Keywords
HIV; rapid diagnostic tests; diagnostic performance; Bayesian meta-analysis; lateral flow immunoassay; enzyme immunoassay; Western Blot; PCR
-
Views
-
Downloads
How to cite this article
Sari AN, Abinawanto A, Nelwan EJ et al. Diagnostic accuracy of rapid lateral flow immunoassay tests in Diagnosing HIV Among Key and General Populations: Six Bayesian Meta-Analyses [version 1; peer review: awaiting peer review]. F1000Research 2026, 15:994 (https://doi.org/10.12688/f1000research.180372.1)
NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.
Export Citation
Sciwheel
EndNote
Ref. Manager
Bibtex
ProCite
Sente
▬
✚
Systematic Review
Diagnostic accuracy of rapid lateral flow immunoassay tests in Diagnosing HIV Among Key and General Populations: Six Bayesian Meta-Analyses
[version 1; peer review: awaiting peer review]
PUBLISHED 23 Jun 2026
OPEN PEER REVIEW
REVIEWER STATUS
This article is included in the Health Services gateway.
Abstract
Corresponding author:
Ayu Nurdiantika Sari
Competing interests:
No competing interests were disclosed.
Grant information:
This research has been funded by the Indonesia Endowment Fund for Education (LPDP), Ministry of Finance, Republic of Indonesia (Scholarship No. 0001135/IPA/D/PARS-I/2025).
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Copyright:
© 2026 Sari AN et al.
This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
How to cite: Sari AN, Abinawanto A, Nelwan EJ et al. Diagnostic accuracy of rapid lateral flow immunoassay tests in Diagnosing HIV Among Key and General Populations: Six Bayesian Meta-Analyses [version 1; peer review: awaiting peer review]. F1000Research 2026, 15:994 (https://doi.org/10.12688/f1000research.180372.1)
First published: 23 Jun 2026, 15:994 (https://doi.org/10.12688/f1000research.180372.1)
Latest published: 23 Jun 2026, 15:994 (https://doi.org/10.12688/f1000research.180372.1)
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Introduction
In 2024, an estimated 40.8 million people lived with HIV. The WHO targets that by 2025, 95% of people with HIV should have a diagnosis, 95% of those enrolled in antiretroviral treatment, and 95% achieving viral load suppression. However, in 2024, these figures only reached 87%, 89%, and 94%, leaving roughly 5.3 million individuals undiagnosed.1 Closing this diagnostic gap remains critical to achieving WHO targets. Third-generation HIV testing introduced lateral-flow immunoassay rapid diagnostic tests (RDTs) that detect antibodies, now the backbone of many national HIV diagnosis algorithms worldwide. RDTs deliver results within minutes and require fewer resources than conventional laboratory testing, making them advantageous for point-of-care use.2 Despite their widespread adoption, diagnosing HIV rapidly remains a significant gap in reaching WHO targets.
Past meta-analyses investigating RDT performance were published before 2020, and new studies have since emerged.3–8 HIV diagnostic research continues to evolve, including advancements such as digital self-tests, though current WHO guidelines still recommend HIV self-testing (HIVST) as optional, with low certainty of evidence.9 This study therefore aims to update and expand prior reviews by establishing reliable sensitivity and specificity estimates across diverse populations, given that test performance variability can meaningfully impact clinical and public health outcomes. It covers adults, children, and UNAIDS-defined key populations, reporting pooled diagnostic accuracy estimates (sensitivity, specificity, diagnostic odds ratio, and summary ROC) against multiple reference standards, with subgroup analyses exploring specimen type, administration method, and testing setting.
Methods
Study eligibility criteria
Studies were eligible if they evaluated the diagnostic accuracy of rapid HIV tests in participants with unknown HIV status, including children (aged 18 months to 18 years) and adults. Eligible populations encompassed the general population, individuals seeking voluntary counseling and testing (VCT), those starting provider-initiated testing and counseling (PITC), pregnant women, and key populations as defined by UNAIDS. The index test included any commercially available blood-based or oral fluid-based RDT from third or fourth-generation testing methods. The target condition was HIV infection, defined by confirmed presence of HIV-1 or HIV-2 antibodies or antigens. Reference standards included national testing algorithms, EIA, ELISA, Western blot, and nucleic acid tests (NAT). Eligible designs were observational studies or RCTs, whether laboratory- or field-based, published in peer-reviewed journals between January 2010 and 31 August 2025, in English. Studies were excluded if they focused on early infant diagnosis, individuals on ART or PrEP, or acute HIV infection prior to seroconversion, or if they failed to report sufficient data to construct 2 × 2 contingency tables.
Information sources, search, study selection
PubMed, Scopus, Cochrane Library, and DOAJ were searched, supplemented by reference lists, reverse citation searching, conference proceedings, and short communications. Search terms combined keywords related to HIV, RDTs, diagnostic performance, and self-testing, covering publications from 2010 to 31 August 2025. Records were imported into Rayyan for deduplication and screened independently by two reviewers (ANS, RRT). Full texts of potentially relevant studies were assessed for inclusion, with disagreements resolved through a third reviewer (AGK).
Data collection
Data were extracted using a piloted, standardized form by two independent reviewers (RT and ANS). Extracted variables included study characteristics, index test details, reference standard, 2 × 2 table data, and diagnostic accuracy outcomes. Study authors were contacted when data were unclear. Risk of bias and applicability were assessed using QUADAS-2 across four domains: patient selection, index test, reference standard, and flow and timing. For studies reporting multiple brands, QUADAS-C was applied. Disagreements were resolved by a third reviewer (AGK).
Data synthesis
The primary measures were sensitivity and specificity. Meta-analysis employed a Bayesian bivariate random-effects model,10 jointly modeling sensitivity and specificity while accounting for their correlation and between-study heterogeneity. Analyses were performed in R using meta4diag, INLA, and tidyverse packages. Summary estimates included pooled sensitivity, specificity, diagnostic odds ratio, and SROC curves with 95% posterior intervals. Heterogeneity was assessed via τ2 values and visual inspection of SROC curves and forest plots. Subgroup analyses were conducted for specimen type, test mode, and test setting. Review is registered in Prospero (PROSPERO 2025 CRD420251244881).
Results
Literature search and study selection
We identified 5,012 records through database searches ( Figure 1). After removing 1,124 duplicates, 3,888 unique records were screened by title and abstract, resulting in 3,695 exclusions. Full texts of 193 articles were assessed, of which 27 could not be retrieved and 112 of the remaining 166 were excluded. A total of 54 studies were included in the meta-analyses,8,11–63 with characteristics detailed in the supplementary materials. Studies originated from sub-Saharan Africa, Asia, and Latin America, with some conducted in laboratory environments using stored serum samples. Populations included adults, adolescents, pregnant women, children, and key groups such as men who have sex with men, sex workers, transgender individuals, people who inject drugs, and serodiscordant couples. Sample sizes ranged from fewer than 100 to over 13,000, and reference standards included EIA, ELISA, Western blot, PCR, or combinations thereof.
Figure 1. PRISMA flow diagram.
The overall assessment of bias showed that only about half of the studies had a low risk of bias. Nearly 75% of the studies were identified as having a low risk in patient selection, as well as flow and timing ( Figure 2). There were significant concerns about the handling of the index test and reference standards. These bias issues mainly stemmed from inadequate reporting of blinding and unclear personnel procedures or assignments for each test.
Figure 2. Risk of bias assessment (QUADAS-2 and QUADAS-C).
Diagnostic accuracy of HIV RDTs compared to various reference standards
Across all six reference standard comparisons, HIV RDTs demonstrated consistently high diagnostic accuracy ( Table 1). Against EIA or ELISA, pooled sensitivity and specificity were 97.9% (95% CrI: 96.6–98.8%) and 99.8% (95% CrI: 99.7–99.9%), with a DOR of 24,043.91 ( Figure 3). Performance improved further against EIA plus Western Blot (Se = 99.2%; Sp = 99.7%; DOR = 48,787.10; Figure 4) and approached near-perfect accuracy against Western Blot alone (Se = 99.8%; Sp = 99.7%; DOR = 126,407.99; Figure 5). Against the national HIV testing algorithm, pooled estimates remained excellent (Se = 99.1%; Sp = 99.6%; DOR = 29,833.83; Figure 6). Comparisons using two EIA/ELISA plus Western Blot and PCR yielded similarly high sensitivity (99.6% and 99.4%, respectively), though specificity was slightly lower under PCR (96.6%), reflecting its higher molecular precision ( Figures 7–8). Across all analyses, HSROC curves clustered near the upper-left region and consistently high DOR values confirmed robust test performance. Moderate between-study heterogeneity was observed, particularly in EIA/ELISA comparisons, likely reflecting differences in study design, specimen types, and population characteristics. The weak sensitivity-specificity correlation and overlapping credible intervals indicate this did not translate into clinically meaningful variability.
Table 1. Bayesian hierarchical meta-analyses of lateral flow immunoassay HIV RDTs compared to various reference standards.
Figure 3. Pooled diagnostic results (Single RDTs vs EIA/ELISA).
A. Overall Summary Receiving Operator Curve (ROC). B, subgroup analysis (Test Mode) ROC. C, subgroup analysis (Specimen). D, subgroup analysis (Setting). E, subgroup analysis (Population). F, pooled sensitivity. G, pooled specificity.
Figure 4. Pooled diagnostic results (Single RDTs vs EIA + WB or ELISA + WB).
A. Overall Summary Receiving Operator Curve (ROC). B, subgroup analysis (Specimen) ROC. C, subgroup analysis (Setting). D, subgroup analysis (Population). E, pooled sensitivity. F, pooled specificity.
Figure 5. Pooled diagnostic results (2 EIA or 2 ELISA, + WB).
A. Overall Summary Receiving Operator Curve (ROC). B, pooled sensitivity. C, pooled specificity.
Figure 6. Pooled diagnostic results (Single RDTs vs national algorithms).
A. Overall Summary Receiving Operator Curve (ROC). B, subgroup analysis (Test Mode) ROC. C, subgroup analysis (Specimen). D, subgroup analysis (Setting). E, subgroup analysis (Population). F, pooled sensitivity. G, pooled specificity.
Figure 7. Pooled diagnostic results (Single RDTs vs PCR).
A. Overall Summary Receiving Operator Curve (ROC). B, pooled sensitivity. C, pooled specificity.
Figure 8. Pooled diagnostic results (Single RDTs vs WB).
A. Overall Summary Receiving Operator Curve (ROC). B, pooled sensitivity. C, pooled specificity.
Subgroup analyses
For EIA/ELISA and national algorithm comparisons, stratification by specimen type, operator, setting, and population consistently demonstrated high accuracy with overlapping credible intervals. Self-testing performed comparably to provider-administered testing, and field-based studies maintained accuracy similar to laboratory settings. Minor reductions in sensitivity were observed in finger-prick samples, but were not clinically significant ( Table 1). Subgroup analyses were not performed for comparisons involving two EIA/ELISA plus Western Blot, PCR, or Western Blot alone, as limited variability in study characteristics, predominantly laboratory-based, provider-administered, general population studies, precluded meaningful stratification.
Discussion
This study demonstrates that HIV RDTs maintain consistently high diagnostic accuracy across multiple reference standards, testing conditions, and populations. The use of multiple reference standards, including serological assays, composite algorithms, and molecular methods, provides a more comprehensive evaluation than a single meta-analysis approach. Overall, HIV RDTs showed strong discriminatory ability, with performance approaching that of established laboratory-based standards, consistent with previous literature.3,6,64 An important finding is the variation in diagnostic performance depending on the reference standard used. Comparisons against serological standards such as EIA/ELISA and Western Blot demonstrated near-perfect accuracy, reflecting their shared immunological basis. In contrast, slightly lower specificity observed against PCR likely reflects the higher analytical sensitivity of molecular assays, which can detect early or low-level infections not yet identifiable by antibody-based RDTs, rather than true false-positive results, an inherent limitation during the serological window period.64
Moderate between-study heterogeneity was observed across several analyses. This might be driven by differences in study design, specimen types, testing environments, and population characteristics. However, this did not translate into clinically meaningful variability, as evidenced by strong clustering of HSROC curves and substantial overlap in credible intervals. The weak correlation between sensitivity and specificity further supports the overall stability of test performance across diverse settings. Subgroup analyses confirmed consistently high performance across specimen types, operators, testing settings, and population groups. Minor variations were observed in finger-prick and oral fluid samples, likely due to lower antibody concentrations or variability in sample collection, but these were not clinically significant.65 Notably, self-testing showed comparable accuracy to healthcare provider-administered testing, supporting its role in expanding access to HIV diagnosis, particularly in resource-limited or high-stigma settings.66 Similarly, comparable performance between laboratory-based and field-based studies underscores the robustness of HIV RDTs under real-world conditions, supporting their use in decentralized and community-based programs.
This study employed a Bayesian bivariate random-effects model, which offers methodological advantages including more intuitive uncertainty estimates, greater stability in handling sparse or heterogeneous data, and better accommodation of the correlation between sensitivity and specificity.67,68 Several limitations should be considered. Variability in study quality, including unclear blinding, differences in reference standard application, and inconsistent reporting, may introduce bias.69,70 Some studies used stored laboratory samples, potentially overestimating real-world performance. Variability in test brands, HIV subtypes, and geographic settings may further limit generalizability.71 At the review level, publication bias toward studies reporting higher accuracy may lead to overestimation of performance, and limited subgroup-specific data restricted more detailed analyses.72
These findings support the continued and expanded use of HIV RDTs in clinical and community settings. Their high diagnostic accuracy, combined with rapid results, ease of use, and cost-effectiveness, make them well-suited for large-scale screening.9 The comparable performance of self-testing and field-based testing further highlights their potential in underserved populations.4 Nevertheless, false-negative results may occur during early infection, and false-positive results, though rare, carry significant psychological and clinical consequences. Confirmatory testing using established algorithms therefore remains essential, particularly in settings with varying HIV prevalence or subtype diversity.73,74
Conclusion
To conclude, this study demonstrates that HIV RDTs maintain consistently high diagnostic accuracy across multiple reference standards, populations, and testing conditions. The robustness of these findings supports their role as a reliable tool for HIV diagnosis and screening. While certain limitations and sources of variability remain, the overall evidence strongly supports the integration of HIV RDTs into both clinical practice and public health strategies aimed at improving early detection and linkage to care.
Download
Export To
metrics
| Views | Downloads | |
|---|---|---|
| F1000Research | - | - |
|
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Citations
CITE
how to cite this article
Sari AN, Abinawanto A, Nelwan EJ et al. Diagnostic accuracy of rapid lateral flow immunoassay tests in Diagnosing HIV Among Key and General Populations: Six Bayesian Meta-Analyses [version 1; peer review: awaiting peer review]. F1000Research 2026, 15:994 (https://doi.org/10.12688/f1000research.180372.1)
NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.
track
receive updates on this article
Track an article to receive email alerts on any updates to this article.
Open Peer Review
Current Reviewer Status:
AWAITING PEER REVIEW
AWAITING PEER REVIEW
Key to Reviewer Statuses
VIEW
HIDE
ApprovedThe paper is scientifically sound in its current form and only minor, if any, improvements are suggested
Approved with reservations
A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.
Not approvedFundamental flaws in the paper seriously undermine the findings and conclusions
Comments on this article Comments (0)