Treatment initiation factors and cognitive outcome in youth with perinatally acquired HIV infection

Abstract

Objectives

Although cognitive outcomes among perinatally infected youth have improved with highly active antiretroviral therapy (HAART), the impact of the age of initiation of treatment and the central nervous system (CNS) penetration effectiveness (CPE) of the regimen on cognitive outcomes is unknown. We aimed to describe the association between initiation age/regimen CPE score and cognitive outcomes in perinatally HIV-infected youth.

Methods

Linear regression was used to retrospectively assess the association between full-scale IQ score (FSIQ) and age of initiation of HAART, regimen CPE, and the presence/absence of an AIDS diagnosis before initiation of HAART in an urban US cohort.

Results

A total of 88 of 181 subjects (48.6%) had an AIDS diagnosis. In 69, AIDS preceded the start of HAART. Mean FSIQ (mean age 155.4 months) was 86.3 [standard deviation (SD) 15.6]. Neither age of initiation of HAART (P = 0.45) nor regimen CPE score (P = 0.33) was associated with FSIQ. Mean FSIQ for patients with an AIDS diagnosis before HAART initiation [82 (SD 17.0)] was significantly lower than for patients initiating HAART before an AIDS diagnosis [90 (SD 13)] (P = 0.001). Of the 129 subjects without AIDS by age 5 years, 41 (31.8%) initiated HAART before age 5 years; four of 41 later developed AIDS, compared with 32 of 88 of those who did not initiate HAART before age 5 years. The relative risk of AIDS if HAART was initiated before age 5 years was 0.19 (95% confidence interval 0.05–0.60).

Conclusions

Earlier age at HAART initiation and higher CPE score of a regimen did not improve cognitive outcomes. However, initiating HAART prior to AIDS protected against AIDS and was associated with a significantly higher FSIQ.

Introduction

In the pre-highly active antiretroviral therapy (HAART) era, neurocognitive complications of perinatally acquired HIV infection were among the most catastrophic effects of HIV infection. Between 20 and 35% of infected children developed progressive encephalopathy (PE), with a similar percentage diagnosed with static encephalopathy (SE) [1–4]. HAART has led to dramatic decreases in the incidence of encephalopathy [5–10]. It is, however, less clear if HAART has improved cognitive outcomes among those not diagnosed with PE/SE. In addition, it is unknown if there is a critical time for initiation of HAART in order to protect the developing central nervous system (CNS) of perinatally HIV-infected children.

Multiple paediatric studies have documented the protective effect of HAART in terms of progression of HIV infection and overall survival [11–17]. Several studies have suggested that earlier HIV therapy might also lead to improvements in cognitive outcomes among children and adults. Studies in adults also suggest that HAART regimens including medications with better CNS penetration might lead to improved cognitive outcomes [18–21]. The CNS penetration effectiveness (CPE) of commonly used HIV treatments has been quantified and can be compared using available scoring systems [22].

The purpose of this study was to evaluate the effect of age of initiation of HAART, and the possibility of an age threshold effect of HAART initiation, along with the regimen CPE and the presence/absence of an AIDS diagnosis before initiation of HAART on cognitive outcomes among a cohort of perinatally HIV-infected children followed longitudinally at an HIV-specialty care site.

Methods

Participants

Eligible subjects were children and adolescents with perinatally acquired HIV infection who received care at an urban hospital-based HIV care centre providing both primary and tertiary care. Subjects had to have at least one full-scale IQ (FSIQ) measurement obtained using an appropriate assessment tool after the age of 4 years. The neurocognitive testing protocol of the site is for all patients to have their first assessment near the time of diagnosis, or after the age of 12 months (for those identified in the first year of life), with repeat testing performed every 1–2 years. Testing is conducted or supervised by a licensed clinical psychologist. Potential subjects were ineligible if they lacked an FSIQ score at age 4 years or older.

Measures

This retrospective cohort study utilized archived neurocognitive test results and clinical data. Data were abstracted from medical records and a clinical database and included: demographics; maternal/birth risk factors; age; clinical status; immune status; viral loads; HIV medications, including initiation and change dates; and cognitive assessment data. The most recent FSIQ was the primary clinical outcome of interest. All subjects had results of one of the following tests, administered over the age of 4 years: the Wechsler Preschool and Primary Scale of Intelligence, revised or third edition (WPPSI-R or WPPSI-III), the Wechsler Intelligence Scale for Children, third or fourth edition (WISC-III or WISC-IV), the Wechsler Abbreviated Scale of Intelligence (WASI), the Wechsler Adult Intelligence Scale, third or fourth edition (WAIS-III or WAIS-IV), or the Stanford-Binet Intelligence Scale (S-B), third revision form L-M or fourth edition. The most recent FSIQ score was recorded as a numerical value ranging from 49 to 150. For subjects with FSIQ scores < 50, a value of 49 was utilized. HAART was defined as a treatment regimen consisting of at least three drugs, from two different classes of antiretroviral agents. Age of initiation of HAART was analysed as a continuous variable, and as a categorical variable, with initiation of HAART prior to age 5 years as a primary analysis. To evaluate a potential specific threshold age effect, initiation of HAART prior to younger annual age cut points (1, 2, 3 and 4 years of age) was also analysed.

CPE scores were calculated for each antiretroviral regimen given before age 5 years based on a previously established ranking system (scored 0–20) [22]. The highest CPE regimen prescribed was used for analysis. The CPE score was analysed both as a continuous variable and also as a categorical variable, where the distribution was broken into high (> 6), medium (5 or 6) and low/none categories based on previous studies and the distribution of our subjects’ scores [6,20,23]. The number of months for which each medication regimen was prescribed was recorded. An analysis was also performed using the interaction of the highest CPE regimen (under age 5 years) and the months on that regimen.

Diagnoses of encephalopathy were abstracted from chart review. The diagnosis of PE was based on published criteria [11], and included a loss of previously acquired developmental milestones, confirmed loss of 15 points or more on FSIQ testing, prolonged plateau without developmental advancement, and evidence of arrested brain growth through either acquired microcephaly or neuroimaging indicating cerebral atrophy [24]. Diagnosis of SE was based on a chronic, nonprogressive significant deficit in cognitive or motor skills. Patients who had significant disability that might or might not have been related to HIV infection, such as cerebral palsy at birth, or the combination of prematurity and delayed development noted from the first visit, were diagnosed with PE/SE although they did not meet criteria for an AIDS diagnosis.

Centers for Disease Control and Prevention (CDC) clinical classification and immune status category were assigned as per published CDC criteria [25]. Viral loads < 400 HIV-1 RNA copies/ml were considered undetectable, and coded as 399 copies/mL. The log₁₀ values of peak viral loads were analysed continuously and were dichotomized as < and ≥ 5.5 log₁₀ copies/mL.

Maternal substance abuse was indicated by the self-reported use of alcohol or drugs (e.g. cocaine or heroin) during pregnancy. Low birth weight was defined as a birth weight < 2500 g (5.51 pounds) and pre-term birth was defined as gestation < 37 weeks. A composite birth risk variable was considered positive if the subject had maternal substance abuse, low birth weight, or prematurity.

Analyses

Normally distributed data were described using means (standard deviations), and medians (interquartile ranges) were used for data that were not normally distributed. Multivariable linear regression models were used to assess the association between FSIQ and each of our three variables of interest (age at initiation of HAART, regimen CPE, and presence/absence of an AIDS diagnosis before HAART initiation). Covariates considered for inclusion in the model (PE/SE, AIDS, presence of an undetectable viral load before age 5 years, peak viral load ≥ 5.5 log₁₀ copies/mL before age 5 years, HAART before AIDS diagnosis, birth risk factors, gender, nadir CD4 count, immune category, and year of birth) were first evaluated univariately for their association with FSIQ and were evaluated in the multivariable model if their univariate P-value was < 0.1.

STATA version 11.2 (Stata Corp., College Station, TX) was used for statistical analysis. For all statistical tests, a two-sided P-value of 0.05 was used to determine statistical significance. Institutional Review Boards (IRBs) at both Widener University and The Children’s Hospital of Philadelphia approved the study.

Results

A total of 265 charts were reviewed. Eight-four patients were ineligible [19 died prior to age 4 years; 12 were too young at study end; 30 were seen in consultation only, or transferred out before age 4 years; 12 had no recorded Neuropsychological Testing (NPT) over the age of 48 months; and in 11 patients, the mode of acquisition of HIV was not clear (possible transfusion and/or sexual transmission)]. Of the 181 eligible subjects, 53% were female, and 78.5% self-identified as African-American, 11.1% as white, and 6.6% as Latino (Table 1). The median age at the last FSIQ measurement was 167.4 months [interquartile range (IQR) 120.8–193.8 months].

Table 1.

Demographics

All subjects	181
Gender [n (%)]
Female	96 (53)
Male	85 (47)
Ethnicity [n (%)]
African-American	142 (78.5)
White	20 (11.1)
Latino	12 (6.6)
Age (months) at FSIQ [mean (SD)]	157.1 (51.1)
AIDS diagnosis [n (%)]	88 (48.6)
Age (months) at AIDS diagnosis [median (IQR)]	49.2 (12.2–82.7)
PE/SE diagnosis [n (%)]	43 (23.8)
Age (months) at PE/SE diagnosis [median (IQR)]	36.0 (15.0–66.0)
Deceased [n (%)]	14 (7.7)
Age (months) at death [mean (SD)]	137.8 (59.4)
Ever on HAART [n (%)]	165 (91.2)
Initiation of HAART (n)	157
Age at initiation of HAART (months) [mean (SD)]	70.6 (51.2)
HAART initiated before AIDS [n (%)]	101 (55.8)
Highest CPE < age 5 years (n)	174
No HIV meds < age 5 years [n (%)]	53 (30.7)
Median (IQR)	6 (0–7)
Peak viral load < age 5 years [n (%)]	92 (50.1)
Median peak viral load (log10) < age 5 years	5.35 (4.85–5.92)
Viral load ever < 400 copies/mL before age 5 years [n (%)]	86 (47.5)
Yes (%)	57
No (%)	43
Nadir CD4 < age 5 years [n (%)]	135 (74.6)
Nadir CD4% before age 5 years [median (IQR)]	22 (14–31)
Nadir CD4 < age 5 years [n (%)]	139 (76.8)
Nadir CD4 count (cells/μL) before age 5 years [median (IQR)]	491 (288–808)

CPE, central nervous system penetration effectiveness score; FSIQ, full-scale IQ at last test; HAART, highly active antiretroviral therapy; IQR, interquartile range; n (%), number of subjects (%) with data available; PE/SE, a diagnosis of progressive or static encephalopathy; SD, standard deviation; viral load, quantitative plasma HIV RNA.

The median age at HIV diagnosis was 8.2 months (IQR 2.0–27.4 months). A total of 48.6% of subjects had a diagnosis of AIDS, at a median age of 49.2 months (IQR 12.2–82.7 months). A total of 7.7% of the subjects were deceased, with a median age of death of 118.5 (IQR 85–165) months. Forty-three subjects (23.8%) had been diagnosed with PE/SE, at a median age of 36.0 months (IQR 15–66 months).

The overall mean FSIQ was 86.3 [standard deviation (SD) 15.6]. Five subjects had FSIQ < 50, which was recorded as 49. Those with an AIDS diagnosis had a significantly lower mean FSIQ [81.9 (SD 15.6)] compared with those without AIDS [90.5 (SD 13.1)] (P < 0.001). However, when the 43 subjects with PE/SE were excluded, this difference was not statistically significant [88.5 (SD 14.1) versus 90.6 (13.1), respectively; P = 0.38].

Overall, 165 subjects (91.2%) had ever been on HAART, with a median age of initiation (for subjects with a known start date; n = 157) of 71.0 (IQR 23.3, 106.5) months. A total of 101 subjects (55.8%) had initiated HAART before an AIDS diagnosis. For those without an AIDS diagnosis before age 5 years, initiating HAART before 5 years of age was not associated with improved FSIQ [mean FSIQ for those on HAART before age 5 years was 89.4 (SD 12.1) versus 87.3 (SD 15.2) for those not initiating HAART by age 5 years; P = 0.4]. There was no significant association between FSIQ and initiation of HAART at any annual age point below 5 years (P > 0.2).

However, initiating HAART at any time before an AIDS diagnosis improved mean FSIQ compared to initiation of HAART after an AIDS diagnosis [89.8 (SD 12.9) versus 81.9 (SD 17.5); P < 0.001] and was protective against later development of PE/SE or other AIDS diagnoses. The odds ratio for development of AIDS for those without an AIDS diagnosis before age 5 years and starting HAART before age 5 years was 0.19 [95% confidence interval (CI) 0.05–0.60]. For subjects without an AIDS diagnosis at the time of last FSIQ measurement [n = 93; mean age 155.0 months (SD 49.3 months)], the mean FSIQ was 90.5 (SD 13.1).

For those with known medication regimens and initiation dates (n = 175), the median highest CPE score was 6.0 (IQR 0–7). The median length of time on the highest CPE regimen was 17 months (IQR 7–34 months). Fifty-three subjects (30.7%) received no medications prior to age 5 years. There was no significant effect of CPE on FSIQ, when CPE was analysed as a categorical or continuous variable, and regardless of including or excluding those subjects who had not initiated HAART prior to age 5 years. There was no relationship between FSIQ and an interaction of highest CPE and length of time on highest CPE regimen.

For the 92 subjects (50.1%) with known peak viral load before age 5 years, the mean log₁₀ viral load was 5.33 (SD 0.9) log₁₀ copies/mL. Among the 87 subjects (47.5%) with known post-treatment viral load measurements before age 5 years, 57% had at least one value < 400 copies/mL.

For 26 of the infants, there were no birth risk data. Of the remaining infants, 105 of 155 had at least one birth risk factor (24.3% had birth weight < 2500 g, 24.3% had gestational age < 37 weeks, and 50.3% had maternal substance abuse).

Significant associations with FSIQ in a univariate analysis were noted for birth risk, sex, a diagnosis of PE/SE and/or an AIDS diagnosis, and, among those with viral load data available, peak viral load at less than age 5 years, and a positive trend towards improved FSIQ was found if the viral load was undetectable at least once by age 5 years (Table 2). However, controlling for these potential confounding variables did not significantly alter the results of the univariate analyses.

Table 2.

Clinical and immunological factors related to last available full-scale IQ (FSIQ) in univariate analysis

	n (%)	FSIQ [mean (SD)]	P-values
All subjects	181	86.3 (15.6)
AIDS diagnosis, ever
Yes	88 (48.6)	81.9 (16.8)	< 0.001
No	93 (51.4)	90.5 (13.1)
PE/SE diagnosis, ever
Yes	43 (23.8)	74.9 (16.6)	< 0.001
No	138 (76.2)	89.9 (13.4)
AIDS diagnosis, but excluding those with PE/SE diagnosis	138 (76.2)
Yes	49 (35.5)	88.5 (14.1)	0.38
No	89 (64.5)	90.6 (13.1)
Initiated HAART before AIDS
Yes	101 (55.8)	89.8 (12.9)	< 0.001
No	80 (44.2)	81.9 (17.5)
Viral load ever < 400 copies/mL before age 5 years [n (%)]	86 (47.5)
Yes	49 (57)	89.2 (14.6)	0.09
No	37 (43)	83.2 (17.8)
Peak viral load before age 5 years [n (%)]	92 (50.8)
< 5.5 log10 copies/mL	52 (57.8)	91.1 (15.0)	0.01
≥ 5.5 log10 copies/mL	40 (42.2)	83.3 (14.2)
Birth risk [n (%)]	155 (83.4)
Yes	105 (67.6)	83.3 (17.2)	0.01
No	50 (33.4)	90.0 (14.2)
Gender
Female	96 (53)	84.1 (16.2)	0.04
Male	85 (47)	88.9 (14.5)

HAART, highly active antiretroviral therapy; n (%), number of subjects (%) with data available; PE/SE, diagnosed with progressive or static encephalopathy.

Discussion

The mean FSIQ of our subjects mirrors that of a recently reported multisite study from the Pediatric HIV AIDS Cohort Study (PHACS). In the PHACS cohort [n = 358 HIV-positive subjects; mean age 144 months (12 years), compared with 157 months in the present study], the mean FSIQ of those with and without a history of AIDS was 81.6 and 86.8 (versus 81.9 and 90.5 in the present study), respectively. Our study differed in that we included any child with an FSIQ measurement after age 4 years, even those who are now deceased. In contrast, the PHACS cohort included only children surviving and consenting to inclusion in the study through adolescence, excluding those with the most severe immunosuppression and illness who did not survive.

We could not identify a HAART initiation age that preserved FSIQ. However, starting HAART before the onset of AIDS was important. In addition, we did find a negative association between FSIQ and peak viral load below age 5 years and a trend towards a positive association of viral load suppression by age 5 years with FSIQ. Peak viral load before age 5 years and undetectable viral load before age 5 years were also associated with the risk of an AIDS diagnosis (positive association with P < 0.01 and negative association with P = 0.05, respectively), further supporting earlier initiation of HAART prior to disease progression.

PE is typically diagnosed early in life, with the majority of diagnoses occurring before 36 months of age [1,3,4]. As reported by others [5,6,11,17,26], we confirmed that initiation of HAART prior to an AIDS diagnosis protected against development of an AIDS diagnosis, including PE/SE.

There have been conflicting results of studies evaluating whether HIV neurological outcomes and survival differ between patients on higher and lower CPE HAART regimens [18–21,27,28]. Fewer data are available on the impact of high CPE regimens in treatment of perinatally infected youth. In a study of 23 infants and children with perinatally acquired HIV infection, McCoig et al. noted that a reduction in cerebrospinal fluid (CSF) HIV RNA was associated with improvement in neurological examination, suggesting that CSF antiviral penetration was an important aspect of the treatment regimen [29]. Patel et al., in a large retrospective analysis, reported a significant decrease in the incidence of PE with the use of HAART versus non-HAART treatments and improved survival for subjects with PE treated with regimens with a CPE score of 6 or more [6]. However, there were no assessments of the association between FSIQ and different CPE regimens. The assignment of CPE scores in a retrospective study such as this is difficult. Subjects in our cohort had been exposed to an average of five different treatment regimens, with widely varying lengths of time on each regimen. Patel et al. analysed CPE score and outcomes based on the highest CPE rated regimen of a subject, regardless of timing or duration of treatment. We chose to concentrate on the highest CPE regimen used before age 5 years, as we were especially interested in the impact of early treatment. We were unable to detect an association of the highest CPE score of a regimen used before age 5 years (including any association with time on highest CPE regimen) and FSIQ.

We attempted to control for the effect of potential maternal/birth risk factors on FSIQ by recording the presence of maternal substance abuse, pre-term birth, and low birth weight (as per maternal report and/or review of birth records, when available). Numerous studies have noted an association between birth factors, such as maternal substance abuse, birth weight and gestational age (including late prematurity of 35–37 weeks of gestation), and cognitive outcome in non-HIV-infected youth [30–32]. Additional factors related to poverty also clearly play a role in mediating FSIQ. We did not have data on parental education level and family income, additional factors known to influence FSIQ in both Perinatally acquired HIV-Infected (PAH) and uninfected youth [31]. In the PHACS referenced above, it is notable that the FSIQ was similar between HIV-infected youth without class C diagnosis and HIV-exposed but uninfected youth. Both groups’ mean FSIQ scores were significantly below general population norms, underlining the impact of environmental and familial characteristics on FSIQ. Similarly, in a longitudinal study of at-risk urban inner city adolescents (mean age 18 years) with and without gestational cocaine exposure, but no HIV exposure, Hunt et al. reported that the combined group mean FSIQ was 86.6 [33].

In our study, subjects without AIDS, on HAART, at a mean age of 13 years, had a mean FSIQ of 90.5 (SD 13.1). For youth with well-controlled HIV infection, without a history of AIDS, it appears that their FSIQ may approach, or mirror, that of their community peers. The interaction of the effects of psychosocial and environmental factions, controlled HIV infection, and even potentially toxic effects of antiretroviral medications remains to be further determined.

The strength of our study is that it was a longitudinal single site study including all eligible patients in care. Cognitive testing performed as a part of routine clinical practice was supervised by one of two psychologists over the length of the study. The limitations of this study include the fact that it was a retrospective study with incomplete data availability for some covariates of interest, such as birth risk factors. Patients presented at different ages, at different stages of immunological and clinical health, and treatment modalities and guidelines changed significantly over the time period covered by this study. In addition, different IQ assessment tools, and versions of those tools, were used over the study period, as dictated by standard neuropsychology practice at that time. Although all IQ assessment tools used were normed to a mean score of 100, the population used to develop those standardized norms varied in sample size, ethnicity and year of study, leading to possible confounding based on the assessment tool utilized. In addition, as noted above, we chose to focus only on the highest CPE regimen prescribed before age 5 years; we did not attempt to incorporate a ‘cumulative’ CPE score for all regimens prescribed prior to age 5 years because of the lack of consensus regarding how to provide meaningful cumulative CPE scores.

Our study supports the present US recommendations to initiate therapy in all infants diagnosed in the first year of life, and the World Health Organization (WHO) guidelines for initiation in all children at diagnosis under the age of five years, and supports the initiation of HAART prior to a diagnosis of AIDS [34,35]. Further studies will be needed to elucidate the effect of specific HAART regimens on FSIQ. The overall goal of anti-HIV treatment is long-term morbidity-free survival. For children with perinatally acquired HIV infection, the use of HAART for controlling virological replication and prevention of progression of HIV disease results in near-normal FSIQ in adolescence.