Using a Multitest Algorithm to Improve the Positive Predictive Value of Rapid HIV Testing and Linkage to HIV Care in Nonclinical HIV Test Sites

Evaluation of a two-test strategy for HIV screening in a low-prevalence setting and the indications for optimizing clinical management

Objectives

To evaluate a two-test strategy for HIV screening in the low-prevalence population and to assess the feasibility of utilizing the optimal signal-to-cutoff (S/CO) threshold on the chemiluminescence immunoassay(CMIA) and an additional rapid test on the gold immune-chromatography assay (GICA) for screening positive patients and optimization of clinical management.

Methods

We conducted a retrospective study of samples analyzed by the fourth-generation Architect HIV Ag/Ab combo assay (CMIA) in a large medical center between June 2017 and August 2020. Reactive samples underwent a second screening test using the rapid test GICA, followed by Western blot (WB) as the confirmatory test. Receiver operating characteristic (ROC) curve analysis was used to determine the optimal S/CO. We calculated sensitivity, specificity, and predictive value based on our population. The performance of the single-test strategy (CMIA) was compared with that of the two-test strategy (CMIA and GICA). Logistic regression was used to analyze the factors of clinical characteristics leading to false positive results.

Results

A total of 220558 samples were screened by CMIA, and 429 patients met the inclusion criteria. Of these, CMIA produced 199 false-positive results with a median S/CO of 1.93(IQR1.45-3.68) and 230 positive results with a median S/CO of 455.1 (IQR169.3-709.7). The optimal S/CO of the single-test strategy was 8.82, which achieved a sensitivity of 100% and a positive predictive value (PPV) of 90.9%. The two-test strategy (CMIA and GICA) provided a sensitivity of 100% and a PPV of 98.7%, which best correlated with the confirmatory test WB. The combination of S/CO 8.82 on the CMIA assay and additional test results of GICA can be defined as four types used to interpret HIV serostatus. The false positive rate (FPR) was high in the female, the age≤18 group, the pre-operative patients, and the patients from the clinical departments of Pediatrics, Gynecology and Obstetrics, and Oncology, etc.

Conclusions

The false positive rate is high in the low-prevalence setting by using CMIA. The two-test strategy (CMIA and GICA) is recommended for HIV screening in hospitals. Hopefully, the clinicians will be able to interpret HIV serostatus and facilitate clinical decision-making while waiting for the confirmatory results.

Seroconversion, seroreversion, and serowaffling among participants initiating antiretroviral therapy in Project DETECT

BACKGROUND

Incomplete HIV seroconversion and seroreversion are increasingly documented by testing and pre-exposure prophylaxis programs more than previously recognized. This analysis reports on incomplete seroconversion and seroreversion by specimen and test type among Project DETECT participants.

METHODS

Project DETECT included a longitudinal study of point-of-care tests. Participants were categorized as having "incomplete seroconversion" if all timepoints had ≥1 nonreactive test at study censoring. Among participants with incomplete seroconversion, we defined "seroreversion" as sustained regression to nonreactive for any test following a reactive result. We define "serowaffling" as any reactive result followed by a nonreactive and then reactive result. We used Fisher's exact tests to explore relationships between Fiebig stage at ART initiation and incomplete seroconversion, seroreversion, and serowaffling.

RESULTS

Twenty of 1940 Project DETECT participants met criteria for this subset. Ten participants had complete seroconversion after a median of 23 (IQR 16-47) days following initial positive tests. Ten participants had incomplete seroconversion, eight of whom had seroreversion. Incomplete seroconversion with persistent nonreactive tests was seen only with oral fluid (OF). Of eight participants with seroreversion, all experienced seroreversion of OF tests if the test was ever reactive (n = 6); seroreversion occurred in fingerstick and venipuncture tests in two participants. Serowaffling occurred in nine (45%) participants. No associations were seen between Fiebig stage at ART start and complete seroconversion, seroregression, or serowaffling in our sample.

CONCLUSIONS

OF tests may be particularly susceptible to providing false-negative results. Seroreversion and incomplete seroconversion among individuals on antiretroviral treatment may represent a growing problem for HIV testing and treatment programs.

Time From HIV Infection to Diagnosis in the U.S., 2014-2018

INTRODUCTION

Understanding the role of sociologic, structural, and biomedical factors that influence the length of time from HIV infection to diagnosis and reducing the time from infection to diagnosis are critical for achieving the goals of the Ending the HIV Epidemic initiative. In a retrospective analysis, the length of time from HIV infection to diagnosis and its association with individual- and facility-level attributes are determined.

METHODS

Data reported by December 2019 to the U.S. National HIV Surveillance System for people with HIV diagnosed during 2014-2018 were analyzed during December 2020. A CD4 depletion model was used to estimate the time from HIV infection to diagnosis.

RESULTS

During 2018, the median time from HIV infection to diagnosis was shorter for those infections diagnosed using the rapid testing algorithm (30.3 months, 95% CI=25.5, 34.5) than those diagnosed using the recommended (41.0 months, 95% CI=39.5, 42.0), traditional (37.0 months, 95% CI=29.5, 43.5), or other (35.5 months, 95% CI=32.5, 38.0) diagnostic testing algorithms. From 2014 to 2018, the time from HIV infection to diagnosis remained stable overall for all testing methods except for the traditional diagnostic testing algorithm. In multivariate analyses, those more likely to have HIV diagnosed closer to the time of infection were younger, were White, had transmission risk factors of injection drug use or heterosexual contact (for female individuals) or male-to-male sexual contact and injection drug use, or had HIV diagnosed at a correctional or screening facility (p<0.01).

CONCLUSIONS

Providing access to expanded testing, including rapid testing in nonclinical settings, is likely to result in a decrease in the length of time a person is unaware of their HIV infection and thus reduce onward transmission of HIV infection.

Rapid Testing Algorithm Performance in a Low-Prevalence Environment

BACKGROUND

The performance of a statewide HIV rapid test algorithm (RTA) in a low-prevalence setting (0.71%) was examined for 3 years.

METHODS

An initial rapid screening by HIV-1/2 Ag/Ab Combo test (RT#1) with Ab verification using a second, different rapid test (RT#2) was conducted. Clinic referral was immediate for antigen-only-positive screens. Antibody-positive screens were confirmed by RT#2. Specimens were collected following discordant RTA results (initially Ab-POS by RT#1, but negative on RT#2) and tested in accordance with the current Centers for Disease Control and Prevention/Association of Public Health Laboratories-based HIV diagnostic algorithm supplemented by a quantitative viral load whenever possible.

RESULTS

Of 310,785 tests performed, 2400 preliminary positive screens were identified; 2191 (91.8%) confirmed by RT#2. Of 13 Determine Combo AG-POS results identified, only 1 confirmed positive. Of the remaining 196 discordant results, 182 (92.9%) were uninfected, including 13 with AG-POS/AB-POS results. Of 14 true positives (7.1%) identified after discordant RTA results, the average quantitative HIV-1 viral load was 277,385 copies/mL, but 5 (35.7%) of 14 had viral loads <1000 copies/mL. Among the 2191 "presumptive positive" by RTA, 3 false-positive (FP) RTAs were reported (both rapid tests having positive results, while the HIV-1/2 Ag/Ab assay and quantitative HIV-1 viral load showed negative results).

CONCLUSIONS

The RTA was effective in predicting true-positive HIV test results and facilitating linkage to care. Discordant results were infrequent. Fingerstick DC Ag detection identified a single early infection. Many discordant cases that were subsequently positive were associated with viral loads <1000 copies/mL.

Differences among diagnostic testing algorithms in the time from HIV diagnosis to care

BACKGROUND

The association between the type of diagnostic testing algorithm for HIV infection and the time from diagnosis to care has not been fully evaluated. Here we extend an earlier analysis of this association by controlling for patient and diagnosing facility characteristics.

STUDY DESIGN

Descriptive analysis of HIV infection diagnoses during 2016 reported to the National HIV Surveillance System through December 2017. Algorithm type: traditional = initial HIV antibody immunoassay followed by a Western blot or immunofluorescence antibody test; recommended = initial HIV antigen/antibody immunoassay followed by HIV-1/2 type-differentiating antibody test; rapid = two CLIA-waived rapid tests on the same date.

RESULTS

In multivariate analyses controlling for patient and diagnosing facility characteristics, persons whose infection was diagnosed using the rapid algorithm were more likely to be linked to care within 30 days than those whose infection was diagnosed using the other testing algorithms (p < 0.01). The median time to link to care during a 30-day follow-up was 9.0 days (95% CI 8.0-12.0) after the rapid algorithm, 17.0 days (95% CI 17.0-18.0) after the recommended algorithm, and 23.0 days (95% CI 22.0-25.0) after the traditional algorithm.

CONCLUSIONS

The time from HIV diagnosis to care varied with the type of testing algorithm. The median time to care was shortest for the rapid algorithm, longest for the traditional algorithm, and intermediate for the recommended algorithm. These results demonstrate the importance of choosing an algorithm with a short time between initial specimen collection and report of the final result to the patient.

Trends in testing algorithms used to diagnose HIV infection, 2011-2015, United States and 6 dependent areas

BACKGROUND

In 2014 the Centers for Disease Control and Prevention (CDC) and the Association of Public Health Laboratories (APHL) issued updated laboratory testing recommendations for the diagnosis of HIV infection.

OBJECTIVES

To examine trends in the use of HIV diagnostic testing algorithms, and determine whether the use of different algorithms is associated with selected patient characteristics and linkage to HIV medical care.

STUDY DESIGN

Analysis of HIV infection diagnoses during 2011-2015 reported to the National HIV Surveillance System through December 2016. Algorithm classification: traditional = initial HIV antibody immunoassay followed by a Western blot or immunofluorescence antibody test; recommended = initial HIV antibody IA followed by HIV-1/2 type-differentiating antibody test; rapid = two CLIA-waived rapid tests on same date.

RESULTS

During 2011-2015, the percentage of HIV diagnoses made using the traditional algorithm decreased from 84% to 16%, the percentage using the recommended algorithm increased from 0.1% to 64%, and the percentage using the rapid testing algorithm increased from 0.1% to 2%. The percentage of persons linked to care within 30 days after HIV diagnosis in 2015 was higher for diagnoses using the recommended algorithm (59%) than for diagnoses using the traditional algorithm (55%) (p < 0.05).

CONCLUSIONS

During 2011-2015, the percentage of HIV diagnoses reported using the recommended and rapid testing algorithms increased while the use of the traditional algorithm decreased. In 2015, persons with HIV diagnosed using the recommended algorithm were more promptly linked to care than those with diagnosis using the traditional algorithm.