Younger age and early puberty are associated with cognitive function decline in children with Cushing disease.

Abstract

Objective:

To investigate the effect of hypercortisolism on the developing brain we performed clinical, cognitive, and psychological evaluation of children with Cushing disease (CD) at diagnosis and one year after remission.

Study design:

Prospective study of 41 children with CD. Children completed diverse sets of cognitive measures prior to and one-year after remission. Neuropsychological evaluation included the Wechsler Intelligence Scale, California Verbal Learning Test, Trail Making Test, the combined subset scores of Wide Range Achievement Test and Woodcock- Johnson Psychoeducational Battery Test of Achievement, and the Behavioral Assessment System for Children.

Results

Comprehensive cognitive evaluations at baseline and one- year following cure revealed significant decline mostly in nonverbal skills. Decrements occurred in most of the various indices that measure all aspects of cognitive function and younger age and early pubertal stage largely contributed to most of this decline. Results indicated that age at baseline was associated with positive regression weights for changes in scores for verbal, performance, and full IQ scores and for subtests arithmetic, picture completion, coding, block design, scores; indicating that older age at baseline was associated with less of a deterioration in cognitive scores from pre- to post-treatment.

Conclusion

Our findings suggest that chronic glucocorticoid excess and accompanying secondary hormonal imbalances followed by eucortisolemia have detrimental effects on cognitive function in the developing brain; younger age and pubertal stage are risk factors for increased vulnerability, while older adolescents have cognitive vulnerabilities like that of adult patients affected with CD.

Summary:

Chronic endogenous glucocorticoid excess and accompanying secondary hormonal imbalances followed by eucortisolemia has detrimental effects on cognitive function in the developing brain; younger age and pubertal stage are risk factors for increased vulnerability, while older adolescents have cognitive vulnerabilities like that of adult patients affected with Cushing disease.

Introduction

Endogenous Cushing’s syndrome (CS) is a rare endocrine disorder characterizd by excessive levels of the glucocorticoid hormone cortisol. Pediatric patients with CS typically display weight gain, impaired growth, fatigue, delayed sexual development, skin changes, hypertension, glucose intolerance, structural and functional changes in the brain, changes in mood and cognitive function ¹. Pediatric patients with endogenous CS present a unique natural model to study the effects of prolonged hypercortisolemia on cognitive function during a sensitive period of neural development. Studies of adults with active CS report structural, functional, and biochemical abnormalities in the brain that improve but do not completely resolve after remission of CS ^2-4. Prior reports of children treated for endogenous CS noted a significant decline in cognitive functions one year after transsphenoidal surgery (TSS) associated with changes in neural structures (i.e. functional alterations in the hippocampus and amygdala, cerebral atrophy) important to cognitive function^5-7. However, these studies examined small numbers of patients with CS and reported summary intelligence quotient (IQ) scores rather than by subsets, therefore it is unclear which specific aspects of cognition are vulnerable to change. We hypothesized that memory tasks may be negatively impacted in children and adolescents with CS, therefore we included a variety of neuropsychiatric tests designed to assess various aspects of memory (short-, long- term-, and working memory) and general academic skills.

In the present study, we analyzed IQ summary and subset scores, as well as general measures of cognitive function with 41 pediatric patients with pituitary tumors causing CS or Cushing disease (CD) before and one year after remission (TSS). We investigated age and Tanner stage at initial presentation as covariates and the effect on cognitive function measures. Also, since CD is associated with elevation of the adrenal steroid hormone dehydroepiandrosterone (DHEA), which is known to have anti-glucocorticoid properties in the liver, skin, and brain ⁸, we investigated the relationship between DHEA levels and cognitive measures prior to and after remission of CS.

Materials and methods

Subjects and Protocol

Forty-one children (seventeen females, twenty-four males, 12.3 ± 3 years; average duration of CS 2.8 ± 1.4 years (based on clinical symptoms and/or attenuation of growth velocity)) referred to National Institutes of Health Clinical Center from 1994-2007 who were diagnosed with CD and underwent TSS for removal of a pituitary adenoma and were in remission at 1-year were studied. The children completed diverse sets of cognitive measures prior to and one-year after transsphenoidal surgery (average time of follow-up post-surgery: 15 ± 9.9 months). The study was approved by the institutional review board at the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD). We obtained written informed consent from the parent(s) and children over the age of 7 years gave their assent.

Endocrine, clinical, and cognitive assessments were performed prior to and one-year post TSS. Tanner stage was assigned according to breast development in girls and testicular volume in boys. Only children with biochemical evidence consistent with remission of CS were included in the final analysis. At baseline, CS was confirmed by biochemical testing as described by Batista ⁹ including the confirmation of elevated urine free cortisol and lack of diurnal variation on serum cortisol levels. At one-year follow-up post-TSS, biochemical and evidence of remission was shown by the following: urine free cortisol within normal range, improved linear growth (if applicable), attenuation of weight gain or BMI and/or weight loss. As per these criteria, 29 subjects had full recovery of HPA function (serum cortisol response of >18ug to ACTH), 2 subjects had diagnosis of panhypopituitarism on hormone replacement(s) and 10 subjects had incomplete recovery of HPA function and were on a tapered dose schedule of GC replacement. Five subjects had secondary hypothyroidism in the immediate post-surgical period and were placed on thyroid hormone replacement. Clinical characteristics of the participants are listed in Table 1.

TABLE 1.

Clinical Characteristics of children with Cushing disease

Age at baseline (pre-surgery) (years)	12.3 ± 3
Female sex, n (%)	17 (41.5)
Duration of Illness (years)	2.8 ± 1.4
Time between baseline and follow up (months)	15 ± 10
Midnight serum cortisol (mcg/dL)	15.9 ± 10.4
Ht SDU (Z-score)	−1.1 ± 1
BMI (Z-Score)	2.1 ± 0.71
24 hour-Urine Free Cortisol (mcg/24hr)	724 ± 2507
Tanner Stage	2 ± 1

Plus-minus values are means ± SD.

Abbreviations: Height standard deviation unit (Ht SDU), Body mass index (BMI)

UFC reference value: 17-129mcg/24hr

Subjects who did not complete baseline and one-year endocrine or cognitive assessments were not included, thus the design of the study was such that each child served as his/her own control. We excluded 6 subjects who were unable to be scheduled for baseline testing and 10 subjects who unable to be scheduled for follow-up testing due to scheduling issues (time constraints of subject) or unavailability of the neuropsychologist.

In the current study, all children were evaluated by a neuropsychologist (E.A.W.) prior to and one year after surgical treatment of CD. Neuropsychological evaluation included: the Wechsler Intelligence Scale for Children (subjects under16 years) or the Wechsler Adult Intelligence Scale (subjects 16 years or older), California Verbal Learning Test- (Child or Adult version), Trail Making Test, and combined certain subset scores of Wide Range Achievement Test and Woodcock- Johnson Psychoeducational Battery Revised: Test of Achievement, and the Behavioral Assessment System for Children (BASC).

For children 6 to 16 years, Wechsler Intelligence Scale for Children (WISC) was used and the Wechsler Adult Intelligence Scale (WAIS) for children older than 16 years. The test calculates intelligence quotient (IQ) through measurements of verbal and performance IQ. Verbal IQ is divided into verbal comprehension index (VCI) and working memory index (WMI), while performance IQ is divided into perceptual organizing index (POI) and processing speed index (PSI). Each of the four indices are then subdivided: VCI into vocabulary, similarities, comprehension, information, and word reasoning; WMI into digit span, letter-number sequencing, and arithmetic; POI into picture concept, block design, matrix reasoning, and picture completion; and PSI into coding, symbol search, and cancellation. The IQ measures are standardized using mean of 100, and standard deviation of 15. The scores were standardized to be representative of the U.S English-speaking population ¹⁰.

In our study, some subtests (comprehension, word reasoning, letter-number sequencing, matrix reasoning, symbol search, and cancellation) were excluded due to the small number of the children who completed both the baseline and one-year follow up of the subtest. Only those subtests with 20 pairs of baselines and first year follow up data (vocabulary, similarities, information, arithmetic, digit span, picture completion, block coding, picture arrangement, and block design) were included in the analysis. The subtests scores are standardized using a mean of 10 and a standard deviation of 3.

California Verbal Learning Test: Second Edition (CVLT-II; adult or child version) assesses verbal short- and long-term memory and learning. It consists of measurements for short delay free call, recognition discrimination index, long delay free recall, semantic clustering, and retroactive interference. The standard duration of the test is 30 minutes of testing, with 30 minutes of delay. The neuropsychologist for our study calculated the total T-score (M= 50, SD=10) and long delay free recall (M=0, SD=1) as the two most important factors for analysis ¹¹.

Trail Making Test A and B (TMT) are associated with numerous cognitive functions. TMT performance is a measure cognitive control, and TMT-B especially is correlated with attentional control during task switches ¹². Scores are derived from age-based norms.

Wide Range Achievement Test (WRAT) assesses academic skills. For this study we used: Word Reading, Spelling, and Math Computation. The scores are standardized based on mean of 100 and standard deviation of 15 using normative age or grade groups ¹³. Both the WRAT and Woodcock Johnson are measures of academic skills, however each test contains subtests that focus on specific skills. Since most subjects did not complete both the WRAT and WJ at pre- and post- time points (due to scheduling/time constraints, or subject fatigue) the missing values of either test were replaced by the other for each subject. The mean and the standard deviation of Reading and Math Computation and Letter-Word Identification and Calculations sections of Woodcock Johnson was the same, and thus, could be used in place of each other. Woodcock Johnson (WJ-IV) Tests of Achievement evaluate individual’s academic, cognitive, and oral skills. WJ-IV consisted of numerous batteries of tests. For our study, Letter-Word Identification and Calculation tests were used. Letter-Word Identification requires participant to recognize and pronounce letters and words. Calculations presents the participant with computational math problems requiring addition, subtraction, division, and multiplication ^{14, 15}.

The Behavioral Assessment System for Children (BASC) Parent Form was administered to assess for changes in the child’s behavior or mood (i.e., externalizing, or internalizing problems, adaptive skills, and school problems). The BASC is an assessment tool frequently used in clinical practice and research settings, that has high internal consistency, test-retest reliability, inter-rater reliability, is norm-referenced¹⁶.

Statistical analysis

Descriptive statistics were calculated for continuous variables IQ scores and all other data are reported as means, standard deviations, delta change scores. Height standard deviation units (Ht SDU), body mass index (BMI) and BMI SD units (BMISDU) were determined using anthropometric reference data for US children. Scores of CD subjects at baseline and 1-year postoperatively were compared by the paired two-sided Student’s t-test. Reliable change index (RCI) was computed to evaluate whether there were reliable changes from pre- to post-tests.

Prior studies have reported that children may experience cognitive effects of endogenous hypercortisolemia that are different from adults with CS; therefore, ANOVA with age and Tanner stage was used as covariates to investigate if there was an association with change in IQ scores. We also examined the relationship between DHEA and change in IQ scores from baseline to follow-up.

Statistical significance was accepted for P-values equal or below 0.05 with Bonferroni correction. Analyses were done using SAS version 9.4.

Results

Cognitive function

Children with CD had average to above-average pro-rated Wechsler Full Scale (FIQ) (N= 39), Performance (PIQ) (N= 39), and Verbal IQ (VIQ) (N= 40) scores prior to and one-year after surgery and a return to eucortisolism (Table 2). ANOVA analysis for the group of subjects as a whole revealed significant decline in scores for verbal IQ, performance IQ, full IQ scores and for subscales: information (N= 36), similarities (N= 36), arithmetic (N= 34), vocabulary (N= 37), block design (N= 34), and coding (N= 34) from baseline to first year follow up (Table 2). No differences were found for digit span (N= 33), picture completion (N= 34), picture arrangement (N= 34), or Trail making test (although smaller number of subjects completed Trail making test pre- and post) (N= 12) scores from baseline and 1-year follow-up. For WRAT/Woodcock Johnson, there was a significant decline in scores for both subtests (Reading/Letter Word Identification, and Math/Calculation) (N= 30). For CVLT-II (N= 26) and TMT (N= 12) there was no difference in delta change of scores (Table 2). We performed Reliable Change Index (RCI) analysis and the results were consistent with ANOVA; the RCI was >1.96 for all tests except digit span, picture completion, picture arrangement, TMT, and CVLT.

TABLE 2.

Comparison of cognitive scores of children with Cushing disease at baseline and one-year post remission

		All patients (n=41)	Patients <14 (n=25)	Patients ≥14 (n=16)
WISC/WASI
Information
	Baseline	11.72 ± 3.4	13.0 ± 2.7	9.87 ± 3.4
	Remission	10.92 ± 3.0*	11.82 ± 2.9	9.5 ± 2.6
Similarities
	Baseline	12.27 ± 3.6	14.45 ± 1.8	9.07 ± 3.1
	Remission	11.08 ± 3.2*	12.55 ± 2.4	8.79 ± 2.9
Arithmetic
	Baseline	11.51 ± 3.6	13.17 ± 2.5	8.79 ± 3.6
	Remission	10.5 ± 3.4*	11.14 ± 3.2	9.6 ± 3.5
Vocabulary
	Baseline	12.37 ± 3.5	14.27 ± 2.0	9.75 ± 3.5
	Remission	11.5 ± 3.1*	12.82 ± 2.6	9.43 ± 2.8
Digit Span
	Baseline	9.75 ± 2.7	10.9 ± 2.1	8.13 ± 2.7
	Remission	9.66 ± 2.8	10.43 ± 2.7	8.5 ± 2.6
Picture Completion
	Baseline	10.51 ± 2.8	11.55 ± 2.6	9.0 ± 2.5
	Remission	10.19 ± 2.8*	9.9 ± 2.5*	10.6 ± 3.3
Coding
	Baseline	12.92 ± 3.8	14.14 ± 3.1	11.0 ± 4.1
	Remission	10.1 ± 3.0*	10.23 ± 3.0	9.87 ± 3.1
Picture Arrangement
	Baseline	10.22 ± 3.2	11.45 ± 2.8	7.5 ± 2.2
	Remission	10.17 ± 2.8*	10.68 ± 2.8	9.2 ± 2.6
Block Design
	Baseline	12.31 ± 3.5	14.14 ± 2.8	9.43 ± 2.6
	Remission	9.78 ± 2.9*	10.59 ± 2.4	8.6 ± 3.2
Verbal IQ
	Baseline	110.22 ± 18.0	120.86 ± 9.5	96.25 ± 17.1
	Remission	104.03 ± 14.3*	110.78 ± 11.5*	93.67 ± 11.9
Performance IQ
	Baseline	110.14 ± 16.1	119.33 ± 8.52	96.36 ± 14.9
	Remission	101.74 ± 14.4*	104.83 ± 12.3*	97.00 ± 16.6
Full IQ
	Baseline	111.92 ± 17.9	122.27 ± 7.5	95.64 ± 17.5
	Remission	103.11 ± 13.8*	108.61 ± 10.5*	94.67 ± 14.3
CVLT-II
Total T- score	Baseline	51.75 ± 15.9	54.0 ± 13.9	48.46 ± 18.6
	Remission	55.79 ± 11.0	56.82 ± 11	53.73 ± 11.1
Long-delay recall	Baseline	0.37 ± 0.9	0.75 ± 0.9	−0.15 ± 0.7
	Remission	0.53 ± 0.9	0.77 ± 0.9	0.05 ± 0.9
Trail Making
Trail A	Baseline	16.44 ± 7.2	11.6 ± 4.1	22.5 ± 5.6
	Remission	19.0 ± 6.6	17.3 ± 6.1	21.83 ± 6.8
Trail B	Baseline	42.4 ± 21	31.1 ± 14.1	56.5 ± 22
	Remission	42.5 ± 17.1	31.8 ± 13.5	57.7 ± 6.9
WRAT/WJ
Read/Letter ID	Baseline	106 ± 23	114 ± 14	93 ± 30
	Remission	99 ± 27	108 ± 16	89 ± 37
Math/Calculation	Baseline	106 ± 20	112 ± 13	95 ± 28
	Remission	97 ± 27	101 ± 26	94 ± 26

Plus-minus values are means ± SD.

^*Significant change in scores from baseline to follow-up, at p=0.05

Abbreviations: Wechsler Intelligence Scale for Children/Weschler Adult Intelligence Scale (WISC/WASI), California Verbal Learning Test (CVLT), Wide Range Achievement Test/ Woodcock Johnson Psychoeducational Battery test of Achievement (WRAT/WJ)

Regression analysis was used to investigate whether age at baseline was a predictor of change in cognitive scores. Results indicated that age at baseline was associated with positive regression weights for changes in scores for VIQ, PIQ, and FIQ and for subtests arithmetic, picture completion, coding, block design, scores (Table 3), indicating that older age at baseline was associated with less of a deterioration in cognitive scores from pre- to post-treatment. Age was not associated with delta change for CVLT, TMT, or WRAT/Woodcock Johnson scores.

Table 3.

The effect of age (continuous) on the change in cognitive test scores

		N	Mean of Change	SD of Change	t	p
WISC/WASI
	Information	36	0.97	0.71	1.36	0.18
	Similarities	35	1.67	0.87	1.91	0.07
	Arithmetic	34	2.65	0.79	3.35	0.002*
	Vocabulary	37	1.04	0.99	1.05	0.30
	Digit Span	33	0.35	0.75	0.47	0.64
	Picture Completion	34	3.56	1.04	3.41	0.002*
	Coding	34	3.19	0.93	3.45	0.002*
	Picture Arrangement	27	2.44	1.26	1.94	0.06
	Block Design	34	2.83	0.84	3.36	0.002*
	Verbal IQ	40	8.92	2.56	3.48	0.001*
	Performance IQ	39	17.58	3.55	4.95	0.001*
	Full IQ	39	14.69	2.58	5.69	0.001*
CVLT-II
	Total T Score	26	5.03	7.09	0.71	0.49
	Long Delay Free Recall	25	−0.07	0.39	−0.17	0.87
TMT
	Trail A	12	−4.33	4.46	−0.97	0.35
	Trail B	12	10.17	8.31	1.22	0.25
WRAT/Woodcock Johnson
	Reading/Letter Word Identification	30	−9.72	10.68	−0.91	0.37
	Math/ Calculation	30	4.33	10.56	0.41	0.69

Based on the results of the regression analysis, subjects were grouped by age at baseline, those younger than 13.9 years (N= 25) or those 14 years and older (N= 16), to investigate which cognitive functions were the most vulnerable to change. In the younger age group, there was a significant decrease in scores from baseline to follow-up for all summary IQ scores and for the subset scores except for picture arrangement; interestingly, scores of the older group of children showed no significant change. For CVLT-II and TMT there was no difference in delta change of scores for either age group. For WRAT/Woodcock Johnson, there was a significant difference in delta scores of younger subjects for Reading/Letter Word Identification, but not for Math/Calculation; however, older subjects had no significant change in scores (Table 3A, 3B). RCI analysis was consistent with the results of regression; for subjects <14yo the RCI was >1.96 for all subtests except digit span, picture arrangement, CVLT, Trail making B, and WRAT/WJ math calculation. For subjects >14yo, RCI index was <1.96 for all subtests except Trail making A test.

Table 3A.

Changes in cognitive scores from baseline to 1-yr follow up (age<14)

Subtest		N	Mean of Change	SD of Change	t	p
WISC/WASI
	Information	22	−1.18	2.26	−2.45	0.02*
	Similarities	21	−1.90	2.45	−3.57	0.002*
	Arithmetic	21	−1.95	2.18	−4.11	0.001*
	Vocabulary	22	−1.91	3.16	−2.83	0.01*
	Digit Span	19	−0.21	2.1	−0.44	0.67
	Picture Completion	20	−1.85	3.07	−2.7	0.01*
	Coding	21	−3.81	2.60	−6.71	0.001*
	Picture Arrangement	18	−0.56	3.31	−0.71	0.49
	Block Design	21	−3.52	2.50	−6.45	0.001*
	Verbal IQ	24	−9.67	7.74	−6.12	0.001*
	Performance IQ	24	−16.38	10.69	−7.50	0.001*
	Full IQ	24	−14.29	7.52	−9.31	0.001*
CVLT-II
	Total T Score	16	1.88	16.87	0.45	0.66
	Long Delay Free Recall	15	0.07	0.88	0.29	0.77
TMT
	Trail A	6	5.83	5.74	2.49	0.06
	Trail B	6	0.17	5.71	0.07	0.95
WRAT/Woodcock Johnson
	Reading/Letter Word Identification	19	−8.37	16.48	−2.21	0.04*
	Math/ Calculation	20	−12.53	30.47	−1.84	0.08

Table 3B.

Changes in cognitive scores from baseline to 1-yr follow up (age>=14)

		N	Mean of Change	SD of Change	t	p
WISC/WASI
	Information	14	−0.21	1.76	−0.46	0.66
	Similarities	14	−0.24	2.66	−0.33	0.74
	Arithmetic	13	0.69	2.32	1.07	0.30
	Vocabulary	15	−0.87	2.64	−1.27	0.23
	Digit Span	14	0.14	2.18	0.25	0.81
	Picture Completion	14	1.71	2.89	2.22	0.045*
	Coding	13	−0.62	2.66	−0.83	0.42
	Picture Arrangement	9	1.89	2.57	2.20	0.06
	Block Design	13	−0.69	2.18	−1.15	0.27
	Verbal IQ	16	−0.75	8.21	−0.37	0.72
	Performance IQ	15	1.20	10.94	0.43	0.68
	Full IQ	15	0.40	8.35	0.19	0.86
CVLT-II
	Total T Score	10	6.90	18.72	1.17	0.27
	Long Delay Free Recall	10	0.00	1.05	0.00	1.0
TMT
	Trail A	6	1.50	9.29	0.40	0.71
	Trail B	6	10.33	19.54	1.30	0.25
WRAT/Woodcock Johnson
	Reading/Letter Word Identification	11	−18.09	41.64	−1.44	0.18
	Math/ Calculation	10	−8.20	18.80	−1.38	0.20

Since there was an association between age and change in cognitive function scores from pre- to post-treatment, we investigated the relationship between Tanner stage and cognitive scores change. At the baseline assessment, the average Tanner stage score for children less than 14 years age was 1.8 +/− 0.14 and the average Tanner stage for children 14 yrs. and older was 3.5 +/− 0.33. We classified subjects by early or mid-to-late puberty as follows: early puberty (Tanner score of I, II) and mid-late puberty (Tanner score of III, IV, V). Regression analysis of change in scores from baseline to follow-up with age as continuous variable adjusted for Tanner stage group, showed a significant change in scores for VIQ (p<0.02), FIQ (p< 0.001), PIQ (p<0.002), and for subscale scores: arithmetic (p<0.02), coding (p<0.001), and block design (P<0.01) (Table 3). A significant association was found for age adjusted for Tanner stage group for long-delay recall scores for CVLT, but not for delta total T scores or WRAT/Woodcock Johnson scores (Table 4).

Table 4.

The effect of age (continuous) on the change in cognitive scores adjusted for Tanner stage

Subtest		N	Beta	SE of Beta	t	p
WISC/WASI
	Information	36	0.14	0.14	0.99	0.33
	Similarities	35	0.16	0.17	0.92	0.36
	Arithmetic	34	0.42	0.16	2.56	0.02*
	Vocabulary	37	−0.03	0.19	−0.15	0.88
	Digit Span	33	0.08	0.15	0.54	0.59
	Picture Completion	34	0.37	0.20	1.86	0.07*
	Coding	34	0.62	0.17	3.56	0.001*
	Picture Arrangement	27	0.04	0.2	0.15	0.88
	Block Design	34	0.39	0.15	2.60	0.01*
	Verbal IQ	40	1.30	0.52	2.48	0.02*
	Performance IQ	39	2.35	0.70	3.37	0.002*
	Full IQ	39	2.00	0.53	3.81	0.001*
CVLT-II
	Total T Score	26	0.33	1.49	0.99	0.33
	Long Delay Free Recall	25	0.15	0.07	2.10	0.05*
TMT
	Trail A	12	−0.93	0.95	−0.98	0.35
	Trail B	12	1.31	1.96	0.67	0.52
WRAT/Woodcock Johnson
	Reading/Letter Word Identification	30	−0.56	2.24	−0.25	0.80
	Math/ Calculation	30	−0.06	2.13	−0.03	0.98

ANOVA analysis demonstrated DHEA levels at baseline were significantly lower in the younger group compared to subjects >14yrs age (490 ng/ml +/− 120 vs. 1123ng +/− 180, respectively, p <0.01) and that higher baseline DHEA levels correlated with less change in PIQ and FIQ scores at follow-up (PIQ: R2= 0.377, p=0.03; FIQ: R2= 0.346, p=0.04).

Behavior and school

Prior to surgical treatment, 12 subjects met the criteria for ‘at risk’ for depression or anxiety and three subjects had diagnosis of anxiety or depression and were prescribed pharmacologic treatment. One-year post remission of CS, 10 subjects showed signs of depression or anxiety disorder and 2 had a diagnosis of anxiety or depression and were prescribed pharmacologic treatment. Seven subjects endorsed suicidal ideations (and one subject committed suicide several years later). At baseline, all subjects reported average to above-average grades in school: 20 patients described themselves as ‘A’ students and taking accelerated classes. At one-year follow-up, 17 of ‘A’ students reported grades were ‘B to C’, with a decrease in honors classes, and 7 students with ‘A/B’ grades at baseline reported a decrease in grades, 7 subjects reported request for individual education plan (IEP) due to change in school performance, 17 subjects reported no change in school performance; no subject reported an improvement in school performance.

Discussion

This study provides important details regarding the cognitive deterioration experienced by pubertal children with CD following cure. Comprehensive cognitive evaluations at baseline and one year following cure revealed significant decline mostly in nonverbal skills. Results of this study are consistent with our previous findings (with a smaller number of patients) that pediatric patients with endogenous CS had average to above average intelligence, with VIQ, PIQ, and FIQ scores at or above the mean prior to and one-year post surgery, with significant deterioration in cognitive function after returning to eucortisolemia ^{5, 7}. Importantly, results of this study identify that younger children with CD who are in early puberty, are more vulnerable to long-term deleterious effects of hypercortisolemia on cognitive function than older children. Although the scores on IQ measures remained within average range, the decrement in scores of younger patients had a significant impact on academic performance, which was a major stressor for the children and parent(s).

For the overall group of subjects, there was a decline in scores between baseline and one- year-follow up for all the subtests of WISC/WAIS and WRAT/Woodcock Johnson tests. Decrements occurred in most of the various indices that measure all aspects of cognitive function and younger age and early pubertal stage largely contributed to most of this decline. The subtests of the WISC/WAIS that were most affected in younger subjects involved cognitive functions related to nonverbal concepts/ learning, visual attention, and coordination, working memory, and reasoning skills.

Our results suggest that the cognitive aspects least affected by exposure to endogenous hypercortisolemia may be verbal/nonverbal measure of visual organization and auditory memory as indicated by a non-significant decline of picture arrangement and digit span (respectively). For subjects 14yrs and older there was no significant change in cognitive function measures, except improvement was noted in the picture completion subscale. These results suggest that that after a return to eucortisolemia, the response of older adolescents is like adult CS patients¹⁷.

Dehydroepiandrosterone (DHEA) has a critical role in human brain development and cognition likely due to effects of this steroid in enhancing brain plasticity and DHEA-precursors, such as progesterone and allopregnanolone have critical roles in neuroprotection¹⁸. Importantly, we found that although DHEA levels were above the average range in all of the subjects at baseline, children in early stages of pubertal development had significantly lower DHEA compared to children with mid-to late puberty; higher DHEA levels were associated with less of a change in overall IQ measures. Puberty is a sensitive developmental period that is associated with rapid structural and functional changes in the brain, for example, cortical white matter increases, while cerebral cortex decreases ¹⁹ and these structural changes are associated with an alteration in the hormonal milieu. Steroid hormones, primarily glucocorticoids, are known to affect the morphology and function of the central nervous system suggesting that the high levels of GC receptors in brain areas involved in higher cognitive processes are target sites for the deleterious effects of hypercortisolemia. Glucocorticoids inhibit neurogenesis and glucose utilization by the brain and occupation of glucocorticoid and mineralocorticoid receptors by supraphysiological levels of glucocorticoids is associated with decreased cell excitability and a reversible phase of atrophy of neurons ²⁰. DHEA has anti-glucocorticoid properties and there is ample evidence to support its role in axonal/dendritic growth and neuroprotection that is independent of sex ^{21, 22}. Thus, our results suggest that excessive cortisol exposure and alterations in DHEA levels during sensitive period of development may impact long-term changes in the structure or function of neural areas regulating attention, working memory, and reading/writing abilities. Also, several secondary hormone imbalances occur with hypo- and hypercortisolemia that may play a role in the structural and functional changes in the brain of patients with active Cushing and after remission.

Studies of adults with CD report conflicting results regarding cognitive function prior to and after surgical treatment. There are reports of poorer performance in tests of attention, executive functioning, nonverbal aspects memory, visual and spatial information, reasoning, verbal ability, and language performance, with limited improvement in these aspects of cognitive function after remission suggesting that these aspects of cognitive function are the most vulnerable to changes of cortisol ^{4, 17, 23, 24}. A recent study of patients with various types of pituitary tumors, including corticotropinomas, reported that patients with CD showed lower verbal recognition memory than other groups of patients with pituitary tumors pre-treatment, with improvement at follow-up after TSS that was associated with a reduction in depression scores ²⁵. Although the number of subjects in this study was small, these findings in patients with CD, compared to other types of pituitary tumors who underwent TSS, are consistent with studies of adults with CS that report an association of mood disorders and cognitive function ^{2-4, 26}.

Previously we reported alteration in amygdala and hippocampal function in children and cognitive decrement in children after remission of CS despite reversal of cerebral and hippocampal atrophy, suggesting that the decline may be due to prior exposure to cortisol excess, a relative cortisol deficiency after remission of CS, recovery of the hypothalamic pituitary axis, or a combination of factors ⁵. There is a paucity of data regarding the cognitive effects of exogenous glucocorticoid use in children treated for acute or chronic medical conditions (e.g., nephrotic syndrome, juvenile arthritis, asthma, etc.). There are reports of behavioral changes, including insomnia, tearfulness, irritability, argumentative, fatigue, psychosis, and externalizing symptoms associated with duration of exogenous steroid use ^27-29.

There are conflicting reports with some studies showing no association of cognitive effects and exogenous glucocorticoid in children, while others note diminished verbal memory with high dose vs. low dose steroid use ^30-34. Alterations in cognitive function in adults treated with high dose steroids were largely reversible, while children showed greater vulnerability, with only partial recovery ³⁵.

While this study includes the cognitive data from the largest number of pediatric Cushing disease reported in the literature, the small sample size limits the generalizability of the findings. Another limitation of this study is the number of children excluded due to missing either pre or post- test results; however, since we did not have a control group of unaffected children we thought it was important to only include children who completed both baseline and follow-up, thus serving as their own control. Since we excluded from analysis some subtests (i.e. comprehension, word reasoning, letters and sequencing, matrix reasoning, symbol search, and cancellation) that had less than 20 completed both pre- and post-tests, it is possible that other differences would have been detected. Because the same tests were used at each time point practice effects may have been observed, possibly underestimating the general decline in scores in younger subjects. This study investigated the short-term cognitive impact of endogenous CS; future studies with longer follow-up are needed to examine long-term outcomes.

Conclusion

It is well established that chronic endogenous hypercortisolism is associated with deleterious effects on physical, metabolic, cognitive, and psychological function, with residual effects that persist after treatment. Our findings suggest that chronic glucocorticoid excess and accompanying secondary hormonal imbalances followed by eucortisolemia have detrimental effects on cognitive function in the developing brain; younger age and Tanner stage are risk factors for increased vulnerability, while older adolescents have cognitive vulnerabilities like that of adult patients, with CS. There are important clinical applications for our findings. Although none of the children in this study had scores less than one standard deviation from the population norms, the significant change in cognitive scores and school function from pre- to post-treatment experienced particularly by younger children, was a significant source of distress. It is important to counsel children and their parents about anticipating possible change in cognitive function, school performance, or mood post-surgery. Families should be encouraged to engage resources in the school, such as neuropsychological testing and individualized education plans, and work with the primary healthcare provider for referral to counseling. Future research is needed to elucidate the long-term effects of CS on neurocognitive function and to develop therapeutic interventions to improve outcomes and quality of life.

Abbreviations

ANOVA

Analysis of variance

BASC

Behavior assessment system for children

BMI

Body mass index

BMI SDU

Body mass index standard deviation units

CVLT

California verbal learning test

CD

Cushing disease

CS

Cushing syndrome

DHEA

Dehydroepiandrosterone

FIQ

Full scale intelligence quotient

GC

Glucocorticoid

Ht SDU

Height standard deviation units

HPA

Hypothalamic pituitary axis

IEP

Individual education plan

IQ

Intelligence quotient

M

Mean

NICHD

National Institute of Child Health and Human Development

POI

Perceptual organizing index

PIQ

Performance intelligence quotient

PSI

Processing speed index

RCI

Reliable change index

SD

Standard deviation

TMT

Trail making test

TSS

Transsphenoidal surgery

VCI

Verbal comprehension index

VIQ

Verbal intelligence quotient

WAIS

Wechsler adult intelligence scale

WISC

Wechsler intelligence scale for children

WRAT

Wide range achievement test

WJ-IV

Woodcock Johnson test of achievement

WMI

Working memory index

Data availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.