Impact of Early Antiretroviral Treatment Initiation on Performance of Cross-Sectional Incidence Assays

Abstract

Antiretroviral therapy (ART) can impact assays used for cross-sectional HIV incidence testing, causing inaccurate HIV incidence estimates. We evaluated the relationship between the timing of ART initiation and the performance of two serologic HIV incidence assays. We analyzed 302 samples from 55 individuals from the RV217 cohort (Early Capture HIV Cohort Study). Participants were grouped by ART start time: ART started <1 year after infection (N = 9); ART started 1–3 years after infection (N = 12); and never received ART (N = 34). Samples were tested using the Sedia LAg-Avidity and Johns Hopkins modified Bio-Rad-Avidity assays. Results were compared with those from the Johns Hopkins HIV Cohort in which participants initiated ART an average of 10 years after infection (N = 17). Participants on ART were virally suppressed at the time of sample collection. The increase in normalized optical density (ODn) values was an average of 2.15 U/year lower in participants who started ART <1 year after infection than in those who did not start ART. Participants who started ART 1–3 years after infection had a decline in ODn values 0.90 U/year faster compared with those who started ART an average of 10 years after infection. Timing of ART initiation did not significantly impact results obtained with the Bio-Rad-Avidity assay. ART initiation <1 year after HIV infection was associated with persistently low limiting antigen (Lag)-Avidity values; this could lead to overestimation of HIV incidence. LAg-Avidity values declined more rapidly the earlier ART was initiated. Bio-Rad-Avidity values were not impacted by the timing of ART initiation.

Introduction

HIV incidence measures the leading edge of the epidemic. Accurate incidence estimates are needed to determine the state of local epidemics and the effectiveness of interventions for HIV prevention.¹ Cross-sectional HIV incidence testing has been used to provide local/regional estimates of HIV incidence² and to determine the outcomes of trials evaluating interventions for HIV prevention.³ Use of cross-sectional methods for HIV incidence estimation is easier and less costly than longitudinal incidence assessments.⁴ In the Population Health Indicators Assessment (PHIA), these methods are being used to determine the impact of the Presidents Emergency Plan For AIDS Relief (PEPFAR).⁵

Unfortunately, cross-sectional HIV incidence assays have several limitations. Most assays used for this purpose measure general features of the antibody response to HIV infection, which usually evolve during the first year of infection.⁶ Many factors have been shown to influence the rate of antibody maturation in HIV infection (e.g., gender,⁷ viral subtype,⁸ and geographic location⁹); differences in antibody responses can lead to misclassification of individuals with late-stage infection as recently infected.¹⁰ Antiretroviral therapy (ART) can also cause downregulation of the anti-HIV antibody response, leading to misclassification of individuals with longer-term infection as recently infected using some serologic incidence assays.^10,11

To address this, most algorithms used for cross-sectional HIV incidence estimation classify all virally suppressed individuals as “nonrecent.”^12–14 This is problematic in many settings, since availability and use of ART is increasing globally,¹⁵ and since current recommendations for HIV care include frequent HIV testing with initiation of ART as soon as possible after HIV diagnosis.¹⁶

The relationship between the timing of ART initiation and the performance of incidence assays remains unclear. Since use of ART is expected to continue to increase globally¹⁵ with a shift toward earlier ART initiation, there is a pressing need to understand the potential impact of the timing of ART initiation on the performance of serologic HIV incidence assays. In this study, we evaluated the performance of two assays that are commonly used for cross-sectional HIV incidence estimation: the limiting antigen (Lag)-Avidity assay and the Johns Hopkins modified Bio-Rad-Avidity assay. The analysis included individuals who started ART at varying times after HIV infection (from <1 month to >10 years).

Materials and Methods

Ethics statement

Participants in the RV217 Early Capture Cohort study provided informed consent before study enrollment; samples were only tested in this study for participants who consented to future use of their specimens. All participants in the Johns Hopkins HIV Clinical Practice Cohort provided written informed consent and the study was approved by the Institutional Review Board of the Johns Hopkins University School of Medicine, which also approved research on cross-sectional incidence testing using stored samples. No new samples were obtained for this study. Research was conducted in accordance with the Declaration of Helsinki.

Samples

Samples were obtained from two cohort studies, the Early Capture HIV Cohort Study (RV217)¹⁷ and the Johns Hopkins HIV Cohort Study.¹⁸ We evaluated 302 samples from 55 individuals in the RV217 cohort (31 participants from Thailand, 19 participants from Kenya, 5 participants from Uganda; Table 1). To be eligible to enroll in the parent study, participants had to have met one of the following four criteria within the 3 months before enrollment: exchanged sex for goods; participated in unprotected sex with a known HIV-positive partner; participated in unprotected sex with three or more partners; or experienced symptoms of a sexually transmitted infection.¹⁹

Table 1.

Participant Demographics in RV217 Cohort

Country	Thailand	Kenya	Uganda	Total
No. of participants	31 (56%)	19 (35%)	5 (9%)	55
Median age (IQR)	23 (18–25)	24 (22–24)	26 (25–26)	23 (20–26)
Cisgender male	19 (100%)	0	0	19
Cisgender female	2 (8%)	19 (73%)	5 (19%)	26
Transgender female	10 (100%)	0	0	10
No ART	19 (56%)	11 (32%)	4 (12%)	34
ART <1 year	3 (33%)	6 (67%)	0	9
ART 1–3 years	9 (75%)	2 (17%)	1 (8%)	12

ART, antiretroviral therapy; IQR, interquartile range.

The 55 participants evaluated included 19 cisgender males, 26 cisgender females, and 10 transgender females (Table 1). The median age was 23 years [interquartile range (IQR): 20–25.5]. Each participant contributed 4–7 samples (mean: 5.5 samples/person); samples were collected 0.15–4.20 years after infection, with an average of four samples in the first year of infection and one sample per year after that. The median time between a participant's last negative HIV test and first positive HIV test was 6 days. The first HIV-positive specimen was defined as the first specimen that tested positive for HIV-1 RNA.¹⁹ Participants were divided into three groups: those who started ART <1 year after infection (N = 9); those who started ART 1–3 years after infection (N = 12); and those who did not start ART (N = 34).

All participants in the RV217 cohort were virally suppressed (viral load <1,000 copies/mL) at the time of sample collection for samples tested after ART initiation. HIV subtype assignments were based on near full-length sequence data generated previously through a single-genome amplification method.²⁰

Results from the RV217 study were compared with previously generated results from participants in the Johns Hopkins HIV Cohort who started ART an average of 10 years after infection (N = 17). Participants in this cohort were from Baltimore, MD, USA. The 17 participants included 7 cisgender males and 10 cisgender females (median age 42 years, IQR: 35–47). The primary risk factors for HIV infection included sexual and parenteral exposure. Each participant contributed 3–7 samples (mean: 3.9 samples/person). All participants in the Johns Hopkins HIV cohort were virally suppressed (viral load <1,000 copies/mL) at the time of sample collection for samples tested after ART initiation. All subjects in this cohort were infected with subtype B.

Laboratory testing

Samples were tested using the LAg-Avidity assay (Sedia Biosciences Corporation, Portland, OR, USA) and the Johns Hopkins modified Bio-Rad-Avidity assay.^18,21 The LAg-Avidity assay was performed using the manufacturer's protocol. Results from the LAg-Avidity assay were reported as a normalized optical density (ODn). Results from the Bio-Rad-Avidity assay were reported as Avidity Index (AI) (i.e., the optical density of the diethylamine-treated well divided by the optical density of the untreated well, × 100). Viral load data were obtained from parent studies.

Data analysis

In the first phase of the analysis, we compared the change in ODn and AI values over time during the first year of infection for participants in each of the three groups from the RV217 cohort (ART started <1 year after infection; ART started at 1–3 years after infection; no ART). In the second phase of the analysis, we compared the change in ODn and AI values over time from the time of infection to the end of follow-up. Three participant groups were included in this analysis: (1) RV217 participants who started ART <1 year after infection, (2) RV217 participants who started ART 1–3 years after infection, and (3) participants from the Johns Hopkins HIV Cohort who started ART an average of 10 years after infection.

Subanalyses included comparisons of assay results from different geographic regions in the RV217 study [Thailand vs. East Africa (Kenya and Uganda)]. Due to the limitations in sample size, we limited this comparison to those who never received ART. The same subgroup (no ART) was used to examine the relationship of age with ODn and AI slope. We also compared assay performance for cisgender males versus transgender females from Thailand (Table 1). Since all 19 cisgender males in the RV217 study were from Thailand and 24 of the 26 (92%) cisgender females were from East Africa (Table 1), we did not compare differences in assay performance for cisgender males versus cisgender females.

Slopes for ODn and AI data were compared using a mixed-effects model with random slope and intercept. This model was chosen for its ability to account for variation between groups that may have reflected differences in assay performance before ART initiation versus the expected variation in slope based on the timing of ART initiation. We did not evaluate estimates of the mean duration of recent infection due to the small sample size in the RV217 cohort (N = 55 participants) and the subtype variation between participants from Kenya, Uganda, and Thailand, which could have impacted those results.²² Analyses were conducted using STATA SE, version 14.2 (StataCorp, College Station, TX, USA).

Results

LAg-Avidity

We first evaluated the impact of the timing of ART initiation on performance of the LAg-Avidity assay (Fig. 1). In the first year after infection, the mean slope of ODn values was 2.15 ODn U/year lower for those who started ART <1 year after infection (Fig. 1A), compared with those who were not on ART (Fig. 1C) [95% confidence intervals (CI): −2.96 to −1.33). In contrast, during the first year after infection, there was no statistically significant difference between the slopes for LAg-Avidity assay values among participants who started ART 1–3 years after infection (Fig. 1B) compared with those who never received ART (Fig. 1C) (p = .31).

FIG. 1.

Impact of the timing of ART initiation on LAg-Avidity assay results. Upper panels: the plots show LAg-Avidity assay results as a function of duration of infection for participants in the RV217 cohort who started ART <1 year (A) or 1–3 years after HIV infection (B), compared with those who did not start ART (C). Lower panels: the plots show LAg-Avidity assay results as a function of time on ART for participants in the RV217 cohort who started ART <1 year or 1–3 years after HIV infection (D and E, respectively); results are also shown for participants in the Johns Hopkins HIV Cohort who started ART an average of 10 years after HIV infection (F). Black lines in the upper panels indicate that participants were not on ART. Red lines indicate that participants were on ART. Horizontal blue lines (ODn = 2) indicate a cutoff used previously to distinguish between recent and nonrecent HIV infection. ART, antiretroviral therapy; LAg-Avidity, limiting antigen avidity assay; ODn, normalized optical density.

Over the entire period of follow-up after ART initiation, participants who started ART 1–3 years after infection (Fig. 1E) had a significantly greater decline in ODn values/year than those who started ART >10 years after HIV infection (Fig. 1F) [0.90 U/year, (95% CI: 0.47, 1.32)]. ODn values also declined more rapidly among individuals who started ART <1 year after infection (Fig. 1D) compared with those who started ART 1–3 years after infection (Fig. 1E) (0.74 U/year, 95% CI 0.28–1.20).

Bio-Rad-Avidity

We next evaluated the impact of the timing of ART initiation on performance of the Bio-Rad-Avidity assay (Fig. 2). In the first year after infection, the mean slope of values before ART initiation was lower for those who later started ART (17.16 AI/year, 95% CI: −17.05 to 51.38; Fig. 2B) than for those who never started ART (33.13 AI/year, 95% CI: 11.89–54.36; Fig. 2C); this difference was not statistically significant (p = .08). The slope of AI values was not significantly different for participants who started ART <1 year after infection (Fig. 2A) compared with those who did not start ART (Fig. 2C) (p = .41). There was also no significant difference in the slope of AI values after ART initiation in those who started <1 year after infection (Fig. 2D) and those who started ART 1–3 years after infection (Fig. 2E) (p = .25), or those who started ART >10 years after infection (Fig. 2F) (p = .66).

FIG. 2.

Impact of the timing of antiretroviral treatment (ART) initiation on Bio-Rad-Avidity assay results. Upper panels: the plots show Bio-Rad-Avidity assay results as a function of duration of infection for participants in the RV217 cohort who started ART <1 year (A) or 1–3 years after HIV infection (B), compared with those who did not start ART (C). Lower panels: the plots show Bio-Rad-Avidity assay results as a function of time on ART for participants in the RV217 cohort who started ART <1 or 1–3 years after HIV infection (D and E, respectively); results are also shown for participants in the Johns Hopkins HIV Cohort who started ART an average of 10 years after HIV infection (F). Black lines in the upper panels indicate that participants were not on ART. Red lines indicate that participants were on ART. Horizontal blue lines (AI = 40) indicate a cutoff used previously to distinguish between recent and nonrecent HIV infection. AI, avidity index.

Subanalyses

In addition to the timing of ART initiation, for the RV217 cohort we evaluated the potential impact of region, gender, and age on the performance of the LAg-Avidity and Bio-Rad-Avidity assays. We identified nine participants who had AI values below 50 for the entire follow-up period; all nine participants were from Thailand (Fig. 3D). Despite this observed difference, there was no significant difference in the mean slope of Bio-Rad-Avidity assay values (AI) for participants from Thailand versus East Africa (p = .69) (Fig. 3C vs. Fig. 3D). There was also no significant difference in the mean slope of LAg-Avidity assay values for participants from these two regions (p = .82) (Fig. 3A vs. Fig. 3B).

FIG. 3.

Comparison of assay results for samples from differing geographic locations. The plots show Lag-Avidity assay results (top panels A and B, ODn) and Bio-Rad-Avidity assay results (bottom panels C and D, AI) as a function of duration of infection for participants in the RV217 cohort from East Africa (left panels A and C) and Thailand (right panels B and D). Black lines indicate that participants were not on ART. Red lines indicate that participants were on ART. Horizontal blue lines indicate cutoffs used previously to distinguish between recent and nonrecent HIV infection (upper panels A and B: ODn = 2; lower panels C and D: AI = 40).

We found no significant difference in the mean slope of AI values (p = .08) or ODn values (p = .58) for cisgender males versus transgender females in Thailand. Finally, there was no significant difference in the mean slope of AI values (p = .07) or ODn values (p = .46) based on age at the time of study enrollment.

We calculated the false recent ratio (FRR) of the RV217 samples using three algorithms. To be included in the FRR calculations participants needed to have at least one sample 2 years after their date of infection. Forty-four of the 55 participants in the study met the criteria. Using a cutoff of 1.5 LAg-ODn there was an FRR of 14.7%, a cutoff of 1.5 LAg-ODn with a VL >1,000copies/mL yielded a FRR of 0%, and a cutoff of 40% AI from the Bio-Rad-Avidity assay resulted in an FRR of 14.7%. Raw data in its entirety can be found in Supplementary Table S1.

Discussion

Previous studies have demonstrated that early ART initiation can lead to a blunted HIV antibody response.^23,24 This study demonstrates that early ART initiation can impact the performance of the LAg-Avidity assay. Individuals who started ART <1 year after infection had persistently low LAg-Avidity values. Also, individuals who started ART 1–3 years after infection had a significantly sharper decline in LAg-Avidity values after initiating ART, compared with those who started ART >10 years after infection. A sharp decline in ODn or AI values is significant as an individual would test recent on the assay sooner and lead to inaccurate incidence estimates.

Our findings are consistent with previous studies that show that there is a greater impact on LAg-Avidity assay performance with earlier ART initiation.^23,24 These results are significant, since global trends toward early ART initiation could confound attempts to estimate HIV incidence using cross-sectional testing algorithms, leading to inaccurate estimates of HIV incidence. Furthermore, classification of those with low viral loads as having nonrecent HIV infection is problematic in settings where ART is started early in infection; this approach may be particularly problematic in settings where the proportion of those starting ART early varies among different subgroups and over time.

In this study, the performance of the Johns Hopkins modified Bio-Rad-Avidity assay was not impacted by the timing of ART, and no difference was observed among those who did versus did not start ART. Previous studies have shown that individuals with longstanding HIV infection tend to reach AI values of around 100 and then remain at that level.^18,25 Our findings are consistent with previous studies that have demonstrated that ART is not associated with a decrease in AI values among individuals who were on ART and maintained a viral load <1,000 copies/mL when using the Johns Hopkins modified Bio-Rad-Avidity Assay.^9,18

The impact of viral rebound after ART failure or noncompliance should be a part of future investigations. Differences in the composition and methods used in the second-generation LAg-Avidity assay and the third-generation Bio-Rad-Avidity assay may account for differences in test performance. Second-generation HIV incidence assays use plates coated with HIV antigens that are targeted by IgG antibodies; in contrast, third-generation assays use plates coated with HIV antigens that are targeted by both IgG and IgM antibodies.

Studies have demonstrated that early ART initiation can blunt the antibody response to HIV infection, and that third-generation assays used for HIV diagnosis (such as the Bio-Rad assay that is modified for HIV incidence testing) are more sensitive for detecting HIV antibodies during ART than second- and fourth-generation assays.^23,24 This, along with previous findings that the Sedia Lag-Avidity assay has been shown to be impacted by a low VL^10,11 could explain why early ART might impact the LAg-Avidity assay more than the Bio-Rad-Avidity assay. The exact mechanism responsible for this variation should be researched further.

Taken together, these findings highlight the importance of assessing ART use and the timing of ART initiation in populations that are assessed using cross-sectional HIV incidence methods. Continued research is needed to identify optimal assays and testing algorithms for cross-sectional HIV incidence estimation in settings with high rates of ART uptake, where individuals have access to frequent HIV testing and early ART initiation.

While we did not observe differences in assay performance in samples from Thailand versus East Africa, the nine participants from Thailand with early ART initiation who had persistently low values with the Bio-Rad-Avidity assay are of particular interest. Persistently low test values could lead to overestimation of HIV incidence when this assay is included in testing algorithms. Prior studies have documented that HIV subtypes can impact the accuracy of cross-sectional HIV incidence assays, including the LAg-Avidity and Bio-Rad-Avidity assays.²⁶ Subtype data were available for the majority of participants in the RV217 study (Table 2). However, we were not able to perform a meaningful assessment of the impact of HIV subtype on assay performance in this study given the small sample size.

Table 2.

HIV Subtype Data by Country

HIV subtype	Thailand, n (%)	Kenya, n (%)	Uganda, n (%)	Total, n (%)
A1	2 (6)	7 (41)	1 (20)	10 (19)
A1C	0	2 (12)	0	2 (4)
A1CD	0	4 (23)	1 (20)	5 (9)
A1D	0	3 (18)	3 (60)	6 (11)
B	2 (6)	0	0	2 (4)
B/CRF01_AE	2 (6)	0	0	2 (4)
C	0	1 (6)	0	1 (2)
C/CRF01_AE	1 (3)	0	0	1 (2)
CRF01_AE	24 (77)	0	0	24 (45)

Future research is needed to determine if HIV subtype impacts antibody maturation or performance of HIV incidence assays in individuals with early ART initiation.

The samples used in this study were a part of an early capture cohort, which provided precise dates of HIV infection and ART initiation. These data allowed us to assess the timing of ART initiation on performance of HIV cross-sectional incidence assays. These data were not available for the Johns Hopkins HIV cohort; this cohort only includes samples collected after ART initiation.¹⁸ One limitation of this study was that the two cohorts differed in age, gender, HIV subtype prevalence, and other factors.

This study suggests that the Bio-Rad-Avidity assay may be superior to the LAg-Avidity assay for assessing HIV incidence in populations where some individuals initiate ART early in infection. Further research is needed to compare the performance of these two assays for HIV incidence estimation in different populations and cohorts, where data are available on ART uptake and timing of ART initiation, and data from longitudinal incidence assessments are available for comparison.

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Departments of the Army or Defense.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was funded by R01-095068 (S.H.E.). Additional funding was provided by the HIV Prevention Trials Network (UM1A068613). Additional funding was provided by the Division of Intramural Research, NIAID. Additional support was provided by the department of Defense through cooperative agreements (W81XWH-07-2-0067 and W81XWH-11-0174) with the Henry M. Jackson Foundation for the Advancement of Military Medicine and by the National Institute of Allergy and Infectious Diseases, NIH, through an interagency agreement with the U.S. Army (Y1-AI-2642-17).

Supplementary Material

Supplementary Table S1