Neuropsychological and neuropsychiatric difference between healthy controls and CD patients.
Patients with Cushing’s disease reported higher depression, anxiety, and higher frequency and severity mental illness than healthy controls. Additionally, CD patients also behaved impaired cognitive ability than healthy controls (see Table 1)
Table 1. Demographic and clinical data from healthy controls and CD patients.
|
|
Healthy controls
(N=57)
|
CD patients
(N=50)
|
Healthy controls vs. CD patients
(p-value)
|
|
|
Mean (SD)
|
Min-Max
|
Mean (SD)
|
Min-Max
|
|
Age(year)
|
|
35.421(11.098)
|
20-63
|
37.960(10.728)
|
15-62
|
0.233
|
Education(year)
|
|
12.350(3.286)
|
5-17
|
11.040(4.015)
|
3-18
|
0.070
|
Gender(m/f)
|
|
0/57
|
-
|
0/50
|
-
|
-
|
Disease duration(month)
|
|
-
|
-
|
43.370(53.698)
|
1-300
|
-
|
Plasma cortisol level (nmol/L)
|
00:00
|
-
|
-
|
603.612(226.104)
|
290.420-11179.600
|
-
|
08:00
|
-
|
-
|
710.909(280.709)
|
221.440-1691.400
|
-
|
16:00
|
-
|
-
|
638.121(236.342)
|
208.990-1534.740
|
-
|
ACTH (nmol/L)
|
00:00
|
-
|
-
|
14.159(7.655)
|
3.640-40.200
|
-
|
08:00
|
-
|
-
|
18.644(15.443)
|
2.310-72.400
|
-
|
16:00
|
-
|
-
|
17.498(10.850)
|
3.220-51.100
|
-
|
UFC (nmol/24h)
|
|
-
|
-
|
2156.4372(1409.0234)
|
423.500-6777.400
|
-
|
Cushing QOL
|
|
-
|
-
|
36.617(8.894)
|
20-55
|
-
|
SDS
|
|
32.000(3.533)
|
25-42
|
49.425(12.379)
|
27-76
|
<0.001
|
SAS
|
|
24.631(3.642)
|
20-34
|
37.893(7.777)
|
24-59
|
<0.001
|
CNPI
|
|
-
|
-
|
11.000(10.836)
|
0-46
|
-
|
MoCA-BJ
|
Visuospatial
/Executive
|
4.777(0.540)
|
3-5
|
3.173(1.141)
|
0-5
|
<0.001
|
Naming
|
2.777(0.421)
|
2-3
|
2.326(0.844)
|
0-3
|
0.002
|
Memory
|
4.305(0.920)
|
2-5
|
2.565(1.544)
|
0-5
|
<0.001
|
Attention
|
5.833(0.447)
|
4-6
|
5.087(1.170)
|
0-6
|
<0.001
|
Language
|
2.472(0.608)
|
1-3
|
2.000(0.788)
|
0-3
|
0.004
|
Abstraction
|
1.722(0.566)
|
0-2
|
1.369(0.770)
|
0-2
|
0.019
|
Orientation
|
6.000(0)
|
6-6
|
5.804(0.542)
|
4-6
|
0.018
|
MoCA-BJ
total score
|
|
27.889(2.135)
|
21-30
|
22.369(4.428)
|
8-29
|
<0.001
|
Note: Independent T test was adopted to test between-group difference (HC vs. CD). Cushing QOL = Cushing quality of life; SDS = Self-Rating Depression Scale; SAS = Self-Rating Anxiety Scale; CNPI = Chinese version of neuropsychiatric inventory; MoCA-BJ = Montreal Cognitive Assessment-Beijing Version.
Functional connectivity within DMN, CEN and SN networks in the four states.
Spatial map of default mode network, central executive network and salience network identified using the group independent component analysis was shown in Fig. 1A. Independent components were grouped based on their anatomical and presumed functional properties: default mode network (ICs, 9, 12, 27, 28, 32, 44, 74), central executive network (ICs, 15, 21, 26, 48, 50, 63, 85, 89, 97), and salience network (ICs, 20, 43, 57, 59, 76, 82, 92). We adopted a k-means clustering algorithm on the dynamic functional connectivity (dFNC) from all subjects into four connectivity states. Fig. 1B shows the cluster centroid and the percentage of occurrences of each state (arranged in the order of emergence).
|---------------Fig.1---------------|
Different temporal properties between HC and CD patients
We firstly compared the mean dwell time between healthy controls and CD patients in each state (Fig. 2, A-D). Using independent T test, we found that the CD patients had higher mean dwell time than HC in State 1 (CD patients: 89.040 ± 59.216 vs. HC: 57.491 ± 40.671; t(105)=3.244, p = 0.002), but less mean dwell time than HC in State 4 (CD patients: 31.300 ± 39.413 vs. HC: 66.438 ± 45.734; t(105)=-4.227, p < 0.001). We did not observe significant difference in State 2 (CD patients vs. HC: t(105)=1.700, p=0.092), nor in State3 (CD patients vs. HC: t(105)=-1.517, p=0.132). For the switch time (i.e., the number of transitions), CD patients revealed less transition number than healthy controls did (CD patients: 6.600 ± 3.187 vs. HC: 7.824 ± 3.059; t(105)=-2.205, p = 0.045; Fig. 2E). Multiple comparisons were corrected by false-discovery rate (FDR), p < 0.05. All contrasts remained the same after FDR correction excepted the results of switch time became marginally significant, FDR corrected p = 0.075. Group difference on fraction of time in each state was similar with the mean dwell time (see Table S1).
|---------------Fig.2---------------|
Correlation between dynamic FNC properties and clinical characteristics.
To examined whether the dynamic FNC properties were associated with clinical characteristics, we did Pearson correlation analyses. Since the group differences were found in State 1 and State 4, we only restricted our analyses on these two states. Notably, we found that the dwell time in State 1 positively correlated with the self-reported anxiety (SAS), and cortisol level at 8:00, 16:00, 00:00, ACTH at 8:00, 16:00, as well as elevated 24-h urinary free cortisol. That is, the longer time spent on State 1 which with more sparsely connected pattern, the worse the mental health and higher cortisol level. We also detected a robust negative correlation between dwell time of State 1 and global cognitive scales (MoCA), which indicated that more time spent in State 1, the worse cognitive ability would be. In the contrary, dwell time in State 4 showed significant negative correlation with the self-reported depression, anxiety, and cortisol level at 8:00, 16:00, 00:00. More dwell time in State 4 predicted better cognitive performance measured by MoCA (all results see Table 2). Multiple comparisons were conducted by FDR, p < 0.05.
Table 2. Correlations between dynamic functional connectivity temporal properties in cognitive control network and clinical data.
|
|
Cushing-QOLa
|
SDS
|
SAS
|
CNPIa
|
MoCA-BJ
|
Cortisol
0000a
|
Cortisol
0800a
|
Cortisol
1600a
|
ACTH
0000a
|
ACTH
0800a
|
ACTH
1600a
|
UFCa
|
Dwell time State 1
|
r
|
-0.015
|
0.210
|
0.238
|
0.219
|
-0.533
|
0.293
|
0.298
|
0.399
|
0.232
|
0.294
|
0.347
|
0.413
|
|
P-value
|
0.919
|
0.054
|
0.028
|
0.126
|
<0.001
|
0.043
|
0.036
|
0.005
|
0.117
|
0.038
|
0.015
|
0.006
|
Dwell time State 4
|
r
|
0.034
|
-0.367
|
-0.323
|
-0.038
|
0.452
|
-0.362
|
-0.404
|
-0.322
|
-0.218
|
-0.267
|
-0.209
|
-0.266
|
|
P-value
|
0.823
|
0.001
|
0.003
|
0.794
|
<0.001
|
0.011
|
0.004
|
0.024
|
0.141
|
0.061
|
0.149
|
0.084
|
a. Correlation performed for the CD patients (N=50). Significant results which were survived after FDR (p<0.05) correction were reported in bold type.
Cushing QOL = Cushing quality of life; SDS = Self-Rating Depression Scale; SAS = Self-Rating Anxiety Scale; CNPI = Chinese version of neuropsychiatric inventory; MoCA-BJ = Montreal Cognitive Assessment-Beijing Version.
Dwell time in State 1 and State 4 within cognitive control network mediate group difference in cognitive performance.
Interestingly, we found the dwell time in State 1 and State 4 significantly mediated the difference between individuals with excessive high cortisol level (CD patients) and healthy controls on cognitive performance. That is, lower cognitive performance in CD patients was linked with more dwell time in State 1 (Fig. 3A), and less dwell time in State 4 (Fig. 3B) within the three networks.
|---------------Fig.3---------------|
Distinct network-based functional connectivity between CD patients and healthy controls and its associations with psychiatric symptoms and cognitive performance.
We have already known that the difference on dwell time in State 1 and State 4 can explain the group difference (i.e., CD patients vs. healthy controls) on cognitive performance. We further characterized the State 1 and State 4 by analyzing functional connectivity between the three networks, as well as the functional connectivity within each network. Results showed that in State 1, the CD patients had weaker connectivity within DMN (t(104)[1] = -2.584, p = 0.011), and the connections between CEN and DMN (t(104) = -5.141, p<0.001), CEN and SN (t(104) = -4.732, p<0.001) were also weaker than healthy controls. And in State 4, CD patients showed weaker functional connections between DMN and SN (t (84)[2] = -4.203, p<0.001), as well as DMN and CEN (t(84) = -3.547, p=0.001). Moreover, in State 4, functional connection between DMN and SN was negatively correlated with anxiety level measured by SAS (r(68) = -0.336, p = 0.005), and depression level measured by SDS (r(68) = -0.320, p = 0.008), but positively correlated with cognitive performance measured by MoCA (r(65) = 0.421, p < 0.001). Since CD patients showed decreased connection between DMN and SN, these results may suggest that the connection between DMN and SN was critical for understanding the psychiatric symptoms and cognitive deficits in CD patients. All significant results reported here were survived after FDR (p<0.05) correction.
We did not find significant associations between functional connectivity of neither inter-network and intra-network and psychiatric symptoms and cognitive deficits in State 1. No significant correlation results were found between the inter-network and intra-network connectivity and physiological indices (i.e., cortisol, ACTH, and UFC) in these two states, which may suggest that the dwell time in specific state would be more sensitive to physiological change.
Classification results based on dynamic FNC features.
The support vector machine (SVM) based on dynamic FNC approach (Fig. 4A, details see Method) showed classification accuracy of 88% for CD patients, 92.98% for healthy controls (Fig. 4B). The classification scores were evaluated using a receiver operating characteristic (ROC) curve aiming to visualize the performance of the classifier. The classification results may further indicate that the dynamic functional connectivity pattern within these three networks would be the potential biomarker of individuals with excessive higher cortisol level.
|---------------Fig.4---------------|
[1] There was one healthy participant did not have State 1.
[2] There were three CD patients did not have State4, and eighteen healthy participants did not have State4.