Children infected with HIV are at risk for developing impairments in cognitive, language, motor, and behavioral functions. The developmental abnormalities characteristic of pediatric AIDS vary among patient subgroups in the time of onset, rate and severity of impairment, and number of domains affected . Specifically, deficits in language function have been frequently reported in HIV-infected children with expressive language skills being especially compromised [2–4].
Neurobehavioral deficits in pediatric AIDS patients appear to be caused primarily by the effects of HIV on the central nervous system (CNS) [2,5,6]. Neuroimaging has shown that brain abnormalities, such as cortical atrophy, white matter hypodensity and basal ganglia calcifications, are frequently found in children with HIV-related encephalopathy . Moreover, these CT brain scan abnormalities are associated with neuro-developmental deficits in HIV-infected children , including impairments in receptive and expressive language .
In a previous cross-sectional study investigating the language functioning of children with symptomatic HIV infection, expressive language was significantly more impaired than receptive language . Encephalopathic children scored significantly lower than nonencephalopathic children in both language domains, yet the magnitude of the receptive-expressive language discrepancy was similar in both patient groups. An uninfected sibling control group, however, did not exhibit a discrepancy between receptive and expressive language and scored significantly higher than the infected patient group. Also, the severity of overall CT brain scan abnormalities were correlated with receptive and expressive language scores for the overall HIV-infected patients. These correlations and the sibling data suggested that the observed language impairments were associated with the direct effects of HIV on the CNS rather than environmental factors. Moreover, in children with encephalopathy, a greater severity of CT brain scan abnormalities was associated with a greater discrepancy between receptive and expressive language, suggesting differential deterioration of expressive skills with advanced CNS disease progression.
Although this cross-sectional study improved our understanding of language impairments in HIV-infected children, questions remained about the development of these deficits in relation to the progression of HIV disease over time. A longitudinal approach would therefore be useful for investigating the long-term effects of HIV on language function. The few studies that followed HIV-infected pediatric patients for more than 1 year were limited to the assessment of general cognitive ability in infants [9–13]. Prospective studies are needed that include long-term follow-up evaluations of specific developmental functions on a wider age range of children.
Currently, most symptomatic children in the United States are routinely given antiretroviral therapy as part of the standard medical treatment for HIV disease. Thus, studying the natural history of pediatric HIV infection unaffected by antiretroviral treatment is not possible. However, the long-term effects of HIV on the CNS, even as influenced by antiretroviral therapy, still needs to be examined since the disease continues to progress and may affect neurodevelopment despite the use of such drugs.
The purpose of this study was to extend the findings from the initial cross-sectional study  by assessing the longitudinal effects of HIV disease on the receptive and expressive language function of symptomatic children. More specifically, this study investigated the development of language deficits over time and the association between language dysfunction and other markers of disease progression. Patients enrolled in the previous baseline language study  were administered follow-up language and cognitive evaluations after 6 and 24 months of antiretroviral treatment. Psychometric test scores were correlated with immunological measures and severity ratings of CT brain scan abnormalities, also obtained at baseline and again after 6 and 24 months of treatment, to assess the relationship between language impairments and measures of disease progression over time.
Children with symptomatic HIV infection, aged 1–13 years, who were consecutively enrolled from June 1990 to June 1992 in antiretroviral treatment protocols at the Pediatric Branch of the National Cancer Institute (NCI), were eligible to participate in this longitudinal language study. Patients received various antiretroviral drugs, including zidovudine (oral or intravenous continuous infusion), didanosine, zidovudine plus didanosine, or lamivudine during their participation on the follow-up language study. Eight (47%) out of 17 patients remaining in the study for 2 years changed to a different antiretroviral drug prior to reaching their 24-month evaluation due to either toxicities or disease progression.
The following exclusionary criteria were utilized to control for extraneous factors that may have confounded the effects of HIV on language functioning. First, patients were excluded if their hearing, as measured by speech reception thresholds by a complete baseline audiological evaluation, was not within normal limits (two patients). In addition, children with known pre-existing non-HIV-related conditions, such as Down's syndrome (two patients), congenital multiple handicaps (two patients), severe prematurity (less than 28 weeks gestation) with complications (one patient), significant perinatal trauma associated with developmental delays (one patient), and severe intraventricular hemorrhage (two patients) were excluded. Two other patients were not included in the study because they were too acutely ill at baseline to participate in the evaluation.
Children with HIV infection were referred to the NCI from throughout the United States to participate in clinical trials of antiretroviral drugs. Intake research nurses initially contacted the child's parent or legal guardian by telephone after the referral to discuss the study and medical status of the child. Patients and their families that appeared eligible for studies then visited the NCI to complete initial comprehensive evaluations. Parents or legal guardians gave informed consent for their children to participate in the research protocols, which were approved by the NCI Institutional Review Board. The patients were treated primarily on an out-patient basis.
Classification of encephalopathy
The patients were classified prior to enrollment as exhibiting HIV-associated encephalopathy by the following criteria: (i) evidence of deterioration in developmental or cognitive functioning (e.g., parental report of regression in the child's developmental milestones or significant decline in standard scores compared to previously administered tests); (ii) general level of cognitive functioning more than 2 SD below the norm (e.g., standard score of less than 70); and (iii) general level of cognitive functioning at least 1 SD below the norm and evidence of brain abnormalities on CT scan, such as cortical atrophy, calcifications, or white matter disease. These criteria are consistently used at the NCI program and have been found to be reliable and valid [14,15]. Other institutions, however, may use somewhat different criteria due to the lack of a generally-accepted classification system for HIV-associated encephalopathy in children [16,17].
The same psychologist generally administered an age-appropriate comprehensive language test to the HIV-infected children as part of their pre-treatment neuropsychological evaluation (baseline) and again after 6 and 24 months of antiretroviral therapy. Patients were not evaluated if they were febrile or experiencing an acute infection. Since no single test assessed the broad age range of the sample, the following two language measures were used as previously described . Patients received the same language measure at each evaluation.
The Reynell Developmental Language Scales  evaluates receptive and expressive language in children aged from 1 to 7 years. Patients aged from 1 year to 4 years 11 months were assessed using this test at baseline and then at the subsequent evaluations 6 and 24 months later. This comprehensive language test is organized in a developmental sequence and is composed of separate receptive and expressive language scales. Overall receptive and expressive language z-scores are derived from these scales. For comparison with the standard scores obtained on the Clinical Evaluation of Language Fundamentals Revised (CELF R), the receptive and expressive language z-scores were transformed into standard scores with a mean of 100 (SD, 15).
CELF-R  evaluates language disorders in children aged 5–16 years by assessing content (semantics), form (syntax and morphology) and memory through a comprehensive battery composed of 11 subtests. Form 1 is administered to children aged 5–7 years and form 2 is for children aged 8–16 years. Patients who initially were assessed with the CELF-R form 1 at baseline were evaluated with either the CELF-R form 1 or form 2 for the subsequent evaluations depending on their age at the time of the assessment.
As part of their baseline, 6- and 24-month evaluation, each child was also assessed with an age-appropriate standardized test of cognitive abilities [20–22] that yielded an overall score of cognitive functioning. Patients were not tested while febrile or experiencing an acute illness.
CT scan studies
CT scans of the children's brains were obtained at baseline, 6 and 24 months as part of their comprehensive medical evaluation. A pediatric neurologist (L.C.), blinded to the clinical status of the patient, rated the scan for presence and severity of various brain abnormalities (ventricular enlargement, cortical atrophy, white matter hypodensity, intercerebral calcifications, and other lesions) as well as overall degree of severity using a 100 mm analog scale, described in detail elsewhere .
CT brain scans were not obtained for two patients at the baseline evaluation and for two other patients at the 6-month assessment for unknown reasons. In addition, the quality of one child's 6-month scan was so poor that it could not be given a valid rating. At the 24-month evaluation, all patients had both a language assessment and a rated CT brain scan.
Age-adjusted CD4 percentage is a measure of immune function, which indicates the clinical status of HIV-infected patients and is associated with the number of opportunistic infections and length of survival in children . CD4 lymphocyte subset percentages were obtained for all children at baseline, and again after 6 and 24 months of antiretroviral therapy. To adjust for rapid age-related changes in normal CD4 levels in the first 4 years of life, a formula was used to transform the data to age-adjusted standard z-scores [24,25].
Analyses of variance (ANOVA) were used to compare psychometric test scores, age-adjusted CD4 percentage, and CT brain scan severity ratings at baseline between the patients who died prior to reaching the 24-month evaluation (n = 21) and the patients who received all three evaluations (n = 17). Repeated measures ANOVA were used to compare receptive and expressive language standard scores of patients who received the baseline and 6-month evaluations (n = 29) and of patients who received the baseline, 6- and 24-month evaluations (n = 17). Similar analyses were performed to compare the following subgroups from baseline to 6 months: encephalopathic versus non-encephalopathic, vertically versus transfusion-acquired infection, previously treated versus untreated, and the two cohorts that varied by age and language test administered. Repeated measures ANOVA were not performed for the subgroups from baseline to the 24-month evaluation because the small number of patients that received all three evaluations resulted in insufficient power. In addition, linear and quadratic trends over time were analysed to assess longitudinal change from baseline to 6–24 months of antiretroviral treatment in the total sample. The Greenhouse-Geisser correction was applied when appropriate. Similar analyses were performed for the indices of general cognitive ability, immune function, and overall CT brain scan abnormalities. The data were not analysed by type of antiretroviral therapy due to the small number of patients on each treatment, and since about one-half of the patients changed to a different antiretroviral regimen before the completion of this 24-month study.
Pearson product moment correlations were used to investigate the relationship between language scores, general cognitive indices, brain imaging measures, and immunological markers from the baseline and 6-month evaluations for the overall group and for the encephalopathic and non-encephalopathic subgroups. Correlations of data from the 24-month evaluations were not performed due to the small number of patients and limited degree of variability in the CT scans and psychometric test scores of these primarily non-encephalopathic patients.
The total sample at baseline consisted of 44 children (36 patients from the initial cross-sectional study  plus eight additional patients subsequently enrolled) with a mean age of 5.4 years (range, 1.2–11.5 years). Seventy-seven per cent (34 out of 44) of the children were infected through vertical transmission, and 55% (24 out of 44) were classified as having evidence of HIV-related encephalopathy.
Twenty-nine children (mean age, 6.3 years; 45% encephalopathic) of the initial 44 patients obtained a follow-up language evaluation after 6 months of antiretroviral therapy. Fifteen patients who received baseline language assessments did not complete the 6-month evaluation for the following reasons: 11 patients died before reaching the 6-month follow-up, one child was suffering from an acute illness, one child missed the scheduled evaluation, and two patients were no longer coming to the NCI for medical care.
Seventeen patients (mean age, 8.5 years; 18% encephalopathic) continued to receive antiretroviral therapy and were administered a third language evaluation at 24 months. Twelve patients having both baseline and 6-month follow-up language evaluations did not receive the 24-month assessment for the following reasons: 10 patients died before reaching the 24-month evaluation and two patients became too old for the norms of the test previously administered so that longitudinal comparison using the same test was not possible. Table 1 summarizes the demographic characteristics of the children with symptomatic HIV infection who were administered language evaluations at baseline (n = 44), after 6 months of antiretroviral therapy (n = 29), and again after 24 months of treatment (n = 17).
Comparison of patient subgroups at baseline
Baseline measures were compared between the group of 21 patients who died before reaching their 24-month assessment (deceased group, 11 patients who died after the baseline evaluation and the 10 patients who died after the 6-month evaluation) and the 17 patients who received all three language evaluations (follow-up group) to investigate possible indicators that may have been related to poor prognosis. The deceased patients had significantly lower receptive language and expressive language standard scores (F = 12.13, P = 0.001) than the 24-month follow-up patients (Table 2). The discrepancy between receptive and expressive language, however, was not significantly different between the deceased and the 24-month follow-up groups (F = 0.25). Full-scale intelligence quotient (FSIQ) score also was significantly lower in the deceased patients (F = 11.33, P < 0.01).
Overall CT brain scan abnormalities at baseline were rated to be significantly more severe in the deceased group compared with the follow-up group (F = 42.86, P < 0.0001). Furthermore, baseline age-adjusted CD4 percentage values were significantly lower in the deceased patients (F = 23.69, P < 0.0001). Age at baseline was not significantly different between the two groups (F = 1.41).
Table 3 shows the mean receptive and expressive language standard scores at the baseline, 6-month, and 24-month evaluations for the overall sample. Mean expressive language scores were significantly lower than mean receptive language scores in the HIV-infected children who were longitudinally assessed at baseline and 6 months (n = 29; F = 49.14, P < 0.0001) and in the patients who received all three assessments at baseline, 6 and 24 months of antiretroviral therapy (n = 17; F = 29.69, P < 0.0001).
No significant changes over time were found in mean receptive or expressive language in patients (n = 29) after 6 months of antiretroviral treatment (F = 0.45). However, a trend analysis was used to account for the sequence and time interval between the three evaluations. This analysis yielded a significant linear time effect indicating that overall language function had declined significantly over 24 months (n = 17; F = 5.56, P < 0.05). Fig. 1 illustrates the consistent discrepancy between receptive and expressive language and the significant decline in overall language over 24 months. No significant interactions were found in this analysis, indicating that the pattern of change over time was similar for both receptive and expressive language function.
Subgroup analyses (encephalopathic versus nonencephalopathic; previously treated versus untreated; vertical versus transfusion-acquired; Reynell versus CELF-R) were only performed for the patients who were assessed at baseline and 6 months (n = 29) due to the small number in each subgroup that received all three evaluations. For all subgroups, expressive language was significantly lower than receptive language from baseline to 6 months (all P values < 0.0001).
Furthermore, the magnitude and direction of the discrepancy between receptive and expressive language scores did not differ significantly between the various subgroups analysed, including encephalopathic versus non-encephalopathic children (F = 0.17); patients previously treated with antiretroviral therapy versus patients who had not been treated (F = 1.39); children infected in utero versus children infected by transfusion (F = 1.25); and the two cohorts of children who were administered the two different language tests due to their age (F = 0.59).
As expected, the encephalopathic group scored significantly lower in overall language than the nonencephalopathic group (encephalopathic, 73.9; non-encephalopathic, 99.0; F = 42.1, P < 0.0001). No differences, however, were found in overall language between the vertical versus transfusion patients (vertical, 84.7; transfusion, 95.6; F = 2.78) or the previously treated versus untreated patients (treated, 85.0; untreated, 91.7; F = 1.21). Furthermore, no significant changes in receptive and expressive language scores or interactions were found after 6 months of antiretroviral therapy in any of the above subgroups. However, a borderline significant interaction (group by language by evaluation) was found (F = 3.0, P < 0.10), indicating that the mean discrepancy between receptive and expressive language scores in the encephalopathic patients increased 4.0 points from baseline to the 6-month evaluation, while the mean discrepancy in the non-encephalopathic children decreased 3.8 points over 6 months.
From baseline to 6 months of antiretroviral therapy, no significant changes in FSIQ score (from 89.0 to 91.0) were found in the overall sample (n = 28; F = 0.94). Furthermore, no significant change was found in FSIQ from baseline to the 6- and 24-month evaluations (from 98.1 to 99.1 to 100. 6) for the total group (n = 17; F = 0.4).
Previously treated versus untreated patients did not differ significantly in FSIQ (F = 1.12; previously treated, 85.8; untreated, 95.8). However, children with vertical transmission had significantly lower FSIQ than children with transfusion-acquired HIV infection (F = 8.39, P < 0.01; vertical, 82.4; transfusion, 109.1). As expected, the encephalopathic children scored significantly lower than the non-encephalopathic children in FSIQ (F = 45.20, P < 0.0001; encephalopathic, 67.6; non-encephalopathic, 106.8). No significant changes in FSIQ from baseline to after 6 months of treatment or interactions were found for any of the above subgroups.
CT scan abnormalities
No significant change was found in the ratings of the overall severity of CT brain scan abnormalities from baseline to after 6 months of antiretroviral therapy (F = 0.34; n = 27; mean CT scan overall severity ratings, 31.3 to 29.6). Furthermore, no significant change was found in the overall CT scan severity ratings from baseline to the 6- and 24-month evaluation (F = 0.49; n = 16; mean overall CT brain scan severity ratings over time, from 19.7 to 17.1 to 17.0).
The encephalopathic patients had significantly higher overall CT scan severity ratings than the nonencephalopathic patients (encephalopathic, 42.9; nonencephalopathic, 20.5; F = 6.47, P < 0.05). No interactions were found indicating that CT brain scan severity ratings did not significantly change after 6 months in either group.
Age-adjusted CD4 percentage increased significantly in the 29 patients assessed from baseline to 6 months of antiretroviral therapy (F = 5.83, P < 0.05; age-adjusted CD4 percentage z-score, −2.8 to −2.4). Furthermore, in the group of 17 patients receiving long-term follow-up from baseline to 6 and 24 months of antiretroviral therapy, a trend analysis found a significant quadratic time effect (F = 8.41, P = 0.01) indicating that immune function improved significantly from baseline to 6 months but remained stable from 6 to 24 months (age-adjusted CD4 percentage z-score, from −2.4 to −1.6 to −1.7).
Age-adjusted CD4 percentage was not significantly different between the encephalopathic and the nonencephalopathic children in the group of 29 patients receiving 6 months follow-up (encephalopathic, −3.2; non-encephalopathic, −2.2; F = 2.77). The magnitude of improvement in age-adjusted CD4 percentage over 6 months did not differ between the encephalopathic and non-encephalopathic patients.
Correlations of language and cognitive function with CT scan abnormalities and immune status
Correlation of CT scan abnormality ratings with language and cognitive test scores
For the total sample, severity ratings of overall CT brain scan abnormalities at baseline were negatively correlated with receptive language (r = −0.41, P < 0.01), expressive language (r = −0.43, P < 0.01), and FSIQ (r = −0.39, P < 0.01) at baseline. Furthermore, overall CT scan severity ratings at 6 months were negatively correlated with expressive language (r = −0.48, P < 0.05) at 6 months.
For the encephalopathic group, ratings of overall CT scan severity at baseline and 6 months were negatively correlated with expressive language at baseline (r = −0.45, P < 0.05) and 6 months (r = −0.66, P < 0.05), respectively, but were not correlated with receptive language at either evaluation (baseline, r = −0.22; 6 months, r = −0.20, not significant). In the nonencephalopathic group, CT scan ratings were not significantly correlated with any language or cognitive measures.
Correlation of age-adjusted CD4 percentage with language and cognitive test scores
In the total sample, baseline age-adjusted CD4 percentage was positively correlated with receptive language (r = 0.36, P < 0.05), expressive language (r = 0.41, P < 0.01), and FSIQ (r = 0.30, P = 0.05) at baseline. Change in immune function was not correlated with change in language or cognitive test scores.
Correlation of age-adjusted CD4 percentage and CT scan abnormality ratings
In the overall group, severity ratings of CT brain scan abnormalities at baseline, 6 months, and 24 months were negatively correlated with age-adjusted CD4 percentage at baseline (r = −0.66, P < 0.0001), 6 months (r = −0.81, P < 0.0001), and at 24 months (r = −0.77, P < 0.001), respectively. Change in immune function was not correlated with change in severity ratings of CT brain scans. Similar correlations were found in the encephalopathic and non-encephalopathic groups.
This longitudinal study confirms the findings of an earlier cross-sectional study  indicating that children with symptomatic HIV infection exhibit differential expressive language deficits. The present data extend previous research by showing that expressive language was more impaired than receptive language at three different evaluations over a 24-month period. Thus, the receptive-expressive language discrepancy in children with symptomatic HIV disease is a consistent and reliable finding.
An important question about the development of language impairments in HIV-infected children, however, remains unanswered. When does the discrepancy between receptive and expressive language appear? The current study did not identify the onset of this discrepancy or any significant increase in its magnitude over time. Additional studies are needed to assess language function in young HIV-infected children who are asymptomatic in order to further investigate when expressive language deficits develop. Such information also may be used to identify the most appropriate time to begin remedial speech and language services.
This longitudinal study also examined the long-term effects of HIV disease on language and overall cognitive function. Results indicated that both the mean receptive and expressive language scores did not change over the first 6 months but showed significant decline by 24 months after baseline, despite treatment with standard antiretroviral therapies. This longitudinal pattern of language scores suggests that antiretroviral drugs initially protect the CNS from the effects of HIV, but their effectiveness becomes less over time as children become refractory or resistant to therapy. The general cognitive function of these children, however, remained stable over 24 months. This differential effect may be due to the fact that the scores from various domains being assessed by general cognitive tests are integrated into one global measure (i.e., FSIQ), which may mask the separate effects. Some domains of functioning may be more vulnerable to the effects of HIV and show a decline, while others remain stable or improve. A global measure of cognitive ability may not therefore be sensitive enough to reflect such differential changes in specific brain functions. In this study, for example, an interaction revealed that verbal IQ scores remained stable over time while performance IQ scores increased 8 points from baseline to 6 months and then decreased 3 points from 6 to 24 months (F = 9.09, P < 0.05). The changes on the performance scale are most likely due to practice effects over the short 6-month test-retest interval, while the decrease is probably related to having less practice effects after the longer time interval. Despite these changes in the performance scale, however, the FSIQ score remained the same. The fact that the overall cognitive index of the symptomatic HIV-infected sample remained stable for over 2 years also may indicate that the use of anti-retroviral drugs protects the CNS from global progressive deterioration. However, antiretroviral therapy may not be able to prevent more subtle selective deficits from developing over time due to impaired growth in specific functions that may be more vulnerable to HIV. These findings suggest that multiple domains of neuropsychological functioning need to be investigated in longitudinal studies.
Previous longitudinal research investigating neurodevelopmental functioning in children with HIV infection primarily assessed children over 6 or 12 months, whereas only a few investigated long-term effects by extending the follow-up evaluations through 24 months [9–13]. Such longitudinal research has tended to evaluate only the effects of HIV on general cognitive function and produced mixed results: some results indicated that overall cognitive function improved [26,27], remained stable [12,28,29] or deteriorated [9,12]. The different findings from these studies may be associated with the type of drug and drug delivery, length of follow-up, attrition of patients, or disease characteristics of the pediatric population, including the age of the children and proportion of patients with HIV-related CNS disease. All of the previous studies with 2-year follow-up assessed HIV-infected infants in the first years of life and found developmental deficits, whereas some reported decline in overall developmental functioning over time. The present study extended previous research by investigating the long-term effect of HIV specifically on language function in an older HIV-infected patient population.
When comparing the performance of encephalopathic and non-encephalopathic children, a trend showed that the discrepancy between receptive and expressive language increased somewhat in the encephalopathic children but decreased slightly in the non-encephalopathic patients after 6 months of treatment. This trend supports a correlation found in our original cross-sectional study  indicating that a greater severity of CT scan abnormalities was associated with a greater discrepancy between receptive and expressive language function in the encephalopathic group. In addition, a significant correlation was found among the encephalopathic patients, indicating that higher CT brain scan severity ratings were associated with more impaired expressive but not receptive language. These findings suggest that children with HIV-related CNS disease may show a gradual increasing discrepancy between receptive and expressive language skills over time. Less impaired receptive skills may respond to antiretroviral treatment, but the more severe deficits in expressive language in children with advanced CNS disease may not respond as well to therapy, possibly due to permanent structural damage. Few encephalopathic patients in our sample reached the 24-month evaluation, which prevented the assessment of possible changes in this discrepancy in children with HIV-related CNS disease over a longer period of time.
In the overall group, baseline levels of immune function and severity of CT scan ratings were significantly correlated with each other and with mean receptive and expressive language scores, indicating a relationship between CNS function and overall disease status, as previously found in other studies [4,24]. However, no significant correlations of change over time were found between CD4 percentage values, CT scan ratings, and psychometric test scores. In addition, mean CD4 percentage increased and mean CT scan abnormalities and FSIQ remained stable, while the mean language test scores declined over 24 months. These findings suggest that HIV and antiretroviral drugs had a differential effect on various systems and functions or varying degrees of impact over time. For example, antiretroviral therapy may provide therapeutic benefit to the immune system but not the CNS, due to the inability of many drugs to cross the blood-brain barrier, leaving the brain more susceptible to the detrimental effects of HIV. Furthermore, functional deficits as measured by psychometric tests may appear before structural brain abnormalities are evident on CT scans.
On the other hand, various medical, environmental, and psychosocial circumstances also may influence development over time and possibly contribute in part to deficits in language function. Such factors may confound the direct effects of HIV on the CNS and should be considered when interpreting these data. First, some HIV-infected patients may have pre-existing medical conditions that may impact upon the developing CNS. To control for such factors, patients with various preexisting non-HIV-related conditions that might affect the CNS, such as cancer, severe prematurity, intraventricular hemorrhage, and Down's syndrome were excluded. HIV-infected children were also excluded if audiological evaluations revealed hearing problems that might significantly impair their language.
In addition, a variety of adverse environmental factors, such as low socioeconomic status, repeated school absences, lengthy inpatient hospitalizations, or psychosocial stresses in the family may contribute to the development of language delays. Low socioeconomic status, however, is not associated with differential expressive language impairments per se [30,31]. Environmental variables were not controlled for in this study, but the characteristics of the 17 patients who received all three evaluations suggest that such circumstances did not appear to significantly effect the findings. First, the median education level (used to indicate socioeconomic status) of the primary female caregiver is comparable with the median level of education of adult women in the United States (12 years) , which indicates that these children were most likely not from impoverished home environments. Furthermore, the primary caregivers of these 17 patients remained the same over time suggesting a relatively stable home situation. In a related longitudinal study, coping and behavioral adjustment were investigated in a comparable sample of patients (unpublished data). Results did not indicate abnormal psychosocial functioning in the patients and no significant detrimental changes in family circumstances were observed over time. However, if one assumes that the infected patients experienced psychosocial problems or adverse environmental conditions that may influence language, then their uninfected siblings would also encounter similar circumstances and be at equally high risk for adjustment difficulties and concomitant developmental deficits. However, as found in the previous cross-sectional study , a group of uninfected siblings of the HIV-infected patients had overall language function in the average range and no discrepancy between receptive and expressive language. Thus, environmental and psychosocial factors did not appear to affect the expressive language skills of the siblings at that point in time. It is unlikely, therefore, that the HIV-infected patients from the same environment would experience significantly more adverse circumstances than their siblings and develop language impairments as a result. Thus, the sibling data from the initial study support our exclusion of environmental and psychosocial factors as primary contributors to the observed expressive language deficits. Finally, this sample of children was treated primarily on an outpatient basis, regularly attended school, and did not experience frequent or lengthy hospitalizations during the 2-year study, according to their medical records. Even so, reports of children with cancer (acute lymphoblastic leukemia), who are frequently hospitalized, have not indicated such a discrepancy between receptive and expressive language . Although the influence of various medical, environmental and psychosocial factors is possible, the characteristics of our sample and results from related research appeared to minimize the impact of such sources of variance language function of these patients. However, since a longitudinal control group was not used in this study, the effects of these confounding variables on receptive and expressive language cannot be completely ruled out.
This study exhibited a high degree of attrition, which may affect internal validity, and thus, should be considered when interpreting the longitudinal findings. Only 39% (17 out of 44) of children who initially received a baseline evaluation were administered a third language assessment after 24 months of antiretroviral therapy. More encephalopathic children died during the study, so that at the 24-month evaluation only 23.5% (four out of 17) were encephalopathic. Thus, the lower language test scores from the more impaired encephalopathic and deceased patients were excluded from the subgroup of 17 children who reached the 24-month evaluation. The overall mean language scores, however, significantly decreased over 2 years, indicating long-term compromise in the language function of children with symptomatic HIV infection, even in healthier children who were classified as having no clinical evidence of HIV-related CNS disease.
In summary, this longitudinal study suggests that the discrepancy between receptive and expressive language, with significantly more impaired expressive skills, is a reliable marker for the effects of HIV on the CNS of children. Additional studies are needed to investigate the genesis of this discrepancy. Furthermore, the long-term follow-up testing indicated that language function significantly declined over 2 years, despite treatment with antiretroviral drugs, and changes in such specific domains may not be reflected in measures of general cognitive function. Thus, longitudinal neuropsychological evaluations that measure specific cognitive functions are important for assessing the dynamic effects of HIV on the CNS. Different types of measures (CT scans versus psychometric testing) that assess CNS functioning may yield distinct but valuable information for monitoring the effects of HIV and antiretroviral therapy on the CNS. Finally, due to the high risk of language deficits in HIV-infected children, an assessment of language skills should be performed as early as possible to identify such impairments and begin remedial speech and language services when indicated.
The authors wish to thank G. Selkin-Gutman, R. Smith, N. Heilman, E. Tassone, S. Bose and A. Pikus, for their assistance with this research. The authors also thank the Rainbow Team at the Pediatric Branch of the National Cancer Institute and the children and families who participated in this study.