Cross-reactivity mediated by SARS-CoV-2 antibodies: Mechanism and impact of elevated HIV ELISA false-positive rates in the late-phase COVID-19 pandemic.

In 2006, the Centers for Disease Control and Prevention issued guidelines advocating routine HIV screening for all patients. However, false-positive results are a potential patient care threat for low-risk populations even with accurate screening assays. A reduction in HIV false-positive screening results can potentially be seen by switching from the third-generation to a more sensitive and specific fourth-generation screening assay. We studied the impact on the false-positive screening rate of a change to a fourth-generation assay at a regional US Department of Veteran Affairs Medical Center. HIV screening tests performed by the laboratory from March 1, 2016 to February 28, 2017, prior to implementation of the new assay were compared with fourth-generation HIV screening tests performed from March 1, 2017 to February 28, 2018. Of 7,516 third-generation HIV screening tests reviewed, 52 were reactive on the screening assay; 24 were true positives, 28 were false positives. The following year 7,802 fourth-generation HIV screening tests were performed and 23 were reactive on the screening assay; 16 were true positives and 7 were false positives. The positive predictive value for the third-generation test was 46% and 70% for the fourth-generation test. There were fewer false-positive results with testing with the more specific fourth- vs third-generation assay (0.09% vs 0.37%, respectively), which was statistically significant (P = .002). This reduction in false-positive screening would reduce the laboratory workload and would save an estimated $3,875 yearly and reduce the adverse effects of false-positive screening results for patients.

Fourth-generation HIV assays are now available in Europe. These screening assays allow the combined detection of HIV antigen and antibody (Ag/Ab) in order to reduce the diagnostic window between infection and antibody detection in primary HIV infection (PHI). We report on a case of PHI that was not detected with one of these available fourth-generation HIV assay in spite of the presence of a high p24 antigen level. A 35-year-old man presented to the emergency room with fever, myalgia and skin rash. Serological tests were performed, including a fourth-generation HIV screening test (Vironostika HIV Uni-Form II Ag/Ab assay, Organon Teknika, Boxtel, the Netherlands) and all the results were negative. Seven days later, a second blood sample was performed and tested for HIV antibodies in another laboratory. The HIV antibody test was positive (Ortho HIV1/2 Ab Capture ELISA, Ortho Diagnostic System, NJ, USA). The Western blot (HIV blot 2.2, Genelabs Diagnostics, Singapore) showed antibodies weakly reactive with p24 and gp160 proteins. p24 antigen (Innotest HIV Antigen mAb, Innogenetics, Ghent, Belgium) was positive at a level of 1300 pg/ml. The plasma HIV-1-RNA load (HIV-1 Amplicor Monitor 1.5, Roche Diagnostics, Branchburg, NJ, USA) was greater than 750 000 copies/ml. All these results together supported a diagnosis of PHI. The patient reported being homosexual and had had a risk contact for HIV 2 weeks before the onset of symptoms. The first sample was then re-analysed using different screening and confirmatory tests. Three antibody screening tests were used: the Murex HIV-1.2.O (Murex Biotech, Dartford, UK), the Ortho HIV1/HIV2 Ab Capture ELISA and the Vitros Immunodiagnostic anti-HIV-1+2 (Ortho Clinical Diagnostic, Amersham, UK). Three Ag/Ab combined assays were tested: the Vironostika HIV Uni-Form II Ag/Ab, which was initially non-reactive, the Vidas Duo (BioMérieux, Marcy l'Etoile, France) and the Murex HIV Ag/Ab combination (Murex Biotech). Quantification of p24 antigen was performed using the Innotest HIV Antigen mAb. Western blot was performed using the HIV Blot 2.2. All the tests were performed and interpreted according to the manufacturers’ instructions. The initially negative result obtained with the Vironostika HIV Uni-Form II Ag/Ab was confirmed. The sample, tested undiluted and at five- and 10-fold dilutions remained negative, excluding the possibility of a prozone effect. On the contrary, the quantitative p24 antigen detection assay showed a high level of p24 antigen. Two out of the three antibody screening tests were positive. The complete results are shown in Table 1.Table 1: Results obtained with the different tests on the two consecutive samples. According to the results of Kabamba et al. (see previous letter), the Vironostika Ag/Ab test seems to be the least sensitive of the three Ag/Ab tests they investigated. In our hands, it was even less sensitive than two classic third-generation antibody tests. The reason for this lack of sensitivity is not clear. The level of p24 antigen was high in the first serum sample and should have been detected by the Vironostika Ag/Ab test, which claims to have a detection limit of 30 pg/ml [1]. A prozone effect was excluded by the negative results obtained after serum dilution. Neither can the failure to detect p24 antigen be explained by the genetic variability of the HIV-1 patient strain, which is a B subtype. It is of note that Laperche et al. [2] also observed a negative Vironostika Ag/Ab test result in a seroconversion serum containing 34000 pg/ml p24 antigen that was detected by the other Ag/Ab tests. In our case, as in the case of Kabamba et al. (see previous letter), the positivity of the Vironostika Ag/Ab test corresponded to the appearance of envelope antibodies in the Western blot assay. Glycoprotein 160 antigen is actually the HIV-1 antigen that is used for the detection of HIV-1 antibodies in the Virostika Ag/Ab test. On the contrary, in our case, the positivity of the Vidas Ag/Ab test correlated with the presence of p24 antigen. It thus seems that the failure to detect early primary infection by the Vironostika Ag/Ab test is caused by its low sensitivity in detecting p24 antigen. Moreover Ag/Ab tests have no gag antigen included among the antigens intended to detect HIV antibodies. In the case of the Vironostika Ag/Ab test, this also could explain the delayed diagnosis compared with classic HIV antibody tests such as the Ortho HIV1/2 Ab Capture ELISA, which includes a gag antigen in their design. Consecutively to the incorporation of HIV antigen detection in HIV antibody tests, the antibody detection part of the fourth-generation assays may be less sensitive than that of third-generation assays. Although the concept of fourth-generation HIV tests presents an improvement in the early diagnosis of PHI, as demonstrated by the evaluation of these tests on seroconversion panels [2–4], their use for large-scale screening purposes needs further assessment. Further evaluation of their sensitivity to the different phases of the viral infection and immune response should be carefully evaluated, together and separately, before replacing classic HIV antibody tests with these new tests. In the meantime, considering their disputable sensitivity for p24 antigen detection and maybe also for HIV antibodies that could appear in early seroconversion, these tests should not be used as sole diagnostic tests. In particular, these tests should not replace p24 antigen tests for confirming an HIV antibody-positive result obtained with a screening assay. Marie-Luce Delforgea Nicole Dolléb Philippe Hermansc Corinne Liesnarda

Fourth-generation assays for the simultaneous detection of human immunodeficiency virus (HIV) antigen and antibody that were available on the international market until now have antigen detection modules with relatively poor sensitivity and produce a higher rate of false-positive results than third-generation enzyme immunoassays (EIAs). The new Cobas Core HIV Combi EIA with an improved sensitivity for HIV p24 antigen was compared to alternative fourth- and third-generation assays, the p24 antigen test, and HIV type 1 (HIV-1) RNA reverse transcriptase PCR (RT-PCR). A total of 94 seroconversion panels (n = 709 sera), samples from the acute phase of infection after seroconversion (n = 32), anti-HIV-1-positive specimens (n = 730) from patients in different stages of the disease, 462 subtyped samples from different geographical locations, anti-HIV-2-positive sera (n = 302), dilutions of cell culture supernatants (n = 62) from cells infected with different HIV-1 subtypes, selected performance panels from Boston Biomedica Inc., 7,579 unselected samples from blood donors, 303 unselected daily routine samples, 997 specimens from hospitalized patients, and potentially interfering samples (n = 1,222) were tested with Cobas Core HIV Combi EIA. The new assay showed a sensitivity comparable to that of the Abbott HIV-1 AG Monoclonal A for early detection of HIV infection in seroconversion panels. The mean time delay of Cobas Core HIV Combi EIA (last negative sample plus 1 day) in comparison to that for HIV-1 RT-PCR for 87 panels tested with both methods was 2.75 days. The diagnostic window was reduced with Cobas Core HIV Combi EIA by between 3.6 and 5.7 days from that for third-generation assays. The specificities of Cobas Core HIV Combi EIA in blood donors were 99.84 and 99.85% (after repeated testing). Overall, 30 repeatedly reactive false-positive results out of 10,031 HIV-negative samples were obtained with Cobas Core HIV Combi EIA. Our results show that a fourth-generation assay with improved specificity such as Cobas Core HIV Combi EIA is suitable for blood donor screening because of its low number of false positives and because it detects HIV p24 antigen with a sensitivity comparable to that of single-antigen assays.

False-positive screening results in the European randomized study of screening for prostate cancer

Developing clinically validated models across institutions is limited by data-sharing restrictions and risks of breaches. Models trained at one institution often require retraining at others due to differences in populations and data distributions. Synthetic datasets offer a promising solution by simulating original data without compromising patient privacy, enabling model development with larger, more diverse datasets. As a pilot, we applied this approach to identify false-positive results in fifth-generation HIV screening, which provides semi-quantitative measurements for HIV-1 antibodies (HIV1Ab), HIV-1 antigen (p24), and HIV-2 antibodies (HIV2Ab). Identifying possible false-positive results allows the lab to delay reporting until consulting with the clinician to decide the next steps, which is especially important in low-prevalence groups like prenatal screening. We aimed to determine if synthetic datasets created from 3 different methods could produce HIV false positivity classification models of similar performance to the original data. We performed a retrospective analysis of fifth-generation HIV screening tests (BioPlex 2200) from 2016-2018 (n = 60,632). Clinical HIV infection status was determined through viral load results and chart review for reactive and selected non-reactive samples (n = 580). Samples were classified as confirmed positive (n = 288), presumed negative (n = 59,909), confirmed negative (n = 269), or excluded due to insufficient information (n = 166). A support vector machine (SVM) with a linear kernel was used to learn models to predict HIV status based on screening test values. Data was split into 80% for training and 20% for testing. Models were first trained using the original data. Then, synthetic datasets were created using Gaussian, generative adversarial network, and variational autoencoder methods (Python Library: The Synthetic Data Vault Project), and models were trained for each synthetic dataset. Classification accuracy for false positives (FP: reported as % correctly classified as FP) and true positives (TP: reported as % correctly classified as TP) were calculated for all models. Using the original data the SVM gave a FP classification accuracy of 92% and correctly classified 94% of true positive (TP) cases. The SVM trained on the Gaussian synthetic data achieved a 100% accuracy for classification of FP but 0% accuracy for classification of TP. The GAN and VAE synthetic data yielded a FP classification accuracy of 78% and 71% respectively with a TP classification accuracy of 94% and 100% respectively. Data on patient HIV status is highly sensitive. Sharing it across institutions to create models that can improve follow-up for patients is challenging. Here we found that some synthetic data generation methods created data which performed similarly to the original data when used in training false positive HIV classification models. Work remains to improve the synthetic data sets and enhance model performance.

ObjectiveTo investigate conceptual knowledge about mammographic screening among Norwegian women.DesignWe administered a cross-sectional, web-based survey. We used multiple-choice questions and a grading rubric published by a research group from Australia.SettingOur...

The influence of genetic variability on the sensitivity of serological diagnosis of human immunodeficiency virus (HIV) infection has, to date, been poorly investigated. The aim of the present study was to assess whether fourth-generation assays for the combined detection of HIV antigen and antibodies to HIV (anti-HIV) permit a reduction of the diagnostic window in comparison to third-generation antibody enzyme immunoassays (EIAs), which so far have shown a poor sensitivity for detection of HIV-1 non-subtype B primary infections. Three patients with primary HIV-1 subtype E (CRF01-AE) infection were tested with different third- and fourth-generation assays, stand-alone HIV antigen (Ag) EIAs and reverse transcription-polymerase chain reaction (RT-PCR). Additionally, virus lysates from HIV-1 Group M and O and HIV-2, at concentrations of p24 Ag close to the detection limit of licensed HIV Ag EIAs, were investigated with fourth-generation EIAs and HIV Ag EIAs. In the first blood donor, the most sensitive fourth-generation assay detected HIV-1 infection 11 days earlier than five of the eight third-generation antibody assays. Fourth-generation EIAs, with a high sensitivity for HIV antigen, detected HIV-1 subtype E infection simultaneously or 4 days later than HIV-1 RT-PCR on pooled samples. Low concentrations of virus lysates of different HIV-1 subtypes A-H and group O, tested positive with fourth-generation EIAs, with a high sensitivity of the antigen-detection module. Fourth-generation EIAs, especially those with a high sensitivity for HIV-1 p24 antigen, reduce the diagnostic window for primary HIV-1 subtype E infection in comparison with third-generation antibody-screening assays. These preliminary data from seroconversions and virus lysates indicate that the genetic diversity of HIV-1 does not represent a major challenge for the most sensitive EIAs of this new assay generation.

BackgroundWe have previously reported on natural humoral immunity against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in a Norwegian cohort between 2020 and 2021. In this study, we evaluated long-term humoral (including vaccination-induced) immunity in the same cohort and assessed predictors of high antibody levels against spike protein, as well as the persistence of antibodies against the virus spike and nucleocapsid proteins.MethodsVaccination data and antibody levels against the spike and nucleocapsid proteins were collected at 12 (only in infected participants) and 24 months (in both infected and uninfected participants) after the participants’ first polymerase chain reaction (PCR) tests for the virus. Antibody levels against spike protein at 24 months were categorized as high or low based on the 50th percentile. Possible predictors of high antibody levels against spike protein were examined using univariate and multivariate logistic regression models.ResultsOf 1119 original participants (400 PCR + and 719 PCR −), 574 responded to our questionnaires and were invited to antibody measurements (median age: 51 years; women: 59%). Vaccination data showed that 11% were fully immunized, and 85% were booster-immunized at 24 months. Antibody levels were evaluated in 72% (287/400) of the PCR + participants at 12 months and 58% (233/400) at 24 months. At 12 and 24 months, we observed that 97% (278/287) and 100% (233/233), respectively, still had antibodies against the spike protein, and 86% (248/287) and 95% (221/233), respectively, against the nucleocapsid protein. Antibody levels were also evaluated in 34% (247/719) of those in the PCR − group, which revealed that 99.5% and 69% had detectable antibodies against spike and nucleocapsid proteins, respectively, at 24 months. Irrespective of pre-vaccination SARS-CoV-2 infection status, the booster-immunized participants were 3.7 × more likely to have high antibody levels against spike protein vs the non-booster-immunized ones. Those aged > 60 years had the highest median antibody levels against the spike protein and were more likely to be booster-immunized.ConclusionsOur findings highlight the benefits of booster vaccinations for humoral immune responses. Long-term antibody levels against the SARS-CoV-2 spike protein were higher in booster-immunized participants vs the non-booster-immunized, irrespective of pre-vaccination infection status.Trial registration.146,469: The COVID-19 study in Telemark and Agder—COVITA. ClinicalTrials.gov ID: NCT04514003.

Evaluation of the virtues and pitfalls in an HIV screening algorithm based on two fourth generation assays – A step towards an improved national algorithm

Passive immunization against COVID-19 by anti-SARS-CoV-2 spike IgG in commercially available immunoglobulin preparations in severe antibody deficiency

SARS-CoV-2 antibody response eight months after vaccination with mRNA vaccines. Influence of prior SARS-CoV-2 exposure

An urgent need for introduction of fourth-generation Ag–Ab based EIA for detection of HIV infection

Editor's evaluation: Cross-reactive antibodies after SARS-CoV-2 infection and vaccination

Current SARS-CoV-2 vaccines are losing efficacy against emerging variants and may not protect against future novel coronavirus outbreaks, emphasizing the need for more broadly protective vaccines. To inform the development of a pan-coronavirus vaccine, we investigated the presence and specificity of cross-reactive antibodies against the spike (S) proteins of human coronaviruses (hCoV) after SARS-CoV-2 infection and vaccination. We found an 11- to 123-fold increase in antibodies binding to SARS-CoV and MERS-CoV as well as a 2- to 4-fold difference in antibodies binding to seasonal hCoVs in COVID-19 convalescent sera compared to pre-pandemic healthy donors, with the S2 subdomain of the S protein being the main target for cross-reactivity. In addition, we detected cross-reactive antibodies to all hCoV S proteins after SARS-CoV-2 vaccination in macaques and humans, with higher responses for hCoV more closely related to SARS-CoV-2. These findings support the feasibility of and provide guidance for development of a pan-coronavirus vaccine.

Decision letter: Cross-reactive antibodies after SARS-CoV-2 infection and vaccination