Skip to main content
. 2018 Sep 7;5(10):1163–1175. doi: 10.1002/acn3.629

Table 2.

ROC analysis of CSF parameters discriminating different diagnostic groups in the discovery cohort

Cut‐off point1 Sensitivity (%) Specificity (%) AUC (95% CI) +LR2 LR3 p value (individually) Coefficient (B) P value (within model)
CSF variables
FTLD (n = 49) vs. CON (n = 23)
MFG‐E8 na na na 0.55 (0.405‐0.702) na na 0.428 −0.0004 0.005
tTau na na na 0.62 (0.487–0.747) na na 0.111 0.068 <0.0001
Aβ42 264 71 69 0.75 (0.632–0.859) 2.29 0.42 0.001 −0.010 <0.0001
FTLD vs. CON model4 0.686 78 83 0.90 (0.8270.976) 4.59 0.27 <0.0001
FTLD (n = 49) vs. non‐FTLD dementia (n = 57)
YKL40 na na na 0.60 (0.491–0.708) na na 0.071 −0.004 0.021
pTau 29.1 77 67 0.80 (0.702–0.877) 2.3 0.3 <0.0001 0.044 <0.0001
p/tTau na na na 0.62 (0.512–0.728) na na 0.034 2.33 0.008
Age na na na 0.61 (0.503–0.719) na na 0.047 0.079 0.010
FTLD vs. non‐FTLD dementia model5 0.3903 90 78 0.86 (0.7810.930) na na <0.0001
FTLD‐TDP (n = 29) vs. FTLD ‐Tau (n = 20)
p/t Tau ratio 0.285 80 59 0.77 (0.641–0.906) 2 0.3 0.001 7.67 0.006
YKL40 na na na 0.64 (0.476–0.794) na na 0.109 −0.009 0.049
MFG‐E8 na na na 0.57 (0.411–0.729) na na 0.406 0.001 0.027
βHexA na na na 0.57 (0.404–0.737) na na 0.414 −0.012 0.021
βHexA/tHex na na na 0.54 (0.370–0.708) na na 0.651 44.330 0.041
Age na na na 0.52 (0.336–0.696) na na 0.851 0.133 0.036
TDP vs. Tau model6 0.4563 80 81 0.87 (0.7720.969) 4.2 0.2 <0.0001
FTLD‐TDP (n = 29) vs. CON (n = 30)
YKL40 25.35 80 62 0.78 (0.665–0.901) 2.1 0.3 <0.0001 0.017 0.001
MFG‐E8 na na na 0.44 (0.283–0.589) na na 0.396 −0.001 0.003
Catalase na na na 0.62 (0.474–0.761) na na 0.122 0.430 0.024
TDP vs. CON model7 0.4431 90 76 0.88 (0.7960.965) 3.8 0.1 <0.0001

AUC, area under the curve; CI, confident interval; LR, likelihood ratio; CON, nondemented controls; FTLD, frontotemporal lobar degeneration.

n.a: not applicable due to the lack of significance.

1

Selected value of the individual biomarker or combination where the two groups of analysis could be discriminated with the reported sensitivity and specificity.

2

Positive likelihood: sensitivity/100‐specificity.

3

Negative likelihood: 100‐sensitivity/specificity.

4

FTLD vs. CON model: y = 2.94 − 0.0004*MFGE8 + 0.07*tTau − 0.01*Aβ 42.

5

FTLD vs. non‐FTLD dementia models: y = −6.621 − 0.004*YKL40 + 0.044*pTau + 2.33*p/tTau ratio + 0.79*Age.

6

TDP vs. Tau model: y = −14.659 − 0.009*YKL40 + 0.001*MFGE8 − 0.012*βHexA activity + 44.33*βHexA/tHex activity ratio + 7.671*p/tTau ratio + 0.133*Age.

7

TDP vs. CON model: y = −3.193 + 0.017*YKL40 − 0.001*MFGE8 + 0.43*catalase activity.

Markers/models achieving sufficient biomarker performance are highlighted in bold.