Reduced time to suppression among neonates with HIV initiating antiretroviral therapy within 7 days of age

Abstract

There are limited data on infants with HIV starting antiretroviral therapy (ART) in the neonatal period. We investigated the association between the timing of ART initiation and time-to-suppression among infants who tested HIV-positive and initiated ART within the first 28 days of life. The effect was estimated using cumulative probability flexible parametric spline models and a multivariable generalized additive mixed model was performed to test non-linear associations.

Forty-four neonates were included. Nineteen (43.2%) initiated ART within 7 days of life and 25 (56.8%) from 8 to 28 days. Infants treated within 7 days were 4-fold more likely to suppress earlier than those treated after 7 days (HR 4.01 [1.7;9.5]). For each week the ART initiation was delayed, the probability of suppression decreased by 35% (HR 0.65 [0.46;0.92]). Age at ART start was linearly associated with time-to-suppression. However, a linear association with normally distributed residuals was not found between baseline VL and time-to-suppression, with no association found when baseline viral loads were ≤ 5 log(10) copies/ml, but with exponential increase in time-to-suppression with > 5 log(10) copies/ml at baseline.

Starting ART within 7 days of life led to 4-fold faster time to viral suppression, in comparison to initiation in 8 to 28 days.

INTRODUCTION

Early combination antiretroviral therapy (ART) initiation in vertically Infants with HIV has had a major impact on clinical outcomes. The landmark Infants with HIV Early anti-Retroviral (CHER) study clearly demonstrated the benefits of starting ART as soon as possible after infant diagnosis compared to deferring ART until clinical criteria are met ^1,2. Moreover, early initiation of ART leads to a significantly smaller reservoir size and a better immune reconstitution ^3–9. Early ART may affect HIV pathogenesis in favourable ways to help achieve future HIV remission in some individuals ^10,11.

Most studies show faster time to viral suppression and longer sustained viral control when compared to treatment that started later ^5,12–15. However, definitions of early vary a great deal, with early ART frequently including infants started in the first 3-6 months of life. Literature reporting on ART initiation in the first year of life rarely reports on outcomes when treatment is initiated in the neonatal period (the first 28 days of life)¹².

A recent review described individual viral trajectories in a handful of infants starting ART in the neonatal period ¹². However, numbers of cases were too few to distinguish outcomes by the timing of ART initiation within this very early period and might have been biased towards the over-representation of infants who actually survived to be reported retrospectively.

Currently, point-of-care diagnosis in maternity wards is increasingly available in high prevalence settings ^16,17. With increasing access to very early ART initiation, there is a crucial need to understand just how rapidly treatment should be started within the neonatal period, and if very early initiation within 7 days of life has any long-term benefit. Here we investigated the association between the timing of ART initiation and time to viral suppression among perinatally HIV-infected neonates treated in the first 28 days of life.

METHODS

Study population

This is a retrospective analysis of longitudinal data pooled from four cohorts: Rome Cohort (Italy) ¹⁸, Imperial College Hospital (London, United Kingdom) ¹⁹, Chulalongkorn University (Thailand) ²⁰, and Spanish National HIV Paediatric Network-CoRISpe (Spain) ²¹. Each cohort has established and maintained standardized databases for paediatric HIV for many years. The analysis was approved by the Ethical Committee of Hospital 12 de Octubre (17/398).

Perinatally HIV-infected infants aged ≤ 28 days at ART initiation, with at least 2 viral load measurements and without interruptions in ART within 2 years of follow up were included. Those infants who received triple-drug postnatal prophylaxis including nevirapine and subsequently transitioned to treatment doses of ART within 15 days were considered for the purpose of analysis as starting ART from date of prophylaxis initiation ²². Intrapartum prophylaxis was defined as intravenous intrapartum AZT administered to the mother during delivery.

Endpoints and secondary variables

The primary endpoints were mortality and time to VL suppression up to 15 years of follow-up. Other virological control markers were also analyzed: virological failure, suboptimal response, blips, and spikes. Definitions were as follows: suppression was ≥2 consecutive measurements ≤50 copies/mL; virological failure was ≥2 consecutive measurements >400 copies/mL preceded by suppression; suboptimal viral response was ≥2 consecutive measurements 50-400 copies/mL preceded and followed by suppression; blips were a single measurement 50-400 copies/mL preceded and followed by suppression, and spikes were single measurements >400 copies/mL preceded and followed by suppression.

Baseline (defined as the closest measurement to start of ART) sociodemographic, clinical, virological and immunological features, were included in the analysis. Baseline VL and CD4 measurements were restricted to those on the day of ART start or before. VL measurements, CD4%, CD8%, CD4:CD8 ratio, and lymphocyte counts were routinely collected at each follow-up visit.

Laboratory Methods:

Before 2003, standard HIV RNA testing (Amplicor HIV-1 monitor; Roche Diagnostic Systems, Branchburg, N.J., USA) was used in all laboratories, with a routine lower limit of detection (LOD) of <400 copies/mL and ultrasensitive LOD of <50 copies/mL. Since 2003, participating laboratories used ultrasensitive measurements with a LOD of 20-50 copies/mL. CD4 and CD8 analyses were performed by flow cytometry at each site.

Statistical Analysis

Baseline characteristics, including ART initiation date, were described with frequencies for categorical variables and means (standard deviation (SD)) or medians (interquartile range [IQR]) for continuous variables, both in the total population and stratified by timing of ART initiation in ≤7 days and >7 days after birth. Chi-squared or Fisher tests were applied to assess differences across groups for categorical variables and Student T-test or U-Mann-Whitney tests for continuous variables.

For time-to-event-analysis, the main effect of ART initiation on time to each virologic event (suppression, virological failure, blip, and suboptimal response) were estimated by cumulative probability flexible parametric spline models with an interval-censored method. This approach, implemented in survival and flexsurv R package ²³, allows to model non-proportional hazards, a feature that is particularly useful in cases were the given outcome variable may have different effect during the time to follow-up. Moreover, this model also allows a built-in-choice to include splines so that it is possible to handle non-linear assumptions. Hazard ratios (HR) and 95% confidence intervals were estimated either for treatment groups or age at ART initiation as a continuous variable. A univariable and multivariable model were designed. The multivariable model was adjusted by baseline VL and ART regimen at treatment initiation. Variable selection for the multivariable model was performed with a stepwise Akaike Information Criterion variable selection. The site/cohort was not selected as confounder according to this criterion.

To test the type of association between age at ART initiation and time-to-suppression, only infants who ever suppressed in the follow-up were selected. To analyze possible non-linear associations, a multivariable generalized additive mixed model implemented in gamm4 package²⁴ was performed. Fixed component of the model was built including baseline VL, baseline CD4, and age at ART initiation with smoothing splines. To model individual treatment differences that provide variability, ART drug combination at initiation was included as a random intercept according to clinical priors knowledge. Backward stepwise elimination was applied to reach the final multivariable model. In those variables with significant smoothing terms, knots were selected according to a smoothing selection optimization implemented in mgcv R package ²⁵, computing an estimate of residual variance based on differing residuals that are near neighbours according to the (numeric) covariates of the smooth. Plots were built with ggplot2 R package ²⁶ and itsadug R package ²⁷. R software was used for all analysis ²⁸.

RESULTS

Participants characteristics

A total of 44 neonates born between 2000 and 2017 were included in this study. Of these newborns, 25/44 (57%) were female and 13/44 (35%) were preterm defined as <37 gestational weeks. Median birth weight was 2440 grams (IQR 2130 ; 2922 grams). The median duration of follow-up was 7.4 [IQR 2.0;11.1] years and the median time between VL checks were 3.7 [2.9-5.6] months. No infant died during follow-up. Median age at ART initiation was 15.5 [IQR 0.0; 24.2] days. Nineteen (43%) infants initiated ART ≤ 7 days of age and 25 (60%) initiated ART 8-28 days. Thirty-two infants received prophylaxis, all of them started triple ART therapy after prophylaxis. Most neonates (30/44, 70%) received their first ART as two Nucleoside Reverse Transcriptase Inhibitors (NRTIs) plus NVP and 30% received NRTIs and protease inhibitors (PI). The group treated within 7 days of life was treated initially more often with NRTIs and NNRTIs than the group treated after 7 days (94% vs. 52%, p=0.008). The combination AZT + 3TC and NVP was the most frequent regimen (27/44; 63%) (Table 1). Infants required ART switches a median of 2.5 times with no differences between groups of age at ART initiation (p=0.219).

Table 1:

Baseline sociodemographic, viral and immunologic characteristics of 44 children starting ART within 7 days versus those starting ART 8 to 28 days of life.

	Overall N=44	ART within 7 days N=19	ART after 7 days N=25	p-value1	Odd Ratio [95% CI]	p-value2
Gender:				0.295
Male	19 (43.2%)	6 (31.6%)	13 (52.0%)		0.44 [0.12;1.51]	0.195
Region				0.036
Spain	23 (52.3%)	11 (57.9%)	12 (48.0%)
Italy	3 (6.82%)	2 (10.5%)	1 (4.00%)
United Kingdom	5 (11.4%)	4 (21.1%)	1 (4.00%)
Thailand	13 (29.5%)	2 (10.5%)	11 (44.0%)
Birth weight
Median [IQR]	2440[2130;2922]	2330 [1300;2720]	2515[2258;2945]	0.157	1.00[1.00;1.00]	0.046
Preterm birth				1.000
No	24 (64.9%)	11 (64.7%)	13 (65.0%)		Ref.
Yes	13 (35.1%)	6 (35.3%)	7 (35.0%)		1.01[0.25;4.08]	0.984
Intrapartum prophylaxis				1.000
No	11 (32.4%)	6 (33.3%)	5 (31.2%)		Ref.
Yes	23 (67.6%)	12 (66.7%)	11 (68.8%)		0.91 [0.20;4.03]	0.906
Prophylaxis type				0.104
No	7 (15.9%)	4 (21.1%)	3 (12.0%)
Monotherapy	5 (11.4%)	0 (0.00%)	5 (20.0%)
Triple therapy	32 (72.7%)	15 (78.9%)	17 (68.0%)
Age at first ART (days)				<0.001
Median [IQR]	15.5 [0.00;24.2]	0.00 [0.00;2.50]	23.0 [19.0;27.0]
ART regimen at initiation				0.008
2NRTI+1 NVP	30 (69.8%)	17 (94.4%)	13 (52.0%)		Ref.
2NRTI+1 PI	13 (30.2%)	1 (5.56%)	12 (48.0%)		0.08[0.00;0.46]	0.003
ART at initiation				0.023
3TC+ABV+NVP	3 (6.98%)	1 (5.56%)	2 (8.00%)
3TC+AZT+NVP	27 (62.8%)	16 (88.9%)	11 (44.0%)
3TC+AZT+LPVr	7 (16.3%)	0 (0.00%)	7 (28.0%)
Other	7 (15.9%)	2 (10.5%)	5 (20.0%)
Baseline CD4 count				0.986
Median [IQR]	2766 [2126;3368]	2810 [2185;3245]	2540 [2106;4179]		0.98 [0.71;1.36]	0.933
Baseline CD8 count				0.589
Median [IQR]	1102 [837;1491]	1050 [854;1265]	1131 [844;1792]		1.00[1.00;1.00]	0.501
Baseline CD4/CD8				0.423
Median [IQR]	2.5 [1.6;3.1]	2.6 [1.7;3.0]	2.3 [1.5;3.1]		1.17 [0.84;1.62]	0.323
Baseline Viral load				0.020
Median [IQR]	17000[962;174882]	26938[14785;259276]	1970[782;26840]		9.2 [1.5;56.7]	0.018
Log(10) Median [IQR]	4.2 [2.9;5.2]	4.4 [4.2;5.4]	3.3 [2.9;4.4]

¹p-value: p-value for the distribution test (either continuous U-Mann Whitney or categorical chi-squared);

²p-value: p-value for the odds ratio (Wald-test).

At ART initiation, the median CD4 count was 2766 [IQR 2126;3368] cells/mm³, CD4 percentage was 46% [IQR 42.2; 52.5], median CD8 count was 1102 [IQR 837;1491] cells/mm³, and the median CD4/CD8 ratio was 2.5 [IQR 1.6;3.1]. No differences were found between groups (Table 1). Median log(10) baseline VL was 4.2 [IQR 2.9;5.2]. The group initiating ART within 7 days had a higher baseline VL than the group initiating ART after 7 days (4.4 vs. 3.3, p=0.020).

Overall, 37/44 (84%) infants suppressed during the follow-up period at a median time of 5.6 months after receiving ART. No differences in the proportions were found between participants treated within 7 days and after 7 days (84% vs. 84%, p= 1.00). Among infants who suppressed, 10/37 (22%) suppressed in the first 3 months after ART initiation, 20/37 (45%) in the first 6 months, and 30/37 (68%) in the first year. Infants spent almost 68% [IQR 36.2;91.5] of the follow-up time suppressed. One-third of infants (12/37) with prior viral suppression experienced subsequent virological failure during the follow-up, 1/37 (3%) patient had a suboptimal viral response, 7/37 (19%) presented any blip, and 8/37 (22%) had a spike. There were no significant differences between groups in virological failure, suboptimal response, blip or spikes (Table 2).

Table 2.

Virologic features of participants during follow-up.

	Overall N=44	ART within 7 days N=19	ART after 7 days N=25	p-value1	Odd Ratio [95% CI]
Follow-up time (years)				0.749
Median [Min, Max]	7.37 [1.91;11.1]	8.31 [2.61;10.8]	6.55 [1.45;11.3]		1.01[0.90;1.13]
Have ever been suppressed ≤50 cp/mL				1.00
No	7 (15.9%)	3 (15.8%)	4 (16.0%)		Ref.
Yes	37 (84.1%)	16 (84.2%)	21 (84%)		1.01 [0.18;6.1]
Time to reach suppression Months				0.053
Median [Min, Max]	5.6 [2.43;11.0]	3.75 [1.6;6.4]	7.6 [3.6;10.8]		0.91 [0.8;0.99]
% time suppressed				0.506
Median [Min, Max]	67.8 [36.2;91.5]	82.7 [46.3;91.3]	59.4 [31.6;91.9]		1.01[0.99;1.02]
Any suboptimal viral response ≥2 consecutive VL≥ 50-400				1.00
No	36 (97.3%)	16 (100%)	20 (95.2%)
Yes	1 (2.70%)	0 (0.00%)	1 (4.76%)
Number of suboptimal viral responses				0.553
Mean (SD)	0.73 (1.49)	0.50 (1.07)	0.86 (1.70)		0.82[0.41;1.66]
Blips Single VL≥ 50-400				1.00
No	30 (81.1%)	13 (81.2%)	17 (81.0%)		Ref.
Yes	7 (18.9%)	3 (18.8%)	4 (19.0%)		0.99[0.16;5.57]
Number of blips / years of follow up)				0.650
Mean (SD)	0.22 (0.2)	0.18 (0.05)	0.26 (0.3)
Spikes Single VL≥ 400				0.254
No	29 (78.4%)	11 (68.8%)	18 (85.7%)		Ref.
Yes	8 (21.6%)	5 (31.2%)	3 (14.3%)		2.61[0.51;15.9]
Number of spikes /years of follow up)				0.476
Mean (SD)	0.1 (0.02)	0.1 (0.03)	0.1 (0.02)
Virological failure				1.000
No	23 (65.7%)	11 (68.8%)	12 (63.2%)		Ref.
Yes	12 (34.3%)	5 (31.2%)	7 (36.8%)		0.79[0.18;3.31]
Time to first virological failure				0.323
Median [Min, Max]	2.40 [1.01;9.61]	2.81 [1.65;8.66]	1.32 [0.62;9.07]		1.01[0.89;1.16]
Number of virological failures /years of follow up)				0.905
Mean (SD)	0.14 (0.1)	0.14 (0.1)	0.14 (0.1)		1.0 [1.0;1.0]

Time to event analysis

Using the flexible parametric spline model, the cumulative probability of achieving virological suppression by 6 and 12 months after ART initiation was estimated at 49% [35, 62.2] and 77% [60.7, 88] respectively. Infants treated within 7 days were 4-fold more likely to suppress earlier than those treated after 7 days (HR 4.01 [1.7;9.5]). (Figure 1, panel A). Most infants of both groups reached suppression after one year of follow-up (Figure 1, panel B).

Figure 1. Time to virological events analysis according to ART timing groups.

Panel A) Probability of suppression according to the time-to-suppression in years in infants treated within 7 days (red) and those treated 8-28 days (blue). Smoothed lines were predicted according to a semiparametric flexible spline model. Hazard Ratio was adjusted by baseline VL and ART regimen at initiation. Panel B) Percentage of suppressed infants according to the time of follow-up in both ART timing groups. P-values were calculated by Fisher exact tests. Panel C) Probability of virological failure according to time-to-first-virological failure, in years, in infants treated within 7 days (red) and those treated from 8-28 days (blue). Smoothed lines were predicted according to a semiparametric flexible spline model adjusted by baseline VL and ART regimen at initiation. Panel D) Probability of blip according to time-to-first-blip in years in infants treated within 7 days (red) and those treated 8-28 days (blue). Smoothed lines were predicted according to a semiparametric flexible spline model adjusted by baseline VL and ART regimen at initiation.

When age at ART initiation was introduced in the model as a continuous variable, age at ART had a HR of 0.65 [0.46;0.92] after adjustment for pre-ART VL and ART regimen. In other words, for each week the ART initiation was delayed, the probability of suppression decreased by 35%.

No significant association was found between the probability of first virological failure and the age at ART groups (HR: 3.0 [0.6;13.8]) (Figure 1, Panel C), even when introducing age at ART initiation as a continuous variable (HR: 0.87 [0.57; 1.32]). Likewise, no significant association was found between the probability of first blip and the age at ART groups (HR: 0.98 [0.2;4.4]) (Figure 1, Panel D), or as a continuous variable (HR: 1.03 [0.62; 1.72]). Since few numbers of events were observed for a suboptimal response (n=1) or spike (n=8) in this population, there was an insufficient sample size to explore an association.

Association between time to suppression and age at ART initiation in infants who suppressed

The type of association between time to suppression and age at ART initiation was tested in those infants who suppressed (n=37) using a generalized additive mixed model. Age at ART initiation and baseline VL as a confounder were included as fixed effects and ART regimen as random intercepts. VL was smoothed by splines with a significant smooth term (p=0.035) with 5 knots according to the smoothness selection.

Evaluating associations between age at ART initiation in weeks and time to viral suppression a normal linear relationship was noted (β = 0.2 (SD 0.1); p=0.048) (Figure 2). Whereas a non-normally linear association was detected in the association between log VL at baseline and time to viral load suppression. Specifically, for infants with a baseline viral load below 5 log(10), no association was noted with time to viral suppression. However, when baseline VL was at or above 5 log(10), a strong effect with an exponential slope was observed.

Figure 2.

Prediction of the association between time to suppression and age at ART initiation or baseline viral load in infants who suppressed during the follow-up (n=37). Association prediction between age at ART initiation (Panel A) and baseline viral load (Panel B) and time to suppression in a total of 37 infants. Generalized additive mixed model with age at ART initiation and baseline viral load as fixed effects and ART regimen at initiation as random effects.

DISCUSSION

In newborns who started ART within the first 28 days of life, those who started ART within 7 days of life had four times higher chance of viral suppression than those who started aged 8 to 28 days. The probability of achieving suppression decreased by 35% for each week ART initiation was delayed in the neonatal period, after adjustment for baseline viral load and ART regimens at initiation. These significant results were mainly driven by the very first months of follow-up. Eventually, there was a similar proportion of suppressed infants after 1 year of follow-up regardless of when ART was started. This implies that there are other factors driving time to virologic suppression beyond the age at ART initiation.

Other virological endpoints examined, including ever achieving viral suppression and markers of lack of sustained viral control after suppression did not differ between the two age at ART start groups. Current clinical practice generally favours rapid initiation of ART, sometimes based on a presumptive diagnosis, and our results support this approach. We hypothesize that a more rapid decline in VL may have a beneficial effect on the size of the viral reservoir. Age at which suppression is attained, which is linked but independent of age at ART initiation, has been shown to be an important predictor of the size of the viral reservoir ²⁹. The other virological endpoints we examined may be less sensitive markers of these important early viral dynamics.

The virological endpoints we observed in this multi-site cohort of perinatally-infected neonates in Europe and Thailand are different than has been reported in other cohorts of very early-treated infants. In a Canadian cohort, four of 12 neonates treated within 72 hours of birth achieved sustained virological control ³⁰. In a South African cohort, only half of the 22 HIV-infected neonates treated remained in care and only two of these had a VL <50 copies when last seen ³¹. No deaths were observed in our cohort which is encouraging. However, we selected only infants with at least two VLs measurements which are likely to be biased towards those surviving and continuing to engage in care for longer.

The association between age at ART initiation, baseline CD4 and baseline VL with time-to-suppression has been previously described ^32–38. However, most of the studies have assumed that these dependent variables followed a normal linear distribution by applying conventional linear or logistic regressions to test this relationship which could provide some loss of interpretability. The use of smoothing functions fitting relaxed assumptions allowed us to identify two different associations in this population. First, we identified a positive linear association between time to suppression and age at ART initiation. Second, the analysis showed a novel finding in the relationship between time to suppression and baseline VL. The dynamics of this relationship suggests that below 5 log(10), baseline VL has no influence in time to suppression. In these infants, time to treatment is the only factor we found influencing time to suppression. However, when baseline VL is over 5 log(10), an exponential linear relationship is observed, suggesting a strong influence in time to suppression. In these infants, not only the age at ART initiation is making a difference in time to suppression, but also the VL. In the study did by Frigati et al.³¹, baseline VL of the 22 participants was 4.444 copies/mL. In our study, VL was higher, especially due to the 19 patients who initiated earlier. The range was also wider, as well as in a South African study³⁹, reaching >6 log(10). The reason for these differences is unclear and may be related with the sample size. In any case, very high VLs seem to have an influence in time to suppression and should be considered in models.One may wonder if these newborns with high VL and longer time to suppression are those infected in utero or might have a more immature immune function. Further studies are warranted to evaluate this possibility.

Our study has several limitations. Data are retrieved from an observational cohort and the age at ART initiation is dependent on a range of clinical and social factors beyond our control. These include, for example, when the HIV-infected mother came to clinical attention, what infant diagnosis programs and treatment recommendations were in place at the time, how the mother engaged with the health service and adhered to medical advice, etc. Due to the nature of the databases, we also have some missing data and very limited information on these factors including markers of the mother’s health and the timing of infant diagnosis. The assessment of suppression is based on having two VL measurement available raising concern as to whether this frequency of measurement can truly reflect the extent of virologic suppression. We cannot rule out that the faster time to suppression observed in the earlier treated infants is due to confounding factors and not to the timing of ART. However, on the parameters on which we have data we did not observe significant differences between neonates starting ART within the first 7 days of life and those starting later, except for baseline ART viral load. Interestingly, baseline ART VL was higher in those starting ART earlier than in those starting slightly later which may reflect the dynamics of primary infection ⁴⁰. This potential bias was addressed adjusting by baseline VL and ART regimens at initiation.

In conclusion, our data support the potential benefits of starting ART within 7 days of birth even in comparison to starting from 8 to 28 days after birth. We observed a faster time to viral suppression which may result in a favourable impact on the viral reservoir. Our results lend support to the clinical guidance to implement universal birth testing to diagnose intrauterine-infected neonates as early as possible as well as to clinical guidance that encourages rapid initiation of ART either presumptively or at diagnosis.