Demographic variables and CSF biomarkers value distribution in the diagnostic groups
Demographic variables and CSF biomarkers' results in the two main diagnostic groups are shown in Table 1. Patients with CJD were significantly younger than those suffering from np-RPD (P<.0001). There was no significant difference between sex distribution across the two diagnostic groups. Results regarding t-tau, NfL, and 14-3-3 were in line with those previously reported (14). P-tau concentrations were not significantly different between CJD and np-RPD cases. Patients with CJD showed significantly higher CSF SNAP-25 (P<.0001) and CSF Ng (P<.0001) levels than those with np-RPD.
CSF biomarkers levels in the diagnostic groups are shown in Table 2. There was no significant difference in CSF SNAP-25 and Ng concentrations between sCJD and gCJD participants. Within the np-RPD group, CSF SNAP-25 and Ng levels did not vary significantly across the three diagnostic categories.
Table 1 - Demographic variables and CSF biomarkers in the CJD and np-RPD cohorts
|
CJD (N = 220)
|
np-RPD (N = 213)
|
P value
|
Age at LP
|
66.5 ± 9.9
|
70.9 ± 11.3
|
<.0001
|
Female (%)
|
48.6
|
52.1
|
.5218
|
CSF t-tau (pg/mL)
|
3287 (1949-8105)
|
680 (381-1346)
|
<.0001
|
CSF 14-3-3 (pg/mL)
|
53350 (26850-101500)
|
10300 (6188-22100)
|
<.0001
|
CSF NfL (pg/mL)
|
6209 (3501-11088)
|
2098 (1013-11838)
|
<.0001
|
CSF p-tau (pg/ml)
|
54 (37-75)
|
59 (35-96)
|
.3363
|
CSF SNAP-25 (pg/mL)
|
582 (240-1250)
|
115 (78-157)
|
<.0001
|
CSF Nga
|
841 (411-1473)
|
390 (260-766)
|
<.0001
|
Abbreviations: CSF, cerebrospinal fluid; LP, lumbar puncture; NfL, neurofilament light chain; Ng, neurogranin; np-RPD, non-prion rapidly progressive dementia; p-tau, phospho-tau181; SNAP-25, synaptosomal-associated protein 25; t-tau, total tau.
Age at LP is expressed as mean (SD), while biomarker data are presented as median (IQR)
aNg levels were assessed only in a subgroup of 215 patients (122 CJD, 93 np-RPD)
Table 2 - Distribution of SNAP-25 and Ng levels in the subgroups of the CJD and np-RPD cohorts
Diagnostic group
|
N
|
CSF SNAP-25 (pg/mL)
|
N
|
CSF Ng (pg/mL)
|
CJD
|
220
|
582 (240-1250)
|
122
|
841 (411-1473)
|
sCJDa
|
183
|
533 (249-1311)
|
106
|
868 (471-1430)
|
• sCJD MM(V)1
|
71
|
798 (485-1559)
|
31
|
1476 (1139-2527)
|
• sCJD VV2
|
43
|
1438 (962-2034)
|
31
|
745 (384-951)
|
• sCJD MV2K
|
53
|
230 (168-338)
|
29
|
531 (320-715)
|
• sCJD MM(V)2C
|
12
|
186 (105-403)
|
11
|
1179 (695-1975)
|
• sCJD MM2T
|
2
|
85.7, 348.3
|
2
|
404, 1751
|
• sCJD VV1
|
2
|
440.5, 512.9
|
2
|
1280, 1716
|
gCJDb
|
37
|
662 (230-1173)
|
16
|
698 (154-1708)
|
• gCJD M1
|
31
|
744 (344-1200)
|
11
|
983 (355-1718)
|
• gCJD M“i“
|
2
|
107.2, 121.7
|
1
|
-, 157
|
• gCJD M2T
|
3
|
37.2, 52.5, 75.3
|
3
|
14, 84, 153
|
• gCJD V1
|
1
|
334.3
|
1
|
1785
|
np-RPD
|
213
|
115 (78-157)
|
93
|
390 (260-766)
|
Non-neurodegenerative
|
98
|
114 (70-162)
|
43
|
416 (228-945)
|
rp-ND
|
101
|
114 (81-156)
|
45
|
354 (273-646)
|
Mixed
|
14
|
125 (70-146)
|
5
|
787 (355-956)
|
Abbreviations: CSF, cerebrospinal fluid; gCJD, genetic Creutzfeldt-Jakob disease; Ng, neurogranin; np-RPD, non-prion rapidly progressive dementia; rp-ND, neurodegenerative np-RPD; sCJD, sporadic Creutzfeldt-Jakob disease; SNAP-25, synaptosomal-associated protein 25.
aBoth patients with a definite diagnosis of a specific subtype and patients with a probable diagnosis and a high level of certainty for a given subtype are included.
bM1 group includes 13 gCJD E200K-129M, 1 gCJD V203I-129M and 17 gCJD V210I-129M; M “i” group includes 2 gCJD E200K-129M; M2T group includes 3 fatal familial insomnia (FFI) (gCJD D178N-129M); V1 group includes 1 gCJD D178N-129V.
Biomarker data are presented as median (IQR)
Diagnostic performance of CSF biomarkers in the differential diagnosis between CJD and np-RPD
To assess the diagnostic performance of CSF biomarkers, we calculated ROC curves, sensitivity, and specificity for all biomarkers, including their ratios with p-tau. Detailed ROC curve analyses for CSF biomarkers are reported in Table 3 and Figure 1.
In ROC curves analysis, CSF SNAP-25 and Ng yielded a diagnostic accuracy of respectively 90% (area under the curve [AUC] 0.902 [0.873-0.931]) and 69% (AUC 0.697 [0.626-0.767]) in discriminating between CJD and np-RPD. The diagnostic performance of CSF SNAP-25 was better than that of t-tau (AUC 0.878 [0.845-0.901]) (SNAP-25 vs. t-tau, P=.0179), 14-3-3 (AUC 0.853 [0.816-0.889]) (SNAP-25 vs. 14-3-3, P=.0005), and NfL (AUC 0.649 [0.593-0.705]) (SNAP-25 vs. NfL, P<.0001). The diagnostic accuracy of CSF Ng slightly exceeded that of NfL (Ng vs. NfL, P=.0209) but was inferior to that of SNAP-25 (Ng vs. SNAP-25, P<.0001), t-tau (Ng vs. t-tau, P=.0002), and 14-3-3 (Ng vs. 14-3-3, P=.0153). The combined use of p-tau increased the diagnostic accuracy of both the synaptic biomarkers, with the CSF SNAP-25/p-tau ratio showing the best diagnostic performance among the single and the p-tau-corrected biomarkers (AUC 0.924 [0.900-0.948]).
In the differential diagnosis between the "atypical," slowly progressive sCJD subtypes (MV2K, MM(V)2C, MM2T, and VV1) and np-RPD, CSF SNAP-25 diagnostic accuracy (AUC 0.792 [0.729-0.854]) was in the range of that of t-tau, t-tau/p-tau ratio, and SNAP-25/p-tau ratio. CSF SNAP-25 and t-tau outperformed CSF Ng in this sub-analysis (SNAP-25 vs. Ng, P<.0001; t-tau vs. Ng, P=.0169).
When we restricted the analysis to the most common CJD cases (MM(V)1 and VV2), CSF SNAP-25 yielded an extremely high diagnostic accuracy (AUC 0.982 [0.971-0.994]). In this case, the correction of SNAP-25 values with p-tau did not significantly impact the diagnostic performance of the biomarker (AUC 0.988 [0.979-0.996]). In this sub-analysis, CSF Ng diagnostic value (as that of the other biomarkers) improved as well (AUC 0.809 [0.727-0.891]), although its diagnostic power remained way lower than that of SNAP-25, t-tau, and 14-3-3.
Diagnostic performance of CSF biomarkers in the differential diagnosis between CJD and non-neurodegenerative np-RPD
When we assessed the biomarkers' diagnostic performance in distinguishing between patients with CJD and those with non-neurodegenerative np-RPD, CSF SNAP-25 outperformed all the other single and p-tau-adjusted biomarkers (AUC 0.887 [0.850-0.925]) (SNAP-25 vs. t-tau, P=.0003; SNAP-25 vs. t-tau/p-tau, P=.0002; SNAP-25 vs 14-3-3, P<.0001; SNAP-25 vs. NfL, P<.0001). CSF Ng diagnostic value (AUC 0.669 [0.573-0.765]) was significantly lower than that of SNAP-25 (SNAP-25 vs. Ng, P=.0003). In this analysis, p-tau-corrected biomarkers yielded a lower diagnostic accuracy than the single correspondent biomarkers.
Diagnostic performance of CSF biomarkers in the differential diagnosis between CJD and rp-ND
When analyzing the biomarkers diagnostic accuracy in discriminating between patients with CJD and those with an rp-ND, CSF SNAP-25 (AUC 0.916 [0.886-0.946]) was outperformed by 14-3-3 (AUC 0.942 [0.916-0.967]) (SNAP-25 vs. 14-3-3, P=.0203). SNAP-25 diagnostic value was in the range of that of t-tau (AUC 0.928 [0.899-0.957]). When considering the p-tau corrected values, the CSF SNAP-25/p-tau, and t-tau/p-tau ratios yielded the highest diagnostic accuracy among the single and the p-tau-adjusted biomarkers (AUC 0.971 [0.954-0.987] and AUC 0.966 [0.943-0.989], respectively). CSF Ng performed worse than all the other biomarkers (Ng vs. t-tau, P<.0001; Ng vs. 14-3-3, P<.0001; Ng vs. SNAP-25, P<.0001), except for NfL (AUC 0.839 [0.782-0.896]). In this analysis, all p-tau-corrected biomarkers showed better diagnostic accuracy than the corresponding single biomarkers.
Table 3 - Diagnostic performance of SNAP-25, Ng, and other surrogate biomarkers
|
CJD vs. np-RPD
|
Atypical CJDa vs. np-RPD
|
|
AUC
(95% CI)
|
Sensitivity (95%CI)
|
Specificity (95% CI)
|
Cut-off (pg/ml)
|
AUC
(95% CI)
|
Sensitivity (95%CI)
|
Specificity (95% CI)
|
Cut-off (pg/ml)
|
t-tau
|
0.878
(0.845-0.901)
|
78.8%
(72.9-83.7)
|
84.9%
(79.5-89.1)
|
1770
|
0.774
(0.719-0.830)
|
84.6%
(73.9-91.4)
|
63.8%
(57.2-70.0)
|
993.5
|
14-3-3
|
0.853
(0.816-0.889)
|
83.8%
(78.3-88.0)
|
75.1%
(68.9-80.4)
|
21750
|
0.733
(0.676-0.791)
|
81.5%
(70.4-9.1)
|
65.7%
(59.1-71.7)
|
14850
|
NfL
|
0.649
(0.593-0.705)
|
89.6%
(84.9-92.9)
|
51.9%
(45.2-58.5)
|
2245
|
0.626
(0.560-0.691)
|
87.5%
(77.2-93.5)
|
51.8%
(45.1-58.5)
|
2245
|
SNAP-25
|
0.902
(0.873-0.931)
|
84.7%
(79.3-88.8)
|
85.4%
(80.0-89.5)
|
198.7
|
0.792
(0.729-0.854)
|
75.3%
(63.6-84.2)
|
78.4%
(72.4-83.4)
|
163.8
|
Ngb
|
0.697
(0.626-0.767)
|
54.1%
(45.2-62.6)
|
77.4%
(67.9-84.7)
|
793.5
|
0.643
(0.545-0.741)
|
84.0%
(70.6-92.0)
|
46.2%
(36.4-56.3)
|
350.0
|
t-tau/
p-tau
|
0.884
(0.851-0.916)
|
92.2%
(87.8-95.0)
|
70.4
(63.8-76.2)
|
17.80
|
0.778
(0.725-0.830)
|
98.4%
(91.5-99.9)
|
63.6%
(56.8-69.8)
|
11.90
|
SNAP-25/p-tau
|
0.924
(0.900-0.948)
|
92.2%
(87.8-95.0)
|
79.6%
(73.6-84.5)
|
2.950
|
0.812
(0.760-0.864)
|
77.8%
(66.0-86.2)
|
79.6%
(73.6-84.5)
|
2.950
|
|
CJD vs. rp-ND
|
CJD vs. non-neurodegenerative np-RPD
|
|
AUC
(95% CI)
|
Sensitivity (95%CI)
|
Specificity (95% CI)
|
Cut-off (pg/ml)
|
AUC
(95% CI)
|
Sensitivity (95%CI)
|
Specificity (95% CI)
|
Cut-off (pg/ml)
|
t-tau
|
0.928
(0.899-0.957)
|
77.9%
(72.0-82.9)
|
94.0%
(87.6-97.2)
|
1808
|
0.826
(0.775-0.877)
|
78.8%
(72.9-83.7)
|
76.5%
(67.2-83.8)
|
1768
|
14-3-3
|
0.942
(0.916-0.967)
|
86.0%
(80.8-89.9)
|
91.0%
(83.9-95.2)
|
18850
|
0.771
(0.712-0.831)
|
83.8%
(78.4-88.0)
|
60.2%
(50.3-69.3)
|
21750
|
NfL
|
0.839
(0.782-0.896)
|
89.6%
(84.9-92.9)
|
75.2%
(66.0-82.6)
|
2245
|
0.512
(0.429-0.595)
|
90.0%
(85.4-93.3)
|
34.0%
(25.4-43.9)
|
15450
|
SNAP-25
|
0.916
(0.886-0.946)
|
80.6%
(74.9-85.3)
|
94.0%
(87.6-97.2)
|
214.2
|
0.887
(0.850-0.925)
|
84.7%
(79.4-88.8)
|
81.6%
(72.8-88.0)
|
195.3
|
Ngb
|
0.733
(0.652-0.813)
|
54.1%
(45.2-62.6)
|
84.4%
(71.2-92.2)
|
787.0
|
0.669
(0.573-0.765)
|
86.0%
(78.8-91.1)
|
46.5%
(32.5-61.0)
|
339.0
|
t-tau/
p-tau
|
0.966
(0.943-0.989)
|
98.6%
(96.0-99.6)
|
88.0%
(80.1-93.0)
|
11.65
|
0.794
(0.736-0.851)
|
63.7%
(57.1-69.8)
|
82.6%
(73.6-89.0)
|
51.40
|
SNAP-25/p-tau
|
0.971
(0.954-0.987)
|
92.2%
(87.9-95.0)
|
93%
(86.2-96.6)
|
2.950
|
0.869
(0.829-0.910)
|
78.4%
(72.5-83.3)
|
83.7%
(74.8-89.8)
|
4.950
|
Abbreviations: AUC, area under the curve; CI, confidence interval; CSF, cerebrospinal fluid; NfL, neurofilament light chain; Ng, neurogranin; np-RPD, non-prion rapidly progressive dementia; p-tau, phospho-tau181; rp-ND, neurodegenerative np-RPD; SNAP-25, synaptosomal-associated protein 25; t-tau, total tau.
asCJD subtypes MV2K, MM(V)2C, MM2T, and VV1 are included.
bNg AUC were calculated considering a cohort of 215 patients (122 CJD, 93 np-RPD)
After stratification according to the sCJD subtype, MM(V)1 and VV2 patients showed significantly higher levels of CSF SNAP-25 compared to other groups (MM(V)1 vs. MV2K, P<.0001; MM(V)1 vs. MM(V)2C, P=.0006; VV2 vs. MV2K, P<.0001; VV2 vs. MM(V)2C, P<.0001). SNAP-25 levels did not differ significantly between MM(V)1 and VV2 cases. Among the genetic cases, M1 patients showed SNAP-25 levels in the range of MM(V)1 and VV2 groups and significantly higher than MV2K (M1 vs. MV2K, P<.0001) and MM(V)2C (M1 vs. MM(V)2C, P=.0276) cases. Eventually, M2T patients had the lowest CSF SNAP-25 levels of the whole CJD cohort, significantly lower than MM(V)1 (M2T vs. MM(V)1, P=.0147), and VV2 (M2T vs. VV2, P=.0005). All findings previously mentioned remained statistically significant after excluding the probable sCJD cases, except for the comparisons between M1 and MM(V)2C and between MM(V)1 and MM(V)2C.
Regarding the distribution of CSF Ng levels according to CJD subtypes, CJD MM(V)1 patients showed the highest values of the whole cohort, with a statistically significant difference when compared to VV2 (MM(V)1 vs. VV2, P=.0248) and MV2K (MM(V)1 vs. MV2K, P<.0001). VV2 cases had CSF Ng concentrations in the range of MV2K patients. MM(V)2C and the two VV1 cases showed a CSF Ng concentration in the range of the MM(V)1 group. M2T cases had the lowest CSF Ng levels of the CJD cohort.
CSF SNAP-25 and Ng distribution in CJD subtypes is shown in Table 2 and Figure 2. The profiles of the remaining CSF biomarkers stratified by prion disease subtypes are shown in Supplementary table 2 (see Additional file 1).
When considering the whole CJD cohort, CSF SNAP-25 was significantly associated with survival (HR 1.71 [1.48-1.99], P<.001), even after accounting for covariates known as prognostic factors in prion disease as codon 129 genotype, age at LP and time from onset to LP. When stratifying for the clinicopathological subtype, CSF SNAP-25 levels correlated with survival in both “typical CJD” (HR 1.36 [1.03-1.80], P=.029) and “slower CJD” (HR 1.69 [1.38-2.06], P<.001) groups. In the multivariate Cox regression, CSF SNAP-25 were significantly associated with survival in both the whole prion cohort (HR 1.71 [1.40-2.09], P<.001), and “typical CJD” (HR 1.52 [1.15-2.02], P=.003), but not in “slower CJD”.
Conversely, we found no significant associations between CSF Ng and survival when considering the whole CJD cohort and the “slower CJD” subtypes. CSF Ng levels significantly correlated with survival in “typical CJD” in both univariate and multivariate Cox regression (HR 1.50 [1.04-2.16], P=.027; and HR 1.81 [1.21-2.93], P=.015; respectively).
Detailed data regarding the association between CSF SNAP-25 and Ng levels and survival are shown in Table 4. Survival curves are shown in Figure 3. Survival analysis results for t-tau, 14-3-3, and NfL are reported in Supplementary table 3 and shown in Supplementary figure 1 (see Additional file 1).
Table 4 - Associations of SNAP-25 and Ng CSF levels with survival time in the whole CJD cohort and after stratification according to the disease subtype
Diagnostic group and biomarker
|
Survival time
|
Univariate Cox regression
|
Multivariate Cox regressiona
|
Median ± IQR (months)
|
HR (95% CI)
|
P value
|
HR (95% CI)
|
P value
|
Whole CJD cohort
|
SNAP-25
(N = 215)
|
Continuous value
|
5.0 (2.8-12.0)
|
1.71 (1.48-1.99)
|
<.001
|
1.71 (1.40-2.09)
|
<.001
|
Low tertile
|
12.0 (6.0-17.5)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
4.0 (2.5-9.0)
|
1.87 (1.34-2.61)
|
<.001
|
1.71 (1.19-2.44)
|
.003
|
High tertile
|
4.0 (2.5-5.5)
|
3.42 (2.38-4.90)
|
<.001
|
3.42 (2.15-5.45)
|
<.001
|
Ng
(N = 119)
|
Continuous value
|
6.0 (3.6-13.0)
|
1.19 (0.99-1.45)
|
.065
|
1.04 (0.82-1.33)
|
.720
|
Low tertile
|
7.5 (5.4-14.0)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
6.0 (3.5-11.8)
|
0.99 (0.63-1.54)
|
.971
|
0.80 (0.49-1.32)
|
.396
|
High tertile
|
4.1 (2.6-12.0)
|
1.23 (0.78-1.92)
|
.374
|
0.93 (0.54-1.60)
|
.808
|
Typical CJDb
|
SNAP-25
(N = 100)
|
Continuous value
|
2.8 (2.1-4.0)
|
1.36 (1.03-1.80)
|
.029
|
1.52 (1.15-2.02)
|
.003
|
Low tertile
|
3.5 (2.4-4.1)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
2.5 (2.0-3.4)
|
0.94 (0.53-1.65)
|
.831
|
1.03 (0.56-1.77)
|
.991
|
High tertile
|
2.5 (2.1-3.3)
|
1.35 (0.76-2.39)
|
.302
|
1.65 (0.91-2.96)
|
.093
|
Ng
(N = 41)
|
Continuous value
|
3.0 (2.0-4.0)
|
1.50 (1.04-2.16)
|
.027
|
1.81 (1.21-2.93)
|
.015
|
Low tertile
|
2.4 (1.8-4.5)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
3.0 (2.1-3.8)
|
2.48 (0.70-8.76)
|
.157
|
7.90 (1.28-48.71)
|
.026
|
High tertile
|
3.0 (2.5-3.9)
|
3.59 (1.01-12.67)
|
.047
|
10.34 (1.71-62.34)
|
.011
|
Slower CJDc
|
SNAP-25
(N = 115)
|
Continuous value
|
11.0 (6.0-16.0)
|
1.69 (1.38-2.06)
|
<.001
|
1.01 (0.73-1.40)
|
.937
|
Low tertile
|
15.0 (11.0-24.0)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
14.8 (10.8-19.7)
|
1.39 (0.88-2.19)
|
.147
|
1.03 (0.58-1.83)
|
.914
|
High tertile
|
5.5 (4.7-6.5)
|
4.68 (2.81-7.78)
|
<.001
|
0.91 (0.36-2.29)
|
.844
|
Ng
(N = 78)
|
Continuous value
|
9.5 (6.0-16.0)
|
0.93 (0.76-1.15)
|
.547
|
0.81 (0.60-1.09)
|
.169
|
Low tertile
|
9.0 (6.3-15.0)
|
Ref
|
Ref
|
Ref
|
Ref
|
Mid tertile
|
9.0 (5.9-15.5)
|
0.69 (0.41-1.15)
|
.164
|
0.58 (0.33-1.01)
|
.058
|
High tertile
|
12.0 (6.0-23.8)
|
0.50 (0.26-0.94)
|
.034
|
0.45 (0.22-0.91)
|
.027
|
Abbreviations: CI, confidence interval; CSF, cerebrospinal fluid; gCJD, genetic Creutzfeldt-Jakob disease; HR, hazard ratio; IQR, interquartile range; Ng, neurogranin; Ref, reference; sCJD, sporadic Creutzfeldt-Jakob disease; SNAP-25, synaptosomal-associated protein 25.
aAll multivariate Cox regression analyses included codon 129 genotype, age at LP and time from onset to sample collection as covariates.
bIncludes sCJD MM(V)1 and gCJD M1.
cIncludes sCJD VV2, sCJD MV2K, sCJD MM(V)2C, sCJD MM2T, sCJD VV1, gCJD M“i”, gCJD M2T, and gCJD V1.
Bold values indicate statistically significant hazard ratios.