ABSTRACT
Introduction HIV assays designed to detect recent infection, also known as “recency assays,” are often used to estimate HIV incidence in a specific country, region, or subpopulation, alone or as part of recent infection testing algorithms (RITAs). Recently, many countries and organizations have become interested in using recency assays within case surveillance systems and routine HIV testing services, and in measuring other indicators beyond incidence, generally referred to as “non-incidence surveillance use cases.”
Methods To identify best methodological and field implementation practices for the use of recency assays to estimate HIV incidence and trends in recent infections for key populations or specific geographic areas, we undertook: 1) a global Call for Information released from WHO/UNAIDS; and 2) a systematic review of the literature to: (a) assess the field performance characteristics of commercially available recency assays, (b) understand the use of recency testing for surveillance in programmatic and laboratory settings, and (c) review methodologies for implementing recency testing for both incidence estimation and non-incidence use cases.
Results and discussion Among the 90 documents ultimately reviewed, 65 (88%) focused on assay/algorithm performance or methodological descriptions, with high-quality evidence of accurate age- and sex- disaggregated HIV incidence estimation at national or regional levels in general population settings, but not at finer geographic levels for prevention prioritization. The remaining 25 documents described field-derived incidence (n=14) and non-incidence (n=11) use cases, including integrating RITAs into routine surveillance and assisting with molecular genetic analyses, but evidence was generally weaker or only reported on what was done, without validation data or findings related to effectiveness of recency assays when used for these purposes.
Conclusions HIV recency assays have been widely validated for estimating HIV incidence in age- and sex-specific populations at national and sub-national regional levels; however, there was a lack of evidence validating the accuracy and effectiveness of using recency assays for non-incidence surveillance use cases. The evidence identified through this review will be used in forthcoming technical guidance on the use of HIV recency assays for surveillance use cases by WHO and UNAIDS; further evidence on methodologies and effectiveness of non-incidence use cases is needed.
Introduction
There are many reasons to identify recently acquired HIV infections on a population level, including: (1) To better understand current transmission of HIV in a country, region, or population subgroup, (2) To evaluate whether specific prevention interventions are having the desired impact, and (3) To focus limited resources for prevention or treatment services on groups of people or geographic locations with the greatest potential benefit (e.g., reducing risk for onward transmission). HIV assays designed to detect recent infection, also known as “recency assays,” can be used to gain an understanding of these epidemic dynamics.
Recency assays discriminate recent from longstanding infection in an individual using one or more biomarkers, typically using an understanding of the typical patterns of immune response maturation following initial infection. Individual recency assay results can be used in a cross-sectional survey to estimate incidence by building on the common epidemiological equation P = I × D (prevalence = incidence × duration of infection).1 However, the accuracy of the incidence estimate is dependent upon accurate knowledge of the performance characteristics of the recency assay or algorithm, specifically mean duration of recent infection (the average time post-infection that individuals are classified as recently infection; MDRI) and false-recent rate (the proportion of long-infected individuals misclassified as recently infected; FRR), and the precision of the estimate is sensitive to these same parameters.2
To date, no recency assay has fully met the target product profile for HIV incidence estimation as set out by the Foundation for Innovative Diagnostics (FIND) and the World Health Organisation (WHO) in 2016.3,4 Numerous factors have been identified that adversely affect recency assay performance and lead to substantial misclassification of longstanding infections as recent (i.e., raise the FRR). Factors that can affect assay performance include natural variability in individual immune responses (in particular, elite control of HIV or natural viral suppression), variability in biomarker progression for different HIV-1 subtypes, the types of specimens collected and storage methods, advanced HIV disease, and treatment with antiretroviral therapy (ART) or use of pre-exposure prophylaxis (PrEP).5-10 The effect of ART on increasing the FRR of recency assays appears to be more pronounced when a person receives treatment very early after initial infection;11,12 which is complicated by rapid improvements in treatment coverage worldwide, as well as uptake of pre-exposure prophylaxis (PrEP). Other factors that may impact assay performance but are not yet well-characterised include sex, pregnancy status and the presence of co-morbidities.13-15
Since the release in 2011 of technical guidance on the use of recency assays to estimate population-level HIV incidence from the WHO and Joint United Nations Programme on HIV/AIDS (UNAIDS),16 the field has changed substantially, motivating release of interim guidance at various times.13,17-20 Numerous examples in the peer-reviewed literature now highlight the necessity of adjustments at a local level to improve the accuracy of incidence estimates derived using recency assays within population-based surveys.14,21-30 Beyond that primary application, however, many countries and organizations have become increasingly interested in using recency assays within HIV case surveillance systems and routine HIV testing services, to measure indicators other than incidence, such as the identification of epidemiologically-linked clusters of recent infections; geographic hotspots; or subpopulations with relatively high, ongoing, or emerging transmission, to inform prioritization of HIV prevention, testing, and partner notification or contact tracing interventions. These types of epidemic monitoring and evaluation strategies are generally referred to as “non-incidence surveillance use cases” for recency assays. However, the non-random nature by which people are included in these types of surveillance systems and programmes requires special attention to characterize and, ideally, mitigate the effect of these selection biases on the accuracy of these non-incidence estimates.
METHODS
To identify methodological and field implementation practices for the use of recency assays for HIV incidence and non-incidence surveillance use cases, we used two strategies. First, a global Call for Information was released as a joint endeavour between WHO and UNAIDS staff working on both surveillance and HIV testing services. This call included a brief questionnaire through which member states and/or regional health jurisdictions were asked to describe current implementation (if any) of recency testing in programmatic HIV testing services in terms of assay performance, clinical utility, and utility for surveillance, including but not limited to incidence estimation.
Second, a systematic review of the literature was conducted, with four primary objectives:
Understand the use of recency testing in surveillance, programmatic and laboratory settings (to provide incidence estimates or for non-incidence surveillance use cases),
Review methodologies for implementing recency testing in population surveys, case surveillance systems and routine monitoring & evaluation activities, and
Highlight use cases that have employed a recency assay or Recent Infection Testing Algorithm (RITA) within specific populations, with special attention to variations in assays, settings, and methods of analysis for calculating HIV incidence estimates and/or employing recency assays for non-incidence surveillance use cases.
Eligibility Criteria for the Systematic Review
The systematic review included two sets of searches, each with a different strategy. Strategy 1 involved looking for articles about recency assay performance in laboratory and field survey settings. To be eligible for inclusion in the review, articles needed to describe some aspect of performance of recency assays/methodologies (e.g., MDRI, FRR, sensitivity, specificity, false positive rate, false negative rate, accuracy, number tested and proportion recently infected; or correlation, R, percent agreement, or kappa related to another standard assay). They also needed to use commercially available assays/methodologies used to determine recency of infection (see Table 1 for this list). Articles reviewing use of a laboratory-developed (“home-grown”) assay that was not commercially available were excluded from the review.
Strategy 2 involved looking for articles about surveillance and programmatic utilization of recency testing; articles reviewed under this strategy was intended to supplement findings from the WHO Call for Information as described above. To be eligible for inclusion, articles needed to describe some aspect of population-level utility (identification of “hotspots”, clusters, case surveillance and/or incidence estimation), using commercially available recency assays/methodologies (e.g., RITAs, adapted assay protocols) used to determine recency of HIV infection. Studies could present either qualitative or quantitative data, and could be descriptive studies lacking a comparator, as long as studies clearly presented outcomes specific to HIV recency testing.
Search Strategy
The literature search for the systematic review was conducted in PubMed and Web of Science, and included literature published in any language and in any indexed journal including preprint servers without peer review, from 1 January 2010 to 5 January 2021, searching title, abstract, and MeSH Terms/author keywords.
For the Strategy 1 search, search terms included HIV; recency assay; incidence assay; test for recent infection (TRI); false recent rate/ratio (FRR); proportion false recent; and mean duration of recent infection (MDRI). For the Strategy 2 search, search terms included recent infection/acute infection; recent infection testing algorithm (RITA); incidence estimates; case surveillance; hotspot identification; hotspot mapping; cluster detection; procedures and protocols; and HIV. See Table S1 for search sets and terms, and Table S2 for our search code.
Given that much of the research output in the field of HIV recency assay utilization is published in formal reports or presented in conference abstracts, we extended the search beyond traditional literature databases to include “grey literature,” i.e., literature that is not formally published in peer-reviewed journals or books. We conducted a search of the grey literature through internet search engines and through websites of major international funders, subject matter conferences, and organizations involved with HIV surveillance (see Table S3) employing the following search terms across sites: “surveillance,” “recency testing,” “case surveillance,” “incidence estimation,” “hotspot,” and “HIV.”
We used a step-wise approach during the screening and reviewing process. After search and duplicate removal, LA screened titles and abstracts to identify papers potentially related to the focus areas and eligibility criteria. After screening was complete, remaining articles for full-text could be obtained were then independently reviewed by SNF and LA to determine if the study met eligibility criteria; ADM served as a tiebreaker for any articles for which the two preliminary screeners were not in agreement about inclusion. Once the full-text review was complete, LA hand-searched the references of all included articles for additional, potentially eligible articles. SNF and LA then reviewed these articles and determined eligibility according to the process outlined above.
Prior to conducting our search, we developed a formal protocol and circulated it among stakeholders at UNAIDS and WHO for approval; we have made the protocol available in unmodified form as supplemental material to this article.
Assessment of Evidence Strength
Both protocols and reports submitted by countries as supporting documentation in response to the WHO Call for Information, and literature included in the systematic review was rated by strength of published evidence using a 23-point rubric (see Figure 1). For each piece of evidence, two team members (SNF and LA) independently rated the strength of evidence through a Microsoft Excel-based scoring rubric designed to implement the grading structure found in Figure 1. If there was disagreement between the two team members, either ADM or EG performed an assessment using the rubric and served as a tiebreaker.
RESULTS AND DISCUSSION
The search was conducted on 5 January 2021, and resulted in 611 records identified via MEDLINE (PubMed), 823 records identified via Web of Science, and 315 records identified through an Internet search of grey literature. An initial “quick screen” round resulted in 1,271 records from the MEDLINE and Web of Science searches being removed from the pool for clearly not meeting inclusion criteria for the review. Of the 478 pieces of evidence remaining after the quick screen, 58 records were identified as duplicate between the two databases and grey literature search and, once removed, resulted in 420 abstracts to be formally screened. Over a 4-month period ending 20 January 2021, 48 survey responses were received to the global Call for Information from five different WHO regions (AFRO, SEARO, WPRO, PAHO, and EURO).
Literature screening steps
After de-duplication, a remaining 420 documents from the search were scanned by LA to identify papers related to our two strategy areas based on the eligibility criteria. A total of 380 records were then excluded; reasons for exclusion are detailed in Figure 2. The remaining 140 documents were then subjected to a full text review, which was conducted independently by both LA and SNF. After excluding full-text articles that did not meet our pre-defined inclusion criteria, a total of 74 studies, reports, or presentations were retained across both focus areas (Figure 2). An additional 16 pieces of evidence were then incorporated into the review from country submissions in response to the global Call for Information, leading to 90 items that were graded for strength of evidence.
Characteristics of Included Studies
Among the 90 pieces of evidence that were identified through country survey and systematic review and that met the inclusion criteria, 65 (88%) 1,4,5,7-15,20,22-73 focused on assay performance, algorithm performance, or methodological descriptions of incidence estimation. The quality of evidence was “very-strong” (40/65), “strong” (13/65), “moderately strong” (11/65), and “weak” (1/65) in these 65 articles. The remaining 25 pieces of evidence described field-derived incidence and non-incidence use cases. Of these, 16 (64%) described use for incidence estimation, 8 describe non-incidence use cases, and 1 described both incidence and non-incidence use cases.
Among the articles describing use of recency assays for estimation of HIV incidence, 12 (75%) 74-85 described national surveillance in the form of population-based surveys (including 10 from the US-supported Population-based HIV Impact Assessment (PHIA) surveys). These population-based incidence use cases are also sometimes known as impact assessment use cases, because they are intended for repeat implementation to assess changes in incidence over time as a result of HIV prevention or care interventions. Most evidence in this category (7/12) was judged to be “moderately strong,” with more details of strength ratings found in Table 2. The remaining five 86-90 articles described calculation of incidence among key or sentinel populations, including those accessing routine HIV or blood donation programs. Key or sentinel population surveillance involves testing within populations that are either of specific interest because they are at higher risk for infection (key) or considered to be representative of a larger population (sentinel). Sentinel and key population surveillance may be facility-based or community-based. For example, needle and syringe distribution programmes are a good point of contact with people who inject drugs, sexual health clinics may provide access to men who have sex with men (MSM) and sex workers, and antenatal clinics are used to sample pregnant women. Most evidence in this category (4/5) was “very strong” quality.
Among non-incidence use cases, seven91-97 described case-based surveillance among new HIV diagnoses in routine HIV programme settings and 2 described recency testing to inform targeted prevention planning, including one from a population-level cohort study90 and one that described use of recency testing within three separate routine HIV country programs.98 The quality of evidence was at least “moderately strong” in all 9/9 non-incidence use cases, each identified from the peer-reviewed literature. Each of these 9 articles used proportion testing recent and/or odds for recently acquired infection as the primary outcome measures.
Tables S4-S5 provide details on each of the 90 pieces of evidence included in this review, including the strength rating and topic of focus for each item.
Use of Recency Assays for HIV Incidence Estimation
In 2015 UNAIDS and WHO released guidelines on monitoring the impact of the HIV epidemic using population-based surveys.99 Since then, 12 population-based surveys with published results have utilized this approach, the majority (n=10) of which were part of the global PHIA assessment.100 In addition to five articles that described efforts to estimate incidence within key or sentinel populations (including those accessing routine HIV programs72,86,89 or blood donation programs87,88), results from 10 completed PHIA surveys with public final reports were also entered into the review.75-84 These surveys involve cross-sectional, household-based, nationally representative sampling of adults and adolescents aged 15 years and older, with some surveys also including children aged 0-14 years. All PHIA countries were located in sub-Saharan Africa, except Haiti (which did not contribute evidence to this review).101 PHIA participants receive home-based HIV testing and counselling. Those who are HIV-positive undergo a laboratory-based RITA. During the first three PHIA surveys in Malawi, Zimbabwe, and Lesotho, the RITA included the Sedia™ HIV-1 Limiting Antigen (LAg) Avidity assay in combination with viral load (VL). The subsequent 7 surveys added antiretroviral (ARV) detection to the LAg and VL tests as an enhanced measure to distinguish recent from long-term infections. Incidence estimates were obtained from the RITA result in accordance with an established cross-sectional incidence estimator2 and performance characteristics specified as a mean duration of recent infection = 130 days (95% CI: 118, 142), time cut-off = 1.0 year and residual proportion false recent = 0.0% (note that an assumption of FRR equal to 0.0% is explicitly not recommended in the 2011 WHO guidance;16 in addition, no uncertainty in the FRR estimate was accounted for in these analyses). No adjustment for subtype-related variation in MDRI was made. Survey weights were utilized for all estimates to account for the complex sampling design. The sample size of PHIA surveys is designed to provide subnational-level (e.g., provinces, regions) estimates of viral load suppression among people living with HIV (PLWH) aged 15 to 49 years with a 95% CI +/- 10% or less, which typically yield reasonably precise estimates of national-level HIV incidence among people aged 15 to 49 years. As a result, these surveys were able to generate HIV incidence estimates disaggregated by sex and high-level region, but not estimates that could be used to target HIV prevention or care to specific districts and/or key populations.
Non-Incidence Surveillance Use Cases of HIV Recency Assays
Though there were 11 peer-reviewed studies reporting on non-incidence use cases identified across a 9-year review period, evidence on non-incidence use cases derived from routine HIV programme settings was scant. Four of these studies were from a single upper-income country (Ireland), three of which reported annual updates on approximately the same method of national implementation.
Beginning in 2016,91,102 Ireland’s Health Protection Surveillance Centre integrated recency testing combined with epidemiological data into national HIV surveillance to better monitor and inform HIV prevention interventions. In this system, normalised optical density (ODn) of new HIV diagnoses (measured using the Sedia™ HIV-1 Limiting Antigen-Avidity EIA) was linked with data captured in the national infectious disease reporting system. People with new diagnoses were classified as recent based on an ODn ≤ 1.5, unless epidemiological or clinical criteria (CD4 count <200 cells/mm3; viral load <400 copies/ml; the presence of AIDS-defining illness; prior ART use) indicated a probable false-recent result. During the pilot implementation of recency surveillance in Ireland in 2016, 448 of 508 (88.1%) new diagnoses nationwide were linked to a recency test result, with 12.5% of new diagnoses classified as recent. People who inject drugs had the highest proportion of recent infections (26.3% of new diagnoses were recent) and recent infection was significantly more likely with a concurrent sexually transmitted infection (aOR 2.59; 95% CI 1.04–6.45). However, data were incomplete for at least one RITA criterion in 48% of cases. Since then, further efforts have been made to improve completeness of the required epidemiological data, with national reports published for 201794 and 2018.95 This example demonstrates the feasibility of integrating a RITA into routine surveillance, along with the challenges of identifying demographic sub-groups with ongoing HIV transmission when surveillance data are incomplete. However, it does not attempt to distinguish populations with high proportions of recent infections due to ongoing transmission from those experiencing improvements in testing frequency (which would lead to more diagnoses early in infection), and no details were available about potential systems modifications that resulted from this change to the surveillance strategy.
We identified two non-incidence, case surveillance reports from countries that used genetic testing combined with RITAs to better understand transmission dynamics.90,92 In one example, researchers in Belgium analysed HIV-1 pol sequences obtained through baseline drug resistance testing of patients newly diagnosed with HIV between 2013 and 2017.92 Information on genetic similarity was combined with demographic data and information on the recency of infection for 927 patients. They found that 48.3% of the patients were genetically linked to others, with 11.4% belonging to a pair and 36.9% involved in a cluster of ≥3 members. Patients of Belgian origin were more frequently involved in transmission clusters (49.7% compared to 15.3%) and diagnosed earlier (37.4% compared to 12.2%) than patients of Sub-Saharan African origin. Of the infections reported to be locally acquired, 69.5% were linked (14.1% paired and 55.4% in a cluster). Interestingly, equal proportions of early and late diagnosed individuals (59.9% and 52.4%, respectively) were involved in clusters, calling into question the added benefit of recency testing for molecular surveillance activities. The researchers argued that identification of a genetically linked individual for the majority of locally infected patients suggested a high rate of diagnosis in this population. However, frequent delays in diagnoses after infection increased opportunities for onward transmission, thus indicating that earlier diagnosis should be prioritized to protected HIV-uninfected members of sexual networks.
Finally, although this review only identified two pieces of evidence related to the targeted prevention use case, more evidence of using HIV recency assays for this purpose will likely emerge from the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) “TRACE” initiative (Tracking with Recency Assays to Control the Epidemic) in the near future. Beginning in Fiscal Year 2019, PEPFAR funded 16 countries (El Salvador, Eswatini, Ethiopia, Guatemala, Kenya, Lesotho, Malawi, Namibia, Nicaragua, Panama, Rwanda, Tanzania, Uganda, Vietnam, Zambia, Zimbabwe) who are nearing the 90–90–90 targets to introduce the TRACE initiative.103 Through TRACE, a lateral flow rapid recency assay is conducted as a supplementary test in routine HIV testing services and/or within HIV case surveillance to detect recent infection among newly diagnosed PLWH in all (or most) facility-and community-based testing sites in a country to drive prevention and care planning.
A variety of terms have been used to describe the analytic goals of TRACE programs including to identify, investigate, and intervene on; “areas with ongoing active transmission,” “hot-spots,” “sub-populations with high levels of HIV recency,” “clusters,” “pockets,” and “outbreaks.” In a 2018 viewpoint article promoting the TRACE initiative, Kim et al. argued that point-of-care recency assays used in conjunction with geographic data will allow jurisdictions to immediately hone in on “hotspot” locations and “sub-populations with high levels of HIV recency,” or to facilitate the identification and investigation of “clusters” of recent infections, triggering a public health response.104 PEPFAR’s 2019 annual report included narrative text stating that, “When used as an ancillary test in all those who are newly diagnosed with HIV, recency testing enables the identification of recent transmission “pockets.”105 More recently, the PEPFAR annual report for 2021 described using recency testing to identify where active transmission is occurring as HIV “outbreak” investigations.106 More precise and/or unified terms and indicators across all settings would improve interpretation and comparability across countries and regions. Furthermore, misidentification of clusters, hotspots and other imprecisely defined indicators through recency testing may result in misdirected or poorly designed prevention plans and missed opportunities for targeting limited resources.
Limitations
There are several limitations to our systematic review. First, while the majority of the literature included in this review (74 of 90 pieces of evidence) came from the systematic review, additional pieces of evidence were submitted in response to the WHO country survey that were not identified during our systematic search; therefore, this portion of the systematic review is not reproducible and may suffer from selection bias. Second, given our search strategy many of the articles included in this review involved findings relevant to the performance of specific commercially available recency assays. However, some of those assays (e.g., the Sedia™ BED HIV-1 Incidence EIA) are technically available but no longer in wide use, due to inferior performance for HIV incidence estimation compared to other available assays. Further, some assays included in this review are not available in all countries globally. Third, as with all systematic reviews, our review process was imperfect and time-limited. Some meaningful literature was noted by our team after the review was complete,107-110 and has not been included to preserve fidelity to our pre-specified protocol.
CONCLUSIONS
Despite widespread use of HIV recency assays for both HIV incidence estimation and non-incidence surveillance use cases, evidence on validated and accurate uses of recency assays for non-incidence surveillance remains weak. Further, more consistent and precise use of terminology is warranted, as is more rigorous validation of non-incidence indicators and methodologies to inform programme management. The evidence identified through this review will be used in a forthcoming technical guidance on the use of HIV recency assays for surveillance use cases, to be released by WHO and UNAIDS. This document is expected to help raise global awareness of benefits and pitfalls of the use of these assays for surveillance purposes, and set clear standards for their appropriate use. Based on the evidence identified through this review, this guidance will be able to provide strong updated recommendations about methods for population-level HIV incidence estimation. However, lack of evidence validating the accuracy and effectiveness of using recency assays for surveillance use cases other than incidence estimation will likely lead to less strong recommendations in those areas, and/or inability to make specific recommendations for use of recency assays for these purposes.
Data Availability
Details of the search terms and results of this systematic review are available as supplemental material.
CONFLICT OF INTEREST STATEMENT
SNF and EG have received consulting income and research support from Sedia Biosciences Corporation and Gilead Pharmaceuticals.
AUTHORSHIP
SNF and LA conducted the systematic review and country survey, with advice and support from ADM and EG. SNF and ADM wrote the initial draft. KM, DL, SD, and MM provided funding and oversight of the project. All authors provided substantive revisions to the manuscript.
ACKNOWLEDGMENTS
We acknowledge Virginia Fonner, Theresa Yeh, Cheryl Case Johnson, Anita Sands, and Rachel Clare Baggaley, who helped with the development and distribution of the global Call for Information from WHO member countries and participants of the WHO HIV Recency Testing Working Group. Thanks also to those who responded to this Call for Information, including WHO country office staff, researchers, public health surveillance staff, and HIV testing programme staff throughout the world.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.
- 7.↵
- 8.
- 9.
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.
- 19.
- 20.↵
- 21.↵
- 22.↵
- 23.
- 24.
- 25.
- 26.
- 27.
- 28.
- 29.
- 30.↵
- 31.
- 32.
- 33.
- 34.
- 35.
- 36.
- 37.
- 38.
- 39.
- 40.
- 41.
- 42.
- 43.
- 44.
- 45.
- 46.
- 47.
- 48.
- 49.
- 50.
- 51.
- 52.
- 53.
- 54.
- 55.
- 56.
- 57.
- 58.
- 59.
- 60.
- 61.
- 62.
- 63.
- 64.
- 65.
- 66.
- 67.
- 68.
- 69.
- 70.
- 71.
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.
- 77.
- 78.
- 79.
- 80.
- 81.
- 82.
- 83.
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.
- 94.↵
- 95.↵
- 96.
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.↵
- 107.↵
- 108.
- 109.
- 110.↵
- 111.