Fourth‐Generation HIV Rapid Tests: Enhanced Sensitivity and Reduced Diagnostic Window for HIV‐1 Primary Infection Screening

ABSTRACT

As most HIV rapid tests (HRT) detect only HIV‐1/2 antibodies, their performance during primary HIV infection is poor. Determine HIV Early detect (Abbott) (Determine) is the only HRT with an HIV‐1 p24‐antigen detection, but the impact of this addition in shortening the diagnostic window remains unclear. A total of 183 HIV‐1 primary infection samples were tested using the HRTs Determine and ONE STEP anti‐HIV (1&2) Test (InTec Products) (One‐Step). The pre‐seroconversion subgroup was defined as p24‐antigen positivity without Western blot nor Liaison XL (fouth generation enzyme immunoassay with distinct signal for p24‐antigen and HIV‐1 antibody) HIV‐1 antibodies. Global sensitivity (95% CI) was 95% (91–97) for Determine versus 80% (74%–85%) for One‐Step (difference p = 1.38e−06). Pre‐seroconversion subgroup sensitivity was lower, at 71.9 (54.6%‐84.4%) for Determine and 9.7% (3.3%–24.9%) for One‐Step. Among the 45 samples with an HIV‐1 infection date, no HRT was reactive up to 2 weeks. Between 2 and 3 weeks, Determine sensitivity was 78% (45%–95%) versus 56% (27%–81%) for One‐Step. From 3 weeks to 1 month Determine sensitivity was 90% (62%–98%) and One‐Step 45% (21%–72%). The last negative sample occurred at 3 weeks for Determine versus 70–90 days for One‐Step. HRT with p24‐antigen detection significantly shortens the diagnostic window from approximatively 3 months to 1 month. HRTs should be used with caution in the first month after HIV infection.

1. Introduction

Since the release of the first FDA‐approved HIV screening assay in 1985, four generations have been developed, each associated with a technical breakthrough [1]. The first generation was based on virus‐infected cells. The second and third relied on recombinant antigens and synthetic peptides, improving both sensitivity and specificity [2]. Finally, the fourth generation introduced the detection of HIV‐1 p24‐antigen, which shortened the diagnostic window from 3 to 6 weeks to around 2 weeks postinfection [3, 4, 5, 6, 7].

As most HIV rapid tests (HRTs) are considered third generation assays, they perform poorly in a primary HIV‐1 infection setting, with sensitivities as low as 70% [8, 9]. Determine HIV Early detect (Determine), the first FDA‐approved fourth generation HRT, was designed to address this issue, with the independent detection of p24‐antigen and HIV‐antibodies. However, there have been few studies with limited sample collections performed regarding its sensitivity in this context [10, 11, 12, 13], and most of them have been conducted on US‐based samples, which may limit their applicability to HIV‐1 subtype‐B samples [14].

Recently, the development of Pre‐exposure Prophylaxis (PrEP) has raised new interest in the relevance of HRTs during HIV‐1 primary infection, as PrEP‐associated HIV diagnosis may rely mostly on serology, even for non‐TDF/FTC PrEP [15, 16].

The objective of this study was to compare the sensitivity and diagnostic windows of the only FDA‐approved fourth generation HRT Determine with a third generation HRT, ONE STEP anti‐HIV (1&2) Test (InTec Products) (One‐Step).

2. Methods

2.1. Sample Collection

Primary HIV‐1 infection samples were collected as part of the routine activities of the Virology laboratory of the Pitié Salpêtrière Hospital (Paris, France). HIV screening was performed using two enzyme immunoassays (EIAs): Alinity I HIV Ag/Ab Combo assay (Abbott, Rungis, France) and, in case of positivity, Liaison XL Murex HIV Ag/Ab (DiaSorin, Antony, France). The Western blot New Lav Blot I (Bio‐Rad laboratories, Marnes‐la‐Coquette, France) was used as a confirmatory assay. HIV‐1 viral load was assessed using Cobas HIV‐1 (Roche Diagnostics, Manheim, Germany) and quantitation of p24‐antigen using the EIA Vidas HIV p24 II (BioMerieux, Marcy l'Etoile, France) (Vidas). Viral loads, subtypes, and recombinant forms of HIV‐1 were assessed as previously described [17, 18]. HIV primary infection was classified into three subgroups based on a simplified Fiebig staging [5]. First, the pre‐seroconversion subgroup, defined as positive p24‐antigen on Vidas or Liaison XL with negative Western blot and negative antibody S/CO signal on Liaison XL, or positive HIV‐1 RNA and negative EIA (Stage Fiebig I). Second, an early seroconversion group, defined by a positive p24‐antigen (Liaison XL or Vidas) with a Western blot pattern compatible with HIV‐1 primary infection and Liaison XL positive for HIV antibodies [19]. Finally, the seroconversion subgroup, defined by a negative p24‐antigen with an HIV‐1 primary infection compatible Western blot and Liaison XL positive for HIV antibodies. HIV‐1 primary infection‐compatible Western blots were defined according to Fiebig's criteria as follows: they consistently presented as incomplete Western blots, either showing only one band among p24, gp41, or gp120/160, or showing at least two bands from the same set (p24, gp41, gp120/160) [20, 21, 22].

Presumptive dates of HIV‐1 infection and prior PrEP use were collected from medical records. Samples were stored at −20°C before use.

2.2. Samples Processing

Samples were analyzed within the same freeze‐thaw cycle using the fourth generation HIV‐screening Determine HIV Early detect (Abbott, Rungis, France) (Determine) and the third generation HIV‐screening One Step anti‐HIV (1&2) (InTec Products, Xiamen, China) (One‐Step). Briefly, both tests are immunochromatography assays. HIV‐specific antibody detection is based on gp41 and gp36 recombinant antigens for both assays, combined with a synthetic peptide of the gp36 for One‐Step [23]. Samples were processed according to the respective manufacturer's guidelines, with results read by two unblinded operators after 40 min for Determine and 20 min for One‐Step. All suspected negative or trace test lines (faint or barely visible line) were adjudicated by a third, blinded operator. Samples with an invalid result were repeated once.

2.3. Statistical Analysis

Statistical analyses and figures were performed using R (R Foundation for Statistical Computing) or GraphPad Prism (GraphPad Software, Inc). Univariate analyses were performed using the Wilcoxon rank sum test for continuous variables and the McNemar test for categorical variables, with p < 0.05 considered to be statistically significant. Confidence intervals were calculated using Wilson's confidence interval for proportions. We determined a priori that considering a 90% sensitivity for each test, 180 samples would allow a clinically relevant 4% accuracy.

2.4. Ethics

This work was approved by the French Infectious Disease Research Ethics Board (IRB00011642), no. 2024‐0107.

3. Results

A total of 183 samples were included in this study. HIV‐1 genetic diversity was relatively high, with approximately 33% of HIV‐1 subtype B, 18% CRF02_AG, and 33% of other subtypes and circulating recombinant forms (Figure 1). One pre‐seroconversion sample was repeatedly invalid by One‐Step and was excluded from further analysis for this assay. The overall sensitivity was 95.1% 95% CI [90.9, 97.4] for Determine and 80.2% 95% CI [73.8, 85.4] for One‐Step. As a consequence, Determine had a significantly higher sensitivity (p = 1.38e−06, using McNemar Test) compared to One‐Step for HIV primary infection screening. As expected, there was an upward trend in sensitivity from pre‐seroconversion samples, with a 71.9% 95% CI [54.6, 84.4] and 9.7% 95% CI [3.3, 24.9] sensitivities for Determine (n = 32) and One‐Step (n = 31), respectively, to 100% 95% CI [95.5, 100] and 92.7% 95% CI [84.9, 96.6] for the early seroconversion subgroup (n = 82), and to 100% 95% CI [94.8, 100] and 97.1% 95% CI [90.2, 99.2] for the seroconversion subgroup (n = 69). The results are summarized Figure 2A. Notably, Determine performed significantly better in both pre (p = 8.57e−05) and early (p = 0.041) seroconversion subgroups. The proportion of reactive results with a unique trace line (either Ag or Ab for Determine) was 12.5% (4/32) and 6.4% (2/31) for Determine and One‐Step, respectively, in the pre‐seroconversion subgroup. These proportions were stable at 6.1% (5/82) and 9.8% (8/82) in the early seroconversion subgroup, but lower at 0% (0/69) and 1.4% (1/69) for the seroconversion subgroup.

Figure 1.

HIV‐1 subtypes and Circulating Recombinant Forms of serum samples included in this study.

Figure 2.

Determine and One‐Step sensitivity (%) according to seroconversion subgroups (A) or time from infection (B).

Of note, the results for confirmed non‐subtype‐B samples were only remarkably consistent with overall results. For the pre‐seroconversion subgroup, Determine had a sensitivity of 66.7% 95% CI [47.4, 82.8] and One‐Step a 9.5% 95% CI [2.7, 28.9], which increased to 100% 95% CI [90.8, 100] (Determine) and 92.1% [79.2, 97.3] (One‐Step) for the early seroconversion subgroup, and to 100% 95% CI [90.1, 100] (Determine) and 94.3% [81.4, 98.4] (One‐Step) for the seroconversion subgroup.

Estimated dates of infection were available for 45 (24%) patients (Figure 2B). During the first 2 weeks after infection, no assay was reactive (n = 2). Between 2 and 3 weeks after infection, Determine accurately identified 78% 95% CI [45%–94%] (n = 7/9) of the positive samples, while One‐Step identified 56% 95% CI [27%–81%] (n = 5/9) of them. From 3 weeks to 1 month, Determine accurately identified 90% 95% CI [62%–98%] of them (n = 10/11), whereas One‐Step identified only 45% 95% CI [21%–72%] (5/11). Beyond a month, Determine accurately identified all of them (100% sensitivity, 95% CI [86%–100%], n = 23/23), while One‐Step was nonreactive for two samples with alleged infection dates of about 1 month and 70–90 days, respectively (91% sensitivity 95% CI [73%–98%], n = 21/23). Overall, the last negative sample occurred at 3 weeks for Determine versus 70–90 days for One‐Step (Figure 2B).

Of note, six of the 183 serum samples included matched four cases of HIV infection among TDF/FTC PrEP users, in all cases due to low treatment adherence. Reported time of infection was available for two of them (three samples) and ranged from 20 to 30 days. Both rapid tests were antibody‐reactive on all six serum samples.

Several reports have highlighted an elevated p24 limit of detection threshold for the Determine assay compared to EIAs [24, 25]. In the present study, p24‐antigen was assessed in a quantitative arbitrary S/CO unit with Liaison XL and in pg/mL with Vidas. After excluding Vidas samples above the limit of quantitation, both assays correlated (r² = 0.81, p < 0.001, Figure 3A) but moderately correlated with viral loads, with a r² = 0.45 (p < 0.0001) and 0.63 (p < 0.0001) for Vidas and Liaison, respectively (Figure 3B,C). Determine p24‐antigen was only reactive with highest viral loads, almost consistently above 6 log copies/mL, or high p24 levels (Figure 3D, Supporting Information S1: Figure 1). Determine p24‐antigen median of positivity was 210.5 pg/mL (IQR [65.35; > 400 pg/mL]). In contrast to this disappointing limit of detection, both Determine and One‐Step had a low limit of detection for HIV antibodies, close to the Liaison XL (Supporting Information S1: Figure 2). Of note, there was no S/CO statistical difference between Determine and One‐Step for negative, trace, low reactive, and reactive subgroups.

Figure 3.

p24 antigen determination in serum is dependent on the assay and is poorly correlated with HIV viral loads. Correlation between Vidas HIV p24 II and Liaison XL p24 (A), between Vidas p24 antigen and Viral load (B), between Liaison XL and Viral load (C), and between Determine and Viral load (D). p24 values above limit of quantification were excluded from the figures. p24 was expressed in pg/mL for the Vidas assay, in a quantitative S/CO value for the Liaison XL assay, and as a visual intensity for Determine. Viral loads are expressed as log(copy/mL). Comparisons were performed using Wilcoxon rank sum (D), and the Spearman coefficient was used for correlation (A).

4. Discussion

Overall, this study highlights the relevance of a fourth HRT generation assay for primary HIV infection screening, with both a higher sensitivity and a shorter diagnostic window compared with a third generation HRT assay, despite an elevated detection threshold for p24‐antigen compared to fourth generation EIA.

Study comparisons of HIV sensitivity in the context of primary infection should always be treated with caution. As emphasized in this study, sensitivity is strongly influenced by the proportion of the pre‐seroconversion subgroup and by how pre‐seroconversion is defined. This definition may be based solely on a negative Western blot [7, 26], which has a relatively high antibody detection threshold, or on a combination of a negative Western blot with a negative‐antibody fourth‐generation EIA, an assay with a low detection threshold. To our knowledge, no study has addressed the sensitivity of the One‐Step assay in this subgroup using this stringent criterion. However, its sensitivity is in line with previous reports from other third generation rapid tests, such as the VIKIA HIV1/2 rapid test (BioMérieux), which was nonreactive on 0% (n = 2) and 15% (n = 13) samples, or the Autotest VIH (AAZ‐Mylan), which had a sensitivity of 0% (n = 13) [27, 28]. Using the other reported definition of a negative Western blot only, third generation HRTs tended to have significantly higher sensitivities, ranging from 7% to 57% [9]. However, these promising results can be misleading, as of Western‐blot negative/antibody‐reactive EIA samples could account for a nonnegligible proportion, up to 80% of the samples tested [9, 29]. With a sensitivity of 71.9% for the Determine assay in the pre‐seroconversion subgroup with a fourth generation EIA and/or positive NAAT as the reference assay, our results are within the ranges of previous studies: 65% in 23 samples [30], 54% in 13 samples [28], 45.5% in 11 samples [11] as summarized in Figure 4. Of note, our higher sensitivity could be explained by a higher proportion of RNA+/EIA− samples in other studies.

Figure 4.

Comparison of pre‐seroconversion subgroup sensitivity of the Determine assay with previous studies.

Few studies have related reactive rapid tests to the diagnostic window. Pavie et al. reported a window period of at least 2 months for five third‐generation rapid tests in two patients [31]. Delaney et al. using commercial seroconversion panels, estimated that the median time from HIV exposure to reactivity was around 18 days for HIV 1/2 Ag/Ab Architect combo (Abbott), a reference EIA assay, 19.2 days for Determine and 26 to 32 days for third generation HRTs [7]. Although our results agree with their estimates, they warn of late false nonreactive samples, which occurred in our study about a week later for Determine and about a month for One‐Step.

Conflicting results have been reported for the Determine p24‐antigen sensitivity in clinical samples, with sensitivities ranging from 23% to 91% compared with laboratory‐based EIAs [10, 29, 30]. In the present study, we chose a 40‐min reading to both maximize p24‐detection and get as close as possible to real‐life settings, where clinicians would use the maximum incubation time to increase the likelihood of detecting a primary HIV infection. Despite our attempt to optimize this parameter, the p24 component was only 55% (63/114) sensitive compared to Liaison XL or Vidas, and was almost exclusively reactive at the highest viral loads or p24 titers. This disappointing result may be due to the high HIV subtype diversity in our samples, with widely varying p24 detection limit for non‐B subtypes, as previously described EIA limitation [32]. Thus, unlike antibody detection, which tends to the limit of detection of EIA assays, further developments are needed to increase the analytical sensitivity of HRT p24‐antigen. Pending these developments, an alternative could be the use of a point‐of‐care NAAT test, which, despite a much higher cost, could be especially useful in the first 2 weeks following a potential HIV‐1 infection [33]. This study has several limitations. Sensitivity was performed on thawed sera, which might lead to an overestimation compared to whole blood [31]. As a consequence, point‐of‐care sensitivity might be lower than laboratory‐based testing. Moreover, we used One‐Step as a surrogate for all third generation HRTs. We have previously reported sensitivities for most CE‐marked third generation HRT [9] on a smaller panel of other seroconversion samples. As Determine and One‐Step were the two main HRT missing in this previous study, there were selected to be included in the present study. Importantly, the One‐Step sensitivity matched well with other third generation HRTs, which may suggest that their lack of sensitivity in this setting is directly related to a lack of HIV antibodies.

5. Conclusion

Despite an elevated threshold for p24 antigen detection, Determine, the first fourth generation HRT has a higher sensitivity and a shorter diagnostic window compared to third generation HRTs.

Author Contributions

Vincent Guiraud and Agnès Gautheret‐Dejean: conceptualization, methodology, validation, data curation, formal analysis, resources, writing–original draft and reviewing. Angèle Naizet: methodology, and data curation. Habiba Khan, Ghizlane Benhafoun, Pierre Hernandez, Luigi Piccin, Agnès Pichon, and Ay Ling Leng: data curation. Léna Yousfi: data curation, writing–review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting information.

Supporting information.

Data Availability Statement

The data that support the findings are available as a Supporting Information Table.