Abstract
Children with sickle cell disease are at risk of cognitive deficits and somatic growth delays beginning in early childhood. We examined growth velocity from age 2 years (height and body mass index progression over time) and cognitive functioning in 46 children with sickle cell disease 4 to 8 years of age. Height-for-age velocity was not associated with cognitive outcomes. Higher body mass index velocity was associated with higher scores on global cognitive and visual-motor abilities but not processing resources or academic achievement. Body mass index progression over time may be a clinically useful indicator of neurocognitive risk in sickle cell disease, as it may reflect multiple sickle cell disease–related risk factors.
Sickle cell disease (SCD) is a genetic disorder that has variable effects on children, from occasional, mild, complications to severe pain, organ damage, and/or death. Many children with SCD experience repeated hospitalizations, with younger children particularly likely to present to Emergency Departments after symptoms have become severe (Sanders et al., 2010). Early identification of children at higher risk of serious complications is critical for developing preventative intervention strategies. Cognitive deficits in SCD have a significant impact on quality of life (McClellan et al., 2008; Schatz et al., 2004), yet little is known about how to predict which children with SCD are at the highest risk of these deficits. This study examines whether a child’s growth trajectory may be helpful in identifying children at risk.
Cross-sectional data indicate a relationship between growth and cognitive development in SCD. In a previous study, we examined the cross-sectional relationship between somatic growth and cognition in young children with SCD compared with demographically matched controls (Puffer et al., 2010). Height-for-age was significantly associated with neurocognitive risk, and higher body mass index (BMI)-for-age appeared to play a protective role for cognitive outcomes. From those results, we speculated that measures of growth over time (i.e. growth velocity) might be even more strongly associated with cognitive abilities than cross-sectional data. This study examines the association between growth velocity for height and BMI and neurocognitive dysfunction.
SCD is characterized by the production of abnormal S-type hemoglobin that causes red blood cells to polymerize in the blood stream. This causes poor oxygen carrying capacity and premature death of cells, resulting in occlusion and stenosis of blood vessels, hypoxic organ damage, anemia, and inflammation (Steinberg, 2008). Risk of cognitive decrements is among the numerous impacts of SCD. Children with the SS or Sβ0 genotypes are at higher risk than less severe genotypes (Schatz et al., 2009). Cognitive deficits are caused by a range of factors including brain damage due to stroke or silent cerebral infarcts (Gold et al., 2008; Pegelow et al., 2002; Schatz and Puffer, 2006; Wang et al., 2001), low blood oxygenation or abnormal blood flow, and sleep disordered breathing (Goldstein et al., 2011; Schatz et al., 2002; Schatz and Roberts, 2007). Common domains affected include general cognitive ability, language ability, and processing resources such as processing speed and working memory (Nabors and Freymuth, 2002; Schatz and Puffer, 2006; White et al., 2000).
Children with SCD are also at higher risk of growth delays that can begin early and increase in severity (Al-Saqladi et al., 2008; Barden et al., 2002; Henderson et al., 1994; Rhodes et al., 2009; Stevens et al., 1986; Zemel et al., 2007). Studies of growth velocity show slower progression on height, BMI, fat-free mass, and puberty onset and progression (Rhodes et al., 2009; Zemel et al., 2007). It should be noted, however, that many children with SCD remain on a normal growth curve or even become overweight (Chawla et al., 2013; Mitchell et al., 2009; Thomas et al., 2000).
Among those with delays, there may be a connection between slowed growth velocity and cognitive deficits, although this is not well understood. General risk factors for all children, such as preterm birth and low socioeconomic status (SES), have been associated with co-occurring growth and cognitive delays, as they can affect both physical and neurological development (Fernald et al., 2006; Nagy et al., 2003; Noble et al., 2005; Pietz et al., 2004). These factors disproportionately affect ethnic minorities, further elevating risk for children with SCD (Reagan and Salsberry, 2005).
SCD-specific processes could also explain associations between these outcomes (Bennett, 2011), as many of the same physiological processes caused by SCD that threaten growth could also affect cognitive development. Thus, growth and cognitive trajectories could mirror one another throughout development with easily observable growth delays serving as a warning sign for current or future cognitive problems. Increased metabolic need is one consequence of SCD that may lead to under-nutrition and inadequate energy resources for both somatic growth and brain functioning (Akohoue et al., 2007; Barden et al., 2000; Dekker et al., 2012; Reid, 2013; Singhal et al., 2002). Cerebral hypoxia and vascular damage (Steen et al., 1999), as well as endocrine abnormalities (Scheepens et al., 2005; Soliman et al., 1997), may also play a role. These processes contribute to severe anemia, red blood cell turnover, and increased infections that could affect both developmental processes. Treatments that reduce sickling and anemia, such as transfusions or hydroxyurea, appear to improve somatic growth in some children (Hankins et al., 2005; Wang et al., 2005) and also have been associated with better cognitive performance (Puffer et al., 2010).
In this study, we examine the relationship between growth velocity and cognitive decrements in young children with SCD who have not yet experienced a stroke to determine whether growth velocity is associated with early neurocognitive risk. We assess growth and cognitive abilities among children with more severe SCD subtypes (i.e. SS and Sβ0) and lower risk subtypes (i.e. SC and Sβ+). We hypothesize that higher growth velocity of children with SCD from age 2 years to the time of cognitive testing, ages 4–8 years, will be associated with better cognitive outcomes in global cognitive ability, language ability, and processing resources. We then explored variables that could help account for this relationship: preterm birth, SES, anemia severity, genotype, cerebral blood flow abnormalities, and sleep apnea.
Methods
Participants
Data were analyzed from 46 children with SCD, 4–8 years of age, recruited from medical clinics as part of a program that provides cognitive screenings for children with SCD at ages 1, 3, 5, and 7 years (Schatz et al., 2009). For this sample, 74 consecutive children were asked to participate, and 64 enrolled and completed testing. Exclusion criteria were history of overt stroke or diagnosis of a major developmental disability. Of the 10 eligible non-participants, two caregivers declined testing and eight reported scheduling difficulties. Of the 64 enrolled, 13 were excluded from analysis because of inadequate records of somatic growth at younger ages. Five cases were excluded because they represented re-assessments of the same child, so only the first assessment was included in these analyses. Only one participant was on a therapeutic dose of hydroxyurea therapy and did not appear to be an outlier for any of the analyses.
Procedures
We obtained informed consent and assent from parents and children, respectively. Participants’ height and weight were measured, and children completed a battery of cognitive tests administered in one session lasting 60–90 minutes. Licensed psychologists or trained graduate students conducted assessments during routine appointments. Children were given books as compensation, and caregivers received assessment results and recommendations from the testing. Institutional review boards of the University of South Carolina and clinics approved all procedures.
Measures
Reviews of medical charts were conducted to collect current growth status and each child’s growth trajectory between age 2 years and the testing date. For most participants, at least one set of measurements (i.e. height and weight) was obtained per year. Data were examined for values that appeared to be very inconsistent with a participant’s growth trajectory in order to exclude data points that were likely incorrect due to recording errors. Data from before the age of two years were not used because data on extent of prematurity were missing for some children, which is needed to calculate growth status by age; growth was also measured using different instrumentation in that age range. Medical records were also reviewed for history of preterm birth, hemotocrit (a measure of anemia severity) from a routine blood draw within 1 week of cognitive testing, sickle cell genotype, history of conditional or abnormal cerebral blood flow at the most recent transcranial Doppler (TCD) examination (Adams, 2005), history of severe complications, and history of sleep apnea diagnosis based on a formal sleep study (Chapman et al., 2002). Two reviewers completed reviews independently; all kappas were .85 or higher, and consensus was established for all discrepancies.
Growth velocity was measured as the linear slope of the relationship between the child’s age and their height-for-age or BMI-for-age percentile. To calculate growth velocity, we used a statistical program from the Centers for Disease Control and Prevention (CDC, 2005) to convert each child’s growth measurements over time into age-adjusted percentiles for height-for-age and BMI-for-age. Only children with at least three growth measurements available between age 2 years and time of testing were included. We plotted height and BMI percentiles to create height and weight growth curves for each child and conducted regression analyses with growth status regressed on age to determine the slope of the line of best fit. This slope indicated the rate of change of the child’s growth status over time compared with same-aged peers. Therefore, a slope of 0 indicated that a child’s growth status compared with same-aged peers remained stable over time, whereas a positive or negative slope indicated that a child’s growth status for age increased or decreased over time, respectively.
We assessed language abilities, processing resources, visual-motor ability, and academic skills with the battery of cognitive tests. Language ability was assessed with the Spoken Language Quotient score of the Test of Language Development (Newcomer and Hammil, 1997). Processing resources was measured with the Woodcock–Johnson Tests of Cognitive Abilities, 3rd ed (WJ-III; McGrew and Woodcock, 2001) with subtests of cognitive speed and working memory. Scores from these were used to compute a Cognitive Efficiency Index, referred to as processing resources (Case, 1985). A composite score was generated from two measures of visual-motor ability: the Beery–Buktenica Developmental Test of Visual Motor Integration (Beery and Beery, 2004) and the Hand Movements subtest of the Kaufman Assessment Battery for Children (Kaufman and Kaufman, 1983). Two subtests of the WJ-III Tests of Achievement, Letter-Word Identification and Applied Problems, were used to assess academic skills in reading and math (McGrew and Woodcock, 2001). Age-adjusted standard scores (M = 100; standard deviation (SD) = 15) were generated for all measures. We calculated the mean of scores for all domains to estimate a global cognitive ability index, which has yielded scores that correlated with Wechsler Intelligence Scale for Children, 3rd edition (WISC-III) Full Scale IQ at r = .85 in previous work (Schatz, 2004).
Statistical analysis
Regression analysis was used to determine whether growth velocity, the slopes of the children’s growth curves, was associated with cognitive outcomes at an alpha level of .05. For significant results, Sobel tests were used to examine whether growth velocity was functioning as a mediator for other variables linked to cognitive outcomes: preterm birth, SES, anemia severity, SCD genotype, TCD cerebral blood flow, and sleep apnea. Follow-up analyses were run to examine whether any observed associations of growth velocity with cognitive outcomes differed for younger (<6 years) versus older (≥6 years) children.
Results
Participants were 46 children, 4–8 years of age, with SCD. Table 1 shows further participant characteristics. The majority (n = 23) had SS disease and others had SC, Sβ+, and Sβ° subtypes. Age at testing ranged from 4 years, 11 months to 8 years, 4 months, with a mean age at testing of 6 years, 4 months (SD = 1.1 years). Mean linear slope of height-for-age was −.09 (SD = .40), and mean linear slope of BMI-for-age was .07 (SD = .45). There was significant heterogeneity within the cognitive performance of children with SCD, with scores across domains ranging from deficient to high average descriptive ranges.
Demographic and disease data for growth velocity analyses (N = 46).
Regression results showed that height-for-age velocity was not significantly associated with any cognitive scores. In contrast, BMI-for-age velocity was found to be related to global cognitive ability scores and visual-motor ability at an alpha level of .05, and a trend toward significance was observed for language abilities (p = .055). Higher BMI-for-age velocity was associated with higher scores in these domains, explaining 8–11 percent of the variance. BMI-for-age velocity was not associated with processing resources or academic achievement abilities (see Table 2). Follow-up analyses were run to examine whether the observed associations of BMI-for-age velocity with cognitive outcomes differed for younger (<6 years) versus older (≥6 years) children based on a median age split of the sample. The interaction effect for BMI-for-age velocity and age group was not statistically significant for either global ability (p = .449) or visual-motor ability (p = .254) and accounted for less than 3 percent of explained variance, suggesting relatively similar effects for younger and older children.
Regression results for associations between BMI-for-age velocity and cognitive outcomes.
BMI: body mass index; SE: standard error.
Global cognitive ability: mean of standard scores of all other domains.
TOLD: Test of Language Development–Primary Version, 3rd ed. (Newcomer and Hammil, 1997).
WJ-III: Woodcock–Johnson Tests of Cognitive Abilities, 3rd ed. (McGrew and Woodcock, 2001).
WJ-III: Woodcock–Johnson Tests of Achievement, 3rd ed. (McGrew and Woodcock, 2001).
Beery Test of Visual Motor Integration (Beery and Beery, 2004).
Kaufman Assessment Battery for Children (K-ABC) (Kaufman and Kaufman, 1983).
p < .05, ⊥p < .10.
The mediation analyses focused on the association between BMI-for-age velocity and the two statistically significant outcome measures: global ability and visual-motor ability. For evidence of mediation to occur, the independent variables of interest must be correlated with the outcome variable. Pearson correlations indicated that for global ability, the variables of anemia severity (r = .32, p = .024), genotype severity (r = .42, p = .002), and sleep apnea (r = .30, p = .031) were potentially important independent variables. For visual-motor ability, none of the potential independent variables were correlated with BMI-for-age velocity (all p values > .10). The mediation path from each independent variable to BMI-for-age velocity indicated associations for all three independent variables (anemia severity, β = .371, t = 2.71, p = .010; genotype severity, β = .304, t = 2.23, p = .010; sleep apnea, β = −.440, t = −3.43, p = .001) with BMI-for-age velocity. Sobel tests examining the mediation pathway showed statistical trends for each of the three variables (anemia severity, p = .084; genotype severity, p = .092; sleep apnea, p = .072), although none of these reached the conventional alpha level of p < .05.
Discussion
Results document associations between BMI velocity and cognitive scores during early childhood on measures of global cognitive ability and visual-motor ability, as well as a trend suggesting that BMI velocity may also be related to language ability. BMI-for-age velocity accounted for approximately 8–11 percent of the variance in these cognitive scores. This relationship was in the positive direction, indicating that children showing stable or increased BMI-for-age status throughout early childhood performed better in these domains than children with decreasing BMI-for-age status over time. These effects were not found for processing resources or academic achievement. Therefore, results suggest that BMI-for-age velocity may be one useful indicator of neurocognitive risk for children with SCD, although likely not uniformly across all domains of cognitive functioning.
The mediation analyses did not identify any specific risk factors for poor cognitive outcomes that might be expressed through growth velocity. However, anemia severity, genotype severity, and history of sleep apnea all showed trends toward significant mediation. Given the limited sample size and statistical power in our analyses, these data may indicate that BMI-for-age velocity is functioning as a proxy variable for a number of disease-related factors. That is, none of these factors individually fully accounts for the association of growth velocity with cognitive ability in our data, but perhaps the aggregate of these effects is responsible for this association.
Examination of growth curves showed that BMI status changed during early childhood for some children with SCD. A small majority of participants, 28 out of 46, exhibited a positive BMI-for-age slope. The child with the steepest positive BMI slope exhibited an increase from the 3rd percentile for BMI at age 2 years to the 25th percentile at age 7 years. Several other children exhibited steady increases in BMI status, with their earliest measurements falling within the first quartile at age 2 years and later measurements falling within the third quartile at age 5–7 years. These patterns show the possibility of significant increases in BMI status that may be related to better cognitive outcomes. In contrast, the children exhibiting negative BMI slopes, which were associated with lower cognitive scores, varied in their growth patterns; some showed fluctuating BMI status reflective of periods of normal weight increases punctuated with periods of less weight gain relative to peers, while other children exhibited steady decreases in BMI status relative to norms. A larger sample of growth trajectories and cognitive outcomes is needed to evaluate whether these more specific growth patterns are associated with poorer cognitive outcomes.
Height velocity across early development was not significantly associated with cognitive performance. Very few participants exhibiting low height status at age 2 years had a steep positive height slope over time; rather, they tended to show slopes close to 0, indicating very little change in status relative to peers. This is consistent with studies that have documented persistent and progressive height deficits in children with SCD throughout childhood and into adolescence (Stevens et al., 1986; Zemel et al., 2007), indicating that catch-up growth may occur later or not at all in some children. Therefore, children in our sample whose height was in the average range at the time of this study may exhibit decreased height status at later ages that could be associated with lower cognitive scores. This is one explanation for the fact that our results are in contrast to the cross-sectional study by Knight et al. (1995) that documented positive correlations between height trajectory from early childhood to adulthood and IQ scores as adults.
Results are also consistent with the findings in other pediatric populations with early growth deficits. For example, a study of children with intrauterine growth retardation demonstrated positive associations between growth measurements over a 10-year period and IQ scores (Fattal-Valevski et al., 2009). Similarly, a meta-analysis of neurodevelopmental outcomes in preterm and very-low-birth-weight infants documented moderate-to-severe deficits during childhood in academic achievement, attention problems, verbal fluency, working memory, and cognitive flexibility (Aarnoudse-Moens et al., 2009). An additional study found that children who were born with very low birth weight after severe intrauterine growth retardation often developed deficits over time in visuospatial ability, nonverbal reasoning, and strategy formation (Smedler et al., 1992).
In our previous case–control study on growth and cognitive ability in children with SCD (Puffer et al., 2010), from which these participants are a subsample, low height-for-age was associated with lower cognitive scores observed in children with SCD compared with peers during early childhood. Data from the current study further suggest that children with SCD are not exhibiting catch-up growth in height during this developmental period. These results do not necessarily mean that height-for-age gains during early childhood would not be associated with cognitive improvement in some young children with SCD if they occurred; rather, this study was unable to answer that question because most children did not show these gains.
Taken together, evidence suggests that at any one point in time, height status is likely a better indicator of neurocognitive risk than BMI status. However, when tracking growth over time, BMI-for-age velocity may be more indicative of risk. Furthermore, in this study, BMI-for-age was a more robust indicator of cognitive abilities in certain domains than other known risk factors (e.g. anemia severity, preterm birth, and clinical complications of SCD), perhaps because it functions as a proxy indicator of multiple risk factors for cognitive outcomes. Although data from this study provide only preliminary evidence for these relationships, understanding the nuances in the ways that growth is related to cognitive development is important for determining how growth should factor into the constellation of clinically useful indicators of neurocognitive risk. Understanding variables that may be related to cognitive development could improve clinical decision making related to identifying children with SCD who may need cognitive testing to identify needs for intervention.
The main limitation of this study is that conclusions on growth velocity are related to growth progress beginning at age 2 years, rather than since birth, as medical records did not all have adequate information about gestational ages of children born preterm. We were, therefore, not able to examine potential effects of early growth during infancy. In addition, since growth velocity curves were created from measurements taken during clinic visits, the exact ages and intervals at which children were measured varied. In future studies, larger samples and more frequent growth measurements from birth would increase the reliability of height and BMI velocity measures and lead to stronger conclusions. Head circumference measurements may also be useful to collect in future work, as smaller head circumference has been correlated with lower cognitive functioning in other child populations (Cheong et al., 2008; Heinonen et al., 2008). Finally, systematic data were not available on any cognitive or academic interventions some children may have received, which could influence the magnitude of growth–cognition relationships.
This is the first study to our knowledge that examines height and BMI velocity in relation to cognitive functioning in children with SCD in this younger age range. Results suggest that higher BMI-for-age velocity during this developmental period may be related to better development of global cognitive abilities, visual-motor ability, and potentially language abilities. Findings provide a rationale for further research on growth velocity and cognitive development that could lead to the development of novel interventions for the early intervention and prevention of cognitive deficits in this population.
Footnotes
Acknowledgements
The authors thank the participants in this study and the staff of the medical and educational settings for data collection. They also thank Carmen Sanchez, Catherine McClellan, Melita Stancil, Christopher Wellbaum, Catherine Macilwinen, Amy Hurt, and Kristy Pruitte for assisting with data collection and management. The authors also acknowledge the contributions of dissertation committee members of the first author, Dawn Wilson-King, PhD and Sandra Kelly, PhD.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded, in part, by a grant from the March of Dimes Birth Defects Foundation, Inc. (Award 12-FY02-109, Principal Investigator: Jeffrey Schatz).