Introduction
Nonclinical testing settings allow HIV testing programs to reach populations at increased risk who otherwise might not engage in regular HIV testing. Confirmation of HIV infection in nonclinical settings can expedite linkage to care for newly infected persons, those unaware of infection, and persons previously diagnosed but not in care [
1,
2]. The 2016 guidelines published for point-of-care (POC) HIV testing in nonclinical settings highlighted three possible testing algorithms: (1) a single rapid test with immediate linkage to clinical provider if initial test is reactive, (2) a single rapid test followed by laboratory-based follow-up testing if initial test is reactive, or (3) a rapid test immediately followed by a second rapid test on-site if the initial test is reactive [
2‐
4]. We evaluated the third nonclinical testing algorithm within National HIV Behavioral Surveillance (NHBS), which conducts anonymous cross-sectional behavioral surveys and HIV testing in annual rotating cycles with men who have sex with men (MSM), persons who inject drugs, and heterosexually active persons at increased risk for HIV. In 2017, NHBS recruited a large sample of MSM using venue-based sampling in 23 U.S. metropolitan statistical areas [
5,
6].
MSM are a population at increased risk for HIV acquisition who could benefit from access to low-barrier HIV testing. In the 2014 NHBS MSM survey, one in five MSM were HIV positive and one in four who were positive were potentially unaware of their infection [
5]. Until 2017, NHBS participants received one RT and, if reactive or preliminary positive, confirmation was performed using laboratory-based HIV IgG/IgM or Ag/Ab immunoassays at local or CDC laboratories. This created a barrier to knowledge of confirmed HIV status as participants needed to call or return in 1–2 weeks for confirmatory results [
7]. In the U.S., an estimated 25–48% of participants who receive conventional HIV laboratory testing do not return for final confirmatory test results [
8,
9]. A meta-analysis found that participants who received a RT were 1.5–2.2-times more likely to obtain their test results than those who received standard laboratory testing [
10].
As indicated in 2016 guidelines for non-clinical settings, a rapid-rapid testing algorithm (RTA) allows participants to receive their results the same day that testing is performed. In NHBS, the use of a RTA was implemented in 2017 to replace laboratory confirmatory testing [
7,
11]. Although the NHBS recommendation was to include two orthogonal RTs, the order (sequential or parallel) in which each RT was performed was not specified. For this analysis, the first RT performed in an RTA we define as RT-1 and the second as RT-2. If RTs were performed in parallel, RT-1 refers to the first RT conducted. Given that false-positive results are possible, employing an RTA allows for establishing concordance between the two tests (RT-1 reactive/RT-2 reactive) or whether they were discordant (RT-1 reactive/RT-2 non-reactive). Laboratory testing was included to address RT false-reactivity. This analysis evaluated the ability of different RTAs in non-clinical settings to provide confirmation of HIV infection status by calculating the frequency of concordant results and false-reactive results.
Results
During the 2017 MSM data collection, 11,232 participants were screened to participate in the 19 (18 s-RTA project areas and 1 p-RTA project area) NHBS project areas included in this analysis. Of the 6500 eligible participants, 6182 received a s-RTA while 318 received a p-RTA. Of 6182 receiving a s-RTA, the frequency at which each test was used for RT-1 was 2825 (45.7%) for DC; 1955 (31.6%) for INSTI; 528 (8.5%) for Unigold; and 874 (14.1%) for Sure Check. Regardless of the RT used 5960 (96.4%) were RT-1 non-reactive and received no further testing; 222 (3.6%) were RT-1 reactive and received a RT-2. Of the 222 RT-1 reactive, 216 (97.3%) were RT-2 reactive (concordant results) and 6 (2.7%) were RT-2 non-reactive (discordant results).
Tables
1 through
3 show the results of specific s-RTAs adopted in 18 project areas with concordance and discordance reported for each combination of RTs. Of 2825 participants testing initially with DC, 2295 (81.2%) were tested with INSTI as a second RT, 404 (14.3%) with Sure Check and 126 (4.5%) with Unigold. There was an overall concordance of 110/115 (95.7%) when DC was followed by any of the RTs while 5.2% participants tested with DC followed only by INSTI had discordant results (Table
1).
Table 1
Variation of sequential RTAs that started with Determine Combo among non-HIV-positive men who have sex with men in 19 U.S. cities—National HIV Behavioral Surveillance, 2017
DC→INSTI | 2295 | 95.8 (2198/2295) | 94.8 (92/97) | 5.2 (5/97) |
DC→Sure Check | 404 | 96.3 (389/404) | 100 (15/15) | 0 |
DC→Unigold | 126 | 97.6 (123/126) | 100 (3/3) | 0 |
For the s-RTA starting with INSTI, 1955 participants were initially tested of which 1134 (58%) participants were tested with DC as a second RT, 469 (24%) with Sure Check, and 352 (18%) with Unigold. There was an overall concordance of 46/47 (97.9%) when INSTI was followed by DC, Sure Check, or Unigold. While 3.3% of participants tested using INSTI followed by DC had discordant results (Table
2).
Table 2
Variation of sequential RTAs that started with INSTI among non-HIV-positive men who have sex with men in 19 U.S. cities—National HIV Behavioral Surveillance, 2017
INSTI→DC | 1134 | 97.3 (1104/1134) | 96.7 (29/30) | 3.3 (1/30) |
INSTI→Sure Check | 469 | 96.6 (453/469) | 100 (16/16) | 0 |
INSTI→Unigold | 352 | 99.7 (351/352) | 100 (1/1) | 0 |
Several project areas performed other variations of a s-RTA (Table
3). Only one s-RTA started with Unigold and was followed by INSTI which 528 (100%) participants were tested. For the s-RTA starting with Sure Check 874 participants were initially tested. Using INTSI as the second RT 851 (97.4%) participants were tested and 23 (2.6%) were tested with OQ. Unigold followed by INSTI and Sure Check followed by INSTI or OQ showed 100% concordance.
Table 3
Sequential RTAs that started with Unigold or Sure Check among non-HIV-positive men who have sex with men in 19 U.S. cities—National HIV Behavioral Surveillance, 2017
Unigold→INSTI | 528 | 94.3 (498/528) | 100 (30/30) | 0 |
SureCheck→INSTI | 851 | 96.7 (823/851) | 100 (28/28) | 0 |
SureCheck→OQ | 23 | 91.3 (21/23) | 100 (2/2) | 0 |
A p-RTA was adopted in one project area which screened 318 participants using Sure Check and OraQuick. In the p-RTA, participants received two orthogonal RT concurrently which returned 16 RT reactive on the first RT used. Fifteen (4.7%) participants had concordant reactive results and 302 (95%) participants had concordant nonreactive results. One (0.3%) participant had discordant results (SureCheck reactive, OraQuick non-reactive). The overall concordance for this algorithm was 99.6%.
Table
4 shows the testing performed to confirm HIV infection in samples from participants with discordant results in s-RTAs/p-RTAs. Although DC can differentiate Ag/Ab reactivity, the individual analyte reactivity data was not collected; only the overall reactive or non-reactive results were recorded. Of the five participants (samples 1–5) with discordant results in the DC-INSTI s-RTA, three consented to DBS storage and had DBS samples sent to CDC for confirmatory testing. The BRC was nonreactive for these and, therefore, Geenius was not performed following the recommended HIV diagnostic algorithm. Undetectable VL results and ARVs in samples 1 and 2 suggested DC false reactivity. Sample 3 had ~ 50 ng/mL of Dolutegravir (DTG) which could not exclude viral suppression. Sample 4 was HIV-1 WB antibody-negative which did not confirm the DC Ag/Ab-reactive result based on local testing, a DBS sample was not sent to CDC. Sample 5 did not receive confirmatory testing locally or at CDC lab as a sample was not received. The s-RTA starting with INSTI followed by DC had one participant with discordant results (sample 6); this sample was BRC reactive but Geenius HIV-1 Ab negative and VL and drug levels were also not detected, likely suggesting false reactivity in both the INSTI and BRC screening tests. In the p-RTA with Sure Check and OraQuick, one participant had discordant results (sample 7). This sample was tested locally and at CDC where WB, BRC, and VL results indicated Sure Check false reactivity.
Table 4
Results of confirmatory testing on dried blood spot specimens from non-HIV-positive men who have sex with men with discordant rapid test results—National HIV Behavioral Surveillance, 19 U.S. cities, 2017
1 | DC-R | INSTI-NR | | Non-Reactive | | Target not detected | Not detected |
2 | DC-R | INSTI-NR | | Non-Reactive | | Target not detected | Not detected |
3 | DC-R | INSTI-NR | | Non-Reactive | | Target not detected | DTG (+ /4) |
4 | DC-R | INSTI-NR | WB HIV-negative DBS not sent | | | | Not done |
5 | DC-R | INSTI-NR | DBS not sent | | | | Not done |
6 | INSTI-R | DC-NR | | Reactive (S/CO = 10.1) | HIV-1 Antibody Negative | Target not detected | Not detected |
7 | Sure Check-R | OQ-NR | WB HIV-negative | Non-Reactive | | Target not detected | Insufficient quantity |
Discussion
Using two orthogonal RTs to confirm HIV infection in non-clinical settings can allow for same day return of results and expedited linkage to care. We demonstrate that the high concordance of RT results in using an RTA supports that testing with two RTs is a viable option for diagnosis in populations at increased risk when conventional laboratory testing settings might not be readily accessible. Most discordant results could be attributed to RT-1 false reactivity in five of the seven cases.
While false-reactivity in HIV RTs has been widely reported, usually leading to discordant RT results, it is not common [
25]. False-reactive results can occur for many reasons including but not limited to technical issues with the test device, mislabeling, improper handling, or misinterpretation of a visually read test [
26]. The s-RTA starting with DC followed by INSTI had 5% (5/97) discordant results, INSTI followed by DC had 3.3% (1/30) discordant results, and Sure Check/OraQuick in the p-RTA had 0.3% (1/318) discordant results. Follow-up laboratory results suggest false-reactivity in few of the discordant samples, although results were inconclusive for others. False- reactivity was confirmed by further laboratory testing in DC followed by INSTI s-RTA (samples 1–4), although the source of the detected ARV in sample 3 remains unknown and could for instance be related to post-exposure prophylaxis. In the p-RTA, Sure Check/OraQuick (sample 7) indicate false reactivity in both screening tests after laboratory confirmation via WB was performed for one sample. VL and drug levels did not confirm reactivity in INSTI and BRC (sample 6). However, we cannot discount that low viremia and ARV drug exposure could have been missed by the assays or that the participant was an elite controller (suppressed HIV-1 RNA levels to below the limit of detection in the absence of ART) [
27]. In contrast, if it were an acute infection, HIV-1 RNA levels would have been higher than the limit of quantification of our validated VL assay. In the p-RTA, Sure Check/OraQuick indicate false reactivity in both screening tests after laboratory confirmation via WB was performed for one sample.
Confirmatory testing limitations include the use of WB for confirmation after screening with DC since WB cannot detect p24 antigen. Moreover, the WB assay faces many challenges such as high rate of uncertainty, low sensitivity, long testing period, and the limitation of detection in the early stages of seroconversion, at a time when cost of the assay is increasing [
28‐
30]. During this cycle, we did not collect the individual analyte reactivity to address the impact of Ag false-reactivity frequently reported from field studies [
31]. Other FDA-approved supplemental tests are available for validated off-label use with DBS or laboratory-developed tests could be used. Another confirmatory testing limitation is the sensitivity of the lab-validated Abbott
RealTime HIV-1 assay using four DBS punches (~ 50 µL), instead of 70 µL whole blood, that detects 66% of 3 log (copies/mL) [
21]; thus participants with low viremia may have been missed.
In following the s-RTA as designed, participants who self-reported being non-HIV-positive and had a non-reactive RT-1 result only received one RT; these participants were considered HIV-negative and did not receive a second RT. These samples were not sent to CDC for testing to evaluate whether any acute infections may have been missed. While RT kits are more expensive and less sensitive than conventional blood test on serum/plasma, research suggests that RTs are more cost-effective when factoring the number of people tested who receive their results when compared to conventional testing [
32,
33]. A randomized trial at an STI clinic found that 88% of MSM preferred RT to standard blood testing and a systematic review showed that 87–97% of clients would choose a RT over standard blood testing [
7,
34]. It is estimated that desirability of RTs would increase uptake of HIV testing three-fold compared to standard blood testing in community-based voluntary counseling and testing settings or when approached by HIV counselors in emergency departments [
7,
35‐
37]. Those who self-reported being HIV-positive with a negative RT result received confirmatory testing at CDC; these findings are being analyzed separately.
Acknowledgements
National HIV Behavioral Surveillance participants; CDC National HIV Behavioral Surveillance Team. National HIV Behavioral Surveillance Study Group. Atlanta, GA: Pascale Wortley, Jeff Todd, David Melton; Boston, MA: Monina Klevens, Rose Doherty, Conall O’Cleirigh; Chicago, IL: Stephanie Masiello Schuette, Antonio D. Jimenez; Dallas, TX: Jonathon Poe, Margaret Vaaler, Jie Deng; Denver, CO: Alia Al-Tayyib, Melanie Mattson; Detroit, MI: Vivian Griffin, Emily Higgins, Mary-Grace Brandt; Houston, TX: Salma Khuwaja, Zaida Lopez, Paige Padgett; Los Angeles, CA: Ekow Kwa Sey, Yingbo Ma; Miami, FL: Emma Spencer, Willie Nixon, David Forrest; Nassau-Suffolk, NY: Bridget Anderson, Ashley Tate, Meaghan Abrego; New Orleans, LA: William T. Robinson, Narquis Barak, Jeremy M. Beckford; New York City, NY: Sarah Braunstein, Alexis Rivera, Sidney Carrillo Newark, NJ: Barbara Bolden, Afework Wogayehu, Henry Godette; Philadelphia, PA: Kathleen A. Brady, Chrysanthus Nnumolu, Jennifer Shinefeld; Portland, OR: Sean Schafer, E. Roberto Orellana, Amisha Bhattari; San Francisco, CA: H. Fisher Raymond, Theresa Ick; San Juan, PR: Sandra Miranda De León, Yadira Rolón-Colón; Seattle, WA: Tom Jaenicke, Sara Glick; Virginia Beach, VA: Celestine Buyu, Toyah Reid, Karen Diepstra; CDC: Monica Adams, Christine Agnew Brune, Qian An, Alexandra Balaji, Dita Broz, Janet Burnett, Johanna Chapin-Bardales, Melissa Cribbin, YenTyng Chen, Paul Denning, Katherine Doyle, Teresa Finlayson, Senad Handanagic, Brooke Hoots, Wade Ivy, Kathryn Lee, Rashunda Lewis, Evelyn Olansky, Gabriela Paz-Bailey, Taylor Robbins, Catlainn Sionean, Amanda Smith, Cyprian Wejnert, Mingjing Xia.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.