Recruitment and enrollment

This was a cross-sectional diagnostic study, conducted between March 2012 to June 2015, which enrolled individuals who had negative rapid HIV results and were 14 years or older. Participants were recruited and enrolled from 5 HIV counseling and testing (HCT) clinics in the Tshwane district of South Africa (SA). Four of these HCT clinics were antenatal clinics and one was a general HCT clinic. Rapid HIV testing was done according to the SA HIV testing guidelines, which recommend a serial HIV testing strategy at the points-of-care [11]. Testing at the HCT clinics was done by HIV counselors, who had received training in HIV testing and counseling. The rapid HIV test that was commonly used for screening during the course of this study was Advanced Quality (Intec Products Inc). Abon (Abon Biopharm) was used for screening in 2014; however, this was replaced with Advanced Quality at the end of 2014. At enrolment, study samples and participants’ cell phone numbers were collected. Plasma was separated from whole blood through centrifugation at 1700 relative centrifugal force (RCF) for 20 minutes, and stored at -70°C within 24 hours after collection until the time of testing.

Ethics statement

Written consent was obtained from all participants before enrollment. The study was approved by the University of Pretoria’s Faculty of Health Sciences Ethics Committee (Protocol number– 295/2015) and by Tshwane Research Ethics Committee (TMREC 2010/26). The legal ages for consenting to HIV testing and medical treatment in South Africa are 12 and 14 years, respectively [12]. During the course of the study we noticed that some people who came for HIV testing were younger than 18 years of age, and came alone without parents or guardians. Hence, we amended our study protocol to include this group, in order to extend the benefit of earlier diagnosis of HIV to them. We applied for inclusion of this group with our Ethics Committee, and were granted approval. So all study participants, including 14–17 years, signed the same written consent forms that were used for participants older than 18 years.

Sample testing

Roche CAP/CTM HIV VL version 2 assay (Roche Diagnostics, Mannheim, Germany) with a lower detection limit of 20 copies/ml was used for NAAT, in a mini-pool of 5 samples, using 200 μl from each sample to constitute a 1 ml sample volume required for testing. An additional volume of about 20 μl from one of the pool samples was used for top-up in order to avoid sample rejection due to insufficient volume. Pools that had undetectable VL were considered negative for HIV. Individual sample testing was done in pools that had detectable VL. A VL threshold of ≥5000 copies/ml in an individual sample was considered as diagnostic for HIV infection. If HIV VL was <5000 copies/ml, a repeat test was done in a follow-up sample in order to exclude a possible contamination in the initial test [5]. All participants were encouraged to voluntarily collect NAAT results; however, those who tested positive on NAAT were contacted on their cell phones to come back to the clinic for further management. During this follow-up visit, HIV counseling and repeat rapid testing were done, follow-up samples collected, and participants referred for appropriate management.

The following serology tests were done in NAAT-positive samples: 3^rd generation Genscreen HIV-1/2 version 2 ELISA (BioRad, Marnes-la-Coquette, France) and HIV Western Blot (Bio-Rad Laboratories, Redmond WA, USA) for antibody detection; p24 antigen (Roche Diagnostics, Mannheim, Germany); and limiting antigen (LAg) HIV avidity assay {Maxim Biomedical Inc., Rockville, USA} for confirmation of early HIV infection in samples with detectable antibodies. LAg avidity assay was repeated on follow-up samples. Acute HIV infection was defined as the presence of HIV RNA with or without p24 antigen in the absence of HIV antibodies. Early HIV infection was defined as the presence of HIV RNA with or without p24 antigen, and presence of HIV antibodies with low avidity as reflected by values <1.5 normalised optical density (OD-n) on LAg avidity assay. Samples found to have HIV RNA with or without p24 antigen, and HIV antibodies with high avidity of >1.5 OD-n were classified as having chronic infection.

Rapid HIV testing was later repeated from stored plasma samples in the laboratory using the same tests that were used at the points-of-care. All the tests were performed and analyzed according to manufacturer’s instructions. CD4 count results were later enumerated from the laboratory information system and patient records for NAAT-positive participants.

Statistical analysis

A descriptive analysis was used to present summary statistics (median, proportions and 95% confidence intervals) for the parameters. This was followed by a comparison between the groups using two sample independent t-tests for proportions. Proportions of acute, early and chronic HIV infections that were missed at the points-of-care were computed. Comparison of the proportions of newly diagnosed HIV infections between males and females was done. In addition, median values of HIV VL and CD4 count were compared between the groups of participants with early and chronic HIV infections. All the statistics were performed on the STATA version 14.1 software (StataCorp LP, College Station, TX, USA). A p-value of ≤0.05 was considered statistically significant.

Demographics and newly diagnosed HIV infections

From March 2012 to June 2015, the study enrolled and tested 6910 participants who had negative rapid HIV test results (Fig 1). Their median age was 27 years (IQR: 23–31). Females formed a large proportion (87%, n = 6011) of the study group, and 88% (n = 5271) of the female participants were pregnant. NAAT detected HIV RNA in 55 samples, resulting in 0.8% of newly diagnosed HIV-infected individuals (95% CI: 0.6–1.0). This showed a negative predictive value (NPV) of 99.2% (95% CI: 99.0–99.4) for rapid HIV testing at the points-of-care. The newly diagnosed HIV infections were detected in all five study clinics, and were slightly higher in pregnant females (0.9%) compared to non-pregnant females and males, 0.7% (p = 0.62) and 0.6% (p = 0.36), respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Algorithm showing enrolment and testing of participants.

Participants’ baseline rapid results were reviewed if they had positive HIV RNA results. Those who had negative rapid test and receiving antiretroviral therapy, and those enrolled by mistake (positive baseline rapid test) were excluded. Ag = antigen, Ab = antibody, W. Blot = Western Blot.

https://doi.org/10.1371/journal.pone.0164943.g001

Serological characterization of NAAT-positive samples

Of the 55 NAAT-positive participants, 52 (95%) tested positive for HIV antibodies on Genscreen HIV ELISA. Western Blot (W. Blot) was positive in 48 (92%) of these antibody-positive participants. P24 antigen testing was performed on 53 participants, and was positive in 12 (23%) participants. The two samples that were insufficient for p24 antigen test were positive on HIV ELISA. Limiting antigen avidity assay identified 16 antibody-positive participants as having early HIV infection, and classified the remainder (n = 36) as having chronic infection. Limiting antigen avidity was repeated on follow-up samples of participants who came for a follow-up visit, and the results remained the same as in initial testing. The combination of HIV RNA, p24 antigen, HIV antibody, LAg avidity and W. Blot test results enabled staging of HIV infections. Amongst those with early HIV infection, few participants were detected very early before the appearance of HIV antibodies, some were detected at the peak of HIV vireamia, and many others were near the time of HIV VL set-point (Fig 2). A significant proportion of HIV-infected individuals were classified as having chronic HIV infection (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Staging of identified HIV infections based on HIV RNA (viral load), p24 antigen, antibodies, limiting antigen (LAg) avidity results and Western Blot (W. Blot).

All the red-blue dots represent participants with HIV infections that were misdiagnosed by rapid HIV tests at the points-of-care. Solid blue line represents HIV viral load during early HIV stage, and dotted blue line represents HIV viral load fluctuation during the chronic stage. Red dotted line represents eclipse stage where all HIV tests are negative. Double red lines represent HIV viral load set-point. Participants with LAg avidity results <1.5 OD-n values had early HIV infections and those with LAg results >1.5 OD-n values had chronic HIV infections. N/A = not applicable, + = positive, - = negative, < = less than, > = greater than, +/- = positive or negative results.

https://doi.org/10.1371/journal.pone.0164943.g002

Serological characterization of NAAT-positive samples showed that 0.04% (95% CI: 0.000–0.001) had acute HIV infection (n = 3); 0.3% (95% CI: 0.1–0.4) had early HIV infection (n = 19, including those with acute infection) (Table 1); and 0.5% (95% CI: 0.5–0.7) had chronic HIV infection (n = 36) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of participants diagnosed with acute or early HIV infection.

https://doi.org/10.1371/journal.pone.0164943.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of participants diagnosed with chronic HIV infection.

https://doi.org/10.1371/journal.pone.0164943.t002

HIV viral loads and CD4 counts of NAAT-positive participants

Median HIV VL was slightly higher in participants with early HIV infection (4.5 log, IQR: 3.7–5.7) compared to those with chronic infection (4.2 log, IQR: 3.8–4.9) (p = 0.33). Most participants (89%) had HIV VLs >1500 copies/ml. CD4 count results were available for 44 participants. Median CD4 count was significantly higher in participants with early HIV infection (434 cells/μl, IQR: 287–706) compared to those with chronic infection (351 cells/μl, IQR: 228–469) (p = 0.03) (Tables 1 and 2).

Repeat rapid testing from stored plasma samples

Rapid tests were repeated in 50 frozen antibody-positive plasma samples with sufficient volumes in the laboratory. Advanced Quality rapid tests were clearly positive in 45 samples, low positive in 2 (faint band noticed), and negative in 3 samples. Abon rapid tests were clearly positive in 48 samples, low positive in 1 sample and negative in 1 sample.

Follow-up of NAAT-positive participants

NAAT-positive participants were recalled to the clinics for follow-up, and they came at different intervals based on their availability. Only 47 (86%) participants came back for follow-up at a median of 4 weeks (IQR: 2–8), and 45 (96%) of them were positive on follow-up rapid tests. Follow-up rapid testing done at the points-of-care improved NPV from 99.2% to 99.8%. Participants were managed according to the SA HIV management guidelines, which recommended immediate initiation of antiretroviral therapy (ART) in pregnant females, and ART eligibility when a CD4 count was <350 cells/μl from 2012–2014 and <500 cells/μl from 2015 [13–16]. Two non-pregnant participants, with confirmed early HIV infection, were still negative on rapid test at a 10 week follow-up interval (Table 1). One of them tested positive on the 4^th generation ELISA, had a low CD4 count of 58 cells/μl, and was initiated on ART. The other participant tested positive on a second follow-up rapid test done 5 months later, and was still not eligible for treatment as his CD4 count was 384 cells/μl (in 2012). All participants with chronic HIV infection had positive HIV ELISA, positive W. Blot and LAg avidity results of >1.5 OD-n values, and all were positive on follow-up rapid HIV testing done at the points-of-care (Table 2). NAAT-positive participants were asymptomatic or had no severe illnesses as they came back to collect results, with the exception of one who was admitted with acute retroviral syndrome.

Frequency of new HIV infections

This study evaluated the detection of early HIV infections in a low resource setting through the use of different HIV assays. The finding of 55 (0.8%) newly diagnosed HIV infections missed by rapid tests shows that a substantial number of HIV-infected individuals are misdiagnosed at the points-of-care in SA. A slightly higher prevalence of these infections in females compared to males was not surprising as HIV prevalence in SA is generally higher in females [17]. The newly diagnosed HIV-infected participants were identified in all the study clinics, highlighting that this could be a bigger problem in SA. Other studies in SA and sub-Saharan African countries have identified individuals with early or chronic HIV infection that were misdiagnosed at the points-of-care [18–23]. This calls for improvement of HIV testing guidelines in HIV hyper-endemic setting in order to facilitate earlier detection of HIV-infected individuals who are misdiagnosed by rapid HIV tests at the points-of-care.

From 2002–2012, SA HIV incidence in people aged 15–49 years ranged between 1.9–2.2% [17]. This maintenance of HIV incidence at around 2% for a 10 year period highlights ongoing transmission of HIV, which is probably driven by highly infectious individuals, particularly those with early HIV infection. Most participants (89%) in this study had HIV viral loads above 1500 copies/ml, which is a known risk factor for sexual transmission of HIV [24]. The 0.3% prevalence of early HIV infections noted in this study is lower compared to a recently published study from SA that showed a prevalence of 1.1% in individuals who were HIV antibody-negative at the point-of-care [21]. This lower prevalence of early HIV infection could be explained by the fact that this study was conducted in an area with a moderate HIV prevalence [17], and at a time where there was higher ART coverage in SA [25].

Earlier detection of HIV-infected individuals misdiagnosed at the points-of-care

The majority of NAAT-positive participants (86%) were successfully recalled to the clinics for further management, showing that it is feasible to detect and manage participants misdiagnosed by rapid HIV testing at the points-of-care. Participants with early HIV infection were detected in all the Fiebig stages of early HIV infection (Fig 2) [26]. The finding that most participants with early HIV infection were near the time of HIV VL set-point could mean that most do not present for HIV testing close to the time of exposure to HIV (Fig 2). Pooled NAAT has been incorporated into the routine screening for HIV in some parts of developed world in order to identify HIV-infected individuals that are misdiagnosed by the rapid tests, particularly those with early HIV infection [27,28]. The availability of resources in the developed world enables detection of these infections, followed by appropriate HIV management [28]. The pooling strategy reduces the costs of NAAT [28]; however, this strategy is expensive to implement for routine diagnosis in low resource settings.

This study showed that follow-up rapid tests done within a 4 week interval are able to detect individuals with early or chronic HIV infection previously misdiagnosed by rapid HIV tests. It is likely that other participants who came for follow-up later than the 4 week interval would have tested positive on rapid HIV tests if they had presented within this interval (Tables 1 and 2). The exception to this was the 2 participants with early HIV infection who were still negative at 10 weeks follow-up (Table 1). Most HIV testing guidelines used in developing countries recommend a follow-up rapid HIV test to be done 12 weeks later after a negative rapid HIV test [16,29,30]. Implementing a 4 week follow-up interval instead of a 12 week follow-up interval would reduce both HIV diagnostic and transmission window periods. Where possible, the 4 week follow-up interval could be incorporated into the existing 12 week follow-up interval to have a 0–4–12 week rapid testing strategy, which could easily be implemented in a setting of frequent clinic visits such as in antenatal clinics. The current WHO HIV testing guidelines recommend follow-up rapid testing at 4–6 weeks only for a minority of individuals who identify a specific recent suspected exposure to HIV [31]. However, it is very difficult to predict or clinically diagnose early HIV infections [5,10]. Hence, a routine follow-up rapid testing at 4 weeks would be a good surveillance tool for early detection of HIV-infected individuals previously misdiagnosed by rapid HIV tests at points-of-care.

Alternative strategies for detecting HIV-infected individuals misdiagnosed at the points-of-care

Parallel testing algorithm, where HIV testing at the points-of-care is done on two different rapid strips concurrently, could be considered for improving the sensitivity of rapid HIV testing at points-of-care. This strategy has been reported to have a higher sensitivity [32,33]. However, some data show that serial testing is equivalent to the parallel testing algorithm with an added cost saving benefit [34,35]. Black et al. have identified HIV-infected individuals who were misdiagnosed at the points-of-care despite the use of the parallel testing algorithm [36]. This highlights that parallel testing may not be the best solution for improving HIV testing at the points-of-care. WHO recommends either serial or parallel testing algorithm for HIV testing at points-of-care [37].

The other alternative strategy for detection of HIV-infected individuals misdiagnosed by rapid tests in resource limited countries would be to screen individuals testing negative on rapid HIV test with the 4^th generation HIV ELISA. However, HIV ELISA is costly and laboratory-based, and thus may also lead to a high rate of loss to follow-up. Molecular tests are now available for use at the points-of-care. These are not readily available in the developing world owing to high costs [38,39]. If NAAT was done at the points-of-care in this study, those misdiagnosed by the rapid HIV test would have been detected and immediately linked to HIV care, without any issues of loss to follow-up.

Implications of misdiagnosing HIV infections at the points-of-care

Advantages of making an accurate diagnosis of HIV infection include opportunities to preserve HIV-specific immune responses if treatment is made available early, and prevent secondary transmission in the population [5]. There is now convincing evidence showing the benefits of early treatment of HIV-infected patients regardless of CD4 counts [40]. Misdiagnosing HIV-infected individuals at the points-of-care means that some people do not benefit from early treatment of HIV infection. This could lead to complications that are associated with delayed ART initiation, such as development of AIDS conditions [40].

Preventing the secondary spread of HIV during the window period is an important public health preventive measure because of extremely high viral loads during this period [41]. Individuals with early HIV infection contribute significantly to the secondary transmission of HIV, and thus play a big role in sustaining new HIV infections [4,6]. Other stages of HIV infection also have a significant contribution to secondary transmission of HIV [6].

Undiagnosed or untreated HIV infection in pregnant women results in a high risk of vertical transmission of HIV, and this risk is much higher with early HIV infection [42,43]. Although prevention of mother to child transmission (PMTCT) programme has led to a remarkable reduction of vertical transmission of HIV in SA and other countries [43,44], more needs to be done to eliminate vertical transmission. Accurate diagnosis of HIV infection during antenatal care creates an excellent opportunity for initiation of PMTCT. Thus, strengthening HIV testing at the points-of-care would contribute towards further decrease and elimination of vertical transmission of HIV.

The high frequency of new HIV infections undermines the global efforts of eliminating new HIV infections. The first 90 of 90-90-90 UNAIDS target might not be properly assessed if there are many HIV-infected individuals that are misdiagnosed by rapid tests at points-of-care in resource-constrained settings. A test and treat strategy might not have a desired impact in low resource setting if point-of-care testing for HIV is not strengthened. Misdiagnosis of HIV infection could lead to inadvertent use of pre-exposure prophylaxis in HIV-infected individuals, thus resulting in emergence and transmission of ARV drug-resistant strains.

Detection of new HIV infections allows a close monitoring of the HIV epidemiology. This monitoring can be used to assess the effectiveness of the existing HIV diagnostic, treatment and prevention policies, and inform necessary changes to such policies.

Possible explanations for misdiagnosis of HIV-infected people at the points-of-care

The finding that the majority of participants (95%) who were misdiagnosed at the points-of-care were positive on HIV ELISA is not surprising as the rapid tests have poorer sensitivity for the detection of HIV antibodies, especially during the early HIV phase [45]. In this study most participants with detectable antibodies had positive W. Blot and also tested positive on repeat rapid testing in plasma. The reason for false negative rapid tests at the points-of-care could be due to human errors during testing. These errors could include insufficient volume of blood and/or incorrect volume of diluent used [31].

Other researchers have previously showed inferior performance of some rapid HIV tests on finger-stick whole blood compared to testing done on serum in a setting where testing was done by skilled personnel [8]. Black et al. showed a remarkably lower sensitivity of rapid tests during a field evaluation done by trained nurses on finger-stick whole blood compared to the laboratory evaluation done on serum samples [36]. A recent study that screened for acute HIV infection in SA also detected a significant number of chronic HIV infections that were missed at the points-of-care [21]. The lower performance of some rapid tests in finger-stick whole blood could be caused by dilution effect from the red blood cells and/or weak antibody binding due to haemolysis [8], and by antigen-antibody complexes predominantly found in early HIV phase before the appearance of free antibodies [41]. These complexes are also found in varying degrees in later stages of HIV infection [46,47]. This shows a need to use rapid HIV tests with higher performance in finger-stick whole blood, as this is a sample type used at the points-of-care. One such test is INSTI, a rapid HIV test approved by the Food and Drug Administration (FDA), which has shown a high sensitivity and specificity in diagnosing HIV infection in whole blood samples [8]. The INSTI rapid test has also been reported to perform well in identifying individuals with early HIV infection [48].

The limitations of this study are that follow-up visits for NAAT-positive participants were done at different intervals and there was no active tracing of their partners. However, all participants identified with HIV infection were encouraged to disclose to their partners, and return with them to the clinics. Repeat rapid testing from plasma samples was not done before NAAT and serology testing. If done earlier, this would have led to faster identification of HIV-infected individuals as most rapid tests were positive in plasma samples.