Genomic Disorders and Neurocognitive Impairment in Pediatric CKD

Abstract

Children with CKD are at increased risk for neurocognitive impairment, but whether neurocognitive dysfunction is solely attributable to impaired renal function is unclear. Data from the CKD in Children Study Chronic Kidney Disease in Children (CKiD) Study indicate that a subset of children with CKD have unsuspected genomic disorders that predispose them to organ malformations and neurocognitive impairment. We therefore tested whether the CKiD Study participants with genomic disorders had impaired neurocognitive performance at enrollment. Compared with noncarriers (n=389), children with genomic disorders (n=31) scored significantly poorer on all measures of intelligence, anxiety/depressive symptoms, and executive function (differences of 0.6–0.7 SD; P=1.2×10⁻³–2.4×10⁻⁴). These differences persisted after controlling for known modifiers, including low birth weight, maternal education, seizure disorder, kidney disease duration, and genetically defined ancestry. The deleterious effect of genomic disorders on neurocognitive function was significantly attenuated in offspring of mothers with higher education, indicating the potential for modification by genetic and/or environmental factors. These data indicate that impaired neurocognitive function in some children with CKD may be attributable to genetic lesions that affect both kidney and neurocognitive development. Early identification of genomic disorders may provide opportunity for early diagnosis and personalized interventions to mitigate the effect on neurocognitive function.

The Chronic Kidney Disease in Children (CKiD) Study is a prospective, observational research initiative comprised of a cohort of children and adolescents with mild to moderate CKD, providing the opportunity for investigation of the etiologies of impaired neurocognitive (NC) functions in this population.¹ Previous data from the study found that between 21% and 40% of children with CKD fell at least one SD below the normative mean for age on intelligence quotient (IQ), attention regulation, and executive functioning, and that the only CKD-specific variables associated with below average NC and academic achievement were lower iohexol GFR and elevated proteinuria.² Previous data have shown that children with CKD also suffer from high rates of depressive and anxiety symptoms.^3–7 The underlying reasons for impaired NC function in pediatric CKD is poorly understood and our ability to identify children at highest risk for NC and psychosocial complications is limited.

Recent studies have shown that genomic disorders (GDs) account for 15%–20% of cases of developmental delay, intellectual disability, and other congenital malformations such as cardiac or renal defects.^8–10 A comprehensive study reported that 14.2% of 15,767 children with intellectual disability or developmental delay carried a pathogenic copy number variation (CNV) larger than 400 kb.⁹ Recent studies have also demonstrated that pediatric patients with CKD have a high burden of GDs compared with normal children. Sanna-Cherchi et al. detected large CNVs diagnostic of known GDs among 10.5% of 522 patients with renal hypoplasia/dysplasia.¹¹ An additional 6% of patients had large gene-disrupting CNVs that were classified as likely pathogenic. Similarly, GDs were detected in 7.4% of the participants in the CKiD Study, with the majority of pathogenic CNVs associated with neuropsychiatric disease.¹² In both studies, the GDs were unsuspected on the basis of the clinical exam, and were detected in children with all forms of CKD. Remarkably, many of the pathogenic CNVs linked with pediatric CKD are also associated with developmental delay in other pediatric populations.⁹

On the basis of these findings, we hypothesized that in a subset of pediatric patients with CKD, GDs may be the biological link between NC impairment and CKD, and consequently, CKiD children with GDs will have a poorer performance on standardized measures of NC and psychosocial functioning. Elucidation of such an association offers the possibility of providing early information regarding risk of neuropsychiatric problems, thereby allowing for more timely and intensive interventions to potentially improve cognitive and socioemotional development.¹³

The analysis included a total of 420 children with CKD, 63% were boys and 68% were white. Clinical diagnoses included a wide spectrum of CKD etiologies, with 80% diagnosed with nonglomerular diseases.¹² Median age was 11.5 years, median disease duration was 9.0 years, median bedside eGFR was 44.3 ml/min per 1.73 m², and 13% of children had a history of seizures. Among their mothers, 57% had college education or some college education. There were 31 children diagnosed with a GD (defined as either a known GD or a likely pathogenic CNV).¹² As previously reported, GDs were associated with a modest increase in proteinuria, but no other significant differences in demographic, hemodynamic, and kidney disease parameters when compared with subjects without GDs (Table 1).¹²

Table 1.

Descriptive statistics and univariate association analyses

Variable	Noncarriers			GDs Carriers
	Na	Mean	SD	Na	Mean	SD	% Difference	t Test P Value	Wilcoxon Test P Value	Pooled SD	Cohen d
IQ score	373	97.90	15.47	30	87.60	15.29	−10.52	1.16×10−3	1.04×10−3	15.46	−0.67
Internalizing Problems score	373	52.52	11.09	28	59.79	8.85	13.83	2.38×10−4	9.54×10−5	10.96	0.66
GEC score	367	53.92	11.25	28	61.00	11.32	13.13	3.22×10−3	1.47×10−3	11.25	0.63
Achievement score	266	95.85	17.81	24	84.13	19.20	−12.24	7.70×10−3	2.87×10−3	17.95	−0.65
Externalizing Problems score	375	50.48	9.50	28	52.29	10.01	3.58	3.63×10−1	3.60×10−1	9.53	0.19
Behavioral Symptoms score	373	51.03	9.95	28	56.14	10.49	10.01	1.83×10−2	7.60×10−3	9.98	0.51
Adapting Problems score	373	47.84	10.06	28	42.32	9.97	−11.53	8.23×10−3	3.58×10−3	10.05	−0.55
Age at diagnosis, yr	383	2.04	4.27	31	2.05	4.20	0.56	9.89×10−1	7.63×10−1	4.26	0.00
Age at neurocognitive testing visit, yr	389	10.99	4.40	31	11.97	4.12	8.96	2.10×10−1	2.46×10−1	4.38	0.23
Disease duration, yr	383	8.92	4.76	31	9.93	5.03	11.39	2.86×10−1	2.65×10−1	4.78	0.21
SBP-adjusted z score	376	0.42	1.11	30	0.30	1.02	−27.08	5.67×10−1	4.95×10−1	1.10	−0.10
DBP-adjusted z score	375	0.59	0.89	30	0.40	0.85	−30.85	2.75×10−1	2.26×10−1	0.89	−0.20
HGB (g/dL)	382	12.57	1.53	31	12.39	1.36	−1.43	4.86×10−1	6.16×10−1	1.52	−0.12
Bedside eGFR, ml/min per 1.73 m2	387	46.91	19.96	31	44.93	18.88	−4.21	5.80×10−1	4.70×10−1	19.88	−0.10
Iohexol GFR, ml/min per 1.73 m2	363	47.40	18.30	27	41.57	14.79	−12.29	6.13×10−2	8.89×10−2	18.10	−0.32
BUN, mg/dl	389	30.25	13.29	31	29.71	14.07	−1.78	8.38×10−1	9.12×10−1	13.35	−0.04
Cystatin C, mg/L	362	1.74	0.65	30	1.83	0.55	5.27	3.96×10−1	2.08×10−1	0.64	0.14
Serum creatinine, mg/dl	387	1.42	0.69	31	1.49	0.62	4.81	5.63×10−1	3.84×10−1	0.69	0.10
Urine protein-to-creatinine ratio	374	0.98	1.87	30	1.87	2.36	90.06	5.34×10−2	1.13×10−2	1.91	0.46
Weight-adjusted z score	389	−0.12	1.33	31	−0.36	1.29	210.67	3.15×10−1	3.16×10−1	1.33	−0.18
BMI-adjusted z score	373	0.37	1.15	30	0.20	1.15	−46.55	4.38×10−1	5.08×10−1	1.15	−0.15

Internalizing, Externalizing, Behavioral, and Adapting Problems scores are parent-reported T-scores from the Behavior Assessment System for Children-II; Achievement score: Achievement total scaled score was derived from Wordread, Numeric, and Spell from the Wechsler Individual Achievement Test-II-Abbreviated Total test; GEC scaled score obtained from the Behavior Rating Inventory of Executive Functions-Parent test, derived from behavior regulation index and metacognitive index; SBP-, DBP-, weight- and BMI-adjusted z scores are z-scores on the basis of age, sex, and height. SBP, systolic BP; DBP, diastolic BP; HGB, hemoglobin; BMI, body mass index.

^aNumber (N) of cases per variable differ from total number of GD carriers (n=31) and noncarriers (n=389) because of missing values in the data set.

The NC and psychosocial evaluation of the study cohort revealed significant correlations between several NC variables: IQ score was correlated with Achievement score (R²=0.49; P=6.5×10⁻⁴⁴); Internalizing Problems score with Behavioral Symptoms score (R²=0.41; P=4.8×10⁻⁴⁷); Global Executive Composite (GEC) score with Externalizing Problems score (R²=0.45; P=2.4×10⁻⁵¹), Behavioral Symptoms score (R²=0.59; P=5.2×10⁻⁷⁶), Adapting Problems score (R²=0.44; P=1.3×10⁻⁵⁰), and Internalizing Problems score (R²=0.24; P=7.1×10⁻²⁵). Other pair-wise correlations were more modest (Supplemental Material, Supplemental Figure 1). In addition, GEC scores have been previously found to be an important feature of NC dysfunction in pediatric CKD.^14,15 We therefore selected IQ, GEC, and Internalizing Problems scores as primary NC outcomes for the genetic association study because they represent clinically relevant measures of NC function.

On univariate analyses, GDs were associated with significantly poorer scores, with 10.5% lower IQ (P=1.16×10⁻³), 13.8% higher Internalizing Problems score (P=2.38×10⁻⁴), and 13.1% higher GEC score (P=3.22×10⁻³) (Figure 1, Table 1). These findings were supported by other NC measures that were correlated with the primary outcomes (e.g., Achievement or Behavioral Symptoms scores). Among GD carriers, there were no significant differences in NC outcome variables between carriers of known GDs and carriers of likely pathogenic CNV (Supplemental Table 2). Examining alert thresholds (i.e., scores falling one SD beyond the normative age-adjusted population means, which would classify the subject as being at risk), 40% of children with GDs were at risk for intellectual disability, 46% for internalizing problems, and 57% for executive function deficits compared with 22%, 23%, and 29% of patients without diagnostic CNVs, respectively (IQ: odds ratio =2.4; 95% confidence interval [95% CI], 1.0 to 5.4; P=0.04, Internalizing Problems: odds ratio =2.9; 95% CI, 1.2 to 6.8; P=0.01, GEC: odds ratio =3.2; 95% CI, 1.4 to 7.8; P=0.004).

Figure 1.

GD carrier status is associated with impaired NC scores. Violin plots show the distribution of IQ, internalizing problems, and GEC scores among GD carriers (gray) and noncarriers (white). Superimposed box plots show median, interquartile range, and whiskers extending to nonoutlier highest and lowest values per group.

The association of GDs with NC measures was robust to adjustment for known covariates, including bedside eGFR, low birth weight, maternal education level, history of seizures, duration of disease, and genetic ancestry (Table 2). After these adjustments, GDs were still associated with significantly lower IQ score (β=−7.6; 95% CI, −2.2 to −13.0; P=0.006), higher Internalizing Problems score (β=6.5; 95% CI, 2.3 to 10.8; P=0.003), and higher GEC score (β=5.8; 95% CI, 1.4 to 10.3; P=0.01).

Table 2.

Multivariate linear regression for NC outcomes

Predictors	IQ Score			Internalizing Problems Score			GEC Score
	Estimate	95% CI	P Value	Estimate	95% CI	P Value	Estimate	95% CI	P Value
GD carrier	−7.62	−13.03 to −2.20	5.94×10−3	6.51	2.26 to 10.76	2.78×10−3	5.82	1.39 to 10.25	0.01
Bedside eGFR, ml/min per 1.732 m2; ln transformed	−0.05	−3.78 to 3.68	0.98	−0.16	−2.97 to 2.66	0.91	−2.47	−5.44 to 0.49	0.10
Low weight at birth	−6.11	−9.79 to −2.42	1.22×10−3	−2.66	−5.45 to 0.13	0.06	1.46	−1.54 to 4.45	0.34
Maternal education
High school or less		Reference			Reference			Reference
Some college	6.17	2.67 to 9.66	5.88×10−4	0.56	−2.10 to 3.22	0.68	−1.39	−4.17 to 1.40	0.33
College	10.64	7.22 to 14.05	2.43×10−9	−3.6	−6.20 to −0.99	6.92×10−3	−3.27	6.00 to −0.55	1.87×10−2
History of seizures	−10.13	−14.2 to −6.06	1.50×10−6	3.71	0.64 to 6.79	1.80×10−2	5.22	2.06 to 8.38	1.26×10−3
Disease duration, yr	−0.05	−0.36 to 0.26	0.75	−0.02	−0.26 to 0.21	0.85	0.08	−0.17 to 0.33	0.53
Principal Component 1	48.18	18.15 to 78.21	1.73×10−3	−18.81	−41.86 to 2.24	0.11	−16.58	−40.47 to 7.31	0.17
Principal Component 2	20.54	−10.70 to 51.77	0.20	−16.73	−39.48 to 6.02	0.14	12.02	−14.19 to 38.24	0.37

As expected, other significant, additive predictors of NC measures included history of seizures, maternal education, and low birth weight (Table 2). In particular, maternal education was significantly associated with attenuated severity of NC impairment in children with GDs: maternal college education and GDs each conferred opposing effects of similar magnitude on NC function (Figure 2). Hence, among children with GDs, offspring of mothers with any college education displayed an average IQ in the low-normal range, whereas offspring of mothers with only high school education scored more than two SDs below the population mean.

Figure 2.

Opposing effect of maternal education level and GD carrier status on IQ. Violin plots show the distribution of IQ among GD carriers (gray) and noncarriers (white) and among children to mothers with any level of college education (left panel) or only high school education (right panel). Superimposed box plots show median, interquartile range, and whiskers extending to nonoutlier highest and lowest values per group.

Neurodevelopmental deficits have long been recognized as a major complication of pediatric CKD.^15–17 Children with mild to moderate CKD have lower IQ and higher (i.e., poorer) GEC T- scores than the normative age-adjusted means,² with low birth weight, history of seizures, and maternal education also identified as significant covariates. As previously reported, a significant subset of pediatric patients with CKD have unsuspected GDs.¹² Because most of the GDs identified in our previous report have been independently associated with neuropsychiatric and neurodevelopmental impairment, we had hypothesized that GDs might impair both renal and NC function and partially account for poor NC performance in children with CKD.^9,11,18–25 Here, we tested this hypothesis by comparing NC scores in the CKiD Study, and provide direct evidence that GD are strongly associated with poorer NC outcomes in patients with CKD. These findings were consistent across multiple standardized measures of NC performance and had a clinical meaningful effect. These results suggest that NC impairment in some pediatric patients with CKD may not be solely attributable to impaired renal function, but may be caused by a genetic lesion that affects both kidney and NC function.

Our findings are consistent with recent studies correlating CNVs with NC outcomes in the general population.^22,23 For example, Stefansson et al.²³ reported that “neuropsychiatric” CNVs affected cognition in apparently healthy controls in the Icelandic population, conferring one SD lower IQ, and 0.7 SD lower score for overall level of NC functioning and ability to carry out activities of daily living. These effects are remarkably comparable to the 0.7 SD decrease in IQ conferred by GDs in the current study.

As expected,^9,20,26–28 NC measures in carriers of 1q21 deletions and duplications, 15q24 deletion, and triple X and Wolf–Hirschhorn syndromes indicated a high risk of intellectual disability. Conversely, carriers of the hereditary neuropathy with liability to pressure palsies deletion, which has not been clearly associated with intellectual disability, had NC scores within the typical range. Moreover, not all patients in this cohort carrying diagnostic CNVs had consistently poorer NC scores (Supplemental Table 1). This could be explained by different expressivity or penetrance of the GDs, age of onset of NC symptoms, and by the opposing effect of other known or unknown background predictors. For example, cognitive function in patients with 22q11.2 deletion or 16p11.2 syndrome are highly associated with parental IQ scores, indicating an effect of background genes and/or environment.^29,30 Consistent with these data, we also found that the effects of GDs on NC are significantly modified by the level of maternal education. Because maternal education integrates many complex dimensions, including genetic, socioeconomic, and environmental factors, these data suggest that the effect of genetic lesions on NC function may be potentially attenuated by postnatal factors, including targeted clinical or educational interventions.

In summary, our findings have significant implications for the clinical management of pediatric patients with CKD, providing a biological explanation for the association of CKD and NC impairment in a subset of patients. Chromosomal microarrays are recommended for many pediatric disorders such as congenital anomalies and/or NC impairment.³¹ These data suggest that microarrays can similarly provide a precise molecular diagnosis in children with CKD and also provide the opportunity for early intervention before NC complications associated with GDs become clinically evident. Future studies will be needed to confirm and extend these results by providing longitudinal measures of NC function and replicate findings in independent cohorts.

Concise Methods

CKiD Study and CNV Analysis

Eligibility criteria for enrollment in CKiD included age 1–16 years and eGFR of 30–90 ml/min per 1.73 m², and exclusion criteria included previous malignancy, transplantation, or dialysis within the previous 3 months, genetic syndromes involving the central nervous system, history of severe to profound intellectual disability, and a limited number of other conditions.¹ We studied 420 CKiD Study participants for which NC and CNV data were available. Analysis of CNV in 424 (419 unrelated) CKiD Study participants was previously reported. We defined GDs as CNV that fulfill criteria for known GDs or likely pathogenic CNV on the basis of the American College of Medical Genetics recommendations.³² Briefly, we defined known GDs when we detected 70% or greater CNV overlap with the coordinates of one out of 131 known, well characterized syndromes. Likely pathogenic CNVs were defined as: (1) CNV size of ≥500 kb, with frequency of 0.02% or less in 21,575 population controls; and (2) partial overlap with a known GD or overlap with likely pathogenic CNVs reported in the International Standards for Cytogenomic Arrays database or in our previous study on renal hypoplasia/dysplasia and/or gene content relevant to kidney development or pathology.¹²

Measures of NC and Psychosocial Function

Baseline measures of NC and psychosocial function were chosen to include broad assessments of multiple cognitive and psychosocial domains. NC outcomes included age-specific measures of intellectual functioning (Mullen Scales of Early Learning,³³ the Wechsler Preschool and Primary Scale of Intelligence-Revised,³⁴ or the Wechsler Abbreviated Scales of Intelligence³⁵; academic achievement using the Wechsler Individual Achievement Test-II-Abbreviated Total score³⁶; and the GEC score from the Behavior Rating Inventory of Executive Functions-Parent³⁷). Psychosocial outcomes included Externalizing Problems composite score, Internalizing Problems composite score, Behavioral Symptoms Index composite score, and Adaptive Skills composite scores measured from the parent-completed Behavior Assessment System for Children-II.³⁸ For the measures of intelligence and academic achievement, scores are reported in age-based standard scores (mean=100, SD=15), with higher scores reflecting a better performance. For the IQ variable, all IQ measures were combined into a single score for data analysis purposes. For the parent ratings of executive function and social-behavior, scores selected for this study are reported in age-based T-scores (mean=50, SD=10), with higher scores reflecting a poorer performance. We defined alert thresholds as one SD beyond the normative age-adjusted population means.

Statistical Analyses

Descriptive statistics for the demographic, NC, and health-status variables were generated. We applied the Student's t- and nonparametric Wilcoxon tests for genetic association analysis of continuous variables and the Fisher exact test for analysis of categorical variables. Multivariate linear regression analysis was performed to analyze association of CNV carrier status with NC outcomes, adjusting for genetically defined ancestry as well as modifiers prospectively selected because of their known association with NC measures (duration of disease, kidney function, bedside eGFR, low birth weight, maternal education level, and history of seizure disorder). All statistical analyses were carried out with Microsoft R Open 3.2.3 in RStudio 0.99.

Disclosures

None.