Demographic data and FDG-PET
A total of 164 subjects were enrolled, including 44 MCI converters (MCI-C), 38 non-converters (MCI-NC), 42 patients with amyloid-positive AD, and 40 cognitively normal controls, all of whom had at least one FDG-PET scan (Table 1), while 37 patients in the MCI group had two or more FDG-PET scans (MCI-C: 24, MCI-NC: 13). The interval from the first FDG-PET scan to the diagnosis of dementia in the MCI-C group and the observation intervals in the MCI-NC group are shown in Fig. 1. FDG-PET hypometabolism preceded conversion to dementia after an average of 3.70 ± 1.68 years (range, 1 year to 8 years). The mean observation duration after the first FDG-PET in the MCI-NC group was 4.34 ± 1.26 years (range, 2 years to 7 years).
Table 1
Demographic data of four groups
|
Control
|
MCI
|
AD
|
|
|
Non-converter
|
Converter
|
|
|
Sample size
|
40
|
38
|
44
|
42
|
Age at first visit FDG-PET
|
62.6 ± 11.6
|
72.1 ± 7.60*
|
74.7 ± 7.04*
|
77.0 ± 7.6*
|
Gender (male/female)
|
23/17
|
15/23*
|
19/25*
|
21/21*
|
Education (years)
|
13.6 ± 3.7
|
6.9 ± 5.0*
|
7.1 ± 4.9*
|
7.5 ± 5.3*
|
APOE Ɛ4 carriers
|
23.8%
|
37.8%*
|
60.9%*§
|
57.9%*
|
First time MMSE
|
28.1 ± 2.1
|
23.5 ± 4.36*
|
21.8 ± 3.7*
|
17.4 ± 2.9*§#
|
Baseline Cognitive ability Screening Instrument
|
Total scores (100)
|
92.6 ± 6.1
|
77.84 ± 13.9*
|
74.9 ± 12.2*
|
63.1 ± 15.3*§#
|
Long term memory (10)
|
9.88 ± 0.7
|
9.2 ± 1.8*
|
9.1 ± 2.6*
|
7.9 ± 3.6*§#
|
Short term memory (12)
|
10.5 ± 1.7
|
7.1 ± 3.6*
|
4.1 ± 4.0* §
|
3.8 ± 1.5*§#
|
Attention (8)
|
7.3 ± 0.8
|
6.8 ± 1.2*
|
6.7 ± 1.4*
|
6.5 ± 1.3*§#
|
Mental manipulation (10)
|
8.8 ± 1.8
|
6.3 ± 2.9*
|
7.0 ± 3.2*
|
6.3 ± 2.5*§#
|
Orientation (18)
|
17.7 ± 1.0
|
15.0 ± 4.0*
|
11.7 ± 5.1* §
|
10.9 ± 2.6*§#
|
Drawing (10)
|
9.8 ± 0.5
|
8.3 ± 2.3*
|
8.1 ± 2.9*
|
7.2 ± 3.6*§#
|
Abstract thinking (12)
|
10.8 ± 1.2
|
8.5 ± 2.3*
|
8.3 ± 2.7*
|
7.5 ± 2.7*§#
|
Verbal fluency (10)
|
8.2 ± 2.1
|
5.7 ± 2.3*
|
5.8 ± 2.8*
|
4.8 ± 1.8*§#
|
Language (10)
|
9.8 ± 0.8
|
9.0 ± 1.5*
|
8.7 ± 2.1*
|
4.9 ± 2.1*§#
|
First visit FDG-PET SUVr
|
|
|
|
|
Frontal lobe
|
1.41 ± 0.10
|
1.45 ± 0.04
|
1.35 ± 0.06* §
|
1.30 ± 0.07*§#
|
Medial temporal lobe
|
1.10 ± 0.07
|
1.06 ± 0.05
|
0.91 ± 0.07 *§
|
0.86 ± 0.02*§#
|
Lateral temporal lobe
|
1.40 ± 0.12
|
1.38 ± 0.04
|
1.23 ± 0.08* §
|
1.19 ± 0.11*§#
|
Posterior cingulate cortex
|
1.61 ± 0.13
|
1.60 ± 0.20
|
1.37 ± 0.14 *§
|
1.20 ± 0.07*§#
|
Precuneus
|
1.61 ± 0.12
|
1.61 ± 0.05
|
1.48 ± 0.09* §
|
1.31 ± 0.08*§#
|
Parietal lobe
|
1.50 ± 0.10
|
1.49 ± 0.04
|
1.36 ± 0.08* §
|
1.21 ± 0.06*§#
|
Occipital lobe
|
1.60 ± 0.11
|
1.59 ± 0.05
|
1.56 ± 0.08
|
1.39 ± 0.12*§#
|
Follow-up (years)
|
4.4 ± 1.8
|
6.4 ± 3.8*
|
6.7 ± 3.6*
|
6.6 ± 3.2*
|
Data presented as mean ± standard deviation. |
* p < 0.05 with control; § p < 0.05 with MCI non-converter; # p < 0.05 with MCI converter. |
Abbreviations: MCI, mild cognitive impairment; AD, Alzheimer’s disease; APOE, Apolipoprotein E; MMSE, mini-mental state examination; FDG-PET, fluorodeoxyglucose positron emission tomography; SUVr, standard uptake value ratio. |
For the first FDG-PET scan, there was no significant difference in mean age between the two MCI groups (p = 0.156). However, the SUVr of seven preselected VOIs, but not the occipital lobe, were significantly lower in the MCI-C group (p < 0.05) (Table 1). The cognitive test scores and SUVr of all VOIs were lowest in the patients with AD.
The percentages of hypometabolism in the MCI-C group were significantly higher than those in the MCI-NC group. For the MCI-C group, the percentages of hypometabolism in the PCC, MTL, precuneus, frontal, parietal, and lateral temporal lobe were 75.2%, 70.0%, 41.2%, 40.0%, 36.7%, and 32.0%, respectively, compared to 12.8%, 5.2%, 8.1%, 5.0%, 5.2%, and 5.0% in the MCI-NC group (Supplementary Fig. 1).
In the two MCI groups and the first FDG-PET, MTL hypometabolism predicted future conversion with 94.7% sensitivity and 80.5% specificity, followed by the PCC (89.5% sensitivity and 73.1% specificity). Thirty-four cases (of 82 MCI cases) were categorized as having normal metabolism status in MTL and PCC areas; all of these cases were in the MCI-NC group. Of note, three cases with MCI showed MTL and PCC hypometabolism but remained cognitively stable (no-conversion) during follow-up.
First Fdg-pet Topography And Its Relationship With Cognitive Performance
The correlation between first FDG-PET Z scores and corresponding MMSE scores, adjusted for years of education, best fit in the AD group (Fig. 2). Among the preselected VOIs, six had significant correlations with MMSE scores in the AD group.
The same analysis was conducted with the CASI and its subdomains to clarify whether hypometabolism in different brain regions was associated with cognitive impairment of different domains (Supplementary results and Supplementary Table 1).
Longitudinal Cognitive Trajectory
Based on the median follow-up duration, we represented the first six consecutive measurements of MMSE or CASI total scores in the four groups (Fig. 3). The MCI-C and AD groups had similar progression trajectories, showing a continuous decline in MMSE and CASI scores as the number of measurements increased. In contrast, the trends of cognitive decline in the MCI-NC and control groups were less obvious. Data of the cognitive features of the two MCI groups are listed in the Supplementary results and Supplementary Table 2.
Evolutional pattern of cortical hypometabolism & cortical atrophy
-individual Level
A representative example of an MCI-C individual depicting the evolutional pattern of three consecutive FDG-PET scans is shown in Fig. 4. Hypometabolism (Z score < -2) started in the MTL and inferior frontal region, followed by lateral temporal, PCC, parietal and frontal cortex regions years before the cognitive changes.
-group Level
The group-level FDG-PET hypometabolism evolution and cortical thickness degeneration in the MCI-C, MCI-NC and AD groups are shown in Fig. 5. According to the predefined time frame, the time effect of FDG-PET hypometabolism and MRI cortical thickness degeneration was divided into early disease stage (Fig. 5A) and disease progression stage (Fig. 5B).
The FDG-PET hypometabolic time effect in the AD group suggested a greater spatial extent of hypometabolism (over the medial prefrontal, lateral temporal, temporal-parietal and precuneus regions; Fig. 5A) in the early disease stage compared to the corresponding cortical thickness degenerative trajectory (only over the hippocampus). As the disease progressed, the cortical thickness degenerative pattern became more consistent with the hypometabolic pattern (Fig. 5B). In the MCI-C group, the evolutional pattern of hypometabolism in the early disease stage (Fig. 5A) was similar to the AD group, but with a relatively sparse distribution. As the disease progressed, the hypometabolism pattern mimicked that of AD (Fig. 5B). One exception was the time effect on the dorsolateral prefrontal cortex (DLFC), which showed progressive hypometabolic changes in the MCI-C group (observed at the disease progression stage) but not in the AD group (in either the early stage or progression stage); instead, in the AD group, cortical atrophy over the DLFC was observed. Hypometabolism or cortical atrophy in the MCI-NC group was inconspicuous at the early stage; however, as the follow-up duration increased, cortical atrophy over the hippocampus and lateral temporal lobe was noted. Of note, there was no signal change in the hippocampus in the FDG-PET time effect model at either the early disease stage or disease progression stage among the three groups. In fact, hippocampal hypometabolism, defined as a Z score < -2, was noted from the first FDG-PET scan in both the MCI-C (70%) and AD (81%) groups.
Distinct Fdg-pet Topographies Of Mci Converters And Non-converters
Repeated-measures analysis was conducted in the MCI subjects who received two FDG-PET scans (MCI-C: 24, MCI-NC: 13). The preselected seven VOIs between the two MCI groups and the two scans were compared (Table 2). There was no significant difference in the mean age at the second FDG-PET scan between the two groups. In the first scan, the MCI-C group had significantly lower frontal, PCC, precuneus, lateral temporal, and MTL Z scores. During follow-up, the Z scores of all VOIs became significantly lower in the MCI-C group. Figure 6 shows the distinct differences in the cortical hypometabolic pattern of FDG-PET in the MCI-C and MCI-NC groups (Fig. 6A, B). The difference could be observed from the first FDG scan and became more evident during follow-up (Fig. 6C, D).
Table 2
Difference of FDG-PET Z scores of two scans between MCI-C (n = 24) and MCI-NC (n = 13)
MCI-C compares to MCI-NC
|
β
|
SE
|
p value
|
Frontal1
|
-0.650
|
0.311
|
0.044*
|
Frontal2
|
-0.952
|
0.337
|
0.008**
|
PCC1
|
-1.467
|
0.319
|
< 0.001***
|
PCC2
|
-1.636
|
0.308
|
< 0.001***
|
Precuneus1
|
-0.969
|
0.342
|
0.008**
|
Precuneus2
|
-1.094
|
0.360
|
0.004**
|
Parietal1
|
-0.696
|
0.358
|
0.060
|
Parietal2
|
-0.966
|
0.374
|
0.014*
|
Occipital1
|
-0.527
|
0.334
|
0.123
|
Occipital2
|
-0.888
|
0.334
|
0.014*
|
Lat. temporal1
|
-1.211
|
0.345
|
0.001**
|
Lat. temporal2
|
-1.577
|
0.328
|
< 0.001***
|
Med. temporal1
|
-1.757
|
0.310
|
< 0.001***
|
Med. temporal2
|
-2.018
|
0.331
|
< 0.001***
|
Superscript 1 indicates the first scan of FDG-PET; 2 indicates the second scan of FDG-PET. * p < 0.05; ** p < 0.01; *** p < 0.001 |
Abbreviations: MCI-NC, non-converter of mild cognitive impairment; MCI-C, converter of mild cognitive impairment; PCC, posterior cingulate cortex; Lat. temporal, lateral temporal; Med. temporal, medial temporal. |
Generalized Estimating Equation Model Revealed The Factors Correlated With Cognitive Decline
To better understand which factors affected cognitive change, a generalized estimating equation model was used for the longitudinal data. As shown in Table 3, the FDG-PET PCC, precuneus and lateral temporal lobe Z scores were inversely correlated with MMSE scores, of which the lateral temporal lobe had the most significant effect. In contrast, the Z scores of the frontal lobe had a positive effect on the MMSE scores. Similar findings were seen when using CASI total scores as the dependent variable.
Table 3
The coefficient of covariates in generalized estimating equation model for MMSE & CASI
|
|
MMSE
|
CASI
|
|
|
β
|
SE
|
p value
|
β
|
SE
|
p value
|
Male gender
|
1.198
|
0
|
.025*
|
4.694
|
1.901
|
0.014*
|
Education year
|
0.415
|
0.074
|
< 0.001***
|
1.376
|
0.252
|
< 0.001***
|
FDG-PET Z score
|
|
|
|
|
|
|
|
Frontal
|
-2.289
|
0.496
|
< 0.001***
|
-9.620
|
1.616
|
< 0.001***
|
|
PCC
|
1.637
|
0.497
|
0.001**
|
4.676
|
1.794
|
0.009**
|
|
Precuneus
|
1.434
|
0.657
|
0.029*
|
4.867
|
2.315
|
0.036*
|
|
Parietal
|
-0.968
|
0.660
|
0.143
|
-1.729
|
2.519
|
.492
|
|
Occipital
|
-0.244
|
0.587
|
0.667
|
-1.582
|
1.759
|
.368
|
|
Lat. temporal
|
1.891
|
0.547
|
0.001**
|
8.733
|
1.938
|
< 0.001***
|
|
Med. temporal
|
-0.011
|
0.266
|
0.967
|
-0.574
|
0.899
|
0.523
|
Measurement number, AD
|
|
|
|
|
|
|
|
2nd time
|
-0.847
|
0.632
|
0.180
|
-1.573
|
1.761
|
0.372
|
|
3rd time
|
-1.020
|
0.765
|
0.182
|
-3.177
|
2.242
|
0.157
|
|
4th time
|
-2.256
|
1.123
|
0.045*
|
-8.341
|
3.017
|
0.006**
|
|
5th time
|
-3.004
|
1.227
|
0.014*
|
-15.092
|
3.618
|
< 0.001***
|
|
6th time
|
-5.325
|
2.154
|
0.013*
|
-21.842
|
4.378
|
< 0.001***
|
Measurement number, MCI-C
|
|
|
|
|
|
|
|
2nd time
|
-0.339
|
0.590
|
0.565
|
-1.260
|
1.637
|
0.441
|
|
3rd time
|
-0.888
|
0.719
|
0.217
|
-2.628
|
2.045
|
0.199
|
|
4th time
|
-1.990
|
0.999
|
0.046*
|
-7.282
|
2.780
|
0.009**
|
|
5th time
|
-2.730
|
1.074
|
0.011*
|
-15.957
|
3.322
|
< 0.001***
|
|
6th time
|
-5.116
|
2.097
|
0.015*
|
-22.565
|
4.022
|
< 0.001***
|
Measurement number, MCI-NC
|
|
|
|
|
|
|
|
2nd time
|
-0.687
|
0.633
|
0.278
|
-1.890
|
2.016
|
0.348
|
|
3rd time
|
0.557
|
0.734
|
0.448
|
4.240
|
2.059
|
0.039*
|
|
4th time
|
-0.076
|
0.912
|
0.933
|
0.248
|
2.375
|
0.917
|
|
5th time
|
-0.237
|
0.942
|
0.802
|
-2.682
|
3.169
|
0.397
|
|
6th time
|
-1.206
|
2.098
|
0.566
|
-8.594
|
3.983
|
0.031*
|
* p < 0.05; ** p < 0.01; *** p < 0.001 |
Abbreviations: MCI-NC, non-converter of mild cognitive impairment; MCI-C, converter of mild cognitive impairment; AD, Alzheimer’s disease; MMSE, mini-mental state examination; CASI, cognitive ability screening instrument total score; FDG-PET, fluorodeoxyglucose positron emission tomography; PCC, posterior cingulate cortex; Lat. temporal, lateral temporal; Med. temporal, medial temporal; SE, standard error. |
Furthermore, using time as a categorical variable, we found negative interactions between the number of measurements and cognitive test scores in the AD and MCI-C groups but not in the MCI-NC group. In addition, the interactions became significant after the third measurement. Interestingly, the MCI-NC group showed some improvement in cognitive test performance during follow-up; for example, the third MMSE scores increased by 0.557 points from baseline, and the third CASI total scores increased by 4.204 points. Nevertheless, the performance worsened from the fourth measurement. The trends in cognitive decline were most prominent in the AD and MCI-C groups.