A total of 468 children (306 HIV+ and 162 HEU) had a valid and complete language assessment, at a median age of 12 years (range 7–16); the children were 52% male, 69% Black and 26% Hispanic. displays demographic and caregiver characteristics by HIV infection status for the 437 children in whom Pri-LI, Con-LI and no LI could be distinguished (excluding 8 children due to missing WISC-IV PRI scores and 23 children with LI and multilingual exposure but no other concurrent hearing or nonverbal cognitive impairment). The children with HIV in our study were more likely to be older, female, Black, and non-Hispanic than HEU children. While there were no significant differences in most caregiver characteristics (education, marital status, or performance IQ) between the HIV+ and HEU groups, children with HIV were less likely to be from a low income household and to have their biological parent as their caregiver. As expected, children with HIV were less likely to have been exposed to ARV drugs in utero, as compared to children in the HEU group (18% vs. 85%).
Demographic characteristics and Language Impairment Category by HIV Status for 437 HIV-infected or HIV-Exposed but Uninfected Children in the Adolescent Master Protocol of the Pediatric HIV/AIDS Cohort Study
Most children (93%) were administered the CELF-4 in English. Of the 437 children, 153 (35%) scored more than one SD below the general population mean on the CELF-4 Core Language Score (CELF < 85). The rates of LI were comparable between the HIV+ and the HEU groups (34% vs 37% respectively) and the mean Core Language scores were also similar (88.5 vs 87.5). Overall, 32 children (7%) had hearing impairment, 88 (19%) had multilingual exposure, and 137 (29%) had low nonverbal cognitive scores (<85).
Among the 437 children for whom LI category could be identified 48 (11%) had Pri-LI and 105 (24%) had Con-LI. Concurrent conditions for the children with Con-LI included a low nonverbal cognitive score for 94 (90%), hearing impairment for 20 (19%) and both low nonverbal cognitive score and hearing impairment for nine (9%). Multilingual exposure was present in combination with either low nonverbal cognitive score or hearing impairment for 18 (17%). Of the children without LI, 43 (15%) had low nonverbal cognitive score, 12 (4%) had abnormal hearing, and 47 (17%) had multilingual exposure. One child (<1%) of the 284 without LI had both a low nonverbal cognitive score and hearing impairment.
The percent within each language impairment category by HIV infection status and demographic factors is summarized in . The prevalence of Pri-LI and Con-LI were similar in the HIV+ and HEU groups, with 10% vs 12% for Pri-LI, and 24% vs 25% for Con-LI, respectively. The mean CELF-4 Core Language standard scores were highest for those with no LI (98.7) and slightly higher for those with Pri-LI as compared to Con-LI (74.2 vs 65.8). The Pearson correlation between nonverbal IQ and CELF Core Language Standard Scores was 0.63 among all subjects, but higher for those with Con-LI (0.51) and those with no language impairment (0.37) than for children with Pri LI (0.29). Most children had consistent levels of nonverbal IQ and language functioning, with 54% scoring above 85 for both the nonverbal WISC and CELF-4 and 24% scoring below 85 for both (i.e., Con-LI). However, 91 children (21%) had inconsistent measures of functioning, with 11% showing only language deficits and 10% observed with normal language functioning but low nonverbal IQ. These inconsistencies in the general continuum of severity motivate separate evaluation of predictor relationships for Con-LI and Pri-LI, to examine possible differences between the two groups of children with LI.
Prevalence of Language Impairment by HIV Status and Demographic Characteristics for 437 HIV-Infected or HIV-Exposed but Uninfected Children in the Adolescent Master Protocol of the Pediatric HIV/AIDS Cohort Study
A summary of HIV disease severity by language impairment group is provided in and illustrated in . Children with Con-LI were significantly more likely than those with Pri-LI or no LI to have initiated ART by 6 months of age, to have a prior CDC Class C diagnosis, a detectable viral load, and CD4% < 25%. Most children were on HAART with protease inhibitor (PI) at the six-month visit (72%) with no differences in ART regimen by LI group.
HIV Disease Markers, Current Antiretroviral (ART) Regimen, and ART History by Language Impairment Category for 284 HIV-Infected Children in the Adolescent Master Protocol of the Pediatric HIV/AIDS Cohort Study
Percent with Certain Indicators of HIV Disease Severity among Perinatally HIV-infected Children within each Language Impairment Category, with 95% Confidence Intervals.
summarizes the results of univariate and final adjusted multivariable logistical regression models for Pri-LI vs no impairment, both overall (upper panel) and within the group of children with HIV (lower panel). For the overall study population, the final adjusted model indicated no significant difference in the odds of Pri-LI vs No LI for HIV+ as compared to HEU groups of children; however, children who were Black had over twice the odds of Primary LI vs. no LI (aOR=2.43, p=0.03).
Logistic Regression Models of Primary Language Impairment vs. No Language Impairment Among All Children and Within HIV-Infected Children
presents the results of analogous models for Con-LI vs no LI groups, again both overall (upper panel) and within the group of children with HIV (lower panel). Univariate logistic models for Con-LI versus No LI among HIV + and HEU revealed a significant effect of caregiver education (p=0.01) and a marginal effect of low caregiver nonverbal IQ (p=0.06), but no significant association with HIV status or other characteristics. However, in the final adjusted model, a significant increase in the odds of Con-LI was observed for children who were Black (aOR=3.7, p=0.003) or Hispanic (aOR=3.2, p=0.01), who had a less educated caregiver (aOR=1.8, p=0.04) or a caregiver with low cognitive score (aOR=3.3, p=0.05) or had a caregiver who was not the biological parent (aOR=1.9, p=0.02). There was no significant effect of HIV infection status on odds of Con-LI vs no LI.
Logistic Regression Models of Concurrent Language Impairment vs. No Language Impairment Among All Children and Within HIV-Infected Children
In a sensitivity analysis based on the continuous CELF-4 scores reported in , linear regression models indicated no difference in adjusted mean scores between HIV+ vs HEU (mean difference =0.44, 95% CI: −3.20, 4.08, p=0.81). However, participants of Black race had an adjusted mean CELF-4 score which was 6.3 points lower than those of non-Black race (p=0.002), and those of lower socioeconomic status had lower mean scores, as reflected by having a caregiver without high school education (4.9 points lower, p=0.02) and an annual household income below $20,000 (3.4 points lower, p=0.06). In addition, low caregiver cognitive score based on PIQ<70 was associated with significantly lower adjusted mean scores (10.0 points lower, p=0.04). Further adjustment for child’s cognitive level led to slight attenuation of differences in adjusted means noted above, but race, caregiver education, and low caregiver nonverbal IQ remained significantly associated with CELF-4 scores while HIV status remained clearly non-significant (mean difference=0.8, p=0.62).
Linear regression models for the continuous CELF score, adjusted for other covariates as needed, for all subjects and restricted to HIV+ subjects
Within the HIV+ group, univariate and adjusted logistic regression models investigating disease severity measures are summarized in (bottom panel) for Pri-LI vs no LI and (bottom panel) for Con-LI vs no LI. In univariate models, none of the disease measures reached statistical significance for predicting the odds of Pri-LI vs no LI. In the multivariate model, only age at ART initiation reached significance indicating a two-fold increase in odds of Pri-LI for those with later ART initiation. In contrast, those with detectable viral load, prior CDC Class C diagnosis, or earlier ART initiation had two to three-fold higher odds of Con-LI vs no LI. Within the HIV+ subgroup, children who were male or had a biological parent as caregiver had significantly lower odds of Con-LI, while those who were Black or had a caregiver with lower education had significantly higher odds. Although children with HIV with lower CD4% (<25%) had significantly increased odds of Con-LI in univariate models, this association did not persist after adjustment for other measures of HIV disease severity. Similarly, in sensitivity analyses based on linear regression models for the continuous CELF-4 score restricted to participants with HIV, only HIV viral load>400 copies/mL and CDC Class C were associated with lower mean CELF scores (5.8 and 4.3 points lower, respectively).While children with low CD4% had marginally lower mean scores in unadjusted models, this finding did not persist after adjustment for viral load and CDC Class. After further adjustment for child’s nonverbal IQ, CDC Class C was no longer associated with CELF scores, but child’s nonverbal IQ was associated with a 19.1 point lower adjusted mean CELF-4-4 score (see ).