do not necessarily represent the official position of the U.S. Centers for Disease Control and Prevention or the World Health Organization.

Author Disclosure Statement

No competing financial interests exist.

References

Kassanjee

, Pilcher

, Busch

, et al.: VL criteria and threshold optimization to improve HIV incidence assay characteristics. AIDS, 2016; 30:2361–2371.

Joint United National Programme on HIV/AIDS (UNAIDS): Recent Infection Testing Algorithm Technical Update: Applications for HIV Surveillance and Programme Monitoring. UNAIDS, Geneva, 2018.

Aghaizu

, Murphy

, Tosswill

, et al.: Recent infection testing algorithm (RITA) applied to new HIV diagnoses in England, Wales and Northern Ireland, 2009 to 2011. Euro Surveill, 2014; 19:pii: 20673.

Nikolopoulos

, Pavlitina

, Muth

, et al.: A network intervention that locates and intervenes with recently HIV-infected persons: The transmission reduction intervention project (TRIP). Sci Rep, 2016; 6:38100.

Shisana

, Rehle

, Simbayi

, et al.: South African National HIV Prevalence, Incidence, and Behaviour Survey, 2012. HSRC Press, Cape Town, 2014.

National AIDS and STI Control Programme (NASCOP): The Kenya AIDS Indicator Survey 2012: Final Report. NASCOP, Nairobi, Kenya, 2014.

Duong

, Kassanjee

, Welte

, et al.: Recalibration of the Limiting antigen avidity EIA to determine mean duration of recent infection in divergent HIV-1 subtypes. PLoS One, 2015; 10:e0114947.

National AIDS and STI Control Programme (NASCOP): Kenya HIV Estimates 2014. NASCOP, Nairobi, Kenya, 2014.

Rehle

, Johnson

, Hallett

, et al.: A comparison of South African national HIV incidence estimates: A critical appraisal of different methods. PLoS One, 2015; 10:e0133255.

10.

Kim

, Mukui

, Young

, et al.: Undisclosed HIV infection and antiretroviral therapy use in the Kenya AIDS indicator survey 2012: Relevance to national targets for HIV diagnosis and treatment. AIDS, 2016; 30:2685–2695.