In utero exposure to HIV and/or antiretroviral therapy: a systematic review of preclinical and clinical evidence of cognitive outcomes

Abstract Introducion With the increasing number of children exposed to HIV or antiretroviral therapy in utero, there are concerns that this population may have worse neurodevelopmental outcomes compared to those who are unexposed. The objective of this study was to systematically review the clinical and preclinical literature on the effects of in utero exposure to HIV and/or antiretroviral therapy (ART) on neurodevelopment. Methods We systematically searched OVID Medline, PsycINFO and Embase, as well as the Cochrane Collaborative Database, Google Scholar and bibliographies of pertinent articles. Titles, abstracts, and full texts were assessed independently by two reviewers. Data from included studies were extracted. Results are summarized qualitatively. Results The search yielded 3027 unique titles. Of the 255 critically reviewed full‐text articles, 25 met inclusion criteria for the systematic review. Five articles studied human subjects and looked at brain structure and function. The remaining 20 articles were preclinical studies that mostly focused on behavioural assessments in animal models. The few clinical studies had mixed results. Some clinical studies found no difference in white matter while others noted higher fractional anisotropy and lower mean diffusivity in the brains of HIV‐exposed uninfected children compared to HIV‐unexposed uninfected children, correlating with abnormal neurobehavioral scores. Preclinical studies focused primarily on neurobehavioral changes resulting from monotherapy with either zidovudine or lamivudine. Various developmental and behavioural changes were noted in preclinical studies with ART exposure, including decreased grooming, decreased attention, memory deficits and fewer behaviours associated with appropriate social interaction. Conclusions While the existing literature suggests that there may be some neurobehavioral differences associated with HIV and ART exposure, limited data are available to substantially support these claims. More research is needed comparing neurobiological factors between HIV‐exposed uninfected and HIV‐unexposed uninfected children and using exposures consistent with current clinical care.


| INTRODUCTION
For the 1.4 million children born annually to mothers living with HIV, there are concerns that HIV or antiretroviral therapy (ART) exposure may negatively impact neurodevelopment, including cognition, language and motor skills [1,2]. A metaanalysis on neurodevelopment in HIV-exposed uninfected (HEU) children highlighted the limitations in the current body of literature including the heterogeneity of the patient populations between studies, limited confounders measured and variability in ART regimens. More recent studies are often limited by sample size and have mixed results [3][4][5].
There are inherent challenges in studying the effects of HIV and ART on neurodevelopment. Children's brains undergo rapid growth and restructuring from birth to adolescence, so functional expressions of neurodevelopmental systems damaged in utero may not be easily detectable early in life or may not manifest until later [6]. For example, most assessments performed on young children measure general categories of neurodevelopment, such as overall cognitive ability. This prevents more refined analysis of processing speed, working memory and fluid reasoning, which can be measured with greater psychometric rigor when children are older. An additional challenge is delineating the direct effects of HIV exposure compared to ART exposure. Prospectively evaluating HIV exposure in utero without ensuring the pregnant mother is also on ART is unethical. However, this delineation could be critical if HEU children were found to have worse neurodevelopmental outcomes.
To overcome some of these challenges, there is a need for both clinical studies using highly sensitive and quantifiable tools and preclinical studies controlling biological states, to delineate the effects of HIV from ART. The objective of this study was to systematically review the clinical and preclinical literature on the effects of in utero exposure to HIV and/or ART on neurodevelopment.

| Search strategy
We conducted a systematic search using a protocol designed by a medical librarian (EW) in accordance with PRISMA guidelines [7]. Ovid MEDLINE, PsycINFO and Embase were searched on 17 October 2017 using a comprehensive search strategy (Table S1). On 20 January 2018, we searched Google Scholar, Cochrane Database for Systematic Reviews and the bibliographies of pertinent articles.
The initial screening of titles and abstracts was performed by two independent reviewers (MM and KB). Articles were excluded if they did not include a population exposed to either ART or HIV in utero or did not look at a neurological outcome. Full texts of the remaining articles were independently reviewed (MM and KB) to determine whether articles met the complete predetermined eligibility criteria, with disagreements between the reviewers settled after discussion.

| Eligibility
Inclusion criteria: 1) a population either exposed to ART in utero or exposed to HIV in utero without contracting the virus, and 2) primary outcome was an objective measure of neurological or cognitive status. For clinical studies, which inherently involve human subjects, this included: measuring brain structure, brain response, or neurobiomarkers with or without a neurodevelopmental assessment. An emphasis was placed on more quantitative measures of brain structure, response and neurobiomakers, in order to minimize variation in the interpretation of neurodevelopmental assessments alone, which primarily assess behaviour that can be culturally dependent and often use different scales for standardization. For preclinical studies involving animal models, this included: measuring brain structure, brain response, neurobiomarkers or neurodevelopmental assessment. Additionally, for preclinical rodent studies, exposure to ART or HIV-related proteins prior to postnatal day 7 was considered prenatal exposure, consistent with known benchmarks of neurological maturation between human and rodent foetuses [8]. Exclusion criteria: studies that only focused on HIV-infected populations or used HEU populations only as a control; studies that only measured neurodevelopment in humans (due to the existence of prior related reviews and known heterogeneity in quality of assessments); review articles; published abstracts without fulltext publications; and case study reports containing < 5 participants.

| Quality assessment
Quality of clinical and preclinical studies was assessed using The Strength of Evidence Tool [9] and the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines [10] respectively. Reviewers independently rated each article, and disagreements were settled after discussion by consensus (Tables S2 and S3).

| Data extraction
Study data were extracted into an electronic table by one reviewer (MM) and cross-checked independently by the second reviewer (KB), including study design, study population/ model organism, exposures, neurological/cognitive/biological outcomes measured, main results and limitations. Data were described qualitatively.

| RESULTS
The searches yielded 3027 unique titles. Post-screening, 255 full-text articles were critically reviewed, and 25 met inclusion criteria for the systematic review ( Figure 1). Five articles included human subjects and studied brain size, structure and function [11][12][13][14][15]. The remaining 20 articles were preclinical studies using animal models mostly focused on behavioural assessments (see Tables 1 and 2 for study characteristics and summary of outcomes).

| Clinical studies
Four studies looked at magnetic resonance imaging (MRI) findings in HEU populations; two were performed in South Africa [12,15], one in France [14] and one in Thailand [11]. In terms of methodological quality, one study was rated as good [12], two were fair-good [13,15] and two were fair [11,14]. Three studies used diffusion tensor imaging (DTI) [11,12,15]. Two noted regionally higher fractional anisotropy (FA) and predominantly lower diffusivity in HEU compared to HUU, in both neonates and children [12,15]. Higher FA was noted in the middle cerebellar peduncles and right posterior corona radiata [12,15], and lower diffusivity was noted in bilateral regions of the corticospinal tract, while the posterior corona radiata showed both significant increases and decreases in different diffusivity metrics [12]. Abnormal Dubowitz neurobehavioral scores were positively correlated with FA in the left uncinate fasciculus and negatively correlated with diffusivity in the right inferior cerebellar peduncle and bilateral hippocampal cingulum in HEU infants [15]. The third study using DTI in agematched HEU and HUU children did not detect group differences in intelligence quotient (IQ) scores, brain volume, or DTI metrics, after controlling for sociodemographic factors [11]. This study found that DTI measures were significantly associated with full scale and performance IQ scores, showing positive associations with FA and negative associations with diffusivity metrics. Subscale analyses showed the strongest effects between perceptual organization scores and diffusivity in the internal capsule, cingulum, and optic/temporal regions, including the uncinate and thalamic radiations [11].
The fourth MRI study only looked at HEU children (n = 49) and found that 50% showed MRI abnormalities, including diffuse hyperintensity in the white matter and pontine tegmentum [14]. Additionally, cerebral volume loss was seen in 10 children [14]. This study was a retrospective chart review of HEU infants and toddlers, the majority (88%) of whom were symptomatic, introducing sample bias towards abnormality [14].
One study looked at brainstem auditory evoked potentials, and showed significant delays of wave I and I-III interwave intervals in HEU infants exposed in utero to zidovudine alone or in combination with lamivudine [13]. The authors suggested   In utero: • 160 mg/kg/day AZT • Vehicle/Control Vehicle or AZT given orally, twice daily, from gestational day 10 to lactation day 10 Pup: • after delivery, the pups nursed from the mother in their respective treatment group until lactation day 10 top rearing at PND 45 (p < 0.05) Experiment 2: AZT-exposed offspring were more active than controls and AZTexposed males displayed more wall rearing at age PND 70 (p < 0.05). AZT exposure was associated with lower grooming frequency at all ages (p < 0.05).

Calamandrei (2002b) [25]
Italy Cohort CD-1 mouse model with two groups In utero: • 160 mg/kg/day AZT • Vehicle/Control Vehicle or AZT given orally, twice daily from gestational day 10 to lactation day 7 Pup: • after delivery, the pups nursed from the mother in their respective treatment group until lactation day 7 Brain-derived neurotrophic factor (BDNF) Nerve growth factor (NGF) Both BDNF and NGF were measured at PND 7, 21, and 60 n/a BDNF levels were increased in the parietal cortex for both males and females exposed to AZT throughout the time points.
In AZT-exposed females, BDNF levels were also increased in the hippocampus on days 7 and 21 (p = 0.0062) and in the hypothalamus on day 21 (p = 0.008). There were no changes in NGF in AZT-exposed females in the cortex, Unclear sample size of the animals SIVexposed in the second trimester reached criterion later than colony normal. Both animals exposed to SIV in the third trimester did poorer than colony norms on a specific cognitive test (forced set breaking), one of the two significantly poorer than colony norms on multiple specific cognitive tasks. AZT-and 3TC-exposed mice.
At PND 14, low muscimol dose enhanced locomotor activity in vehicle and 3TC but not in AZT-exposed pups.
At PND 28, no prenatal treatment effect was seen in locomotor activity.
AZT increased nociceptive sensitivity at all time points, especially in female pups. that these findings may indicate toxicity in the lower regions of the brainstem in HEU infants [13].

| Cognition/memory
Two studies reported that zidovudine-exposed mice (prepubertal and young adult) showed either significant impairment or trends towards impairment during the acquisition session of the passive avoidance task, a memory test [19,20]. However, retention of passive avoidance, spatial learning, and memory were not impacted by zidovudine exposure [20,27], and when a longer intra-session delay was introduced in a spatial learning and memory task, zidovudine-exposed mice (100 mg/kg/day) had improved performance over controls (p < 0.05) [27].

| Motor skills/nociception
In utero exposure to zidovudine alone [19,31], or with the addition of a dopamine receptor D1 agonist [34], did not affect locomotor activity of offspring. However, zidovudineexposed neonatal mice showed increased locomotor activity in response to GABAergic agonist treatment early in life, but not persisting into adulthood [31]. Zidovudine-exposed mice also demonstrated increased nociceptive sensitivity at all ages that was not dependent on GABA-regulated nociceptive mechanisms [31].

| Anxiety/sociability
Studies measuring sociability and anxiety-related behaviours report mixed findings. When given amphetamine to mimic a stress response, two studies found that zidovudine-exposed rats did not exhibit developmental delays that prohibited them from reacting to the stress [16,18]. Zidovudine-exposed rats had less wall-hugging behaviours (possibly indicative of lower anxiety) and an increased locomotor response, but no difference in rearing or sniffing. The increased locomotor response to amphetamine was only seen in females in one study [16], but in both sexes in the other [18]. One study noted that prenatal exposure to zidovudine was associated with reduced exploration of objects at all ages considered, increased wall-and top-rearing at specific ages, and hyperactivity in adulthood [24]. However, another study reported no significant differences in locomotion or rearing between zidovudine-exposed mice and controls, even after dopamine agonist injection [34].
Two studies had an additional focus on aggressive behaviours. In one, zidovudine exposure was associated with less digging (p < 0.05), higher number of aggressive bouts in both sexes (p < 0.05), and a tendency towards more inter-male aggressive behaviours [19]. A second study looking at intermale aggressive behaviours reported alterations of both aggressive and defensive actions, with zidovudine-exposed mice having significantly more frequent aggressive behaviours compared to controls in adolescent but not adult mice [32].
Several studies noted lower grooming frequencies in zidovudine-exposed mice [18,24,34]. Lower grooming frequency was also seen in zidovudine-exposed mice following administration of dopamine receptor D1 agonist or amphetamine, which was not the anticipated effect [18,34].
Exposure to zidovudine in utero was associated with a dosedependent increase in peak latency in the acoustic startle response and enhanced tactile stimuli response [28,29]. Perinatal zidovudine exposure enhanced startle responses following injection of amphetamine [28], apomorphine and serotonin agonist [29]. These results may suggest abnormal nerve conduction velocity and long-term functional alterations within the startle reflex pathways [28,29].

| Biomarkers
One study measured brain-derived neurotrophic factor (BDNF) within various areas of the brain at multiple time points [25]. In zidovudine-exposed mice, BDNF levels were increased in the parietal cortex, hippocampus and hypothalamus compared to controls [25]. Sex differences in BDNF concentrations were observed in various brain regions, with zidovudine-exposed females having increased BDNF levels in the hippocampus, cortex and hypothalamus at various time intervals and males having increased BDNF in the cortex at one time interval [25].

| Learning/memory interventions
One study hypothesized that L-acetylcarnitine, an antioxidant and neuroprotective factor, might improve ART-induced mitochondrial dysfunction and associated neuropathies [35]. The study reported zidovudine-induced impairment of spatial learning and memory that was counteracted by L-acetylcarnitine treatment (p < 0.05). In addition, zidovudine-exposed mice had reduced expression of hippocampal metabotropic and ionotropic glutamate receptors, which was counteracted by L-acetylcarnitine administration during pregnancy [35]. Of note, while pups exposed to zidovudine in utero had undetectable levels of the drug in plasma at birth, they had significant levels of zidovudine in their brain (70.2 AE 6.1 ng/mg) [35].

| Cognition/memory
Studies investigating the impact of lamivudine exposure on cognition and memory reported mixed results. In one study, lamivudine exposure in mice did not appear to affect acquisition and retention performance in passive avoidance tasks, although there was a trend towards lamivudine-exposed mice requiring more trials to reach acquisition (p = 0.066) [21]. A study assessing long-term neurobehavioral effects reported that lamivudine-exposed mice had a longer escape latency in the reversal phase in the Morris water maze, suggesting impaired reversal learning, but otherwise working and reference memory were not negatively impacted [23].

| Motor skills/nociception
One study reported that lamivudine-exposed mice (regardless of dose) had lower locomotor activity compared to controls (p < 0.05) [21]. Male mice exposed to the highest lamivudine dose tested showed decreased habituation in the locomotor activity test (p < 0.05) [21]. Additionally, no gross changes were seen in somatic and sensorimotor development [21].
A study examining the impact of nucleoside analogues on the GABAergic system reported increased locomotor activity in lamivudine-exposed mice following administration of a GABA receptor agonist at eight days of age, but this effect was not seen later in life [31].

| Anxiety/sociability
One study looked at sociability in lamivudine-exposed mice. Mice exposed in utero to lamivudine (500 mg/kg, the highest dose tested) had non-specific alterations in sensitivity to a muscarinic cholinergic antagonist (scopolamine), showing increased sniffing behaviour (p = 0.0161) and decreased immobility (p < 0.05) [23]. Lamivudine exposure in utero also influenced maternal behaviour, with lamivudine-exposed mothers showing higher rates of aggressive behaviour towards foster pups, manifesting in cannibalism [23].

| Combination therapy (zidovudine/lamivudine)
The first of two studies looking at a combination of zidovudine and lamivudine examined the effect of in utero exposure on cholinergic muscarinic neuroregulation in adulthood [22]. Responsiveness to a muscarinic cholinergic antagonist did not differ significantly between ART-exposed mice and controls, with similar habituation and response inhibition. However, ART-exposed mice displayed a higher frequency of rearing activities (p < 0.05) and a lower frequency and duration of self-grooming behaviour (p < 0.05) [22], both behaviours that involve the dopaminergic system [36,37].
The second study examined the effect of in utero exposure to a combination of zidovudine and lamivudine on a variety of neurobehavioral endpoints. ART exposure had a small but marked delayed effect on somatic and sensorimotor development, such as forelimb placing (p = 0.0025), forelimb stick grasping (p = 0.05), level screen (p = 0.0093) and pole grasping (p = 0.0038) [33]. Additionally, ART-exposed mice had selective alterations in social interaction tests, with females displaying a significant reduction in affiliative interactions, such as mutual circling (p = 0.0325) and allogrooming (p = 0.0055) [33]. There was no effect on passive avoidance or locomotor activity [33].

| HIV-derived proteins or SIV
One study looked at the impact of HIV-derived peptides on the neurotoxicity of excitatory amino acid agonists. Rat pups were administered the excitatory amino acid agonist Nmethyl-D-aspartate (NMDA) intracerebrally with or without recombinant fusion peptide envelope gag (env-gag) on postnatal day 7, the equivalent of third trimester exposure for a human foetus [17]. Histopathological scoring and measurements of hippocampal cross-sectional areas showed that the env-gag/NMDA injection rats had significantly more severe brain injury (p < 0.003) compared to those with NMDA injections alone [17]. A similarly designed study looked at two additional HIV-1 proteins, Tat and gp120 [26]. This study found that Tat had an overall transient effect on many behavioural assessments early in development and on pre-attentive processes and spatial memory in adulthood, while gp120 had more selective effects on negative geotaxis in neonates and locomotor activity in adults [26]. This study also assessed the relationship between behaviour and hippocampal anatomy, and found 81% of the variance in spatial memory (searching behaviour in the probe test) was explained by the estimated number of neurons and astrocytes in the hilus of the dentate gyrus [26].
One very small study using pigtailed macaques looked at behaviour in offspring born to SIV-infected mothers [30]. Five of the offspring were exposed but uninfected, and of these, most completed various object permanence and cognitive testing tasks slower than colony norms [30].

| DISCUSSION
This is the first systematic review of clinical and preclinical studies focused on the cognitive implications of in utero HIV or ART exposure. The few clinical studies in HEU children have been mixed. Some found no difference in white matter integrity relative to HUU children, while others reported higher FA and lower diffusivity values, correlating with lower neurobehavioral scores. The preclinical data provided a more comprehensive view of differences in brain structure, behaviour, and biomarkers in models exposed to HIV proteins or two antiretrovirals, with some adding information about how the effects of exposure may persist into adulthood.
While the clinical data were limited, they did reveal some intriguing findings for HEU children. In two of the three clinical studies using DTI, regional increases in FA were noted in HEU children, generally reflecting more densely packed axons, greater axon diameter, or greater myelination. While increased fractional anisotropy may represent higher white matter connectivity [38,39], it has also been observed in pathological conditions, such as autism [40] and attention-deficit disorder [41]. Lower diffusivity values [12] correlating with lower neurobehavioral scores [12,15] were also seen in HEU children. Lower diffusivity is thought to be related to denser white and gray matter [39]. These findings may be due to other potential confounders, such as poor maternal nutritional status and other in utero stress, as these studies only adjusted for the children's sex and age. In another study [11], correlations between DTI metrics and cognitive function fell in the more typically expected direction, with better cognitive functioning correlating with higher FA and lower diffusivity. One of the preclinical studies showed that neuronal and astrocyte density in certain areas of the brain accounted for a majority of the distribution in certain behavioural tasks [26]. Difference in methodologies used to measure white matter and study participant characteristics (e.g. age) make it difficult to speculate on mechanisms underlying these potential differences between HEU and HUU children and whether HIV and/or ART play a role. More studies measuring white matter changes and their relationship with clinical neurodevelopmental outcomes are needed to determine if differences exist between HEU and HUU. Future research in this area would provide the most benefit by having the following characteristics: an adequately powered sample size in a longitudinal cohort or a cross-sectional cohort with a large age band; appropriate technical methodologies; and associated clinical neurocognitive assessments. If a relationship between white matter connectivity and axon density, cognitive delays, and HIV and/or ART exposure was confirmed, it might allow us to identify children at risk for neurodevelopmental impairments at an earlier age, enabling timely intensive interventions providing maximal benefits.
Given increasing access, cognitive studies of HEU children should consider including advanced neuroimaging. This will help reconcile apparent discrepancies in the available studies and clarify brain structure and function relationships within this population. If relationships demonstrated in the preliminary work are replicated, these could be very informative in terms of understanding potential later consequences for HEU children. For example, the corona radiata is a white matter structure connecting the brainstem and cerebral cortex, critical for sensorimotor function, while the uncinate fasciculus and hippocampal cingulum are important for learning and memory. Abnormalities in such white matter pathways may therefore have important clinical implications for effective development of such functions.
The preclinical data also revealed some important findings regarding exposure to ART. For both zidovudine-and lamuvidine-exposed rodents, some studies suggested that memory and learning may be impaired at early ages [19,23,27]. More recent clinical studies with newer ART regimens have not found such deficits in young children [2,3,42]. Fewer data are available about specific cognitive functions in older HEU children and adolescents, but studies have found a higher frequency of reading and math impairments in HEU children compared to the general population [43]. While this review does not clearly identify HIV or ART exposure as being associated with deficits in cognition or memory, it does raise concerns regarding cognitive development in HEU. The inconsistency of results between preclinical data and clinical data in young children using neurodevelopmental assessments should encourage researchers to carefully consider methodologies used to assess neurodevelopment in young children, specifically focusing on high-quality, culturally and age-appropriate assessments and the use of technologies such as MRI to best measure cognitive function. Since some preclinical data reveal defects in younger mice that do not persist into adulthood, longitudinal follow-up of HEU children in future research is merited.
Social behaviours were another area of concern within the pre-clinical data. Data suggest that ART-exposed mice displayed less social behaviours, with significantly less affiliative interactions [33], higher number of aggressive bouts [19,32], or increased rates of maternal cannibalism of pups [23]. ART exposure differed between studies, so the aetiology of the findings remains unclear. However, little is known about social behaviours in HIV-and ART-exposed children and adults. Higher rates of autism spectrum disorder have been found in HEU populations; it is therefore hypothesized that mitochondrial dysfunction may be a contributing factor [44]. With the growing population of HEU children worldwide, further research on their long-term cognitive and social functioning is critically important.
Preclinical research has valuable potential to direct research in humans. Animal models allow for mechanistic studies, which yield clearer results compared to efforts in deciphering specific neurobiological pathways in humans. By including preclinical studies, this review further explores potential effects resulting from ART versus HIV exposure. Most preclinical studies investigated the effects of ART monotherapy, specifically zidovudine and lamivudine, on cognitive function, illuminating potential impact from the individual components of many combination therapies. Other preclinical studies used HIV proteins to mimic HIV-exposure in the absence of ART exposure, deepening the understanding of the interplay between HIV and ART exposure and brain development and function, which is not possible in human studies. For example, one study determined that zidovudine-exposed pups had high concentrations of the drug in brain, but undetectable levels in plasma [35]. Another study found elevated BDNF levels in the parietal cortex of mice exposed to zidovudine in utero, with females having particularly high levels [25]. These preclinical studies generate important hypotheses to inform the design of clinical studies, such as appropriate anatomical locations for sampling or the possible influence of oestrogen on BDNF. However, because of the limited number of studies within this review, we were unable to disentangle the effects of ART exposure from HIV exposure. Future preclinical research in this area would benefit from choosing assessments that have comparable measures in clinical research, which may be later evaluated. This will strengthen our knowledge of potential infant outcomes of ART or HIV exposure. This review has some limitations. We did not expand our search criteria to capture clinical studies looking at mitochondrial disorders and microcephaly in HEU children due to the heterogeneity of this literature. This may hinder us from comprehensively describing other conditions impacting neurodevelopment in HEU children. Also, our interpretation of the clinical data within this review is limited by the heterogeneity of the techniques used to determine structural or neurobiological outcomes, as well as the age of the study participants. While we were unable to hypothesize on a specific mechanism for the structural differences present between HEU and HUU children, we believe that the fact that structural differences were found within the studies warrants further investigation into brain structure and its relationship with clinical neurodevelopmental outcomes as children mature. Additionally, this review was limited by the lack of power calculations, small sample sizes, and the absence of point estimate and 95% confidence interval reporting. These statistical issues limited confidence in the reported results and understanding of how the small sample size may have impacted the results. The potential effects of modern ART regimens were not considered, as most preclinical studies were > 15 years old.

| CONCLUSIONS
Due to the complexities of cognitive assessments and confounding variables impacting neurodevelopment, the current literature on neurodevelopment in HEU and HUU children does not clearly indicate whether there are differences between these children. This review summarizes objective data from both clinical and preclinical studies. The combined literature suggests possible differences in white matter connectivity in clinical studies and memory and sociability differences in preclinical studies, although the applicability of these data is limited. However, due to the abnormalities in brain structure, function and biochemistry found in HEU compared to HUU, this is an area where more systematic and translational work is needed. Future preclinical studies should consider looking at specific mechanisms of neurobiological changes and use ART exposures and neurobehavioral assessments that harmonize with current clinical standards. More clinical research is needed comparing the neurobiological factors between HEU and HUU, with a focus on domains found to be impacted in preclinical research, such as memory and sociability. In doing this, we will take steps in determining the clinical implications of in utero HIV and ART exposure.

A U T H O R S ' C O N T R I B U T I O N S
All six authors meet criteria for authorship and have made substantial contributions to various facets of the study including study design, data collection and analysis, and writing and editing of the manuscript. MM first conceptualized this systematic review. EW designed the search criteria for each database, with input from MM. MM and KB reviewed all articles and performed data collection. BM, RV and LS provided expert consultation on paediatric neuroimaging, paediatric HIV, and preclinical studies involving ART/HIV exposure respectively. MM wrote the first draft with considerable input from the coauthors. All six authors take responsibility for the reported research, have critically reviewed this final manuscript, approved its submission and take full responsibility for the manuscript.

R E F E R E N C E S SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article: