Evaluation of Blood-Based Antibody Rapid Testing for HIV Early Therapy: A Meta-Analysis of the Evidence

Background Western blot (WB) assay is considered the gold standard test for HIV infection confirmation. However, it requires technical expertise and is quite time-consuming. WHO recommends blood-based rapid diagnosis to achieve same-day test and treatment. However, this rapid testing strategy has not been promoted worldwide due to inadequate research evaluating the effectiveness of rapid tests (RTs) as an alternative confirmatory HIV test for WB. This study aims to compare the diagnostic performance of rapid HIV tests compared with WB. Methods PubMed and Web of Science were searched for publications on rapid HIV tests using blood specimen. A meta-analysis was performed to quantitatively evaluate the diagnostic performance of rapid HIV tests compared with the WB assay in terms of pooled sensitivity, specificity, area under summary receiver operating characteristic (SROC) curve, and diagnostic odds ratio (DOR). Results Twenty articles involving 27,343 fresh specimens for rapid HIV tests were included in the meta-analysis. Regarding Capillus HIV-1/HIV-2, the pooled sensitivity, specificity, area under SROC curve, and DOR derived from six studies were 0.999 (95% CI, 0.956–1.000), 0.999 (95% CI, 0.991–1.00), 1.00 (95% CI, 0.99–1.00), and 1.0 × 106 (95% CI, 2.6 × 104–3.9 × 107) compared with the WB assay, respectively. With respect to Determine HIV-1/2, the pooled sensitivity, specificity area under SROC, and DOR derived from eight studies were 1.00 (95% CI, 0.789–1.000), 0.992 (95% CI, 0.985–0.996), 1.00 (95% CI, 0.99–1.00), and 1.8 × 106 (95% CI 406.049–7.8 × 109) compared with the WB assay, respectively. Regarding two-step serial RTs, the pooled sensitivity, specificity area under SROC, and DOR derived from eight studies were 0.998 (95% CI, 0.991–1.000), 0.998 (95% CI, 0.994–0.999), and 1.00 (95% CI 0.99–1.00) compared with the WB assay, respectively. Conclusion Our meta-analysis results may provide evidenced-based support for substituting RT for WB. Blood-based rapid HIV tests have comparable sensitivity and specificity to WB for HIV early therapy.

. Moreover, WB tests can only be performed by lab-professionals, which greatly restrict the accessibility of HIV testing. Surveys indicate that roughly 40-50% of HIV cases progressed to AIDS within the first year of diagnosis largely due to late diagnosis (3,4). According to the Chinese integrated HIV/ AIDS database, about 25% of newly diagnosed HIV/AIDS cases reported in China were already AIDS patients when they were first identified (5). Such late diagnoses are mainly caused by lack of access to fast and accurate HIV testing, particularly in settings that continue to rely on traditional WB confirmatory testing (6,7). However, simple and rapid blood-based HIV tests other than ELISA that target HIV antibodies have become available, enabling testing in outreach settings outside the laboratory at the point of client contact (8). Compared with WB assay, a rapid and accurate diagnosis in 15-30 min strategy is required. Blood-based rapid tests (RTs) targeting HIV antibody are convenient tests that require minimal laboratory infrastructure and expertise training compared with WB assay. Test results can be available in less than 30 min, so WHO recommends its serial use in different clinical settings to achieve same-day test and treatment (9,10). These tests can offer prompt test results and receipt of a "preliminary positive" test result, and therefore have greatly enhanced both the awareness rate and result notification rate of the infected and the coverage of antiretroviral therapy (ART) (11). The FDA recommends that the sensitivity and specificity of RTs should exceed 98% (12). Health authorities have recommended diagnosis strategies to replace traditional WB confirmatory test (13)(14)(15), using either combinations of HIV antibody RT reagents or combinations of RT reagents and ELISA reagents. China relies on a traditional HIV testing algorithm that requires WB confirmation to date. It is of great need to explore a fast and accurate strategy of HIV diagnosis by using RT centered testing. However, the effectiveness of HIV diagnosis using RTs has not been adequately explored. Therefore, the purpose of this paper is to evaluate the diagnostic performance of blood-based rapid HIV tests compared with the WB assay in terms of pooled sensitivity and specificity by meta-analysis, aiming to summarize the overall diagnostic accuracy of blood-based RTs as alternative algorithms for HIV early therapy.

metHODS
This study aims to compare the diagnostic performance of rapid HIV tests compared with WB assay in terms of pooled sensitivity, specificity, area under summary receiver operating characteristic (SROC) curve, and diagnostic odds ratio (DOR) among different populations. The work was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis PRISMA guidelines [see RPISMA checklist in Supplementary Material (16)].

Search Strategy
PubMed and Web of Science were searched using combinations of the following search keywords: [(HIV or AIDS) and ("rapid test" or "fast test")]. Reference lists of selected articles and related review articles were further screened. Additional searches in Google Scholar and manual searches of related journals (e.g., AIDS; Journal of Clinical Virology and Lancet Infectious Diseases) were also performed. Articles were limited to those published from January 2000 to May 2016. The search was limited to English language publications.

Study Selection
Studies were identified as eligible if RTs were conducted on whole blood sample.
Studies were excluded if (i) sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were not reported or could not be calculated from the information provided; (ii) RTs were conducted on saliva; (iii) RT results were not confirmed by WB; or (iv) they were the reviews of rapid HIV antibody tests.
Two independent reviewers (Jieqing Chen and Xiaojie Huang) without prior consideration of the results initially selected search results based on titles and abstracts. The remaining articles were further selected by full-text assessment. Disagreements between reviewers about eligibility were resolved by discussion.
The procedure of study selection and numbers of included and excluded studies is shown in Figure 1.

Data extraction
Two independent researchers (Bin Su and Zhiying Liu) extract data of each study into a spreadsheet. The following information was extracted from each selected study: information about the article (name of the first author, journal title, and the publication year), information about the RT (assay type, specimen type, and number of the test samples), information about the HIV-infected patients, and the performance outcomes, such as sensitivity, specificity, PPV, and NPV or other relevant data, to construct 2 × 2 tables of test results.
If there were several RTs performed on the blood samples from the same subjective populations in a study, just one RT could be included in a certain meta-analysis to ensure the independence of the included studies in one meta-analysis.
The quality of individual studies was rated by two independent reviewers (Xinchao Liu and Yaokai Chen) using a Quality Assessment of Diagnostic Accuracy Studies (QUADAS) scale (17).

Statistical analysis
Diagnostic meta-analysis was conducted by using command "midas" in StataSE 12.0, where the command required the number of observations of greater than or equal to 4. The presence of between-study heterogeneity when combining the study results was assessed using the chi-square-based Q-test of heterogeneity and the I 2 statistic (18). Heterogeneity was considered as moderate to large when P < 0.1 for Q-test or I 2 > 50%. Due to the paucity of trials in each comparison, we relied primarily on the I 2 statistic. Fixed-effect models were adopted to combine the performance measures derived from studies not presenting  moderate to large heterogeneity, assuming that all differences in observed effects were due to sampling error, and that there was one true effect size (i.e., the fixed effect) which underlined all the studies in the analysis. Otherwise, meta-analyses were conducted by random-effects models, because the true effect could vary from study to study and the true effect sizes were assumed to represent a random sample (i.e., the random effects) of effect sizes of individual studies. The combined performance indexes included the pooled sensitivity, specificity, summary receiver operating characteristic (SROC) curve, and diagnostic odds ratio (DOR) (19). In a single diagnostic test, sensitivity and specificity are defined as TP/ (TP + FN) and TN/(TN + FP), respectively, where TP, FP, FN, and TN represent the number of true positives, false positives, false negatives, and true negatives, respectively. Then, DOR is defined as: DOR is more suitable to compare diagnostic performance for two or more diagnostic tests. A DOR value ranges from 0 to infinity, with a higher value indicating a better discriminatory performance. DOR value of 1 means that a test does not discriminate between patients with and without the disorder. DOR also equals to the ratio of positive likelihood (LR+) and negative likelihood ratio (LR−), where a diagnostic test with LR+ >10 and LR− <0.1 (i.e. DOR > 100) is considered to be excellent discriminatory performance.
For those more used RT assays, one meta-analysis would be conducted on each of them to evaluate their performances, while one meta-analysis would be conducted on a mix of different assays that were just used in one or two RTs.

ReSULtS general Study information
The search strategy identified 3,884 papers from electronic databases and other sources in total. By reviewing the titles and abstracts of these papers, 3,727 irrelevant papers were ruled out. Of the remaining 157 studies that assessed rapid HIV testing, 111 articles without final retest results confirmed by WB were excluded after reviewing abstracts; 2 studies were excluded after detailed reviewing of the full text, because they did not provide sufficient information to construct the diagnostic 2 × 2 tables. Another 24 articles were excluded due to duplicate data or lack of data. Therefore, 20 papers involving 27,343 fresh specimens for RTs were included in the meta-analyses (Figure 1).
Among the 27,343 specimens from 20 studies, 6,753 specimens were HIV-positive and 20,590 were HIV-negative by WB.   The rapid HIV antibody detection assays used in these studies included Capillus HIV-1/HIV-2 ® (Trinity Biotech, Ireland or Cambridge Diagnostics, Galway, Ireland), Determine HIV-1/2 ® (Abbott Laboratories, Tokyo, Japan or Inverness Medical Japan Co. Ltd., Japan), Uni-Gold HIV test ® (Trinity Biotech, Inc., Wicklow, Ireland), and other independent assays. Combinations of two independent assays were also used in six studies. A summary of the studies and their key features is listed in Table 1.

DiScUSSiON
The prevention and control of HIV/AIDS as a chronic infectious disease should focus on timely detection and treatment. To further increase the coverage and availability of testing, traditional testing methods will not suffice (11,39). To achieve widespread testing coverage and rapid result acquisition, RT is a promising diagnostic approach that should be promoted. Though in many studies specificity and sensitivity were reportedly high, single studies were unable to give sufficient statistical power (20,37). While global guidance from WHO promotes a RT strategy, it has not been successfully implemented globally. One hypothesized reason for the lag in implementation is the inadequate research evaluating the effectiveness of RTs as an alternative to WB as a confirmatory HIV test. Moreover, it is difficult to validate trustworthy testing algorithms using different RTs with varying sensitivity and specificity. Currently, RT reagents can   be used for blood (including whole blood, serum, or plasma) and saliva. Several surveys conclude that RT reagents have lower specificity and higher false positive rate in saliva than in blood (40). Therefore, we conducted a meta-analysis to evaluate the specificity and sensitivity of blood-based RT compared with the WB assay in terms of pooled sensitivity, specificity by meta-analysis in a large number of samples. All the studies with appropriate reference (confirmed with WB) reported excellent estimates of sensitivity and specificity. We evaluated the performance of these RT strategies in the context of multiple diagnostic algorithms. Our meta-analysis evaluated the effectiveness of using different FDA-approved RTs on diagnostic accuracy. Areas under the SROC curves for the two most popular rapid assays (Capillus HIV-1/HIV-2 and Determine HIV-1/2) were all above 0.99 with sensitivities above 99.9%. In addition to the most popular RTs, 15 other kits have been evaluated in 15 studies (see mixed assays in Table 2). The pooled sensitivity, specificity, and area under the SROC curve of these assays were 0.998, 0.999, and 1.00, respectively, which demonstrates bloodbased rapid HIV test has comparable accuracy to WB for HIV early therapy. HIV screening via successive or simultaneous RT reagents has been widely adopted in Africa (41). Two successive RT reagents have lower costs than simultaneous RT reagents and are widely used for HIV screening (24). A study in Tanzania indicates that a good pair in combination is Korean SD and US Abbott Determine (20). SD can be used for screening and Determine can be used to recheck positive results. Both the sensitivity and specificity of this combination can reach up to 100% (23). Our meta-analysis also studied serial testing strategies (the second test is done only if the first test is positive). Overall, the pooled sensitivity and specificity were 0.998 and 0.998, respectively. Therefore, a serial two-step testing strategy has comparable accuracy to single test strategies.
The FDA regulation for manufacturers seeking licensure of tests recommends that the lower bound of the one-sided 95% confidence interval for sensitivity and specificity exceed 98% (12). Our review suggests that blood-based RT have high diagnostic accuracy, with comparable estimates when using a two-step or single testing strategy. It leads to early diagnosis and treatment of HIV and better clinical results. These data have the potential to change recommendations on voluntary counseling and testing from using originally ELISA based testing to RTs, and furthermore to replace the confirmatory WB test for HIV early therapy at the same day of detection. Particularly in countries and regions with high HIV/AIDS prevalence, timely actions should be taken to develop the relevant policies, technical protocols, and quality assurance systems to ensure the widespread implementation of RT.
Although the sensitivity and specificity of RT reagents both exceed 99.5%, they could be compromised due to unstandardized operations in non-laboratory settings (42, 43). The sensitivity of RT can be reduced in the absence of quality assurance and evaluation system (21). Unstandardized operations may lead RT false negative rate of up to 5.4% (29). RT test inevitably faces other challenges, such as inability of rechecking the same sample, and relatively low sensitivity for early HIV infection (42).
Our meta-analysis has several strengths. Algorithms either using serial RT testing strategies or single FDA-approved RT have been proved with satisfactory results. In addition, there has been an expansion in suitable specimen types (finger stick whole blood). We performed a comprehensive search of sources to identify studies that adopt different kinds of RTs. Several meta-analyses addressing the efficacy of so-called "rapid HIV testing" have been published recently, some of which focused on the fourth-generation ELISA test (Ag/Ab combination) that takes several hours to get testing result instead of "real" rapid HIV tests (43)(44)(45). And small sample sized meta-analysis has showed that rapid HIV voluntary counseling and testing improves the receipt rate of HIV test results among clients who seek HIV counseling and testing (45). Therefore, our current meta-analysis contributed uniquely to the field with greater sample size and trustworthy results. In some countries and regions, traditional tests still prevail, particularly in China. Thus, it is feasible to have RT performed by trained nonmedical professionals outside laboratories, which can promote HIV testing services among high risk groups such as MSM population more easily and greatly enhance both the awareness rate and result notification rate of the infected and the coverage of ART.
There also have been several limitations. For example, statistical comparison between subgroups (i.e., different populations) was not possible due to lack of data. Only English language studies were included in this meta-analysis, which may lead to a potential reporting bias. Different areas can determine the combinations based on the performance of reagents, costs, HIV prevalence, and risk behaviors of populations. Especially in low epidemic areas, indicating the need for counselors and clients to understand the limitations of RT positive results and the necessity to receive confirmatory tests. WB confirmatory test can be applied mainly for confirmation of indeterminate result.
Overall, our study indicated that RT would function as well as the WB and RT can and should be accessible and extensively used for HIV early therapy at the same day of detection if it is carried out by standardized management protocols. Quality assurance and quality control of RTs are extremely important and non-trivial in a devolved environment where such practices are not engrained in the culture, thus it is of great importance to acquire the knowledge and skills required for conducting quality control at a RT site. The time is ripe for RTs to play a more central role in increasing timely identification of HIV infections.

cONcLUSiON
Our meta-analysis results may provide evidenced-based support for substituting rapid test for WB. Blood-based rapid HIV tests have comparable sensitivity and specificity to WB for HIV early therapy.