Clinical and Immunologic Outcomes After Immediate or Deferred Antiretroviral Therapy Initiation During Primary Human Immunodeficiency Virus Infection: The Sabes Randomized Clinical Study

Abstract

Background

In addition to demonstrated public health benefits on reducing transmission, it remains unclear how early antiretroviral therapy (ART) must be started after acquisition of human immunodeficiency virus (HIV) to maximize individual benefits.

Methods

We conducted an open-label randomized clinical study in Lima, Peru among adult men who have sex with men and transgender women with acute (HIV-antibody negative/HIV-1 RNA positive) or recent (confirmed negative HIV-antibody or RNA test within 3 months) HIV infection, who were randomized to start ART immediately versus defer by 24 weeks. We evaluated outcomes by treatment arm and immunologic markers by days since estimated date of detectible infection (EDDI).

Results

Of 216 participants, 105 were assigned to immediate arm and 111 to deferred arm (median age 26.8 years, 37% with acute HIV). The incidence of non-ART-related adverse events was lower in immediate versus deferred arm (83 vs 123/100 person-years, IRR 0.67 (95% confidence interval [CI] .47, .95; P = .02), the difference dominated by fewer infections in those treated immediately. After 24 weeks of ART, between-group differences in CD4/CD8 cell ratio lessened (P = .09 overall), but differences between those initiating ART ≤ 30 days from EDDI (median 1.03, interquartile range [IQR] 0.84, 1.37), and those initiating > 90 days (0.88, IQR 0.61, 1.11) remained, P = .02. Principal components analysis of 20 immune biomarkers demonstrated distinct patterns between those starting ART > 90 days from EDDI versus those starting within 30 or 90 days (both P < .001).

Conclusions

To our knowledge, this is the only evaluation of randomized ART initiation during primary HIV and provides evidence to explicitly consider acute HIV in World Health Organization recommendations for universal ART.

Clinical Trials Registration

NCT01815580.

In this randomized study of 216 persons with primary HIV in Lima, Peru, those who began ART within 30 days of estimated date of infection experienced the fewest symptoms and had improved immune profiles compared to those starting ART later.

INTRODUCTION

Universal treatment of human immunodeficiency virus (HIV) infection is now recommended at time of diagnosis based on findings of reduced HIV-associated morbidity and mortality [1, 2] and reduced transmissibility [3] with the early initiation of antiretroviral therapy (ART) in persons with prevalent HIV infections [4, 5]. Best practices for clinical management of acute HIV infection are unknown, with limited data from mainly observational studies. With improved ability to identify individuals with acute HIV infection, the virologic and immunologic benefits of treating persons with acute HIV infection are becoming clearer [6].

Recently, particular attention has been given to the dynamic CD4/CD8 cell ratio during primary HIV infection. Early ART (within 40 or 120 days of estimated infection date) has been associated with a significant increase in the CD4/CD8 cell ratio compared to delayed ART [7, 8]. An inverted CD4/CD8 cell ratio is linked both to CD4 cell depletion and to expansion of activated HIV-specific cytotoxic T cells [9]. Initiation of ART soon after infection may prevent persistent immune activation, result in less T-cell activation, smaller latent HIV reservoir size during long-term therapy [10–12], and normalization of the coagulation cascade and systemic inflammatory biomarkers [13]. Even in the “test-and-treat” era, individual health benefits of starting ART in acute infection have not been rigorously evaluated. In the Sabes trial, we hypothesized that immediate ART initiation following incident HIV infection would yield better clinical and immunologic outcomes than deferring ART by just 24 weeks.

METHODS

Design and Participants

Between 2013 and 2017, the Sabes study evaluated an HIV treatment-as-prevention intervention among men who have sex with men (MSM) and transgender women (TW), the populations most affected by HIV in Lima, Peru. This 3-step screen, rescreen, and treat study design has been published. In brief, of 2685 MSM and TW who were HIV-uninfected but at high-risk for HIV, 2109 entered a longitudinal cohort and were tested monthly by point-of-care third-generation HIV immunoassay and for human immunodeficiency virus type 1 (HIV-1) RNA if negative [14]. Acute HIV infection was defined as a positive plasma HIV-1 RNA test and negative third-generation antibody test. Recent HIV infection was diagnosed as a positive third-generation rapid HIV test (confirmed by separate immunoassay) with confirmation of a negative HIV test within 3 months. An additional 12 participants meeting criteria for recent HIV infection from another local clinic were invited to enroll.

Participants were randomized 1:1, stratified by acute/recent HIV, to start ART immediately or defer ART initiation by 24 weeks (henceforth “Immediate” or “Deferred” ART). Either coformulated efavirenz 600 mg/emtricitabine 200 mg/tenofovir disoproxil fumarate 300 mg (EFV/FTC/TDF) or elvitegravir 150 mg/cobicistat 150 mg/emtricitabine 200 mg/tenofovir disoproxil fumarate 300 mg (EVG/cobi/FTC/TDF) were first-line regimens provided. ART was offered prior to week 24 for Deferred participants reaching contemporary Peruvian criteria for ART initiation (CD4 count ≤ 350 cells/mL before December 2014; ≤500 cells/mL thereafter) or by clinician discretion, generally for symptomatic acute retroviral syndrome (ARS); the study was completed before Peruvian guidelines adopted the World Health Organization (WHO) “test-and-treat” approach in 2018. Alternate ART regimens were allowed for virologic failure, regimen toxicity, or intolerance. Forty-eight weeks after randomization, participants were transferred to the Peru Ministry of Health ART Program to continue ART.

This study was approved by the Institutional Review Board at the Fred Hutchinson Cancer Research Center and the Bioethics Committee at Asociación Civil Impacta Salud y Educación. All participants provided written consent in Spanish, study implementation was authorized by the Peruvian National Institute of Health, and the study is registered with ClinicalTrials.gov (NCT01815580). Sponsors played no role in study design, execution, or results preparation.

Laboratory Methods

At enrollment, and every 12 weeks thereafter, we performed CD4 and CD8 cell counts and quantitative HIV-1 RNA. Virologic suppression was defined as ≤ 40 copies/mL, the lower limit of detection. We evaluated markers of immune activation on plasma obtained at enrollment, ART initiation visit (if not enrollment), and 24 weeks after ART initiation from a subset of participants (n = 100) with representative distribution of time since estimated date of infection. We used 2 custom Meso Scale Discovery U-PLEX chemiluminescent immunoassay panels (Meso Scale Diagnostics, Rockville, MD, USA) to test 20 soluble biomarkers selected to encompass innate and T-cell responses to viral infections. We also evaluated soluble CD14 (sCD14), high-sensitivity C-reactive protein (hsCRP), and D-dimer in plasma (eMethods).

Analysis Methods

Intent-to-treat analyses included all participants randomized to Immediate or Deferred ART. We compared median and interquartile ranges (IQR) for CD4 counts, CD4/CD8 ratio, and log₁₀-transformed HIV-1 RNA at each 12-week interval with Kruskal-Wallis tests. We evaluated cumulative incidence of early study discontinuation, noninitiation of ART, virologic suppression, and adverse events (AEs) with multiple-failure survival functions, incidence rate ratios (IRR), and log-rank tests. Because we were interested in the effect of ART on HIV-associated symptoms, we excluded bacterial sexual transmitted infections (STI) detected by quarterly screening from our primary analysis of AEs.

Given variable intervals between HIV acquisition, enrollment/randomization, and actual ART initiation, we also analyzed outcomes by time between estimated date of detectable infection (EDDI) and ART initiation. EDDI was calculated using a published calculator that incorporates all HIV testing data from each participant, including assay-specific uncertainty windows [15]. We categorized participants into 3 as-treated groups: ≤30 days, 31–90 days, or > 90 days between EDDI and ART initiation. Persons who did not initiate ART during study were excluded from these analyses. Figure 1 depicts how serostatus at diagnosis and randomization arm align with as-treated category (see Lama et al for published Sabes CONSORT diagram [14]).

Figure 1.

Assignment of randomized participants to as-treated groups, by days elapsed between estimated date of detectable infection (EDDI) and actual start of antiretroviral therapy (ART). To account for different lengths of time between presumed human immunodeficiency virus (HIV) acquisition, diagnosis, and randomization, as well as ART initiation outside of assigned study visit, we also performed “as treated” analyses, in which participants were reclassified as having started ART within 30, within 90, or > 90 days after EDDI. Acute or recent HIV was determined by seropositivity (or absence thereof) on date of diagnosis and not necessarily date of enrollment/randomization. *Of 40 potential participants with eligible incident infections, 7 were not contactable, and 13 eligible by HIV infection criteria had other medical exclusions such as chronic hepatitis B infection, severely elevated transaminases, or psychiatric impairment. Twenty were screened for enrollment but declined participation.

Immune activation biomarkers were analyzed by as-treated category. A priori, we defined interleukin (IL)-6, induced protein (IP)-10, tumor necrosis factor (TNF)-α, sCD14, hsCRP, and D-dimer as analytes of interest. All analytes were log₁₀ transformed, with the exception of hsCRP and D-dimer. We performed generalized linear models (GLM) with Gaussian assumptions to evaluate the relationship between analytes and predictors and used binary assumptions and logistic GLM for prediction of CD4/CD8 ratio normalization. Change in biomarkers after 24 weeks of ART and between-group differences were tested with linear mixed effects models clustered on participant and group. Principal components analysis (PCA) evaluated multivariate differences in 20 biomarkers (hsCRP, D-dimer, and sCD14 excluded due to limited endpoints) and between-group differences tested with MANOVA [16]. We employed Bonferroni adjustments for multiple comparisons. All analyses were performed in STATA version 15.1 (Statacorp, College Station, TX, USA).

RESULTS

Of 256 eligible participants with acute or recent HIV infections, 216 participants enrolled and randomized to Immediate ART (n = 105, 37 acute, 68 recent infections) or to Deferred ART (n = 111, 34 acute, 77 recent infections). Characteristics of eligible new diagnoses, including reasons for exclusion have been previously described [14] and demographics of enrolled participants described in Table 1. Participants in the Immediate arm began ART at a mean of 0.1 week (range 0, 0.9) and 106 Deferred participants began ART at a mean of 22.6 weeks (range, 0, 28.6) after enrollment. Thirty-two Deferred participants initiated ART outside of the assigned 24-week visit window including 22 (23.4%) who initiated ART early: 16 (14.4%) due to CD4 cell count ≤ 350 cell/mL, 5 participants (4.5%) had symptomatic ARS or an opportunistic infection, and 1 participant requested early ART initiation. Five participants never initiated ART during this study (4.5% vs 0% in Immediate arm, hazard ratio [HR] incalculable, P = .028); 5 Deferred participants (4.5%) failed to start ART on time but started before week 30. There was no difference in study completion by arm, although Deferred participants were more likely to end participation before 24 weeks (intended ART initiation visit, P = .08).

Table 1.

Demographics and Baseline Clinical Characteristics of Sabes Participants, by Randomization Arm

	Immediate N = 105	Deferred N = 111
Age (median, IQR)	26.8 (22, 31)	24.5 (21, 30)
Gender identity (N, %)
Cisgender male	95 (90.5)	92 (82.9)
Transgender female	10 (9.5)	19 (17.1)
Education (N, %)
Primary	4 (3.8)	5 (4.5)
Secondary	21 (20.0)	31 (27.9)
Postsecondary	80 (76.2)	75 (67.6)
Income (median, IQR)   in Peruvian soles/month	750 (200, 1000)	750 (200, 950)
HIV diagnosis (N, %)
Acute	37 (35.2)	34 (30.6)
Recent	68 (64.8)	77 (69.4)
Initial ART (N, %)
EFV/FTC/TDF	92 (87.6)	96 (86.5)
EGV/co/FTC/TDF	13 (12.4)	15 (13.5)
Days from enrollment to ART (mean, range)	0 (0, 6)	158 (0, 200)
CD4 count at enrollment (median, IQR)	449 (272, 586)	406 (280, 544)
CD8 count at enrollment (median, IQR)	920 (625, 1314)	938 (639, 1450)
Initial HIV-1 RNA (median log  10   copies/ml, IQR)	5.94 (5.04, 6.79)	5.76 (5.18, 6.51)

Randomization was stratified by acute (seronegative) or recent (seropositive, but negative test within 90 days) at time of diagnosis.

Abbreviations: ART, antiretroviral therapy; EFV/FTC/TDF, coformulated efavirenz, emtricitabine, and tenofovir disoproxil fumarate; EVG/cobi/FTC/TDF, coformulated elvitegravir, cobicistat, emtricitabine, and tenofovir disoproxil fumarate; HIV, human immunodeficiency virus; IQR, interquartile range.

In total, 277 AEs were reported, of which 95 were bacterial STI, leaving 182 events for analysis (Table 2). The IRR for any AE during the study was 0.83 (95% confidence interval [CI] .62, 1.11; P = .22) for Immediate versus Deferred participants. The incidence rate for non-ART-related events was lower in Immediate versus Deferred participants (83 vs 123 per 100 person-years [PY], IRR 0.67; P = .02). Limiting nonrelated AEs to time on ART to limit bias in drug-attribution of events, the effect was more pronounced (83 vs 133 per 100 PY, IRR 0.62; P = .003). The Immediate arm experienced a greater number of ART-related AEs due to longer observation on ART, although the incidence rate was similar (35 vs 39 per 100 PY, P = .13), including common ART-related events, such as rash, neuropsychiatric, and gastrointestinal complaints. Immediate arm participants experienced significantly fewer non-STI infections (P = .005), notably fewer upper and lower respiratory infections. Opportunistic infections (OIs) in Deferred arm participants included herpes zoster (n = 2), oral candidiasis (n = 3), oral hairy leukoplakia (n = 1), pulmonary tuberculosis (n = 2), and tuberculous pleural effusion (n = 2). One case each of pulmonary and pleural tuberculosis were diagnosed in the Immediate arm, but no other classic OIs occurred. Symptoms associated with ARS, including pharyngitis, fever, and headache were also more frequent in Deferred participants (eTable 1). Laboratory abnormalities (related and unrelated to ART) were more common among Immediate participants (P = .03), all but one of which were transaminase elevations. Most transaminase elevations were attributed to primary HIV rather than ART, although it is likely that ART initiation during acute infection was associated with greater risk of hepatotoxicity, as it occurred more frequently in the Immediate arm (8 vs 1 events). Three serious AEs occurred, including one noncompleted suicide in a Deferred participant not yet on ART, one unrelated episode of hypotension, and one grade 3 transaminase elevation (on EFV/FTC/TDF, not attributed to ART) in Immediate participants (eTable 2). In as-treated analysis, we confirmed that the earliest-treated group (≤30 days from EDDI) had fewest overall events and non-ART-related AEs, with a trend toward fewer ART-related events (Table 2), whether or not Deferred participants who initiated ART ahead of assigned week were excluded in sensitivity analyses (eTable 3).

Table 2.

Adverse Events

Adverse Event Type	Immediate Arm N (%)		Deferred Arm N (%)	P value (log-rank)
Bacterial STI	38 (40.0)		57 (60.0)	.17
Total events (non-STI)	78 (42.9)		104 (57.1)	.19
Neurologic/psychiatric disorders	18 (48.6)		19 (51.4)	.63
Gastrointestinal disorders	17 (40.5)		24 (59.5)	.43
Infections/infestations (non-STI)	13 (25.8)		35 (74.2)	.005
Skin/subcutaneous	12 (57.1)		9 (42.9)	.44
Systemic/general disorders	3 (25.0)		9 (75.0)	.12
Laboratory abnormalities	8 (80.0)		2 (20.0)	.03
Intent to treat  Incidence rate of adverse events per 100PY (95% CI)
	Immediate Arm	Deferred  Arm	Incidence Rate Ratio for Immediate Arm	P value (log rank)
All adverse events	117 (93, 146)	141 (116, 170)	0.83 (.61, 1.13)	.22
Nonrelated events  (entire study)	83 (63, 108)	123 (100, 151)	0.67 (.47, .95)	.02
ART-related events (during ART)	35 (23,52)	39 (23, 68)	0.88 (.42, 1.89)	.13
Nonrelated events (during ART)	83 (64, 108)	133 (99, 179)	0.62 (.41, .95)	.0003
As Treated  Incidence of adverse events per 100PY (95% CI)
	≤30 days from EDDI to ART  N = 38	31–90 days from EDDI to ART  N = 76	>90 days from EDDI to ART  N = 97	P value (log-rank)
All adverse events	85 (56, 129)	163 (129, 205)	125 (102, 154)	.03
ART-related events	35 (18, 67)	32 (19, 54)	44 (25, 75)	.07
Nonrelated events	50 (29, 86)	131 (101, 170)	107 (85, 133)	.007

Adverse events were evaluated with a multiple-failure survival model. To evaluate the effects of immediate ART initiation vs deferral for 6 months after diagnosis in early infection, we excluded bacterial STI from further analysis, understanding that their incidence would more be related to behavior than direct biologic effects of ART. Total AE, ART-related AE and non-study-related AE were evaluated overall during the study, including stratification by time under exposure to ART to evaluate separately whether being on ART modified reporting of non-ART-related events. Overall, persons randomized to Deferred ART experienced a higher rate of nonrelated events overall and while on ART, as well as a higher risk of infections. All respiratory system events were included in the infection category, as all reported AEs attributed to that organ system were infections. Persons who were randomized to Immediate ART had more lab abnormalities (8 vs 2), all of which except 1 (isolated hyperbilirubinemia in a participant changed to second-line ART including atazanavir) were transaminase elevation; 8 of 9 transaminase elevations were in persons on efavirenz (EFV), of which only 2 were considered EFV-related. ART was temporarily held and then ultimately exchanged for ritonavir-boosted lopinavir (LPV/r) in both cases. The other 7 events were considered non-ART-related, likely secondary to acute HIV infection, although hepatoxicity was likely exacerbated by starting ART, considering that they were disproportionally represented in those immediately treated. The last event occurred in a person on EVG/cobi/FTC/TDF and began prior to ART initiation. All other transaminase elevations not attributed to EFV resolved spontaneously. P values that are statistically significant (<0.05) with bold.

Abbreviations: AE, adverse events; ART, antiretroviral therapy; CI, confidence interval; EDDI, estimated date of detectible infection; EVG/cobi/FTC/TDF, elvitegravir 150 mg/cobicistat 150 mg/emtricitabine 200 mg/tenofovir disoproxil fumarate 300 mg; PY, person-years; STI, sexually transmitted infection.

The proportion of Immediate participants who achieved viral suppression exceeded that of the Deferred arm, likely because treatment time was longer by week 48 (78.1 vs 63.1%, P = .016; Figure 2, eTable 4). However, there was no difference in virologic suppression at 24 weeks after assigned ART initiation in either arm (64.8 vs 63.1%). Although at enrollment, there was no between-arm difference in CD4 count or CD4/CD8 ratio, by week 12, there was a nearly 100-point difference in median CD4 count that narrowed somewhat by week 48 (615 vs 533, P = .03). CD4 count did not differ by arm 24 weeks after intended ART initiation visit. At enrollment, CD4/CD8 ratio was low in both arms, median 0.49 (IQR 0.25, 0.77) in Immediate participants and 0.47 (IQR 0.26, 0.77) in Deferred participants, P = .85. By week 24, and at end of study, the median CD4/CD8 ratio had normalized (ratio ≥ 1.0) in Immediate participants, 1.07 (IQR 0.83, 1.37), but remained notably lower in Deferred participants at 0.71 (IQR 0.47, 1.03), P = .0001. Twenty-four weeks after respective ART-initiation visits, 58 (55.2%) Immediate participants and 44 (39.6%) Deferred participants had normalized their CD4/CD8 ratio, HR 3.40 (95% CI 2.04, 5.66; P < .0001).

Figure 2.

HIV-1 RNA and CD4 T-cell count in participants randomized to initiate ART immediately or 24 weeks after diagnosis of early human immunodeficiency virus (HIV) in the Sabes Study. Quantitative HIV-1 RNA and CD4 + T-cells measurements were performed cross-sectionally with blood from within-visit windows at each analysis point (intent to treat analyses with all participants randomized to that arm included). A, Change in HIV-1 RNA viral load (VL) by arm, throughout 48 weeks of the Sabes Study. Box plots represent median and interquartile ranges. Equivalent time-under-ART comparisons for 24-weeks on ART lie at study week 24 in the Immediate arm and at 48 weeks in the Deferred arm, as outlined per-protocol. Virologic suppression was calculated as a single-failure model. B, Proportion of participants with suppressed VL, defined as <40 copies/mL, by analysis week. After 48 weeks on study, the difference in cumulative proportion with virologic suppression was 84.8 vs 70.3% in Immediate vs Deferred arms. C, Box plots demonstrating median and interquartile range in absolute CD4 count by arm at each analysis point. D, CD4+/CD8 + ratio at each analysis point. In (A), (C), and (D), the P value from Kruskal-Wallis test is shown in black for the week 48 between-arm comparison, and the blue bar shows the comparison between arms 24 weeks after intended ART initiation visit (week 24 for Immediate arm, week 48 for Deferred arm). As-treated analysis by actual ART initiation since estimated date of infection presented in Figure 3. Abbreviations: ART, antiretroviral therapy; CI, confidence interval; HIV-1, human immunodeficiency virus type 1.

As-treated Immune Activation Analyses

Because as-treated groups differed in time between EDDI and study enrollment/randomization, CD4 count and CD4/CD8 ratio differed between the 3 groups at enrollment (P = .0001) and at time of ART initiation (P = .0001), Figure 3. At ART initiation, the median CD4/CD8 cell ratio was 0.75 (IQR 0.46, 0.96) in those starting ≤ 30 days from EDDI, 0.37 (IQR 0.18, 0.65) in those starting 31–90 days after EDDI, and 0.52 (0.29, 0.68) in those starting > 90 days (corrected P-values P < .01 for each comparison). After 24 weeks of ART, the difference between the ≤ 30-day group (median 1.03, IQR 0.84, 1.37), and > 90-day group (median 0.88, IQR 0.61, 1.11) remained, P = .02, but other comparisons were not significantly different.

Figure 3.

CD4+/CD8 + T-cell ratio at ART initiation and 24 weeks after ART initiation, analyzed as treated by interval between estimated date of detectable HIV infection (EDDI) and ART initiation. Overall, the median CD4+/CD8 + T-cell ratio was different across all 3 groups (1.03 vs 0.96 vs 0.88, respectively, P = .0001 by Kruskal-Wallis test) at time of ART start, and the ratio of those starting ART after 90 days (CD4+/CD8 + T-cell ratio of 0.52) and within 90 days (0.37) differed from those starting ART within 30 days (0.76). After 24 weeks of ART, the CD4+/CD8 + T-cell ratio was not different (P = .09) overall, but those treated > 90 days from EDDI had a significantly lower ratio than those treated within 30 days (0.88 vs 1.03). Abbreviations: ART, antiretroviral therapy; HIV, human immunodeficiency virus.

At ART initiation, persons starting > 90 days from EDDI had lower levels of IL-8, interferon (IFN)-α2a, IP-10, IL-6, IL-7, IFN-γ, and IL-10 compared with persons who started ≤ 30 days from EDDI (eTable 6 and eFigure1). Overall, most evaluated biomarkers declined significantly in all 3 treatment groups after 24 weeks of ART. In unadjusted analyses after 24 weeks of ART, normalization of CD4/CD8 ratio (≥1.0) was only predicted by CD4/CD8 ratio at diagnosis, with trends seen for interval from EDDI to ART, and levels of IP-10, IFN-γ, and D-dimer at enrollment. In multivariable analysis, initial CD4/CD8 ratio continued to be the strongest predictor, but IFN-γ level at enrollment was inversely associated with CD4/CD8 normalization and hsCRP was positively associated. Adjusted for biomarkers at enrollment, time to ART initiation was no longer predictive; fewer associations were seen between enrollment biomarkers and absolute CD4 count (eTable 8a/b). Further analyses between cytokines, CD4/CD8 ratio, VL, and EDDI can be found in eTables 5–9.

At enrollment, immune profiles in the as-treated groups (n = 94) were distinct from each other by PCA (using 8 PCs with ~80% of sample variance; Pillai trace test statistic, F = 2.77; P = .009, Figure 4). At ART initiation, the groupwise comparison remained distinct (n = 85; F = 156.0, P = .001). The > 90-day group was distinct from the 31–90-day group (P = .02, corrected) and the ≤ 30-day group (P < .001), although 31–90 and ≤ 30-day groups did not differ from each other (P = 1.0). After 24 weeks on ART, the 3 groups separated even further (n = 69; F = 15.1, P < .0001), and those who started > 90 days from EDDI continued to separate from those started 31–90 days (P < .0001) and those started ≤ 30 days (P < .0001). However, these latter 2 groups remained indistinct from each other (P = 1.0).

Figure 4.

Principal component analysis (PCA) plot showing multivariate variation in immune activation markers between as-treated groups in the Sabes Study. PCA was undertaken on all 23 biomarkers tested in plasma at time of study enrollment/randomization (A), at time of ART initiation (B), and after 24 weeks of ART (C). The principal components (PC) 1 and 2 are displayed, and between-group tests for difference in multivariate outcome were performed for PC1–8 (describing ~80% in total variance in biomarkers) using multivariate analysis of variance, as all PCs conformed to normalcy assumptions. At randomization, the only statistically different pairwise comparison was between the ≤30 and >90-day group (P value <.05, corrected by Bonferroni for multiple comparisons). At both time of ART start and after 24 weeks on ART, the >90-day group was different from each of the more-immediately treated groups, but the ≤30 and 31–90 day groups were not different from each other. Abbreviations: ART, antiretroviral therapy; EDDI, estimated date of detectable infection.

DISCUSSION

We performed a randomized study to evaluate the impact of initiating ART during or just after acute HIV. We here evaluated clinical and immunologic outcomes and, in a separate publication [14], described the feasibility of initiating ART during acute infection and model the public health implications thereof. Subsequent to completion of Sabes enrollment, the WHO and Peruvian national guidelines now recommend initiating ART for all people living with HIV within 1 week of diagnosis, although recommendations largely stem from treatment studies without information on duration of HIV infection. Whether beginning ART during acute infection confers individual health benefits above and beyond reducing transmission risk, was previously less clear. Overall, persons who waited only 24 weeks to initiate ART after diagnosis of early HIV infection experienced more symptoms, including both AIDS-defining and non-AIDS-defining opportunistic infections and common respiratory infections. Persons who started ART during acute infection were no more likely to experience ART-associated AEs, although there was a small excess risk of severe transaminase elevations compared to those who started ART in postacute infection. Almost one-quarter of persons randomized to Deferred ART initiated treatment early due meeting contemporary immunologic or clinical criteria for ART initiation within just weeks after randomization, and 4.5% never started ART during the first year after diagnosis. Every participant randomized to start ART immediately did so within 6 days. Participants initiating ART immediately also had improved CD4 cell counts and were more likely to normalize their CD4/CD8 cell ratio by 24 weeks on ART. In as-treated analysis, persons who started ART > 90 days from EDDI were less likely to normalize their CD4/CD8 cell ratio compared with those who started ART within 30 days, and their inflammatory profiles remained distinct after 24 weeks of ART. Adverse events were lowest in those starting ART within 30 days of EDDI. Although CD4 counts and inflammatory profiles improved most markedly in those treated within 30 days of detectable infection, differences between those treated within 30 versus 31–90 days were less striking. Taken together, the intent-to-treat and as-treated analyses demonstrate that deleterious effects can be seen in those starting ART more than 1 month from incident HIV infection.

To our knowledge, this is the only modern study that randomized time of ART initiation in participants with incident HIV infection. Considering these results, current WHO recommendations for universal ART, and strong evidence that treating primary HIV interrupts HIV transmission [17, 18], it is unlikely that such a randomized study would ever be replicated. Other large cohorts with acute infections are unable to answer many questions addressed by this study due to bias introduced by diagnosis of acute HIV in persons presenting with known recent exposure or symptomatic ARS, and because treatment of incident infections previously was guided by clinical indication. In Sabes, randomized ART initiation following monthly HIV screening allowed for homogeneous demographics and treatment allocation at study entry. Longitudinal monthly HIV screening also allowed us to precisely time HIV acquisition, often within a range of days, even in participants who were seropositive at diagnosis. Additionally, we used PCA to evaluate the joint outcomes of many biomarkers in a biologic system where a standard multivariate model would attempt to adjust co-linear and synergistic immune signals.

The main limitation of this study is the short duration of 48 weeks, which allowed only comparisons after 24 weeks of ART for as-treated groups, or 48 versus 24 weeks of intended ART for the randomization arms, respectively. The follow-on MERLIN study (NCT02744040) will address longer-term outcomes over 5 years, including whether immediate ART has lasting impact on the HIV latent reservoir and gastrointestinal microbiome. We lacked sufficient evaluations of sCD14, D-dimer, and hsCRP at week 24 in persons who started ART > 90 days from EDDI, such that we were unable to include these biomarkers some analyses. Because persons treated within 1 month of EDDI started with higher levels of HIV-1 RNA and many inflammatory biomarkers compared to persons starting ART > 90 days, even though they had steeper declines in these analytes, their levels were overall similar after 24 weeks of ART. Longer follow-up will be important because biomarkers are highly kinetic during early infection. However, even though the latest-treated participants had improved biomarkers prior to ART initiation, they did not recover their CD4/CD8 ratio as the earliest-treated participants did. Finally, because we intentionally enrolled persons assigned male at birth in the high-risk HIV-negative cohort, we could not evaluate cisgender women in this study.

This study recapitulates many of the findings of the TEMPRANO and INSIGHT-START studies, but in the setting of incident HIV infection; persons assigned to immediate ART were significantly more likely to start ART at the prescribed time or at all, and experienced fewer non-AIDS events [1, 2]. Although this study had few severe non-AIDS events, including no cases of cardiovascular or cancer outcomes, as predictable in a young population with early HIV, decrements in CD4/CD8 cell ratio and elevated IL-6 have previously been strongly associated with risk of future severe non-AIDS events [19–21]. We conclude that early ART has clinical and immunologic benefits, and those benefits are maximal when ART is started within 1 month of infection. This study reinforces WHO recommendations for immediate ART initiation with data explicitly to support this strategy for incident infection.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgements. The authors thank Shelley Facente and the Consortium for the Evaluation and Performance of Human Immunodeficiency Virus [HIV] Incidence Assays (CEPHIA group) for use of their online HIV incidence estimation tool. The authors gratefully acknowledge antiretroviral therapy donation from Merck & Co and Gilead Sciences Inc.

Disclaimer. Neither pharmaceutical company participated in the study design, execution, collection or interpretation of data, or decision to publish results. The study team and corresponding author had access to all data and final responsibility for decision to submit for publication. A de-identified dataset, data dictionary, and study protocol are available as of the publication date of this manuscript, on request to Dr Ann Duerr, aduerr@fredhutch.org. We request an interinstitutional data access agreement and brief plan/disclosure of intended use of said data requested.

Financial support. Funding for this project was supported by the National Institute on Drug Abuse R01DA032106 (A. D.); National Institute of Allergy and Infectious Diseases K23AI129659 (R. B. I.). Antiretroviral drugs were donated by Gilead Sciences, Inc, and Merck & Co.

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.