Longitudinal patterns in indeterminate HIV rapid antibody test results: a population-based, prospective cohort study

ABSTRACT Rapid HIV tests are critical to HIV surveillance and universal testing and treatment programs. We assessed longitudinal patterns in indeterminate HIV rapid test results in an African population-based cohort. Prospective HIV rapid antibody test results, defined by two parallel rapid tests, among participants aged 15–49 years from three survey rounds of the Rakai Community Cohort Study, Uganda, from 2013 to 2018, were assessed. An indeterminate result was defined as any weak positive result or when one test was negative and the other was positive. A total of 31,405 participants contributed 54,459 person-visits, with 15,713 participants contributing multiple visits and 7,351 participants contributing 3 visits. The prevalence of indeterminate results was 2.7% (1,490/54,469). Of the participants with multiple visits who initially tested indeterminate (n = 591), 40.4% were negative, 18.6% were positive, and 41.0% were indeterminate at the subsequent visit. Of the participants with two consecutive indeterminate results who had a third visit (n = 67), 20.9% were negative, 9.0% were positive, and 70.2% remained indeterminate. Compared to a prior negative result, a prior indeterminate result was strongly associated with a subsequent indeterminate result [adjusted prevalence ratio, 23.0 (95% CI = 20.0–26.5)]. Compared to men, women were more likely to test indeterminate than negative [adjusted odds ratio, 2.3 (95% CI = 2.0–2.6)]. Indeterminate rapid HIV test results are highly correlated within an individual and 0.6% of the population persistently tested indeterminate over the study period. A substantial fraction of people with an indeterminate result subsequently tested HIV positive at the next visit, underscoring the importance of follow-up HIV testing protocols. IMPORTANCE Rapid HIV tests are a critical tool for expanding HIV testing and treatment to end the HIV epidemic. The interpretation and management of indeterminate rapid HIV test results pose a unique challenge for connecting all people living with HIV to the necessary care and treatment. Indeterminate rapid HIV test results are characterized by any weak positive result or discordant results (when one test is negative and the other is positive). We systematically tested all participants of a Ugandan population-based, longitudinal cohort study regardless of prior test results or HIV status to quantify longitudinal patterns in rapid HIV test results. We found that a substantial fraction (>15%) of participants with indeterminate rapid test results subsequently tested positive upon follow-up testing at the next visit. Our findings demonstrate the importance of follow-up HIV testing protocols for indeterminate rapid HIV test results.

individuals with treatment for their own individual health and slowing the spread of HIV.Rapid HIV antibody tests are quicker and easier to use than previous methods that required clinicians to obtain and send participant blood samples to centralized testing centers (1).Rapid HIV tests are now among the most widely used diagnostic tests globally.In 2021, 176 million rapid HIV tests were distributed worldwide (M.S. Jamil, WHO Personal Communication 2022).Following a positive rapid HIV test result, confirmatory testing is required to diagnose people with HIV.Among high-risk and high-burden populations, individuals with a negative rapid HIV test result are recommended to be re-tested at a follow-up time point (2).The preferred subsequent timepoint is dependent upon risk behaviors, known HIV exposure, or pregnancy in high-burden areas (2).With the increased availability and efficiency of these tests, the World Health Organization (WHO) recommends using rapid HIV tests, as they provide same-day test results and expedite HIV diagnoses and linkage to care (1).
As of 2021, the WHO guidelines for HIV testing recommend the use of multiple rapid tests (or enzyme immunoassays) to diagnose individuals with HIV (1).The number of parallel reactive rapid HIV tests required for official diagnosis varies by country preva lence of HIV (1).A three-test strategy is recommended in settings of HIV prevalence under 5%, and a two-test strategy is satisfactory in countries above this threshold (1).In either protocol, the first test should prioritize sensitivity, while the second and third tests should prioritize specificity, ultimately having an algorithm with at least 99% positive predictive value (the combination of tests should also have ≥99% sensitivity and ≥98% specificity) (1).The eight most common commercial rapid HIV tests have high sensitivity estimates ranging from 99.1 to 100% and high specificity estimates ranging from 98.9 to 100% (3).Given the potential for conflicting results from parallel rapid tests, the WHO provided additional guidelines for retesting in the event of an indeterminate rapid test result (2).An indeterminate rapid result is when one test is negative and the other positive or the inclusion of any weak positive result (2).It is recommended that the sample should be immediately tested again with the same testing algorithm, and if the results remain indeterminate, a new sample should be tested in 2 weeks (i.e., retesting) (2).If the individual continues to provide indeterminate rapid test results after 2 weeks, previous guidance from the WHO suggests it should be referred to a higher testing facility for confirmatory testing (2).
A lack of confirmatory testing and contradictory rapid test result interpreta tions can result in the underreporting of new HIV cases and in false-positive HIV diagnoses (4); thus, confirmational and follow-up testing is recommended to improve diagnostic accuracy (2).In tandem with additional testing, counseling services have been used to help identify participants who are most at risk and in need of additional monitoring, as indeterminate results may occur during HIV seroconversion (5,6).The persistence of indeterminate results may highlight a prolonged window of potential transmission and development of HIV infection (6).However, the popula tion-based frequency and determinants of indeterminate rapid HIV test results are not well understood.This information may help optimize protocols for handling indeterminate results for HIV surveillance and diagnostic purposes.Accordingly, it is essential to understand the population-based prevalence of and patterns in indeterminate rapid HIV test results.
Our study aimed to describe the prevalence of and longitudinal patterns in indeterminate rapid HIV test results among a population-based cohort of individuals in Rakai, Uganda.Specifically, we examined rapid HIV test results among 54,469 person-visits from 31,405 individuals across 3 rounds of the Rakai Community Cohort Study (RCCS).Prospective population-based rapid HIV testing data in a setting with high HIV prevalence and incidence may reveal important patterns in indeterminate results (7).

Study population
The current study is an analysis of previously collected data from the RCCS, an open, population-based study situated in the Rakai District of Uganda.RCCS is an open prospective cohort study that evaluates participants aged 15-49 years about every 18 months (i.e., rounds), across 50 communities, through censuses, sociodemographic and health surveys, and blood sample collection (8).This open cohort enrolls in-migrants and recently age-eligible participants each survey round.The current study used data from three RCCS rounds (9)(10)(11) conducted between July 2013 and May 2018.

Laboratory testing
Rapid test procedures were conducted per manufacturer protocol for both the Determine (Determine HIV-1/2 Alere Medical Company Limited, Chiba, Japan), Stat-Pak (HIV ½ Stat-PakR Dipstick, Chembio Diagnostic Systems, Medford, NY, USA), and Uni-Gold (Uni-Gold Recombigen HIV-1/2, Trinity Biotech, Bray, Ireland) diagnostic kits (12)(13)(14).Rapid antibody tests were conducted in the field per manufacturer protocol; a finger prick was applied for blood collection, and blood sample was immediately transferred to the manufacturers sample pad for testing.Each rapid test diagnostic kit had high specificity and sensitivity (>99%) to fingertip blood samples (12)(13)(14).The participants provided fingertip blood samples to two parallel rapid tests (Determine and Stat-Pak), and an additional third rapid test was applied if initial parallel tests were discordant (Uni-Gold) (12)(13)(14).Following the rapid test protocol, if participants continued to test indeterminate, additional ELISA (Murex HIV-1, 2.O, Murex Biotech Ltd., Dartford, UK, if reactive followed by Vironistika HIV Uni-Form II plus O Mircoelisa System, BioMer ieux: Marcy I'Etoile, France), western blot (GS HIV-1 Western Blot, Bio-Rad Laboratories, Redmond, WA, USA), and Abbott RealTime RNA (Abbott Molecular, Abbott Park, IL, USA), assays were utilized with specimens collected at the same visit (12)(13)(14).Final HIV status outcomes (i.e., "confirmed" HIV status) were determined by RCCS clinical and survey records, as well as all available lab data.For the survey rounds investigated (i.e., [16][17][18], all participants were systematically tested by the rapid HIV tests regardless of previous HIV diagnoses or laboratory test results.This policy was subsequently changed to not test individuals who were known to be living with HIV.Other variables were self-repor ted by participants using sociodemographic surveys (e.g., age, sex, pregnancy status, educational attainment, and occupation).

Statistical analysis
This descriptive analysis estimated the proportion of study visits with rapid HIV indeterminate test results overall and by sociodemographic characteristics.Relative to negative and positive rapid HIV test results, odds ratios (OR) and 95% CI of rapid HIV indeterminate test results were estimated using logistic regression models with generalized estimating equations (GEE), an exchangeable correlation structure, and robust variance estimation (15).A separate model was used for each outcome compar ison group (negative or positive).We also conducted multivariate analyses including sociodemographic covariates determined a priori (i.e., sex, age, education, occupation, residential community type, and survey round).
Additional analyses were conducted for participants with two or more person-visits.In this sample, we assessed the association of an individual's rapid HIV test result at the prior visit (indeterminate, positive, or negative) with the probability of an indeterminate rapid HIV test result (at the subsequent visit).We also separately examined the associa tion of an individual's "final" confirmed HIV test result at the prior visit with the probabil ity of an indeterminate rapid HIV test result (at the subsequent visit).Prevalence ratios (PR) and 95% CI of rapid HIV indeterminate test results were estimated using univari ate and multivariate Poisson regression models with GEE, an exchangeable correlation structure, and robust variance estimation (16).The multivariate model similarly included adjustment for sociodemographic covariates.In a separate analysis, we also estimated the probability of being confirmed HIV positive conditional on the individual's rapid HIV test result at the prior visit.
Among the participants with three person-visits, we conducted descriptive analyses to assess longitudinal patterns in indeterminate rapid HIV test results over the entire study period.We assessed the prevalence of longitudinal patterns in rapid HIV test results and estimated conditional transition probabilities of rapid HIV test outcomes across the three study visits.
All statistical analyses were conducted in Stata/SE 14.2 (StataCorp LLC, College Station, TX, USA).

Baseline characteristics of the study sample
There were 31,405 RCCS participants who contributed 54,469 person-visits (Table 1).A total of 15,692 participants were tested at only 1 visit, 8,362 participants were tested at 2 visits, and 7,351 participants were tested at 3 visits (i.e., 15,713 participants contributed multiple visits).In the overall sample, 53% (n = 16,766) were female, of which 1,900 (11.3%) self-reported being pregnant (Table 1).The overall prevalence of "confirmed" HIV infection at baseline was 16.4% (n = 5,164).Baseline sociodemographic characteris tics remained fairly consistent across all analytic samples based on visits contributed; however, the prevalence of HIV varied substantially.

Overall prevalence of indeterminate rapid HIV test results
The proportion of all person-visits with an indeterminate result was 2.7% (1,490/54,469; Table 2).Of the total 1,490 indeterminate rapid test results, 40.2% (n = 599) were classified as indeterminate due to discordant (negative and positive) parallel rapid tests, and 59.8% (n = 891) were classified as indeterminate due to at least one weak positive result on parallel rapid tests (Table S2).Of the total 1,490 indeterminate rapid test results, 1,104 (74.1%) were ultimately "confirmed" as HIV negative at that visit, and 386 (25.9%) were ultimately "confirmed" as HIV positive at that visit based on additional data.
The proportion of participants who ever had an indeterminate rapid test result was 3.9% (1,227/31,405), and this estimate increased by the number of person-visits contributed by a participant: 2.7% (422/15,692) for participants who contributed one person-visit, 4.4% (364/8,362) for participants who contributed two person-visits, and 6.0% (441/7,351) for participants who contributed three person-visits.Among partici pants with two person-visits, 1.1% (88/8,362) had indeterminate rapid test results at both person-visits.Among participants with three person-visits, 0.6% (47/7,351) had indeterminate rapid test results at all three person-visits.

Sociodemographic associations with indeterminate rapid HIV test results
The prevalence of indeterminate rapid test results varied by sociodemographic factors (Table 2).The prevalence of indeterminate results increased over the three study rounds, from 2.4% in round 16 to 3.0% in round 18.In particular, participants were significantly more likely to have an indeterminate result than a negative result in rounds 17 and 18, as compared to in round 16.Compared to men, women were significantly more likely to have an indeterminate result than negative [OR, 2.1 (95% CI: 1.9-2.4)]and more likely to have an indeterminate result than positive [OR, 1.3 (95% CI: 1.1-1.5)].However, among women, pregnancy status was not associated with an indeterminate result [vs negative, OR, 1.0 (95% CI: 0.8-1.1);vs positive, OR, 1.1 (95% CI: 0.9-1.3)].The prevalence of indeterminate results significantly varied by age group with persons aged 15-19 years having the lowest prevalence (1.5%).Compared to 15-19 year olds, all older age groups were significantly more likely to have an indeterminate result than negative and significantly less likely to have an indeterminate result than positive.Similar associations were observed in multivariate analyses.

Longitudinal patterns of indeterminate rapid HIV test results
We further analyzed 15,713 participants with two or more visits (n = 22,064 person-vis its).The probabilities of transitioning from a prior indeterminate rapid test result (n = 591) to HIV negative, HIV positive, or remaining indeterminate based on the rapid test at the subsequent visit were 40.4% (n = 239), 18.6% (n = 110), and 41.0% (n = 242), respectively (Table 3).The probability of transitioning from a prior indeterminate rapid test result subsequently to a 'confirmed' HIV-negative result were 76.0% (n = 449) and to a "confirmed" HIV-positive result were 24.0%(n = 142) (Table 4).The probability of transitioning from a prior negative rapid test (n = 20,022) to an indeterminate rapid test result was 1.8% (n = 364), and the probability of transition ing from a prior positive rapid test (n = 2,451) to an indeterminate rapid test result was 2.9% (n = 70) (Table 3).Compared to participants who had a negative rapid test result at their prior visit, participants with an indeterminate rapid test result at their prior visit were over 20 times more likely to have a subsequent indeterminate rapid test result [PR, 25.8 (95% CI: 22.5-29.6)];this association was observed even after adjustment for sociodemographic covariates [aPR, 23.0 (95% CI: 20.0-26.5)].Participants with a positive rapid test result at their prior visit were also more likely to subsequently have an indeterminate rapid test result [PR, 1.6 (95% CI: 1.2-2.0);aPR, 1.5 (95% CI: 1.2-2.0)]than participants who had a negative HIV test result at their prior visit.Additionally, participants with a "confirmed" HIV-positive result at the prior visit were also more likely to subsequently have an indeterminate rapid test result (4.0% vs 2.8%) than participants who had a "confirmed" HIV-negative result at their prior visit [PR, 1.5 (95% CI: 1.2-1.8);aPR, 1.3 (95% CI: 1.0-1.7)];participants with a "confirmed" HIV-positive result at the prior visit were more likely to have had indeterminate rapid results at the prior visit than the participants with a "confirmed" HIV-negative result at the prior visit (Table S1).
Of the 7,351 participants with 3 person-visits, 92.1% (n = 6,769) had a negative rapid test result, 5.4% (n = 400) had a positive rapid test result, and 2.5% (n = 182) had an indeterminate rapid test result at their first visit (Table 1; Fig. 1).The prevalence of a negative rapid test result at all three visits was 87.3% (n = 6,416); 0.6% (n = 47) of the participants had an indeterminate rapid test result at all three visits; and 5.1% (n = 378) of the participants had a positive rapid test result at all three visits (Fig. 1).Participants with indeterminate rapid test results at their first and second visit (n = 67), were more

DISCUSSION
Population-based data on longitudinal patterns in rapid HIV diagnostic results can provide insight into how to manage indeterminate rapid test results in everyday practice.These longitudinal and population-based data in which all participants were systemat ically tested regardless of their prior testing history allowed tracking rapid HIV test outcomes over an extended 5-year period.In a setting with high HIV burden, the overall population-level prevalence of HIV indeterminate rapid test results was 2.7%.The strongest predictor of an indeterminate result was a prior indeterminate result; such individuals were 20 times more likely to test indeterminate again on their next test compared to those with an initial negative test.We also found that 0.6% of individuals (n = 47) observed at three study visits remained persistently indeterminate over the 5-year study period.Notably, among the participants who had a prior indeterminate rapid result, 24% were confirmed HIV positive following additional follow-up testing.
Our study documented the high within-person correlation of indeterminate rapid HIV tests and provided evidence that a non-negligible fraction of individuals remain persistently indeterminate.For instance, 0.6% were indeterminate at three separate longitudinal visits over a 5-year period.Prior evidence of persistently indeterminate HIV test results across multiple studies and different serodiagnostic HIV tests lead us to consider the prolonged factors that might influence these results (9,10,17,18).For example, persistent indeterminate ELISA results (Murex HIV-1, Vironostika HIV Uni-Form) were previously identified among RCCS participants (17).Cross-reactivity from other infections has been found to be associated with indeterminate results and, therefore, might explain why some subjects remain indeterminate over a significant follow-up period (11,19,20).Understanding the factors that induce persistent indeterminate outcomes could further inform distinctions within the follow-up testing protocol to explain and manage people with continual indeterminate results.In attempts to manage these people with persistently indeterminate results, previous studies have implemented partner testing to assess participant risk profiles and probable status of the person (18,21).Confirmatory testing after 2 weeks should also be performed to resolve indetermi nate rapid HIV antibody test results (1,2).PCR testing should also be considered to resolve samples with persistently indeterminate rapid HIV antibody test results.
The initial rapid HIV indeterminate test results were most likely to turn negative, but 18% of these participants rapidly tested positive at the next visit, and following confirmatory testing, 24% of participants were determined to be living with HIV.The high HIV burden in this cohort may impact the transportability of this finding outside of this population.However, the prevalence of confirmed participants living with HIV who rapidly tested indeterminate was not negligible and elucidates the importance of identifying and following initially indeterminate patients to identify those amid seroconversion.Under current "test and treat" guidelines, newly diagnosed people with HIV should be immediately linked to care to optimize HIV control (1).
Current rapid HIV tests have high sensitivity and specificity (22,23), but they remain imperfect as indicated by the data in this study.Rapid test results are subject to individual interpretation of faint lines characterized as weak positive.The greatest inter-reader variability among rapid test results has been in distinguishing samples as "weak positive" vs "truly positive" (24).There are significant tradeoffs in readers' decision to classify results as "indeterminate" vs negative.The parallel testing protocol can help to nullify any contradictory rapid test results, as one study reported zero false-positive or false-negative results across more than one rapid HIV test (22,23).The parallel testing protocol confirms the prior test outcome and, if discordant, subjects the participant to higher-level testing.The parallel testing protocol can similarly aid in the event of indeterminate outcomes including weak positive results (10,24).As previously noted, PCR testing may be used to resolve indeterminate rapid HIV antibody test results, but this may not be feasible in all low-and middle-income country (LMIC) settings or when resources are constrained (1).Providing readers model test results may help significantly in ensuring consistent classifications between readers, test kits, and clinics alike to most accurately describe prevalence of indeterminate results in a given population (25).Given the recurrence of subsequent indeterminate rapid test outcomes, further research is needed to develop more specific guidelines for the consistent interpretation and handling of indeterminate results.
The population-based, longitudinal study design with systematic rapid HIV testing regardless of prior test results or HIV status is a key strength of this study.This allowed us to uniquely quantify follow-up rapid HIV test results even among individuals with previous positive confirmed HIV results.It is notable that of the 2,573 person-visits among people confirmed to be living with HIV, 103 (4%) subsequently had an indetermi nate rapid test result at the next visit.The reason for this phenomenon remains unclear and suggests there may be some reductions in antibody levels.However, it should be noted that in this subset of people previously laboratoryconfirmed to be living with HIV, 33/103 (32%) were also rapidly indeterminate at the prior visit (i.e., they were persistently rapid HIV indeterminate).
The prevalence of indeterminate results also varied by sociodemographic characteristics.For instance, women were twice as likely to test indeterminate than were men.Sex differences have been seen in the immunologic response to HIV, and women have been considered better viral controllers than men (26,27).The conceivable role of increased antibody function in women's viral control of HIV is not known, but a higher antibody concentration may impact the interpretation of weak positive rapid test results (28,29).In our data, females were more likely than males to be discordant and more likely to be weak positive (Table S2).Conversely, a prior analysis of RCCS data of HIV test outcomes by ELISA reported a greater number of indeterminate outcomes among men than women (17).In ELISAs, weak positive results may manifest from higher concentration of antigen in male serum (26,30).Further research is needed regarding sex differences in immunological responses to HIV and their impact on HIV serodiagnostics.
Overall, our findings support the current WHO guideline that HIV surveillance programs should report and provide further analysis on indeterminate test results to avoid underestimation of the HIV burden.Again, 24% of initially indeterminate participants were determined HIV positive following confirmational testing at their next visit.Rapid HIV tests have been essential in the expansion of diagnostic care in LMIC, but further attention to the performance and utilization of these tools is needed.The implementation of a protocol to utilize multiple HIV rapid tests and confirmatory tests is necessary for the proper identification of all new HIV infections and ensure prompt HIV diagnosis and linkage to care.

TABLE 1
Baseline characteristics of the study sample a Reflects the final confirmed HIV result after additional laboratory testing (e.g., HIV western blot) and triangulation with other survey data.

TABLE 2
Sociodemographic characteristics associated with indeterminate HIV rapid test results among 31,405 RCCS participants in Rakai, Uganda (2013-2018) (n = 54,469 person-visits) c Odds ratios (OR) comparing an indeterminate rapid test result were estimated by logistic regression with generalized estimating equations, an exchangeable correlation structure, and robust variance estimation.A separate model was used to compare indeterminate rapid test results to negative rapid test results and positive rapid test results.

TABLE 3
Association of prior rapid HIV test result with indeterminate rapid HIV test result for 15,713 RCCS participants in Rakai, Uganda (2013-2018) with 2 or more visits (n = 22,064 person-visits) c Prevalence ratios (PR) of an indeterminate rapid test were estimated by modified Poisson regression with generalized estimating equations, an exchangeable correlation structure, and robust variance estimation.b The multivariate model (aPR) included adjustment for sociodemographic variables (sex, age, education, occupation, residential community type, and survey round).CI, confidence interval; n, number of person-visits; PR, prevalence ratio; RCCS, Rakai Community Cohort Study.
a c

TABLE 4
Confirmed HIV status by prior rapid HIV test result for 15,713 RCCS participants in Rakai, Uganda (2013-2018) with 2 or more visits (n = 22,064 person-visits) b a Reflects the final confirmed HIV result after additional laboratory testing (e.g., HIV western blot) and triangulation with other survey data.b n, number of person-visits; RCCS, Rakai Community Cohort Study.FIG 1 Transition probabilities in rapid HIV test results given prior rapid test results and the prevalence of individual-level longitudinal patterns among 7,351 RCCS participants with 22,053 person-visits across 3 study visits in Rakai, Uganda (2013-2018).