The levels of Aβ40, Aβ42, p-tau and t-tau were measured with a fully automated chemiluminescent platform (Lumipulse G600-II, Fujirebio inc.) on two aliquots of the same CSF sample, one analyzed immediately after the sampling (“fresh”) and the other after 30 days of storage at -80°C. To investigate the relation between fresh and frozen biomarker levels, Passing Bablok regressions were performed, the results of this analysis are plotted in Figure 2 (panels A-E). To assess the significance of the deviations from identity, the relative differences on Aβ40, Aβ42, Aβ42/Aβ40, t-tau and p-tau were also calculated, the results are represented in Figure 2, panel F.
Figure 2 A-E) Passing Bablok regression analyses with 95% CI of core biomarkers measured in fresh CSF samples and after 1 month of deep-freezing. Correlations have been calculated in terms of Spearman’s correlation coefficients (ρ). Fitted slopes (α) and intercepts (β) with their 95% CI are also shown. F) Mean relative differences of Aβ40, Aβ42, Aβ42/Aβ40, t-tau and p-tau between fresh and frozen samples are plotted together. Boxes representing data distributions are centered on the mean values, with the internal horizontal line representing the median. Boxes heights are equal to the standard deviations of the relative differences while whiskers represent the 5%-95% data range.
As shown in the panels A-E, fitted intercepts are consistently equal to zero within their CI for all the measured quantities. The largest deviation from identity was obtained for Aβ40. Bland Altman plots (Figure S1) showed higher concentration values for Aβ40, Aβ42, p-tau and t-tau in fresh samples and a trend of larger differences in magnitude for higher values of biomarkers. However, about 95% of the samples fell within 2 SD from the mean fresh vs frozen difference, confirming the high correlation obtained with Passing-Bablok analysis. As shown in Figure 2 F and Table 1, mean relative differences greater than 0.06 have not been observed for any of the tested biomarkers. Even if small, the differences on Aβ40, Aβ42/Aβ40 ratio, t-tau and p-tau were significant according to Student’s t-test (see Table 1). Although the relative differences on Aβ42/Aβ40 ratio were significant, this biomarker showed less variability in terms of standard deviation (SD) of relative differences with respect to Aβ42 and Aβ40 alone (0.075 vs 0.11). The highest SD of relative differences was obtained for t-tau (0.16) while the smallest for p-tau (0.06), the latter showed also the smallest mean relative difference (0.017) and the highest correlation between measurements (ρ = 0.99). Despite of the presence/absence of the freezing step, for each analyte, the measured SD of relative differences between fresh and frozen samples was of the same order of magnitude of the inter-assay variability (CV) of the QC sample (Table S4).
Table 1 mean relative differences between fresh and frozen samples with their 95%CI, standard deviations of relative differences (SD) and p-values obtained by applying one-sample Student’s t-test.
Biomarker
|
Mean relative difference (95%CI)
|
SD
|
p-values t-test
|
|
Aβ40
|
0.059 (0.033, 0.084)
|
0.10
|
< 0.001
|
|
Aβ42
|
0.023 (-0.004, 0.051)
|
0.11
|
0.10
|
|
Aβ42/Aβ40
|
-0.039 (-0.059, -0.020)
|
0.075
|
< 0.001
|
|
t-tau
|
0.044 (0.001, 0.086)
|
0.16
|
0.04
|
|
p-tau
|
0.017 (0.0015, 0.032)
|
0.059
|
0.03
|
|
For each aliquot, Aβ42/Aβ40, p-tau and t-tau have been classified according to the A/T/(N) criteria [6] by using standard cutoffs and tolerances (reported in the Methods section). The results of this classification are shown in Table 2.
Table 2 Classification of core AD biomarkers and in fresh and frozen samples by using the A/T/(N) criteria. Number of changes in classification between biomarkers measured in fresh and stored at -80°C CSF samples are also reported. Samples with both pathologic Aβ42/Aβ40 ratio and pathologic p-tau (A+/T+) have been classified positive for AD. For “robust” AD diagnosis and AD exclusion (non-AD) a ± 10% tolerance was applied on the cut-offs of Aβ42/Aβ40 ratio and p-tau.
|
+ (%)
|
− (%)
|
N° of changes (%)
|
Aβ42/Aβ40 frozen
|
39 (67%)
|
19 (33%)
|
1 (2%)
|
Aβ42/Aβ40 fresh
|
40 (69%)
|
18 (31%)
|
p-tau frozen
|
34 (59%)
|
24 (41%)
|
0 (0%)
|
p-tau fresh
|
34 (59%)
|
24 (41%)
|
t-tau frozen
|
34 (59%)
|
24 (41%)
|
2 (3%)
|
t-tau fresh
|
36 (62%)
|
22 (38%)
|
AD frozen
|
34 (59%)
|
24 (41%)
|
0 (0%)
|
AD fresh
|
34 (59%)
|
24 (41%)
|
robust AD frozen
|
32 (55%)
|
26 (45%)
|
1 (2%)
|
robust AD fresh
|
33 (57%)
|
25 (43%)
|
robust non-AD frozen
|
24 (41%)
|
34 (59%)
|
1 (2%)
|
robust non-AD fresh
|
23 (40%)
|
35 (60%)
|
Considering the classification (Table 2), in some cases, Aβ42/Aβ40 ratio and t-tau changed their classification between fresh and frozen samples. However, when looking at AD diagnosis, which mainly relies on the values of Aβ42/Aβ40 ratio and p-tau (A/T), none of the samples changed classification directly from + to - or the other way around considering the raw cut-offs. Only one sample changed from + to – by considering a + 10% tolerance on Aβ42/Aβ40 ratio and p-tau cut-offs (robust AD diagnosis) and another one when considering a - 10% for excluding AD.
To further investigate the impact of these differences in diagnostics, we recalculated the cut-offs for Aβ42/Aβ40 ratio, p-tau and t-tau for fresh samples. One way to do that is to transfer frozen-defined cut-offs by applying a linear transformation based on the parameters estimated with the Passing Bablok regression (cut-off fresh R in Table 3). Another way can be to consider the raw classification of frozen samples as a reference and maximize Youden’s index for fresh samples (cut-off fresh Y in Table 3). The results of both procedures are shown in Table 3.
Table 3 Recalculated cut-offs for freshly measured CSF core AD biomarkers with their 95% CI. Cut-off fresh R: cut-off transferred by applying Passing Bablok regression. Cut-off fresh Y: cut-off calculated maximizing the Youden’s index for fresh samples, by taking as reference the classification performed on frozen ones.
Biomarker
|
cut-off frozen
|
cut-off fresh R (95%CI)
|
cut-off fresh Y (95%CI)
|
Aβ42/Aβ40
|
0.069
|
0.067 (0.060, 0.075)
|
0.062 (0.061, 0.075)
|
p-tau
|
56.5
|
57.2 (55.7, 59.7)
|
55.1 (52.3, 60.9)
|
t-tau
|
404
|
419 (379, 473)
|
419 (406, 581)
|
As it is shown in Table 3, only for t-tau the recalculated cut-off from direct diagnostic information (Youden) does not include the cut-off defined in frozen samples within its 95% CI.