A longitudinal analysis of the effect of anemia on executive functions in children with mild to moderate chronic kidney disease

Abstract

Background:

Children with chronic kidney disease (CKD) are at risk for cognitive dysfunction. The aim of this study was to investigate associations between executive functions (EF), anemia, and iron deficiency.

Methods:

A total of 688 children > 6 years of age enrolled in the Chronic Kidney Disease in Children (CKiD) study who underwent evaluation for EF were included. Hemoglobin (Hgb) was characterized as low (1st–5th percentile) or very low (< 1st percentile) compared to normative values for age, sex, and race irrespective of erythropoiesis-stimulating agent (ESA) usage. Longitudinal analysis was conducted using consecutive visit pairs, with anemia status defined as new onset, resolved, or persistent. Linear mixed models with random intercept were used and adjusted for key covariates.

Results:

Anemia was present in 41% of children, and median Hgb was 11.8 gm/dl. New onset anemia was associated with lower digit span total score (− 0.75, 95% CI − 1.36, − 0.15, p = 0.01). Persistent anemia was associated with lower scores on color-word inhibition/switching (β = − 0.98; 95% CI − 1.78, − 0.18, p = 0.02). Errors of omission were significantly higher (worse) in those with persistent anemia (β = 2.67, 95% CI 0.18, 5.17, p = 0.04). Very low Hgb levels were significantly associated with lower color-word inhibition/switching scores (β = − 1.33, 95% CI − 2.16, − 0.51; p = 0.002). Anemia and low GFR were associated with lower category fluency scores compared to non-anemic subjects with higher GFR (β = − 1.09, 95% CI − 2.09, − 0.10, p = 0.03).

Conclusion:

The presence of anemia, in addition to its severity and duration in children with CKD, is associated with poorer scores on select measures of EF.

Graphical Abstract

Introduction

Children with chronic kidney disease (CKD) are at risk for neurocognitive difficulties, with early studies of neurocognitive outcomes reporting that infants with kidney failure were most likely to have poor neurodevelopmental outcomes [1, 2]. Advances in the care of children with CKD including improved nutrition, better dialysis, elimination of aluminum as a phosphate binder, and subsequent transplant have resulted in more favorable outcomes and a decreased prevalence of severe developmental delay. Whereas children with mild to moderate CKD do not demonstrate major neurocognitive deficits, data from the Chronic Kidney Disease in Children (CKiD) study has provided evidence of an increased risk for neurocognitive difficulties, especially in the area of executive function (EF) [3–5]. These patients are also at risk for lower academic achievement [6]. In a previously published meta-analysis of studies examining neurocognitive outcomes in children with CKD, deficits in neurocognitive functioning were more prevalent in children with CKD than in the general population, with executive functions (e.g., attention regulation, working memory) being more commonly affected [7].

Various CKD-related variables have been studied to evaluate potential mechanisms to explain the associations between cognitive deficits and CKD. Some evidence suggests that underlying risk factors may include longer duration of CKD, high blood pressure, blood pressure variability, presence of proteinuria, and severity of CKD[3, 8–11]. In addition, children with genetic disorders have significantly lower scores on measures of intelligence and executive function, as well as increased symptoms of anxiety and depression.[11]

Anemia is a common complication of children with CKD and worsens as CKD progresses in severity. The estimated prevalence of CKD-related anemia based on data from the North American Pediatric Renal Trials and Collaborative Studies (NAPRTCS) was 73% for children with CKD stage 3, 87% for stage 4, and > 93% for stage 5 [12]. Even in children prescribed an erythropoiesis-stimulating agent (ESA), over 40% of children with stage 5 CKD have a persistently low hemoglobin (Hgb) level, demonstrating a long-standing exposure to anemia in CKD patients[12, 13]. Previous noteworthy studies have demonstrated that the presence of both anemia and iron deficiency affects psychomotor development in infancy[14, 15]. Lozoff et al. showed that children who had moderate iron-deficiency anemia as infants had lower scores on tests of mental and motor functioning at 5 years of age despite correction of anemia[16].

A relationship between anemia and neurocognition has been detected in the CKD population. In adult patients with advanced CKD, the presence of anemia has been associated with poor neurocognitive outcome, with improvement noted after treatment with an ESA[17–19]. Marsh et al. demonstrated significantly improved neuropsychological test scores in adult patients in the domains of memory, attention, and learning once the participants’ average hematocrit increased from 23 to 36% with use of an ESA. Similarly, in a study of adult CKD and ESKD patients by Singh et al., neurophysiological events improved after treatment with an ESA [20]. In pediatrics, a study conducted by Slickers et al. [21] evaluated this relationship in three children with CKD and a Hgb level < 11.0 g/dL. Although the patients did demonstrate low mean scores for memory and attention, the small number of patients precluded any meaningful conclusions and emphasized the gap in knowledge pertaining to the possible association between these two common clinical complications in children with CKD.

In this study, neurocognitive data from the CKiD study were analyzed to determine the following: (1) if children with anemia and CKD demonstrate decreased performance on measures of EF compared to children with CKD and no anemia; (2) if the severity of anemia affects EF in children with CKD; and (3) if iron deficiency with or without anemia affects EF. We hypothesized that the presence of anemia and iron deficiency would be associated with poorer performance on measures of EF.

Methods

Participants

The CKiD study is a multicenter, longitudinal, observational cohort study of pediatric patients with mild to moderate CKD with participating centers from across North America. Participants had an estimated glomerular filtration rate (eGFR) of 30–90 ml/min/1.73m² at the time of study enrollment. Children with central nervous system dysfunction related to genetic syndromes and those with severe intellectual disabilities were excluded from participation in CKiD. Details of the study design, inclusion and exclusion criteria, and methods have been published previously [22]. Hemoglobin values were collected at baseline and at each annual visit; evaluation of iron status was performed at baseline and at each annual visit for a subset of participants. The current analysis included participants 6 years of age and older who contributed hemoglobin levels and participated in EF testing for at least two consecutive visits so that any change in anemia status and EF status over 1-year intervals could be observed. A total of 688 participants contributed a median follow-up of 2.1 years (IQR 0.0–4.0) creating an adequate database of longitudinal observations. The CKiD study protocol was approved by the Institutional Review Board at each participating center.

Measures

Anemia was defined as Hgb < 5^th percentile for age, sex and race according to the CDC guidelines, independent of the use of an ESA. Those with anemia were classified into two groups to capture severity of anemia: very low (Hgb <1^st percentile) and low (Hgb 1^st to <5^th percentile). Iron deficiency was defined as serum ferritin < 100 ng/ml and/or transferrin saturation less than 20%; the severity of iron deficiency was further classified into two groups: severe (transferrin saturation < 10% and/or serum ferritin < 50 ng/ml) and moderate (iron deficiency, but not meeting the severe criteria).

The CKiD study includes assessment of EF at 6 months after study entry and every 2 years thereafter [22]. Tests are standardized, age appropriate, and administered by trained personnel at each site. Tests were selected to measure different EF components including working memory, inhibitory control, set-shifting, problem solving, and attention regulation. The neurocognitive measures for this study included the Digit Span subtest from the Wechsler Intelligence Scale for Children-IV (WISC-IV) or Wechsler Adult Intelligence Scale-IV (WAIS-IV) depending on participant age; the Color-Word Interference, Design Fluency, Verbal Fluency, and Tower subtests from the Delis-Kaplan Executive Function System (D-KEFS); and the Errors of Omission and Errors of Commission variables from the Connors’ Continuous Performance Test-Second Edition (CPT-II). Personnel administering the assessments were not aware of the subjects’ anemia status. The study also included the Behavior Rating Inventory of Executive Function-Second Edition (BRIEF), a parent rating scale of children’s executive functioning that includes three index scores: Behavior Regulation Index (BRI), Metacognitive Index (MI), and a Global Executive Composite (GEC). Estimated GFR (eGFR) was determined using the published CKiD U25 eGFR equation[23]. Demographic and medical history data were collected using standardized forms. Variables of interest for this analysis included age, sex, glomerular versus non-glomerular kidney disorder, level of proteinuria, maternal education (high school or less, some college, or ≥16 years of education), hypertension (casual BP ≥90^th percentile, or self-reported diagnosis of hypertension plus antihypertensive medication use), abnormal birth history (low birth weight <2500 gm, premature birth or small for gestational age), and history of seizures.

Statistical Analyses

In order to study associations between anemia status and EF longitudinally, two different models were used for statistical analysis. Analyses that included anemia used visit pairs (two consecutive study visits), with the primary outcome being the EF measures. The presence or absence of anemia was characterized based on both visit X and X+1 1 year apart in the following manner: new onset (non-anemic at X, anemic at X+1); resolved (anemic at visit X, non-anemic at visit X+1); or persistent (anemic at both X and X+1).

In order to assess associations with severity of anemia and iron deficiency, these pairwise definitions could not be utilized, so a traditional model with separate predictors for low Hgb and very low Hgb as defined above was used. A secondary model included measures of iron deficiency as defined above. This was analyzed separately from the main models as there was considerably less iron data available.

Linear mixed models with random intercept were used for statistical analysis. Models were adjusted for variables of interest, which included age, sex, glomerular versus non glomerular kidney disorder, level of proteinuria, maternal education (high school or less, some college, or ≥ 16 years of education) which was included as a proxy of socioeconomic status, hypertension (casual BP ≥ 90th percentile, or self-reported diagnosis of hypertension plus antihypertensive medication use), abnormal birth history (low birth weight < 2500 gm, premature birth, or small for gestational age), and history of seizures. All analyses were performed in SAS 9.4 (SAS Institute Cary, NC); graphics were produced in R 4.0.3. P values < 0.05 were considered statistically significant.

Results

Sample characteristics

Descriptive statistics for the 688 participants included in the study are shown in Table 1; neurocognitive data was initially obtained 6 months after study entry. Participants had a median follow-up of 4 years and the median follow-up period to study the effect of anemia was 2.1 (IQR 0.0–4.0)years. Median age of the participants at study entry was 13 years; 61% were male. The median duration of CKD at study entry was 8 years and the median eGFR was 51 ml/min/1.73 m². Of the sample, 28% of the participants had a glomerular diagnosis and 53% had hypertension. Participants contributed a median of 2 visit pairs to the analysis and the median hemoglobin was 12.9 gm/dl. Forty-one percent of the participants met the definition of anemia, among which 8% had new onset, 25% had persistent, and 8% had resolved anemia, while 11% of all participants were being treated with an ESA.

Table 1.

Descriptive statistics at first available visit (n = 688)

Characteristic	Median [IQR] or n (%)
	No anemia at baseline (n = 462)	Anemia at baseline (n = 226)
Age	13.0 [9.2, 16.6]	14.3 [10.8, 17.0]
Male sex	279 (60%)	144 (64%)
African American race	112 (24%)	34 (15%)
Hispanic ethnicity	55 (12%)	40 (18%)
Maternal education
High school or less	160 (36%)	94 (42%)
Some college	129 (29%)	61 (27%)
College or more	157 (35%)	69 (31%)
Glomerular diagnosis	115 (25%)	79 (35%)
Years with CKD at study entry	7.8 [3.9, 11.7]	8.3 [4.1, 12.5]
U25eGFR, ml/min|1.73 m2	54 [43, 67]	35 [23, 51]
Urine protein/creatinine ratio	0.2 [0.1, 0.8]	0.6 [0.2, 1.7]
Anemia status
None	406 (88%)	0 (0%)
Acquired	56 (12%)	0 (0%)
Persistent	0 (0%)	169 (75%)
Resolved	0 (0%)	57 (25%)
Hemoglobin percentile*	44 [15, 75]	2 [0, 11]
ESA use	3 (1%)	74 (33%)
Hypertension**	242 (52%)	122 (54%)
Seizure history	57 (12%)	32 (14%)
Abnormal birth history***	150 (33%)	55 (25%)
Iron data (n = 283)
TSAT	25 [19, 32]	25 [15, 37]
Ferritin	41 [26, 83]	60 [28, 105]
Moderate iron deficiency	44 (24%)	30 (29%)
Severe iron deficiency	109 (60%)	48 (47%)

1.

^*Adjusted for age, sex, and race.

^**Casual systolic or diastolic BP ≥ 90th percentile, or self-reported diagnosis of hypertension plus use of antihypertensive medication.

^***Premature birth, low birth weight, or small for gestational age

Among participants on an ESA, 73% had anemia (57 of 78) had anemia, with 55% (43 of 78) having a very low Hgb. Follow-up data of anemia status for participants who were prescribed an ESA was available for 41 of 78 subjects, of which 78% (32 out of 41) remained anemic. Median Hgb levels were 11.8 gm/dl [IQR 11.4, 12.5] in the low Hgb group and 10.9 gm/dl [ IQR 10.3, 11.7] in the very low Hgb group. Among the subset of 297 participants with measured iron, 82% of the sample had iron deficiency, among which 55% had severe iron deficiency; median serum ferritin level was 46 ng/ml.

The relationship between anemia, iron deficiency and EF outcomes was assessed. Table 2 displays the results of adjusted linear mixed models on visit-pairs, using dynamic anemia status as the primary predictor. Errors of omission scores were significantly higher (i.e., worse) in subjects with persistent anemia (β=2.67, 95% CI 0.18, 5.17, p=0.04), but lower in those with resolved anemia compared to persistent anemia (β=−3.69, 95% CI −6.93, −0.46, p=0.03). Subjects with new onset anemia had significantly lower scores (i.e., worse) on digit span total (β=−0.75, 95% CI −1.36, −0.15, p= 0.01); however, the same effect was not seen in patients with persistent anemia (β=−0.37; 95% CI −0.92, 0.17, p= 0.18). Color-word inhibition/switching scores were significantly lower for patients with persistent anemia (β=−0.98; 95% CI −1.78, −0.18, p= 0.02) compared to non-anemic participants. There was no significant difference between groups in the other EF variables. Table 3 displays the results of adjusted linear mixed models on visits, with Hgb levels and ESA use taken concurrently with the EF measurements. Digit span total was lower in those with low Hgb levels (β=−0.53, 95% CI −1.06, −0.00, p=0.048), though this effect was marginal and was not similarly observed in those with very low Hgb (β=−0.26, 95% CI −0.81, 0.30, p=0.36). Color-word inhibition/switching was considerably lower in subjects with very low Hgb levels (β=−1.33, 95% CI −2.16, −0.51; P= 0.002). No other effects of low Hgb or ESA use were observed.

Table 2.

Longitudinal analysis of effect of dynamic anemia status

EF outcome	n used	Acquired anemia (vs. none)		Persistent anemia (vs. none)		Resolved anemia (vs. persistent)
		Estimate (95% CI)	p-value	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Errors of omission	1029	− 0.75 (− 3.86, 2.60)	0.64	2.67 (0.18, 5.17)	0.04	− 3.69 (− 6.93, − 0.46)	0.03
Errors of commission	1036	− 0.46 (− 2.73, 1.81)	0.69	0.02 (− 1.84, 1.89)	0.98	1.16 (− 1.21, 3.53)	0.34
Tower total achievement	938	− 0.19 (− 0.86, 0.47)	0.57	− 0.01 (− 0.54, 0.52)	0.97	− 0.09 (− 0.82, 0.63)	0.81
Digit span total	740	− 0.75 (− 1.36, − 0.15)	0.01	− 0.37 (− 0.92, 0.17)	0.18	0.03 (− 0.70, 0.76)	0.94
Category fluency	467	− 0.21 (− 1.38, 0.97)	0.73	− 0.38 (− 1.29, 0.53)	0.41	0.65 (− 0.59, 1.89)	0.30
Category switching	465	0.37 (− 0.82, 1.57)	0.54	0.03 (− 0.87, 0.93)	0.94	0.07 (− 1.20, 1.34)	0.91
Total switching	464	0.40 (− 0.69, 1.50)	0.47	0.04 (− 0.78, 0.86)	0.93	0.08 (− 1.08, 1.24)	0.90
Design fluency	461	− 0.08 (− 1.09, 0.93)	0.88	0.10 (− 0.68, 0.88)	0.80	− 0.21 (− 1.31, 0.88)	0.70
Design switching	460	0.49 (− 0.55, 1.53)	0.36	0.13 (− 0.66, 0.93)	0.74	− 0.21 (− 1.33, 0.90)	0.70
Color-word inhibition/switching	455	− 0.21 (− 1.24, 0.82)	0.68	− 0.98 (− 1.78, − 0.18)	0.02	0.45 (− 0.62, 1.53)	0.41
BRI	1304	0.69 (− 1.08, 2.46)	0.44	− 0.45 (− 2.06, 1.16)	0.59	− 0.47 (− 2.51, 1.57)	0.65
MI	1299	0.26 (− 1.55, 2.08)	0.78	− 0.57 (− 2.23, 1.08)	0.50	0.20 (− 1.89, 2.29)	0.85
GEC	1301	0.41 (− 1.39, 2.21)	0.65	− 0.64 (− 2.28, 1.00)	0.44	− 0.21 (− 2.28, 1.85)	0.84

^1.Models adjusted for age, sex, glomerular diagnosis, duration of CKD, eGFR, proteinuria level, maternal education, hypertension, abnormal birth history, and history of seizures

Table 3.

Longitudinal analysis of effect of anemia severity

EF outcome	n used	Low HGB (1st to < 5th pctl)		Very low HGB (< 1st pctl)		ESA use
		Estimate (95% CI)	p-value	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Errors of omission	1074	2.03 (− 0.52, 4.57)	0.12	1.12 (− 1.42, 3.66)	0.39	− 1.76 (− 5.14, 1.61)	0.31
Errors of commission	1081	− 0.68 (− 2.53, 1.18)	0.47	− 0.47 (− 2.35, 1.41)	0.63	0.27 (− 2.27, 2.81)	0.84
Tower total achievement	986	− 0.07 (− 0.63, 0.49)	0.82	− 0.07 (− 0.62, 0.47)	0.79	0.11 (− 0.67, 0.88)	0.78
Digit span total	767	− 0.53 (− 1.06, − 0.00)	0.048	− 0.26 (− 0.81, 0.30)	0.36	0.51 (− 0.26, 1.28)	0.19
Category fluency	488	− 0.50 (− 1.46, 0.47)	0.31	− 0.52 (− 1.49, 0.44)	0.29	0.37 (− 1.31, 2.04)	0.66
Category switching	486	0.12 (− 0.85, 1.10)	0.80	− 0.22 (− 1.16, 0.72)	0.64	0.94 (− 0.67, 2.55)	0.25
Total switching	485	0.29 (− 0.61, 1.19)	0.53	− 0.39 (− 1.25, 0.48)	0.37	0.19 (− 1.29, 1.66)	0.80
Design fluency	483	− 0.41 (− 1.25, 0.43)	0.33	0.39 (− 0.43, 1.22)	0.35	0.10 (− 1.25, 1.46)	0.88
Design switching	482	0.11 (− 0.74, 0.96)	0.80	0.34 (− 0.49, 1.17)	0.42	0.02 (− 1.35, 1.38)	0.98
Color-word inhibition/switching	476	0.06 (− 0.76, 0.88)	0.88	− 1.33 (− 2.16, − 0.51)	0.002	0.77 (− 0.67, 2.20)	0.29
BRI	1371	0.74 (− 0.77, 2.26)	0.33	− 0.09 (− 1.67, 1.49)	0.91	1.71 (− 0.41, 3.83)	0.11
MI	1366	0.41 (− 1.12, 1.93)	0.60	− 0.55 (− 2.16, 1.05)	0.50	0.76 (− 1.41, 2.93)	0.49
GEC	1368	0.59 (− 0.92, 2.10)	0.45	− 0.44 (− 2.04, 1.15)	0.58	1.23 (− 0.93, 3.39)	0.26

^1.Models adjusted for age, sex, glomerular diagnosis, duration of CKD, eGFR, proteinuria level, maternal education, hypertension, abnormal birth history, and history of seizures

In addition, we examined the effects of iron deficiency with or without anemia and eGFR on EF outcomes. There was no significant difference between the non-anemic, no iron deficiency reference group compared to anemic, iron deficient participants on any EF outcome. Anemia was also treated separately as an adjustment covariate and iron deficiency was classified as moderate or severe; no effects of iron deficiency were observed on EF outcomes. Finally, table 4 displays the effect of eGFR and presence or absence of anemia on EF outcomes. Having anemia and low GFR was associated with lower category fluency scores compared to non-anemic subjects with higher GFR (β=−1.09, 95% CI −2.09, −0.10, p=0.03). No other effects of anemia and GFR were observed.

Table 4.

Longitudinal analysis of effect of anemia and GFR category

EF outcome	n used	No anemia, GFR < 45		Anemia, GFR ≥ 45		Anemia, GFR < 45
		Estimate (95% CI)	p-value	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Errors of omission	1074	− 1.33 (− 3.70, 1.04)	0.27	0.04 (− 3.17, 3.25)	0.98	0.31 (− 2.23, 2.85)	0.81
Errors of commission	1081	− 0.84 (− 2.59, 0.91)	0.35	− 0.96 (− 3.33, 1.41)	0.43	− 1.37 (− 3.26, 0.53)	0.16
Tower total achievement	986	− 0.42 (− 0.94, 0.09)	0.11	− 0.29 (− 0.95, 0.37)	0.38	− 0.22 (− 0.77, 0.33)	0.43
Digit span total	767	− 0.22 (− 0.75, 0.32)	0.42	− 0.38 (− 1.05, 0.28)	0.26	− 0.38 (− 0.94, 0.18)	0.19
Category fluency	488	− 0.22 (− 1.18, 0.75)	0.66	− 0.07 (− 1.04, 0.90)	0.89	− 1.09 (− 2.09, − 0.10)	0.03
Category switching	486	0.18 (− 0.79, 1.16)	0.71	0.67 (− 0.30, 1.65)	0.17	− 0.67 (− 1.64, 0.30)	0.18
Total switching	485	0.26 (− 0.64, 1.16)	0.57	0.49 (− 0.41, 1.39)	0.28	− 0.28 (− 1.18, 0.61)	0.53
Design fluency	483	− 0.65 (− 1.50, 0.20)	0.13	− 0.32 (− 1.15, 0.52)	0.46	− 0.39 (− 1.25, 0.47)	0.37
Design switching	482	− 0.16 (− 1.02, 0.70)	0.71	− 0.02 (− 0.87, 0.83)	0.96	0.09 (− 0.78, 0.96)	0.83
Color-word inhibition/switching	476	0.27 (− 0.57, 1.10)	0.53	− 0.59 (− 1.43, 0.25)	0.17	− 0.32 (− 1.19, 0.55)	0.47
BRI	1373	− 0.52 (− 2.02, 0.98)	0.50	0.83 (− 1.13, 2.79)	0.41	− 0.48 (− 2.05, 1.10)	0.55
MI	1368	− 0.21 (− 1.73, 1.31)	0.79	0.68 (− 1.30, 2.66)	0.50	− 0.45 (− 2.06, 1.16)	0.58
GEC	1370	− 0.35 (− 1.86, 1.16)	0.65	0.77 (− 1.20, 2.74)	0.44	− 0.52 (− 2.12, 1.09)	0.53

^1.Models adjusted for age, sex, glomerular diagnosis, duration of CKD, proteinuria level, maternal education, hypertension, abnormal birth history, and history of seizures

Discussion

Anemia is a common complication of pediatric CKD and is associated with adverse clinical outcomes such as an increased rate of hospitalization and mortality, progression of cardiovascular disease risk factors, and lower health-related quality of life [24–26]. Despite significant advances in the care of children with CKD, studies continue to demonstrate that some children with CKD remain at risk for cognitive dysfunction and limited information is available pertaining to the possible contribution of anemia to this outcome. Our study assessed associations between anemia and targeted executive functions, those which are most likely to be affected in children with CKD. We used a large sample of pediatric patients with mild to moderate CKD to investigate this association longitudinally. We hypothesized that the presence of anemia would be associated with decreased EF performance in patients with mild to moderate CKD. This hypothesis is based on multiple studies showing that anemia affects cognition and behavioral function in healthy children and in adults with CKD and ESKD[14–19, 27–29].

Our study demonstrated that the presence of anemia impacted select measures of EF, while there was no association between anemia and the majority of EF measures, including parental ratings, during the follow-up period of 2.1 years. It should be pointed out that this time frame may be inadequate to completely assess the effect of anemia on EF, especially in the presence of mild anemia and slow CKD progression. It is also unclear why anemia was associated with worse outcomes on only three EF measures; however, we speculate that these direct measures of EF (as opposed to parent ratings) are assessing higher-order EF (i.e., working memory, set-shifting) and thus may be more vulnerable to the disruptive effects of anemia on cognitive abilities. The EF measures impacted by anemia consisted of a measure of cognitive flexibility, attention regulation, and response inhibition (D-KEFS Color-Word Inhibition/Switching); a measure of attention efficiency and verbal working memory (Wechsler Digit Span); and a measure of attention regulation (CPT–II Errors of Omission), with lower performances being associated with the presence of anemia. In addition, there was an association between the scores for Color-Word Inhibition/Switching and the severity of anemia, with children in the most severe category of anemia demonstrating disproportionately lower scores on this measure.

Anemia was persistent in over half the patients with anemia, but only affected two of the thirteen EF variables in the persistent anemia group. Persistent anemia was associated with a negative outcome on attention regulation, as indicated by the CPT–II Errors of Omission variable. High omission scores indicate that the participant is either easily distractible or has a sluggish response and therefore implies decreased attention regulation. In our study, a small but significant effect size was present for subjects with persistent anemia, showing a 2.67-point higher score T-score (worse performance) than those with no anemia. Then, we compared the scores of subjects with persistent anemia to those with resolved anemia. Again, a small but significant effect size was present showing that the performance of participants with resolved anemia was better than those with unresolved anemia as reflected by a 3.69-point lower T-score.

The results of our study are similar to those of Hooper et al. which previously identified the presence of anemia as a significant predictor of Developmental Level/Intelligence Quotient risk status in patients with mild to moderate CKD [30]. Choi et al. reported brain white matter volume loss and lower Full-Scale IQ scores proportional to the presence and severity of anemia in patients with sickle cell disease (SCD) as well as in non-SCD anemic controls [31]. Due to the overlap between the two groups, the authors suggested that the neurological injury patterns commonly found in patients with SCD are primarily due to their low hemoglobin levels rather than the abnormal red blood cells themselves; in turn, they speculated that similar changes may be extended to patients with chronic anemia from other causes..

Several studies have established that anemia and iron deficiency affect cognition and behavioral function in children[15, 16]. Prior studies have demonstrated long-term negative consequences of early-life iron deficiency, even in those cases that have not yet reached the point of anemia, making it likely that iron deficiency, and not simply hypoxia from iron deficiency anemia, is influential [32]. In our study, among the participants that had iron studies available, a significant proportion of them (82%) were iron-deficient, and despite receiving an ESA, three-quarters of the patients were anemic at follow-up visits. Still, there was not a significant difference in EF performance between the iron-deficient and the non-iron-deficient groups in our study. Similarly, the severity of iron deficiency also did not affect the EF performance. A possible explanation for the lack of effect of iron deficiency on EF performance in our sample may be that the median age of our participants was 13 years; in contrast, prior studies which have demonstrated a significant impact of iron deficiency primarily involved infants and preschool children with mild to moderate CKD who may be more susceptible to the exposure [30]. This implies that brain growth and associated cognitive development may be ongoing, yet progressing at a slower rate at this older age and thus less vulnerable to disruptions like those studied here. Given the mild spectrum of anemia and CKD captured in our cohort, there also is a possibility that the late and long-term effects of anemia and iron deficiency as mediators of EF have not yet occurred; thus, it will be important to follow the neurocognitive functioning of these patients, especially those with progressive worsening of their kidney function.

The Kidney Disease Improving Global Outcomes (KDIGO) clinical practice guidelines for anemia use World Health Organization (WHO) age-specific Hgb values to define the level at which an evaluation for the cause of anemia in children with CKD should be initiated[33]. As Hgb levels vary by age and sex, application of adult normative Hgb values underestimates the prevalence of anemia in pediatric CKD patients. While there is no well-defined Hgb level at which ESA therapy should be initiated in children, therapy is usually implemented in clinical practice for Hgb < 10 gm/dl and titrated to a goal of Hgb 11–12 gm/dl. This Hgb goal for ESA therapy is recommended based on multiple clinical trials conducted in adults which, in turn, have raised concerns regarding the safety of ESA therapy in patients with high Hgb levels, because of the higher risk of adverse cardiovascular outcomes and stroke risk in those patients [34–36]. However, in a retrospective study of pediatric hemodialysis patients, a Hgb level > 12 gm/dl was not associated with increased cardiovascular visits, mortality, or all-cause and cardiovascular-related hospitalizations, likely due, in part, to a lower prevalence of comorbidities in these patients when compared to adults with stage 5 CKD[37].

In our study, anemia was defined as a Hgb value less than 5^th percentile based on age, sex, and race, with a median Hgb value of 11.8 gm/dl for the anemic subsample. Based on current clinical practice, ESA therapy would not be initiated at these levels. However, our study demonstrated lower scores in select executive functions at these levels of Hgb, raising the important question of whether a higher and potentially normal age-related hemoglobin level achieved with the use of ESA therapy would be more desirable in children with CKD. Since our study did not assess the relationship between cardiovascular outcomes and Hgb levels, additional studies are clearly needed to better define a target hemoglobin level in pediatric patients with CKD with particular attention to both cardiovascular and neurocognitive outcomes..

Our main study strengths include our capacity to study a large, well-defined sample of children with mild to moderate CKD followed longitudinally and the use of well-validated measures of EF in accordance with a multidimensional model. Another strength of the study is our ability to adjust for potential covariates which can influence cognitive outcomes such as the presence of hypertension, reduction in kidney function, prematurity, seizures, and maternal education. Our study has several limitations as well. As noted previously, the mean duration of follow-up of 2.1 years may be an inadequate time frame to assess the effect of anemia on EF, especially in the presence of mild anemia and slow CKD progression. The higher median age of our cohort may have precluded our ability to discern any relationship between anemia and cognitive outcome as some cognitive milestones may have already been achieved in a substantial percentage of the cohort. Also, a majority of the patients had mild anemia where the median Hgb level would be considered in the accepted goal range for children with CKD. In addition, follow-up anemia data were only available for 48 patients, and we did not consider the potential variability of Hgb status in between the two consecutive study visits. Finally, due to the observational nature of this study, we cannot infer causality.

In conclusion, we report that the presence of anemia was associated with adverse outcomes of select higher-order executive functions in children with mild to moderate CKD, specifically working memory, inhibitory control/set-shifting, and attention regulation. These findings were reflected in performance-based measures as opposed to parent ratings of EF. Further studies with longer follow-up, particularly for those children with worsening kidney function, are needed to advance our understanding of the role of anemia as a potential modifiable risk factor for variable and/or abnormal cognitive performance. In addition, a well-defined treatment goal for Hgb range in pediatric patients with CKD, taking into consideration a variety of clinical outcomes including neurocognition, needs to be further examined.