Computerized Analysis of Brain MRI Parameter Dynamics in Young Patients With Cushing Syndrome—A Case-Control Study

Abstract

Background

Young patients with Cushing Syndrome (CS) may develop cognitive and behavioral alterations during disease course.

Methods

To investigate the effects of CS on the brain, we analyzed consecutive MRI scans of patients with (n = 29) versus without CS (n = 8). Multiple brain compartments were processed for total and gray/white matter (GM/WM) volumes and intensities, and cortical volume, thickness, and surface area. Dynamics (last/baseline scans ratio per parameter) were analyzed versus cortisol levels and CS status (persistent, resolved, and non-CS).

Results

Twenty-four-hour urinary free cortisol (24hUFC) measurements had inverse correlation with the intensity of subcortical GM structures and of the corpus callosum, and with the cerebral WM intensity. 24hUFC dynamics had negative correlation with volume dynamics of multiple cerebral and cerebellar structures. Patients with persistent CS had less of an increase in cortical thickness and WM intensity, and less of a decrease in WM volume compared with patients with resolution of CS. Patients with resolution of their CS had less of an increase in subcortical GM and cerebral WM volumes, but a greater increase in cortical thickness of frontal lobe versus controls.

Conclusion

Changes in WM/GM consistency, intensity, and homogeneity in patients with CS may correlate with CS clinical consequences better than volume dynamics alone.

In 1932 Dr Harvey Cushing described a syndrome of “painful obesity, hypertrichosis and amenorrhea with overdevelopment of secondary sexual characteristics”(1). This syndrome, known today as Cushing syndrome (CS), is caused by hypercortisolemia, either due to iatrogenic glucocorticoid administration or to endogenous cortisol over-secretion. Among the latter, the majority of patients harbor pituitary adrenocorticotropic hormone (ACTH)-secreting adenomas, as initially described by Dr Cushing, whereas others have increased cortisol secretion by the adrenal gland(s), or rarely, ACTH/corticotropin-releasing hormone (CRH) secretion by an extra-pituitary tumor (2).

Cushing Syndrome is a multisystemic disease, with deleterious effects on body metabolism, cardiovascular parameters, bone and mineral equilibrium, endocrine function, and other systems (3). Importantly, the hypercortisolemic state has a strong impact on brain functions, both affective and cognitive. Adult patients with active CS have a high prevalence of depression (4, 5), bipolar disorder (6), and suicidality (7), as well as altered cognitive function (8), expressed as memory impairment and the difficulty to concentrate (7).

Similar alterations were reported in pediatric patients with active CS, which had a high prevalence of obsessive–compulsive behavior, personality changes, and mood disorders (9). Despite successful treatments leading to the remission of CS, brain functions may not return to baseline and may even deteriorate (10). Among adults in remission of CS, residual anxiety (11), depression (4, 11) and suicidality (4) were reported, as well as impaired cognitive function. Similarly, children after remission of CS have residual affective syndromes, suicidal ideation (12), and cognitive deficits even 1 year after the resolution of CS (13, 14). In fact, during the active disease and in relapse of hypercortisolemia, patients seem to be more focused and better functioning, and deteriorate in the post-treatment periods, with cognitive decline and new psychiatric alterations.

Considering these observations, several studies tried to assess the association between hypercortisolemia and brain imaging parameters (10). In children with CS, the cerebrum cerebral and amygdala had lower volumes, and brain ventricles volumes were larger in CS patients versus controls (13).

Analysis of brain imaging parameters after CS resolution was reported in adults, with negative correlation between 24hUFC measurements and hippocampal volume dynamics (15–18), and positive correlation with third ventricle volume, brain atrophy, and bicaudate diameter (19). However, this has not yet been analyzed in young patients.

Following our previous studies (13, 14, 20), we aimed to analyze changes in brain MRI parameters in different anatomic sites, in patients with persistent hypercortisolemia, among those with disease resolution, and in controls in order to better define the effects of cortisol excess on the brain.

Materials and Methods

Population

This is a retrospective analysis of patients with CS and a control group of patients that underwent brain MRI for other indications. All patients were recruited through clinical protocols 97-CH-0076, 95-CH-0059, and 00-CH-0160 conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and gave written informed consent. The diagnostic process for children and adults with CS in these protocols was described in details previously (21, 22).

Biochemical analysis

All patients were hospitalized and underwent screening tests for CS, including serial 24hUFC collection and plasma cortisol measurements. To control for different reference ranges, the UFC is presented as the ratio between the measurement and the upper limit of normal range. The change in 24hUFC between the scans is presented as the ratio between the UFC values in the time of the second/first scans. Disease activity was defined as 24hUFC above the upper limit of the reference range.

Imaging protocol

All subjects underwent 2 MRI scans on the same 1.5Tesla Philips Achieva Scanner (Best, Netherlands), with a 1.3 year interval between the scans (median, interquartile range 1.0–2.6). In-plane resolution of the images were recorded as 0.5 × 0.5 mm^2 with slice thickness of 5 mm. Images were acquired using a 3D spoiled gradient-recalled echo in the steady state, with standard head positioning (13). A multichannel SENSE head coil was used during scanning. T1-weighted SE (spin echo) images were reconstructed with a TR (repetition time) of 425, TE (echo time) of 15 ms, and flip angle = 76 degrees. The image analysis team was blinded to age, gender, and diagnosis of the patients while doing volumetric and surface-based analysis of MR scans. The change in each parameter between the scans is presented as the ratio between the second and first measurements.

Quantification of the parameters per brain regions

Freesurfer (23), a software tool for the quantitative analysis of brain surface was utilized for the MRI analyses in our experiments, both for subcortical and cortical segmentation.

Subcortical segmentation.

For each scan, MRI volume was first registered with the MNI305 atlas using an affine registration. An initial labeling was performed to generate a prior distribution. Bias field correction was done, followed by a nonlinear registration to the MNI305 template. Next, skull stripping was employed to generate the brain mask to which labeling was performed. The volumes of the subcortical structures were used as a metric for our analysis (23).

Surface parcellation.

After registering the MRI volume with the MNI305 atlas using an affine registration, bias field correction and skull stripping were performed. Voxels were classified as WM or other tissue types based on the intensity and neighbor constraints. The cerebellum and brain stem were removed, and hemispheres were separated from each other via a cutting planes approach. An initial surface, white surface, was then generated for each hemisphere and refined to follow the intensity gradients between the WM and GM. The pial surface was generated from this surface by following the intensity gradients between the GM and the cerebrospinal fluid.

For brain surface labeling, the spherical representation of the brain was generated and registered to a manually labeled training set. A spherical statistical atlas was used to label the cortical surfaces into neuroanatomical regions of interest. The methods for spherical labeling were previously reported and validated (24).

Cortical surface measurements included cortical thickness (calculated from the distance between the WM surface and pial surface), surface area (computed from the white surface data), and WM surface, volume, and intensity values (computed from the white surface). The methods have been fully described and validated previously (25).

Computational protocol

Seven parameters were computed: brain compartment volumes, WM volume and intensity, cortical thickness, mean intensity, and surface area. Per each patient, the ratio of the parameter between the second and first MRI scans was calculated and used for the analysis, using each patients’ scan as a baseline control.

Statistical analysis

Statistical calculations were performed with SPSS 20.0 software (SPSS Inc., Chicago, Illinois), and figures were produced using R studio (version 1.0.153). Results are expressed as mean ± standard deviation (SD) unless otherwise indicated. Generally, the independent Student’s t-test was used to analyze differences in numerical variables, the chi-square test was employed to analyze differences in categorical variables, and the Pearson product was used for analysis of correlations between variables. The Mann–Whitney U test was used to compare the change in MRI parameters between patients with CS, with remission versus without remission, and for comparing patients achieving remission versus controls. The two-tailed P-value for significance was set at less than 0.05.

Among patients with CS, correlation analyses were performed between 24hUFC collections and MRI parameter dynamics between the last versus first MRI scans. Patients older than 30 years (n = 6) were excluded from the brain dynamics analysis. We then compared the dynamics of brain MRI parameters among patients with CS at baseline, comparing those with remission versus persistent disease at the second evaluation. In addition, we compared those parameters between patients with CS achieving remission versus controls.

The required minimal sample size was n = 14 for the between groups comparisons to detect a difference of 0.3 mL between the groups, with an alpha of 0.05 (two-tailed) and a beta of 0.3, considering a ratio of 1:3.5 between sample sizes. For the correlation analysis, a sample size of n = 37 was required to reject the null hypothesis using a threshold of 0.35 for the r correlation coefficient, with an alpha of 0.05 (one-tailed) and beta of 0.3.

Results

The current analysis includes 37 patients (median age 14 years, interquartile range [IQR] 10–20) and 21 female patients. Twenty-nine patients were diagnosed with CS (12 with Cushing disease, 15 with primary adrenal CS, and 2 with ectopic ACTH/CRH-secreting tumors), and 8 controls were evaluated for a pituitary mass (n = 5) or for tall stature (n = 3) without clinical symptoms of CS. The groups were comparable in terms of gender and age. Patients’ characteristics by the clinical groups are detailed in Table 1 (26).

Table 1.

Characteristics of Patients Included in the Study

	Cushing Syndrome			P-value
	Not-resolved	Resolved	Control	NR vs R	CS vs. control
n	8	21	8
Age at diagnosis	15.5 [10.8]	21.0 [10.5]	14.5 [25.5]	0.12	0.32
Female n (%)	6 (75.0)	12 (57.1)	3 (37.5)	0.67	0.25
Ethnicity n (%)				0.55	0.56
African-American	2 (25.0)	0	0	-	-
Asian	0	1 (4.8)	0	-	-
Hispanic	0	2 (9.5)	1 (12.5)	-	-
Caucasian	6 (75.0)	18 (85.7)	7 (87.5)	-	-
Bilateral adrenalectomya n (%)	0	8 (38.1)	0	0.76	1.0
First UFC (xULN)	1.4 [5.3]	1.3 [4]	0.5 [0.5]	0.59	0.017
Last UFC (xULN)	2.5 [1.7]	0.3 [0.5]	0.3 [0.3]	<0.001	0.35
UFC ratio	1.2 [3.5]	0.3 [0.6]	0.9 [1.4]	0.06	0.56

Continuous variables are presented as median [interquartile range], and categorical variables are presented as n (%).

Abbreviations: CS, Cushing Syndrome; NR, not resolved; R, resolved; UFC, urinary free cortisol; xULN, times the upper limit of the normal range.

^a Between the scans included in the current analysis.

Of the 29 patients with CS, 8 had persistent disease during the second MRI scanning, and among the remaining 21 patients, 6 were in remission of CS before the first MRI scan, with a median time interval between the maximal UFC and scan of 2 years (range 1–8 years).

24hUFC levels versus MRI parameters

When analyzing all the scans together, 24hUFC was positively correlated with the lateral ventricles volumes (left and right, r = 0.51 and r = 0.61, respectively, P < 0.001 both) and negatively correlated with the intensity of several subcortical structures (Fig. 1A), and of the corpus callosum in independent measures of its different sections (Fig. 1B). Similarly, late night plasma cortisol values correlated negatively with the intensity of WM in gyri located in the frontal and parietal lobes (Fig. 2).

Figure 1.

Correlation analysis between A, brain MRI intensity of subcortical structures, and B, of corpus callosum substructures and 24-hour urinary free cortisol measurements in the entire cohort.

Figure 2.

Correlation analysis between 24-hour urinary free cortisol and white matter intensity in the frontal and parietal lobes.

Urinary free cortisol changes versus MRI parameter changes

Urinary free cortisol changes had an inverse correlation with the change in total brain volumes (r = -0.62, P = 0.04, n = 14), as well as with the volume change in the supratentorium, cerebellar cortex, hippocampus, and putamen (Fig. 3A). In line with this finding, the change in 24hUFC had a negative correlation with the surface area of several anatomic sites in the frontal and temporal lobes (Fig. 3B).

Figure 3.

Correlation analysis between A, the change in 24-hour urinary free cortisol versus the change in volumes of the supratentorial brain, the cerebellar cortex, hippocampus, and putamen, and B, between the change in cerebral cortex surface area and the change in 24-hour urinary free cortisol in the frontal (superior frontal gyrus and insula) and temporal lobes (fusiform and inferior-temporal gyri and pars orbitalis) in patients with Cushing syndrome.

MRI parameter dynamics—CS patients with remission versus without remission

Patients with persistent disease had an increase in WM volume and a decrease in cortex thickness and WM intensity compared with patients achieving remission of CS (Fig. 4). These findings were found in multiple anatomic sites, mainly in the frontal and parietal lobes (Table 1).

Figure 4.

Dynamics in brain MRI parameters of different anatomic sites among patients with Cushing syndrome; comparison between patients with disease remission versus without disease remission in: A, cortical thickness; B, white matter volume; and C, intensity.

Patients with resolving CS versus controls

When compared with a control group of patients not diagnosed with CS, patients recovering from CS had a decrease in subcortical GM volume, an increase in cortical thickness, and a decrease in WM volume in multiple anatomic sites (Fig. 5).

Figure 5.

Brain MRI parameter dynamics; comparison between patients recovering from Cushing syndrome and controls.

Discussion

In the current study we analyzed MRI parameters in young patients with CS versus patients that underwent brain MRI for other indications. We compared the dynamics of several brain MRI parameters in 2 consecutive scans between the groups and between patients with remission versus without remission of CS. In addition, we performed a correlation analysis between 24hUFC and brain MRI parameters.

Urinary free cortisol was negatively correlated with the intensity of subcortical GM structures and of the corpus callosum, and with the cerebral WM intensity. Changes in UFC had an inverse correlation with the volume of several structures in the cerebrum and cerebellum, including the hippocampus. Patients with persistent CS had less of an increase in cortical thickness and WM intensity between the scans, and less of a decrease in WM volume compared with patients with resolution of CS. Finally, patients with resolution of their CS had less of an increase in subcortical GM and of cerebral WM volume, but a greater increase in cortical thickness of the frontal lobe compared with controls.

Changes in brain structure have been shown to be associated with clinical findings in patients with CS. Hippocampal volume was positively correlated with learning (16) and memory capacity (27), reduced WM integrity in the uncinated fasciculus was associated with depressive symptoms severity (28), GM volume negatively correlated with visual memory and quality of life (29), and an increase in right caudate head volume was associated with improved mood and ideation (30). Hence, our findings may explain both affective and cognitive functional deficits in patients with CS and after disease remission.

We compared the differences between patients with CS achieving versus not achieving remission, and between CS patients achieving remission versus controls. In both comparisons, most of the differences found were located anatomically in the frontal lobes, and various parameters were different in the same anatomic regions. This strengthens our results, and may point to the areas mainly affected by the dynamics in cortisol levels. While this manuscript was under revision we became aware of another study (31) that is largely confirmatory of our data.

To build a clearer picture of cognitive impairment in CS, a multimodal study can be a possible direction to move forward. With the foundation of the changes in brain structure associated with CS, the emotional and neurocognitive dysfunctions can be further studied using DTI and fMRI modalities. DTI studies can help identify axonal loss, alterations to fiber coherence and/or demyelination (32, 33) and whether any observed changes are reversible. Any microstructural features thus identified could be indicative of biological changes. Resting-state fMRI studies investigating the altered spontaneous brain activity in patients with CS found significant differences in the precuneus and thalamus regions (34), supporting our current findings on alterations in the same sites.

The effect of age and gender on brain structure was studied previously, demonstrating an age-related decrease of brain volume and thickness, most pronouncedly in the parietal and frontal lobes (35). Surface area was larger in men versus women, but it had much lower effect than gender in the multivariable analysis and did not show anatomic site-specific effect. In comparison to these changes, that were calculated for lifetime (age range 18–94 years), the dynamics in our study were calculated in much shorter time periods, and hence are not expected to distort the results.

In our analysis we found more pronounced increase in WM volume in patients with persistent CS versus those in remission, and decreasing WM volume in patients achieving remission versus controls. Glucocorticoids exert a trophic effect on oligodendrocytes (36), which are responsible for myelin production. In contrast, when CS resolves, the sharp decrease in cortisol levels may lead to a local cortisol depletion, with consequent demyelination (37). This may explain both the differences in WM volume and intensity between the groups, and among patients with CS with different UFC levels.

The current analysis has several limitations that should be taken into consideration when interpreting the results. First, the small sample size increases the chance for statistical errors, type I and II. The sample size required for the correlation analysis was borderline, but the consistent findings using several parameters (that are not measured in statistical tools) strengthen our confidence in the reliability of the results. Second, the patients included in the analysis are heterogeneous in terms of their etiology for CS. Since the pattern of cortisol dynamics is highly related with the etiology and with the related intervention (eg, bilateral adrenalectomy for PPNAD vs. adenoma resection in Cushing disease), these factors could not be effectively controlled in the analysis. Third, in the current analysis we did not assess the impact of cortisol dynamics on clinical parameters.

In conclusion, our study shows that in addition to the volumetric changes in the brain structures of patients with CS, there are widespread changes in the consistency of the brain in terms of the WM/GM ratio and in terms of WM intensity and homogeneity. Hence, the change in WM volume by itself might not reliably correlate with brain functions, while the combination of the change in volume, WM intensity, and the density of WM hyper/hypointensities may better correlate with the clinical consequences. Our analysis may also suggest, that the dynamics in cortisol levels, in addition to the hypercortisolemia per se, may alter the brain structure and function.

Additional Information

Disclosure statement: Dr. Stratakis holds patents on the PRKAR1A, PDE11A and GPR101 genes and/or their function and his laboratory has received research funding from Pfizer Inc.