Early antiretroviral therapy reduces severity but does not eliminate neurodevelopmental compromise in children with HIV

Abstract

Background:

Early antiretroviral therapy (ART) during infancy reduces cognitive impairment due to HIV, but the extent of benefit is unclear.

Setting:

Children were recruited from hospital and health centers providing HIV care and treatment in Nairobi, Kenya.

Methods:

Cognitive, behavioral and motor outcomes were assessed in children with HIV (CWHIV) and early-ART (<1 year), late-ART (1.5–6 years) and children HIV-unexposed uninfected (CHUU). Domain z-scores and odds neurobehavioral impairment (≤15th percentile in CHUU) were compared in adjusted analyses.

Results:

CWHIV initiated ART at median age 0.4 (early-ART) and 3.5 years (late-ART). Children were assessed at median age 6.9 (CHUU; N=61), 6.9 (early-ART; N=54) and 13.5 (late-ART; N=27) years. Children with late- vs. early-ART had significantly lower z-scores in 7 domains, specifically, global cognition, short-term memory, visuospatial processing, learning, nonverbal test performance, executive function and motor (adjusted mean differences [AMD], −0.42 to −0.62, P-values ≤0.05), and had higher odds impairment in 7 domains (adjusted odds ratios [aORs]. 2.87 to 16.22, P-values ≤0.05). Children with early-ART vs. CHUU had lower z-scores in 5 domains (global cognition, short-term memory, delayed memory, processing speed, behavioral regulation [AMD, −0.32 to −0.88, P-values <0.05]), and higher impairment for 2 (short-term memory [aOR, 3.88] and behavioral regulation [aOR 3.46], P-values <0.05). Children with late-ART vs. CHUU had lower z-scores in 8 domains (AMD, −0.57 to −1.05, P-values ≤0.05), and higher impairment in 7 domains (aORs 1.98 to 2.32, P-values ≤0.05).

Conclusion:

Early ART in the first year of life attenuates but does not eliminate neurodevelopmental compromise of HIV.

INTRODUCTION

In 2021, an estimated 1.7 million children were living with HIV, most in sub-Saharan Africa (SSA) [1]. Pediatric HIV infection can result in a range of neurodevelopmental impairments, including reduced brain volume, brain lesions, lost or delayed developmental milestones, motor impairment and poor cognition [2–4]. Historically, children on long-term antiretroviral treatment (ART) have had below norm average cognitive scores [5–7] and high prevalence of learning disorders (25–42%) [7, 8]. These estimates are difficult to contexualize for current populations of CWHIV because they are based on cohorts with heterogeneous treatment histories, who were often treated based on older guidelines that initiated ART later in childhood [9]. In US studies, CWHIV with earlier vs. later viral suppression or with never vs. ever having had an AIDS defining illness prior to ART had improved cognitive outcomes [5, 8, 10, 11], suggesting longer duration of unsuppressed virus or symptomatic disease prior to ART in early life contributes to worse neurodevelopmental outcomes in CWHIV.

The extent of benefit of ART provided in infancy for cognitive outcomes is based on limited studies. In a large multi-site trial, children who initiated ART at median age 1.2 years had lower scores in several domains vs. children HIV-unexposed uninfected (CHUU), including global cognitive ability, short-term memory, visuospatial processing, nonverbal test performance, learning, and planning, delayed recall, attention, and processing speed [12]. In contrast, in the Children with HIV Early Antiretroviral (CHER) Study, CWHIV had differences in visual perception (executive functioning) but otherwise had similar cognitive scores in several domains vs. CHUU at 7 and 9 years of age [13]. Children in CHER had either initiated ART while asymptomatic (mean age 7 weeks) or had ART deferred until meeting stringent immune and clinical criteria (mean age at ART, 6.8 months) [14, 15]. In a third study with most children (79%) initiating ART at <6 months and with mean CD4 percentage of 24%, CWHIV had significantly lower socres for several domains vs. CHUU, including global cognitive ability, short-term memory, visuospatial processing, nonverbal test performance, learning, and planning [16]. Interpretation of this latter study is limited by higher maternal education level in the CHUU group.

Antiretroviral coverage in CWHIV has lagged adults, and remains poor [17]. It is important to define the extent of benefit of early ART and detriment of delayed ART. Here, we compared long-term neurocognitive outcomes in CWHIV in Kenya with early- vs. late-ART. We hypothesized CWHIV and early-ART would have better neurocognitive outcomes compared with children with late-ART. Secondarily, both groups were compared with CHUU. We hypothesized CWHIV and late-ART would have worse neurocognitive outcomes vs CHUU and those with early-ART would have similar or subtle neurocognitive differences vs CHUU.

METHODS

Study population

This was a prospective study comparing neurocognitive outcomes in children with early- (ART at age <1 year) vs. late-ART (ART at age 1.5–6 years). Both cohorts had similar catchment and detailed pre-ART and long-term follow-up data. Children were enrolled in Nairobi, Kenya. Written informed consent for study procedures was obtained from participants and study procedures were approved by the University of Washington Human Subjects Division and the University of Nairobi/Kenyatta National Hospital Ethics and Research Committee.

Early antiretroviral therapy

CWHIV had taken part in the Optimizing Pediatric HIV-1 Therapy 03 (OPH03) randomized clinical trial (RCT) (NCT00428116) of continued vs. interrupted ART [18]. Initial age inclusion criteria were 0–4 months of age and were subsequently expanded to <13 months. Infants were identified (2007–2009) at maternal child health (MCH) clinics or hospital wards and either initiated ART at or prior to entry. Infants initiated nevirapine (NVP) or ritonavir-boosted lopinavir (LPV/r) (if NVP exposed following prophylaxis for prevention of mother-to-child-transmission) plus zidovudine (AZT) or abacavir (ABC) and lamivudine (3TC) [18]. Infants were followed monthly for up to 42 months, then 3-monthly until assessment.

Late antiretroviral therapy

CWHIV were recruited (2004–2006) from the paediatric wards and HIV treatment clinic of Kenyatta National Hospital for an RCT of medication diaries for treatment adherence [19] and extended follow- [20]. Children initiated ART (either NVP- or efavirenz [EFV] plus AZT or stavudine [d4T] plus 3TC, and were followed 3-monthly for up to 7 years. Children were exited in 2012, and in 2015, caregivers were re-contacted and recruited for follow-up with annual assessment. At re-entry, all children initiated ART between ages 1.5 to 6 years and were age 5–17 years. Methods for CD4+ T cell count and percentage and plasma HIV RNA have been described for both CWHIV cohorts ^[21].

Children HIV-unexposed uninfected

CHUU were identified at the Mathare North Health Centre in Nairobi (2013–2016). This site was selected because its clientele was anticipated to have similar demographics to the CWHIV cohorts. Key inclusion criteria for CHUU were age 5–12 years and confirmed HIV seronegative status for child and biological mother.

Neuropsychological and behavioral assessments

Assessments were performed by local staff with either bachelor’s or graduate level training in psychology. Testers did not begin to test for the study until they had successfully administered the full battery to a child not participating in the study, with evaluation via video and a standard rubric. Testers participated in routine internal video reviews as a group, individual reviews by a neuropsychologist (PB) using video, and periodic site visits by a neuropsychologist for in person teaching and quality control. All children underwent the same neuropsychological test (NPT) and behavioral assessment battery, including the Kaufman Assessment Battery for Children 2^nd edition (KABC) [22], the Test of Variables of Attention (TOVA) [23], the Behavior Rating Inventory of Executive Functioning (BRIEF) [24], and Bruininck’s-Oseretsky Test of Motor Proficiency 2^nd edition Brief Form (BOTMP) [25]. These assessments have been used extensively in SSA [12].

Analysis

Scores were standardized using US norms. To accommodate assessments with differing norm scale means and standard deviations, scores are presented as z-scores. Scores were reversed for the BRIEF because higher BRIEF standard scores reflect worse performance. Neurobehavioral impairment was defined as below the 15^th percentile for CHUU by domain [26]. Using data from KABC, z-scores are presented for global cognitive ability (Mental Processing Index), short-term memory (Sequential Scale), visuospatial processing (Simultaneous Scale), learning (Learning Scale), planning (Planning Scale), delayed memory (Delayed Recall), and nonverbal test performance (Nonverbal Index). The KABC Mental Processing Index comprises the Sequential, Simultaneous, Learning and Planning Scales. The Nonverbal Index comprises subtests with minimized involvement of verbal ability. Only children ages 7 and above receive a score for Planning. Using data from TOVA, scores for attention (ratio of the hit rate [correct responses to a target stimulus] to the false alarm rate [d’prime]) and processing speed (average time for correct response to a target [Response Time] are presented. For the BRIEF scores for executive function (Global Executive Composite), behavioral regulation (Behavior Regulation Index) and metacognition (Metacognition Index) are presented. The Global Executive Composite is comprised of the Behavior Regulation and Metacognition Indices. For the BOT, a single z-score (Total Point Score) is presented.

Descriptive statistics are provided for demographic, social, health, growth and immune parameters, and z-scores for NPTs and behavioral assessments for each cohort. Growth parameters (z-scores for weight-for-age (WAZ), height-for-age (HAZ) and body mass index (BMI) were calculated using WHO population norms for children aged 5–19 years [27]. Children were considered an orphan or vulnerable child (OVC) if one parent was deceased or alive and not fulfilling parental duties. Household hunger was evaluated using a validated 3 question scale [28]. Differences in demographics and indicators of socioeconomic status (caregiver age, marital status, and education, household monthly rent, one-room housing, household hunger, and child OVC status) were evaluated between groups. Chi-square or Fisher’s exact (in case of sparse cell counts) tests (for dichotomous variables) or rank sum tests (for continuous variables) were used to compare characteristics and proportion with neurobehavioral impairment between groups. Similar to prior studies, analyses were restricted to CHUU with no self-reported history of neurologic conditions (seizures and developmental delay) [12, 13]. Post-hoc analyses examined differences between cohorts with these individuals included. Univariable and multivariable linear regression analyses were used to compare z-scores for each NPT and behavioral assessment domain between children with early-vs. late-ART, and between both early- and late-ART groups vs CHUU. Multivariable logistic regression was used to compare odds of neurobehavioral impairment in CWHIV with early- vs. late-ART and between both CWHIV groups vs. CHUU. Caregiver age, marital status, and education, household monthly rent, and one-room housing were considered as potential confounders. Child sex, child OVC status and household hunger were considered as potential mediators. Univariable analyses examined these cofactors for domain z-scores within each cohort (P <0.1). Potential cofactors were evaluated for co-linearity first by examining pairwise correlation (P <0.1), and then in regression models with and without each cofactor for differences of >10% in standard errors. In cases of co-linearity, the more plausible cofactor with the least missingness was retained. The Benjamini-Hochberg procedure with a false discovery rate of 0.25 [29, 30] was used to account for multiple comparisons.

RESULTS

Among 62 children with early-ART and 27 children with late-ART, 54 and 27 respectively underwent NPT and behavioral assessment. Seventy CHUU were enrolled and 67 underwent testing. CHUU with history of developmental delay (n=4), seizures (n=1) and both seizures and developmental delay (n=1) were excluded, with 61 CHUU remaining in analysis (Table 1). At ART initiation, children with late-ART were more immunosuppressed (median CD4%, 8%) vs. early-ART (median CD4%, 19%, P=0.0001) and had more severe growth compromise (median LAZ/HAZ, −2.35 vs. −0.99; P=0.0001, See Table, Supplemental Digital Content 1). No children with late-ART and 25 (46.3%) children with early-ART initially received a protease inhibitor-based regimen (P <0.0001).

Table 1.

Child, caregiver and household characteristics at assessment for each cohort.

Child characteristics	CHUUa (N=61)	Early-ARTb (N=54)	Late-ARTc (N=27)	Early-ART vs. CHUU P	Late-ART vs. CHUU P	Late- vs. Early-ART P
Male	24 (39.3)	29 (53.7)	15 (55.6)
Age (years)	6.9 (6.1, 8.3)	6.9 (6.5, 7.6)	13.5 (12.5, 14.8)		<0.0001	<0.0001
Years on ART	-	6.5 (6.1, 7.0)	10.1 (9.0, 10.7)	-	-	<0.0001
WAZ	−0.73 (−1.33, −0.10)	−0.78 (−1.35, −0.28)	-		-	-
HAZ	−0.03 (−0.79, 0.50)	−0.74 (−1.39, 0.27)	−1.26 (−1.71, −0.84)	0.009	<0.0001	0.006
BMI z-score	−1.22 (−1.84, −0.69)	−0.50 (−0.97, −0.07)	−1.01 (−1.75, −0.2)	<0.0001
Plasma HIV RNA log10 c/ml	-	1.88 (1.88, 3.02)	1.88 (1.88, 1.88)	-	-
Plasma HIV RNA suppressedd	-	31 (57.4)	22 (81.5)	-	-	0.03
CD4%	41 (35, 44)	36 (30, 41)	34 (29, 43)	0.005	0.03
CD4 count	1224 (1,065, 1,664)	1,148 (799, 1,431)	908 (689, 1,116)	0.02	<0.0001	0.03
Prior seizures	0	3 (5.6)	2 (7.4)
Prior meningitis	0	3 (5.6)	1 (3.7)
Caregiver and household characteristics
Age (years)	29 (26, 34)	32 (28, 39)	40 (36, 44)	0.01	<0.0001	0.0008
Married	46 (75.4)	30 (55.6)	14 (51.9)	0.03	0.03
Education (years)	8 (8, 12)	9.5 (8, 12)	8 (8, 12)
Monthly rent (KES)	2,000 (1,500, 3,000)	3,000 (2,000, 6,000)	4,500 (3,000, 7,000)	0.0009	<0.0001
One room house	53 (86.9)	29 (54.7)	7 (25.9)	<0.001	<0.001	0.01
Moderate or severe household hunger	18 (29.5)	4 (7.6)	2 (7.4)	0.004	0.03
OVC	15 (24.6)	23 (42.6)	15 (57.7)	0.04	0.003

Data presented as number (percentage) or median (25^th and 75^th percentiles). ART, antiretroviral therapy, CHUU, children HIV-unexposed uninfected; WAZ, weight-for-age z-score; HAZ, height-for-age z-score; BMI, body mass index, c, copies; KES, Kenya Shilling; OVC, orphan or vulnerable child. P-values >0.05 are not shown.

^aCHUU, WAZ, N=60; Caregiver education, N=60; Monthly rent, N=58.

^bEarly-ART, Caregiver education, N=52; Monthly rent, N=47; One-room house, N=53; Household hunger, N=53.

^cLate-ART, Caregiver education, N=26; Monthly rent, N=23; OVC, N=26.

^dDefined as 150 copies/ml (limit of detection).

At assessment, the median age for children with early-ART and for CHUU were similar (medians, 6.9 years) (Table 1). Children with late-ART were older (median 13.5 years, P <0.0001). WAZ scores were similar for children with early-ART (median, −0.73) and CHUU (median, −0.78). The early-ART cohort had lower HAZ (median, −0.74) than CHUU (median, −0.03, P=0.009), as did the late-ART cohort (median −1.26, P <0.0001 vs. CHUU). The late-ART cohort had lower HAZ than the early-ART cohort (P=0.006). Fewer children with early-ART (57.4%) had suppressed plasma virus levels vs. late-ART (81.5%) (P=0.03).

Most children with early-ART and CHUU were cared for by their biological mother (98.4% and 92.5% respectively), whereas only 63.0% of children with late-ART were (vs. CHUU, P <0.001; and vs. early-ART, P=0.002). Both early-ART (42.6%) and late-ART (57.7%) groups were more likely to be OVCs vs. CHUU (24.6%, P=0.04 and P=0.003, respectively). Most CHUU (86.9%) lived in a one-room house vs. 54.7% for early-ART (P <0.001) and 25.9% for late-ART (P <0.001, Table 1). A substantial proportion of caregivers of CHUU (29.5%) reported moderate (n=17) or severe (n=1) household hunger, which was less common for both children with early-ART (7.6%, P=0.004) and late-ART (7.4%, P=0.03).

Domain z-scores are summarized in Table 2 for CHUU and each group of CWHIV. In univariate analyses comparing children with late- vs. early-ART, children with late-ART had significantly lower scores for 5 domains, specifically, global cognitive ability (mean z-score difference [beta], −0.35 (95% confidence intervals [CI], −0.69, −0.00; P=0.05), short-term memory (−0.43, 95% CI, −0.79, −0.07; P=0.02), nonverbal test performance (−0.43, 95% CI, −0.78, −0.07; P=0.02), executive function (−0.54, 95% CI, −1.04, −0.05; P=0.03) and motor skills (−0.62, 95% CI, −0.98, −0.25; P=0.001; See Table, Supplemental Digital Content 2). In multivariable analyses adjusted for one-room housing, associations remained and children with late-ART also had significantly lower scores for visuospatial processing (−0.42, 95% CI, −0.81, −0.03, P=0.04) and learning (−0.54, 95% CI, −1.00, −0.08, P=0.02; Table 3) (7 domains total).

Table 2.

Summary of neuropsychological and behavioral assessment z-scores for each cohort.

Domain	CHUU (N=61)a	Early-ART (N=54)b	Late-ART(N=27)c
Global cognitive ability	−1.67 (−2.13, −1.13)	−1.87 (−2.33, −1.4)	−2.37 (−3.00, −1.27)
Short-term memory	−1.13 (−1.53, −0.60)	−1.33, (−1.93, −1.00)	−1.83 (−2.47, −1.33)
Visuospatial processing	−1.87 (−2.33, −1.13)	−1.87 (−2.33, −1.53)	−2.40 (−3.20, −1.53)
Learning	−0.93 (−1.47, −0.20)	−0.93 (−1.47, −0.20)	−1.00 (−2.20, −0.40)
Planning	−1.97 (−2.40, −1.33)	−2.07 (−2.40, −1.87)	−2.47 (−3.07, −1.33)
Nonverbal test performance	−1.73 (−2.33, −1.40)	−2.07 (−2.40, −1.73)	−2.37 (−3.00, −1.60)
Delayed memory	−0.80 (−1.53, −0.40)	−0.80 (−1.53, −0.50)	−0.80 (−1.33, −0.40)
Attention	−1.00 (−1.60, −0.60)	−1.40 (−1.73, −0.87)	−1.67 (−2.27, −1.07)
Processing speed	−0.67 (−2.00, −0.13)	−1.47 (−2.47, −0.87)	−1.90 (−3.20, −0.27)
Executive function	0.10 (−0.60, 0.60)	0.10 (−0.90, 0.50)	−0.35 (−1.60, 0.20)
Behavioral regulation	−0.20 (−0.70, 0.40)	−0.55 (−1.15, 0.40)	−0.50 (−1.60, 0.50)
Metacognition	0.40 (−0.30, 0.80)	0.40 (−1.00, 0.85)	−0.45 (−1.30, 0.30)
Motor	−0.50 (−0.90, −0.1)	−0.80 (−1.10, −0.30)	−1.30 (−2.00, −0.70)

Data are presented as median (25^th and 75^th percentiles). Z-scores presented for executive function, behavioral regulation and metacognition were calculated by reversing the standard scores. CHUU, children HIV-unexposed uninfected; CWHIV, children with HIV; ART, antiretroviral therapy.

^aCHUU, Global cognitive ability, Short-term memory, Visuospatial processing, Learning, Nonverbal, N=59; Planning, N=28; Delayed memory, N=50; Attention, Processing speed, N=53; Motor, N=60.

^bEarly-ART, Planning N=26; Nonverbal, N=53; Delayed memory N=44; Attention, Processing speed N=45; Executive function, N=51; Behavioral regulation, Metacognition, N=52; Motor, N=53.

^cLate-ART, N=27 children were assessed; N=26 except for Delayed memory, N=22; Attention, N=25; Motor, N=25.

Table 3.

Multivariable analyses of neuropsychological and behavioral assessment z-scores for comparison of each cohort.

Domain	Adjusted beta (95% CI) Early-ART vs. CHUUa	P	Adjusted beta (95% CI) Late-ART vs. CHUUb	P	Adjusted beta (95%CI) Late- vs. early-ARTc	P
Global cognitive ability	−0.32 (−0.59, −0.04)	0.02	−0.60 (−1.06, −0.14)	0.01	−0.46 (−0.81, −0.12)	0.009
Short-term memory	−0.41 (−0.71, −0.11)	0.009	−0.73 (−1.16, −0.29)	0.001	−0.51 (−0.88 –0.14)	0.008
Visuospatial processing	−0.24 (−0.56, 0.08)	0.1	−0.57 (−1.13, −0.01)	0.05	−0.42 (−0.81, −0.03)	0.04
Learning	−0.09 (−0.49, 0.31)	0.7	−0.22 (−0.86, 0.43)	0.5	−0.54 (−1.00, −0.08)	0.02
Planning	−0.29 (−0.68, 0.09)	0.1	−0.39 (−0.99, 0.20)	0.2	−0.18 (−0.60, 0.23)	0.4
Nonverbal test performance	−0.26 (−0.59, 0.06)	0.1	−0.59 (−1.09, −0.08)	0.02	−0.50 (−0.86, −0.13)	0.009
Delayed memory	−0.36 (−0.69, −0.04)	0.03	−0.33 (−0.85, 0.20)	0.2	−0.30 (−0.75, 0.15)	0.2
Attention	−0.29 (−0.67, 0.09)	0.1	−0.46 (−1.00, 0.09)	0.1	−0.16 (−0.56, 0.25)	0.4
Processing speed	−0.88 (−1.47, −0.29)	0.004	−1.05 (−2.06, −0.03)	0.04	−0.21 (−0.92, 0.50)	0.6
Executive function	−0.29 (−0.68, 0.09)	0.1	−0.61 (−1.19, −0.03)	0.04	−0.51 (−1.03, 0.01)	0.05
Behavioral regulation	−0.39 (−0.77, −0.02)	0.04	−0.47 (−1.05, 0.11)	0.1	−0.19 (−0.75, 0.38)	0.5
Metacognition	−0.29 (−0.72, 0.13)	0.2	−0.65 (−1.25, −0.06)	0.03	−0.50 (−1.10, 0.07)	0.09
Motor	−0.18 (−0.46, 0.10)	0.2	−0.69 (−1.13, −0.26)	0.002	−0.62 (−1.00, −0.24)	0.002
Number domains P ≤0.05		5 (38.4)		8 (61.5)		7 (53.8)

Multivariable analyses were adjusted for living in a one room house and household hunger for analyes of each cohort of CWHIV vs. CHUU. Multivariable analyses were adjusted for living in a one room house only for analyses of children with late- vs. early-ART. Z-scores presented for executive function, behavioral regulation and metacognition were calculated by reversing the standard scores. CI, confidence interval. CHUU, children HIV-unexposed uninfected; CWHIV, children with HIV; ART, antiretroviral therapy. Bold text denotes P ≤0.05.

^aN=111 except Planning, N=54; Delayed memory, N=94; Nonverbal, Executive function, N=110 except Attention, Processing speed, N=97.

^bN=85 except Planning, N=54; Delayed memory, N=72; Attention, N=78; Processing speed, N=79; Executive function, Behavioral regulation, and Metacognition, N=87.

^cN=79 except Planning, N=52; Nonverbal, N=78, Delayed memory, N=66; Attention, N=70, Processing speed, N=71; Executive function, N=76; Behavior regulation, Metacognition, Motor, N=77.

Children with late- vs. early-ART had higher proportion with impairment for 7 domains, specifically global cognitive ability (50.0% vs. 22.2%; P=0.01), visuospatial processing (42.3% vs. 9.3%; P=0.001), learning (34.6% vs. 7.4%; P=0.004), planning (34.6% vs. 3.9%; P=0.01), nonverbal test performance (38.5% vs. 5.7%; P=0.001), executive function (46.2% vs. 23.5%; P=0.04), and motor (44.0% vs. 15.1%; P=0.006; See Table, Supplemental Digital Content 3). In multivariable analyses adjusted for one-room housing, results were largely similar and children with late vs. early ART additionally had significantly higher odds of impairment for short-term memory but not executive function (adjusted odds ratios (aORs) ranging from 2.87 to 16.22; all P-values ≤0.05) for 7 domains (Table 4). Here and below, separate models adjusted for additional potential confounders gave similar results, as did models adjusted separately for child sex, OVC status and household hunger.

Table 4.

Multivariable analysis of neurobehavioral impairment comparing each cohort.

Domain	aOR (95% CI) Early-ART vs. CHUU (N=112)a	P	aOR (95% CI) Late-ART vs. CHUU (N=87)b	P	aOR (95% CI) Late- vs. Early-ART (N=79)c	P
Global cognitive ability	1.89 (0.70, 5.15)	0.2	2.12 (1.11, 4.06)	0.02	5.28 (1.65, 16.90)	0.005
Short-term memory	3.88 (1.45, 10.36)	0.007	2.22 (1.15, 4.28)	0.02	2.87 (0.98, 8.40)	0.05
Visuospatial processing	0.97 (0.28, 3.35)	1.0	2.16 (1.09, 4.29)	0.03	10.23 (2.58, 40.51)	0.001
Learning	1.00 (0.25, 3.99)	1.0	2.28 (1.08, 4.81)	0.03	10.07 (2.28, 44.37)	0.002
Planning	0.23 (0.02, 2.94)	0.3	1.78 (0.78, 4.11)	0.2	16.22 (1.74, 151.24)	0.01
Nonverbal test performance	0.57 (0.14, 2.32)	0.4	1.98 (1.0, 3.92)	0.05	11.38 (2.57, 50.38)	0.001
Delayed memory	1.72 (0.61, 4.90)	0.3	0.87 (0.40, 1.90)	0.7	1.15 (0.32, 4.16)	0.8
Attention	1.28 (0.40, 4.09)	0.7	1.79 (0.86, 3.76)	0.1	2.04 (0.63, 6.61)	0.2
Processing speed	1.62 (0.57, 4.55)	0.4	1.71 (0.85, 3.43)	0.1	1.82 (0.62, 5.35)	0.3
Executive function	1.79 (0.62, 5.18)	0.3	1.80 (0.94, 3.44)	0.08	2.41 (0.85, 6.82)	0.1
Behavioral regulation	3.46 (1.27, 9.40)	0.02	2.32 (1.19, 4.54)	0.01	1.50 (0.56, 4.04)	0.4
Metacognition	1.75 (0.64, 4.83)	0.3	1.44 (0.75, 2.77)	0.3	1.77 (0.64, 4.93)	0.3
Motor	1.12 (0.35, 3.54)	0.9	2.21 (1.11, 4.38)	0.02	5.12 (1.59, 16.51)	0.006
Number domains P ≤0.05		2 (15.4)		7 (53.8)		7 (53.8)

All multivariable models were adjusted for living in a one room house. aOR, adjusted odds ratio, CI, confidence interval; CHUU, children HIV-unexposed uninfected; CWHIV, children with HIV; ART, antiretroviral therapy. Bold text denotes P ≤0.05.

^aPlanning, N=54; Nonverbal, N=80; Delayed memory, N=94; Attention, Processing speed, N=98; Executive function, N=111.

^bGlobal cognitive ability, Short-term memory, Visuospatial processing, Learning, Nonverbal, Motor, N=85; Delayed memory, N=72; Attention, N=78; Processing speed, N=79; Planning, N=54.

^cPlanning, N=52, Nonverbal, N=78; Delayed memory, N=66; Attention, N=70, Processing speed, N=71; Executive function, N=76; Behavioral regulation, Metacognition, Motor, N=77.

Compared with CHUU, the early-ART group had significantly lower scores for 3 domains in univariable analyses, specifically, short-term memory (Fig. 1; beta, −0.34, 95% CI, −0.61, −0.07, P=0.01), processing speed (−0.79, 95% CI, −1.30, −0.27, P=0.003, and behavioral regulation (−0.37, 95% CI, −0.70, −0.04, P=0.03; See Table, Supplemental Digital Content 2). In multivariable analyses adjusted for both one-room housing and household hunger, these associations remained and children with early-ART additionally had lower scores for global cognitive ability (−0.32, 95% CI, −0.59, −0.04, P=0.02), and delayed memory (−0.36, 95% CI, −0.69, −0.04, P=0.03) (5 domains total) (Table 3). Children with early-ART also had significantly higher proportion with neurobehavioral impairment for short-term memory (33.3%) and behavioral regulation (34.6%,) vs. CHUU (P=0.02 and P=0.007, respectively; see Table, Supplemental Digital Content 3). Adjusted analyses gave similar results. Children with early-ART had higher odds impaired short-term memory (adjusted odds ratio [aOR] 3.88, 95% CI, 1.45, 10.36, P=0.007) and behavioral regulation (aOR, 3.46, 95% CI, 1.27, 9.40, P=0.02; Table 4). In sensitivity analyses including CHUU with self-reported history of developmental delay and siezures, results were similar.

Fig 1.

Comparison of neuropsychological and behavioral assessment z-scores for CWHIV and early- and late-ART vs. CHUU. Symbols indicate unadjusted z-score betas for children with early-ART (black) and late-ART (gray) vs. CHUU (referent). Lines indicate 95% CI.

CWHIV, children with HIV; ART, antiretroviral therapy; CHUU, children HIV-unexposed uninfected; CI, confidence intervals.

Compared with CHUU, children with late-ART had significantly lower scores for all domains assessed except learning, planning and delayed memory (Fig 1; 10 domains with significant differences, all P-values <0.05; See Table, Supplemental Digital Content 2). These associations remained for 8 domains in multivariable analysis, specifically global cognitive ability, short-term memory, visuospatial processing, nonverbal test performance, processing speed, executive function, metacognition, and motor (all P-values ≤0.05; Table 3). Children with late-ART had significantly higher proportion with neurobehavioral impairment for 9 domains (all P-values <0.05; see Table, Supplemental Digital Content 3). In adjusted analyses, the odds of neurobehavioral impairment were approximately 2-fold higher for 7 domains, specifically global cognitive ability, short-term memory, visuospatial processing, learning, nonverbal test performance, behavioral regulation and motor skills (all P-values ≤0.05; Table 4).

DISCUSSION

Here we describe long-term neurocognitive benefit of ART in the first year of life (early-ART) for CWHIV compared with children who received late-ART. Children with early-ART had less severe impairment than children with late-ART. In spite of this benefit, children with early-ART still had evidence of compromise and had lower z-scores than CHUU for global cognition, short-term memory delayed memory, processing speed and behavioral regulation. These score differences exceeded one-third of a standard deviation, suggesting clinically meaningful differences in functioning. Our findings differ from those in the CHER study, which noted limited neurocognitive compromise with early treatment. The CHER study included even earlier ART and children were either asymptomatic or carefully monitored prior to ART [13, 15]. Cognitive outcomes in the early-ART group in our study were more similar to those in a study from Cameroon, in which most children initiated ART before 6 months of age. Taken together, these data suggest a very short window of opportunity for optimizing neurocognitive outcomes, and underscore the importance of early diagnosis and prompt ART initiation during infancy, prior to onset of symptoms and CD4 decline. Given CWHIV in the early-ART group often initiated treatment after 5 months of age, our study does not directly address the protective benefit of very early ART; however our study emphasizes the importance of rapid ART in infancy and the need to support developmental outcomes in children who may have even a short duration of delayed ART. These data also highlight the importance of careful monitoring of neurocognitive outcomes in CWHIV and both early and middle childhood interventions to promote healthy neurodevelopment.

Numerous studies have demonstrated deficits in memory [10, 31–34] and processing speed [31, 32, 34] in CWHIV vs. uninfected children. A meta-analysis concluded processing speed, working memory, and executive functioning were domains most affected by perinatal HIV [35]. Our data are consistent with these studies, given that processing speed, memory and behavioral regulation, an aspect of executive functioning, were among the outstanding domains with observed deficits in children with early-ART. Both the early- and late-ART groups had similarly marked deficits in processing speed, suggesting this capacity may be sensitive to insult, despite early-ART.

Residual deficits in cognitive functioning in children with early-ART may have been due to early brain insult prior to ART or early establishment of HIV reservoir in the CNS followed by ongoing neuronal injury in spite of ART or both. Systemic immune activation may also drive entry of additional HIV target cells [36], allowing increased entry of HIV [37, 38]. In a prior study of the early-ART group, we found infants with higher systemic monocyte activation markers had later age at attainment of developmental milestones, despite having virologic suppression [39]. Prolonged poor viral suppression, growth, or immune function or increased co-morbidity may have also worsened cognitive outcomes in children with early-ART.

At start of ART, both CWHIV groups had immune compromise and undernutrition; however, immunosuppression was more severe in the late-ART group. Children with late-ART had profound differences in scores for several domains vs. CHUU. Poor neurocognitive outcomes in this group may have been amplified due to more severe or prolonged immune and growth compromise prior to ART [8, 10, 11]. Our findings on severe neurocognitive impairment in children with late-ART are not surprising but suggest high future risk of broad neurocognitive compromise in children with late diagnosis and treatment and further underscore the importance of investing in interventions for these children. Given that children with early-ART also had quite severe immune and growth compromise prior to ART, it is encouraging that their prospects appeared better. It remains important to emphasize the critical benefit of expedient ART during infancy, and continued monitoring for cognitive outcomes. Cognitive outcomes were generally were lower for early-ART vs. CHUU. Whether children with early-ART will experience further cognitive declines in future remains to be determined.

Social and economic vulnerability at our CHUU catchment site may also have led to underestimation of differences between early-ART and CHUU groups. Single parenthood, lower rent, and higher household hunger were observed in the CHUU cohort. These differences may reflect geographic differences in catchment and better social and health services, such as nutritional supplementation, for the CWHIV in research. Neither local Kenyan, nor regional African reference population data for cognitive outcomes assessed in this study are available for CHUU to help evaluate whether this group was representative of the general Kenyan population. According to the 2014 Kenya Demographic and Health Survey, 29.1% of Nairobi households reported lacking food or money to purchase food in the past 7 days [40], suggesting some generalizability.

Study strenths include comparison of two cohorts of CWHIV, recruited through similar catchment, and parallel assessments of CHUU, recruited from a site with a relatively high HIV prevalence and clientele with demographic similarities to the CWHIV. At the same time, this study is limited by certain key differences between groups. Although survivor bias may be expected to correlate with improved cognitive outcomes, we note CWHIV were born in different HIV treatment policy eras, and children with late-ART had both survived HIV infection for longer without ART and lacked access to PI-based regimens. Children with late-ART lacked a similarly aged CHUU or early-ART comparator. To accommodate different ages, scores were standardized by z-scores; however, this approach involved use of US norms which may lack cultural relevance and have differential relevance across age ranges. Children with late-ART had additional time for environmental or ART toxicity or HIV disease related neurodevelopmental insult to occur. Children with late-ART had substantially higher orphanhood, likely due to having been born prior to widespread ART access in Africa. Additional unmeasured HIV-related social characteristics may have also contributed to neurocognitive differences between groups. Finally, CWHIV but not CHUU were enrolled in the context of clinical trials. In particular, the early-ART group participated in a RCT involving a brief treatment interruption; it is difficult to evaluate impact of the interruption, given small numbers and short duration [18]. We cannot rule out that these and other differences may have impacted our findings.

Additional strengths were the detailed, widely used NPT and behavioral assessment battery [12]. Numerous steps were taken to ensure rigorous testing. In addition to inherent cohort differences, this study was limited by small sample size, lack of a local CHUU norm population, and lack of detailed data on the home environment and caregiver behavior during infancy and early childhood.

In conclusion, we observed a less broad and severe neurocognitive compromise in children with early- vs. late-ART. Deficits in children with early-ART included differences in global cognition, memory, processing speed and behavioral regulation. It remains important to develop interventions to support brain development in these children. Most critically, considering that access to ART for CWHIV is persistently poor [17], improving early infancy HIV diagnosis and treatment remains an profoundly important prong in efforts to optimize quality of life in CWHIV.

Supplementary Material

Supplemental Digital Content 1

Child characteristics at ART initiation for each cohort of CWHIV

Supplemental Digital Content 2

Summary of univariable linear regression analyses comparing neuropsychological and behavioral assessment z-scores for each cohort

Supplemental Digital Content 3

Summary of proportions with neurobehavioral impairment for each cohort