The impact of viraemia on inflammatory biomarkers and CD4 cell subpopulations in HIV-infected children in sub-Saharan Africa

Abstract

Objective

To determine the impact of virological control on inflammation and CD4 depletion among HIV-infected children initiating antiretroviral therapy (ART) in sub-Saharan Africa.

Design

Longitudinal cohort study.

Methods

In a substudy of the ARROW trial (ISRCTN24791884), we measured longitudinal HIV viral loads, inflammatory biomarkers (CRP, TNF-α, IL-6, soluble CD14), and (Uganda only) whole blood immunophenotype by flow cytometry in 311 Zimbabwean and Ugandan children followed for median 3.5 years on first-line ART. We classified each viral load measurement as consistent suppression, blip/post-blip, persistent low-level VL or rebound. We used multi-level models to estimate rates of increase or decrease in laboratory markers, and Poisson regression to estimate incidence of clinical events.

Results

Overall, 42% children experienced viral blips, but these had no significant impact on immune reconstitution or inflammation. Persistent detectable viremia occurred in one-third of children and prevented further immune reconstitution, but had little impact on inflammatory biomarkers. Virological rebound to ≥5000 copies/mL was associated with arrested immune reconstitution, rising IL-6 and increased risk of clinical disease progression.

Conclusions

As viral load testing becomes more available in sub-Saharan Africa, repeat testing algorithms will be required to identify those with virological rebound, who need switching to prevent disease progression, whilst preventing unnecessary second-line regimen initiation in the majority of children with detectable viremia who remain at low risk of disease progression.

Introduction

Of the 1.8 million children living with HIV globally, over 80% live in sub-Saharan Africa. Although antiretroviral therapy (ART) has transformed clinical outcomes in HIV-infected children[1], only half were receiving treatment in 2017[2]. World Health Organization (WHO) recommendations[3] for universal treatment mean that large numbers of HIV-infected children still need to start ART in sub-Saharan Africa.

HIV infection is characterised by inflammation and CD4 depletion, which underlie mortality in both children[4] and adults[5–8]. Virological suppression following ART leads to increases in CD4 count and reductions in morbidity and mortality[1]. Although there has been considerable research in adults[9–16], much less is known about the relationships between CD4 reconstitution, inflammation and viral load (VL) dynamics following ART initiation in children, particularly in sub-Saharan Africa.

We therefore characterised the interplay between VL suppression and loss of suppression, immune reconstitution, inflammation and clinical disease progression in a large cohort of HIV-infected children starting ART in Uganda and Zimbabwe.

Methods

This was a substudy of the ARROW trial (ISRCTN24791884), in which 1206 children (age 3 months-17 years), eligible for first-line ART using WHO 2006 criteria[17], were recruited in Uganda and Zimbabwe[18]. The trial tested four HIV management strategies[18–20]. Children initiated ART with a regimen containing lamivudine+abacavir plus a non-nucleoside-reverse-transcriptase-inhibitor (NNRTI), and had CD4 counts assayed 12-weekly. Long-term, two-thirds received 2NRTI+NNRTI maintenance, and one-third 3NRTIs.

Immunology substudy

From June 2008, 316 children (97%) were enrolled in this substudy (Figure, Supplemental Digital Content 1). HIV VL was assayed retrospectively on cryopreserved plasma taken pre-ART, at weeks 4, 24, 36 and 48 post-ART initiation, then every 24 weeks; additional VLs were assayed in children randomized to once- versus twice-daily lamivudine+abacavir[20]. The lower limit of detection was 80 copies/mL because many low-volume samples were diluted 1:2. No VLs were used in real-time to make treatment decisions[18]. Children fulfilling WHO clinical or immunological criteria for failure switched to boosted protease-inhibitor-containing second-line ART.

Inflammatory biomarkers (CRP, TNF-α, IL-6, soluble CD14) and IL-7 were assayed by ELISA on cryopreserved plasma taken pre-ART and then every 24 weeks. In Uganda, whole blood immunophenotyping by flow cytometry was done pre-ART and at weeks 4, 12, 24, 36 and 48 post-ART initiation, then 24-weekly. Assay details are in Text, Supplemental Digital Content 2.

Analysis

Analysis considered all measurements on first-line ART. To estimate the impact of long-term VL dynamics on immune responses, we first classified each individual VL measurement as consistent suppression, blip/post-blip, persistent low-level VL (pLLVL) or rebound[21]. We pre-defined consistent suppression as <10,000 copies/mL or >1 log10 drop at week-4, <5000 copies/mL at week-24, declining or <400 copies/mL at week 36 and subsequently <80 copies/mL at all measurements. VL measurements ≥80 copies/mL were classified as a blip if the child had a previous VL <80 copies/mL and either subsequently re-suppressed <80 copies/mL (one or two VL≥80 copies/mL allowed; n=150) or their last measurement was a single value ≥80 copies/mL (n=31; Supplementary Methods). Those not re-suppressing <80 copies/mL but remaining <5000 copies/mL were defined as persistent low-level VL (pLLVL), whereas rebound was defined as confirmed VL ≥5000 copies/mL (following WHO 2010 guidelines[22]). We recorded the greatest degree of viral replication observed to have occurred, meaning children could move from response to blip, pLLVL and rebound states but not backwards through these states, i.e. if a child had ever experienced a VL blip they were subsequently classified as having previously ‘blipped’.

Compared to those on 2NRTI+NNRTI long-term, children on 3NRTI maintenance were more likely to experience pLLVL and rebound[21]; however, as our goal was to estimate the impact of VL dynamics on biomarkers regardless of cause, we did not adjust for first-line regimen or other randomizations. We used multi-level models to estimate rates of increase or decrease for each of the following outcomes: inflammatory biomarkers; CD4%; CD4-for-age (the ratio of observed:expected count for the child’s age[23]); CD8-for-age; CD4/CD8 ratio; CD4 subpopulations; weight-for-age; and height-for-age (details in Text, Supplemental Digital Content 2.).

Incidence of clinical events (WHO stage 3/4 events and deaths) were estimated using Poisson regression, and the impact of time-dependent biomarker values on clinical outcomes estimated using time-updated Cox regression. All analyses used Stata 15.1 (StataCorp). All p-values are two-sided.

Ethical approvals

Caregivers gave written informed consent for all children; if older children (8-17 years) were aware of their HIV status, they gave additional assent or consent following national guidelines. ARROW was approved by Research Ethics Committees in Uganda, Zimbabwe and the UK.

Results

Of 316 children in the immunology substudy, 3 (0.9%) died before week 4. Of the remaining 313 children, 311 (99.4%) achieved an initial VL response <5000 copies/mL at week-24 and were included in analyses; 2 children never achieved VL response. Median (IQR) age at ART initiation was 5.4 years (2.2,9.4) and median CD4% 13% (8%,19%) (Table 1). Median follow-up on first-line ART was 3.6 years (IQR 3.5,3.7). Only 4 (1.3%) children switched to second-line regimens during follow-up. Cytokine measurements were available for 99.4% (n=309) children pre-ART and for 83.6% of subsequent 24-weekly measurements (median (IQR) 7 (7,8) time-points per child).

Table 1. Characteristics at ART initiation of all children with initial VL response, and those subsequently experiencing viral load blips, persistent low level viral load and/or rebound.

Factor (at ART initiation)	Total with initial VL response N=311 Median (IQR) or n (%) [number missing]	Ever had VL blip N=130 Median (IQR) or n (%) [number missing]	P (ever vs never had VL blip)	Ever had persistent low level viral load N=59 Median (IQR) or n (%)	P (ever vs never had pLLVL)	Ever had rebound N=59 Median (IQR) or n (%)	P (ever vs never had rebound)
Male	146 (46.9%)	62 (47.7%)	0.82	25 (42.4%)	0.43	29 (49.2%)	0.71
Country/centre      Uganda/A      Uganda/B      Uganda/C      Zimbabwe/D	50 (16.1%) 79 (25.4%) 78 (25.1%) 104 (33.4%)	14 (10.8%) 26 (20.0%) 26 (20.0%) 64 (49.2%)	<0.0001	8 (13.6%) 18 (30.5%) 15 (25.4%) 18 (30.5%)	0.75	13 (22.0%) 15 (25.4%) 18 (30.5%) 13 (22.0%)	0.15
Age (years)	5.4 (2.2, 9.4)	5.3 (2.3, 8.4)	0.17	6.5 (2.7, 10.2)	0.34	7.8 (1.7, 12.3)	0.14
WHO stage 1/2 3/4	100 (32.2%) 211 (67.8%)	39 (30.0%) 91 (70.0%)	0.49	17 (28.8%) 42 (71.2%)	0.54	17 (28.8%) 42 (71.2%)	0.54
Real-time monitoring CD4 monitoring No CD4 monitoring	154 (49.5%) 157 (50.5%)	64 (49.2%) 66 (50.8%)	0.93	30 (50.8%) 29 (49.2%)	0.82	30 (50.8%) 29 (49.2%)	0.82
Long-term ART      2NRTI+NNRTI      3NRTI	204 (65.6%) 107 (34.4%)	93 (71.5%) 37 (28.5%)	0.06	20 (33.9%) 39 (66.1%)	<0.0001	28 (47.5%) 31 (52.5%)	0.001
Viral load (copies/mL)	216800 (67000, 641200)	273400 (94300, 704300)	0.10	249100 (92200, 726200)	0.30	275100 (94800, 1039200)	0.22
TNFa (pg/mL)	23.9 (19.3, 29.7) [3]	25.4 (21.0, 30.1) [2]	0.02	25.5 (19.8, 30.5)	0.29	24.8 (18.7, 32.9)	0.60
IL-6 (pg/mL)	5.8 (4.5, 8.2) [5]	6.3 (4.9, 9.0) [3]	0.007	5.5 (4.6, 7.2)	0.51	6.3 (4.7, 10.6) [1]	0.20
CRP (mg/L)	4.0 (1.4, 13.5) [2]	4.3 (1.5, 13.2) [2]	0.63	4.4 (1.6, 14.4)	0.24	6.3 (1.6, 29.4)	0.03
sCD14 (mg/L)	2.1 (1.6, 2.5) [2]	2.1 (1.6, 2.5) [2]	0.38	2.1 (1.6, 2.4)	0.81	2.3 (1.8, 2.8)	0.03
IL-7 (pg/mL)	9.8 (3.4, 17.5) [14]	10.6 (4.1, 18.7) [3]	0.09	9.8 (5.1, 18.5)	0.51	8.7 (4.4, 16.1) [4]	0.89
CD4%	13 (8, 19)	13 (10, 18)	0.78	13 (9, 16)	0.73	10 (5, 15)	0.003
CD4-for-age	0.3 (0.2, 0.5)	0.3 (0.2, 0.5)	0.36	0.3 (0.2, 0.4)	0.37	0.2 (0.1, 0.3)	0.0005
CD8-for-age	2.1 (1.4, 3.0)	2.0 (1.3, 2.9)	0.65	2.0 (1.4, 2.8)	0.40	2.0 (1.4, 2.9)	0.94
CD4:CD8	0.3 (0.1, 0.4)	0.3 (0.2, 0.5)	0.37	0.2 (0.2, 0.4)	0.49	0.2 (0.1, 0.3)	0.0005
% CD4 CD45RA+*	48.4 (31.2, 62.2) [14]	53.3 (35.0, 64.0) [5]	0.20	46.5 (27.2, 57.1) [3]	0.15	42.6 (23.0, 54.2) [4]	0.009
% CD4 CD45RA+CD31+*	36.9 (22.4, 49.7) [14]	40.9 (28.3, 50.8) [5]	0.22	33.6 (20.0, 45.7) [3]	0.22	30.6 (18.0, 42.5) [4]	0.02
% CD4 CD45RA+CD31-*	7.5 (4.2, 14.1) [14]	9.5 (4.4, 15.0) [5]	0.19	6.6 (3.9, 13.8) [3]	0.39	5.7 (3.0, 9.0) [4]	0.02
% CD4 CD45RA-CD31-*	34.5 (24.6, 45.7) [14]	32.5 (24.8, 41.8) [5]	0.21	38.8 (25.9, 44.8) [3]	0.31	40.9 (27.7, 47.2) [4]	0.0504
% CD4 Ki67+*	8.7 (4.8, 14.3) [12]	9.7 (5.6, 14.1) [4]	0.44	8.0 (4.5, 16.7) [4]	0.84	13.2 (5.7, 23.5) [4]	0.004
% CD4 HLADR+*	8.7 (5.1, 12.9) [16]	8.7 (4.8, 12.5) [5]	0.58	8.5 (4.8, 12.5) [4]	0.60	9.4 (6.1, 15.3) [6]	0.20
Weight-for-age	-2.2 (-3.3, -1.3)	-2.2 (-3.3, -1.5)	0.84	-2.1 (-3.9, -1.5)	0.25	-2.2 (-3.3, -1.5)	0.45
Height-for-age	-2.5 (-3.4, -1.6)	-2.8 (-3.4, -1.9)	0.04	-2.5 (-3.4, -1.8)	0.83	-2.1 (-3.3, -1.3)	0.10

^*Uganda only.

Note: Numbers in square brackets indicate missing data (Uganda only for immunophenotyping). Children experiencing more than one of viral load blips, persistent low-level viral load and rebound included in all relevant columns, see Methods for definitions. Chi-squared or rank-sum tests used to compare categorical and continuous factors at ART initiation, respectively. Some of these data have previously been shown in²⁵

Overall changes in CD4 populations and inflammatory biomarkers on first-line ART

Children showed rapid CD4 reconstitution on ART (Figure 1A). CD4% increased by 10% [95%CI 9-10%] over the first 12 weeks on ART, 7% [6-8%] through week 72, then 0.2% [0.2-0.3%] every 12 weeks throughout follow-up (all p<0.001). CD4-for-age and CD4/CD8 ratio increased similarly, with mean CD4/CD8 ratio stabilising at just over 1. CD8-for-age declined more slowly, stabilising at around 1.6 by week 96.

Figure 1. Overall changes in biomarkers and VL on first-line ART.

(a) CD4 and CD8; (b) CD4 subpopulations; (c) IL-7 proliferating and activated CD4 cells; (d) inflammatory biomarkers; and (e) VL.

CD4 reconstitution was driven by increases in the proportion of naïve (CD45RA+) CD4 cells over the first 24 weeks (p<0.001), most of which were CD31+, with slower rises subsequently (p<0.001) (Figure 1B). Conversely, the proportion of CD45RA-CD31-CD4 cells decreased over the first 24 weeks (p=0.001), then more slowly throughout follow-up (p<0.001).

IL-7 decreased substantially over the first 24 weeks (p<0.001), then more slowly throughout follow-up (p<0.001) (Figure 1C). The proportion of proliferating (Ki67+) CD4 cells dropped strikingly by 8% [7-9%] over the first 48 weeks (p<0.001) (Figure 1C), then rose by 1% [1-2%] to week 72 (p<0.001), driven by increases in memory (CD45RA-) CD4 cells, and did not change significantly thereafter (p=0.58).

The proportion of activated (HLA-DR+) CD4 cells decreased more slowly and over a longer period, by 5% [4-6%] over the first 72 weeks on ART (p<0.001), and then by 0.4% [0.3-0.5%] every 24 weeks thereafter (p<0.001) (Figure 1C).

Different trajectories on first-line ART were observed for each of the inflammatory markers (Figure 1D). TNF-α declined over the first 48 weeks then rose again to week 72, before subsequently declining more slowly (all p<0.001). CRP also dropped substantially over the first 48 weeks, but then rose through week 96 and subsequently declined (p≤0.001). IL-6 dropped by 28% [24-32%] over the first 24 weeks (p<0.001), but did not change significantly thereafter (p=0.69). sCD14 declined much more slowly throughout follow-up (p=0.001).

At ART initiation, CD4% was strongly associated with cell populations and biomarkers[4]. Despite this, initial responses to ART were broadly similar, and normalization of each pathway occurred regardless of CD4 count (Figure, Supplemental Digital Content 3) or age at ART initiation (data not shown).

Taken together, the post-ART dynamics showed a sharp rise in CD4 cells, particularly over the first 12 weeks, driven predominantly by recent thymic emigrants, and an initial decline in activated and proliferating CD4 cells over 24-48 weeks, accompanied by substantial decreases in IL-7 and inflammatory biomarkers (except sCD14). After 48 weeks of ART, the proportion of proliferating CD4 cells, and concentrations of TNF-α and CRP, began to rise.

Despite good immunological responses, a proportion of children experienced VL blips and loss of VL control, to either pLLVL (<5000 copies/mL) or rebound (confirmed ≥5000 copies/mL) (Figure 1E). Fastest transitions occurred between consistent VL suppression and blip, and between pLLVL and rebound (Figure, Supplemental Digital Content 4). We therefore investigated how changes in CD4 subpopulations and biomarkers related to consistent VL suppression or VL non-suppression after initial virologic response.

Consistent VL suppression

Of 311 children achieving an initial VL response, 103 (33.1%) consistently sustained VL<80 throughout follow-up, 93 (29.9%) experienced pLLVL or rebound, and 115 (37.0%) had one or more VL blips (Figure 1E). Whilst VL was consistently maintained <80 copies/mL, increases in CD4%, CD4-for-age and CD4/CD8 ratio on ART (Figure, Supplemental Digital Content 5) were only slightly greater than the whole cohort, which included children with blips, pLLVL and rebound (Figure 1).

After children were stable on ART from 48 weeks, those with consistent VL suppression had ongoing increases in CD4%, CD4-for-age and CD4/CD8 ratio (Figure 2), while their CD8-for-age, IL-7 and TNF-α continued to decrease (Figures 2-3, and Figure, Supplemental Digital Content 6). There was no evidence of changes in other inflammatory biomarkers (IL-6, CRP, sCD14) once children were stable with consistent VL suppression (p>0.3). The proportion of naïve (CD45RA+) and recent thymic emigrant (CD45RA+CD31+) CD4 cells continued to increase significantly, while the proportion of CD45RA-CD31- and CD45RA+CD31-CD4 and HLA-DR+ cells decreased (Figure, Supplemental Digital Content 7).

Figure 2. CD4 and CD8 over time with consistent VL suppression, previous VL blips, pLLVL and rebound.

Figure 3. Inflammatory cytokines over time consistent VL suppression, previous VL blips, pLLVL and rebound.

Impact of VL blips

Overall, 130 (41.8%) children experienced blips to a median (IQR) 200 (110, 750) copies/mL; 19 (14.6%) had already experienced a blip by week-48 and 43 (33.1%) had more than one blip on first-line ART. After their first blip, children still had significant increases in CD4%, CD4-for-age and CD4/CD8 ratio, and non-significant decreases in CD8-for-age, which did not differ from those during consistent suppression (p>0.7, Figure 2, Table 2). There was also no evidence that changes in IL-6, sCD14, TNF-α and IL-7 differed between children with blips and children with consistent suppression (p>0.2; Table 2). If anything, CRP actually decreased faster in those with blips compared to consistent VL suppression (p=0.03). After their first blip, children still had non-significant increases in CD45RA+CD31+ and naïve CD4 cells, and decreases in CD45RA-CD31- and HLA-DR+ CD4 cells, which did not differ from those during consistent suppression (p>0.3). Ki67+ CD4 cells decreased (p=0.02), whereas they did not change during suppression (suppression vs. blips p=0.01).

Table 2. Comparisons of trajectories in CD4 subpopulations, inflammatory biomarkers, weight-for-age and height-for-age according to VL dynamics.

	Consistent suppression vs. blips vs. pLLVL vs. rebound	Consistent suppression vs. blips	Consistent suppression vs. pLLVL	Consistent suppression vs. rebound	Rebound vs. pLLVL	Consistent suppression+blips vs. pLLVL+rebound
CD4%	p<0.0001	0.94	0.003	<0.0001	0.85	<0.0001
CD4-for-age	0.004	0.75	0.046	0.003	0.34	0.0001
CD8-for-age	0.17	0.71	0.053	0.17	0.42	0.01
CD4/CD8 ratio	<0.0001	0.72	0.02	<0.0001	0.45	<0.0001
% CD4 CD45RA+	0.83	0.61	0.64	0.54	0.43	0.33
% CD4 CD45RA+CD31+	0.89	0.64	0.96	0.50	0.62	0.32
% CD4 CD45RA+CD31-	0.41	0.61	0.59	0.15	0.14	0.32
% CD4 CD45RA-CD31-	0.31	0.41	0.12	0.83	0.16	0.84
TNF-α	0.49	0.68	0.35	0.30	0.90	0.07
IL-6	0.03	0.20	0.17	0.04	0.008	0.17
CRP	0.04	0.03	0.34	0.28	0.11	0.18
sCD14	0.0006	0.23	0.0003	0.98	0.002	0.02
IL-7	0.21	0.94	0.26	0.07	0.73	0.01
% CD4 Ki67+	0.02	0.01	0.02	0.29	0.68	0.34
% CD4 HLADR+	0.53	0.34	0.42	0.93	0.66	0.98
Height-for-age Z-score	0.03	0.049	0.85	0.01	0.102	0.13
Weight-for-age Z-score	0.37	0.92	0.98	0.09	0.15	0.38

Note: Shading indicates significance of p-values: dark grey: p ≤ 0.05; light grey: 0.05 < p ≤ 0.1. Modelled trajectories shown in Figures 2-5.

Impact of pLLVL

Overall, 59 (19.0%) children experienced pLLVL after median 72 weeks on ART (IQR 36,120), with median (IQR) VL 720 copies/mL (IQR 150,2430) and CD4% 31% (24%,37%) during pLLVL. Over half (53.6%) of time in pLLVL was spent with most recent VL measurement <1000 copies/mL. VL did not increase over time during pLLVL (p=0.32). Although pLLVL did not lead to CD4 depletion, it arrested further immune reconstitution (Table 2): none of CD4%, CD4-for-age, CD4/CD8 ratio or CD8-for-age changed significantly over time during pLLVL (Figure 2), although the proportion of memory (CD45RA-CD31-) and activated (HLA-DR+) CD4 cells continued to decrease (Figure, Supplemental Digital Content 6; and Figure, Supplemental Digital Content 7). Similar to findings after a first blip, Ki67+ decreased (p=0.04), whereas it did not change during suppression (suppression vs. pLLVL p=0.02); there was no evidence that changes in any other CD4 sub-population during pLLVL were different to those during consistent VL suppression (p>0.1). IL-7, IL-6, CRP and TNF-α did not change significantly over time during pLLVL (Figure 3). By contrast, during pLLVL sCD14 decreased significantly, albeit by a relatively small absolute amount. Although the absolute level of IL-6 was generally higher in pLLVL than in VL suppression (or after experiencing blips), there was no evidence of differences in how these biomarkers changed over time (p>0.1, Table 2).

Impact of rebound

Overall, 59 (19.0%) children experienced rebound after median 72 weeks on ART (IQR 36,120), with median (IQR) VL 31910 copies/mL (11190,86430) and CD4% 23% (16%,31%) during rebound (Table 1). Having experienced rebound led to a significant effect on the overall levels of CD4, CD8, CD4 subpopulations, and proportions of proliferating (Ki67+) or activated (HLA-DR+) cells, but only a modest impact on inflammatory cytokines (Figures 2-3; Supplemental Digital Content 8). There was no evidence that VL changed over time in rebound (p=0.14). Similarly to pLLVL, there was no evidence of progressive declines in CD4% during rebound (0.1% decrease/year [95%CI 1.0% decrease-0.9% increase], p=0.91), or in CD4-for-age and CD4/CD8 ratio, but changes were significantly smaller than the increases seen during suppression (p<0.001, Figure 2, Table 2). There was no significant change in CD8-for-age or Ki67+, with no evidence of any differences from changes seen in consistent suppression or pLLVL (p>0.1). HLA-DR+ and TNF-α continued to decrease despite VL rebound, with no evidence of difference in trajectory compared with consistent VL suppression (p>0.3). In contrast, IL-6 increased significantly faster during rebound than during pLLVL or suppression (p<0.05); CRP changed similarly to IL-6 but differences were not significant. There was no evidence of a change in sCD14 or IL-7 during rebound.

Clinical events and growth by VL dynamics

After week-48, clinical events (WHO stage 3/4 events and deaths) were relatively rare (n=10) but occurred at a greater rate during rebound (53/1000 child-years [95%CI 20-116]) than during suppression (5/1000 child-years [95%CI 1-17]) and after blips (9/1000 child-years [95%CI 1-34]; rebound vs. suppression: p=0.003); there were no events in children with LLVL [95%CI 0-46]. Both deaths after week-48 (1 pulmonary TB, 1 pneumonia) occurred in children with rebound. Combining LLVL (no events) and blips, the increased risk associated with rebound persisted after adjusting for time-updated CD4-for-age (rebound vs. suppression: hazard ratio (HR)=5.94 [95% CI 0.91-38.64], p=0.06). There was no evidence of additional effects of inflammatory cytokines (p>0.17). In the subset of 205 children with immunophenotyping, there was some evidence that high HLA-DR+ was associated with increased risk of clinical events (HR=1.13 [0.98-1.30], p=0.09, adjusting for VL group and CD4-for-age; p=0.04 adjusting for VL group alone).

After week-48, height-for-age increased regardless of VL suppression (all p<0.0001, Figure, Supplemental Digital Content 9), but increases were slower during rebound than suppression (p=0.01) (Table 2). Weight-for-age changes were much smaller after week-48, with no evidence of changes in weight-for-age during rebound (p=0.39).

Discussion

In this study of HIV-infected children in Uganda and Zimbabwe, most had excellent virological, immunological and clinical responses on first-line ART. Children showed rapid immune reconstitution with naïve CD4 cells, and a rapid decline in activated and proliferating T-cells and most inflammatory biomarkers. Despite good initial virological suppression, only one-third remained consistently undetectable long-term; 42% had one or more blips, and almost one-third had persistent detectable viraemia. We found no evidence that blips were deleterious, whereas persistent detectable viraemia prevented further immune reconstitution, and virological rebound (≥5000 copies/mL) was associated with more inflammation and clinical disease progression.

CD4 counts rose rapidly following ART initiation, with concomitant declines in IL-7, peripheral T-cell proliferation and activation. Immune reconstitution was mainly driven by CD45RA+CD31+ CD4 cells, which are predominantly recent thymic emigrants[24, 25], together with a decline in memory CD4 cells. Most inflammatory biomarkers (IL-6, CRP, TNF-α) also declined over the first 48 weeks on ART. Normalization of each pathway did not strongly depend on the baseline CD4 count or age, suggesting that in settings where children continue to present with advanced disease[26–29], excellent restoration of immune and inflammatory pathways is achievable.

The causes of inflammation in HIV-infected children are poorly understood, but may include co-infections and enteropathy as well as HIV itself[30]. There are complex inter-relationships between inflammatory biomarkers; for example, pre-ART levels of IL-6 or CRP predict mortality on ART, whilst levels of TNF-α and sCD14 do not[31]. In this analysis, the pattern of changes in inflammatory markers showed distinct differences. All biomarkers declined on ART over 48 weeks, although reductions in soluble CD14 were very small, similar to other cohorts[32, 33]. After 48 weeks, IL-6 remained stable, whilst the concentrations of TNF-α and CRP rose at the population level, before slowly declining again. Changes were similar in children who remained virologically suppressed, suggesting that these population-level effects could not be explained by temporary or ongoing loss of virological control. It is striking that detectable viremia only impacted inflammatory marker levels once VL was quite high (≥5000 copies/mL); at lower levels, loss of virological control had little effect on the overall level or subsequent evolution of inflammatory markers. This suggests that inflammatory pathways are controlled by a complex system of homeostatic regulation, which tends to maintain an inflammatory equilibrium, similar to the stability of the HIV viral load set-point; it is only once the perturbation is sufficiently great, due to high-level viral load rebound, that control of the system is disrupted. Once virological rebound occurred, and inflammatory markers stabilised at higher levels, WHO stage 3/4 events and deaths occurred at a greater rate, independently of CD4-for-age. In HIV-infected adults, higher levels of inflammation on ART are associated with elevated morbidity and mortality and there is considerable interest in evaluating adjunctive therapies to reduce inflammation[34, 35]. In children we did not find evidence that inflammation was associated with morbidity independently of VL, but few children had serious morbid events on ART.

Despite limited laboratory (and no VL) monitoring, children showed a good initial virological response to ART, as reported in other African cohorts[36], but frequently had episodes of detectable viraemia during median 3.5 years of follow-up. This was often due to blips – i.e. isolated detectable HIV RNA followed by a return to virological suppression[37]. We found that blips were not deleterious: CD4 populations continued to normalize during and following blips and there was no rise in inflammatory biomarkers or increase in the rates of subsequent progression to pLLVL or rebound. To our knowledge, no previous studies have evaluated the impact of blips in HIV-infected children in sub-Saharan Africa. The mechanisms underlying blips may relate to assay variability at the limits of detection[15], statistical variation[11], release of virus from long-lived reservoirs[14, 16], intercurrent infections or vaccination[9, 10] and incomplete ART adherence[12], although whether the drivers are the same in adults and children is unknown. It is interesting that whilst the overall increase in CD4 cells was comparable to that found in suppressed individuals, we saw a reduction in inflammation during blips, with declines in CRP and Ki67+ CD4 cells that were greater than among those with virological suppression.

One-third of children had persistent detectable viraemia on ART; however, this was only deleterious for disease progression when ≥5000 copies/mL – a level higher than recent guidelines suggest for switch to second-line therapy[38]. In cross-sectional and longitudinal cohort studies, detectable viraemia has been reported in 24-34% children on ART[39–42] and switch to second-line therapy for ART failure is often delayed[43]. We show here that while persistent viral replication (<5000 copies/mL) did not cause a decline in CD4 cells and had no substantial impact on inflammatory biomarkers or clinical disease progression, it prevented further immune reconstitution and rise in CD4/CD8 ratios. This indicates that persistent low-level virus prevents children from realising their full potential to achieve normal or high levels of CD4 cell immune reconstitution. This may become particularly important when children enter adulthood.

Children with higher levels of viraemia (≥5000 copies/mL) had poorer immune reconstitution and increased levels of IL-6, a key inflammatory biomarker associated with mortality[31]. These children had greater risk of clinical disease progression (WHO stage 3/4 events or death), and slower height gain than children with virological suppression. Where virological monitoring is available, it clearly has a role in identifying children with persistently detectable viraemia at a reasonably high level, in whom immunological and clinical disease progression and increased inflammation will develop over time. Our results suggest that switching treatment at lower VL thresholds may not necessarily provide advantages over the long-term given the limited treatment options for children in low-income countries, and the challenges of adherence[44].

This study had strengths and weaknesses. ARROW is the largest trial of children initiating ART in sub-Saharan Africa. Since VL was not measured in real-time, our data provide a realistic impression of the long-term impact of blips and virological failure in a large cohort where adherence counselling was not undertaken following detectable viraemia. However, several different VL platforms were used across trial sites, which can affect the chance of detecting low-level blips[15, 45]. We had good clinical data, including independently adjudicated WHO stage 3/4 disease and deaths, but event rates on ART were low, limiting our ability to establish clinical sequelae of each virological state. However, the low clinical event rates also support no major clinical harms being associated with persistent viraemia.

In summary, we show that most children starting ART in sub-Saharan Africa maintained consistent virological suppression or only experienced viral blips, which were not deleterious and do not warrant switch to second-line ART. Ongoing recovery of CD4 and CD8 parameters occurred at the same rate in children with blips as those who were consistently suppressed, and both groups maintained stable levels of inflammatory markers. In contrast, persistently detectable viraemia occurred in one-third of children. Persistent low-level viremia limited further immune reconstitution, and once ≥5000 copies/mL, children had increased inflammatory biomarkers and clinical disease progression. This highlights the importance of strategies to enhance ART adherence, since even low-level viremia hinders long-term immune recovery. Children with confirmed VL ≥5000 copies/mL should be prioritised for adherence support, due to the risk of disease progression, and switched to second-line regimens. Given the push to scale-up virological monitoring across Africa, this study shows the challenges that are likely to arise in interpreting VL results, especially if only conducted infrequently. Repeat testing algorithms are required to identify those with virological rebound, who need switching to prevent disease progression, whilst preventing unnecessary second-line regimens in the majority of children with detectable viraemia who remain at low risk of disease progression.